Derived Relational Responding Applications For Learners (2009) PDF

Derived Relational Responding
Applications for Learners

with Autism and Other
Developmental Disabilities
A Progressive Guide to Change
EDITED BY
RUTH ANNE REHFELDT, PH.D., BCBA
& YVONNE BARNES-HOLMES, PH.D.
Context Press
New Harbinger Publications, Inc.
Publisher’s Note
This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold with
the understanding that the publisher is not engaged in rendering psychological, financial, legal, or other professional services. If expert
assistance or counseling is needed, the services of a competent professional should be sought.
Distributed in Canada by Raincoast Books
Copyright © 2009 by Ruth Anne Rehfeldt and Yvonne Barnes-Holmes

New Harbinger Publications, Inc.
5674 Shattuck Avenue
Oakland, CA 94609
www.newharbinger.com
All Rights Reserved

Printed in the United States of America
Acquired by Catharine Sutker; Cover design by Amy Shoup;

Edited by Karen Stein; Text design by Tracy Carlson
A copublication of New Harbinger Publications and Noetic Books
Library of Congress Cataloging-in-Publication Data
Derived relational responding : applications for learners with autism and other developmental disabilities / edited by Ruth Anne
Rehfeldt and Yvonne Barnes-Holmes ; foreword by Steven C. Hayes.
p. ; cm.
Includes bibliographical references and index.
ISBN-13: 978-1-57224-536-5 (hardcover : alk. paper)
ISBN-10: 1-57224-536-0 (hardcover : alk. paper)
1. Autistic children--Rehabilitation. 2. Developmentally disabled children--Rehabilitation. I. Rehfeldt, Ruth Anne. II. Barnes-
Holmes, Yvonne.
[DNLM: 1. Autistic Disorder--rehabilitation. 2. Association Learning. 3. Conditioning, Operant. 4. Developmental Disabilities-
-rehabilitation. 5. Reinforcement (Psychology) WM 203.5 D598 2009]
RJ506.A9D457 2009
362.198’9285882--dc22
2008052211
11 10 09
10 9 8 7 6 5 4 3 2 1 First printing

Contents
Series Editor Letter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Foreword
An Applied Behavioral Psychology of Language and Cognition . . . ix
Introduction and Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 1
PART 1
Establishing the Prerequisites for Normal Language
Chapter 1
Reinforcer Identification Strategies and Teaching Learner
Readiness Skills. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Thomas S. Higbee, Utah State University
Chapter 2
The Assessment of Basic Learning Abilities (ABLA) and Its
Relation to the Development of Stimulus Relations in Persons
with Autism and Other Intellectual Disabilities. . . . . . . . . . . . . . . . 25
W. Larry Williams and Marianne L. Jackson, University of Nevada, Reno
Chapter 3
Observing Responses: Foundations of Higher-Order
Verbal Operants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Dolleen-Day Keohane and Jo Ann Pereira Delgado, Columbia University
Teachers College and CABAS; and R. Douglas Greer, Columbia University
Graduate School of Arts and Science Teachers College
Chapter 4
Joint Attention and Social Referencing in Infancy as
Precursors of Derived Relational Responding . . . . . . . . . . . . . . . . . 63
Martha Peláez, Florida International University
Chapter 5
Establishing Mand and Tact Repertoires. . . . . . . . . . . . . . . . . . . . . 79
Linda A. LeBlanc and Courtney M. Dillon, Western Michigan University;
and Rachael A. Sautter, Y.A.L.E. School
Part 2
Speaking with Meaning and Listening with Understanding
Chapter 6
Nonrelational and Relational Instructional Control. . . . . . . . . . . . 111
Jonathan Tarbox, Center for Autism & Related Disorders; Rachel S. F.
Tarbox, Chicago School of Professional Psychology at Los Angeles; and
Denis O’Hora, National University of Ireland, Galway
Chapter 7
Naming and Frames of Coordination . . . . . . . . . . . . . . . . . . . . . . 129
Caio F. Miguel, California State University, Sacramento; and
Anna I. Petursdottir, Texas Christian University
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
iv Derived Relational Responding

Chapter 8
Acquiring the Earliest Relational Operants: Coordination,
Difference, Opposition, Comparison, and Hierarchy. . . . . . . . . . . 149
Carmen Luciano, Miguel Rodríguez, Israel Mañas, and Francisco Ruiz,
University of Almeria, Spain; Nicholas M. Berens, Center for Advanced
Learning, Reno; Sonsoles Valdivia-Salas, University of Albany, New York
Chapter 9
Applying Relational Operants to Reading and Spelling. . . . . . . . . 171
Deisy G. de Souza, Julio C. de Rose, and Camila Domeniconi, Universidade
Federal de São Carlos, Brazil
Chapter 10
Syntax, Grammatical Transformation, and Productivity:
A Synthesis of Stimulus Sequences, Equivalence Classes,
and Contextual Control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Harry A. Mackay, Northeastern University and Praxis Inc.; and Lanny Fields,
Queens College and the Graduate School of the City University of New York
Chapter 11
Extending Functional Communication Through
Relational Framing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Rocio Rosales and Ruth Anne Rehfeldt, Southern Illinois University
Part 3
Self, Reasoning, Problem Solving, and Creativity
Chapter 12
Training Analogical Reasoning as Relational Responding. . . . . . . 257
Ian Stewart, National University of Ireland, Galway; Dermot Barnes-Holmes,
National University of Ireland, Maynooth; and Tim Weil, University of
Nevada, Reno
v
Chapter 13
Understanding and Training Perspective Taking as
Relational Responding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Louise McHugh, University of Wales, Swansea; and Yvonne Barnes-Holmes
and Dermot Barnes-Holmes, National University of Ireland, Maynooth
Chapter 14
Establishing Empathy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Sonsoles Valdivia-Salas, University of Albany, New York; Carmen Luciano, University
of Almeria, Spain; Olga Gutiérrez-Martinez, University of Central Barcelona, Spain;
and Carmelo Visdómine, Justice Administration, Madrid, Spain
Chapter 15
Mathematical Reasoning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
Chris Ninness, James Holland, Glen McCuller, Robin Rumph, Sharon Ninness,
and Jennifer McGinty, Stephen F. Austin State University; and Mark Dixon,
Southern Illinois University
Chapter 16
Developing Self-Directed Rules. . . . . . . . . . . . . . . . . . . . . . . . . . . 335
Carmen Luciano, University of Almeria, Spain; Sonsoles Valdivia-Salas,
University of Albany, New York; Francisco Cabello-Luque, Universigy of
Murcia, Spain; and Monica Hernandez, Universigy of Jaén, Spain
Chapter 17
Teaching Flexible, Intelligent, and Creative Behavior. . . . . . . . . . . 353
Catriona O’Toole, Carol Murphy, and Dermot Barnes-Holmes, National University
of Ireland, Maynooth; Jennifer O’Connor, ABACAS, Kilbarrack, Ireland
Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
vi Derived Relational Responding

Dear reader,
Welcome to New Harbinger Publications. New Harbinger is dedicated to publishing
books based on acceptance and commitment therapy (ACT) and its application to spe-
cific areas. New Harbinger has a long-standing reputation as a publisher of quality, well-
researched books for general and professional audiences.
Most existing books in the ACT series focus on teaching either practitioners or
members of the general public how to develop acceptance and mindfulness skills, and
how to apply those skills to various aspects of their experience. Although theoretical con-
structs and concepts are discussed to some extent in these books, they never take center
stage. This book is different. It explicitly deals with relational frame theory, a new behav-
ioral psychology of language and cognition that has emerged on many fronts over the
past thirty years. Human behavior is qualitatively different from and more complex than
animal behavior, and many would agree that it is the complexity of human language and
cognition that sets us apart from nonhuman animals. So making advances in understand-
ing how language and cognition work to our advantage and to our detriment is truly
important.
One of the goals of acceptance and commitment therapy is to help people move
beyond the tyranny of the mind that brings so much unnecessary suffering into the lives
of virtually all of us—not just people with clinical diagnoses. Because RFT provides an
understanding of how the mind works, it is helpful for an ACT therapist (and in fact
any other therapist) to know and understand RFT. However, psychotherapy is only one
application of RFT, and this book makes it clear that RFT has much broader practical
implications and applications. A glance at the table of contents will immediately show
the incredibly broad scope of RFT. Indeed, what sets this book apart from other books
on RFT is the formidable range of areas and applied topics from education and clinical
psychology that are covered in it. Several of the chapters examine specific skills that are
essential to all human functioning, such as reasoning (including mathematical reasoning),
perspective taking, and establishing empathy. All chapters present a balance of theory,
empirical data, and specific applications that bring theoretical concepts to life—often in
the form of real-life or case examples. The authors also offer useful suggestions on how to
apply RFT knowledge in a variety of contexts.
As part of New Harbinger’s commitment to publishing books based on sound, scien-
tific, clinical research, we oversee all prospective books for the Acceptance and Commitment
Therapy Series. Serving as ACT series editors, we comment on proposals and offer guid-
ance as needed, and use a gentle hand in making suggestions regarding the content,
depth, and scope of each book.
Books in the Acceptance and Commitment Therapy Series:
Have an adequate database, appropriate to the strength of the claims

being made.
Are theoretically coherent. They will fit with the ACT model and under-
lying behavioral principles as they have evolved at the time of writing.
Orient the reader toward unresolved empirical issues.
Do not overlap needlessly with existing volumes.

Avoid jargon and unnecessary entanglement with proprietary methods,
leaving ACT work open and available.
Keep the focus always on what is good for the reader.
Support the further development of the field.
Provide information in a way that is of practical use to readers.
These guidelines reflect the values of the broader ACT community. You’ll see all of
them packed into this book. They are meant to ensure that professionals get information
that can truly be helpful, and that can further our ability to alleviate human suffering by
inviting creative practitioners into the process of developing, applying, and refining this
approach. Consider this book such an invitation.
Sincerely,
Georg H. Eifert, Ph.D., John Forsyth, Ph.D., and Steven C. Hayes, Ph.D.
viii Derived Relational Responding

Foreword
An Applied Behavioral Psychology of

Language and Cognition
Behavior analysis is a unique tradition in psychology. It was based on the bold idea that
a set of analytic principles could be developed from work with nonhuman animals that
would eventually enable the understanding and analysis of human complexity. Studying
the relatively simple acts in relatively simple contexts of organisms with relatively well-
known histories was a strategy, not the final goal of the research. B. F. Skinner said his
approach was to create “a promising conception of human behavior” by a research focus
“which began with simple organisms in simple situations and moved on, but only as its
growing power permitted, to the complexities of the world at large” (Skinner, 1938 p. xiv).
Behavioral psychology aspired to give a comprehensive account of psychological phenom-
ena. It was just taking a bottom-up approach to getting there.
Behavior analysts did not know if this strategy would succeed. It was not a certainty—
it was a hope. There is nothing in evolutionary processes that ensures that the behavior of
nonhumans must provide a good guide to the analysis of complex human behavior.
Remarkably, it often does. Indeed, the behavior analytic strategy worked almost too
well. Functional principles of behavior proved to be powerful tools with vast applied
relevance, but they did not seem to provide a fully adequate set of tools for the empiri-
cal analysis of language and cognition. That combination was a bad one for behavioral
psychology. Progress was good enough to keep adherents happy but poor enough to lead
others to abandon hope of ever succeeding.
By the mid 1970s the mainstream had moved elsewhere, convinced that the original
dream of a comprehensive psychology was either unattainable or unlikely to be success-
ful within the behavioral tradition. Psychology focused on cognitive science. Clinicians
moved on to clinical theories of cognition. Microtheories were the norm, and grand
visions fell out of favor.
And there it stood for a few decades. But in the last fifteen years something remark-
able has occurred. A true behavioral psychology of language and cognition has begun to
form. Initially it resided almost entirely in the basic behavioral laboratories, but ever so
gradually it has emerged in the applied arena as well. As befitted the inductive behavior
analytic tradition, each step built on the work of the past. Little needed to be thrown
away; instead, more and more was added until, finally, it became reasonable to suppose
that behavioral psychology could successfully and empirically move on, as its growing
power permitted, to the complexities of human language and cognition.
One began to see the implications of this shift first in clinical behavior analysis as
approaches such as acceptance and commitment therapy, dialectical behavior therapy, and
functional analytic psychotherapy began to revitalize the behavioral wing of cognitive
behavioral therapy. But these clinical extensions were not yet point-to-point empirical
extensions of the new basic behavioral work in language and cognition. That would take
more time.
This book is the next giant step in that process. It is the first to demonstrate a com-
prehensive set of applied behavior analytic training approaches for language and cogni-
tion that directly addresses most of the key areas within that domain. The chapters avoid
needless quarrels between competing factions within basic behavior analysis; theories in
this volume are treated more as useful tools than as distinctions between warring camps.
Even if individual chapters largely adopt a particular perspective, considered as an entire
set they give testimony to the emergence of an increasingly unified behavior analytic
approach that is now ready to walk, step-by-step, from the simplest learning tasks all the
way through empathy, self, and creativity.
That is a notable achievement, and one that may be a first in applied psychology. I
know of no other book that extends a single approach within basic experimental psychol-
ogy into intervention programs across the full range of issues that need to be addressed in
applied work in human language and cognition.
This book, intended for parents and a variety of professionals working with individu-
als with autism and other developmental disabilities, who understand the principles of
applied behavior analysis, begins with the earliest steps needed to establish the prereq-
uisites for normal language, helping applied workers think through how to determine
reinforcers; how to establish observation, attention, and simple discriminations; and how
to establish a simple repertoire of Skinnerian verbal operants. That section is closest to
traditional applied behavior analysis and it deals with issues that many cognitive scientists
do not address because they deal with higher-functioning participants in their studies of
reading or problem solving. The book then steps firmly and boldly into derived relational
responding and the issues of meaning and understanding. Chapters address instructional
control, naming, acquisition of relational framing, syntax, reading, and functional com-
munication. These chapters represent important steps forward within behavior analysis,
bringing together research that is reasonably well-known but also showing in the total-
ity how much progress has been made. Finally, the book dances into some of the most
complex issues of all as it considers self, reasoning, problem solving, and creativity, with
chapters on analogy, perspective taking, empathy, self-rules, mathematics, and creativity.
Especially in this last section, the volume reveals how bold and cutting-edge it is. Yet even
in this final section, and throughout the entire book, all of the chapters have empirical
support.
What is almost startling is that this book is not mere interpretation and logical exten-
sion, or a broad vision to be tested in some distant future. It is also not a volume that tries
to declare by fiat that a limited empirical analysis obviates an analysis of more complex
issues. We have seen such volumes before. What we have not seen before is a compre-
hensive empirical book that covers the full range of applied topics that educators and
clinicians can begin to use now. In its scope, practicality, and empirical base, this volume
declares that a comprehensive applied behavioral psychology of language and cognition is
here, is real, and is moving ahead.
Will this approach fully succeed as measured against the bold goals of behavior anal-
ysis? That we do not know. But a volume like this is needed to help applied workers
take the steps to find out; within the inductive, technique-building tradition of behavior
x Derived Relational Responding

analysis, each step forward is likely to create progress that is sustained, since even when
well-crafted steps fail, they provide important information.
Applied behavior analysis is a sophisticated and vigorous area with thousands of
sophisticated and creative applied professionals. I can’t wait to see what all of the wonder-
ful behavioral educators and practitioners do with this bold new approach.
Steven C. Hayes
University of Nevada
References
Skinner, B. F. (1938). Behavior of Organisms. New York: Appelton-Century-Crofts.
xi
Introduction and Acknowledgments
In the 1970s, Murray Sidman and colleagues made an important, albeit unplanned, dis-
covery that would serve as the impetus for a prosperous research program in behavior
analysis for years to come. Sidman (1971, 1977, Sidman & Cresson, 1973) found that,
after researchers explicitly taught individuals with developmental disorders and limited
language skills to match dictated names to the corresponding pictures and the pictures to
their corresponding printed words, the individuals proved capable of naming the pictures,
orally reading the text, and matching words to pictures and pictures to words—a skill
seemingly indicative of reading comprehension—all in the absence of direct instruction.
Sidman (1971, 1977) termed these novel, emergent relations equivalence relations, as the
untrained skills seemed to represent symbolic or referential behavior, phenomena for which
a behavior analytic explanation had previously been lacking. These findings, and those
that would follow in years to come, were exciting for the field of behavior analysis, for
they identified an economic and efficient means of establishing new academic and pre
academic repertoires in individuals with significant learning challenges, as each desired
skill did not have to be directly taught. But what was perhaps even more important is that
these developments, even where they were academic, have led to significant transforma-
tions in our understanding of the basic processes of human language and cognition—
making it possible that in years to come there will be no areas of learning that cannot be
taught.
In the years since Sidman’s pioneering discoveries, a voluminous body of laboratory
research has been conducted that further elucidated the conditions necessary and suf-
ficient for the establishment of stimulus equivalence, multiple stimulus relations, and
indeed whole networks of derived or untrained stimulus relations. This research led to
the formulation of several major theories regarding the nature of relational learning (such
as relational frame theory and the naming hypothesis), which have inspired even more
sophisticated types of experimental research and innovative methodologies. Of course,
it is not of paramount importance that we determine which is the best theory at one
level or another. But the best test is one that shows which theory delivers most into the
lives of those who need it, either scientifically and indirectly in the form of the advance-
ment of knowledge or directly in the form of the development of new educational tech-
nologies. And on both counts we have much to inspire us. For example, we know that
establishing a history of reinforced relational responding in individuals with a variety
of learning challenges is an effective and efficient means of programming for the emer-
gence of such educationally relevant skills as reading and spelling (Hanna, de Souza, de
Rose, & Fonseca, 2004), recognizing names and faces of caregivers (Cowley, Green, &
Braunling-McMorrow, 1992), requesting preferred items (Rosales & Rehfeldt, 2007), and
understanding basic numerical concepts (Lynch & Cuvo, 1995), to name a few. Thus,
the incorporation of a technology based on derived stimulus relations into learning cur-
ricula for individuals with developmental disabilities would seem to hold great promise in
helping such individuals acquire functional and meaningful goals.
The present book serves as a compilation of instructional strategies based on decades
of basic and applied research on derived stimulus relations from prominent, world-
renowned researchers who attest to different theoretical frameworks. The book is intended
for parents and a variety of professionals working with individuals with autism and other
developmental disabilities. These professionals include but are not limited to teachers,
developmental therapists, adult service providers, speech-language pathologists, and
behavior analysts, all of whom have some basic understanding of the principles of applied
behavior analysis.
The book is divided into three parts. Part 1 focuses on the establishment of prerequisite
skills necessary for individuals to participate meaningfully in a curriculum based upon,
or including components of, derived relational responding. Part 2 emphasizes instruction
that will lead to the production of such intermediate skills as naming, reading, spelling,
and requesting. Part 3 aims to help the practitioner establish more complex skills in learn-
ers, including perspective taking and empathy, higher-order intelligence, and mathemati-
cal competence. Each chapter contains a variety of practitioner tools, such as sample data
sheets, step-by-step instructions, training notes, and problem-solving strategies. The reader
need not work through the entire book for it to be of value. Some learners may be more
appropriate candidates for the strategies and techniques presented in one or more parts of
the book only. Thus, the chapters can be used in isolation or in combination with other
chapters, depending on the particular learner’s educational needs. It is also not necessary
for the reader to be committed to one particular theory regarding derived stimulus rela-
tions or verbal behavior, since the chapters represent an eclectic mix of theoretical orienta-
tions. Rather, our intention is that the strategies in this book can be incorporated, if not
made the basis of, educational curricula for learners with mild or significant communica-
tion and intellectual deficits due to autism, mental retardation, or other developmental
disabilities. As a result, we hope that practitioners and their clients will benefit from the
material presented in this book, and that future years will see the implementation of this
technology in schools, clinics, and habilitation settings around the world.
We wish to thank Anna Neises and Char Burrell for editorial assistance, and we
gratefully acknowledge the many contributors to this book for their thoughtful and per-
sistent work. Many of the authors who contributed to this book are our long-standing
colleagues, friends, and mentors and we have been proud to be associated with their
creativity and expertise. In this vein, we also acknowledge the many researchers, research
assistants, students, and research participants whose work over many years has been an
inspiration for this book.
Ruth Anne Rehfeldt, Ph.D., BCBA Yvonne Barnes-Holmes, Ph.D.

Southern Illinois University National University of Ireland,
Carbondale, Illinois Maynooth Ireland
2 Derived Relational Responding

References
Cowley, B. J., Green, G., & Braunling-McMorrow, D. (1992). Using stimulus equiva-
lence procedures to teach name-face matching to adults with brain injuries. Journal of
Applied Behavior Analysis, 25, 461–475.
Hanna, E. S., de Souza, D. G., de Rose, J. C., & Fonseca, M. (2004). Effects of delayed
constructed-response identity matching on spelling of dictated words. Journal of
Lynch, D. C., & Cuvo, A. J. (1995). Stimulus equivalence instruction of fraction-decimal
relations. Journal of Applied Behavior Analysis, 28, 115–126.
Rosales, R., & Rehfeldt, R. A. (2007). Contriving transitive conditioned establishing
operations to establish derived manding skills in adults with severe developmental
disabilities. Journal of Applied Behavior Analysis, 40, 105–121.
Sidman, M. (1971). Reading and auditory-visual equivalences. Journal of Speech and
Hearing Research, 14, 5–13.
———. (1977). Teaching some basic prerequisites for reading. In P. Mittler (Ed.), Research
to practice in mental retardation: Vol. 2. Education and training. Baltimore, MD:
University Park Press.
Sidman, M., & Cresson, O. (1973). Reading and crossmodal transfer of stimulus equiva-
lences in severe retardation. American Journal of Mental Deficiency, 77, 515–523.
Introduction and Acknowledgments 3

PART 1
Establishing the Prerequisites

for Normal Language
Chapter 1
Reinforcer Identification Strategies and

Teaching Learner Readiness Skills
Thomas S. Higbee,
Utah State University
Behavioral intervention programs provide students with autism and other developmental
disabilities with opportunities to practice and acquire important skills. The success or
failure of these programs often depends on the quality of reinforcement that is provided
for appropriate student behavior. Though the goal of behavioral intervention programs,
including those based upon derived stimulus relations, is to teach students complex verbal
and social skills, necessary learner readiness skills, such as sitting in a chair and attending
to the instructor and instructional materials, often need to be taught first. This chapter
focuses on strategies for identifying reinforcers and methods for teaching learner readiness
skills.
Reinforcer Identification Strategies

Positive reinforcement is the engine that drives any behavioral education program. For
practitioners working with students with autism, identifying effective reinforcers is often
both the most challenging and the most important part of the intervention program.
Over the past several years, a behavioral technology called stimulus preference assessment
(SPA) has been developed that allows practitioners to identify potentially effective rein-
forcers for students with autism and other disabilities. Modern SPA techniques involve
systematically providing students with choices between potential reinforcing items and
activities and then measuring their choices. Although multiple methods of conducting
preference assessments may be effective, the two that are most commonly used will be
discussed here: the paired stimulus (forced choice) method and the multiple stimulus
without replacement method.
Paired Stimulus (Forced Choice) Method

The paired stimulus method of preference assessment was developed by Fisher and
colleagues (1992). In the paired stimulus method, potentially reinforcing items and activi-
ties are presented two at a time and the student is asked to make a choice between the
two available items. Each item is presented at least twice (once on the left and once on the
right, to control for side preferences) with every other item. A selection response is scored
when the individual makes physical contact with one of the available items (although other
choice responses could be identified for individuals with motor difficulties). Following all
choice trials, a selection percentage is calculated for each item by dividing the number of
times an item was selected by the number of times that it was available for selection and
then multiplying by one hundred (Fisher et al., 1992). Items are then ranked according to
selection percentages. Items with selection percentages of 80 percent or greater are likely
to function as reinforcers.
Fisher and his colleagues (1992) compared the paired stimulus method of assessing
preference to a method called the single item presentation method (Pace, Ivancic, Edwards,
Iwata, & Page, 1985), in which items or activities were presented one at a time to par-
ticipants, and whether or not the participant made contact with or approached each item
was recorded. Although the single item presentation method had been demonstrated to be
effective in classifying items as being either preferred or not preferred, it was not capable
of directly comparing items to one another (because items were never presented together)
and because of this was thought to overestimate preference in at least some cases. Fisher
and others (1992) conducted preference assessments using both the single item and paired
stimulus methods with the same individuals using the same set of items or activities
for both assessments. They then evaluated the reinforcing effects of items that had been
ranked as highly preferred using both preference assessment methods as well as items that
were ranked as highly preferred by the single item method but not by the paired stimulus
method. The paired stimulus method was shown to more accurately identify potential
reinforcers when compared to the single item method.
The paired stimulus method is generally considered to be the gold standard method
of identifying potential reinforcers. It allows for items or activities to be directly compared
to one another in a controlled and systematic way in order to determine preference. The
only potential disadvantage to the paired stimulus approach is the length of time that can
be required to complete the preference assessment. Pairing each item with every single
other item for two trials can be a very time-consuming process even if only a few items
are compared. For example, twenty preference assessment trials would be required in
order to complete a paired stimulus preference assessment involving five items or activi-
ties. If more items are compared, the number of trials goes up exponentially. Thus, even
though it is likely the most accurate form of preference assessment, it may not be practi-
cal to conduct paired stimulus preference assessments during each instructional session or
even on a daily basis. Nevertheless, if accuracy is the most important consideration, then

the paired stimulus preference assessment is the most appropriate choice. It may also be
most appropriate for assessing larger numbers of stimuli than can be concurrently assessed
using the MSWO method described below.
The paired stimulus preference assessment data sheet shown below, which is for five
items, can be used to record and analyze the data from the assessment.
The following are the procedures for running the paired stimulus preference assess-
ment (Fisher et al., 1992):
1. Identify four items or activities that the student has requested in the past or has
been observed to interact with during free-choice times, as well as one new item
or activity. If edibles are among the choices, break them up into small bite-size
pieces before presenting them to the student. If a drink is offered present only a
small amount in the cup so the student can drink this amount quickly. If toys
or activities are used, make sure that interacting with the toys or activities will
be meaningful if it only occurs briefly.
2. Allow the student to briefly sample each item by allowing him or her to eat
or drink a small portion of edibles or briefly (for example, for ten to fifteen
seconds) engage with nonedibles.
3. For each trial, place two items on the table about one foot apart in front of the
student. The student should be seated in front of the table with easy access to
the pair. Present each item twice with every other item.
4. Tell the student to pick one. Allow student to approach or select one item.
Circle the item number of the selected item (the first item touched) on the data
sheet (for example, circle 2 if item 2 was chosen over item 3). If the student
simultaneously approaches both items, block access to both items.
5. Remove the nonselected item from the student’s reach. Allow the student to
consume or have thirty seconds of interaction with the selected item.
6. If the student does not approach or select either item within five seconds,
verbally prompt the student to sample each item. After the sampling period,
present the two items again as directed in steps 2 through 4. If the student still
does not approach or select either item, remove both items, circle N on the data
sheet, and begin the next trial.
7. Repeat steps 3 through 6 until each item has been paired twice with every other
item.
The following are the instructions for scoring the paired stimulus preference
assessment:
1. Record the number of times each item was selected by totaling the number of
circles in each column of the data sheet.
2. Divide the number of times each item was selected by the number of times it
was presented in the SPA (each item is available eight times in a five-item assess-
ment) and multiply by 100 to get a percentage.
Chapter 1 9
Paired Stimulus Preference Assessment Data Sheet
Student: Assessed by: Date: Time
Overall rank
Stimulus items: (List largest percentage first):
1.
2.
3.
4.
5.
Record item with corresponding item number: Circle item selected:

1. 2. 1 2 3 4 5 N
5. 4. 1 2 3 4 5 N
3. 1. 1 2 3 4 5 N
2. 4. 1 2 3 4 5 N
4. 5. 1 2 3 4 5 N
3. 2. 1 2 3 4 5 N
1. 5. 1 2 3 4 5 N
3. 4. 1 2 3 4 5 N
5. 1. 1 2 3 4 5 N
1. 4. 1 2 3 4 5 N
2. 3. 1 2 3 4 5 N
3. 5. 1 2 3 4 5 N
4. 2. 1 2 3 4 5 N
5. 2. 1 2 3 4 5 N
4. 3. 1 2 3 4 5 N
2. 5. 1 2 3 4 5 N
1. 3. 1 2 3 4 5 N
4. 1. 1 2 3 4 5 N
5. 3. 1 2 3 4 5 N
2. 1. 1 2 3 4 5 N
Times selected:
1. ÷ x 100 = %

2. ÷ x 100 = %
3. ÷ x 100 = %
4. ÷ x 100 = %
5. ÷ x 100 = %

3. Rank the items based on the percentages, putting the largest percentage first,
the next largest second, and so forth.
4. Items selected during 80 percent or more of trials are most likely to function as
reinforcers.
Multiple Stimulus Without Replacement Method

In an attempt to reduce the amount of time required for preference assessments,
researchers have investigated the effects of offering choices between multiple items or
activities at the same time rather than in pairs. One way to present multiple items has been
termed the multiple stimulus without replacement (MSWO) method. This method was first
developed by DeLeon and Iwata (1996) and later streamlined by Carr, Nicholson, and
Higbee (2000). In a MSWO assessment, multiple (usually five through eight) items or
activities are presented simultaneously. The items or activities are typically placed, equally
spaced, in a row (often called a stimulus array) in front of the student. The student is then
allowed to choose between the items or activities. The item touched first is scored as the
selection and the individual is allowed to consume or interact with the item or activity
for a brief period of time. The selected item is then removed from the stimulus array
and the remaining items are resequenced and the individual is allowed to make another
selection. This process continues until all items have been selected or no item is selected
within a brief period of time (usually five to ten seconds). Usually, this entire process is
repeated three (Carr et al., 2000) to five (DeLeon & Iwata, 1996) times, although com-
parable results may be obtained in some cases with only completing the selection process
once (Carr et al., 2000). A selection percentage is calculated for each item or activity in
the same fashion as the paired stimulus method, where the number of times an item or
activity is selected is divided by the number of times an item or activity was available
for selection and then multiplied by one hundred. Items are then ranked according to
selection percentage. In contrast to selection percentages obtained from paired stimulus
preference assessments, which can be used to predict reinforcing effectiveness (because the
denominator of the ratio generally remains constant for all items or activities), selection
percentages obtained from MSWO assessments cannot be used to predict reinforcing
effectiveness because the order in which item selections take place can influence selec-
tion percentages. For example, if the same item or activity was selected first during each
of three stimulus arrays, its selection percentage would be 100 percent (selected three
times ÷ available three times = 100%). In this example, if another item or activity was
selected second during each of three stimulus arrays, its selection percentage would be
50 percent (selected three times ÷ available six times = 50%). Thus, selection percent-
ages in the MSWO procedure are used only for ranking stimuli. Items ranked first in
MSWO preference assessments are most likely to function as reinforcers (Carr et al.,
2000; Higbee, Carr, & Harrison, 2000). Data obtained by Higbee, Carr, and Harrison
(1999) also suggest that items ranked second and third also may function as reinforcers
in many cases.
DeLeon and Iwata (1996) compared the MSWO assessment to the paired stimu-
lus procedures developed by Fisher and colleagues (1992). They conducted both types
Chapter 1 11
of preference assessments using the same set of items or activities and with the same
participants. In general, they found that the MSWO preference assessment produced
results comparable to those of the paired stimulus assessment but in about half the time.
Carr et al. (2000) attempted to reduce the amount of time required to complete the
MSWO assessment by reducing the number of stimulus arrays from five to two. They
conducted these “brief” MSWO procedures with three students with autism and then
examined the reinforcing effectiveness of items or activities identified as being high-,
medium-, and low- preference by the brief MSWO, by delivering these items or activities
contingent on student academic behavior. They found that the brief MSWO procedure
accurately predicted reinforcer effectiveness as contingent delivery of high-, medium-, and
low-preference reinforcers produced responding that corresponded to the degree of pref-
erence. In a secondary analysis, Carr and colleagues (2000) calculated correlation coef-
ficients for the stimulus rankings produced by student selections in the first stimulus array
compared to the rankings produced by the combined results of the three arrays and found
that the correlations were high, indicating that conducting an MSWO preference assess-
ment with one stimulus array may be sufficient to accurately rank items or activities. The
authors reported that the brief MSWO assessments could be completed in ten minutes
or less when three stimulus arrays were used. The time could be further decreased if only
one stimulus array was used.
The brief MSWO assessment data sheet shown on the next page can be used to
record and analyze the data from the assessment.
Below are guidelines for conducting the brief MSWO preference assessment (Carr et
al., 2000):
1. Identify four items or activities that the student has requested in the past or
has been observed to interact with during free-choice times, as well as one new
item or activity. If edibles are in the array, break them up into small bite-size
pieces before presenting them to the student. If a drink is offered, present only
a small amount in the cup so the student can drink this amount quickly. If toys
or activities are used, make sure that interacting with them will be meaningful
even if it only occurs briefly.
2. Allow the student to briefly sample each item by allowing him or her to eat
or drink a small portion of edibles or briefly (for example, for ten to fifteen
seconds) engage with nonedibles.
3. Place the items on the table or desk in front of the student with equal distance
between them.
4. Provide a brief instruction, such as “Pick the one you w ant,” to the student, and
allow him or her to choose one item. If the student attempts to grab more than
one item, block access to the other items. You can do this by either pulling the
table or desk out of the student’s reach or quickly removing all of the nonchosen
items. Write the number next to the item on the data sheet according to the
order in which it was chosen (for example, write a “1” next to “soda” if soda was
chosen first).

Brief MSWO Preference Assessment Data Sheet
Rank by Trial
Overall Rank (list
Stimulus Items 1 2 3 Sum of 1, 2, and 3 smallest sum first)

Rank by Trial
Overall Rank (list

Rank by Trial
Overall Rank (list

Chapter 1 13
5. When the student has finished consuming the edible or the specified period
of interaction has elapsed for engagement with the nonedible item or activity,
remove the nonedible item or activity and place it out of sight. Arrange the
remaining four items as in step 3 and center them in front of the student.
6. Steps 3 through 5 will be repeated until all items have been selected and no
items are left, or until the student does not select an item within ten seconds. If
the student fails to select an item within ten seconds, score all of the remaining
items as “5” on the data sheet.
7. Repeat the entire procedure two additional times using the same items (step 2 is
not necessary after the first array). Record data in column 2 for the second array
and in column 3 for the third array. If you find that the student is responding
in similar fashion to all three stimulus arrays, you may be able to shorten the
procedure to one array in subsequent assessments.
The following are the instructions for scoring the the brief MSWO preference
assessment:
1. Add the ranks for each item in columns 1, 2, and 3 and then record this number
in the “Sum of 1, 2, and 3” column.
2. Rank the items based upon the numbers in the “Sum of 1, 2, and 3” column,
with the smallest number being ranked first, the next smallest being ranked
second, and so on.
Answers to Common Questions About

Preference Assessments
Q: How often should I run the preference assessment?
A: Preferences have been shown to be relatively stable for some students and fluctuate
greatly for others (Carr et al., 2000). A conservative approach would be to conduct prefer-
ence assessments at least daily. It would be preferable to complete a preference assessment
multiple times per day, such as before each teaching session or when the student’s perfor-
mance starts to deteriorate, if possible.
Q: How do I choose the items to use in the preference assessment? Why should I
include a new item each time?
A: Watching what your student interacts with during free play is a good way to
select items for inclusion in the preference assessment. Informal interviews with parents
or other caregivers can also provide information about what to include in the assessment.
It is important to include new items so that the student is exposed to them during the
stimulus sampling procedure. Continuing to try new items in a search for new potential
reinforcers is also important.

Q: Is it okay to assess edibles and nonedibles together in the same preference
assessment?
A: Combining edibles and nonedibles in the same preference assessments may be
problematic in some cases, since some students tend to select edible items before noned-
ible items, even though the nonedible items may actually function as reinforcers (DeLeon,
Iwata, & Roscoe, 1997). Thus, if your student appears to be selecting all of the edible
items before the nonedible items, you may consider conducting separate preference assess-
ments for edibles and nonedibles.
Q: Do I need to use the actual items or activities in the preference assessment? Is it

okay to use symbols or pictures of the items or activities or conduct an entirely verbal
preference assessment?
A: Several studies have been conducted in an attempt to answer these questions.
In these studies, preference assessments have been conducted using pictures of stimulus
items or symbols, either representing these items instead of the items themselves (for
example, presenting pairs or arrays of pictures of potential reinforcers and asking students
to choose which one they would most like to earn as a reward for working) or presenting
potential reinforcers in a verbal forced-choice format (for example, “Would you like to
work for candy or music?”). Results of research on the use of verbal-, picture-, or symbol-
based preference assessments have been mixed, with some studies showing positive effects
(Graff & Gibson, 2003) and others not (Higbee et al., 1999). A critical variable appears
to be whether or not access to the chosen item or activity is provided immediately follow-
ing a selection response in the preference assessment. Preference assessments appear to be
more accurate when access to the chosen item is provided following a selection response
(Tessing, Napolitano, McAdam, DiCesare, & Axelrod, 2006). In summary, when pos-
sible, it is best to use the actual items or activities in preference assessments. Pictures,
symbols, and verbal representations in preference assessments should be used with caution
until further research determines the conditions under which they can most effectively
be used.
Summary: Strategies for Reinforcer Identification

The identification of effective reinforcers is a critical task for practitioners attempting to
provide instruction to students with autism and other developmental disabilities. Finding
reinforcers for children with autism can be particularly difficult due to the restricted
patterns of interest that often accompany the disorder. It is also particularly important,
however, because the most successful educational interventions for these children rely
heavily on procedures based on positive reinforcement. Students in intense behavioral
programs, including those using the procedures described in this book, are presented
with hundreds of learning opportunities per day. Powerful positive reinforcers are neces-
sary to provide motivation for learning and positive consequences for correct responding.
Stimulus preference assessments, like those discussed in the present chapter, are a valuable
tool for helping practitioners identify effective reinforcers for students with autism and
other developmental disabilities.
Chapter 1 15
Teaching Basic Learner Readiness Skills
Now that they have powerful and effective reinforcers in hand, the next task for practitio-
ners working with students with autism and other developmental disabilities and hoping
to ultimately establish derived relational responding skills is to start teaching basic readi-
ness skills. Before tackling such complex skills as language, social, and intellectual skills,
many students will need to learn prerequisite skills that will allow them to be active
participants in the learning process. These learner readiness skills include such behaviors
as sitting appropriately in a chair, attending to the instructor and instructional mate-
rials, and imitating simple responses. Variations on two basic behavioral techniques—
shaping or differential reinforcement and prompting—are typically employed to teach
these important prerequisite skills.
Differential Reinforcement and Prompting Techniques

Differential reinforcement is the act of delivering reinforcement for certain specific
responses while not reinforcing other responses. The most common form of differential
reinforcement that is used in behavioral education programs is called shaping. Shaping is
a specific differential reinforcement procedure where successive approximations to a target
behavior are differentially reinforced. Instead of expecting the student to perform the
final target behavior, which may be beyond the student’s capabilities at the moment, rein-
forcement is provided for performing an approximation or intermediate form of the target
behavior. Once the student can reliably produce the approximation of the target behavior,
the reinforcement criterion is changed so that the student must produce a behavior that
more closely resembles the target behavior. In this way, responding is gradually shaped
into forms that more and more closely resemble the target behavior until the ultimate
target behavior is reached. Shaping allows the student to come in contact with the rein-
forcer for the target response early and often and helps lead the student to the ultimate
target behavior. For example, if we were teaching a student to respond to the simple direc-
tion “Raise your hand” by fully extending his or her arm and hand upward (the target
response), one might initially reinforce any hand movement in response to the verbal
instruction. Once this behavior was occurring reliably, we might then require the student
to raise his or her hand to chest level with the hand open in order to receive reinforce-
ment. Finally, we could then require the complete action of fully extending the arm and
hand in response to the teacher instruction. If the student had difficulty with any of these
shaping steps, we could create smaller approximations.
Prompting is another instructional technique that is used to teach learner readiness
behavior. Prompting involves providing the student with additional help (prompts) beyond
the initial instruction, which promotes correct responding. Prompts come in all shapes
and sizes and fall into two basic categories: response prompts and stimulus prompts
(Cooper, Heron, & Heward, 2007). Response prompts are most common and include pro-
viding the student with additional verbal instructions, a model of the correct response,
or physical guidance or assistance. For example, if we were teaching a student to clap his
or her hands in response to the instruction “Clap your hands,” we could provide a model

prompt by clapping our hands following the instruction (and before the student makes
an error). The student could then imitate the response. Over time, we would fade out, or
gradually eliminate, the model prompt so that the student would respond to the instruc-
tion alone. One way to do this would be to use a time-delay procedure in which we
gradually introduced an increasing delay between our instruction and the model prompt
(Halle, Marshall, & Spradlin, 1979).
Stimulus prompts involve highlighting or changing some aspect of the instruction or
instructional stimuli provided to the student, such as using position prompts (where the
correct stimulus is placed closer to the student), or altering the instructional stimuli to
make the correct response more apparent. For example, when teaching a student to select
the appropriate letter from an array of three letter cards in response to the instruction
“Give me the letter A,” the A on the card could be made larger than the letters on the
other cards, with the size of the letter gradually reduced across instructional trials (but
see Doran & Holland, 1979; Sidman & Stoddard, 1967; Stoddard & Sidman, 1967;
Touchette, 1968). It should be noted that the prompting-and-fading procedure chosen
should be individualized for the particular student. When working with students with
severe disabilities, practitioners may find that one procedure (such as using added bright-
ness or size prompts that are then faded out gradually) is ineffective for a particular student
but effective for another. Moreover, responding may come under exclusive control of the
particular prompt that was added and it may not be clear that this has occurred until
errors occur in the final step(s) of fading when the prompt becomes unavailable. For this
reason, prompting strategies must be chosen with careful attention paid to the student’s
prior history with prompting systems and idiosyncracies in his or her attention to certain
relevant or irrelevant aspects of stimuli.
Teaching Appropriate Sitting Behavior

Sitting in a chair is often one of the first skills targeted in behavioral programs for
students with autism and other developmental disabilities. Not only is sitting an impor-
tant prerequisite behavior, but the procedures used for teaching it, physical prompting and
differential reinforcement, are relatively straightforward. Thus, it is often a “quick win”
for both the instructor and the student and allows the student to contact positive rein-
forcement quickly. It is important to note, however, that this is not always the case, and
that the physical prompting involved in teaching students to sit in a chair may cause some
students to engage in disruptive behavior such as aggression, screaming, and noncompli-
ance. Although the specific procedures that will be most effective in decreasing these
negative behaviors will vary for each individual student, the general strategy typically
involves shaping appropriate sitting by providing reinforcement for increasingly longer
periods of time where the student is sitting appropriately without engaging in disruptive
behaviors. What is initially most important is that the student obtains reinforcement
while engaging in appropriate sitting behavior (in other words, the reinforcer is delivered
before the student engages in negative behavior). The initial time intervals may be very
short (a few seconds) but can usually be increased relatively quickly if the reinforcer used
is powerful and the student regularly gains access to it.
Chapter 1 17
Below are guidelines for teaching sitting:
1. Sitting is usually taught by first placing two chairs facing one another (one for
the student and one for the teacher), one to two feet apart.
2. The student is guided to stand in front of his or her chair, facing the instructor
while the instructor is seated. To prevent tipping, the instructor may place the
student’s chair with its back against a wall or may use his or her legs and feet to
keep the chair on the ground.
3. The instructor then gives a verbal instruction such as “Sit” or “Sit down” and
then immediately provides a physical prompt to assist the student in sitting.
4. When the student sits appropriately, even if prompts were provided, reinforce-
ment is delivered. Reinforcement should be delivered while the student is
sitting.
5. Prompts should be faded as quickly as possible as the student becomes more
independent at sitting. The amount of reinforcement provided should roughly
correspond to the amount of independence the student demonstrates in sitting.
In other words, the less prompting required to get the student to sit, the higher
the quality and quantity of reinforcement that should be provided.
For more detailed information about teaching the skill of sitting, see chapter 9 in O. I.
Lovaas’s book Teaching Individuals with Developmental Delays (2003).
Teaching Eye Contact and Other Attending Behaviors

The ability to attend to an instructor and instructional materials is another impor-
tant prerequisite skill for students with autism and other developmental disabilities to
learn. Historically, attending behaviors have been explicitly taught by presenting verbal
cues such as “Look at me” and then providing differential reinforcement for increasingly
longer durations of eye contact (Lovaas, 1981; Taylor & McDonough, 1996, p. 74). When
students in these studies did not readily provide eye contact, increasing levels of prompt-
ing were provided, which included displaying a reinforcer (usually an edible item) and
then pulling the reinforcer up close to the instructor’s face. When the student provided
eye contact for the specified period of time, the reinforcer was delivered. This procedure
was typically repeated several times in succession.
Although procedures such as this can be effective in teaching students to provide eye
contact in response to a verbal instruction, they seem rather contrived. Most typically
developing three- and four-year-olds would not look at a parent for five to ten seconds in
response to their name or following an instruction such as “Look at me.” In fact, working
on eye contact in isolation may actually take valuable instructional time that could be
used for teaching other skills. More recent approaches to teaching eye contact and attend-
ing involve teaching these skills within the context of other instructional programs. For
example, instead of teaching the student over repeated trials to give eye contact for longer
and longer intervals in response to a verbal instruction (the “Look at me” approach),
many behavioral interventionists work on eye contact while they are concurrently teaching

imitation, receptive object identification, matching, and other tasks. The prompting pro-
cedures used in this contextual approach are often the same as in the “Look at me”
approach, although prompting procedures are often used instead of directly showing the
student a reinforcer to attract his or her attention. Such prompting procedures include
time delay (waiting for a few seconds for the student to give eye contact), using instruc-
tional materials to attract the student’s attention (placing a flash card or other object in
the student’s line of sight near his or her face and then slowly moving the object back near
the instructor’s face), and light physical prompting (lightly touching the student’s face and
then directing attention to the instructor’s face by the instructor pulling his or her hand
back toward his or her eyes).
The primary difference is in the consequences provided for giving eye contact. In
the “Look at me” approach, reinforcement is provided when eye contact is given. In the
contextual approach, when the student gives appropriate eye contact in the instructional
setting, he or she is presented with an opportunity to respond to a direction and then
reinforcement is provided, contingent upon his or her responding correctly to the direc-
tion. Thus, eye contact is conceptualized as a necessary part of the instructional sequence
rather than a separate skill to be taught in isolation. In addition to requiring eye contact
before presenting the student with an opportunity to respond to an instruction, practitio-
ners should also provide reinforcement when the student makes eye contact spontaneously
in the course of instruction or play. In many cases, simply requiring students to be either
attending to instructional materials or looking at the instructor before he or she gives an
opportunity to respond is sufficient to produce the necessary amounts of eye contact and
attending behavior.
Another contextually appropriate situation for teaching students to make eye contact is
in teaching students to mand for, or request, preferred edibles, objects, or activities. Mand
training takes advantage of naturally occurring student motivation to request preferred
items or activities (often edibles in early stages of training). The student is prompted to
make an appropriate communicative response via vocal speech, sign, or picture exchange
(depending on the communication level of the student) and then the requested item is pro-
vided contingent on the response. Once the student reliably makes requests, the instruc-
tor can add an eye contact requirement in addition to the communicative response. Thus,
in order to obtain the requested item, the student must provide eye contact while making
the communicative response. Using differential reinforcement, the instructor can teach
the student to provide eye contact when making requests.
Two important factors can influence student eye contact and attending behaviors:
motivation and pacing of instruction. If the instructor is providing frequent enough access
to powerful reinforcers contingent on correct student responding, the student will likely
be attentive to the instructor and materials because he or she is motivated to gain further
access to the reinforcers. If the instructor is providing reinforcement too infrequently
or if the items or activities being delivered are not potent reinforcers, the instructor will
likely be struggling to keep the student’s attention. Also, appropriately paced instruction
helps students to maintain attention to the instructor and instructional materials. Though
student attention can be lost when instruction is happening too rapidly, it is most often lost
when the pace of instruction is not rapid enough. Thus, a good strategy for maintaining
student interest and attention is to be sure to have potent reinforcers (identified using a
preference assessment), deliver these reinforcers on a rich enough schedule, and keep the
pace of instruction high.
Chapter 1 19
Below are guidelines for teaching and maintaining eye contact and attending
behavior:
1. Teach eye contact and attending within the context of other instructional
programs.
2. Before presenting an instruction, require the student to look at you or at the

materials.
3. Start with a time-delay prompt. Wait up to five seconds for the student to
provide eye contact on his or her own.
4. If the student does not readily provide eye contact, use instructional materials
or a light physical prompt to obtain the student’s attention.
5. When the student provides attention, immediately provide him or her with an
opportunity to respond to a direction.
6. Provide reinforcement when the student responds correctly to the direction

provided.
7. Also provide reinforcement when eye contact is provided spontaneously during

play or instruction.
8. Work on eye contact while teaching mands for preferred objects by requiring
students to provide eye contact when they are making requests.
9. Use high-quality reinforcers and keep the pace of instruction rapid enough.
Teaching Generalized Imitation Skills

Typically developing children learn many simple and complex skills by observing
and imitating the behavior of others. Many students with autism and related disorders,
however, do not readily imitate the behavior of others. Thus, teaching the student to
imitate a model is another important area of learner readiness programming. Because
model prompts are often used in the teaching of more advanced skills, learning to imitate
is critical for the development of more complex behaviors. The goal of imitation training
is to produce what has been called generalized imitation, where the student will reliably
imitate new behaviors without specific training on the modeled response (Lovaas, 2003;
Peterson & Whitehurst, 1971). Imitation skills are often broken down into two categories:
nonverbal (motor) imitation and verbal imitation.
Nonverbal imitation is typically one of the first instructional programs implemented
with young children with autism. The only prerequisite skills required for participation in
nonverbal imitation training are the ability to sit in a chair and attend to the instructor,
although, as discussed previously, these skills can be taught concurrently during imitation
training. As is the case with other instructional programs, nonverbal imitation training
usually starts with simple motor responses (such as putting a block in a bucket or clapping
one’s hands) and then progresses to more complex behaviors (such as play and other social

behaviors) as the student acquires basic imitative skills. During initial imitation training,
many students have more success learning to imitate motor actions that involve objects,
such as putting a block in a bucket or hitting the table with a toy hammer, as opposed to
learning to imitate actions that only involve body movements, such as waving good-bye or
nodding the head. When teaching targets that involve the manipulation of objects, however,
it is important to also have distracter objects on the table to ensure that the student is imi-
tating the instructor’s behavior rather than simply learning that when the block and the
bucket are together on the table, the block goes in the bucket (Lovaas, 1981).
The typical teaching arrangement is similar to others that have been discussed: The
teacher and student are seated and facing one another. Often, there is a table for materials
that is either in between the instructor and the student or to the side. Each instructional
trial begins with the instructor providing a verbal cue such as “Do this” while simul-
taneously performing the motor action that the student is being asked to imitate. For
example, if the motor response involves placing a wooden block in a bucket, the instructor
states, “Do this,” while dropping the block into the bucket. The student is then physically
prompted to imitate the motor response and reinforcement is provided when he or she
does so. As with other teaching procedures discussed previously, the amount and quality
of reinforcement should correspond to the quality of the behavior. Thus, more-independent
responses should produce higher-quality reinforcement. In some cases, it may be necessary
to have a second instructor prompt the student while the primary instructor models the
correct response. Over time, prompts are faded until the student is reliably imitating the
motor response. The amount of target responses being concurrently taught will vary from
student to student and may be as small as one or as high as four or five responses. As target
responses are mastered, new targets are then introduced. When several motor responses
have been learned, untrained items should be presented to test for generalized imitation.
Verbal imitation training teaches the student to imitate a verbal model (such as a
sound, word, or sentence) rather than a motor response. By teaching verbal imitation, the
teacher can bring the student’s vocal behavior under instructional control. This means
that the student will produce sounds in response to the model provided by the instruc-
tor rather than for other reasons, such as self-stimulation. According to Lovaas (2003),
bringing the student’s vocalizations under the control of the teacher’s model and external
reinforcement accomplishes two goals: it puts the teacher in a better position to shape the
student’s vocal behavior into appropriate speech, and it can teach the student that vocal
sounds can be used to produce reinforcers from the social environment, which may facili-
tate the development of appropriate social behaviors that replace tantrums and other inap-
propriate behaviors. Before beginning verbal imitation training, the student should be
able to sit in a chair and attend to an instructor. While generalization between verbal and
nonverbal imitative repertoires should not be unequivocally expected, it may be helpful
for individuals beginning verbal imitation training to have demonstrated mastery of many
nonverbal imitation items, particularly those that involve oral-motor movements.
Verbal imitation training procedures are similar to those used in nonverbal imita-
tion training except that, instead of presenting models of motor actions, they use verbal
models of sounds, words, and eventually sentences. A primary difference in the teaching
procedures for verbal imitation has to do with the way target responses are prompted.
While physical prompting is the primary strategy with nonverbal imitation, only verbal
prompting methods can be used in teaching verbal imitation, since there is no way to
physically prompt a verbal response. A typical verbal imitation teaching trial involves the
Chapter 1 21
instructor giving a verbal instruction such as “Say ‘ball.’” For students who engage in
echolalia, or the repetition of vocalizations, the initial instruction of “Say” may be omitted
or stated with less volume so that the student does not repeat it. Reinforcement is then
provided for the correct imitation of the model. Just as is done with nonverbal imita-
tion, verbal imitation training starts with simple responses such as sounds and progresses
to more complex responses such as words and sentences. Initial verbal imitation train-
ing often involves bringing sounds that the student already produces, such as humming
and babbling, under instructional control. Before beginning verbal imitation training, a
teacher may find it helpful to observe the student and record the sounds that he or she
produces repeatedly. Then the teacher can attempt to bring one or more of these sounds
under instructional control by modeling it and providing reinforcement when the student
produces the sound in response to the model.
The following are guidelines for teaching generalized imitation:
1. Choose target behaviors that match the motor or verbal abilities of the student.
In nonverbal imitation training, many individuals find it easier to learn non-
verbal imitative behaviors that involve the simple manipulation of objects rather
than those that only involve the movement of body parts. In verbal imitation
training, consider starting with sounds that the student already produces and
bringing them under instructional control.
2. Sit across from the student and have a table available for materials if the target
response requires them.
3. If the target response involves the manipulation of an object, include distracter

objects on the table.
4. Before presenting an instruction, require the student to look at you or at the

materials.
5. Give the verbal instruction “Do this” while simultaneously performing the
motor response for nonverbal imitation. For verbal imitation, give the verbal
instruction “Say ” followed by the sound, word, or sentence
that the student is to imitate.
6. For nonverbal imitation, provide physical prompting as necessary to help the
student imitate the motor response. In some cases, a second person may be
required to provide prompting during the initial stages of imitation training.
7. Provide reinforcement for correct imitation of the motor or verbal response
(in general, the more independent the response, the greater the quality of the
reinforcer).
8. Fade prompts as quickly as possible.
9. Test for generalized imitation periodically (after the student has mastered several
targets) by presenting novel targets.
For more detailed information about teaching imitation, see chapters 13 and 22 in O.
I. Lovaas’s book Teaching Individuals with Developmental Delays (2003).

Summary: Teaching Basic Learner-Readiness Skills
Through prompting and differential reinforcement, students with autism and related dis-
abilities can learn to sit, attend to instructional materials, make eye contact with their
instructors, and imitate simple oral and gross motor movements. Often, these skills can
be taught within the context of other instructional programs. These skills, while useful
in their own right, have more importance because they prepare students to learn more-
complex language, social, and intellectual skills, including those that may emerge from a
reinforced history of relational responding, as will be discussed in the coming chapters.
References
Carr, J., Nicholson, T., & Higbee, T. (2000). Evaluation of a brief multiple-stimulus
preference assessment in a naturalistic context. Journal of Applied Behavior Analysis,
33, 353–357.
Cooper, J., Heron, T., & Heward, W. (2007). Applied behavior analysis (2nd ed.). Upper
Saddle River, NJ: Pearson Merrill Prentice Hall.
DeLeon, I., & Iwata, B. (1996). Evaluation of a multiple-stimulus presentation format for
assessing reinforcer preferences. Journal of Applied Behavior Analysis 29, 519–533.
DeLeon, I., Iwata, B., & Roscoe, E. (1997). Displacement of leisure reinforcers by food
during preference assessments. Journal of Applied Behavior Analysis, 30, 475–484.
Doran J., & Holland, J. G. (1979). Control by stimulus features during fading. Journal of
the Experimental Analysis of Behavior, 31, 177–187.
Fisher, W., Piazza, C., Bowman, L., Hagopian, L., Owens, J., & Slevin, I. (1992). A
comparison of two approaches for identifying reinforcers for persons with severe and
profound disabilities. Journal of Applied Behavior Analysis, 25, 491–498.
Graff, R., & Gibson, L. (2003). Using pictures to assess reinforcers in individuals with
developmental disabilities. Behavior Modification, 27, 470–483.
Halle, J. W., Marshall, A. M., & Spradlin, J. E. (1979). Time delay: A technique to
increase language use and facilitate generalization in retarded children. Journal of
Higbee, T., Carr, J., & Harrison, C. (1999). The effects of pictorial versus tangible
stimuli in stimulus preference assessments. Research in Developmental Disabilities, 20,
63–72.
Higbee, T., Carr, J., & Harrison, C. (2000). Further evaluation of the multiple-stimulus
preference assessment. Research in Developmental Disabilities, 21, 61–73.
Lovaas, O. I. (1981). Teaching developmentally disabled children: The me book. Austin, TX:
PRO-ED.
Lovaas, O. I. (2003). Teaching individuals with developmental delays. Austin, TX:
PRO-ED.
Chapter 1 23
Pace, G., Ivancic, M., Edwards, G., Iwata, B., & Page, T. (1985). Assessments of stimu-
lus preference and reinforcer value with profoundly retarded individuals. Journal of
Peterson, R. F., & Whitehurst, G. J. (1971). A variable influencing the performance of
generalized imitative behaviors. Journal of Applied Behavior Analysis, 4, 1–9.
Sidman, M., & Stoddard, L. T. (1967). The effectiveness of fading in programming a
simultaneous form discrimination for retarded children. Journal of the Experimental
Analysis of Behavior, 10, 3–15.
Stoddard, L. T., & Sidman, M. (1967). The effects of errors on children’s performance
on a circle-ellipse discrimination. Journal of the Experimental Analysis of Behavior, 10,
261–270.
Taylor, B., & McDonough, K. (1996). Selecting teaching programs. In C. Maurice,
G. Green, & S. Luce (Eds.), Behavioral intervention for young children with autism.
Austin, TX: PRO-ED.
Tessing, J., Napolitano, D., McAdam, D., DiCesare, A., & Axelrod, S. (2006). The effects
of providing access to stimuli following choice making during vocal preference assess-
ments. Journal of Applied Behavior Analysis, 39, 501–506.
Touchette, P. E. (1968). The effects of graduated stimulus change on the acquisition of a
simple discrimination in severely retarded boys. Journal of the Experimental Analysis
of Behavior, 11, 39–48.

Chapter 2
The Assessment of Basic Learning Abilities

(ABLA) and Its Relation to the Development
of Stimulus Relations in Persons with Autism
and Other Intellectual Disabilities
W. Larry Williams and Marianne L. Jackson,

University of Nevada, Reno
The objective of this chapter is to acquaint the reader with a practical clinical tool, the
Assessment of Basic Learning Abilities (ABLA). This assessment informs the teacher
or therapist working with children with intellectual disabilities about the appropriate
form and difficulty level of the materials and methods for teaching basic concepts and
communication skills.
Discrimination Learning
Learning to discriminate between relevant stimuli in the environment is crucial to learn-
ing many functional skills, including communication and social skills, and is an assumed
skill in many psychometric tests. Matching to sample (MTS) is one of the most com-
monly used paradigms for teaching and assessing discrimination skills and is also the
training protocol commonly used to assess relational responding for communication skills,
elementary reading, and equivalence relations (see chapter 8 of this volume). Therefore,
if individuals do not demonstrate the ability to match to sample, it renders them basi-
cally untestable on many psychometric tools. MTS is also frequently used in many class-
room and preacademic training settings and has been instrumental in communication
and language training. Teaching an individual to match to sample appears to be basic for
increasing the array of basic concepts that an individual might learn.
There are a number of types of discriminations that are relevant and even vital to
many educational skills. These discriminations could be described as existing on a con-
tinuum of difficulty or complexity. Nonrelational discriminations may represent some of
the least complex discriminations, and arbitrary conditional discriminations could be said
to represent a more complex type of discrimination.
Nonrelational discriminations involve a simple, simultaneous discrimination that does
not require the presence of a sample stimulus. A basic relational discrimination is said to
have occurred when a learner comes to respond to the presence of a given environmental
event or stimulus (discriminative stimulus) and does not respond in the absence of that
event or stimulus. For example, a simple, nonrelational discrimination would be learn-
ing that in the presence of a plate the correct response is to put it in the dishwasher. In
the case of a simple discrimination, this response will occur regardless of whether or not
the plate is dirty and needs to be washed. The presence of the plate itself functions as the
stimulus for the response of placing it in the dishwasher.
In relational responding, the individual must attend to two stimuli and respond on
the basis of their relationship (in MTS, the sample and correct comparison). A conditional
discrimination requires that the function of the comparison stimuli change from trial to
trial depending upon the sample stimulus, such that a given stimulus is presented over
trials as both the correct and incorrect comparison. This type of discrimination represents
a logical if-then rule. Following on the previous example of a simple discrimination, this
would represent a conditional discrimination if the individual responded by only putting
a used or dirty plate (sample stimulus) in the dishwasher but not putting a clean plate
in the dishwasher. A simpler example of a conditional discrimination would be match-
ing a red sample to a red comparison and matching a blue sample to a blue comparison
stimulus.
Conditional discriminations may occur within and across any sensory modality,
including visual, auditory, olfactory, and tactual. Moreover, they may involve the fea-
tures of formal physical similarity, referred to as nonarbitrary conditional discrimina-
tions, or they may be formally dissimilar and thus be completely arbitrary conditional
discriminations.
The ABLA and Conditional Discriminations

The Assessment of Basic Learning Abilities (ABLA), which was developed by Kerr,
Meyerson, and Flora in 1977, tests an individual’s ability to form some of these various
types of discriminations. The ABLA is typically administered to individuals with some
degree of intellectual disability and examines the ease or difficulty with which an indi-
vidual can learn one simple motor task and five two-choice discrimination tasks. The
ABLA is a short test that can be administered relatively easily and quickly using a small
number of stimuli.
Performance on the ABLA provides information that allows inferences to be drawn
regarding individuals’ prior histories with such tasks. These inferences can be drawn

largely because the tests are short and hence provide few opportunities for learning to
occur throughout the assessment. In addition, research shows that levels at which the
learner initially fails are very difficult for them to ultimately master, often requiring a
large number of trials (Meyerson, 1977).
ABLA Procedures and Materials

The ABLA consists of six levels of testing, each examining different discrimination
abilities. Testing at each level begins with a demonstration trial, a guided trial, and a
practice trial, and reinforcement is provided for every independent correct response. The
passing criterion for each level is eight consecutive independent correct responses; the
failing criterion is eight cumulative errors on a given level. Each progressive level presents
a more complex type of discrimination and builds on the skills tested at previous levels.
Testing is terminated upon the failure at any level, because research has shown that an
individual who fails a particular level will not pass higher levels (Kerr, et al., 1977).
The materials used in the ABLA are a large yellow can of approximately 6 inches in
diameter and 7 inches in height, a large red box approximately 5.5 by 5.5 by 4 inches, a
small yellow cylinder approximately 1.5 inches in diameter and 3 inches in height, a small
red cube with approximate dimensions of 2 by 2 by 2 inches, and two small, shapeless
pieces of gray foam approximately 2 inches in diameter.
All correct responses should be reinforced with items that are appropriate for that
individual. These should be assessed prior to testing. Incorrect responses should result
in a neutral “no” from the tester and no other differential consequence, followed by the
prompting hierarchy outlined in Figure 2.1, which depicts the operationalization of each
instructional trial. It provides descriptions of the appropriate consequences for correct
and incorrect responses and the prompting hierarchy as necessary for error correction.
Table 2.1 delineates the relevant instructional materials, instructions given to the learner,
and response requirements for each level. By using the materials and procedures listed in
table 2.1 and the flowchart in figure 2.1, a variety of individuals should be able to use the
ABLA as a simple and relatively quick assessment tool.
Level 1
In level 1 the tester presents one of the two containers, either the yellow can or the
red box, on the table in front of the individual. Both the tester and the individual being
tested have a piece of gray foam and the tester gives the instruction “Do this” while
placing his or her piece of foam in the container. The individual responds correctly when
he or she places his or her piece of foam in the same container. This procedure is then
conducted with the other container so as not to develop differential histories with respect
to each. Passing and failing criteria are as described above.
Level 2
In level 2 the tester presents both the red box and the yellow can on the table, with the
yellow can on the individual’s left side and the red box on the right. The individual being
Chapter 2 27
tested is given a small piece of gray foam and is provided with the standard demonstration
trial, a guided trial, and a practice trial before being asked for an independent response.
The instruction given is “Where does it go?” and a correct response is the placement of
the foam in the yellow container, on the individual’s left side. This level tests the indi-
vidual’s ability to learn a positional discrimination.
Level 3
Level 3 is similar to level 2, with the exception that the positions of the two contain-
ers are randomly alternated between trials. During any given trial the correct response is
the placement of the piece of gray foam in the yellow can, regardless of its position on the
table. This tests the individual’s ability to learn a simple discrimination.
Level 4
Level 4 presents the first test of conditional discriminations. Both the yellow can
and the red box are present on the table in front of the participant and their position
alternates randomly between trials. The participant is given either the small yellow cyl-
inder or the small red cube and the instruction “Where does it go?” After being pro-
vided with a demonstration trial, a guided trial, and a practice trial on both stimuli, the
individual responds correctly by placing the small red cube into the large red box, or
the small yellow cylinder into the large yellow can. The presentation of the small yellow
cylinder or small red cube alternates randomly between trials. This level tests the indi-
vidual’s ability to learn a visual-visual conditional discrimination (sometimes referred to
as a quasi-identity match, since the sample and correct comparison are formally similar
but not identical).
Level 5
Testing in level 5 introduces cross-modal discriminations. These discriminations
involve samples and comparisons from different modalities, specifically visual and audi-
tory, whereas all previous levels tested discriminations within the visual modality. In level
5 both the yellow can and the red box are placed on the table and their positions do not
vary during this level. The individual is given a small piece of gray foam and asked to “put
it in the red box” (said in a rapid or staccato fashion) or “put it in the “yehlllloooow…
caaan” (said in a slow, drawn-out fashion). This assesses the individual’s ability to learn a
simple auditory-visual discrimination.
Level 6
Level 6 is similar to level 5 with the exception that the yellow can and red box alter-
nate positions randomly on each trial. The instructions given are the same. This assesses
the individual’s ability to perform a conditional auditory-visual discrimination. As stated
previously, testing is terminated upon the failure of any particular level.

Table 2.1. ABLA Materials and Methods
Level Materials Instruction Correct Response
1 Yellow can or red box in “Do this” (as the Individual picks
center of table (but only assessor picks up a up his or her
one of them on any given piece of gray foam piece of foam
trial). and places it in and places it in
the container on the container on
Two pieces of gray foam
the table, either the table.
(one in front of the assessor
the yellow can or
and one in front of the
the red box)
individual being tested).
2 Yellow can and red box, “Where does Individual

with yellow can positioned it go?” (as the takes the piece
to the left of the individual piece of foam is of foam and
being tested and the red presented to the places it in the
box to the right. Both will individual) yellow can,
be at an equal distance which is always
from the individual. positioned to his
or her left.
Positions of the yellow can
and red box do not change
over trials.
One piece of foam is
handed to the individual to
be tested.
3 Yellow can and red box, “Where does Individual takes

both positioned in front of it go?” (as the the piece of
the individual and at equal piece of foam is foam and places
distances. presented to the it in the yellow
individual) can (regardless
At this level the left and
of its left or
right positions of these
right position on
stimuli will alternate
the table).
randomly on each trial,
so starting position is not
important.
One piece of foam is
handed to the individual
being tested.
Chapter 2 29
4 Yellow can and red box, “Where does it When presented
both placed in front of go?” (as either the with the yellow
the individual at equal yellow cylinder cylinder, the
distances. or red cube is correct response
presented to the is to place it in
Again, at this level their
individual) the yellow can.
positions will alternate
randomly. When presented
with the red
One small yellow cylinder
cube the correct
and one small red cube
response is to
(only one presented to
place it in the
the individual on any
red box.
given trial).
5 Yellow can and red box “Put it in the Individual places

both placed in front of yehlllloooow… the foam in the
the individual at equal caaan,” said in a container that
distances. slow voice with corresponds to
a hand covering the instruction.
Both stay in the same
the mouth.
position on each trial.
“Put it in the
One piece of gray foam
red box,” said
is placed in front of the
quickly with a
participant.
hand covering
the mouth
6 Yellow can and red box “Put it in the Individual places

both placed in front of yehlllloooow… the foam in the
the individual at equal caaan,” said in a container that
distances. slow voice with corresponds to
a hand covering the instruction.
The positions of the
the mouth.
containers alternate
randomly on each trial. “Put it in the
red box,” said
One piece of gray foam
quickly with a
is placed in front of the
hand covering
participant.
the mouth

Demonstration trial:
“When I say (insert appropriate instruction
for level), you do this (demonstrate correct
response).”
Guided trial:
“Now let’s do it together.”
Give instruction for that level, then take the
individual’s hand and guide him or her
through the correct response.
Independent trial:
“Now you try.” Give instruction for that level
and then allow 3-5 seconds for response.
No response or incorrect
Correct response
Begin testing
Give instruction for level and

allow 3-5 seconds for response
No response or incorrect Correct response
Score as incorrect
Score as correct
Demonstration trial
Level is passed after 8
consecutive correct
Guided trial responses
Independent trial
Begin next level
No response or incorrect
Testing ends
Level is failed after 8 cumulative errors
Figure 2.1. ABLA flowchart.
Chapter 2 31
Common Problems and Suggested Strategies
Given the range of individuals who may be assessed using the ABLA, there are some
common problems that may arise. Most of these will occur during the earlier stages of
testing and are dealt with before higher levels of testing. Furthermore, it is often the case
that individuals who progress to testing on higher levels (such as levels 4, 5, and 6) have
experienced a variety of similar training or testing situations, so these issues may not
arise.
In the early stages of testing, there are three common problems. The first is that indi-
viduals may not respond at all to the materials or the instructions given. In this situation,
the assessor should implement the prompting hierarchy described in the flowchart. The
assessor will first provide a demonstration of the correct response, guide the individual to
make the correct response, and then repeat the instruction and provide an opportunity
for the individual to perform the response independently. If the individual responds cor-
rectly, the response will be reinforced appropriately and the next trial will begin. If the
individual does not respond correctly or fails to respond, the assessor will record this as
one incorrect response and repeat the prompting hierarchy. Each time the final step in
the prompting hierarchy is implemented (in other words, an opportunity to make an
independent response), a correct or incorrect response will be recorded. This cycle will
continue until the passing or failing criteria are met.
Another common problem encountered in the earlier stages of assessment occurs
when individuals throw materials or try to hand them to you. Throwing of materials
should be considered an incorrect response, and be followed by a neutral “no,” rapid
replacement of materials, and implementation of the described prompting hierarchy. The
assessor should not directly address the throwing of materials in any other way. In the
event that the individual tries to give the assessor the materials, the assessor should not
accept the materials but should, again, respond with a neutral “no” and implementation
of the prompting hierarchy.
One final problem often encountered is that the individual demonstrates a position
bias, meaning that he or she constantly responds to any stimulus presented in a specific
position (the right or left side) regardless of the task or instructions given. In addition,
the individual may respond to one of the stimulus items and not the other. Although
this may simply reflect a general lack of ability to form discriminations beyond level 2
or 3 (positional discrimination, or simple discrimination), the effect may be reduced or
avoided by using a simple procedural intervention early in testing. During level 1 it may
be advantageous to use one container (such as the yellow can) for some trials and the
other stimulus (such as the red box) for other trials. Changing the container used may
help prevent overselectivity of one container over the other in later levels of testing.
Although these are only a few examples of the possible errors that individuals may
display throughout testing, they represent some of the more common areas of diffi-
culty experienced. The strategies used to address these problems can also be applied to
almost any error that occurs. In general, re-presenting the trial and prompting a correct
response will be required. Assessors should be sure to record all incorrect responses and
independently made correct responses throughout these procedures.

A Predictive Hierarchy of Discrimination Abilities:
ABLA Findings
When Kerr, Meyerson, and Flora developed this test in 1977, they administered it to
117 individuals. Rising out of this project were a large number of research projects that
began to investigate the nature of these learning abilities and the practical implications of
them (for reviews see Martin & Yu, 2000, and Martin, Yu, & Vause, 2004). One of the
earliest findings of ABLA research was that the test discriminations were hierarchically
ordered according to their difficulty; the findings showed that failure at a particular level
predicted failure at higher levels of the hierarchy. Another finding of the 1977 study by
Kerr, Meyerson, and Flora was that failure at a level occurred relatively quickly and that
failed levels were extremely difficult to teach. Furthermore, this hierarchy of discrimina-
tions and the difficulty with which they can be learned has also been found to hold for
typically developing children.
ABLA as a Predictive Assessment

Research has shown that the ABLA not only predicts an individual’s ability to make
particular types of discriminations but is also highly predictive of success on educational,
prevocational, and vocational tasks (Stubbings & Martin, 1995). In fact, performance
on the ABLA was found to be a more accurate predictor of success on a variety of train-
ing tasks than the assessments of experienced teachers with direct knowledge of a given
individual.
Many studies have investigated the relationship between the ABLA levels and language
abilities. More specifically, it has been shown that individuals who passed ABLA levels
2 and below but failed higher levels typically had no formal verbal abilities. Individuals
who passed up to level 4 but failed higher levels were typically able to communicate
using single words or sounds. Language skills involving two or more words typically only
occurred in individuals who passed level 5 or 6 (Ward & Yu, 2000).
More recently, Richards, Williams, and Follette (2002) examined performance on
the ABLA and the communication, daily living, and social skills subdomains of the
Vineland Adaptive Behavior Scale (Sparrow, Balla, & Ceccetti, 1984). The Vineland
Adaptive Behavior Scale is a broad measure of an individual’s adaptive functioning in
these areas. They reported a positive correlation between these tests but found that the
Vineland was not sensitive to differences between individuals below ABLA level 6. Such
individuals showed low age-equivalent scores on the Vineland. Moreover, they found that
the Wechsler Adult Intelligence Scale-Revised (WAIS-R) not only failed to differentiate
individuals below ABLA level 6 but also scored them as untestable. This research suggests
that the ABLA may be a more sensitive and predictive assessment for individuals who
have severe impairments.
Marion and colleagues (2003) found that ABLA performance was a better predictor
of performance on assessments of verbal operants (mands, tacts, and echoics; Skinner,
1957) than diagnosis. Furthermore, their results suggested that ABLA level 6 may be a
prerequisite or bridging task for teaching such verbal operants.
Chapter 2 33
Further Refinements of the ABLA
The discriminations tested at levels 5 and 6 are arbitrary conditional discriminations,
but they are cross-modal because they involve auditory and visual stimuli. Until Sakko,
Martin, Vause, Martin, and Yu (2004) developed such a testing level, there was no exist-
ing test of arbitrary conditional visual-visual (or within modal) discriminations.
The visual-visual nonidentity match (VVNM) includes the same yellow can and red
box used in the ABLA, but the sample stimuli used are two three-letter words cut from
wood and thick enough to be described as three-dimensional. The three-dimensional
words bear no formal similarity to the yellow can and red box and so the relationship
between them is arbitrary and defined by the experimenter. All other procedures, includ-
ing prompting strategies and passing or failing criteria, are the same as those used in other
levels of the ABLA. Sakko et al. (2004) found that the VVNM applied after level 4 and
before level 5 of the standard ABLA hierarchy. However, Jackson (2006) suggests that
the ability to make nonidentity or arbitrary visual discriminations may represent a higher
level of difficulty than ABLA level 6. Further replication is needed on a larger scale in
order to resolve this issue and include the VVNM as a level of the ABLA hierarchy.
The Practical Utility of the ABLA for Programming Learning

Given the widespread use of simple and conditional discriminations that are formed
both within and across modalities, it is clear that the ABLA can be used as an assessment
tool to make decisions about programming for learning. Failure on a particular educa-
tional, vocational, or daily living task can often be readily explained by the use of the
ABLA. For example, an individual who is unable to complete a sorting task may be found
to fail ABLA level 4, because it involves making visual-visual identity or quasi-identity
conditional discriminations. Alternatively, an individual who does not respond correctly
to verbal instructions after many training trials may be found to fail ABLA levels 5 and
6. It then becomes obvious that the ABLA not only provides a picture of that individual’s
behavioral repertoire with respect to discrimination abilities involved in the task, but it
also suggests a course of action for remediation of such problems. This feature of the
ABLA in revealing repertoire limitations may be particularly relevant to descriptive and
experimental procedures for determining the function of challenging behaviors (Iwata,
Dorsey, Slifer, Bauman, & Richman, 1994). For example, if escape from task demands
is identified as the function of disruptive behavior, the ABLA may quickly demonstrate
that the individual cannot make auditory-visual discriminations and therefore cannot
comprehend spoken instructions. It then seems plausible to assume that an instructional
environment in which the individual does not understand the instructions, and has no
way to communicate this, may become unpleasant and the individual may attempt to
escape the situation. Although other functional assessment procedures will be useful in
such a situation, the ABLA will highlight such areas of difficulty and suggest a starting
point for programmed learning.
In the acquisition of new skills, an individual’s difficulty learning one particular skill
may initially seem puzzling, but further analysis of the discriminations involved and
assessment of the individual’s ABLA level may shed light on the problem. For example,
the failures of an individual who has experienced thousands of trials attempting to teach

a letter sound with a single letter printed on a flash card may initially seem strange;
however, the ABLA may reveal that the individual is not able to make auditory-visual
discriminations. This would clarify the nature of the problem, ruling out the possibility
that it could be a hearing or motivation problem and showing that it is actually a skill
building issue. Furthermore, the ABLA results will identify the individual’s level of com-
petence, and the next discrimination in the hierarchy that should be taught. In addition,
it may be appropriate for individuals to learn other skills that involve discriminations that
they are capable of, allowing them to be successful and acquire new skills within that
domain.
Other areas where the ABLA has general utility involve skills that seem comparable
in the natural environment but in actuality, when analyzed in view of ABLA level, are
not. For example, an individual may appear to be able to follow some simple spoken
instructions in certain situations but seem unable to respond to others. This may initially
seem like a problem of motivation, distraction, or a variety of other issues. However,
an analysis of ABLA level may reveal that the individual is only able to make visual
conditional discriminations and is in fact not able to make cross-modal auditory-visual
discriminations. Often further examination of the instructions to which the individual
responds correctly may allow the analyst to see that these are given in close proximity
to the individual, and with the addition of visual elements such as gestures and facial
expressions. When responding to these instructions the individual is able to use the visual
cues to make visual discriminations. The accuracy of these is of course a function of the
consistency of gestures and facial expressions across people and instances, although these
can be surprisingly consistent. The instructions to which the individual does not respond
correctly may be somewhat stripped of such visual features (for example, they are given
from a distance, the instructor is not looking at the individual, there is a lack of gestures
or facial expressions, or instructions are given using some type of speaker device where the
instructor is not physically present).
In all of these situations, the ABLA proves useful in at least four ways: First, it encour-
ages an analysis of tasks with respect to the discriminations that are actually involved.
Second, it provides a measure of areas of competence with types of discriminations and
tasks involving these areas of competence. Third, it shows types of discriminations with
which the individual can be expected to have difficulty. Fourth, it provides direction for
future learning. Given that these discrimination types appear as a hierarchy of complex-
ity, instructors should begin with the lowest level at which the individual was unsuc-
cessful and begin teaching there. Although research suggests that learning in these areas
may take thousands of trials, it may be effective and worthwhile for that individual.
Furthermore, such efforts may lead to the discovery of additional techniques to accelerate
this learning.
The ABLA and Equivalence Relations

As we have seen, the ABLA is a practical tool that involves some of the building blocks
for the formation of more complex stimulus relationships. The different discriminations
within the hierarchy of the ABLA can be described as (1) a simple motor response, (2)
a simple position discrimination, (3) a simple visual discrimination, (4) a quasi-identity
Chapter 2 35
conditional visual discrimination, (5) an auditory-position or auditory-visual conditional
discrimination, and (6) an auditory-visual conditional discrimination. (It is tempting to
consider level 1 a motor imitation response. For persons with that repertoire, that is prob-
ably an accurate description of the variables controlling their behavior. However, what
then would explain performance on the same difficulty of matching at level 4 if the
person fails that level? An alternative analysis is that the person without visual identity
matching skills passes level 1 due to the demonstration and repeated overcorrection for
error responses, together with the “errorless” feature of placing only one container on the
table. Both of these explanations could be tested empirically.)
Virtually all demonstrations of stimulus equivalence or other forms of derived rela-
tional responding involve conditional discriminations (see chapter 8 of this volume). The
original demonstrations of equivalence (Sidman, 1971; Sidman & Tailby, 1982) and the
great majority of studies that followed involved cross-modal (auditory-visual) conditional
discriminations. Indeed, a subsequent development was the controversy surrounding
the role of language ability for the formation of equivalence or other such derived rela-
tional responding (Hayes, 1989; Horne & Lowe, 1996, 1997; see also chapter 4 of this
volume).
Spoken language typically involves cross-modal equivalence relations emerging from
a history of cross-modal conditional discriminations. However, if language is not a nec-
essary condition for such equivalence relations, it follows that equivalence may be pos-
sible within other sensory modalities. Although the literature on this subject is sparse,
within-sensorial-modality equivalence has been demonstrated for tactile-tactile relations
(Belanich & Fields, 1999, visual-visual relations (Green, 1990), and auditory-auditory
relations (Sidman, 1994). The fact that equivalence is possible within one sensory modal-
ity and may not be dependent upon language suggests that perhaps it is simply condi-
tional discrimination performance that is necessary for the formation of derived stimulus
relations. With the necessity of language having been widely debated in the literature,
matching to sample is the most common training and testing technique in the stimulus
equivalence literature. The formation of conditional discriminations as the basis for a
variety of derived relations renders the ability to make such discriminations central to this
area. Although it is a straightforward conclusion, the relationship between equivalence
and conditional discriminations makes salient the findings of studies on the ABLA for
derived stimulus relations and language acquisition (Jackson, 2006; Jackson, Williams, &
Biesbrouck, 2006).
Future Research
The ABLA has prompted research that has produced consistent outcomes regarding a
hierarchy of discrimination skills, accurate prediction of performance on similar skills,
and a practical testing instrument for determining basic functioning levels. Nonetheless,
questions remain concerning the relationship among the discriminations represented in
the ABLA as well as the nature of discriminations beyond the ABLA level 6.
One feature of the ABLA that may require some clarification is the difference between
true visual-visual identity conditional discrimination performance and quasi-identity per-
formance. Most researchers have accepted this discrepancy in the ABLA level 4 materials

and, for the most part, have reported research participants’ responding as identity con-
ditional discrimination formation. However, we have worked with participants who can
match black and white materials identical in size but for whom a change of size or color
without detailed fading has resulted in errors.
In a similar vein, refinement is needed in comparing the discriminative abilities of
different populations when these populations are presented with human-produced audi-
tory cues as opposed to more simple pure tones for auditory-visual conditional discrimi-
nations and auditory-auditory conditional discriminations. This information is directly
relevant for procedures that involve computer presentation of tasks as opposed to tabletop
presentations with three-dimensional objects. Human speech as an auditory stimulus is
extremely complex compared to a simple auditory tone. In addition, it has been our expe-
rience that many lower-functioning individuals do not respond well or at all to computer
screen presentations of information; this is similar to the difficulty of some individuals
with responding to pictures of objects as opposed to actual objects (Dixon, 1981; Dixon
& Dixon, 1978).
Although some information is currently available due to the ABLA research on visual
and auditory discriminations, more research is needed on the relative positioning of
simple auditory skills, visual-visual nonidentity matching skills, auditory-auditory iden-
tity matching skills, and auditory-auditory nonidentity matching skills. Further research
on the role of other sensory modalities, such as tactile, proprioceptive, olfactory, and taste
discriminations, and their relative positioning, if any, with auditory and visual skills is
also in order. Such information would be relevant for the development of meaningful
relational responding repertoires as observed in the development of language, and many
social conventions involving those modalities.
Directly related to these issues is the need for clarification of the role of conditional-
ity in any or all of these discriminations and any hierarchical ordering observed across
sensory channels. To date, the major feature of the ABLA outcomes has been the appar-
ent necessity of ABLA level 4 skills in order to observe auditory-visual performances.
Whereas this has large practical implications for clinicians and educators, it also has gen-
erated a certain uneasiness in researchers and clinicians alike in its suggestion that basic
visual skills precede auditory skills. This possibility also appears to fly in the face of the
concept of dominance or at least preference of auditory discriminations over visual dis-
criminations as seen in developed adult language repertoires.
Finally, expansion of the ABLA to include discrimination skills beyond ABLA level 6
would also be of practical value. Although the current ABLA is predictive of testability on
standard intelligence tools (Richards et al., 2002), there remains a huge gap that needs to
be filled with respect to the nature of responding and its possible hierarchical complexity
between ABLA level 6 and the repertoires associated with lower and then higher intel-
ligence scores. The identification, for example, of “milepost” skills that could be demon-
strated to correlate with existing standardized measures of intellectual functioning would
represent the first empirical behavioral tools for the assessment of cognitive development.
Because these skills may be describable by more complex arrangements such as those
involved in relational responding and the formation of more complex relational networks,
this volume represents the first efforts in such a direction.
Chapter 2 37
References
Belanich, J., & Fields, L. (1999). Tactual equivalence class formation and tactual-to-visual
cross-modal transfer. Psychological Record, 49, 75–91.
Dixon, L. (1981). A functional analysis of photo-object matching skills of severely retarded
adolescents. Journal of Applied Behavior Analysis, 14, 465–478.
Dixon, M. H., & Dixon, L. S. (1978). The nature of standard control in children’s match-
ing-to-sample. Journal of the Experimental Analysis of Behavior, 30, 205–212.
Green, G. (1990). Differences in development of visual and auditory-visual equivalence
relations. American Journal on Mental Retardation, 95, 260–270.
Hayes, S. C. (1989). Nonhumans have not yet shown stimulus equivalence. Journal of the
Experimental Analysis of Behavior, 51, 385–392.
Horne, P. J., & Lowe, C. F. (1996). On the origins of naming and other symbolic behav-
ior. Journal of the Experimental Analysis of Behavior, 65, 185–241.
Horne, P. J., & Lowe, C. F. (1997). Toward a theory of verbal behavior. Journal of the
Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S. (1994).
Toward a functional analysis of self-injury. Journal of Applied Behavior Analysis, 27,
197–209.
Jackson, M. (2006). An examination of the relations between ABLA performance, language
ability, and within modal stimulus equivalence. Master’s thesis, University of Nevada,
Reno.
Jackson, M., Williams, W. L., & Biesbrouck, J. (2006). Conditional discrimination
ability, equivalence formation and mental retardation: Implications for development
in children with developmental disabilities. Journal of Speech Language Pathology and
Behavior Analysis, 1, 27–42.
Kerr, N., Meyerson, L., & Flora, J. A. (1977). The measurement of motor, visual, and audi-
tory discrimination skills [Monograph issue]. Rehabilitation Psychology, 24, 95–112.
Marion, C., Vause, T., Harapiak, S., Martin, G. L., Yu, C. T., Sakko, G., et al. (2003).
The hierarchical relationship between several visual and auditory discriminations and
three verbal operants among individuals with developmental disabilities. Analysis of
Verbal Behavior, 19, 91–105.
Martin, G. L., & Yu, C. T. (2000). Overview of research on the assessment of basic learn-
ing abilities test. Journal on Developmental Disabilities, 7, 10–36.
Martin, G. L., Yu, C. T., & Vause, P. (2004). Assessment of basic learning abilities test:
Recent research and future directions. In W. D. Williams (Ed.), Developmental dis-
abilities: Etiology, assessment, intervention and integration. Reno, NV: Context Press.
Meyerson, L. (1977). AVC behavior and attempts to modify it [Monograph issue].
Rehabilitation Psychology, 24, 119–22.

Richards, D. F., Williams, W. L., & Follette, W. C. (2002). Two new empirically derived
reasons to use the Assessment of Basic Learning Abilities. American Journal on Mental
Retardation, 107, 329–339.
Sakko, G., Martin, T. L., Vause, T., Martin, G. L., & Yu, C. T. (2004). A visual-visual
nonidentity matching assessment is a worthwhile addition to the Assessment of Basic
Learning Abilities test. American Journal on Mental Retardation, 109, 45–52.
Sidman, M. (1994). Equivalence relations and behavior: A research story. Boston: Authors
Cooperative.
Sidman, M., & Tailby, W. (1982). Conditional discrimination vs. matching to sample: An
expansion of the testing paradigm. Journal of the Experimental Analysis of Behavior,
37, 5–22.
Skinner, B. F. (1957). Verbal behavior. New York: Appleton-Century-Crofts.
Sparrow, S., Balla, D., & Ceccetti, D. (1984). Vineland adaptive behavior scales. Circle
Pines, MN: American Guidance Service.
Stubbings, V., & Martin, G. L. (1995). The ABLA test for predicting performance of
developmentally disabled persons on prevocational training tasks. International
Journal of Practical Approaches to Disability, 19, 12–17.
Ward, R., & Yu, D. C. T. (2000). Bridging the gap between visual and auditory discrimi-
nation learning in children with autism and severe developmental disabilities. Journal
on Developmental Disabilities, 7, 142–155.
Chapter 2 39
Chapter 3
Observing Responses: Foundations of

Higher-Order Verbal Operants
Dolleen-Day Keohane and Jo Ann Pereira Delgado,

Columbia University Teachers College and CABAS;
R. Douglas Greer,
Columbia University Graduate School of Arts
and Science Teachers College
Observing responses associated with listener and speaker repertoires is the foundation of
certain aspects of early language. Observing responses consist of the operant responses of
looking, listening, tasting, smelling, and touching. The observing operants are selected
out by the consequences that reinforce observation, and the stimuli that reinforce them
are established by reinforcement conditioning processes. These observing responses and
their reinforcers lead to the development of more complex behaviors (Donahoe & Palmer,
2004; Greer & Ross, 2008). Observing responses are critical to production responses,
both of which are inherent in a variety of cultural practices, including art, music, and lan-
guage. We have identified some of the subcomponents of language as a result of working
inductively toward a hierarchy of verbal development (Greer & Keohane, 2005; Greer &
Ross, 2008), drawing on Skinner’s verbal behavior theory and extensions of that theory
(Hayes, Barnes-Holmes, & Roche, 2001; Horne & Lowe, 1996; Skinner, 1957, 1989). In
this chapter we will concentrate on prelistener and early-listener capabilities, because they
lead to the emergence of new verbal capabilities, or cusps, in children with autism and
other developmental disabilities. Prelisteners are defined as children who do not observe
their environment, do not participate in the social community, and are completely depen-
dent on others for their very survival. Early listeners are children who have basic observing
skills, are able to participate in the social community to some extent, can follow a number
of directions, and are less dependent on others for their everyday needs.
We propose that the emergence of imitation through observation, conditioned rein-
forcement for listening to voices, looking at stimuli and print, and matching stimuli across
the senses or across sensory modalities may be prerequisites for the development of observ-
ing responses as related to early language acquisition. Such early language acquisition
occurs across repertoires of listener (for example, a child who follows simple directions),
speaker-listener (for example, a child with vowel-consonant auditory discrimination skills
that result in the production of speech), speaker-as-own-listener (for example, a child who
is able to speak and listen to himself or herself, as in “thinking”), and cross-modal capac-
ity for sameness (for example, a child who discriminates what is the same and what is
different across sensory modalities).
We have been on an applied behavior analytic journey of sorts, beginning with
Skinner’s theoretical framework of verbal behavior (1957) and the recent expansions of
that theory (Greer & Keohane, 2005; Hayes et al., 2001; Horne & Lowe, 1996). Along
the way we have incorporated theories, research, and practices related to the basic science
and, when relevant, the infant developmental literature and animal social learning theory.
The cumulative body of literature was very useful to us in our development of instruc-
tional protocols for children with significant language delays due to autism or other
developmental disorders. These protocols are based on a progression of complex language
functions or cusps (Baer, 1983; Hart & Risley, 1999; Premack, 2004; Rosales-Ruiz &
Baer, 1997). We also reviewed evidence that nonhumans could be taught certain noncom-
plex features of language (Epstein, Lanza, & Skinner, 1980, 1981; Premack & Premack,
2003; Savage-Rumbaugh, Rumbaugh, & Boysen, 1978), and that teaching, as distinct
from acquiring repertoires based on modeling, is unique to the human species (Premack,
2004). As we moved forward, it became increasingly clear that listener, speaker-listener,
and speaker-as-own-listener repertoires make complex verbal behavior possible and are
unique to humans.
As part of the process of developing a comprehensive systems-based behavior analytic
approach to teaching and learning over the last twenty-five years, the Comprehensive
Application of Behavior Analysis to Schooling (CABAS) team and colleagues in the
applied behavior analysis programs at Columbia University Teachers College have com-
piled instructional components based upon new research findings. The CABAS system
includes empirically based curricula and protocols for teaching new operants, the training
and motivation of teachers and other school professionals, and the design of functional
curricula (Greer, 2002). Our work has also allowed us to identify and remediate missing
developmental cusps in children with autism and related developmental disabilities.
Inherent in this system is the conceptualization of learn units, or instructional presenta-
tions that provide yoked or interlocking contingencies between teacher and child. Learn
units include (1) the need to know (for example, a motivating condition; (2) the child’s
attention to the relevant antecedent stimulus, such as the teacher’s instructions; (3) an
opportunity for the child to respond; (4) reinforcing consequences immediately follow-
ing correct responses; and (5) noncoercive corrections that immediately follow incorrect
responses and require the child to repeat the correct response (see Greer, 2002, and Greer
& Ross, 2008, for the extensive research base).
Our early-language training programs focus on self-awareness, or the ability to observe
ourselves, an essential component of language acquisition. When children begin to respond

to adult voices, make sustained eye contact with visual stimuli, learn to imitate through
observation, and match stimuli across the senses, they demonstrate early evidence of self-
awareness. They are able to distinguish between themselves and others in their environ-
ment. This distinction is elemental to all observing and producing responses and is the
basis for participating in such culturally evolved activities as music, art, or language.
Building Listener Literacy for Children with

Language Delays
In our initial efforts to bridge the educational gap for children with significant communi-
cation delays we focused on speaker repertoires, following Skinner’s lead (1957), or what
is often traditionally described as expressive language. It was a logical approach and was
closely tied to Skinner’s emphasis on the speaker, since the children we worked with were
frequently unable to speak. However, when the children began to display some speaker
behavior, it was often context specific, scripted, and directly taught. Even though our
assessments focused on both speaker and listener repertoires, including listener skills that
involved responding appropriately to the language of others (as in orienting toward or
identifying environmental stimuli labeled by others and complying with spoken instruc-
tions [Skinner, 1957]), we were much better able to measure and produce responses asso-
ciated with speaking skills.
As a result of our research and our success (as well as our lack of success) with indi-
vidual children, we reconsidered Skinner’s comments on the role of the listener. Skinner
described the relationship between the speaker and the listener as follows: “When the
listener looks to the speaker for an extension of his own sensory capacities, or for contact
with distant events, or for an accurate characterization of a puzzling situation, the speak-
er’s behavior is most useful to him if the environmental control has not been disturbed
by other variables” (Skinner, 1957, p. 418). This description provided us with a way to
reconceptualize the components of early-language development and design protocols for
assessment and intervention related to listener skills among children with autism and
related disabilities.
Classifying Children’s Prerequisite Foundations of

Verbal Behavior as Listener Developmental Cusps
As soon as we identified some of the capabilities that children needed in order to prog-
ress, these were organized into a hierarchy of verbal developmental milestones or cusps
(Greer & Keohane, 2005). This facilitated the identification of prerequisite skills and
the design of protocols to provide a means for children to advance across these cusps.
Compliance with adult instructions and the establishment of the teacher as a source of
conditioned reinforcement are prerequisites for the three earliest cusps and corequisite
cusp listed below.
Shown in tables 3.1 and 3.2 are listings of prerequisite and corequisite developmental
cusps and the instructional protocols used to establish them.
Chapter 3 43
Table 3.1. Prelistener: Very Early
Verbal Developmental Cusps
Verbal Developmental Cusp Protocol

Adult voices as conditioned reinforcers Conditioning listening to adult voices
Conditioned reinforcement for visual Conditioning visual stimuli as
stimuli reinforcement for tracking/print
The capacity for sameness across the Cross-modal sensory matching
senses
Table 3.2. Listener-Speaker:

Early Developmental Corequisite
Verbal Developmental Corequisite Protocol

Imitation The mirror procedure
Prelisteners are entirely dependent upon others for everything in their lives. Entrance
into the social community is not possible. Early listeners and early speakers are able to
participate in some aspects of the social community. When children reach speaker-as-
own-listener levels of verbal capability, they are able to participate in and contribute to the
social community in more comprehensive and independent ways.
The Development of Very Early Observing Responses

Responding to human voices appears to be an important component of children’s acqui-
sition of both listener and speaker repertoires (Peláez-Nogueras, Gewirtz, & Markham,
1996) and, not surprisingly, occurs very early in typically developing infants. Novak and
Peláez (2004) found that even newborn infants are able to discriminate human language
sounds. Since language is arbitrarily defined (for example, consonant-vowel combina-
tions are arbitrarily given meaning), the individual’s responding must come under control
of the acoustical properties of speech. DeCasper and Fifer (1980) conducted a study in
which newborns could occasion the playing of a recording of either their mother’s voices
or the voice of an unknown female by sucking on a nonnutritive nipple at differential
rates. Once the response requirement was established, it was found that the infants caused
the recordings of their mothers’ voices to be played more frequently than the recording
of the unknown female’s voice, suggesting that certain voices may become conditioned

reinforcers early in life. We propose that the mothers’ voices selected out the attention of
the children because they were conditioned reinforcers for observing responses.
Most typically developing infants begin to differentiate between consonant-vowel com-
binations, such as “pa” and “ba,” within their first month (Eimas, Siqueland, Jusczyk, &
Vigorito, 1971); in fact, vocal imitation of phonetic units appears to play a critical role
in infants’ acquisition of language (Kuhl & Meltzoff, 1996). For children with significant
language delays, however, adult voices do not serve as reinforcers for listening and, as a
result, prelistener skills may not develop. If voices do not select out attention, discrimina-
tions of the “pa” and “ba” sounds are not likely to occur. Field (1987) found that children at
higher risk for developmental disabilities did not develop positive social interactions through
typical mother-infant play, and they needed more intensive stimulation than did typically
developing children before responding positively to a caregiver’s touch and voice. Our work
suggests that the attainment of conditioned reinforcement for listening to adult voices is one
of the earliest capabilities associated with language development and social functioning.
Children who are selected for the conditioning listening to adult voices protocol,
outlined in table 3.3, are assessed to be prelisteners. They do not typically respond to
visual or auditory stimuli in the environment and as a result cannot progress toward their
curricular goals. Pre- and post-test measures of the children’s levels of responding to learn
units across all areas of the curriculum are taken before and after mastery of the protocol,
as are measures of nonfunctional self-talk. Observing responses associated with sustained
eye contact and observation of the activities of other individuals in the immediate envi-
ronment are also measured before and after mastery of the protocol. The measures of the
numbers of learn units required in order to master components of the curricula are tests
of the rate of learning.
In this chapter, for the protocols we describe based on a pair and test trial format (con-
ditioning listening to adult voices and conditioning print stimuli), we use the Pavlovian
second-order conditioning procedure. The teacher delivers edibles and noninterfering
unconditioned reinforcers (such as edibles) or conditioned reinforcers (such as praise or
a token if they are truly conditioned reinforcers) when the child is attending appropri-
ately to the target item or activity. Data are recorded using a pair and test trial format.
Initially, the first short-term objective begins with a five-second pair trial interval, and
when completed successfully it is followed by a five-second test trial interval. During the
pairing trial, the teacher delivers a reinforcer either two or three times contingent upon
the child’s emitting the target behavior. The trials of two and three pairings are alter-
nated. However, if the child does not engage in the target activity or item and/or emits
stereotypy (repetitive or ritualistic movements) or passivity at any time during the pairing
trial, the pairing trial is immediately restarted. The test trial begins only when the pairing
trial is completed without stereotypy or passivity.
During the test trial, the teacher records whether or not the child engages in the
activity or item according to the definition of the target behavior for the duration of the
interval used (five-, ten-, or fifteen-second intervals or longer). No reinforcement is deliv-
ered during the test trial. Whole interval recording is used, in which a correct response is
recorded if the child engages appropriately for the entire interval. If at any point the child
is not engaging in the activity or item appropriately, an incorrect response is recorded,
the test session is ended immediately, and the next pairing interval is started. Data are
recorded and graphed out of twenty for the number of correct test trials. Generally,
criterion is set at 90 percent for two consecutive pair and test trial sessions. When the
Chapter 3 45
child does not meet the 90 percent criterion for the five-minute free operant test of listen-
ing to the recorded voices, we increase the duration of the pair and test intervals (such as
ten seconds, fifteen seconds, and twenty seconds), until the child meets criterion.
Table 3.3. Conditioning Listening to Adult

Voices Protocol: Conditioning Adult Voices as
Reinforcement for Observing Responses
Rationale This protocol is indicated if a child does not orient toward adult voices
and/or look at speakers, particularly those holding sources of reinforce-
ment. Be certain the child does not have a major hearing deficit before
attempting to teach this stimulus control. If voices do not select out
or attract the child’s attention, the child is unlikely to be prepared to
discriminate vowel-consonant sounds and other aspects of speech that
come to have listener and speaker effects. If adult voices are condi-
tioned reinforcers for observing responses, the child will learn at a
significantly faster rate.
Pre- and Twenty experimental probe trials (no consequences) should be com-
Post- pleted using duration recording of each trial lasting one or more
intervention seconds. These probe trials should consist of a variety of novel oppor-
Probes to tunities for the child to respond to an adult’s presence (for example,
Test for the the child turns toward an adult when her name is called, looks toward
Acquisition an adult entering the room, looks toward an adult speaking to a child
of the Cusp nearby, or looks toward an adult rearranging the child’s environment,
such as moving a toy or other tabletop materials; please see “Pre- and
Post-probes of Observing Responses” below), measured in three selected
environments (one-to-one, small group, and unstructured settings).
Pre- and post-protocol probes of total learn units to criterion across
subject area lessons (such as match/duplicate, point/show) based on a
minimum of 1,000 learn units and 1 criterion per category should be
completed.
Pre- and post-probes of observing responses: In a 20-trial format (a
trial should continue for at least 1 second to meet the response defi-
nition criterion), the duration of the following responses should be
measured:
Data collection settings: In one-to-one (such as teacher and child),
small group (for example, 2 to 6 children), and unstructured settings
(such as a play area), the duration of the following responses should be
measured:
1. Child orienting toward a speaker when her name is called from a
distance of 1 to 4 feet

2. Child orienting toward a speaker when her name is called from a
distance of 5 to 8 feet
3. Child orienting toward a speaker when the child is given a direction
from a distance of 1 to 4 feet
4. Child orienting toward a speaker when the child is given a direction
5. Child orienting toward a speaker when another child is spoken to
6. Child orienting toward a speaker when another child is spoken to
7. Child orienting toward an adult rearranging the child’s materials on
the desk
8. Child orienting toward an adult removing the child’s materials from
the desk
9. Child orienting toward an adult who is entering the room and
speaking.
10. C
hild orienting toward an adult entering the room who is not
speaking
Materials Use nursery rhymes or selections from children’s books recorded by the
child’s mother, teacher, and someone the child does not know. Tapes or
CDs should be 5 minutes in duration and should not include singing.
Have 6 or more recordings available that are rotated in equal measure
across speakers.
Special Note Academic lessons other than those associated with an expanded com-
munity of activities and interests (such as looking at books, playing
with toys, manipulating puzzles, and so on) are suspended during the
implementation of the protocol. A multiple probe format is used.
Conditioning Use a tape recorder or other recording device that reproduces the
Procedure selected voices when the button is manually held down by the
child and automatically stops when the child takes her hand off the
button. Electric switches from speech therapy catalogues can be used.
Alternatively, you may substitute a laminated circle or square that will
act as a simulated on/off button or switch for the child to touch while
the teacher controls the progress of the recording device. (If the child
has her hand on the simulated disk or switch, the tape is played; if the
child takes her hand off the disk or switch, the tape is stopped immedi-
ately.) Note that touching the disk or switch, is the most direct measure
of the auditory observational response, and this allows us to determine
if the child is listening or not listening. If the child emits stereotypy,
the tape should be stopped as well. A pair-test conditioning procedure
with pre- and post-tests of the child’s orienting to voices is described
Chapter 3 47
above under “Pre- and Post-test Probe Trials.” In the pairing segment,
edibles are typically paired with listening to adult voices until the child
listens with no observable stereotypy (stereotypy is a competing rein-
forcer). During the pairing intervals, 2 and 3 pairings of edibles should
be rotated (the number stays the same as the pairing intervals graduate
from 5 seconds to 10 seconds, then 15 seconds, and so on). No rein-
forcement procedures are used during test trials.
Sessions are typically 5 minutes in duration and whole interval
continuous 5-second intervals constitute measurement of the student’s
progress in achieving criterion on the conditioning intervention.
Criterion 90% of 5-second whole interval recordings (ninety 5-second intervals)
over two 5-minute consecutive sessions. Criterion for meeting the
test of conditioned reinforcement is 90% + intervals for 2 consecutive
5-minute sessions with no observation of stereotypy or passivity. The
voice is a conditioned reinforcer when the child will touch the disk or
hold down the button continuously for 90% of the observation inter-
vals recorded in 5-second intervals within a 5-minute session.
This protocol has been repeatedly shown to be effective in increasing a child’s listen-
ing to adult voices (Greer, Keohane, & Delgado, 2006), conditioning listening to specific
music (Greer, Dorow, & Hanser, 1973; Greer, Dorow, Wachhaus, & White, 1973) and
acquisition of more complex verbal, academic, and social skill sets (Tsai & Greer, 2006).
Conditioning Observing Visual Stimuli

An early challenge for those who work with children with severe language delays is
ensuring that the children visually attend to relevant environmental stimuli. Many of
the children we work with do not make sustained eye contact with other individuals or
stimuli in the environment. Sustained attention to visual stimuli is a critical early step
in language acquisition, as these children may otherwise be unable to develop correspon-
dence between what they hear and what they see. Prior research reported that visual
training may actually enhance tactile discrimination skills in young children (Krekling,
Tellevik, & Nordvik, 1989), and that reinforcement in the form of music or other pre-
ferred activities may improve visual tracking skills in infants (Darcheville, Madelain,
Buquet, Charlier, & Miossec, 1999). Because of our many years of work with children
with disabilities, we knew that language acquisition was hindered when the children did
not make sustained eye contact with stimuli in the environment. As a result we designed
a protocol using a form of conjugate reinforcement to address the problem. Collier and
Bitetti-Capatides (1979) defined conjugate reinforcement as continuous reinforcement
during which the intensity of the reinforcement is directly related to the intensity of the
response (for example, in vigor, rate, or duration). For example, the longer a child gazes
at a colorful toy, the longer the duration of the reinforcer for gazing at the toy—the
sight of the colorful toy itself). That is, the degree to which the child’s attention to the

visual stimulus is selected out is a measure of the reinforcement by that stimulus of the
child’s operant observing behavior. We frequently find that when a child is having dif-
ficulty acquiring prelistener verbal capabilities, the problem is related to a lack of attend-
ing behaviors associated with specific and general observation of the environment. The
objective of the visual tracking protocol, outlined in table 3.4, is to condition sustained
observation of visual stimuli by pairing unconditioned or conditioned reinforcers (such
as edibles and toys) with the child’s attention to visual stimuli (Keohane & Greer, 2005).
In this way, we transfer the reinforcement control from the unconditioned or conditioned
reinforcement to the previously neutral stimulus (in other words, attending to the visual
stimuli that is now a reinforcer for the operant observing response). See the introduction
to Tsai and Greer (2006) for references to the extensive laboratory research on conjugate
reinforcement.
Prelisteners who attend to visual stimuli inconsistently, do not imitate teacher model-
ing, do not match to sample, do not follow basic directions, and fail to meet short-term
and long-term curricular objectives are candidates for the visual tracking protocol. Pre-
and post-test measures of responding to learn units across all areas of the curriculum,
and the observing responses associated with sustained eye contact to stimuli and other
individuals in the environment, are measured before and after mastery of the protocol.
Table 3.4. Visual Tracking Protocol: Observing 3-D

Tabletop Stimuli as Conditioned Reinforcement
Rationale This protocol should be implemented if a child does not attend to

visual stimuli or look at adults or children in the environment and
emits low numbers of correct responses across visual matching pro-
grams and other curricula requiring attention to visual stimuli. See
Greer and Ross (2008) for prerequisites.
Experimental These probes are measures of the child’s rate of learning instructional
Probes objectives that entail visual observing. If 3-D visual stimuli are condi-
tioned reinforcers for observing responses, the child will learn at a signifi-
cantly faster rate than he would if the stimuli do not reinforce looking.
Pre- and post-protocol probes of total learn units to criterion, visual
(e.g., match/duplicate) learn units to criterion, and visual-listener (e.g.,
point/show) learn units to criterion based on a minimum of 1,000 learn
units and 1 criterion per category should be completed.
Pre- and post-protocol probes of eye contact and visual tracking of
items and individuals in three selected environments (one-to-one, small
group, and unstructured) should be completed using duration recording
of each trial lasting 1 or more seconds. This could be done by starting
timing when the child begins observing the stimulus, stopping timing
when the child stops sustained observing of the stimulus, regaining the
child’s attention, and beginning another duration recording trial.
Chapter 3 49
Pre- and post-protocol probes of sustained eye contact with stimuli:
A 20-trial probe measuring the duration of sustained eye contact with
a neutral or a nonpreferred stimulus should be completed for each trial
lasting 1 or more seconds.
Criterion for probes: If the child emits 160 (or more) cumulative
seconds of sustained eye contact with the stimulus in a maximum of 20
trials, criterion for the developmental cusp has been achieved.
Pre- and post-probes of observing responses: In a 20-trial format (a
trial should continue for at least 1 second to meet the response defi-
nition criterion), the duration of the following responses should be
measured.
Data collection settings: In one-to-one (teacher and child, for
example), small group (2 to 6 children, for example), and unstructured
settings (such as a play area), the duration of the following responses
should be measured:
1. Child making sustained eye contact with a speaker when his name
is called from a distance of 1 to 4 feet
2. Child making sustained eye contact with a speaker when his name
is called from a distance of 5 to 8 feet
3. Child looking toward a speaker when the child is given a direction
4. Child looking toward a speaker when the child is given a direction
5. Child looking toward a speaker when another child is spoken to
6. Child looking toward a speaker when another child is spoken to
7. Child making sustained visual contact as an adult rearranges the
child’s materials on the desk
8. Child making sustained visual contact as an adult removes the
child’s materials from the desk
9. Child looking toward an adult entering the room who is speaking
10. C
hild making sustained visual contact with an adult moving about
the room who is not speaking
Materials Use 2 or 3 identical transparent containers (cups or other containers).
Vary size and shape of identical sets for each presentation. A variety of
neutral items for the pre- and post-probes (paper clips, unfamiliar shapes,
and other items) and preferred items (edibles, tokens, toys, and the like)
that can be placed under the target container should be available. A stop-
watch is used to record duration of sustained eye contact.

Special Note Programs other than those associated with self-management and an
expanded community of activities and interests (such as looking at
books, playing with toys, manipulating puzzles, and so on) should be
suspended during the implementation of the visual tracking protocol.
We suggest this because the child is not likely to progress with visual
instructional programs at this point, and it is counterproductive to
continue those programs while you are attempting to induce the cusp.
However, programs for conditioning or expanding the child’s commu-
nity of reinforcers or interests can be conducted simultaneously since
these are also reinforcers that increase a variety of observing responses.
General Data Collection: During implementation of the protocol, pairing trials
Procedure (of 1 second or more) are recorded as duration in each 20-trial session.
When the child looks away from the stimulus the duration record-
ing is stopped for that trial, and a new trial is begun after the child’s
attention is regained. Duration criterion for the long-term objective is
160 or more seconds of sustained eye contact in a 1- to 20-trial session
(in other words, it is possible for the child to reach the 160-second
cumulative objective in a session containing less than 20 trials). Each
20-second trial is an opportunity to test the duration of stimulus
control and a convenient way of delineating sessions.
Pairing procedure: Edibles, tokens, toys, and the like placed under
transparent containers should be used. The child may be prompted to
visually track the preferred item under the container during the first
few pairing trials.
Short-term objectives as tactics: Short-term objectives include incre-
mental increases in targeted duration of sustained eye contact and
interspersal of known items (such as running pairing trials until it
becomes difficult to regain the child’s attention). At that point, inter-
sperse known items (for example, have the child respond to the kinds
of instruction he has already mastered, ensuring that you do not lose
the child’s attention). In addition, use the preferred item the child
is tracking as a reinforcer for correct responding to the interspersed
trials for previously mastered stimuli). Rate of rotation of stimuli and
number of stimuli rotated (2 or 3) may be used as short-term objec-
tives in this program (for example, stimuli may be rotated 1 to 4 times
at progressively increasing rates of rotation as short-term objectives
become more complex).
Note: The child is not given an opportunity to receive the preferred
item after any pairing trial. We do this because we want only the visual
stimulus control of the object to be the conditioned reinforcer.
Criterion Two consecutive sessions with 20 (or fewer) trials resulting in 160
cumulative seconds of sustained eye contact with the target stimulus.
All programs should be reintroduced and post-probe data collected
when the child meets criterion.
Chapter 3 51
Pre-test and post-test measures of selected programs should be conducted after each
short-term objective is achieved. When criterion is achieved the child’s full schedule of
programs or academic lessons should be resumed and post-test data collected across all
areas of the curriculum.
Conditioning Print Stimuli

Orienting toward instructional materials and the table at which instruction is conducted
is an important early observing response. Additionally, matching tasks provide impor-
tant educational goals because successful matching performance demonstrates that a child
can visually attend to stimuli, a prerequisite to learning, as well as discriminate among
the relevant properties of environmental stimuli. Basic matching involves responding
to the sameness of identical objects, identical pictures, objects to pictures, and pictures
to objects. More advanced matching may include abstractions such as matching colors,
shapes, irrelevant dimensions (such as a spotted dog versus a black dog), numbers, and
letters. The ultimate goal is for children to acquire generalized matching skills so that
they can match anything without direct instruction, including novel items that they have
never experienced before. Children with significant developmental disabilities often fail to
master both basic and generalized matching skills despite the employment of numerous
remedial techniques such as stimulus or response prompts. We have identified an effec-
tive protocol, the conditioning print stimuli on a page protocol, outlined in table 3.5, for
inducing generalized matching in such children. For children at an earlier stage of verbal
development, the visual tracking protocol (Keohane, Greer, & Ackerman, 2006) can be
used to teach looking at three-dimensional stimuli on a table, and as a result, three-
dimensional visual matching responses may be acquired. However, for other children
we find that print stimuli do not serve as conditioned reinforcers. These children may
acquire basic matching skills with three-dimensional stimuli only and are thus appropri-
ate candidates for the conditioning print stimuli on a page protocol. Our research and
clinical work has shown that when print stimuli and the pictures on pages of books are
established as conditioned reinforcers for observing, children readily acquire generalized
matching skills. In addition, we find that the children are more attentive to in a variety
of settings during and outside instructional sessions.
Observing print stimuli is a critical cusp for a child to have. When a child does not
attend to such stimuli, more complex instructional tasks cannot be targeted. Attention
to stimuli serves as a building block for complex skills, such as joint attention (the capac-
ity to use gestures and eye contact to coordinate attention with another person in order
to share the experience of some object or event in the environment; Mundy, Sigman, &
Kasari, 1994; see also chapters 4 and 13 of this volume).
The conditioning stimuli on a page protocol involves delivering preferred items or
edibles while the child attends to two-dimensional stimuli on a page using the condition-
ing procedure described earlier in this chapter. The duration of the child’s attention to
such stimuli is measured on pre- and post-test probes.

Table 3.5. Conditioning Print Stimuli on a
Page Protocol: Conditioning Sustained Eye
Contact with Print Stimuli
Rationale Use this protocol if the child fails to attend to print stimuli and has a
high number of learn units to criterion on matching programs.
Pre- and Post- Pre- and post-test probe learn units to criterion on visual (match/
test Trials duplicate) and visual-listener (point/show) learn units to criterion based
on a minimum of 1,000 learn units and 1 criterion.
Pre- and post-test probes of looking at stimuli on a page. These probes
occur following criterion on each short-term objective of conditioning.
The teacher records whether the child looks at a single page of stimuli
for 10 consecutive seconds. Five individual pages are presented 1 page
at a time and these same pages are reserved for post-conditioning probes
only (do not use during conditioning sessions).
Materials Materials include a variety of 15 to 20 nonpreferred 2-D stimuli
(printed letters, numbers, or pictures) on 8.5-by-11-inch sheets of paper.
Special Note All matching programs and point-to programs or academic lessons are
suspended during the implementation of this protocol. Only return to
these programs when the child meets the long-term objective (LTO) for
this protocol. (The term LTO indicates that the child has acquired the
developmental cusp of conditioned reinforcement for print stimuli.)
General Deliver edibles or noninterfering conditioned reinforcers as the child
Procedure looks at various pages of stimuli with no observable stereotypy or
passivity. Conduct probe sessions prior to and after each short-term
objective of the conditioning procedure, until the LTO is achieved
during the probes for looking at stimuli on a page.
Criterion Conditioning procedure = 90% for 2 consecutive sessions.
Post-probe sessions for looking at stimuli on a page = 80% or 4 out
of 5. This is the LTO for the protocol.
Development of a Capacity for Sameness

Across the Senses
Having the developmental capacity for sameness across the senses helps children advance
to new behavioral developmental cusps. Acquisition of skills that combine the audi-
tory, visual, tactile, olfactory, and gustatory senses is a critical developmental milestone.
Children cannot fully function in a verbal environment without coordination of the
various sensory observations and discrimination skills. Children who have not met the
Chapter 3 53
sensory matching developmental milestone are at the prelistener level of verbal capability
and do not reliably attend to sensory stimuli. Most children acquire this foundation of
verbal development early in life in the absence of specialized instruction. However, chil-
dren with certain disabilities may not acquire it without special behavioral developmental
interventions. The sensory matching protocol, outlined in table 3.6, provides children
with rotated multiple-exemplar experiences across five critical sensory modalities. The
rotated exposure to sensory matching experiences provided within this protocol supports
the development of this capability as well as more-complex listener behaviors (Keohane &
Greer, 2005). Greer and Ross (2008) argue that developing a capacity for sameness may
be the fundamental step toward becoming verbal. When a child matches across all senses,
she learns an arbitrarily applicable cross-modal response of sameness.
The objective of this program is to provide children with the capacity for sameness
across different sensory modalities. If children master this protocol, they typically have
the foundation for the abstraction of sameness across sensory stimuli.
Table 3.6. The Sensory Matching Protocol:

Matching Across the Senses
Rationale We use this protocol if adult voices and visual stimuli are conditioned
reinforcers but children are not meeting short-term and long-term
objectives at an adequate rate and do not have the capacity to match
across the senses. Assuming 20-learn-unit instructional sessions, an
adequate rate of learning would be 80 to 120 learn units to mastery
of instructional objectives.
Pre- and Post- The target of this protocol is to accelerate learning rates, just as in the
intervention visual tracking goal. Increased attention is often a collateral effect and
Probes to Test should be measured also.
for Acquisition Pre- and post-probe measures of learn units to criterion, long-term
of the Cusp and short-term objectives achieved, and observing responses associated
with listener and visual sensory modalities (see “Pre- and post-probes of
observing responses,” below) should be conducted. Probes of noncontex-
tual self-talk (such as palilalia) may also be conducted if this is
a problem for the child.
Pre- and post-probes of observing responses: In a 20-trial format
(a trial should continue for at least 1 second to meet the response
definition criterion) the duration of the following responses should be
measured.
Data collection settings: In one-to-one (e.g., teacher-child), small
group (e.g., 2 to 6 children), and unstructured settings (e.g., play area),
the duration of the following responses should be measured:

1. A child orientating toward adults or other children calling her name
2. A child orientating toward an adult or child in the immediate envi-
ronment as the first step in a conversational unit
3. A child emitting sustained eye contact with a stimulus relocated by
an adult or other child in the immediate environment
4. A child responding to instructions given by an adult in the immedi-
ate environment
5. A child emitting functional self-talk in a play area or other appropri-
ate setting
Materials Select items and presentation formats that assure that the child can
identify the matching items only through the targeted sensory modality.
Select 2 exemplars across each of the 5 senses. Some suggestions for the
preparation of pairs of sensory stimuli are as follows: (1) auditory sense
(a dog barking and water running), (2) visual sense (picture of a cow
and picture of a house), (3) tactile sense (sandpaper and velvet placed in
separate sacks, one with the exemplar, and the other with the exemplar
and a nonexemplar), (4) smelling or olfactory sense (vanilla or orange
scents in matching containers), and (5) tasting or gustatory sense
(sweetened juice versus water). Each learn unit presentation includes
a correct match exemplar (the positive or target) and 2 incorrect or
negative exemplar stimuli in a matching-to-sample format.
Special Note For the auditory matching component, place a device on the table with
prerecorded sounds of the selected auditory stimuli with accurate or
positive exemplars and one nonexemplar or inaccurate exemplar per
matching-to-sample procedure. Alternate orientation or positions to
ensure that position is not a factor. For the tactile component, a
comparison can be made only when the child is touching the contents
of each sack simultaneously. The child must not be able to see the
tactile stimuli. All instructional programs other than those associated
with self-management and conditioning of preferred interests and
activities should be placed on hold or discontinued until the cusp is
achieved. It is counterproductive to continue unsuccessful instruction.
General Presentation of stimuli across the senses is rotated across 20-learn-unit
Procedure sessions. For example, an olfactory learn unit is followed by a tactile
learn unit, and then a gustatory learn unit, followed by an auditory one,
and finally a visual one. Place one exemplar and one nonexemplar on a
table; rotate nonexemplars across instructional presentations. Continue
to use this format across 20 instructional presentations per session.
Rotate the sequence of instruction and the nonexemplars so that the
child does not learn to respond in a particular order, and also rotate
orientation of positive and negative exemplars per learn unit.
Chapter 3 55
Criterion Given one positive exemplar (an accurate match) and two nonexemplars
(nonmatching stimuli), children will match identical exemplars of the
target items, rotated across gustatory, visual, olfactory, and tactile senses
until they achieve standard mastery criterion (typically 90% across 2
sessions). Once the child has met the long-term objective, you may
return to the full schedule of subject-area lessons. The rate of learning
should have accelerated such that children have significantly reduced the
numbers of learn units required to master instruction.
In the studies we have conducted to date, children’s levels of observing responses

have been shown to increase across the areas targeted when this protocol is implemented
(Greer, Keohane, Ackerman, et al., 2006). After completion of the sensory matching
protocol, children are often able to master new skills, including pointing and echoing.
We also assume that the steps required for the prelistener level of verbal capability have
now been achieved, and the child is ready to begin the next set of protocols as he moves
toward a more comprehensive listener status. (See Greer & Ross, 2008, for additional
information on the sensory matching protocol.) The original focus of this procedure was
simply to develop a capacity for sameness, but we found that the procedure frequently had
collateral effects on observing responses. We do not yet know why this appears to be the
case for some children and not others.
Generalized Imitation as the Next Step

The acquisition of imitation skills is critical for the development of language and social
repertoires (Rogers & Pennington, 1991). Baer, Peterson, and Sherman (1967) defined
imitation as a behavior that closely follows another individual’s behavior in which the
form is controlled by the behavior of other individuals. For example, a child who imi-
tates will likely clap his hands when a model claps his hands if the response has been
directly trained. Imitation is a see-and-do relationship, which involves a point-to-point
correspondence between behavior of the model and the behavior of the observer (Greer
& Ross, 2008). Generalized imitation is present when a child imitates novel behavior
without direct reinforcement of that particular response. Generalized imitation, unlike
observational learning, is the result of direct reinforcement of a class of responding and
is not controlled by the observation of the contingencies experienced by another (Greer,
Singer-Dudek, & Gautreaux, 2006). Generalized imitation thus involves the continued
performance of a response for which the child has never received reinforcement (Brigham
& Sherman, 1968). Once a child has acquired generalized imitation, he does not need to
be explicitly taught each motor action separately with response prompts. For example, a
child may sit down, clap his hands, and wave bye-bye once he has learned the behavior
of imitating—each response does not need to be explicitly shaped. Generalized imita-
tion can thus be regarded as a higher-order operant (Greer & Ross, 2008). A teacher can
simply demonstrate a behavior and then teach the relevant antecedent and consequent
relations that are needed to teach a new operant.

Much research has been devoted to directly taught imitation and generalized imi-
tation in both typically developing children and children with developmental disabili-
ties. Typically developing children at age four have shown generalized imitation skills
(Baer & Deguchi, 1985), but infants of one and two years of age have not (Horne &
Erjavic, 2007). Children with severe developmental delays often have deficits in their imi-
tative repertoires. Research over the past few decades has identified successful tactics to
teach imitative responses to such children. These include simultaneous stimulus prompts
(Wolery, Holcombe, Billings, & Vassilaros, 1993) and shaping procedures (Garcia, Baer,
& Firestone, 1971). These tactics are often successful in teaching an explicitly defined
set of responses. For some children such strategies lead to the development of general-
ized imitation, but for other children such techniques are often tedious and unsuccessful
because the class of responding or higher-order operant has not been formed.
In our schools we use learn units, as described earlier in this chapter, to measure
teacher and child responses. In many cases learn units combined with the above-men-
tioned strategies do not result in generalized imitation. For example, we may teach chil-
dren to imitate the following set of actions to mastery using the learn unit: stand up,
clap hands, touch head, and wave. Even though the children are explicitly taught these
responses, they may not imitate novel actions, such as sit down, tap lap, and touch shoul-
ders, which would be indicative of generalized imitation. Children who do not display
generalized imitation repertoires are considered prelisteners. Imitation and generalized
imitation are two of the five basic attention programs taught in our schools. They are
prerequisites for learning basic listener responding. Until a child acquires listener literacy
(Greer, Chavez-Brown, Nirgudkar, Stolfi, & Rivera-Valdes, 2005), he will not achieve
more-advanced developmental milestones. Listener literacy includes the ability for one’s
behavior to be governed by the speaker behavior of others and results in a greater degree
of independence for the child (Greer, 2002).
The Mirror Protocol

Pereira-Delgado, Greer, and Speckman-Collins (2006) evaluated the use of a mirror
in the emergence of generalized imitation. In the first experiment, the participants were
three- and four-year-old children diagnosed with developmental disabilities who had
acquired numerous motor actions through direct learn-unit presentations combined with
various other instructional tactics. However, the children did not show generalized imita-
tion during unreinforced test trials. The intervention consisted of teaching the children to
imitate teacher actions using a mirror, in which the child faced the mirror and the teacher
sat behind the child. Learn-unit presentations were used throughout the mirror proce-
dure, where the child was required to look in the mirror to observe the teacher’s presenta-
tion of a given motor action. The target behavior included the child modeling or copying
the teacher’s motor actions with a point-to-point correspondence. We found that, after
being taught a set of four previously acquired motor actions in the mirror to mastery,
the participants acquired generalized imitation. In the second experiment, the partici-
pants were also three- and four-year-old children with developmental delays. However,
these children were only taught a few motor actions directly. Test sessions for generalized
imitation included a specified set of twenty different motor actions that were not taught
Chapter 3 57
during instruction (for example, tap fingers on desk, rub hands together, lift up one leg).
Subsequently, the mirror procedure was implemented to determine if the children would
acquire generalized imitation. The participants were closely matched based on levels of
verbal behavior with peers who did not receive the mirror procedure. The findings paral-
leled the results of the first experiment; moreover, the matched peers who did not receive
the mirror intervention did not acquire generalized imitation even when the numbers of
learn units were controlled for.
The results of both experiments can be explained in terms of the correspondence
between “see” and “do,” a theory suggested by Catania (1998) as something that could
be established using a mirror. It seems plausible that the opportunities the children had
to view their responses in the mirror enhanced this see-and-do correspondence. Prior
research suggests that deficits in imitation may result when an individual is not able to
observe a visible endpoint (Meltzoff & Moore, 1983). For example, a child may have no
difficulty imitating clapping because the response is visible to him. Touching one’s head,
however, may be more difficult for him to imitate because the response is out of the
child’s sight. The mirror provides children with the opportunity to see their responses. As
noted previously, prior research has found that children with disabilities typically require
an extensive amount of shaping to develop imitation of discrete behavior, yet this does not
guarantee that they will acquire generalized imitation. For example, Baer and colleagues
(1967) found that children with developmental disabilities did not acquire generalized
imitation until their imitative repertoire consisted of at least forty to sixty directly trained
imitative responses! The mirror protocol, outlined in table 3.7, has shown promise in
establishing generalized imitation skills efficiently and expediently.
Table 3.7. Generalized Imitation

through the Mirror Protocol
Rationale If the child does not show generalized imitation, use this protocol.
Long-Term Given a set of 20 unreinforced test trials, which consist of novel fine
Objective and/or gross motor actions presented by the instructor, the child will
imitate these actions with 80% accuracy for 1 session.
Materials A full-length unbreakable mirror measuring approximately 13.5 by 50

inches is used for this protocol. Note: Breakable mirrors may be hazard-
ous to the health of children and professionals.
Special Note For the probe sessions, reserve a set of 20 novel motor actions that are
not in the student’s repertoire (do not teach these actions in the mirror).
For the teaching sessions with the mirror, if the child has mastered
several directly taught actions prior to the implementation of this proto-
col, teach these previously mastered actions in the mirror; otherwise, you
can teach novel actions in the mirror (not part of the probes).

Ensure that the child is looking at the mirror and not at the teacher
by pointing to the mirror as part of the antecedent during instructional
sessions. Suspend all other instructional programs on imitation.
General Conduct unconsequated probes (20 novel actions) directly facing the
Procedure student (do not use the mirror). Begin with the first short-term objective
of 4 target actions to be taught in the mirror using 20-learn-unit ses-
sions until criterion is achieved. Conduct a post-probe session for novel
actions. Continue with a new short-term objective of 4 novel target
actions to be taught in the mirror until generalized imitation is estab-
lished or the long-term objective is achieved during the probe sessions.
Criterion 90% correct learn units for two consecutive sessions for teaching actions
in the mirror. 80% correct trials for probes for one session, which is the
LTO for this protocol.
Research to a Theory of Verbal Development

Dinsmoor (1985) noted that greater observing and attending to specific environmental
stimuli resulted in an increase in control over behavior by those stimuli. In this chapter
we have presented what we believe to be strong evidence of the establishment of children’s
observing responses as a result of the implementation of the conditioning procedures and
protocols described. After implementation of these protocols, the children with whom
we have worked have advanced to higher levels of verbal capability. We also believe that
the evidence illustrating the efficacy of these protocols presents a more complete view of
the foundations of verbal behavior. In conclusion, we submit that the role of the listener
as observer and particularly the interrelationships among listener, listener-speaker, and
speaker-as-own-listener functions are the foundations basic to the development of early
language capabilities (Greer & Keohane, 2005).
References
Baer, D. M. (1983). Can you decode a code? Behavioral and Brain Sciences, 1, 138–139.
Baer, D. M., & Deguchi, H. (1985). Generalized imitation from a radical-behavioral
viewpoint. In S. Reiss & R. Bootzin (Eds.), Theoretical issues in behavior therapy. New
York: Academic Press.
Baer, D. M., Peterson, R. F., & Sherman, J. A. (1967). The development of imitation by
reinforcing behavioral similarity to a model. Journal of the Experimental Analysis of
Behavior, 10, 405–416.
Brigham, T. A., & Sherman, A. J. (1968). An experimental analysis of verbal imitation in
preschool children. Journal of Applied Behavior Analysis, 1, 151–158.
Catania, A. C. (1998). Learning (4th ed.). Englewood Cliffs, NJ: Prentice Hall.
Chapter 3 59
Collier, C. K., & Bitetti-Capatides, J. (1979). Positive behavioral contrast in three-month-
old infants on multiple conjugate reinforcement schedules. Journal of Experimental
Darcheville, J. C., Madelain, L., Buquet, C. Charlier, J., & Miossec, Y. (1999). Operant
conditioning of the visual smooth pursuit in young infants. Behavioural Processes, 46,
131–139.
DeCasper, A. J., & Fifer, W. P. (1980). Of human bonding: Newborns prefer their
mothers’ voices. Science, 208, 1174–1176.
Dinsmoor J. A. (1985). The role of observing and attention in establishing stimulus
control. Journal of the Experimental Analysis of Behavior, 43, 365–381.
Donahoe, J. W., & Palmer, D. C. (2004). Learning and complex behavior. Richmond, VA:
Ledgetop Publishing.
Eimas, P. D., Siqueland, E. R., Jusczyk, P., & Vigorito, J. (1971). Speech perception in
early infancy. Science, 171, 303–306.
Epstein, R. R., Lanza, P., & Skinner, B. F. (1980). Symbolic communication between two
pigeons (Columbia livia domestica). Science, 207, 543–545.
Epstein, R. R., Lanza, P., & Skinner, B. F. (1981). “Self-awareness” in the pigeon. Science,
212, 695–696.
Field, T. (1987). Affective and interactive disturbances in infants. In J. D. Osofsky (Ed.),
Handbook of infant development (2nd ed.). Oxford, England: Wiley.
Garcia, E., Baer, D. M., & Firestone, I. (1971). The development of generalized imitation
within topographically determined boundaries. Journal of Applied Behavior Analysis,
4, 101–112.
Greer, R. D. (2002). Designing teaching strategies: An applied behavior analysis system
approach. San Diego, CA: Academic Press.
Greer, R. D., Chavez-Brown, M., Nirgudkar, A. S., Stolfi, L., & Rivera-Valdes, C. L.
(2005). Acquisition of fluent listener responses and the educational advancement of
young children with autism and severe language delays. European Journal of Behavior
Analysis, 6, 88–126.
Greer, R. D., Dorow, L. G., & Hanser, S. (1973). Music discrimination training and the
music selection behavior of nursery and primary level children. Bulletin of the Council
for Research in Music Education, 35, 30–43.
Greer, R. D., Dorow, L. G., Wachhaus, G., & White, E. (1973). Adult approval and stu-
dents’ music selection behavior. Journal of Research in Music Education, 21, 293–299.
Greer, R. D., & Keohane, D. D. (2005). The evolution of verbal behavior in children.
Behavior Development Bulletin, 1, 31–47.
Greer, R. D., Keohane, D. D., Ackerman, S., O’Sullivan, D., Park, H., Longano, J., et
al. (2006). Sensory matching protocol: Providing children with the capacity for sameness
across the senses as a component of basic listener literacy. Paper presented as part of a

symposium at the thirty-second annual convention of the Association for Applied
Behavior Analysis International, Atlanta, GA.
Greer, R. D., Keohane, D. D., & Delgado, J. P. (2006). Conditioning adult voices to induce
listener capabilities. Paper presented as part of a symposium at the thirty-second
annual convention of the Association for Applied Behavior Analysis International,
Atlanta, GA.
Greer, R. D., & Ross, D. E. (2008). Verbal behavior analysis: Inducing and expanding new
verbal capabilities in children with language delays. New York: Allyn and Bacon.
Greer, R. D., Singer-Dudek, J., & Gautreaux, G. (2006). Observational learning.
International Journal of Psychology, 41, 486–499.
Hart, B., & Risley, T. R. (1999). The social world of children learning to talk. Baltimore,
MD: Paul H. Brooks Publishing.
Hayes, S., Barnes-Holmes, D., & Roche, B. (2001). Relational frame theory: A post-
Skinnerian account of human language and cognition. New York: Kluwer Academic/
Plenum.
Horne, P. J., & Erjavec, M. (2007). Do infants show generalized imitation of gestures?
Journal of the Experimental Analysis of Behavior, 87, 63–87.
Keohane, D. D., & Greer, R. D. (2005). Teachers’ use of a verbally governed algorithm
and student learning. International Journal of Behavioral and Consultation Therapy, 1,
252–271.
Keohane, D. D., Greer, R. D., & Ackerman, S. A. (2006). The effects of conditioning
visual tracking on the acquisition of instructional objectives by prelisteners and prespeak-
ers. Paper presented as part of a symposium at the thirty-second annual convention of
the Association for Applied Behavior Analysis International, Atlanta, GA.
Krekling, S., Tellevik, J. M., & Nordvik, H. (1989). Tactual learning and cross-modal
transfer of an oddity problem in young children. Journal of Experimental Child
Psychology, 47, 88–96.
Kuhl, P. K., & Meltzoff, A. N. (1996). Infant vocalizations in response to speech: Vocal
imitation and developmental change. Journal of the Acoustical Society of America, 100,
2425–2438.
Meltzoff, A. N., & Moore, M. K. (1983). Newborn infants imitate adult facial gestures.
Child Development, 59, 702–709.
Mundy, P., Sigman, M., & Kasari, C. (1994). Joint attention, developmental level, and
symptoms presentation in autism. Development and Psycholopathology, 6, 389–401.
Novak, G., & Peláez, M. (2004). Child and adolescent development: A behavioral systems
approach. Thousand Oaks, CA: Sage Publications.
Chapter 3 61
Peláez-Nogueras, M., Gewirtz, J., & Markham, M. (1996). Infant vocalizations are
conditioned both by maternal imitation and motherese speech. Infant Behavior and
Development, 19, 670.
Pereira-Delgado, J., Greer, R. D., & Speckman-Collins, J. (2006). The effects of using a
mirror to induce generalized imitation. Paper presented as part of a symposium at the
thirty-second annual convention of the Association for Applied Behavior Analysis
International, Atlanta, GA.
Premack, D. (2004). Is language the key to human intelligence? Science, 303, 318–320.
Premack, D., & Premack, A. (2003). Original intelligence: Unlocking the mystery of who we
are. New York: McGraw-Hill.
Rogers, S. J., & Pennington, B. F. (1991). A theoretical approach to the deficits in infan-
tile autism. Development and Psychopathology, 3, 137–162.
Rosales-Ruiz, J., & Baer, D. M. (1997). Behavioral cusps: A developmental and pragmatic
concept for behavior analysis. Journal of Applied Behavior Analysis, 30, 533–544.
Savage-Rumbaugh, E. S., Rumbaugh, D. M., & Boysen, S. (1978). Science, 201, 64–66.
Skinner, B. F. (1989). The behavior of the listener. In S. C. Hayes (Ed.), Rule-governed
behavior: Cognition, contingencies, and instructional control. New York: Plenum Press.
Tsai, H., & Greer, R. D. (2006). Conditioned preference for books and faster acquisition
of textual responses by preschool children. Journal of Early and Intensive Behavior
Interventions, 3, 35–61.
Wolery, M., Holcombe, A., Billings, S. S., & Vassilaros, M. A. (1993). Effects of simul-
taneous prompting and instructive feedback. Early Education and Development, 4,
20–31.

Chapter 4
Joint Attention and Social Referencing

in Infancy as Precursors of
Martha Peláez,
Florida International University
Joint attention and social referencing appear to be critical features of parent-child or

teacher-learner exchanges, which help the learner gather information to guide his own
thoughts, feelings, and behavior. These two related abilities also appear to be necessary
precursors for the emergence of derived relational responding, which underpins much
of higher cognition and language development. The current chapter focuses on teaching
strategies for establishing these core skills with very young learners or learners with devel-
opmental delay, including those with autism spectrum disorder. The first section of the
chapter discusses strategies, including a protocol, for the establishment of the prerequisite
skills for joint attention and social referencing in terms of training conditional discrimi-
nations and identity matching. The second section describes behavioral interventions and
strategies for establishing the core features of joint attention and social referencing.
Establishing the Prerequisite Skills for Joint

Attention and Social Referencing
As well as constituting core social skills, joint attention and social referencing are
codependent abilities that are derived from basic visual discriminations, which have long
been known to be prerequisites for derived relational responding. This early integration of
cognitive and social processes is clearly reflected in the overlap between learners’ abilities to
form conditional discriminations, derive relations, develop language, and interact socially
with others. Numerous studies offer empirical support for integration of these core abili-
ties. For example, Devany, Hayes, and Nelson (1986) demonstrated a correlation between
language and equivalence when only the children in their research with no verbal skills
failed to derive equivalence relations. Furthermore, the more severely language-disabled
children also required more extensive training of the target conditional discriminations
than the other children did, thus suggesting that prerequisite abilities in this regard were
also deficient.
A subsequent replication of the study by Devany and colleagues (1986) provided
further evidence of the importance of conditional discrimination abilities to equivalence
and language. In research by Peláez, Gewirtz, Sanchez, and Mahabir (2000), nine nor-
mally developing infants, aged twenty-one to twenty-five months, were assessed on the
Receptive-Expressive Emergent Language Scale (REEL-2) and then exposed to a series
of visual-visual conditional discriminations. These involved matching animal-like figures
presented in a match-to-sample (MTS) training format. There were four conditional dis-
criminations: if A then B; if A then C; if D then E; and if D then F. Hence the trained
relations were A-B, A-C, D-E, and D-F. All of the children readily demonstrated the target
conditional discriminations and eight of the children demonstrated transitivity (B-C and
E-F); however, five performed below chance on the symmetry tests (for example, B-A
and F-D). As expected, there was a significant negative correlation between the number
of conditional discrimination training trials and the learners’ language quotient (in other
words, higher language means less training). These findings highlighted the relationship
between the level of explicit conditional discrimination training necessary for class forma-
tion and language competence, and they suggested some degree of distinction between
the various component skills in equivalence.
A Protocol for Establishing Conditional Discriminations

In the relevant literature, there are few studies that have described the explicit train-
ing of conditional discriminations in very young learners. This most likely stems from
difficulties in adapting existing methodological paradigms to this population, rather than
weaknesses in the underlying concepts. Indeed, developing experimental methodologies
for use with infants is always extremely challenging, but where they are available they
may offer useful methodologies for training populations with severe developmental dis-
abilities or delays. In one of the only existing studies, Peláez, Lubián, McIlvane, and Dube
(2001) attempted to train and test conditional discriminations in infants who had not yet
fully developed language. One child of eighteen months and two children of twenty-two
months participated. The step-by-step protocol that comprised the training and testing
of these skills is presented below, along with the procedural problems encountered and a
sample data set. The protocol comprises six basic stages that guide instruction from simple
touch-screen response training to generalized identity matching. These, along with an
example of one trial from each stage, are presented in figure 4.1.

Illustration of the 6 training phases:
A D G APP B BAB E
Touch-Screen Response Training Fading Discrimination
APP APP
APP
BAB
Identity Matching Mixed Identity Matching Generalized Identity Matching
Figure 4.1. The discrimination training protocol reported by Peláez et al. (2001), with an
example of each of the six stages.
Training Notes: Preparing to Start

Below are some notes and preparations for conditional discriminations training.
Before you even begin, think very clearly about conducting training of
this sort with very young or very developmentally disabled learners. On
a good day, it can be frustrating. It is extremely difficult to adapt even
the most articulate and systematic methodology to the attention spans of
individuals from these categories, particularly with regard to appropri-
ate responding and attaining a meaningful accuracy criterion. Also, it is
very difficult to preserve their participation and that of their parents or
assistants in a single-learner training environment. These are not small
concerns. These matters raise important questions about the feasibility of
any type of formal training with these learners. They also raise questions
about how to interpret their performances, especially where the method-
ologies have been adapted from basic laboratory studies. For example, in
Chapter 4 65
MTS training programs in which I have been involved, it is not uncom-
mon to find that a learner appears to fail in one stage but then recovers
or responds well in the next. Does that mean that appropriate responding
in the first stage is present? Also, the prompting and cuing involved are
usually more effective when the mother or teaching assistant is present,
compared to when the learner is working with an unfamiliar adult. In
light of these issues, one minimal benefit to be obtained from the current
protocol is that it offers easy start-up training and confidence building
with young learners who have some limited communicative and instruc-
tional histories.
Ensure that training does not interrupt a regular episode involving sleep-
ing, eating, or changing, because learner attention is frequently influ-
enced by organismic conditions (such as fatigue or hunger).
Try to make training sessions as short as possible. Fifteen minutes is the

maximum for infants or severely disabled learners if you are to avoid
distraction and fatigue. In my experience, sessions longer than thirty
minutes can contribute to drastic changes and decline in responding.
It is wise to test all reinforcers prior to training. There are numerous stan-
dard and simple procedures for doing this (see chapter 1 of this volume).
One type of reinforcer available for automated procedures that has been
found to be useful is the blinking of target stimuli and an accompanying
musical sound (for example, for three seconds).
Try to ensure that interventions are maximally effective by further

enhancing the value of the chosen reinforcers to prevent satiation or
habituation. Two ways to do this include alternating with other reinforc-
ers (Higbee & Peláez-Nogueras, 1998) and ensuring that all reinforcers
are short in duration. Social stimuli such as the caregiver’s touch, smiles,
and verbal praise have been shown to be very effective with young infants
(Gewirtz & Peláez-Nogueras, 2000).
Training Notes: Presenting Trials

Training is usually presented in blocks of ten trials. The mastery-train-
ing criterion typically consists of eight consecutive correct responses.
However, with certain learners this criterion may still be too high, and
it is feasible to adjust the criterion (at least early on) to seven consecutive
correct responses. But remember that this is only two responses above
half of all responses being wrong, so try to move back up to the higher
(more stringent) criterion as soon as you can. Do not proceed to the next
stage until the learner has attained this criterion.
Where there are several comparison stimuli, their locations should be

randomized across trials within a block. However, presenting more than
two comparison stimuli in the teaching of conditional discriminations

does not work well with infants, although it usually works well with
learners two years of age and older (see Augustson & Dougher, 1992).
Reinforcement is provided contingently and immediately on all correct

responses.
Trials in which the learner makes an incorrect response are not

reinforced.
Additional interventions (such as shaping and task analyses) are usually

necessary when learners fail to reach criterion after three consecutive
blocks of the same type of trial (see also chapter 8 in this volume).
Each training block is generally followed by a single block of five ran-

domized probe trials (without feedback).
Prompts can be used throughout all training trials. However, it is impor-

tant to determine that the reinforcers, rather than the prompts, are con-
trolling the learner’s responses, so you must conduct subsequent training
trials with no prompts, and responding should remain the same.
Another important aspect to consider is whether the learner can name

the target stimuli. When such a repertoire is not in place, more train-
ing trials are usually required (at least with MTS procedures). It may be
useful (though not always an essential or required condition) to teach
stimulus names prior to further training. Naming the objects may facili-
tate the matching (recognition of the object). See chapter 7 in this volume
for further discussion and instructional strategies for establishing naming
repertoires.
Learners who fail to respond correctly on all five probes must start train-
ing again from the beginning.
There are some reasons to believe that, for some learners, it may be more
effective to conduct identity matching prior to (rather than after) dis-
crimination training.
Training Notes: Identity Matching

The sequence of stages described below are taken from the research by Peláez and
colleagues (2001) and should facilitate the training and testing of conditional discrimina-
tions in infants or other persons who have not yet fully developed language. The step-by-
step protocol mentions procedural problems that may be encountered and a sample data
set. The six basic stages guide instruction from simple touch-screen response training to
generalized identity matching.
Stage 1: Touch-screen response training. This phase of training simply involves teach-
ing the learner to touch a computer screen when a stimulus appears. Young children,
learn faster when seated on their mother’s lap. This is because the caregiver is helpful in
Chapter 4 67
shaping the touch-screen response by modeling, prompting, and signaling appropriate
responding. A familiar assistant may perform these functions for a learner who is severely
disabled. Three stimuli that appear on the screen are directly trained: a picture of an
apple (A), a picture of a baby (D), and a sketch of a bear in a box (G). Each stimulus is
presented in a separate block of trials.
Stage 2: Fading. During this phase of training, comparison stimuli should be gradu-
ally faded in, so that each trial begins to more closely resemble the MTS format. That is,
while A, D, or G appear, two comparison stimuli also gradually appear (fade in) until all
three stimuli are clearly visible and the learner can select or point to the target stimulus.
Pointing is reinforced by contingent stimulation (movement and sounds coming from the
stimulus, while the mother or assistant also touches and praises the learner).
Stage 3: Discrimination training. During this stage, all stimuli should appear simulta-
neously on the screen and the learner must select the appropriate sample. Again, explicit
training of A, D, and G remain in separate blocks.
Stage 4: Identity matching. During identity matching (also called reflexivity training),
A, D, or G each appear individually as a sample with all three stimuli presented as com-
parisons. The learner is required to select the comparison that is an identity match with
the sample (for example, A-A). Again, each target sample appears within a separate block
of trials.
Stage 5: Mixed identity matching. During this stage, the sample stimuli are randomly
presented within one block of trials. A novel B stimulus (for example, the word “apple”)
is also introduced as an alternative comparison.
Stage 6: Generalized identity matching. This is a testing stage with no training. A

series of novel stimuli (B, C, E, F, H, and I) appear as random samples within a block of
twelve trials. Accurate identity matching of these stimuli (for example, B-B and H-H) is
deemed evidence of generalized identity matching because none of these had previously
been included during explicit reflexivity training.
Sample Training Results

In the research reported by Peláez et al. (2001), one child (a twenty-two-month-old
male) readily reached criterion in the initial response training (thirty out of thirty correct
responses) but failed to proceed through discrimination training. Specifically, across nine
blocks of trials, his performance systematically deteriorated from seven out of ten to two
out of ten, at which point the child was removed from the study. In this case, it was
clear that difficulties resulted primarily from fatigue after a training session of longer
than thirty minutes. Notably, this infant demonstrated a significant decline in respond-
ing during discrimination training with the introduction of each new stimulus, thus also
suggesting possible habituation.
For illustrative purposes, the data from a second infant (eighteen months old) are
presented in figure 4.2. This child required more extensive response training (sixty trials)
to reach criterion. His performances during discrimination training began well (seven out

of ten) but then became erratic with the introduction of each new stimulus. However,
he did eventually produce seven out of ten correct responses again. Training in identity
matching was extensive but comprised relatively good performances for at least five blocks
of trials. Nonetheless, performance toward the end declined, again suggesting habituation
Response Discrimination Identity Mixed Generalized

Training Training Matching Identity Identity
Matching Matching
10
7
No. of
6
Correct
Responses 5
4
3
Stimulus A
2
Stimulus D
1 Stimulus G
Stimuli A, D, G
0
Each Block of 10 Trials
and in this case a possible decline in reinforcer efficacy.

Figure 4.2. The data recorded with one learner at each stage of the protocol used by Peláez
et al. (2001).
In summary, training conditional discriminations and training identity matching are

critical precursors to language development, but they are difficult, and there is simply no
easy way to make this happen. The protocol above has been used with some success with
very young or very disabled learners, and common problems encountered in this context
have been noted. The key is to ensure that the target skills can be generalized to novel
stimuli, a capability that will be essential if the learner is to make the crucial transition
between conditional discriminations and derived relations based upon them.
The Concept of Joint Attention

The current chapter offers a conceptual and functional distinction between joint atten-
tion and social referencing, with the view that the former is a necessary prerequisite for
the development of the latter. Joint attention describes the capacity to use eye contact
and cues to coordinate attention with another person in the sharing of an experience
(such as an interesting object or event; Mundy, Sigman, & Kasari, 1994). Put simply, it
comprises shared awareness of a stimulus. Joint attention begins to emerge between nine
and twelve months of age and initially comprises of gaze shifts between a target object
and a familiar person (Bakeman & Adamson, 1984). Consider a three-year-old girl and
her mother visiting family friends. As the adults sit in the living room and chat, the girl
plays with a puzzle on the floor. Suddenly, a kitten runs into the room and the little girl’s
Chapter 4 69
face lights up with surprise and pleasure. However, her next action is not to engage the
kitten in play, but to look up at her mother’s face while pointing to the kitten, to see if her
mother had also witnessed the animal’s dramatic entrance. Gaze shifts may subsequently
be combined with gestures toward the object within the visual field of the familiar face.
Behavioral researchers (including myself) have proposed that an operant (rather than
age-based) history guides the emergence of the skills of joint attention (Dube, MacDonald,
Mansfield, Holcomb, & Ahearn, 2004; Holth, 2005). Specifically, these behaviors nor-
mally result from environmental contingencies that operate during early mother-child
verbal and gestural communications (Peláez, Gewirtz, & Wong, 2007). From this per-
spective, gaze shifts in joint attention incorporate (1) the selective effects of environmental
stimuli that set the occasion for the response class, (2) stimuli that support joint attention
behavioral chains in dual roles as discriminative and reinforcing stimuli, (3) the conse-
quences that lead to the choice of experiencing a stimulus together with the adult versus
experiencing it independently of the adult, and (4) relevant and plausible environmental
conditioning histories. The analysis also identifies the function of reinforcers and suggests
various classes of socially mediated stimuli that maintain joint attention behavior. Indeed,
the most common function of the reinforcers appears to be face-to face interactions with
an adult (Peláez-Nogueras, Field, Hossain, & Pickens, 1996). Put simply, reinforcers are
initially produced by the activity related to the stimulus in question (for example, playing
with a toy) and then increased by adult-generalized social reinforcers such as vocalizations
and smiling, gestures of approval, or demonstrations of affection while engaged. In other
words, it is often more reinforcing for a child to play with a toy or look at a book when
the caregiver participates in the event than it is when the caregiver is absent.
Joint Attention Deficits in Autism

Interest in the concept of joint attention has increased because of its putative role in
developmental disabilities (Carpenter, Pennington, & Rogers, 2002), and Dawson and
colleagues (2004) have even argued that joint attention deficits alone can differentiate
between normally developing learners and those with autism. Specifically, learners with
autism appear to lack prerequisites for joint attention that include orienting to speech
sounds and other social stimuli (for example, when someone points) and show more
direct evidence of deficiencies in joint attention behaviors. For example, Charman and
colleagues (1997) demonstrated that children with autism looked at a mechanical toy
when it was activated but did not exhibit gaze switches between the toy and an adult who
was present.
Deficits in joint attention have also been associated with abnormalities in language
development (Mundy, Sigman, Ungerer, & Sherman, 1986). Specifically, in children with
autism, correlations have been recorded between low frequencies of adult-object gaze
switching at twenty months, limited language gains, and diminished social communica-
tion at forty-two months (Charman et al., 1997). One explanation for the relationship
between language and joint attention suggests that the rapid vocabulary expansion of
typical preschool development depends in part on the learner’s ability to determine, via
observation of adult-attending stimuli, which object in the immediate environment is
related to the adult’s speech.

According to this view, joint attention should warrant a potentially important place
in early intervention programs, especially those aimed at establishing critical language
prerequisites. And yet it is not often given such an important place in these programs.
Indeed, despite its pivotal developmental significance, there are few effective interventions
for ameliorating deficits in joint attention in the literature. The section below sets out a
training sequence for this purpose that may be used with very young learners or those
with developmental delay.
A Protocol for Establishing Joint Attention

The section below contains a description of the key components of a protocol for
establishing joint attention, followed by empirical evidence to support the use of training
regimes such as these in children with autism.
Establishing social reinforcers. The training of joint attention critically requires the
existence of social stimuli, such as nods or smiles, as reinforcers at an early age. This can
be accomplished with the teacher and learner sitting face-to-face, with ten edible reinforc-
ers spread across the table between them. Block any attempts to remove the reinforcers
from the table until the learner is sitting quietly; then nod and smile before allowing the
learner to take one. It is important to emphasize that the learner is only allowed to take a
reinforcer when the teacher nods and/or smiles (in order to make these gestures function
as discriminative stimuli). In addition, you should emit an occasional verbal cue, such as
“yes” or “Look at that” to further improve the learner’s general communication skills.
Of course, this type of training may lead to the possibility that nods and smiles func-
tion as conditioned reinforcers only when treats are available but fail to do so in other
situations. Naturally, this would mean that the learner may not recognize the nods and
smiles of other adults in other contexts. But this situation seems unlikely, or at least rela-
tively easily rectified.
Gaze following. When teaching a child the skill of gaze following, the teacher and
learner should again sit at opposite ends of a small table. First, show the learner a rein-
forcer of choice, and then ask her to turn around while you place the reinforcer under one
of two opaque cups. Then say, “ready,” and allow her to turn around again to observe the
cups. Ask her to point to the cup that she thinks contains the treat. Lift the chosen cup,
and if the treat is there the learner can have it. If the empty cup has been selected, simply
remove the treat and start again.
On a subsequent trial, place your face close to the cup with the treat while maintain-
ing eye contact with the learner, such that she comes to rely on this cue for discriminating
the cup that holds the treat. Continue with this type of training until the learner looks
at your face and consistently chooses the cup with the treat. Next, across trials, fade out
your proximity to the cup, so that eventually the learner can choose the right cup after
only a brief glance on your part.
Learners such as those with developmental disorders may experience difficulty simply
attending to others’ faces. In this case, getting the learner to attend to your face, even
when it is near the cup, will be difficult. In such a situation, it is possible to establish this
skill by saying the learner’s name, holding the treat up to your eyes, and then tracing
Chapter 4 71
a visual path from your eyes to the treat as you place it under a cup. This can also be
trained explicitly. This is to be repeated until the learner chooses the right cup.
Joint object attention. With joint object attention, the learner orients quickly or directly
toward an object once another person’s attention to the object has been discriminated. At
best, the learner should also initiate your attention once a novel object has been identified.
Consider the following scenario described by Jones, Carr, and Feeley (2006). Position a
toy of choice less than five feet away from the learner, activate the toy, and turn and look
at the learner while pointing to the toy and commenting upon it (for example, “Look at
what the car is doing”). It should be possible to get the learner to orient toward the toy
within as little as two seconds of your comment. Then, to improve initiation on behalf of
the learner, have him attend to the toy for several seconds, and encourage him to point
to the object while he looks at you. To reinforce this pointing response, you can simply
model it or physically form the learner’s hand to point at the object.
Mutual object orienting with gestures. Holth (2005) described the following steps for
establishing mutual object orienting with gestures. Attach five or six envelopes to a wall
in a horizontal line. In view of the learner, who is seated approximately ten to thirteen
feet away from you, place an edible reinforcer in one of the envelopes. In order to access
the snack, the learner must guide you through the envelopes. That is, you will begin by
pointing to the envelope farthest away from the one containing the snack, and prompt
the learner to guide you with simple directives such as “left” and “right,” and “stop” when
you reach the correct envelope. You can also arrange the envelopes in a vertical line and
include prompts such as “up” and “down.” Ultimately, you should be able to arrange the
envelopes in a semirandom sequence (some side-by-side and others above and below) and
all direct prompts to the learner should be faded. It is interesting to note some empirical
evidence suggests that this type of intervention not only improves mutual orienting and
gesturing, but is also associated with language gains (Jones et al., 2006).
Empirical Evidence
There is some empirical evidence to support the use of training regimes for the estab-
lishment of joint attention in children with autism. In one study, MacDonald and col-
leagues (2006) investigated joint attention initiations in twenty-one typically developing
children (ages two to four) and twenty-six children with autism. As expected, the children
with autism demonstrated relatively minor deficits in joint attention responding and more
severe deficits in joint attention initiation. While the majority (78 percent) demonstrated
gaze shifts, 44 percent demonstrated use of gestures, and only 22 percent were capable of
related vocalizations. However, after one year of participation in a comprehensive treat-
ment program, all of the children with autism demonstrated gaze shifts, all had gestures,
89 percent could vocalize, and levels of joint attention were now commensurate with the
normally developing counterparts.
A study by McClannahan and Krantz (2006) also demonstrated the remediation of
deficits in joint attention in three children with autism (ages two to five). In this research,
photographic activity schedules were used to cue learners to play with toys in three
locations a puppet theater, toy shelves, and a toy box. When learners initiated use of toys,

they were manually guided to point to the toy while orienting to the teacher. Across trials,
manual prompts were faded (from graduated guidance to spatial fading and shadowing),
and the teacher’s proximity was decreased gradually. The results indicated that all three
children learned to point and orient for attention and could do so with novel stimuli.
The Concept of Social Referencing

Although numerous authors integrate the concepts of joint attention and social referenc-
ing, the current chapter argues that they are distinct and that joint attention essentially
precedes social referencing. Specifically, what social referencing adds to joint attention is
that it also involves the learner reacting to the novel stimulus in a manner that is in accor-
dance with the other’s expression (Peláez-Nogueras & Gewirtz, 1997). Consider again the
previous example of the three-year-old’s surprise when the kitten runs into the room. As
part of her joint attention skills, the child looks up at her mother while pointing to the
kitten, but then she engages in social referencing when she sees her mother make a fearful
face and as a result avoids approaching the kitten.
As well as incorporating the component of concordant responding (in other words,
using the reactions of others as discriminative stimuli for one’s own responding), social
referencing also appears to comprise an emotional component. In other words, it extends
beyond the simple sharing of information and also facilitates the learner’s emotional reac-
tion to stimuli. This emotional aspect of social referencing appears to make up a four-stage
process that involves recognizing emotional expressions, understanding emotional expres-
sions, responding to emotional expressions as cues, and altering behavior in accordance
with changes in emotional expression.
Cognitive-developmental psychologists view the informational and emotional com-
ponents of social referencing as separate processes. Specifically, they distinguish between
instrumental social referencing, which involves the learner’s use of knowledge from others
as indicators of how to “understand” stimuli (Feinman, 1982), and affective social refer-
encing, which involves the learner’s use of others’ emotional facial expressions to deter-
mine how to feel about ambiguous events (Klinnert, Campos, Sorce, Emde, & Svejda,
1983).
Learning theorists such as myself, however, have argued that both types of social
referencing are outcomes of the same conditioning process, because the cues that convey
affective components also contain instrumental information, and vice versa. The study
by Gewirtz and Peláez-Nogueras (1992) provided some empirical support for this view,
as well as examples of how the emotional aspects of social referencing can be explicitly
trained.
In contexts of ambiguity, we identified two originally meaningless maternal facial
expressions and then trained them with standard conditioning procedures to denote
opposite consequences for responses where infants reached for objects. Hence, one mater-
nal hand-to-face expression was trained to predict positive auditory-kinetic consequences
of the infant reaching for ambiguous objects (see figure 4.3), while the other maternal
hand-to-face expression was trained to predict negative auditory-kinetic consequences of
the infant reaching for ambiguous objects (see figure 4.4).
Chapter 4 73
Figure 4.3. Mother signals joyful cue to infant that predicts pleasant musical sound,
contingent on infant reaching for object.
Figure 4.4. Mother signals fearful cue to infant that predicts loud sound and movement
of object, contingent on infant reaching for object.

What this research demonstrates is that the extent to which an infant orients to the
mother’s face for cues in contexts of uncertainty depends on past success in obtaining
such information, its validity, and its utility. For training purposes, therefore, either with
very young learners or with those who are developmentally disabled, the cues of others
must be consistently contingent on the learner’s object-referencing behavior and must
reliably predict environmental consequences for the learner’s approach or avoidance. This
interpretation is summarized in Figure 4.5.
Figure 4.5. A learning approach to a social referencing paradigm.
From this perspective, social referencing is an example of social knowledge with an

emotional component. Put simply, the experienced learner gains the knowledge that if
another is smiling when a stranger approaches, reinforcement is likely; but if the other
person is cringing, for example, reinforcement is not likely. Thus, the facial expression
of the other becomes a setting event that establishes the function of the stranger as
being discriminative for positive or negative reinforcement or aversive consequences for
approaching. In line with this analysis, it should then be possible to establish learners’
responses to the basic emotions displayed by others and how they should act on this basis.
Once these have been established, it is likely that a whole array of more subtle emotional
reactions and appropriate response patterns will be trainable within the context of simple
conditioning paradigms.
Why Are Joint Attention and Social Referencing

Important for Derived Relational Responding?
Joint attention and social referencing would seem to have an important role in the
establishment of derived relational responding, thus forming the core of language and
Chapter 4 75
higher cognition. For example, it seems likely that the emotional and social aspects of
social referencing form the basis of the later development of perspective taking. That
is, reciprocal conversation, cooperative play, and displays of sympathy and empathy for
others are all social abilities that require the basics of joint attention and social referenc-
ing, because without them you would not use the ongoing cues of others to determine
how they were feeling and to act accordingly. Hence, it is not surprising that individuals
with autism who present with deficits in social referencing, for example, subsequently
develop considerable delays in their social and emotional skills.
Concluding Comments
Joint attention and social referencing are an intricate part of the tapestry of social interac-
tions that comprise normal development. Not only are they critical to the development
of social and related emotional repertoires, but they also appear to be essential precursors
to conditional discriminations and identity matching, which are also important precur-
sors to language development and its core process of derived relational responding. The
current chapter described teaching strategies for establishing conditional discriminations,
joint attention, and social referencing in young and developmentally disabled learners.
Despite the importance of these skills, such training is far from easy. But there is simply
no way around this—if language and social and emotional development are desired and
potentially within the capabilities of the learner, then the difficulties must be endured
and the teacher must generate increasingly clever and creative ways to make the train-
ing work. Although empirical evidence in support of the various teaching strategies out-
lined is still scant, they offer good first steps toward the establishment of these essential
building blocks of human development.
References
Augustson, K. G., & Dougher, M. J. (1992). Teaching conditional discrimination to
young children. Experimental Analysis of Human Behavior Bulletin, 9, 21–24.
Bakeman, R., & Adamson, L. (1984). Coordinating attention to people and objects in
mother-infant and peer-infant interaction. Child Development, 55, 1278–1289.
Carpenter, M., Pennington, B. F., & Rogers, S. J. (2002). Interrelations among social-
cognitive skills in young children with autism. Journal of Autism and Developmental
Disorders, 32, 91–106.
Charman, T., Swettenham, J., Baron-Cohen, S., Cox, A., Baird, G., & Drew, A. (1997).
Infants with autism: An investigation of empathy, pretend play, joint attention, and
imitation. Developmental Psychology, 33, 781–789.
Dawson, G., Toth, K., Abbott, R., Osterling, J., Munson, J., Estes, A., et al. (2004). Early
social attention impairments in autism: Social orienting, joint attention, and atten-
tion to distress. Developmental Psychology, 40, 271–283.

Devany, J. M., Hayes, S. C., & Nelson, R. O. (1986). Equivalence class formation in
language-able and language-disabled children. Journal of the Experimental Analysis of
Behavior, 46, 243–257.
Dube, W. V., MacDonald, P., Mansfield, R. C., Holcomb, W. L., & Ahearn, W. H. (2004).
Toward a behavioral analysis of joint attention. Behavior Analyst, 28, 197–207.
Feinman, S. (1982). Social referencing in infancy. Merrill Palmer Quarterly, 28,
445–470.
Gewirtz, J. L., & Peláez-Nogueras, M. (1992). Infant social referencing as a learned
process. In S. Feinman (Ed.), Social referencing and the social construction of reality in
infancy. New York: Plenum.
Gewirtz, J. L., & Peláez-Nogueras, M. (2000). Infant emotions under the positive-
reinforcer control of caregiver attention and touch. In J. C. Leslie & D. Blackman
(Eds.), Issues in experimental and applied analyses of human behavior. Reno, NV:
Context Press.
Higbee, T. S., & Peláez-Nogueras, M. (1998). Reinforcer identification in infants.
Behavioral Development Bulletin, 7, 10–14.
Holth, P. (2005). An operant analysis of joint attention skills. Journal of Early and Intensive
Behavioral Interventions, 2, 160–175.
Jones, E. A., Carr, E. G., & Feeley, K. M. (2006). Multiple effects of joint attention inter-
vention for children with autism. Behavior Modification, 30, 782–834.
Klinnert, M., Campos, J. J., Sorce, J. F., Emde, R. N., & Svejda, M. (1983). Social ref-
erencing: Emotional expressions as behavior regulators in emotion. Theory, Research,
and Experience, 2, 57–86.
MacDonald, R., Anderson, J., Dube, W. V., Geckeler, A., Green, G., Holcomb, W.,
et al. (2006). Behavioral assessment of joint attention: A methodological report.
Developmental Disabilities, 27, 138–150.
McClannahan, L. E., & Krantz, P. J. (2006). Teaching conversation to children with autism:
Scripts and script fading. Bethesda, MD: Woodbine House.
Mundy, P., Sigman, M., & Kasari, C. (1994). Joint attention, developmental level and
symptom presentation in autism. Development and Psychopathology, 6, 389–401.
Mundy, P., Sigman, M. D., Ungerer, J., & Sherman, T. (1986). Defining the social defi-
cits of autism. The contribution of non-verbal communication measures. Journal of
Child Psychology and Psychiatry and Allied Disciplines, 27, 657–669.
Peláez, M., Gewirtz, J. L., Sanchez, A., & Mahabir, N. M. (2000). Exploring stimulus
equivalence formation in infants. Behavior Development Bulletin, 9, 20–25.
Peláez, M., Gewirtz, J. L., & Wong, S. E. (2007). A critique of stage theories of human
development: A pragmatic approach in social work. In B. A. Thyer (Ed.), Comprehensive
handbook of social work and social welfare: Vol. 2. Human behavior in the social envi-
ronment. New York: Wiley and Sons.
Chapter 4 77
Peláez, M., Lubián, J., McIlvane, W., & Dube, W. (2001). Training discrimination, reflex-
ivity, mixed identity matching, generalized identity matching, and arbitrary matching in
infants: Towards the “emergence” of stimulus equivalence. Presented at the first interna-
tional conference of the Association for Behavior Analysis, Venice, Italy.
Peláez-Nogueras, M., Field, T., Hossain, Z., & Pickens, J. (1996). Depressed mothers’
touch increases infant positive affect and attention in still-face interactions. Child
Development, 67, 1780–1792.
Peláez-Nogueras, M., & Gewirtz, J. (1997). The context of stimulus control in behavior
analysis. In D. M. Baer & E. M. Pinkston (Eds.), Environment and behavior. Boulder,
CO: Westview Press.

Chapter 5
Establishing Mand and Tact Repertoires
Linda A. LeBlanc and Courtney M. Dillon, Western Michigan

University; and Rachael A. Sautter, Y.A.L.E. School
Children with autism typically exhibit substantial speech delays during early childhood,
with up to 50 percent failing to develop any speech (Charlop & Haymes, 1994). Those
who do develop speech often engage in echolalia, repeating words or phrases heard previ-
ously, or do not speak for social purposes such as engaging in conversation (Smith, 1999).
Because one of the best predictors of outcome for children with autism is the develop-
ment of spontaneous language before six years of age (Szatmari, Bryson, Boyle, Streiner,
& Duku, 2003), it is essential to develop effective programs to teach language to children
with autism. This chapter describes behavioral techniques that can be used to teach basic
language to young children with autism in accordance with B. F. Skinner’s analysis of
language.
Skinner’s Analysis of Verbal Behavior

As discussed in chapter 3 of this volume, Skinner (1957) defined language or verbal
behavior as behavior reinforced through the responses provided by other people, and he
defined several verbal operants according to the relevant antecedents and consequences.
He classified language according to specific functional operants (for example, antecedent-
behavior-consequence groups), rather than the traditional units of language (such as words
and phrases). This chapter will focus on two of Skinner’s elementary verbal operants, the
mand and the tact. In common terms, a mand is a request (such as “May I have that?”)
or an imperative (such as “Give it to me”). The mand tells the listener what response or
item will serve as a reinforcer, and it occurs because of a highly specific motivation called
an establishing operation (EO; Michael, 1988). An EO has been defined as a change in
the environment that momentarily increases or decreases the power of a reinforcer and
evokes behaviors that have previously produced access to that reinforcer (Michael, 1982).
For example, hunger increases the reinforcing value of food and evokes behaviors that
have produced access to food in the past (for example, asking Mom for lunch or cooking).
A tact is a label or description that occurs in the presence of a nonverbal discriminative
stimulus or condition that results in social reinforcers. For example, in the presence of
an apple, a child who says the word “apple” would be praised; if the child said the word
“orange,” he or she would not. Much of early language development consists of these two
verbal operants as children learn to ask for things that they want and describe things that
they see. Although they are not the focus of this chapter, other elementary verbal operants
such as echoics (or verbal imitation) and textual behavior (the term for reading), are some-
times used as prompts in training mands and tacts. These verbal operants are conceptual-
ized as functionally independent such that a child who mands “cookie” when he wants
one may not be able to tact, or name a cookie when he sees a picture of one, a pattern that
has been repeatedly demonstrated to occur in children with language disorders (Nuzzolo-
Gomez & Greer, 2004; Sigafoos, Reichle, Doss, Hall, & Pettitt, 1990).
Applying Skinner’s Analysis of Language

to Instruction
Sundberg and Michael (2001) outline several benefits of using Skinner’s analysis of lan-
guage to teach children with autism. First, Skinner’s identification of the functional inde-
pendence of the verbal operants indicates that each one has to be independently and
directly targeted or it may not develop. In other words, children with autism can only be
expected to develop functional mands or requests if they are taught in the presence of an
important motivational variable (for example, learning to ask for water when thirsty) and
not just as a product of training to tact (label) a glass of water (Twyman, 1996). Sundberg
and Michael (2001) also identify the mand as the type of operant to teach first to children
with autism because the mand is uniquely valuable to the child and doesn’t require social
reinforcers. That is, a child learns to mand for things he or she desires when the desire
exists, allowing the child some control over his or her environment (for example, “I ask
for an item and receive it”). In addition, the child receives a powerful reinforcer follow-
ing a verbal response (in other words, the child receives what he or she asks for), which
can increase the reinforcing properties of speaking, resulting in a greater likelihood of
language development in the future.
When one is preparing to teach mands and tacts, it is critical to assess the child’s
existing repertoires for two purposes. First, one must determine the current profile of
functional language in order to identify which operants should be targeted and to deter-
mine which existing operants can be used to prompt weaker or emerging operants.
Second, one must determine the extent of existing prerequisite repertoires, which could
affect your selection of topography of the target response. Each of these areas is covered
briefly below.

Functional Language Abilities
The Assessment of Basic Language and Learning Skills (ABLLS) is one tool based
on Skinner’s analysis that allows therapists to identify a child’s specific language deficits
in terms of verbal operants (Partington & Sundberg, 1998). Another tool is the Verbal
Behavior Milestones Assessment and Placement Program (VB-MAPP; Sundberg & Hale,
2007). Each tool evaluates Skinner’s verbal operants to determine whether a child has a
specific verbal operant in his or her repertoire and allows you to track development via
repeated administration. Each verbal operant is assessed under the relevant antecedent
conditions and with the corresponding consequences in place to determine if the child
readily emits that verbal operant. The information obtained from the ABLLS can inform
parents and teachers about how to prioritize language targets. However, generally chil-
dren should learn mand (requesting) and echoic (verbal imitation) skills before moving
on to tacts (labels), questions (mands for information), and conversation (intraverbals;
Sundberg & Partington, 1998). The obtained profile can also allow the teacher or parent
to identify strong repertoires (such as echoics) that can be used in teaching other operants
(such as mands). (See the section “Transfer of stimulus control,” below.)
Response Topography
Selecting a modality for responding is critically important to success and maintenance
in language instruction with children with autism. Language, or verbal behavior, does not
have to be vocal (in other words, spoken) to be meaningful or functional. Verbal behavior
can also include manual signing (Tincani, 2004; Bartman & Freeman, 2003), selection
responses such as picture exchange (Charlop-Christy, Carpenter, Le, LeBlanc, & Kellet,
2002; Chambers & Rehfeldt, 2003) or use of voice output devices (Mirenda, 2003). All
potential communication modalities should be considered for all children, with modality
selection based on three important factors: available audiences, practicality or portability
of the system, and child repertoires.
Vocal responses are ultimately portable and have broad potential listener communities,
assuming adequate articulation, but they have the drawback of requiring complex vocal
musculature manipulations that cannot be directly prompted. Picture exchange commu-
nication systems (PECS) and voice output communication aids (VOCA) can be used
with almost any listener community but must be portable, well-organized, and constantly
available. Researchers have proposed several drawbacks associated with selection-based
systems such as PECS. Sundberg and Michael (2001) distinguish between topography-
based systems, where there is a unique topographical response (such as a spoken word
or sign) for each communicated idea, and selection-based systems, where the series of
responses (such as scanning, selection, and picture delivery) is identical for each com-
municative event. Sundberg and Partington (1998) argue that selection-based systems
often require more complex skills than initial appearance would suggest, with increased
probability of difficulties in acquisition as language concepts become more abstract and
difficult to depict visually. Sign language has the benefits of being topography based and
amenable to modeling and manual guidance, but it has the drawback of a restricted
verbal community (such as a deaf signing community and those trained to “listen” to
Chapter 5 81
the child) and may result in idiosyncratic or incomplete signs (the sign equivalent of poor
articulation or volume control).
Individual abilities and deficits that have an impact on failure to develop the mand
repertoire should also guide selection of a communication modality and mand training
procedures (Tincani, 2004; Wraikat, Sundberg, & Michael, 1991; Bourret, Vollmer, &
Rapp, 2004). Children may fail to exhibit mands due to dearth of items with reinforcing
properties, lack of a meaningful trained mand response, limited reinforcement opportuni-
ties (for example, poor articulation, and so few viable listeners), or because EOs were not
present during training and do not occasion responding. Bourret and colleagues (2004)
found that mand training is more effective when the training procedures target specific
problems. For example, one might target a lack of response by selecting a topography
based on the child’s existing repertoires; vocal sounds and verbal imitation may indi-
cate that spoken language should be pursued, whereas limited vocal imitation repertoire
with good fine motor and motor imitation skills may suggest sign language, PECS, or
VOCA.
The Mand: Importance and Types

Development of a functional mand repertoire can decrease problem behavior in children
with autism (Durand & Merges, 2001) by replacing inappropriate behaviors (such as tan-
trums and aggression) with less effortful and more appropriate means to obtain the same
functional reinforcers (such as attention, specific tangibles, or a break from an aversive
stimulus). Thus, teaching children to mand early in their verbal training can allow them
some control over their environment (terminating aversive stimuli, for example) and may
decrease the occurrence of problem behavior if the vocal response that is taught is less
effortful to emit than the problem behavior that is maintained by the same functional
reinforcers.
For the purposes of this chapter, mands will be divided into simple mands and mands
for information. Initial language training should focus primarily on establishing a robust
repertoire of simple mands across different environments and with different people.
Simple mands are those basic requests for preferred items (such as toys, food, or
drink), people, changes in the environment (for example, “Let’s go outside”), and termi-
nation of unpleasant situations (such as “Stop, please” or taking a break). These mands
should be targeted as soon as a clear preference or motive is identified, and targeted con-
tinuously until a strong and spontaneous repertoire exists across environments and with
multiple people. A spontaneous repertoire means that mands occur under EO control (for
example, hunger, thirst, or boredom) rather than as the result of a prompt from another
person (for example, “What do you want?”), such that the child is able to use language
to meet his or her needs when they occur rather than waiting for someone to notice his
or her distress.
As training continues with preferred items, mands for missing preferred items should
be periodically tested to ensure that the relevant EO (such as hunger), rather than a
nonverbal discriminative stimulus (such as the sight of food), is controlling the child’s
requests. For example, requests for a meal that occur only when food is visible or can be

smelled and not when the child is hungry would not be particularly functional. Mands for
missing items should be gradually introduced by conducting trials with known mands in
the absence of the discriminative stimulus (in other words, the actual preferred item). For
example, a child who likes cookies and is hungry might be given a small visible portion of
a cookie upon request followed by an opportunity to mand for additional cookies that are
not visible. Certain manipulations can be employed in order to evoke a mand for an item
in the absence of the actual item. For example, placing the child in a contextually relevant
environment (for example, bringing the child to the kitchen and orienting him or her
toward the cabinet where the cookies are located), but keeping the actual item that is to
be requested out of sight, may evoke this mand. The PECS also includes a phase for train-
ing mands in the absence of the actual item. This occurs in phase two of training, where
the items and the communication book are moved farther away from the learning and
typically out of the direct line of vision (Bondy & Frost, 2002). Mands that occur when
an item is missing are considered pure mands because of the certainty one can have that
the EO, rather than a visible discriminative stimulus, is the relevant antecedent variable.
Mands for removal of aversive stimuli should be taught in naturally occurring contexts
as soon as the aversive stimuli can be readily identified and removed. Often practitioners
fail to recognize the importance of teaching a specific mand for a break, for help, or to
stop until problem behavior maintained by escape from aversive situations has been well
established and must be subsequently replaced. Often, one to two generally applicable
mands for removal of aversive stimuli will suffice across many situations, while a large of
number of specific mands for items are required to produce functional repertoires. Mands
for removal of aversive stimuli should be targeted exclusively through captured EO pro-
grams, as opposed to contrived EO programs, to avoid creation of unnecessary distress
and increased likelihood of the therapist becoming a conditioned aversive stimulus.
A mand for information is a subtype of mand (Sundberg, Loeb, Hale, & Eigenheer,
2002) that specifies certain information as the desired reinforcer (for example, “Who has
my Elmo?” “What time is it?” “Where are my keys?” or “What is your name?”). There
are often at least two pertinent reinforcers (Sundberg et al., 2002): the information itself
and the outcome that the information allows the person to achieve. The child benefits
from the information because it allows quicker and more efficient problem solving than
a trial-and-error strategy (for example, approaching each person in the room to look for
the doll) would provide.
A mand for information is more complex than a simple mand because it requires
a child to identify the necessary information and a person who might have it, and to
formulate a reasonable and understandable question. Some mands for information occur
in a purely social context and have only one reinforcer (in other words, the informa-
tion itself). For example, “Where did you go on vacation?” produces information that
facilitates ongoing conversation and social connection but does not provide any other
tangible stimulus change in the environment. Mands for information should be targeted
when a child has an extensive tact and developing intraverbal repertoire. Initial training
should focus on mands for concrete and powerful secondary reinforcers in addition to
information (for example, “Where is my favorite toy?”), while mands for purely social
information should be targeted later because children with autism may not strongly value
social information or small talk.
Chapter 5 83
Procedures for Teaching Mands
The goal of mand training is to teach a child to request an item or activity in response
to the relevant EO without a prompt (for example, “What do you want?”). This can be
accomplished by ensuring that the relevant EO is present during learning trials (Sundberg
et al., 2002).
EO Manipulations
As stated previously, mands taught in the presence of an EO are more likely to occur
whenever this EO is present, even if the EO is present in novel environments or with a
variety of people (LeBlanc, Esch, Sidener, & Firth, 2006). Two primary strategies for
ensuring that training occurs in the presence of the EO are to capture naturally occurring
EOs for learning opportunities, and to create the EO as a specific part of the instructional
plan (Shafer, 1994). Most practitioners use both of these approaches at different points in
a child’s day.
Captured EOs: Incidental teaching. Several studies illustrate the beneficial effects of
incidental teaching with children with autism (Fenske, Krantz, & McClannahan, 2001;
McGee, Morrier, & Daly, 1999), in which the child “initiates” the learning trial as he
or she navigates the natural environment and the therapist monitors the child’s interests
and creates related learning opportunities. Incidental teaching involves five steps (Hart
& Risley, 1980). First, allow the child to interact naturally with his or her environment.
Second, watch the child closely to identify an opportunity for the child to request a
desired or needed item in the environment. Third, respond to the opportunity by requir-
ing that the child request the item in his or her communication modality (for example,
vocal, sign, PECS, or VOCA). Fourth, use shaping or transfer of stimulus control pro-
cedures (see below) to facilitate an appropriate response from the child if the child does
not immediately emit the appropriate mand. Finally, immediately give the requested item
to the child when he or she gives an appropriate response and, if necessary, praise the
child for emitting the request (for example, say, “Good job asking for the cookie”). As
the mand response becomes stronger, eliminate the praise and provide only the requested
item. See the “Captured Establishing Operations: Incidental Teaching with Time Delay”
program for a detailed list of steps for using incidental teaching to teach mands which
can be used with the “Mands Data Sheet.” A program for teaching mands for removal
of aversive stimuli can be found at the end of the chapter (see “Capturing Establishing
Operations: Terminating Aversive Stimuli”).
Contriving EOs. Another way to ensure the presence of the relevant EO during mand
training is to intentionally create or contrive situations in which an item or event becomes
momentarily highly reinforcing immediately before you prompt responding. Contriving
an EO essentially consists of “tripping a child up” so that the child must request an
item or event to ensure that a preferred or needed activity or event can occur. Similar to
incidental teaching, the therapist or parent has to be continually vigilant in watching for
situations that would make teaching opportunities. However, contriving EOs differs from
incidental teaching in that the parent or therapist is arranging the environment to occa-
sion the trial rather than relying on naturally occurring situations to occasion the child’s

interest. One might contrive thirst as an EO by providing salty foods (thereby momen-
tarily increasing the reinforcing properties of a drink) and blocking access to a drink until
the child asks for it. One might also cover a light switch with a hand as the child enters
a room in order to increase his or her motivation to request that the light be turned on.
Placing preferred toys on a high shelf can set the occasion for mands as well. Finally, one
might hold a child’s swing to block the forward progress until the child requests (either
with or without prompts), “Let go!” or “Swing!”
Several studies illustrate the beneficial effects of identifying behavioral chains and
contriving an EO by interrupting the chain at some point (Sigafoos, Kerr, Roberts, &
Couzens, 1994). Interrupted chain procedures typically begin with teaching a child to
complete a chain or targeting a previously learned chain (for example, eating cereal and
milk with a spoon, or cutting out a circle from construction paper). At some point in
the chain the instructor prevents access to a component required for the chain (such as
a spoon or scissors), creating an EO relevant to a mand trial for the relevant item. Note
that these interrupted chain procedures are particularly useful for testing and targeting
pure mands for missing items because the child requests the item when no visual cue for
the item is present.
There are four steps to the interrupted chain instructional procedure to teach mands
(Duker, Kraaykamp, & Visser, 1994). First, a behavioral chain requiring several steps
(such as making a sandwich or preparing and eating cereal) that the child can do inde-
pendently should be identified. Second, a step of the chain should be identified as the
targeted mand (for example, “Knife, please,” when all other items needed to make the
sandwich are available, or “Spoon, please,” when all other components for serving and
eating breakfast cereal are available). Third, the EO is contrived by creating a learning
trial where the child has all the items needed to complete a task except for one item. In
the sandwich example, the child would sit at a table with peanut butter, jelly, bread, and
a plate, and the teacher or parent would say, “Let’s make a sandwich!” Allow the child to
initiate the chain and potentially request the needed item independently before model-
ing the response or using some other transfer of stimulus control technique to facilitate
responding (see below). Finally, immediately provide the requested item upon request.
See the “Contrived Establishing Operations: Interrupted Chains” program for the steps
involved in using an interrupted chain procedure to teach mands and the “Interrupted
Chains: Sample Chains” list for potential chains. Use with the “Mand Training Data
Sheet.” A program for teaching mands for a missing item in a game-type context also can
be found at the end of this chapter (see “Contrived Establishing Operations: The ‘What’s
in the Bag?’ Program”).
Supplemental Teaching Procedures

In addition to the EO manipulation procedures described above, general instructional
strategies such as shaping and transfer of stimulus control procedures should be incorpo-
rated to increase the effectiveness of mand training. Errorless learning procedures are also
recommended such that early trials have immediate prompts sufficient to produce the
desired response, which are subsequently delayed (in other words, time delay) or altered
(in other words, partial vocal prompt) and gradually eliminated. See Green (2001) for
a review of errorless learning and other stimulus control technology procedures. These
supplemental procedures are equally appropriate for tact training.
Chapter 5 85
Shaping. Shaping is defined as differential reinforcement of successive approximations or
providing reinforcers for behaviors that closely resemble the behavior that the child needs
to learn (Baum, 2005). Thus, in teaching the response “bubble,” an instructor might first
reinforce the sound “buh,” “bah,” or even “beh” for several trials while listening closely for
variable responses, reinforcing sounds that even more closely resemble the word “bubble”
(such as “bub” and “buhl”), and placing prior responses on extinction (in other words, no
longer providing reinforcement). This process continues until the child’s response closely
matches the target response, in this case, “bubble.”
Harris (1975) discussed the use of shaping to teach language to nonverbal children.
She said that the five steps in using shaping to teach language are the following: First,
ensure that the child is attending to the teacher, and that the child has the ability to
sustain this attention for a short period of time. Second, Harris suggests that the therapist
begin teaching by reinforcing all of the child’s vocalizations. Third, the therapist should
reinforce only vocalizations that occur within six seconds of the model. Fourth, the thera-
pist should reinforce all vocalizations that occur within six seconds of the model that
resemble the model. Finally, the therapist should provide a novel model and repeat the
steps described previously. The therapist should continue to quiz the child on previously
learned words to make sure his or her skills on the newly learned word do not decrease
because they are not being used.
Transfer of stimulus control. All of the procedures listed above can be enhanced with
the introduction of prompts to occasion the target response or a close approximation.
When prompts are used, stimulus control over the behavior is established with the bridge
stimulus (echoic prompt, picture prompt, and so on) and stimulus control is gradually
transferred to the relevant EO as those prompts are removed. Selection of the prompt
type should be based on evaluation of the strength of other verbal operants, with strong
operants used to support the emergence of weaker operants. Children with strong echoic
repertoires (in other words, they display ready responses to a variety of vocal models)
would benefit from echoic prompts. For example, a teacher might use incidental teaching
to capture a naturally occurring EO of interest in a nearby sink to instigate teaching trials
for water (turning it on, playing, or drinking) by providing the relevant vocal models at
the right times (in other words, saying “on” or “water” while standing with his or her
hand on the faucet handle, or saying “drink” while water is running and the teacher is
holding a glass). Children with strong tact repertoires (in other words, they display ready
responses to an object or picture) would benefit from tact prompts (Arntzen & Almas,
2002). For example, a therapist might prompt a child who can tact a glass of water (saying
“water” when a glass of water is present) in the presence of the EO by giving him or her
salty foods, presenting the glass of water, and responding to the child’s tact “water” with
a glass of water to establish reduction of thirst as the maintaining reinforcer, and thirst as
the controlling antecedent stimulus.
After several successful prompted trials, stimulus control can be transferred by delay-
ing the vocal model (echoic prompt) or altering the prompt (for example, the tact prompt
changes from a picture of a glass to a faint outline of a glass, or a partial echoic prompt).
A commonly used strategy for transferring stimulus control with children with autism is
institution of a time delay before the presentation of the prompt (Charlop, Schriebman,
& Thibodeau, 1985). After successfully establishing stimulus control with the prompt,
the therapist pauses before the presentation of the vocal model or prompt to allow an

opportunity for the child to respond to the EO in isolation, rather than the combined EO
and echoic or tact prompt condition used in previous trials, often termed a spontaneous
request. Time delay can be implemented as a fixed delay (for example, five or ten seconds
on all trials) or a gradually increasing delay across trials (for example, one second, two
seconds, four seconds, six seconds) until the child responds during the delay interval.
Fixed time delay is favored for mand training with visible, tangible stimuli and after
many mands have been mastered; graduated time delay is recommended for initial mand
training (Charlop-Christy, LeBlanc, & Carpenter, 1999).
The Tact: Importance and Types

In common vernacular, tacts are labels or descriptions emitted by the speaker. Skinner
defined a tact as a response that is “evoked by a particular object or event or property of
an object or event” (1957, p. 82). Thus, a tact is occasioned by the presence of a discrimi-
native stimulus and is maintained by social reinforcers (such as praise or continued inter-
action with the listener). For example, in the presence of an apple, the responses “apple”
or “fruit” would be praised, while the response “orange” would likely be corrected (“No,
that’s an apple!”).
The tact differs from the mand in that the controlling antecedent is a nonverbal
stimulus rather than an EO, and the reinforcer is a general social one rather than specific
access to a preferred event or stimulus change. For example, a child who sees an apple
and says “apple” in response to a parent query “What do you call this?” has emitted a
tact. The hungry child who says “apple” and immediately consumes the provided snack
has manded for the apple. Unlike mands, which are directly beneficial to the speaker,
tacts are primarily beneficial for the listener, as they provide descriptive information that
may enhance understanding or provide specificity. A child who tells a peer, “Look in the
tree!” (mand) may provide additional specificity by describing, “That black kitty is going
to jump” (tact). The listener now knows exactly what to attend to and can respond in a
social interchange, “That’s too high. It might get hurt!” Unfortunately, generalized social
reinforcers are notoriously unmotivating for young children with autism compared to
other types of reinforcers (in other words, tangibles); however, tacts are a basic building
block for conversation and are critical for success in general education environments.
The tact also differs from listener behavior, which is often referred to by various terms
including receptive discrimination, receptive labeling, or receptive vocabulary. Tacts refer to
the behavior of the speaker (in other words, naming or describing a critical aspect of the
environment), while listener behavior refers to one’s ability to respond in various ways to
the important aspects of the environment specified by others. In the example provided
above, the speaker is directing a listener to observe a specific aspect of the environment,
the black kitty in the tree (tact), while the selection response of touching the picture of
the black kitty in an array of pictures of different animals would be a receptive discrimi-
nation (listener behavior). These two repertoires are related but functionally distinct and
will not necessarily develop without the direct targeting of each. Tacts are beneficial in
that they can facilitate social interactions, they facilitate development of generalized con-
ditioned reinforcers by pairing weak social reinforcers (such as praise) with potent pre-
ferred items (Partington, Sundberg, Newhouse, & Spengler, 1994), and they may reduce
Chapter 5 87
the amount of nonfunctional language emitted by children with autism (Karmali, Greer,
Nuzzolo-Gomez, Ross, & Rivera-Valdes, 2005).
Several basic tact targets should be distinguished and ordered with respect to curricu-
lum. Initial training should focus on naming of familiar three-dimensional objects (for
example, toys, animals, clothing, foods, and body parts) and people (for example, teacher
and siblings), followed by naming of two-dimensional representations of those objects
and people. Subsequent targets include features or aspects of familiar objects (such as size,
color, and shape), followed by specific location types (for example, kitchen or playground).
More advanced tact targets include actions (such as jumping, rolling, or flying) and func-
tions and classes followed by relational tacts including prepositions (for example, in/out
of the box, above/below, and in front of/behind) and relational descriptors (big/little, less/
more, or slow/fast).
Procedures for Teaching Basic Tacts

Procedures for transfer of stimulus control are well suited to teaching tacts (Barbera
& Kubina, 2005) when echoic and mand repertoires are previously established and when
shaping and errorless learning procedures are incorporated (Partington et al., 1994).
Specific procedures are described below for structured work situations and play-based
interventions such as the Natural Language Paradigm (NLP; Koegel, O’Dell, & Koegel,
1987).
Transfer of Stimulus Control

Procedures for transfer of stimulus control establish responding under some version
of multiple control with either echoic prompts (for example, “G.I. Joe”), EO control (in
other words, manding), or both, and the presence of the nonverbal stimulus (in other
words, tacting). Targeting multiple controlled responses rather than pure operants may
result in higher compliance and greater enjoyment for the learner without sacrificing speed
of acquisition or overall strength of the learned repertoire (Braam & Sundberg, 1991).
Once responding is established, the extraneous controlling variables are faded until only
the nonverbal stimulus occasions responding. The clinician should initiate training with
the fewest possible extraneous controlling variables, because fading each variable requires
additional time and effort. Thus, a child with a strong echoic repertoire and reasonable
levels of compliance might only require echoic prompts (for example, “Say ‘boat’” in the
presence of a toy boat or a picture of a boat) without the need for manipulation of EOs
to occasion responding. Children with more limited echoic repertoires or who consis-
tently emit problem behavior in response to demand situations might benefit more from
inclusion of EO control in establishing initial responding and might also perform more
consistently in a play context.
To use transfer of stimulus control procedures to teach tacts in a structured teaching
environment, first identify a robust verbal operant repertoire (for example, mand, echoic,
or intraverbal) and base your prompting strategies on that operant. Next, have the child
sit at a table and present the nonverbal discriminative stimulus (in other words, show the
item to the child). If you are using errorless prompting strategies, immediately provide the

relevant prompt (for example, if the verbal operant is echoic, the prompt might be “red”).
Respond to accurate responses with immediate praise and, if needed, either unrelated
reinforcers or the item (in other words, mand transfer). Note that as the tact repertoire
progresses beyond simple naming, you must present the relevant intraverbal prompt (such
as “What is it?” or “What color is it?”) simultaneously with multiple exemplars of the
nonverbal stimulus (for example, red ball, red square, and red car) and appropriate nonex-
emplars (for example, blue ball, pink ball, orange ball) to ensure that the relevant stimulus
feature will control responding. See the “Tact Training: Names, Features, and Actions”
program for a sample of specific instructions, though you will need to create other specific
programs for specific tact targets. Programs for advanced tacts (such as prepositions or
tacts of sensory experiences) can be found at the end of this chapter (see “Tact Training:
Prepositions” and “Tact Training: Sensory Experiences”). These programs often work well
when receptive program trials are intermixed with tact trials (see chapter 7 for additional
information). See the “Tact Training Data Sheet” for a model data sheet for all tact
programs.
Tact responses to intraverbal prompts (such as “What is that?” or “What color is
that?”) constitute a functional repertoire; however, do not expect tacts taught exclusively
as responses to intraverbal prompts to occur as independent descriptions of the envi-
ronment (for example, “There’s your red ball”) in conversations. Conduct training (see
the “Play-Based Tact Training: Natural Language Paradigm” program and data sheet) in
play situations with natural change agents (for example, parents) to facilitate this use of
language (Gillett & LeBlanc, 2007; Koegel et al., 1987; Laski, Charlop, & Schreibman,
1988).
Procedurally, Natural Language Paradigm (NLP) has several steps (Charlop-Christy
et al., 1999; Gillett & LeBlanc, 2007; LeBlanc et al., 2006). Adult and child face each
other with fun, common objects and toys (such as cup, fish, boat, and car). The adult
presents an array of objects from which the child may select (in other words, a prefer-
ence assessment) and restricts item access after selection. The adult models an appropriate
play activity and spoken phrase (such as “Fish swims”) using the item. Echoic response
approximations are shaped by providing contingent item access with continued models
of the descriptive phrase while the child plays with the toy for twenty to thirty seconds.
The adult retrieves the item and presents a different exemplar by modeling a different
phrase for the same object (for example, “blue fish”) or presenting a new stimulus array
for selection. The adult inserts a five-second delayed echoic prompt while modeling the
action in order to allow the opportunity for responses without any direct vocal model. See
the “Play-Based Tact Training: Natural Language Paradigm” program for instructions
and create your own user-friendly data sheet that allows you to track targets and whether
responses occur before or after echoic prompts.
Responses during initial trials occur in response to a vocal model and the EO created
by restricted access (in other words, echoic and mand response features). As the procedure
progresses and the five-second delay begins to occur prior to the vocal model in response
to restricted access and the nonverbal stimulus, the mand and tact features of the child’s
response become more evident. The therapist is advised to begin incorporating neutral to
only mildly preferred stimuli over time with the goal of minimizing the EO control over
responding to ensure that the ultimate response form is a pure tact (and not an impure
mand or partial tact). The response requirement can be shifted over time in several ways
so that children learn to make longer responses (in other words, two- or three-word
Chapter 5 89
tterances) or incorporate specifically targeted grammatical features such as use of actions
u
or verb tacts (for example, “The ball rolls” or “The cheetah is running”). Conversational
interchanges during the child’s access to the item can also become the context for trials
of other verbal operants (for example, “The cheetah ran” [tact]; “Can you make it run?”
[receptive]; “Can an elephant jump?” [intraverbal]), with the child’s responses during
access resulting in praise and additional attention in the form of conversation.
Summary and Recommendations for Teaching

Mands and Tacts
In summary, follow several important recommendations for effective mand and tact train-
ing. Target mands as the first verbal operant (Sundberg & Michael, 2001) to allow the
child control over the environment and to create a history with communication responses
and potent reinforcing consequences. Target simple mands (in other words, mands for
preferred items, for termination of aversive stimuli, and for missing items) first, and delay
training mands for information until tact and beginning intraverbal repertoires exist. Train
mands in the presence of the relevant EO by either capturing or contriving EOs. Teach
tacts in response to specific questions and also as unprompted descriptions of meaningful
events and aspects of the environment. Choose initial training targets that are interesting
common items (for example, for mands, choose a favorite food or toy; for tacts, choose
commonly available items and people) in order to increase the likelihood that the target
stimuli will be encountered outside of specific training situations. Provide continuous or
near-continuous reinforcement during early trials with gradual schedule thinning. Finally,
use a mixture of structured teaching trials and naturally occurring learning opportunities
to facilitate rapid acquisition, make learning fun, and enhance generalization.

Captured Establishing Operations:
Incidental Teaching with Time Delay
PURPOSE: To capture naturally occurring establishing operations to teach functional

requests (in other words, mands).
NEEDED: Data sheet, keen observation of the environment, any relevant communica-
tion tools (such as PECS or VOCA).
PROCEDURE
Step 1: Allow the child to interact within his or her environment (for example, move
about or explore) and ensure that the environment includes many interesting things.
Step 2: Watch for indications that the child is interested in or needs something (for
example, looking at an item or reaching for it).
Step 3: Prevent access to the item and make eye contact, with an expectant look.
Step 4: Use the designated delay interval before prompting. Note that zero seconds
means immediate prompt.
a. If the child provides a response that is reinforceable (see below) within the delay
interval, immediately provide the requested item or event and praise (for example,
“Great job! Nice words!”). Note zero seconds means to reinforce responses after
the prompt.
b. If the child does not provide an appropriate response within the specified time
delay, use prompts as indicated below. Provide the prompt once, and if the child
still does not provide an appropriate response, move to step 5 without providing
access to the item or event.
Step 5: Return to observing the child in the natural environment and watching for
the next teaching opportunity.
REINFORCEABLE RESPONSES DATE
1st consonant sound (for example, “buh” for ball)

PROMPTS TIME DELAY DATE
ECH 5 seconds

Prompt Type
ECH: Echoic—say what the child should say.
TACT: Show a picture of the item or point to the item the child should request.
IV: Intraverbal—provide a partial or full carrier phrase (for example, “I need a ”).
Chapter 5 91
Capturing Establishing Operations:
Terminating Aversive Stimulation
PURPOSE: To capture naturally occurring establishing operations to teach functional

requests (in other words, mands) to remove or terminate unpleasant situations.
NEEDED: Data sheet, keen observation of the environment, any relevant communica-
tion tools (such as PECS or VOCA).
PROCEDURE
Step 1: Allow the child to interact within his or her environment (in other words,
explore).
Step 2: Watch for indications that the child is frustrated or seeking the removal of
an aversive stimulus (for example, termination of a demand, distress at a loud noise, or
agitation at someone being too close). Do not attempt to contrive aversive situations; if you
do so, you risk establishing yourself as an aversive stimulus.
Step 3: Make eye contact, with an expectant look.
interval, immediately remove the aversive stimulus (for example, say, “Let’s take a
break” or “Let’s go somewhere quiet”).
delay, use prompts as indicated below. Provide the prompt up to three times.
If the child still does not mand, either move to step 5 or remove the aversive
event if safety and comfort indicate.
This child does/does not (circle one) have a protocol for problem behavior that
takes precedence over this protocol if problem behavior occurs.
Step 5: Return to observing or interacting with the child in the natural environment
and watching for the next teaching opportunity.
1st consonant sound (for example, “buh” for break)

ECH 5 seconds

Prompt Type
TACT: Show a picture of or symbol indicating “break,” “stop,” “leave,” and so on.
IV: Intraverbal—provide a partial or full carrier phrase (for example, “I need a ”).

Contrived Establishing Operations:
Interrupted Chains
PURPOSE: To create a situation for the child to ask for a needed item, activity, or event
using a previously mastered chain.
NEEDED: Data sheet, items needed to complete the chain (see sample chains, next page).
PROCEDURE
Step 1: Ask the child to complete a previously mastered task requiring a series of
steps. If the child is unable to complete all steps, use chaining procedures to teach it.
Step 2: Remove an item that is critical to completing the chain (initially place it out
of reach but visible, later hide it altogether).
Step 3: When the child finishes the step prior to the targeted step (for example,
pouring cereal and milk into bowl prior to using the target item of spoon), look expec-
tantly at the child.
interval, immediately provide the requested item or event and praise (for example,
say, “Great job! Nice words!”). Note zero seconds means reinforce responses after
the prompt.
delay, use prompts as indicated below. Provide the prompt once, and, if the child
still does not provide an appropriate response, then reset the scene and wait for
the child to indicate need for the object again (for example, by pointing, whining,
looking for the item, or other action.)
1st consonant sound (for example, “sp” for spoon)

ECH 5 seconds

Prompt Type (indicate all appropriate and order of delivery above)

TACT: Show a picture of the item or point to the item the child should request.
IV: Intraverbal—provide a partial phrase (for example, “Where’s the ?” or “Give
me the ”).
Chapter 5 93
Interrupted Chain: Sample Chains
Making a bowl of cereal
Target: “Spoon”
1. Pick up box of cereal.
2. Pour cereal in bowl and put box down.
3. Pick up milk.
4. Pour milk on cereal and put milk down.
5. Pick up spoon. (Step is interrupted when therapist makes spoon unavailable
[visible in early stages, hidden after the child acquires the response when the
item is visible] and waits for child to say “spoon.”)
6. Eat cereal.
Making a sandwich
Target: “Knife”
1. Put bread on plate.
2. Open peanut butter.
3. Use knife to put peanut butter on bread. (Step is interrupted when therapist
makes knife unavailable [visible at first, hidden later] and waits for child to say
“knife.”)
4. Open jar of jelly.
5. Use knife to put jelly on bread.
6. Put knife down.
7. Put pieces of bread together to make sandwich.
Putting on your coat

Target: “Coat”
1. Go to closet where coat is kept.
2. Open closet doors.
3. Take coat off hanger. (Step is interrupted when therapist makes coat unavailable
[visible at first, hidden later] and waits for child to say “coat.”)
4. Put coat on

Putting together an object
Target: “Scissors,” “Glue”
1. Provide child with figures to cut out.
2. Provide child with scissors to cut out figure. (Step is interrupted when therapist
makes scissors unavailable [visible at first, hidden later] and waits for the child
to say “scissors.”)
3. Ask child to glue items together.
4. Provide child with glue. (Step is interrupted when therapist makes glue unavail-
able [visible at first, hidden later] and waits for child to say “glue.”)
5. Child’s picks up glue and squeezes glue on figure.
6. Child places figures together as asked.
Other:
1.
2.
3.
4.
5.
6.
Chapter 5 95
Contrived Establishing Operations: The
“What’s in the Bag?” Game
PURPOSE: To increase requests for information and to teach persistence in question

asking, and to introduce the program to the child as a game.
NEEDED: An opaque bag (for example, a brown grocery bag or backpack), and inter-
esting toys and objects (such as action figures, a baseball, or a favorite video) or favorite
food items.
PROCEDURE
Part 1: Asking
Step 1: Have the child close his or her eyes or place a hand over them.
Step 2: Hide an interesting object in the bag and say, “Open your eyes.”
Step 3: Hold the bag in front of the child, shake it for three to five seconds, and then
look inside with an interested expression.
a. If the child reaches for the bag, move it out of reach and shake your head as if
saying no.
b. If the child asks a question such as “What’s that?” “What’s in the bag?” or “What
is it?” then:
Praise (e.g., “Good asking”).
Answer the question (e.g., “It’s an apple”).
Give the child the item.
c. If the child does not respond or makes a different statement (such as “Give me the
bag”), prompt him or her, saying, “Ask me a question.” If the child asks a relevant
question, go to step b above. If the child makes no response or an inappropriate
response, prompt with “Say, ‘What’s in the bag?’” and repeat up to two times. If
the child repeats the question, go to step b above. If the child still gives no correct
response, open the bag, take out the item, and label the item, but don’t give it to
the child.
Step 4: Say, “Let’s try again,” and repeat steps 1 through 3 with a new item.

Part 2: Persistence
Step 1: Have the child close his or her eyes and place a hand over them.
Step 2: Hide an interesting object in the bag and say, “Open your eyes.”
Step 3: Hold up the bag, shake it for three to five seconds, and then look inside with
an interested expression.
a. If the child reaches for the bag, move it out of reach and shake your head as if
saying no.
b. When the child asks a question (such as “What’s in the bag?”) about the object,
say, “You’ll have to guess. Ask me another question.”
c. If the child asks a question such as “What color is it?” or “Is it blue?” then praise
the child, answer the question, and give the child the item.
d. If the child does not respond or repeats the same question, prompt the child
with “Say, ‘What color, shape, or size is it?’ or ‘Is it a food or animal?’” If the
child repeats the question, follow step b above. If the child still gives no correct
response, repeat the prompt up to two times. If the child continues to not give
a correct response, open the bag, take out the item, and say, “It’s yellow—it’s a
Pokémon!” but do not give it to the child.
Step 4: Say, “Let’s try again” and repeat steps 1 through 3 with a new item.
Chapter 5 97
Tact Training: Names, Features,
and Actions
PURPOSE: To teach appropriate tacts such as names of objects, features (such as, blue
or round), and descriptions of actions (for example, it’s racing, he is crying) of familiar
objects.
NEEDED: A variety of objects, or photographs or pictures of objects, that the child can
label.
PROCEDURE
Work on only the types of tacts that are circled on the Tact Training Data Sheet.
Step 1: Have the learner sit at a desk or table. Present an item or picture.
Step 2: Teach names of items before features or actions and require the child to emit
a response that closely resembles the target (and record the exact response) before provid-
ing reinforcers.
Step 3: Present a label, feature, or action class query about the item.
a. Name examples: “What is this?” “What do you call this?”
b. Feature examples: “What color is it?” “What shape is it?”
c. Action examples: “What is it (or he or she) doing?”
Step 4: Provide consequences and prompts.
a. For correct responses, provide praise and tickles or other tangible reinforcers (only
if needed and fade as soon as you can).
b. For incorrect responses or no response, repeat the question and provide an echoic
prompt (e.g., “That’s an apple”) after the designated delay.
Praise correct echoic responses.
If still incorrect, repeat the echoic a final time and begin a different
trial.
Step 5: Score trial type, stimulus, prompts, and responses on the Tact Training Data
Sheet.
Step 6: Rotate trial types and stimuli. For example, ball and action (“What is that ball
doing?”); ball and feature (“What color is it?”); candy and feature (“What shape is it?” or
“What flavor does it have?”)
We are currently working on the following tacts:

Names Features Actions Date
Ball, boat, mom, Shape (round, square,
dad, horse, car, apple, oval, triangle), color
airplane (red, blue, green,
yellow)
Chapter 5 99
Tact Training: Prepositions
PURPOSE: To teach appropriate tacts for prepositions, such as under, above, and beside.
Receptive trials are interspersed.
NEEDED: A variety of small, interesting objects (such as a box and a toy that can be put
on the box, next to the box, and so on) or pictures that illustrate relational prepositions.
PROCEDURE
Step 1: Have the learner sit at a desk or table. Provide a situation in which the child
can complete two tasks requiring the use of a preposition. (For instance, place two boxes
on the table, one right side up and one upside down. Then ask the child to put an item in
the box; on the next trial, ask the child to put something on top of the box.)
Step 2: Give the child an item and require the child to provide a receptive response
using a preposition. For example, say, “Put the block in the box.”
Step 3: If the child responds correctly, provide praise. If the child does not, provide
the prompt again and physically guide the child to complete the task.
Step 4: Next, require the child to complete the other task (for example, put the block
on the box) identified as a target at the beginning of the exercise.
Step 5: When the child is reliably responding in the receptive format, place the item
in one of the locations (on or in the box) and ask, “Where is the toy?”
a. If the child responds correctly with an appropriate response (for example, “in
box”), praise the child.
b. If the child does not respond within the designated time delay, provide an appro-
priate model (for instance, “It is in the box”), and provide the child with an
opportunity to imitate this response. Repeat the prompt up to three times before
moving on to the next trial.
TARGET RESPONSES
For example, “in the box,” “on the box”

ECH 5 seconds


Tact Training: Sensory Experiences
PURPOSE: To teach a child to label sensory experiences, such as things that are smelled,
seen, or heard.
NEEDED: Various items that provide sensory experiences. Examples include certain
things that have identifiable smells (such as orange or chocolate) or identifiable sounds
(for example, an audio recording of animal sounds or of vehicle sounds). Teaching can
also occur in the natural environment (for example, “What do you hear?” “What do you
see?” “What do you smell?”).
PROCEDURE
Step 1: Have the learner sit at a desk or table. For feeling, smelling, and hearing tasks,
either blindfold the child or turn off the lights. This is not needed for the seeing task
because the child will need to experience the stimulus visually.
Step 2: Give the child an item and, if needed, prompt the child to interact with the
stimulus (for example, say, “Touch it!” or prompt the child to touch it physically).
Step 3: Ask the child a question appropriate to the stimulus (such as “What do you
see?” or “What does it feel like?”). If the child responds correctly, provide praise. If the
child does not respond correctly or does not respond within the designated time delay,
provide the appropriate prompt up to three times. If the child responds correctly at any
time, provide praise.
Step 4: Do another trial with a different stimulus. It may be beneficial to intersperse
seeing, hearing, smelling, and feeling trials so that the child does not begin to respond
only to the aspect of the stimulus that has been recently targeted (such as all “feeling”
responses). Note that some stimuli have multiple relevant sensory tacts that should be
queried back-to-back (such as “What does it feel like?” and “What does it smell like?”).
TARGET RESPONSES

ECH 5 seconds

Chapter 5 101

Play-Based Tact Training: Natural
Language Paradigm
PURPOSE: To increase conversational descriptive tact responses.
NEEDED: A wide variety of common fun toys.
PROCEDURE
Step 1: Place three toys in front of the child and ask him or her to choose one (for
example, pointing, placing a hand on the item, or directing eye gaze at the object for a
sustained period of time).
Step 2: Remove the nonchosen items and keep the chosen item (you may keep two of
the items and play with them interactively).
Step 3: Repeatedly model an appropriate action either with or without a simultaneous
vocal model of a phrase:
a. Model an action and a phrase simultaneously during initial training trials,
when you want to change the target phrase, and when the response criterion in
changed.
1. Provide a simultaneous model of a play action and a descriptive tact as a vocal

model. If the child echoes your vocal model or makes the target approxi-
mation, provide access to the item and praise and record the exact response
scored as imitative.
2. If the child does not echo, try two or three more prompts and then present a
new array to find a more highly preferred stimulus.
b. Model with a delayed vocal model once the frame for NLP has been established
by waiting approximately 5 seconds before providing a verbal model of an appro-
priate tact (for example, “bear” or the sign for bear). This time delay allows the
child an opportunity to emit a tact spontaneously.
1. If the child says any word related to the item or action or a target approxi-
mation during the delay, provide access to the item and praise. Record the
child’s response or response approximation exactly and score as “spontaneous.”
Proceed to step 4.
2. If no response occurs during the time delay, provide a simultaneous corre-

sponding tact as a vocal model (just like in step a above).
a. Provide a simultaneous model of a play action and a descriptive tact as
a vocal model. If the child echoes your vocal model or makes the target
approximation, provide access to the item and praise, and record the exact
response scored as imitative.
b. If the child does not echo, try two or three more prompts and then return to
step 1 and present a new array to find a more highly preferred stimulus.

Step 4: As the child plays with the toy, repeat the modeled tact a few times or model
a new, similar tact.
Step 5: Say, “my turn,” remove the item, and do a new trial with that item.
TARGET RESPONSE CRITERION DATE
Two-word phrase with verb (for example, “Ball rolls”)

Chapter 5 103

Mand Training Data Sheet
Therapist Targeted Child Correct Time
104
Sample: Date Time Program Prompts
Initials Item/Event Response Response Delay
Captured EO,
ECH
verbal model, IT CON IC
C.D. 06/26/07 2:15pm “Drink” Dr- Yes/No Tact 2s
teaching “drink” Visible Hidden
IV
Sample: ECH
Contived EO, IT CON IC
C.D. 08/25/07 10:30am “Coat” Co- Yes/No Tact 2s
verbal model, teaching Visible Hidden
IV
“coat” when it is not
visible (advanced mand) ECH
IT CON IC
Yes/No Tact
Visible Hidden

IV
ECH
Program Type IT CON IC
Yes/No Tact
IT = incidental Visible Hidden
teaching IV
CON = contrived EO ECH
IT CON IC
IC = interrupted Yes/No Tact
Visible Hidden
chain IV
ECH
IT CON IC
Prompt Type Yes/No Tact
Visible Hidden
ECH = echoic prompt IV
Tact = showing child ECH
IT CON IC
a picture of needed Yes/No Tact
item Visible Hidden
IV
IV = providing a
ECH
partial statement IT CON IC
Yes/No Tact
(“I need a ”) Visible Hidden
IV
Tact Training Data Sheet
Therapist Targeted Child Correct Time
Date Time Program Prompts
Initials Item/Event Response Response Delay
Sample: FFC NLP ECH
C.D. 06/26/07 2:15pm “Drink” Dr- Yes/No 2s
Prep Sen IV
FFC NLP ECH
Yes/No
Prep Sen IV
FFC NLP ECH
Yes/No
Prep Sen IV
FFC NLP ECH
Program Type Yes/No
Prep Sen IV
FFC = feature,
function, class FFC NLP ECH
NLP = natural Yes/No
Prep Sen IV
language paradigm
Prep = prepositions FFC NLP ECH
Yes/No
Sen = sensory tacts Prep Sen IV
FFC NLP ECH
Yes/No
Prep Sen IV
Prompt Type
ECH = echoic FFC NLP ECH
prompt Yes/No
Prep Sen IV
IV = providing a
partial statement FFC NLP ECH
Yes/No
(“It’s a ”) Prep Sen IV
Chapter 5
FFC NLP ECH
Yes/No
Prep Sen IV
105
References
Arntzen, E., & Almas, I. K. (2002). Effects of mand-tact versus tact-only training on the
acquisition of tacts. Journal of Applied Behavior Analysis, 35, 419–422.
Barbera, M. L., & Kubina, R. M., Jr. (2005). Using transfer procedures to teach tacts to
a child with autism. Analysis of Verbal Behavior, 21, 155–161.
Bartman, S., & Freeman, N. (2003). Teaching language to a two-year-old with autism.
Journal on Developmental Disabilities, 10, 47–53.
Baum, W. (2005). Understanding behaviorism. Malden, MA: Blackwell Publishing.
Bondy, A., & Frost, L. (2002). A picture’s worth: PECS and other visual communication
strategies in autism. Bethesda, MD: Woodbine House.
Bourret, J., Vollmer, T. R., & Rapp, J. T. (2004). Evaluation of a vocal mand assess-
ment and vocal mand training procedures. Journal of Applied Behavior Analysis 37,
129–144.
Braam, S. J., & Sundberg, M. L. (1991). The effects of specific versus nonspecific rein-
forcement on verbal behavior. Analysis of Verbal Behavior, 9, 19–28.
Chambers, M., & Rehfeldt, R. A. (2003). Assessing the acquisition and generalization
of two mand forms with adults with severe developmental disabilities. Research in
Charlop, M. H., & Haymes, L. K. (1994). Speech and language acquisition and inter-
vention: Behavioral approaches. In J. L. Matson (Ed.), Autism in children and adults:
etiology, assessment, and intervention. Pacific Grove, CA: Brooks/Cole Publishing.
Charlop, M. H., Schreibman, L., & Thibodeau, M. G. (1985). Increasing spontaneous
verbal responding in autistic children using a time delay procedure. Journal of Applied
Charlop-Christy, M. H., Carpenter, M., Le, L., LeBlanc, L. A., & Kellet, K. (2002).
Using the picture exchange communication system (PECS) with children with
autism: Assessment of PECS acquisition, speech, social-communicative behavior, and
problem behavior. Journal of Applied Behavior Analysis, 35, 213–231.
Charlop-Christy, M. H., LeBlanc, L. A., & Carpenter, M. H. (1999). Naturalistic teach-
ing strategies (NATS) to teach speech to children with autism: Historical perspective,
development, and current practice. California School Psychologist, 4, 30–46.
Duker, P. C., Kraaykamp, M., & Visser, E. (1994). A stimulus control procedure to
increase requesting with individuals who are severely/profoundly intellectually dis-
abled. Journal of Intellectual Disability Research, 38, 177–186.
Durand, V. M., & Merges, E. (2001). Functional communication training: A contempo-
rary behavior analytic intervention for problem behaviors. Focus on Autism and Other

Fenske, E. C., Krantz, P. J., & McClannahan, L. E. (2001). Incidental teaching: A not-
discrete-trial teaching procedure. In C. Maurice & R. M. Foxx (Eds.), Making a dif-
ference: Behavioral intervention for autism. Austin, TX: PRO-ED.
Gillett, J. N., & LeBlanc, L. A. (2007). Parent-implemented natural language paradigm
to increase language and play in children with autism. Research in Autism Spectrum
Disorders, 3, 247–255.
Green, G. (2001). Behavior analytic instruction for learners with autism: Advances in
stimulus control technology. Focus on Autism and Other Developmental Disabilities,
16, 72–85.
Harris, S. L. (1975). Teaching language to nonverbal children with emphasis on problems
of generalization. Psychological Bulletin, 82, 565–580.
Hart, B., & Risley, T. R. (1980). In vivo language intervention: Unanticipated general
effects. Journal of Applied Behavior Analysis, 13, 407–432.
Karmali, I., Greer, R. D., Nuzzolo-Gomez, R., Ross, D. E., & Rivera-Valdez, C. (2005).
Reducing palilalia by presenting tact corrections to young children with autism.
Analysis of Verbal Behavior, 21, 145–153.
Koegel, R. L., O’Dell, M. C., & Koegel, L. K. (1987). A natural language paradigm for
teaching non-verbal autistic children. Journal of Autism and Developmental Disorders,
17, 187–199.
Laski, K. E., Charlop, M. H., & Schreibman, L. (1988). Training parents to use the
natural language paradigm to increase their autistic children’s speech. Journal of
LeBlanc, L. A., Esch, J., Sidener, T., & Firth, A. (2006). Behavioral language interven-
tions for children with autism: Comparing applied verbal behavior and naturalistic
teaching approaches. Analysis of Verbal Behavior, 22, 49–60.
McGee, G. G., Morrier, M. J., & Daly, T. (1999). An incidental teaching approach to
early intervention for toddlers with autism. Journal of the Association for Persons with
Severe Handicaps, 24, 133–146.
Michael, J. (1982). Distinguishing between discriminative and motivational functions of
stimuli. Journal of the Experimental Analysis of Behavior, 37, 149–155.
Michael, J. (1988). Establishing operations and the mand. Analysis of Verbal Behavior, 6,
3–9.
Mirenda, P. (2003). Toward a functional augmentative and alternative communication
for students with autism: Manual signs, graphic symbols, and voice output commu-
nication aids. Language, Speech, and Hearing Services in Schools, 43, 203–216.
Nuzzolo-Gomez, R., & Greer, R. D. (2004). Emergence of untaught mands or tacts of
novel adjective-object pairs as a function of instructional history. Analysis of Verbal
Behavior, 20, 63–76.
Partington, J. W., & Sundberg, M. L. (1998). The assessment of basic language and learning
skills. Pleasant Hills, CA: Behavior Analysts, Inc.
Chapter 5 107

Partington, J. W., Sundberg, M. L., Newhouse, L., & Spengler, S. M. (1994). Overcoming
an autistic child’s failure to acquire a tact repertoire. Journal of Applied Behavior
Analysis, 27, 733–734.
Shafer, E. (1994). A review of intervention to teach a mand repertoire. Analysis of Verbal
Sigafoos, J., Kerr, M., Roberts, D., & Couzens, D. (1994). Increasing opportunities for
requesting in classrooms serving children with developmental disabilities. Journal of
Autism and Developmental Disabilities, 24, 631–645.
Sigafoos, J., Reichle, J., Doss, S., Hall, K., & Pettitt, L. (1990). “Spontaneous” trans-
fer of stimulus control from tact to mand contingencies. Research in Developmental
Disabilities, 11, 165–176.
Smith, T. (1999). Outcome of early intervention for children with autism. Clinical
Psychology: Science and Practice, 6, 33–49.
Sundberg, M. L., & Hale, L. M. (2007). The VB-MAPP: An overview and field test data
from Oregon and California. Symposium conducted at the annual convention of the
Association for Behavior Analysis, San Diego, CA.
Sundberg, M. L., Loeb, M., Hale, L., & Eigenheer, P. (2002). Contriving establishing
operations to teach mands for information. Analysis of Verbal Behavior, 18, 15–29.
Sundberg, M. L., & Michael, J. (2001). The benefits of Skinner’s analysis of verbal behav-
ior for children with autism. Behavior Modification, 25, 692–724.
Sundberg, M. L., & Partington, J. W. (1998). Teaching language to children with autism
and other developmental disabilities. Pleasant Hill, CA: Behavior Analysts, Inc.
Szatmari, P., Bryson, S. E., Boyle, M. H., Streiner, D. L., & Duku, E. (2003). Predictors
of outcome among high functioning children with autism and Asperger syndrome.
Journal of Child Psychology and Psychiatry, 44, 520–528.
Tincani, M. (2004). Comparing the picture exchange communication system and sign
language training for children with autism. Focus on Autism and Other Developmental
Twyman, J. S. (1996). The functional independence of impure mands and tacts of abstract
stimulus properties. Analysis of Verbal Behavior, 13, 1–19.
Wraikat, R., Sundberg, C. T., & Michael, J. (1991). Topography-based and selection-
based verbal behavior: A further comparison. Analysis of Verbal Behavior, 9, 1–17.

Part 2
Speaking with Meaning and

Listening with Understanding
Chapter 6
Nonrelational and Relational

Instructional Control
Jonathan Tarbox, Center for Autism & Related Disorders;

Rachel S. F. Tarbox, Chicago School of
Professional Psychology at Los Angeles; and
Denis O’Hora, National University of Ireland, Galway
Direct instructional control is among the simplest forms of environmental control of

human behavior. Establishing direct instructional control of nonvocal behavior is often
one of the first priorities in establishing verbal repertoires in individuals lacking them.
Deficits in “receptive” language are characteristic of autism and many other develop-
mental disorders (Gillum & Camarata, 2004). This chapter will define various sorts of
receptive language skills, describe various procedures that are commonly used to establish
them, and conclude by relating the establishment of direct instructional control to the
larger endeavor of establishing a generalized relational repertoire. We begin by proposing
a crude but (we hope) useful distinction between two categories of direct instructional
control that are commonly the targets of rudimentary language intervention for individu-
als with developmental disabilities: nonrelational instruction following, and nonrelational
matching to sample.
The categories of behavior upon which this chapter focuses are usually referred to as
“receptive” among a number of clinical professions, but for behavior analysts the term has
little to recommend it. The origin of the term “receptive” is cognitive, in the sense that
an individual is assumed to take language into the brain, process it, and then act upon it.
From a behavioral perspective, the individual is assumed to behave in direct interaction
with environmental stimuli; the stimuli are not assumed to be “received” by the individ-
ual for processing. Perhaps more important, the term “receptive” does not appear to refer
to any actual environmental features other than the presence of an auditory antecedent
stimulus. In other words, in most cases of nonrelational instructional control, the behav-
ior in question is under the antecedent control of the vocal stimuli produced by someone
else’s verbal behavior. For example, a person hears someone else say, “Sit down,” and
then he or she sits down. Such behavior is sometimes referred to as “manded” behavior
or “mand compliance,” but these terms refer to an analysis of the behavior of someone
else—that is, the speaker to whose verbal behavior the person of interest responds. It
would seem inconsistent to name the behavior of the person of interest according to the
function of someone else’s behavior in the case of instructional control, whereas virtually
all other behavior is referred to in terms of its own function (for example, mand, tact,
and so on). In short, a satisfactory or consistently agreed upon term for the behavior of
central concern in this chapter does not seem to exist. We therefore will use the terms
“nonrelational instruction following” and “nonrelational matching to sample” as defined
and exemplified below.
Nonrelational Instruction Following

Nonrelational instruction following typically involves straightforward, successive dis-
criminations. Specifically, it involves a single antecedent stimulus that occasions a single
response, which is typically followed by some form of generalized reinforcer. The dis-
criminative stimulus for such behavior is typically vocal. For example, the discrimina-
tive stimulus may consist of the command “Turn off the light,” where the nonrelational
instruction-following behavior would consist of simply turning off the light. A history of
direct reinforcement is responsible for establishing such behavior.
From a conceptual standpoint, there is nothing peculiar about nonrelational instruc-
tion following that necessitates a vocal discriminative stimulus; it is simply a matter of
convention in our verbal community given the ubiquity of vocal verbal behavior. In
hearing-impaired verbal communities, the discriminative stimuli involved in nonrela-
tional instruction following would likely be visual in modality—either the presentation
of manual signs or lip movements emitted by another person. In language training for
individuals with developmental disabilities, some have advocated for the use of nonvocal
discriminative stimuli, such as those used in picture communication systems (Bondy &
Frost, 1994) and sign language (Sundberg, 1992). However, the majority of instructions
in our culture that occur in interpersonal interactions are vocal in modality. For this
reason, we will limit our discussion to vocal instructions. Below we describe several types
of nonrelational instruction following that are commonly used in educational programs
for individuals with autism and other developmental disabilities.
One-Step Instructions
One-step instructions typically involve the teacher’s simple, clear instruction and rein-
forcement of compliance with it (Craighead, O’Leary, & Allen, 1973). Other common
clinical terms for such instruction include receptive commands and compliance training.
Common examples during discrete trial instruction include relatively simple responses,
such as “Touch table,” “Clap hands,” “Arms up,” “Stomp feet,” “Wave bye-bye,” “Stand

up,” “Sit down,” “Turn around,” and the like (Maurice, Green, & Luce, 1996). Instructions
sometimes include “acting out” more abstract actions or emotions, such as pretending to
cry in response to the instruction “Pretend you are sad,” or running in place in response
to the instruction “Show me running.”
Multiple-Step Instructions
After a child masters a variety of one-step instructions, behavioral intervention pro-
grams often proceed to establishing two-step instructions. In such cases, the teacher
typically states two instructions in a row, such as “Clap hands and stomp feet,” and rein-
forces compliance with the two-step instruction (Maurice et al., 1996). After a variety of
two-step instructions are under reliable stimulus control, clinicians sometimes proceed to
establishing three-step instructions in a similar manner.
Body Parts
Receptive body parts typically refers to a teaching situation in which the teacher
states the name of a body part and delivers reinforcement when the student touches that
body part—for example, “Touch nose” (Lovaas, 1981). The instruction may be stated as
simply the name of the body part or it may include some other word or phrase, such as
“Touch ,” “Show me ,” or “Where’s ”?
Increasing Complexity of Instructions

Other features of the instructional environment can be manipulated in order to
increase the complexity and sophistication of the nonrelational instruction-following
behavior targeted. First, distance between the learner and the relevant environmental
object is sometimes introduced. For example, instructions can be given requiring the
student to respond to stimuli that are across the room (such as “Open the door”). Second,
instructions can be given requiring the student to respond to a stimulus that is not visu-
ally present. For example, when the teacher and student are standing in the kitchen, the
instruction could be given to “touch the couch” (which is located in the living room).
One could presumably increase the complexity of instruction following and increase the
number of instructions taught ad infinitum; there is currently no research of which we
are aware that has determined the optimal number of instructions to teach. However, see
our discussion immediately below on the broader purpose of teaching these instructions
and consider constructing your treatment plans accordingly.
Relevance of Nonrelational Instruction Following

It may not be immediately obvious why one might spend valuable intervention time
teaching children with autism to follow instructions in a manner as rote as that described
Chapter 6 113

above. In other words, it may not be readily apparent how these skills are functional
for clients in their everyday lives. As is emphasized throughout this book, the goal of
interventions based upon derived stimulus relations is not the establishment of rote per-
formances; it is the establishment of generalized, flexible operant classes of relational
behavior. Teaching nonrelational instructional control (both instruction following and
matching to sample) is a first step toward teaching broader generalized operant classes,
which include responding to the stimuli involved in the training described above (for
instance, the vocal names of actions, body parts, and the like). We describe the relevance
of this more thoroughly below, in the section titled “Relevance to Generalized Relational
Repertoires.”
We are not suggesting that one teach children with autism to memorize long lists of
commands simply for the sake of mastering them. When selecting which instructions to
teach, consider which ones will be included in larger repertoires of relational behavior in
the future. For example, teaching a child to follow the instruction “Clap hands” is not,
in itself, going to be functional for that child in his or her everyday life. However, if the
goal of language intervention for that child is the establishment of a full verbal repertoire,
then one will sometime soon also teach the child to say “clapping” when he or she sees
someone clapping (a tact), to say “Clap for me” when he or she does something well and
wants attention for it (a mand), and eventually to say “clap” when asked, “What are you
supposed to do when your teammate scores a goal?” (intraverbal).
In cases of more severe cognitive impairment, where it is not reasonable to expect to
establish a full verbal repertoire, one might do well to teach only instructions that will
be useful in the everyday life of the client (“Go get your shoes,” “Wash your hands,” and
“Come here,” for example). Unfortunately, there is currently no science-based criterion
for making the distinction between clients who may be able to achieve fully elaborated
relational verbal repertoires and those who will not. The general consensus within the
field of behavior analysis is that young children who receive intensive behavioral interven-
tion for at least a year or two may have a chance at reaching age-appropriate language
development. Therefore, the current approach appears to be to aim for this goal in young
children with autism, for whom intensive behavioral intervention can be provided.
Recommendations for Teaching Nonrelational

Instruction Following
1. Begin with one-step instructions. Once a variety of instructions are mastered,
move onto multiple-step instructions.
2. Provide the same instructions on each trial.
3. State the instructed behavior in the delivery of verbal praise for correct responses
(for example, “Nice job sitting down!”).
4. Introduce training opportunities for the student to follow instructions by

responding to stimuli at increasing distances from the teacher.
5. Ensure that the instructional repertoires established are relevant for the every-
day life of the student.

Nonrelational Matching to Sample
Matching-to-sample (MTS) arrangements are common in educational settings for indi-
viduals with developmental disabilities. MTS involves the presentation of a single stimu-
lus, often referred to as the sample stimulus, followed by the presentation of two or more
stimuli from which the individual can choose, often referred to as the comparison stimuli,
followed by the opportunity for the individual to respond to one of the comparison
stimuli by selecting it. Sample stimuli can be visual (in other words, a picture or object)
or auditory (a dictated name of a picture or object). The correct comparison stimulus on
any given trial is conditional upon the particular sample stimulus presented on that trial,
which is why the term conditional discrimination learning is used to describe the behavior
process involved in matching to sample (Sidman & Tailby, 1982; see also chapter 8 of
this volume).
The type of MTS procedure we will be describing here is one in which the sample
stimulus consists of the instructor vocally stating the name of a stimulus and present-
ing comparison stimuli consisting of the stimulus being named, as well as one or more
other stimuli. For example, an instructor may say, “car,” and present comparison stimulus
pictures or objects consisting of a car, an airplane, and a bicycle; the correct response on
the part of the student would be to indicate the picture of the car in some way, such as
pointing to it or handing it to the instructor.
Common Comparison Stimuli

A wide variety of stimuli are included in nonrelational MTS training in behavioral
approaches to educating individuals with autism and other developmental disabilities.
Comparison stimuli are typically either pictures (often referred to as 2-D receptive ID) or
objects (3-D). Pictures often depict familiar objects, familiar people, people engaging in
actions (such as running or eating), rooms in the child’s house, emotional facial expres-
sions (for example, a picture of someone frowning), professions (for example, a picture of
a firefighter), colors, shapes, letters, numbers, and printed words, to name a few. Objects
included in such training vary widely, presumably only being limited by what can fit on a
table and what one desires the student to learn. Existing curricula for discrete trial train-
ing for children with autism include a variety of recommended target stimuli (Lovaas,
1981; Maurice et al., 1996; Leaf & McEachin, 1999).
Variations on Sample Stimulus Presentation

Several variations on the presentation of the sample stimulus during nonrelational
MTS training are commonly used. Sometimes only the stimulus itself is presented once,
as when the teacher simply says, “car.” Variations involve including other words in the
presentation, such as “Show me car,” “Where’s car?” “Touch car,” “Point to car,” and
“Give me car.” Little or no research has systematically compared these different options.
However, it seems that adding irrelevant stimuli to the sample stimulus presentation
would only enhance the difficulty of the task. Therefore, it may be wise to omit phrases
Chapter 6 115

such as “point to,” and “give me” from the sample stimulus, unless one is explicitly trying
to teach the student to point or give. For example, Green (2001) recommends presenting
only the sample stimulus, stated clearly.
Prompting Procedures
A variety of procedures are used for prompting correct responses in both of the tasks
described in this chapter. All are relatively common to applied behavior analysis (Martin
& Pear, 2003) and will be described briefly here.
Physical Guidance
Physical guidance is also commonly referred to as manual guidance and hand-over-
hand prompting. With physical guidance, the teacher typically places his or her hands on
the student and guides the student’s hands or other body parts in some manner (Cooper,
Heron, & Heward, 2006). In the case of nonrelational MTS training, a common physical
prompt involves the teacher holding the student’s hand in his or her own and guiding the
student’s hand to make the correct response. For example, the teacher might present the
conditional stimulus “Touch fork” while guiding the student’s hand to a fork and touch-
ing the student’s hand or fingers to the fork. In an example of a physical guidance prompt
in a nonrelational instruction-following teaching situation, the teacher might say, “Turn
on the sink,” and take the student’s hand in his or her own, place it on top of the faucet,
and guide the student’s hand through the motion of turning on the faucet. There is no
nonarbitrary rule regarding the amount of force that should be used in a physical guidance
prompt, but most agree that the smallest amount of force necessary to produce a correct
response should be used. Completely guiding a correct response is sometimes referred to
as full physical guidance, whereas a lighter amount of force, which does not necessarily
entirely control a correct response, is often referred to as partial physical guidance.
Modeling
With modeling prompts, the teacher demonstrates the correct behavior to the student.
For example, Roark, Collins, Hemmeter, and Kleinert (2002) used a model prompt to
teach four young adults with multiple disabilities to identify food items via a nonrela-
tional MTS procedure. In this study, they presented instructions such as “Point to bread,”
and they implemented a modeling prompt by having the instructor point to the bread. In
an example of using a modeling prompt while teaching nonrelational instruction follow-
ing, a teacher would present the instruction “Touch your head” and touch his or her own
head (the modeling prompt).
Gesture
Gestural prompts are aptly named, in that they entail just what you might expect—
providing a gesture, typically in the direction of a stimulus to which a correct response
should occur. For example, in the case of nonrelational MTS, a teacher might present the

conditional stimulus “running” and then gesture in the direction of a picture of someone
running, and not in the direction of the other comparison stimuli present. In the case
of nonrelational instruction following, a teacher might present the instruction “Close the
door” and then give the gestural prompt of pointing toward the door.
Proximity
Proximity prompts are sometimes also referred to as position prompts. A proximity
prompt involves manipulating the position of a stimulus so that a correct response is more
likely. An example of the use of a proximity prompt in nonrelational MTS training is to
place the correct comparison stimulus closer to the student than the incorrect comparison
stimuli. An example of a proximity prompt in nonrelational instruction following could
be to have the student stand next to the door when the clinician gives the instruction
“Open the door,” rather than giving the instruction when the student is sitting at his or
her table at the other side of the room.
Intrastimulus
Intrastimulus prompts are sometimes referred to as within-stimulus prompts. These
are prompts in which some aspect of a stimulus to which the student is to respond is
altered or exaggerated in some way so as to make a correct response more probable. In
the case of nonrelational MTS, an intrastimulus prompt could consist of making the
correct comparison stimulus larger than the incorrect comparison stimuli when pic-
tures are presented as comparison stimuli. An example of an intrastimulus prompt in
nonrelational instruction-following training could be to exaggerate the relevant aspect
of a vocal instruction, such as saying “Touch your head ” on some trials and “Touch
your nose” on other trials.
Some have argued that intrastimulus prompts may be particularly favorable because
they do not require the student to attend to an additional stimulus (Schreibman, 1975).
For example, in the nonrelational MTS example above, the fact that the correct com-
parison stimulus is larger than the incorrect ones does not require the student to attend
to anything other than the correct stimulus, whereas a point prompt used instead of
the intrastimulus prompt would require the student to attend to both the point and the
correct comparison. Therefore, it is possible that transfer of stimulus control from the
prompt to the intended discriminative stimulus may occur more readily with intrastimu-
lus prompts than with some others, particularly in the case of individuals with autism
who are likely to display stimulus overselectivity (in which responding comes under
control of only one, often irrelevant, aspect of a stimulus). Thus, the transfer of stimulus
control from a prompt to an intended discriminative stimulus may be impeded if the
student is overly selective in his or her attention and therefore attends only to the prompt
and not to the intended discriminative stimulus. This problem may be avoided by embed-
ding the prompt in the intended discriminative stimulus from the outset. For example,
Schreibman (1975) compared the use of intrastimulus prompts (in which the relevant
visual dimension of a stimulus was exaggerated) to the use of a point prompt in teaching
discriminations to children with autism. Schreibman found that intrastimulus prompting
was significantly more effective and could be faded out more rapidly.
Chapter 6 117

Exclusion
Exclusion is the technical term for prompting via the process of elimination. That is,
in nonrelational MTS training, an exclusion prompt involves including only one com-
parison stimulus that is not already known (Clark & Green, 2004). This procedure may
result in a larger probability that the student will respond to the unknown (and, in this
case, the correct) comparison stimulus, because selecting the other comparison stimuli
has already been brought under the control of other conditional stimuli in the past. Carr
(2003) utilized exclusion to teach nonrelational MTS discriminations to children with
autism and elaborated on it by showing that children who did not demonstrate the ability
to exclude could then be trained to do so via multiple-exemplar training.
Prompt-Fading Procedures
Regardless of which prompting procedures are used, prompts must be faded once
the desired behavior has occurred. That is, stimulus control of correct responding must
be transferred from the prompt to the intended discriminative stimulus (Cooper et al.,
2006). Prompt-fading procedures used in establishing nonrelational instructional control
are common to applied behavior analysis and will be described briefly below.
Intrusiveness
No consensus exists regarding the definition of intrusiveness. However, physical guid-
ance is generally considered among the most intrusive, whereas vocal or intrastimulus
prompts tend to be less intrusive. A prompt may be faded along the lines of intrusiveness.
For example, moving from a vocal to a gestural to a physical prompt constitutes fading from
a less-intrusive to a more-intrusive prompt. Similarly, a prompt consisting of full physical
guidance may be gradually faded until the teacher is scarcely touching the student.
Proximity
The proximity of a prompt to the learner or to a stimulus may be systematically faded
up or down. For example, a teacher may fade the proximity of a point from very near the
correct comparison stimulus to farther away from it, eventually eliminating it altogether
(Schreibman, 1975).
Delay
Rather than, or in addition to, fading a prompt along the lines of intrusiveness or
proximity, the delay from the discriminative stimulus to the onset of the prompt can be
systematically increased or decreased. For example, in nonrelational MTS, a teacher might
present three comparison stimuli, state a vocal instruction (such as “happy”), and point to
the correct comparison stimulus. Over successive trials with correct responding, the delay
between the instruction “happy” and the prompt of pointing is gradually increased, given
the student’s continued correct responding (Touchette & Howard, 1984).

Least to Most vs. Most to Least
Regardless of the dimension along which prompts are faded, one may choose to
fade prompts out or in—that is, from most to least (MTL) or least to most (LTM).
MTL prompt fading is also sometimes referred to as errorless learning, because it tends to
produce fewer errors, particularly if the clinician waits to fade out a prompt until correct
responding is stable at the current level of a prompt. This method of prompt fading
involves much prompting at first and may therefore involve more prompting than a par-
ticular client needs, particularly if one’s fading criteria is overly conservative. However,
MTL prompt fading is highly likely to produce correct responding rapidly and to produce
fewer errors overall.
MTL prompt fading can be implemented either across successive sessions or within a
particular session. When fading across successive sessions, one might conduct one session
of physical prompting, followed by a session of partial physical prompting, followed by
a session of gestural prompting, ending with a session with no prompting, provided that
correct responding continues to occur. When implementing MTL fading within a session,
one might, for example, implement a full physical prompt on the first trial of a session, a
partial physical prompt on the second trial, a gestural prompt on the third trial, and no
prompt on the fourth trial, provided that correct responding continues to occur across
trials.
LTM prompt fading, on the other hand, is typically implemented within a session,
with prompts faded in contingent on incorrect responding. For example, on the first trial
of a session, the vocal discriminative stimulus may be presented and may be followed by
a gestural or model prompt if an incorrect response occurs; this may then be followed by
a physical prompt if an incorrect response continues to occur.
Substantial research has demonstrated both LTM and MTL prompt fading to be
effective in establishing discriminations in individuals with developmental disabilities;
however, very little research has systematically compared the two approaches (MacDuff,
Krantz, & McClannahan, 2001). A recent study comparing LTM to within-session MTL
prompt fading, for teaching how to answer general knowledge questions (intraverbal
behavior) to individuals with a history of head injury, found that LTM was more effective
for teaching answers to easier questions, whereas MTL was more effective for teaching
around more difficult questions (Riley & Heaton, 2000). Given the absence of defini-
tive comparative research, clinicians would be advised to be flexible when programming
prompt-fading strategies and to attempt to empirically determine which procedure is best
for each client.
Other Variables
A nonexhaustive list of other variables that may be relevant to training individuals with
developmental disabilities in nonrelational instructional control are described below.
Chapter 6 119

Observing Response
An observing response is a response that in some way indicates to the teacher that the
student has attended to the relevant stimuli involved in the learning situation, typically
either the conditional stimulus (in nonrelational MTS) or the instruction (in nonrela-
tional instruction following). An example of an observing response in the case of nonre-
lational MTS using auditory sample stimuli is to require the student to repeat the vocal
sample stimulus back to the teacher before the teacher presents the comparison stimuli.
One might apply the same example to nonrelational instruction following by requiring
that the student repeat the instruction back to the teacher before complying with it. The
rationale behind the use of an observing response is that discrimination learning cannot
occur if the student is not attending to the relevant stimuli (Green, 2001), and it can
be difficult for a teacher to determine whether the student is attending to such stimuli
if an observing response is not required. In a study by Dube and McIlvane (1999) on
establishing delayed nonrelational MTS in adolescents with developmental disabilities,
prompting learners to actively observe sample stimuli before they selected a comparison
stimulus significantly increased their accuracy. Specifically, participants were prompted to
match identical stimuli to sample stimuli before being given an opportunity to respond
to the comparison stimuli, thereby ensuring that they had, at a minimum, attended to
the sample stimuli.
Repeating the Instruction

A recommendation that is sometimes made for training auditory discriminations is
to repeat the auditory discriminative stimulus several times each time a trial is presented,
rather than simply presenting it a single time. The rationale for this procedure is that
it may be more difficult for a student to respond to an auditory stimulus if it has not
occurred in the very immediate past, for example, one second (Green, 2001). It is impor-
tant to note that the auditory stimulus essentially comprises a prompt, in that an indi-
vidual cannot expect to have all auditory stimuli repeated to him or her several times on
an everyday basis. Therefore, if used, it should be faded out after correct responding is
established.
Mass Trialing
Mass trialing is a procedure that is commonly used in nonrelational MTS training.
Mass trialing involves presenting only one comparison stimulus for the student to respond
to when first teaching a new discrimination (Lovaas, 1981; Wynn & Smith, 2003). For
example, when teaching a student to choose between apple, orange, and pear (“receptively
identify”) in a field of three pieces of fruit on a table, mass trialing would consist of pre-
senting only the apple during initial training trials, and then presenting the conditional
stimulus “apple” and reinforcing responses to the apple. Such a performance constitutes a
simple discrimination, rather than a conditional discrimination, because the student need
only touch whatever is on the table whenever the teacher speaks (it does not require a

discrimination between the apple and anything else, since the apple is the only choice on
the table). After correct responding is established to the apple, the position of the apple
could be altered after each trial by moving it around the table. After correct responding
is established in this way, another comparison stimulus could be introduced but only the
conditional stimulus “apple” may be presented on each trial. After correct responding is
produced in this manner, the entire process might be restarted with the next fruit in isola-
tion. The end goal would be to present two or more comparison stimuli simultaneously,
with their positions being randomly rotated on each trial and each stimulus being asked
for an equal number of times.
At present, clinicians associated with various autism centers and schools have widely
divergent opinions regarding the utility of mass trialing. Many argue that nonrelational
MTS training should not include mass trialing (Green, 2001). Little research of which the
current authors are aware has systematically evaluated when, if ever, mass trialing should
and should not be conducted. On one hand, mass trialing was a central component to
large-scale, long-term, comprehensive intervention programs for individuals with autism
that have been empirically demonstrated to produce dramatic clinical gains (Lovaas,
1987). On the other hand, the mass-trialing procedure as an individual component has
not been compared to the results of its absence, so it is possible that it is not necessary.
Additionally, many studies have demonstrated the acquisition of conditional discrimi-
nations by individuals with developmental disabilities without the use of mass trialing
(Green, 2001). One possibility is that students who have not previously learned MTS
discriminations may benefit from mass trialing early on but may not continue to need it
as they acquire more conditional discriminations. In any case, clinicians would probably
be wise to be aware of this procedure and empirically evaluate whether it is helpful for the
particular clients with whom they work.
Recommendations for Teaching Nonrelational

Matching to Sample
Consider using a placement board so that the distance between compari-
son stimuli and the learner is consistent.
Ensure that the prompting strategy that is relevant for the learner and his
or her respective setting is used. For example, physical prompts might be
considered too intrusive for a child in an integrated school setting.
Consider using intrastimulus prompts when working with individuals
with autism when at all possible.
Consider using exclusion techniques to enhance accuracy.
Remember that the ultimate goal with any prompting strategy is to even-
tually eliminate it, but the fading procedure must be individualized.
Require observing responses to sample stimuli for individuals who have
trouble attending to instructional stimuli
Consider using mass trialing only for individuals who are new to
matching-to-sample tasks.
Chapter 6 121

Relevance to Generalized Relational Repertoires
Although the types of behavior-environment relations described in this chapter are not,
in themselves, relational in nature, they are likely to be critical initial building blocks
of generalized relational repertoires. The simplest relational repertoire is likely a general-
ized symmetry repertoire. A generalized symmetry repertoire consists of the tendency
to derive the concept that, if one thing equals another, then the second thing equals
the first. The relational frame theory (RFT) account of symmetry suggests that a gen-
eralized class of symmetry is acquired via a history of multiple exemplar training for
responding to one thing as equal to another, and vice versa. For example, reinforcement
might be delivered for an infant’s pointing to familiar people (such as a parent) when
the person’s name is said. Later, when the child begins emitting vocal speech, reinforce-
ment would likely be provided for producing the name of the person when visually
confronted with them (for example, saying “Mommy” when Mommy is present). That
is, one is first taught to respond to vocal stimulus A (the spoken word “Mommy”) by
responding to nonvocal stimulus B (the appearance of Mommy). One is then taught to
respond in the opposite direction—to the nonvocal stimulus B by producing the vocal
stimulus A. Hundreds of examples of the same type of learning are conceivable, given
that young humans are constantly asked to respond to the vocal names of stimuli by
responding to the nonvocal stimuli to which they refer (toys, animals, people, bedding,
food, and so on) and vice versa. The RFT account suggests that a learning history of
this type, across sufficient exemplars, will eventually result in the ability to derive the
relation in one direction when it is directly learned in the other. Little research has
directly evaluated this possibility, but initial evidence from work with typically devel-
oping children supports the RFT account (Barnes-Holmes, Barnes-Holmes, & Roche,
2001a, 2001b).
If a history of multiple-exemplar training is necessary for establishing a general-
ized class of symmetry, then nonrelational instructional control is likely one of the
primary sources of such training. Nonrelational instruction following likely comprises
one side of the symmetrical relation between motor actions and their names. Tacting
such actions likely comprises the other side of the symmetrical relation. For example,
the student walking in response to the teacher saying, “Walk,” might constitute A-B
training, whereas the student saying, “walking,” when the teacher points to someone
walking may constitute B-A training. Similarly, nonrelational MTS training is likely
one half of the multiple-exemplar training needed to establish symmetry in relation to
tabletop stimuli.
If the above analysis is accurate, then an instructional program that includes non-
relational instructional control (MTS or otherwise) should likely be closely coordinated
with an instructional program that includes naming the stimuli that are included in the
“receptive” program (see chapter 7 of this volume). For example, if a student is taught to
respond to a stimulus when its name is said, but that student is never taught to say its
name in its presence, then an opportunity for multiple-exemplar training with respect to
a generalized symmetry repertoire will be lost. Indeed, if it is accurate that the founda-
tion of language is flexible, generalized relational responding, then the establishment of
a flexible, generalized symmetry repertoire is likely the most important first step and
therefore should not be taken for granted.

Relational vs. Nonrelational Instructional Control
An important feature of both kinds of direct instructional control, which we are describ-
ing here, is that they are not relational in nature. A few examples may be illustrative.
Suppose a caregiver instructs her student, “Sit down in your seat,” after which the student
sits down in the seat. If the reason that the student sits down is because he had received
reinforcement for doing so when presented with the same instruction in the past, then
this would constitute an example of direct instructional control. The same behavior could
potentially be relational in nature if the reinforcement history had not involved direct
reinforcement for following that same instruction. For example, if the student had been
explicitly taught to comply with the instruction “Sit down in your chair” and was then
taught the synonymous relationship between the words “chair” and “seat,” then comply-
ing with the instruction “Sit down in your seat” could only occur via the derived stimulus
relation established between the words “chair” and “seat.” The same would be true if the
child was a native Spanish speaker and had a history of reinforcement for following the
same instruction stated in Spanish, and if this child was told the meaning of the instruc-
tion in English and subsequently followed the instruction the next week, despite never
having followed the instruction in the past. That is, the reason why the native Spanish
speaker follows the instruction is because of the stimulus relations established between the
phrases “Sit down in your chair” and “Sientate en tu silla.” Following instructions using
novel words or phrases thus constitutes examples of relational instructional control.
Similarly, auditory-visual simultaneous conditional discriminations (in other words,
“receptive identification”) of the sort we are describing in this chapter are nonrelational.
Consider the example of a teacher stating the instruction “Touch car,” followed by the
student selecting a picture of a car from among three pictures. If the student selects the car
because of a history of reinforcement for doing so in response to the instruction “Touch
car,” then the behavior would be of the sort we are describing in this chapter—that is,
nonrelational. It would constitute a directly trained conditional discrimination. However,
if the student in this example touched the car without such a history of direct reinforce-
ment, then it could potentially be an example of the relational sort. For example, if the
student had previously learned to touch the picture of the car when given the instruction
“Touch automobile” and later learned that the words “car” and “automobile” are synony-
mous, then touching the picture when given the instruction “Touch car” would be an
example of relational instructional control. Likewise, if the student was a native Spanish
speaker and had previously received reinforcement for naming a picture of a car “coche”
and then received reinforcement for saying “car” in response to the stimulus “coche,” then
this would likely be an example of derived equivalence.
Establishing Relational Instructional Control

Little or no previous research has attempted to establish relational instructional
control in individuals with developmental disabilities who do not already possess it. Since
it is not possible to look to the empirical literature for demonstrations of how to teach
relational instructional control, we will describe a conceptual account based on RFT, as
well as some observations from our clinical experience.
Chapter 6 123

Relational instructional control is more commonly known as rule-governed behavior.
Conceptual accounts of rule-governed behavior based on RFT have much to recommend
them; perhaps most important, they provide specific and testable predictions regarding
how rule-governed behavior is acquired. According to RFT, rule following, like all other
relational framing, is built upon generalized classes of operant behavior. In particular,
rule following depends upon an individual’s ability to respond in accordance with the
relation between the words stated in a rule and the relation between those words and
other stimuli in the environment. A full conceptual analysis of rule-governed behavior
is beyond the scope of this chapter, but we will provide an analysis of what is likely the
simplest form of rule-governed behavior and we will describe clinical procedures that are
commonly used to establish it.
Perhaps the simplest form of rule-governed behavior involves behaving with respect to
a rule that contains a single if-then relation (Barnes-Holmes, O’Hora, et al., 2001). The
if-then relation may relate to an opportunity to respond and the response itself (in other
words, “If X is present, then do behavior Y”). This type of rule specifies the antecedent
and behavior in a contingency. For example, a teacher may tell his or her class, “If you
are wearing a white shirt, then jump up.” Alternatively, the if-then relation in a simple
rule may relate to a behavior and the consequence it will produce (“If you do behavior
X, then you will receive consequence Y”). This type of rule specifies the behavior and
consequence in a contingency. For example, a parent might tell his or her child, “If you
clean your room, then I’ll take you to the playground.” A slightly more complex rule
necessitates relational responding with respect to three events, in that it specifies all three
terms of the contingency: the antecedent, behavior, and consequence. For example, “If
it is raining outside (antecedent), and you bring an umbrella (behavior), then you won’t
get wet (consequence).” The important feature of rule-governed behavior, or relational
instructional control, is that it is relational. Relational instructional control is not directly
trained behavior; it is a generalized class of relational responding that is characterized by
the fact that specific instances do not have to be directly trained once the class as a whole
is acquired. Therefore, when training relational instructional control, the goal is not for
the learner to acquire or memorize many individual instructions. The goal is for the
learner to acquire a generalized ability to follow novel instructions.
The RFT account of how relational instructional control is acquired coincides per-
fectly with anecdotal observations from our clinical experience. In practice, the way we
(and many clinicians working with young children with autism in intensive behavioral
intervention programs) establish relational instructional control is by teaching many
examples of particular instructions, each of which is different, but where each includes
the same relational cues. That is, across many different instances of instruction follow-
ing being reinforced, the particular antecedents, consequences, and behaviors described
in the rules change, but the contextual cues remain constant and the contingencies that
the rules describe are always followed through. Just as in any multiple-exemplar training,
initially acquired behaviors are under direct discriminative control—they are not rela-
tional. However, after sufficient exemplars have been trained, a generalized class forms.
For example, one might teach compliance with rules such as “If you are wearing a white
shirt, then clap your hands,” “If you are sitting on the ground, then jump,” “If you are
four years old, then stick out your tongue,” and so on. On each occasion, the antecedent
and behavior change, but the if-then contextual cues remain constant, and reinforcement
is always delivered when the rule is followed correctly. A sufficient history of such training

eventually results in the contextual cues of if-then controlling relational instruction fol-
lowing. Moreover, this is likely to occur with combinations of antecedents and behaviors
that have never before been presented. The basic teaching and learning process for estab-
lishing relational instructional control with respect to (1) rules that describe behavior and
consequence and (2) rules that describe antecedent, behavior, and consequence is presum-
ably the same. Although there is little or no research evaluating this process, children
with autism in our clinical practice are taught in this manner and acquire these skills
regularly.
Increasing Complexity
The complexity of an individual’s relational instruction-following repertoire can
presumably be enhanced via the same basic approach—multiple-exemplar training. The
complexity of an instruction-following repertoire may be defined as the degree of its elab-
oration, similar to adding multiple families of relations. For example, a frame of negation
could presumably be added to a rule by teaching a child to follow instructions such as “If
you are not wearing a white shirt, then jump.” Frames of coordination might be added
by teaching rules such as “If you are wearing the same color as your neighbor, then stand
up.” Frames of comparison might be added by teaching rules such as “If you are bigger
than your neighbor, then raise your hand.” Multiple frames may presumably be combined
in any particular sequence, all of which would likely be taught via multiple-exemplar
training. Again, research evaluating the effectiveness of this training process is only in
its infancy, but these are relatively common practices among those providing intensive
behavioral intervention for children with autism and related disabilities. More research is
clearly needed to further specify the details of such instructional procedures.
Recommendations for Teaching Relational

Instruction Following
1. Teach (in other words, directly reinforce) numerous instances of instruction
following.
2. Maintain an active curriculum in conditional discrimination training for estab-
lishing relationships between numerous words and objects.
3. Occasionally test for new instances of instruction following (in other words,
with novel objects or responses and in novel settings).
4. Target instructions with multiple families of relations (in other words,

negation).
Conclusion
Nonrelational instructional control is one of the most fundamental areas of language
intervention for individuals with autism and other developmental disorders who possess
significantly delayed verbal repertoires. Most forms of nonrelational instructional control
Chapter 6 125

can be grossly classified as either nonrelational instruction following or nonrelational
matching to sample. In either case, teaching consists of prompting and reinforcing par-
ticular behaviors in the presence of particular stimuli, and fading out prompts, all while
ensuring generalization and maintenance. Improvements in nonrelational instructional
control can significantly enhance an individual’s ability to respond in a functional way to
the spoken words of others. Relational instructional control involves relational framing.
Establishing relational instructional control in those who do not already have it has not
been researched significantly, but RFT provides a useful conceptual framework. The basic
components comprising an RFT approach to establishing relational instructional control
are (1) prompting and reinforcing compliance with very simple rules, (2) teaching multiple
exemplars, all of which vary in every manner from circumstance to circumstance, except
for the fact that they all contain the same contextual cues relevant to the rule, and (3)
expanding the complexity of the rules taught when untrained rule following emerges.
References
Barnes-Holmes, Y., Barnes-Holmes, D., & Roche, B. (2001a). Exemplar training and
derived transformation of function in accordance with symmetry. Psychological
Record, 51, 287–308.
Barnes-Holmes, Y., Barnes-Holmes, D., & Roche, B. (2001b). Exemplar training and
derived transformation of function in accordance with symmetry II. Psychological
Record, 51, 589–603.
Barnes-Holmes, D., O’Hora, D., Roche, B., Hayes, S. C., Bisset, R. T., & Lyddy, F.
(2001). Understanding and verbal regulation. In S. C. Hayes, D. Barnes-Holmes, &
B. Roche (Eds.), Relational frame theory: A post-Skinnerian account of human language
and cognition. New York: Kluwer Academic Plenum.
Bondy, A. S., & Frost, L. A. (1994). The picture exchange communication system. Focus
on Autistic Behavior, 9, 1–19.
Carr, D. (2003). Effects of exemplar training in exclusion responding on auditory-visual
discrimination tasks with children with autism. Journal of Applied Behavior Analysis,
36, 507–534.
Clark, K. M., & Green, G. (2004). Comparison of two procedures for teaching dictated
word/symbol relations to learners with autism. Journal of Applied Behavior Analysis,
37, 503–507.
Cooper, J. O., Heron, T. E., & Heward, W. L. (2006). Applied behavior analysis (2nd ed.).
Colombus, OH: Merrill Prentice Hall.
Craighead, W. E., O’Leary, K. T., & Allen, J. S. (1973). Teaching and generalization
of instruction-following in an “autistic” child. Journal of Behavior Therapy and
Experimental Psychiatry, 4, 171–176.
Dube, W. V., & McIlvane, W. J. (1999). Reduction of stimulus overselectivity with
nonverbal differential observing responses. Journal of Applied Behavior Analysis, 32,
25–33.

Gillum, H., & Camarata, S. (2004). Importance of treatment efficacy research on lan-
guage comprehension in MR/DD research. Mental Retardation and Developmental
Disabilities Research Reviews, 10, 201–207.
16, 72–85.
Leaf, R., & McEachin, J. (1999). A work in progress: Behavior management strategies and
curriculum for intensive behavioral treatment of autism. New York: DRL Books.
Lovaas, O. I. (1981). Teaching developmentally disabled children: The me book. Austin, TX:
PRO-ED.
Lovaas, O. I. (1987). Behavioral treatment and normal educational and intellectual func-
tioning in young autistic children. Journal of Consulting and Clinical Psychology, 55,
3–9.
MacDuff, G. S., Krantz, P. J., & McClannahan, L. E. (2001). Prompts and prompt-
fading strategies for people with autism. In C. Maurice, G. Green, & R. M. Foxx
(Eds.), Making a difference: behavioral intervention for autism. Austin, TX: PRO-ED.
Martin, G., & Pear, J. (2003). Behavior modification: What it is and how to do it. Upper
Saddle River, NJ: Prentice Hall.
Maurice, C., Green, G., & Luce, S. C. (1996). Behavioral intervention for young children
with autism: A manual for parents and professionals. Austin, TX: PRO-ED.
Riley, G. A., & Heaton, S. (2000). Guidelines for a selection of a method for fading cues.
Neuropsychological Rehabilitation, 10, 133–149.
Roark, T. J., Collins, B. C., Hemmeter, M. L., & Kleinert, H. (2002). Including manual
signing as nontargeted information when using a constant time delay procedure to
teach receptive identification of packaged food items. Journal of Behavioral Education,
11, 19–38.
Schreibman, L. (1975). Effects of within-stimulus and extra-stimulus prompting on dis-
crimination learning in autistic children. Journal of Applied Behavior Analysis, 8,
91–112.
Sidman, M., & W. Tailby. (1982). Conditional discrimination vs. matching to sample: An
37, 5–22.
Sundberg, M. (1992). Selecting a response form for nonverbal persons: Facilitated com-
munication, pointing systems, or sign language? Analysis of Verbal Behavior, 11,
99–116.
Touchette, P. E., & Howard, J. S. (1984). Errorless learning: Reinforcement contingen-
cies and stimulus control transfer in delayed prompting. Journal of Applied Behavior
Analysis, 17, 175–188.
Wynn, J. W., & Smith, T. (2003). Generalization between receptive and expressive lan-
guage in young children with autism. Behavioral Interventions, 18, 245–266.
Chapter 6 127

Chapter 7
Naming and Frames of Coordination
Caio F. Miguel, California State University, Sacramento;

and Anna I. Petursdottir, Texas Christian University
The study of derived stimulus relations has generated a large and fruitful body of research
and application within behavior analysis. It has also generated some controversy regard-
ing the origin of derived relational responding. The controversy centers on the observa-
tion that performance on tests of derived relations appears to be highly correlated with
verbal ability. As a result, it has been debated whether or not such performances depend
on the occurrence of covert behavior in the form of naming during training and testing
(see Clayton & Hayes, 1999; Stromer & Mackay, 1996). For example, it has been noted
that children who fail tests of derived relational responding may later pass them follow-
ing explicit training to name the stimuli to be related (Dugdale & Lowe, 1990; Eikeseth
& Smith, 1992). In addition, the nameability of stimuli (for example, how easy it is for
the learner to pronounce them) seems to affect performance (Arntzen, 2004; Randell &
Remington, 2006). However, derived relations have also been observed with animals, and
individuals with minimal verbal repertoires, in which the occurrence of naming appears
unlikely (Carr, Wilkinson, Blackman, & McIlvane, 2000).
The debate has not yet been resolved, but it seems apparent that naming may greatly
facilitate derived relational responding. When teaching children with autism or other dis-
abilities, clinicians may be able to use training protocols for various curriculum compo-
nents so that naming is employed to the learner’s advantage. In this chapter, we consider
how naming has been conceptualized within behavior analysis, describe how naming
repertoires may be established, and present several examples of how naming might be
incorporated into intermediate language intervention curricula.
What Is Naming?
In 1957, B. F. Skinner proposed that individual terms such as naming could be replaced
with a functional taxonomy of verbal operants. For example, a child saying “cat” upon
seeing a cat would be emitting a tact response, evoked by the sight of the cat, due to
a history of that response being reinforced in the presence of cats. By contrast, upon
hearing another person say, “Can you name an animal that has whiskers?” the same child
saying “cat” would be emitting an intraverbal response, which would require a reinforce-
ment history separate from that which established the tact response. Skinner’s analysis
suggests that such verbal operants are functionally independent of what he referred to as
the “behavior of the listener,” which largely involves responding receptively to the lan-
guage emitted by others. Therefore, a child capable of responding to an adult’s saying
“Look at the cat!” by orienting toward or pointing to a cat would not necessarily be able
to say “cat” unless that skill had also been directly taught. However, it appears that, for
verbally competent individuals, acquiring particular speaker or listener skills contributes
to the development of more complex skills.
Naming as a Bidirectional Relation

Recently it has been suggested that naming is a higher-order operant that involves,
at minimum, a bidirectional relation between a spoken word and a particular stimulus
(Hayes, Barnes-Holmes, & Roche, 2001; Horne & Lowe, 1996). Naming is said to exist
when the reinforcement of a listener relation is accompanied by the emergence of a speaker
relation, or vice versa. Hayes, Barnes-Holmes, and Roche (2001) described naming as a
simple frame of coordination, a type of relational response that is contextually controlled
by cues of sameness or similarity. In the case of naming, the stimuli to be related consist
of a word and its referent—for example, the sight of a cat and the word “cat”—whereas
the relevant contextual cues may include “is,” “called,” and “name of.” According to this
analysis, children come to respond relationally to words and their referents in a frame of
coordination when contextual cues are present that in the past have resulted in reinforce-
ment for symmetrical responding. For example, when a child is told that the animal that
is running around the room “is” a cat, “is” may serve as a contextual cue for symmetrical
responding; the next time the cat appears, the child may say “cat,” provided that the child
has the relevant instructional history with the word “is.” Early in a child’s development,
speaker relations, such as the tact, may be functionally independent of listener relations.
In other words, the child may say “cat” when in the presence of a cat but may not orient
toward the cat when a caregiver says “cat,” and vice versa. Then, following exposure to
multiple instances of reinforcement for both tacts and listener relations involving the
same objects in the presence of appropriate contextual cues, the child may begin to show
untrained speaker and listener relations (Barnes-Holmes, Barnes-Holmes, & Cullinan,
2000).
Horne and Lowe (1996) propose a detailed history by which bidirectional naming
may arise, initially through the incidental reinforcement of orienting responses that occur
when a tact is reinforced, and through the reinforcement of self-echoic responses that
occur when a listener relation is reinforced. For example, when a parent asks her child to

“go get the ball,” social reinforcement for retrieving the ball is also delivered contingent
upon looking at the ball as well as repeating “ball” in the presence of the object. Repeated
exposure to similar situations involving a variety of objects will create an interlocking set
of speaker and listener relations. It is thus argued that naming is a higher-order operant
involving a bidirectional relation consisting of two component relations: a speaker compo-
nent (tact) and a listener component (receptive discrimination). Naming occurs when just
one of these components, speaker or listener, suffices to establish both relations.
Recent evidence suggests that bidirectional naming relations may be important
in facilitating derived relational responding (Horne, Hughes, & Lowe, 2006; Miguel,
Petursdottir, Carr, & Michael, 2008). Thus, there may be good reason for language inter-
vention programs for children with disabilities to focus on the establishment of bidirec-
tional naming skills. In the following sections, we will use the term tact training to refer
to the reinforcement of a (typically vocal) response in the presence of a nonverbal stimu-
lus; for example, reinforcing a child’s saying “cat” in the presence of a picture of a cat. We
will use the term listener training to refer to the reinforcement of a selection response (for
example, pointing to or touching one of a number of nonverbal stimuli) in the presence
of a (typically vocal) verbal stimulus; asking a child to “find the cat” and then reinforcing
the selection of a picture of a cat would be an example. (This procedure is also known as
receptive discrimination training [see Miguel, Petursdottir, & Carr, 2005].) We will use the
term naming to refer to the emergence of a derived tact or listener component following
the direct establishment of the other component. That is, a child who receives listener
training with respect to cats and dogs and is then able to tact cats and dogs, even though
the tacts have never been reinforced, is demonstrating naming.
Teaching Naming
Traditionally, manuals that describe early language interventions for children with
disabilities have listed the completion of listener training protocols, often referred to as
the training of receptive skills, as prerequisites to teaching corresponding tacts or other
expressive language skills (Leaf & McEachin, 1999; Maurice, Green, & Luce, 1996). For
example, it is recommended that before a child is taught to vocalize color names in the
presence of color stimuli, the child should be able to respond to vocally presented color
names by selecting color stimuli from an array. This recommendation is consistent with
what seems to occur in typical language development (Fraser, Bellugi, & Brown, 1963).
However, it may not be the most efficient sequence if the goal is to establish bidirectional
naming relations. In fact, the reverse sequence may be more productive. Specifically, among
children both with and without disabilities, it appears that tact training is more likely
than listener training to generate both tact and listener skills (Connell & McReynolds,
1981; Wynn & Smith, 2003). In addition, tact training may consume fewer training trials
than listener training does (Cuvo & Riva, 1980), and prior listener training may not nec-
essarily facilitate the acquisition of new tacts (Miller, Cuvo, & Borakove, 1977). Thus, it
may be wise to begin tact instruction early on in language intervention programs.
As mentioned earlier, it seems that in typical language development children acquire
naming via a history of multiple-exemplar training. Applying this notion to teaching
children with developmental delays, Greer, Stolfi, Chavez-Brown, and Rivera-Valdes
Chapter 7 131

(2005) found that children who did not initially tact objects when they were trained in
listener relations did so following a history of multiple-exemplar training. Likewise, lis-
tener responses may well emerge as a result of tact training, as was shown in a study by
Guess and Baer (1973), who found that failure of a tactlike training condition to generate
appropriate listener responding was overcome when tact trials were interspersed with rein-
forced listener trials. Some practitioners have, in fact, long recommended interspersing
tact and listener trials during training (Sundberg & Partington, 1998) rather than adher-
ing to the more commonly proposed receptive-before-expressive instructional sequence.
Additional strategies for promoting the acquisition of naming include reinforcing echoic
(vocal imitation) responses during listener training trials. That is, a child who does not
tact objects following listener training, despite being able to echo the names of those
objects, might benefit from being prompted to echo vocal stimuli presented on listener
trials during future listener training (see Horne & Lowe, 2000).
Recommendations for Establishing Vocal Naming

1. Assessing prerequisite skills:
a. Ensure that the child has acquired a generalized echoic (vocal imitation)
repertoire.
b. Ensure that the child has acquired a number of basic tacts and listener rela-
tions involving the same stimuli. In other words, will she label a number of
objects and identify those same objects in an array that contains at least one
other object by pointing or touching?
c. Ensure that instructional control of those skills has been established. In other
words, will the child respond to the question “What’s this?” when shown an
object or picture, and identify an object or picture by pointing or touching
when told to find the item?
2. Tact training:
a. Begin instruction on new naming targets with tact training. For example, a
first step to teaching color naming might be to reinforce the vocal response
“yellow” in the presence of a yellow stimulus. If it appears that a child has
difficulty acquiring new tacts but not listener relations, it may be advisable to
begin with listener training and test for the emergence of tacts. (Although we
generally recommend beginning with tact rather than listener training, some
children may more easily acquire bidirectional naming relations with listener
than with tact training [Wynn & Smith, 2003].)
3. Listener test:
a. Following the mastery of each new tact, test for the emergence of a listener
relation. For example, present a yellow stimulus along with a different colored
stimulus and ask, “Which one is yellow?”
4. Multiple-exemplar training and testing:

a. If the child consistently responds correctly on listener trials (in other words,
touches the yellow stimulus), continue training other tacts (such as “black”),

and testing the corresponding listener relation (for example, “Which one is
black?”).
b. If, following tact training, the child typically responds correctly on listener
trials, this means that the tact training resulted in bidirectional naming.
c. If the child does not consistently respond correctly on listener trials following
tact training on the first relation, train the listener relation directly. Then go
on to train another tact, and test listener relations while continuing to allow
the child to practice both of the already mastered “yellow” relations.
5. Troubleshooting:
a. If, after training a number of exemplars, the child does not begin to respond
correctly, additional manipulations might involve interspersing tact and lis-
tener trials (see Greer et al., 2005), in which a tact training trial is immedi-
ately followed by the corresponding listener trial throughout training. This
should be done until the training of one of the skills results in the emergence
of the other.
Ultimately we would hope to see a decrease in the amount of direct tact training
required in order to establish the new naming relations. That is, the eventual goal should
be for the child to acquire at least some new naming relations as a result of merely being
taught the name of an object.
We now go on to consider how naming may be used to facilitate the acquisition of
other frames of coordination and categorization skills. In the following sections, when
we speak of tact training, we will generally assume a learner for whom tact training may
already establish bidirectional naming relations.
Frames of Coordination
The relational responding that comprises what we have called a frame of coordination has
been the object of study for stimulus equivalence researchers for many years (see Sidman,
1994). This research has focused on the study of stimulus classes whose members are not
physically similar but serve similar behavioral functions, such as the relationship between
an object or picture and its corresponding spoken name and printed word. Learning to
substitute these stimuli for each other under specific conditions seems to be of practical
concern to those working with children with disabilities. When asked to point to a ball,
for example, a child should be able to point to either a ball, a picture of a ball, or the
printed word ball. If, however, a child is asked to kick the ball, only one stimulus (the
actual ball) should function as an effective discriminative stimulus for this response.
The matching-to-sample procedure (MTS) has long been used in the study of stimu-
lus equivalence (see chapter 8 of this volume). It is also a common method used to teach
numerous skills to children diagnosed with autism and other developmental disabilities
(Maurice et al., 1996; Sundberg & Partington, 1998). In MTS, a single stimulus (sample)
is presented followed by two or more other stimuli (comparisons). For instance, a teacher
may show a picture of a ball as a sample, and the printed words ball, car, and hammer as
comparisons. Selecting one of the comparisons is always reinforced in the presence of one
Chapter 7 133

specific sample, but not in the presence of others. In the previous example, selecting the
printed word ball would only be reinforced in the presence of the picture of the ball. As
a result, the discriminative function of the comparison stimuli for the selection response
becomes dependent upon the presence of a specific sample. When individuals learn to
respond this way, they are said to have learned conditional discriminations (Green &
Saunders, 1998). If, during a series of MTS tasks, individuals learn to match comparisons
with sample stimuli, it is suggested that stimulus substitutability between sample and
comparison has been achieved.
Such substitutability between the stimuli does not always seem to occur for learners
with autism and other developmental disabilities; however, it may be achieved by merely
teaching individuals to tact stimuli. It has been reported that performance on matching-
to-sample tasks (Eikeseth & Smith, 1992) and on relational responding can be greatly
improved by teaching individuals to tact related stimuli with a common name (Goyos,
2000). These results imply that, for learners for whom tact training already establishes
bidirectional naming relations, tact training alone may be sufficient for the establishment
of frames of coordination, as shown by mastered performance on matching-to-sample
tasks. Listener training alone, however, may also be sufficient (Miguel et al., 2008). In
one of our clinical studies, two preschoolers diagnosed with autism demonstrated substi-
tutability among spoken words, pictures, and printed words after receiving either listener
or speaker training (Stone, Miguel, & Gould, 2006). Listener training consisted of a
typical MTS preparation involving reinforcement contingent on AB relations (selecting a
picture after hearing the name of the object in the picture) and AC relations (selecting the
printed word after hearing the name). Speaker training consisted of reinforcement of tact
(BD) relations (saying the name of the picture in its presence) and textual (CD) relations
(reading the textual stimulus when presented). Both speaker and listener training proce-
dures took approximately the same number of trials in order to be mastered. In addition
to the BC (selecting the correct printed word in the presence of its picture referent) and
CB (selecting the correct picture referent in the presence of the word) derived relations,
participants were able to select the correct pictures when hearing the names of the objects
in the pictures (listener behavior) after learning the tact and textual relations (speaker
behavior) and vice versa, which suggested that naming skills had been acquired.
These studies have important clinical implications. Once bidirectional naming skills
are acquired, as previously described, speaker training alone may serve to establish frames
of coordination such as the ones involved in reading comprehension. Thus, for a child to
comprehend that the printed words ball and pelota and a picture of a ball have the same
meaning, it may be sufficient to teach her to label all of these stimuli as “ball.” If the child
were to only label the printed word ball (textual behavior), this could not be considered
reading since there is no guarantee that such training would establish the printed word,
object, and spoken word as equivalent. Reading comprehension can only be achieved
when word and referent are related to one another in a frame of coordination (see figure
7.1).

Object
Auditory Label
Picture
Printed
Word
Figure 7.1. Relations involved in a frame of coordination among auditory stimuli (words
spoken by experimenter), objects, pictures, printed words, and labels (words spoken by
the participant). Solid lines depict the components of the listener and speaker relations
(naming). Dotted lines depict derived, or untrained, relations.
Recommendations for Establishing Frames of Coordination

via Speaker Training
1. Tact training (BD):
a. Select at least three pictures of objects and their respective printed words (six
stimuli total). These pictures can be labeled B1, B2 and B3. Examples may
include a picture of a bike (B1), a car (B2), and a truck (B3).
b. Each trial begins with the presentation of a sample picture (B). Stimuli may
be presented one at a time. Say, “What is this?” while pointing to the picture.
Use a prompt-delay or errorless training procedure (echoic prompt) for the
child’s vocal responses (D).
2. Textual training (CD):

a. Repeat steps a and b above, replacing B1, B2, and B3 with the printed words
bike (C1), car (C2), and truck (C3).
3. Mixed training (BD/CD):

a. Repeat the procedures described above, mixing tact (oral labeling, BD) and
textual (“reading,” CD) trials.
4. Derived relations test (CB and BC):
Chapter 7 135

a. Each trial begins with the presentation of either a picture (B) or a printed word
(C). The child should be required to point at the sample (observing response)
prior to the presentation of the comparison (either B or C stimuli).
b. Presentation of comparison stimuli should be randomized so that each picture
serves as a correct choice the same number of times in the right, middle, and
left positions.
c. These trials should not be reinforced; therefore, BD (tact) and CD (textual)
trials can be interspersed and reinforced to keep the child motivated.
5. Troubleshooting:
a. If the child fails to match the correct printed word to the picture (CB) and
the correct picture to the printed word (BC) stimuli, this skill should be
taught directly using the same errorless training procedure as in 1 and 2
above.
Recommendations for Establishing Frames of Coordination via

Listener Training
1. Receptive discrimination of pictures (AB):
a. Select at least three pictures of objects and their respective printed words
(six stimuli total). These pictures of objects can be labeled B1, B2, and B3.
Examples may include a picture of a bike (B1), a car (B2), and a truck (B3).
b. Each trial begins with the presentation of a sample spoken by the experi-
menter. Sample stimuli can be labeled A1, A2, and A3. Examples include
the words (spoken by the experimenter) “bike” (A1), “car” (A2), and “truck”
(A3).
c. If the child can echo, require that she repeat the sample. This guarantees that
the child is attending to the sample (observing response).
d. After the child repeats the sample, present the three comparison stimuli: B1,
B2, and B3 (pictures).
e. Use a most-to-least prompt-delay or errorless training procedure to teach the
child to point to the correct comparison stimulus.
f. Presentation of comparison stimuli should be randomized so each picture
left positions.
2. Receptive discrimination of printed words (AC):

a. Repeat steps A through F above, replacing B1, B2, and B3 with the printed
words bike (C1), car (C2), and truck (C3).
3. Mixed training (AB-BA):

a. Repeat the procedures described above mixing spoken-word-to-picture (AB)
and picture-to-spoken-word (BA) trials.

4. Derived relations test (CB and BC):
a. Each trial begins with the presentation of either a picture (B) or a printed word
(C). The child should be required to point at the sample (observing response)
prior to the presentation of the comparison (either B or C stimuli).
b. Presentation of comparison stimuli should be randomized so each picture
left positions.
c. These trials should not be reinforced; therefore, picture-to-word (BC) and
word-to-picture (CB) trials can be interspersed and reinforced to keep the
child motivated.
5. Troubleshooting:
a. If the child fails to match the correct printed word to picture (BC) and the
correct picture to printed word (CB), this skill should be taught directly
using the same errorless training procedure as in 1 and 2 above. For more
details regarding this type of training see Green and Saunders (1998).
Naming and Stimulus Categorization

The process of determining how to group objects or events together is usually called
categorization or classification, while those objects or events that cohere may be regarded
as a category or a class. It is often assumed by cognitive scientists that the categorization
process is dependent upon the acquisition of specific concepts (Quilliam, 1968; Rosch,
1975). These concepts are said to be units of mental representation that are independent
of any behavior-environment relation (Zentall, Galizio, & Critchfield, 2002). A concept,
however, should be defined as a group of objects (such as stimuli or actions) that control
similar responses. The concept “chair,” for instance, involves a collection of objects that
evoke sitting behavior. As a result, concepts may be equated to stimulus classes, or groups
of stimuli that exert the same function. It is believed that stimulus equivalence serves as
a useful model for what cognitive psychologists refer to as language categories (Galizio,
Stewart, & Pilgrim, 2001; Pilgrim & Galizio, 1996).
When different objects produce the same speaker-and-listener behavior, they are said
to acquire the same meaning (Horne & Lowe, 1996). For example, a child may learn to
react as a listener when hearing the word “animal” by looking at the animal (such as a
bird). Later, when learning that a giraffe, which she has never seen before, is also called
an “animal,” the child would be able to appropriately label the giraffe by saying “animal”
in its presence. When she hears herself saying “animal” (either overtly or covertly), all of
the behaviors previously associated with hearing someone say “animal” would occur. This
would include categorizing or putting together all of the stimuli or objects previously
called “animals.”
A series of studies conducted in the last five years (Horne, Lowe, & Randle, 2004;
Horne et al., 2006; Lowe, Horne, Harris, & Randle, 2002; Lowe, Horne, & Hughes,
2005; Miguel et al., 2008) have evaluated effective procedures for producing categori-
zation skills in young children: children between one and six years old were taught to
Chapter 7 137

either label two sets of objects with a common name (speaker behavior; for example, label
pictures of northern and southern American state maps as either “north” or “south”),
or point to the objects given a common name by the experimenter (the listener behav-
ior; for example, select the correct picture of a state map when hearing either “north”
or “south”). After this training, children were tested to see whether they could catego-
rize these same objects. The categorization test included presenting both sets of visual
stimuli (for example, pictures) to the participants. One of the stimuli was then with-
drawn from the array and held up in front of the participants (for example, a northern
state). Participants were then asked to select the remaining stimuli belonging to the same
set (the two remaining northern states). Results from all studies suggest that children
who acquired bidirectional naming relations (listener and speaker) also demonstrated
untrained stimulus sorting skills. In other words, it was only when children were able to
label the pictures as “north” and “south” and receptively select the pictures from an array
when hearing either “north” or “south” spoken by the experimenter that they were able to
match the pictures by category—a skill that is targeted and expanded upon in many early
childhood settings. By contrast, children who failed to acquire tacts as a result of listener
training (or vice versa) did not correctly match the stimuli.
These studies have important applied implications. The skill of sorting or matching
objects or pictures by category (for example, sorting pictures of maps of northern and
southern states) may develop with no direct training when children learn to tact pictures
and objects with a common category name. In our own clinical practice, we have observed
derived categorization or sorting after teaching common tacts to children diagnosed with
autism (Miguel, 2006; Miguel, Kobari, & Findley, 2009). After children were taught to
tact the name and the category of several pictures belonging to three different stimulus
sets, they demonstrated derived sorting skills. In other words, after being taught to label
pictures of foods, clothing, and animals by name and category (for example, “This is a
bear and an animal”); children were able to sort these pictures into categories by placing
them into different piles according to the category to which they belonged. Hence, emer-
gent sorting or categorization skills may be obtained if children are first taught to tact the
pictures to be sorted (see figure 7.2).
Recommendations for Establishing Derived Categorization Skills

1. Simple tact training (naming stimuli):
a. Select nine to twelve pictures or objects belonging to three different catego-
ries. Possibilities include animals, furniture, and clothing.
b. Stimuli may be presented in an array or one at a time. Say, “What is
this?” while pointing to the item. Use a prompt-delay or errorless training
procedure.
2. Multiple-tact training (naming stimuli and categories):

a. With three items on the table, say, “What is this?” while pointing to one
item. After the initial response, state, “Right. This is a [item], and [prompt
the category name using the same prompt hierarchy as in 1 above].” Use
prompt delay or errorless training procedure until the child can reliably tact
both the name and category of each stimulus.

Visual Visual
Exemplar 1 Exemplar 2
Auditory Auditory Auditory

1 Common 2
Label 1 Label Label 2

Common
Figure 7.2. Relations typically involved in programs that teach categorization. For sim-
plification, a category is shown that contains only two exemplars. A frame of coordi-
nation is shown among visual stimuli, common auditory stimuli (words spoken by the
teacher), and common labels (words spoken by the learner). Hierarchical relations are
shown between common auditory stimuli and exemplar-specific labels, and vice versa. In
addition, relations are shown between different exemplar-specific labels. Solid lines depict
the components of the listener and speaker relations (naming).
3. Category test:
a. Place one item from each category in a clear bin, one per bin, or on the table in
front of the child (the child should be able to see the picture or object if it is in a
bin). Present the remaining stimuli to the student and state, “Put these in catego-
ries,” “Put these in the right bins,” “Match,” or “Sort.” (Use familiar instructions
and be consistent across trials; to establish instructional control, the child should
be exposed to similar procedures with familiar items.)
4. Troubleshooting:
a. If the child fails to categorize, this skill should be taught directly using the
same errorless training procedure as in 1 and 2 above. (Sample data sheets
are provided in this chapter’s appendix.)
Although the success of this methodology has been extensively demonstrated with
typically developing children, we are currently evaluating its effectiveness with chil-
dren diagnosed with autism; therefore, the above recommendation should be taken with
caution. However, if individuals with disabilities, many of whom require numerous trials
to learn how to sort pictures and objects, are able to categorize after learning to tact
Chapter 7 139

stimuli, then categorization would not need to be taught directly but would emerge as an
outcome of these language training procedures.
In addition to teaching children to categorize visual stimuli and name categories, cur-
ricula designed for use with individuals with autism (Leaf & McEachin, 1999; Maurice
et al., 1996; Sundberg & Partington, 1998) typically include programs designed to estab-
lish hierarchical intraverbal relations between category names and exemplar names (for
example, to answer questions such as “What are some animals?” or “What is a cat?”).
It is generally assumed that these skills will be taught directly, typically subsequent to
teaching exemplar naming, categorization of visual stimuli, and common naming. It is
possible, however, that the categorization curriculum can be structured such that those
skills will emerge in the absence of direct training. It appears, that it should be possible to
structure the training of the other components of a categorization curriculum in such a
way that those skills would emerge without direct training given the appropriate contex-
tual cues (see figure 7.2). For example, if a child has learned to respond with both “cat”
and “animal” upon seeing a cat, we might expect the emergence of intraverbal relations
between “cat” and “animal.” Or, if a child has learned “cat” as a tact and then learns
to respond with “animal” to the question “What is a cat?” we might expect to see the
child select a cat from among other stimuli in response to the question “Which one is an
animal?” So far, such effects have not been demonstrated. Studies on the effects of com-
bining tact and/or listener training of exemplar names and category names have generally
demonstrated minimal effects on typically developing children’s intraverbal responding
(Miguel et al., 2005; Partington & Bailey, 1993; Petursdottir et al., 2008). In addition,
we found little evidence that direct intraverbal training, in combination with training of
exemplar names, resulted in the emergence of other relations (Petursdottir et al., 2008).
For one of two children we worked with, one instance of training resulted in the emer-
gence of bidirectional relations with respect to common category names; however, the
effects of these different training sequences remain to be further investigated.
Conclusion
We have described the naming repertoire as composed of the bidirectional relation
between listener and speaker behavior, or, in other words, the frame of coordination
between words and their referents. The emergence of either speaker or listener behavior
after the direct training of one of those skills could be referred to as derived naming.
Like any other relational frame, naming is a higher-order operant established via a history
of direct reinforcement of the word-object bidirectionality with multiple stimuli. This
history encompasses the contingencies involved in child-caregiver interactions at early
stages of language development (Horne & Lowe, 1996). However, when language fails
to develop as a consequence of this typical history of reinforcement (as is the case with
many children with developmental disabilities), the task of teaching complex linguistic
skills such as derived naming lies with educators.
As previously mentioned, the establishment of naming skills is important for many
reasons. First, it allows us to overcome the functional independence of listener (receptive)
and speaker (expressive) behavior that may be observed in early speakers with and without
developmental delays. Second, naming may be important in facilitating the acquisition

of additional frames of coordination, such as the one involved in reading comprehension.
Third, it may serve to remedy failures observed during conditional discrimination tasks.
Finally, it facilitates the development of derived nonverbal responding such as stimu-
lus categorization. We should stress that, although we have made suggestions for teach-
ing language to children with autism and other developmental disabilities, the existing
data on naming and derived relational responding in such special populations is limited.
Although these data are promising, further work is necessary to translate findings from
basic experimental research into application. The suggestions presented here are direct
extrapolations from a body of experimental research that has included mostly typically
developing children as participants, or laboratory experiments involving educationally
irrelevant stimuli. Thus, we hope that our suggestions will inspire future applied research
that will pave the way for effective application.
References
Arntzen, E. (2004). Probability of equivalence formation: Familiar stimuli and training
sequence. Psychological Record, 54, 275–291.
Barnes-Holmes, D., Barnes-Holmes, Y., & Cullinan, V. (2000). Relational frame theory
and Skinner’s Verbal Behavior: A possible synthesis. Behavior Analyst, 23, 69–84.
Carr, D., Wilkinson, K. M., Blackman, D., & McIlvane, W. J. (2000). Equivalence classes
with individuals with minimal verbal repertoires. Journal of the Experimental Analysis
Clayton, M. C., & Hayes, L. J. (1999). Conceptual differences in the analysis of stimulus
equivalence. Psychological Record, 49, 145–161.
Connell, P. J., & McReynolds, L. (1981). An experimental analysis of children’s generaliza-
tion during lexical learning: Comprehension or production. Applied Psycholinguistics,
2, 309–332.
Cuvo, A. J., & Riva, M. T. (1980). Generalization and transfer between comprehen-
sion and production: A comparison of retarded and nonretarded persons. Journal of
Dugdale, N., & Lowe. C. F. (1990). Naming and stimulus equivalence. In D. E. Blackman
& H. Lejeune (Eds.), Behaviour analysis in theory and practice: Contributions and con-
troversies. Hove, England: Erlbaum.
Eikeseth, S., & Smith, T. (1992). The development of functional and equivalence classes
in high-functioning autistic children: The role of naming. Journal of the Experimental
Fraser, C., Bellugi, U., & Brown, R. (1963). Control of grammar in imitation, comprehen-
sion, and production. Journal of Verbal Learning and Verbal Behavior, 2, 121–135.
Galizio, M., Stewart, K. L., & Pilgrim, C. (2001). Clustering in artificial categories: An
equivalence analysis. Psychonomic Bulletin and Review, 8, 609–614.
Chapter 7 141

Goyos, C. (2000). Equivalence class formation via common reinforcers among preschool
children. Psychological Record, 50, 629–654.
Green, G., & Saunders, R. R. (1998). Stimulus equivalence. In K. A. Lattal & M. Perone
(Eds.), Handbook of research methods in human operant behavior. New York: Plenum
Press.
Greer, R. D., Stolfi, L., Chavez-Brown, M., & Rivera-Valdes, C. (2005). The emergence
of the listener to speaker component of naming in children as a function of multiple
exemplar instruction. Analysis of Verbal Behavior, 21, 123–134.
Guess, D., & Baer, D. M. (1973). An analysis of individual differences in generalization
between receptive and productive language in retarded children. Journal of Applied
Hayes, S. C., Barnes-Holmes, D., & Roche, B. (2001). Relational frame theory: A post-
Skinnerian account of human language and cognition. New York: Kluwer Academic
Plenum.
Horne, P. J., Hughes, J. C., & Lowe, C. F. (2006). Naming and categorization in young
children: IV. Listener behavior training and transfer of function. Journal of the
Horne, P. J., & Lowe, C. F. (2000). Putting the naming account to the test: Preview of
an experimental program. In J. C. Leslie & D. Blackman (Eds.), Experimental and
applied analysis of human behavior. Reno, NV: Context Press.
Horne, P. J., Lowe, C. F., & Randle, V. R. L. (2004). Naming and categorization in
young children: II. Listener behavior training. Journal of the Experimental Analysis of
Behavior, 81, 267–288.
Leaf, R., & McEachin, J. (1999). A work in progress: Behavior management strategies and a
curriculum for intensive behavioral treatment of autism. New York: DRL.
Lowe, C. F., Horne, P. J., Harris, F. D. A., & Randle, V. R. L. (2002). Naming and cate-
gorization in young children: Vocal tact training. Journal of the Experimental Analysis
Lowe, C. F., Horne, P. J., & Hughes, J. C. (2005). Naming and categorization in young
children: III. Vocal tact training and transfer of function. Journal of the Experimental
Maurice, C., Green, G., & Luce, S. (Eds.). (1996). Behavioral intervention for young chil-
dren with autism: A manual for parents and professionals. Austin, TX: PRO-ED.
Miguel, C. F. (2006). Language development: A behavior analytic research agenda.
Paper presented at the annual convention of the California Association for Behavior
Analysis, Burlingame, CA.
Miguel, C. F., Kobari, V., & Findlay, K. (2009). The effects of multiple tact training on
the emergence of naming and categorization by children with autism. Paper presen-

ted at the annual convention of the California Association for Behavior Analysis,
Burlingame, CA.
Miguel, C. F., Petursdottir, A. I., & Carr, J. E. (2005). The effects of multiple-tact and
receptive-discrimination training on the acquisition of intraverbal behavior. Analysis
of Verbal Behavior, 21, 27–41.
Miguel, C. F., Petursdottir, A. I., Carr, J. E., & Michael, J. (2008). The role of naming
in stimulus categorization by preschool children. Journal of the Experimental Analysis
Miller, M. A., Cuvo, A. J., & Borakove, J. (1977). Teaching naming of coin values:
Comprehension before production versus production alone [Abstract]. Journal of
Partington, J. W., & Bailey, J. S. (1993). Teaching intraverbal behavior to preschool chil-
dren. Analysis of Verbal Behavior, 11, 9–18.
Petursdottir, A. I., Carr, J. E., Lechago, S. A., & Almason, S. M. (2008). An evaluation
of intraverbal training and listener training for teaching categorization skills. Journal
of Applied Behavior Analysis, 41, 53–68.
Pilgrim, C., & Galizio, M. (1996). Stimulus equivalence: A class of correlations or a cor-
relation of classes. In T. R. Zentall & P. M. Smeets (Eds.), Stimulus class formation in
humans and animals. Amsterdam, Holland: Elsevier.
Quilliam, M. R. (1968). Semantic memory. In M. Minsky, (Ed.), Semantic information
processing. Cambridge, MA: MIT Press.
Randell, T., & Remington, B. (2006). Equivalence relations, contextual control, and
naming. Journal of the Experimental Analysis of Behavior, 86, 337–354.
Rosch, E. (1975). Cognitive representations of semantic categories. Journal of Experimental
Psychology: General, 104, 192–233.
Sidman, M. (1994). Equivalence relations and behavior: A research story. Boston: Authors
Cooperative.
Stone, A. M., Miguel, C. F., & Gould, D. D. (2006). The effects of conditional discrimi-
nation and tact-textual training on the development of equivalence classes. Poster
presented at the annual convention of the Association for Behavior Analysis, Atlanta,
GA.
Stromer, R., & Mackay, H. A. (1996). Naming and the formation of stimulus classes. In
T. R. Zentall & P. M. Smeets (Eds.), Stimulus class formation in humans and animals.
Amsterdam, Holland: Elsevier.
Sundberg, M. L., & Partington, J. W. (1998). Teaching language to children with autism or
other developmental disabilities. Pleasant Hill, CA: Behavior Analysts, Inc.
Wynn, J. W., & Smith T. (2003). Generalization between receptive and expressive lan-
guage in young children with autism. Behavioral Interventions, 18, 245–266.
Chapter 7 143

Zentall, T. R., Galizio, M., & Critchfield, T. S. (2002). Categorization, concept learn-
ing, and behavior analysis: An introduction. Journal of the Experimental Analysis of
Behavior, 78, 237–248.
Authors’ Notes
The preparation of this chapter has been partially supported by the California State
University, Sacramento, RCA Summer Fellowship awarded to the first author. We would
like to thank Katharine Findlay for helping to create the data sheets included in the
appendix.

APPENDIX
Data Sheets for Categorization Procedure

Categorization Testing1
Instruction: “Match.”
Correct trial: Child responds by matching the sample to the comparison from the same category → “Okay.” → (+)
Incorrect trial: Child responds by matching the sample to a comparison from another category → “Okay.” → (-) or does not respond
within 5 seconds → “Okay.” → (-)
Block 1 Circle one: Pre-Test/Post-Test Block 2 Circle one: Pre-Test/Post-Test

Teacher Date IVI 2 Teacher Date IVI
Trial Sample L M R +/- Ø Sr+ LMR Trial Sample L M R +/- Ø Sr+ LMR
1 B3 B1 A3 C2 1 A2 B2 A1 C3
2 A1 C1 B2 A3 2 C3 C1 B2 A3
3 C2 A1 C3 B2 3 B1 A1 B3 C2
4 A3 B2 A1 C3 4 B2 B1 A3 C2
5 B2 A2 B3 C1 5 C1 A3 C2 B1
6 C3 B3 C1 A2 6 B3 C3 A1 B2
7 B1 B3 C2 A1 7 C2 B1 A2 C3
8 A2 A3 C2 B1 8 A1 A2 C1 B3
9 C1 A2 B1 C3 9 A3 A2 B3 C1
# correct / 9 = % # correct / 9 = %
Block 1 Circle one: Pre-Test/Post-Test

Teacher Date IVI
Trial Sample L M R +/- Ø Sr+ LMR
1 C1 B3 A1 C2
2 B2 C1 B3 A2
3 A3 A2 C1 B3
4 A1 B1 C3 A2
5 B3 B2 C3 A1
6 A2 C2 B1 A3
7 C3 A3 C2 B1
8 B1 B2 A3 C1
9 C2 A1 B2 C3
# correct / 9 = %
Prompting: All blocks should be completed in one session. Instructions should be given without any visual or physical cues.
Reinforcement: Trials are not reinforced. Reinforced maintenance trials may be interspersed (every 2 to 3 trials)
Scoring: + Indicates correct responding without prompt. – indicates incorrect or response after 5 seconds of the delivery of the
instruction.
Mastery: N/A
Categories: 1 2 3
Exemplars: A1 A2 C3
B1 B2 C3
C1 C2 C3
1 Categorization Testing is conducted under extinction. Teacher should acknowledge response (e.g., “okay”) regardless of accuracy. Reinforced mastered trials
can be interspersed with categorization trials.
2 The last two columns of each block are for collection of independent variable integrity (IVI) data. Teacher should make a + or – on whether reinforcement
was/was not delivered (Ø Sr+) and whether the comparisons were presented in the correct position in front of the participant (LMR)

Multiple-Tact: Category Mixed
Instruction: “What is this?”
Correct Prompt: Child responds with the correct exemplar name. —> “That’s right! And what else? [Category name]” Child
responds with correct category name. —> “Good job!” —> +p.
Correct Ind: Child responds with the correct exemplar and category name before prompt. —> “Good Job!” +
Incorrect Trial: Child responds with an incorrect exemplar or category name, or does not respond within 5 seconds. —> Prompt:
“What is this?” [Exemplar and Category names]
a. Child responds with the correct exemplar and category name —> “Good Job!” —> -.
b. Child responds with only exemplar name, only category name, an incorrect exemplar or category name, or does not
respond within 5 seconds. —> [Exemplar and Category.] —> -.
Block 1 Block 2
Teacher Date IVI Teacher Date IVI
Trial Exemplar Category +p/+/- Sr+ Exe Trial Exemplar Category +p/+/- Sr+ Exe
1 B2 2 1 A1 1
2 A1 1 2 C2 2
3 A3 3 3 B3 3
4 C2 2 4 A2 2
5 B3 3 5 C1 1
6 A2 2 6 B2 2
7 B1 1 7 C3 3
8 C3 3 8 B1 1
9 C1 1 9 A3 3

Block 1 Block 1
Trial Exemplar Category +p/+/- Sr+ Exe Trial Exemplar Category +p/+/- Sr+ Exe
1 A3 3 1 C3 3
2 C1 1 2 B3 3
3 A2 2 3 C1 1
4 B2 2 4 C2 2
5 C3 3 5 A1 1
6 A1 1 6 B2 2
7 B3 3 7 A3 3
8 B1 1 8 A2 2
9 B2 2 9 B1 1
Prompting: Begin with two blocks at a 0-second delay. Proceed to prompting at a 1-second delay, until the child has 2 consecutive
blocks of +8/9. Proceed to prompting at a 2-second delay, until the child has 2 consecutive blocks of +8/9. Proceed to prompting at a
3-second delay, until the child has 2 consecutive blocks of +8/9. Proceed to prompting at a 4-second delay, until the child has 2 consecu-
tive blocks of +8/9. Instructions should be given without any visual or physical cues. If the child makes three consecutive errors within
a block, return to previous second delay and repeat trials.
Reinforcement: Trials are reinforced at the prescribed prompt. Differentially reinforce independent responses (prior to the prompt) by
increasing the magnitude of the reinforcer and by being more enthusiastic.
Scoring: +p indicates correct answer was prompted (i.e., all 0-second delay). + indicates correct answer without prompting. - indicates
incorrect answer, or delay in delivering correct answer
Categories: 1 2 3
Exemplars: A1 A2 C3
B1 B2 C3
C1 C2 C3
Chapter 7 147

LISTENER TESTING3
Instruction: “Give me the [Category].”
Correct Trial: Child responds by choosing the comparison from the correct category —> “Okay.” —> +
Incorrect Trial: Child responds by choosing a comparison from another category —> “Okay.” —> - Or does not respond within 5
seconds —> “Okay.” —> -

Teacher Date IVI 2 Teacher Date IVI 2
Trial L M R Category +/- Ø Sr+ LMR Trial L M R Category +/- Ø Sr+ LMR
1 A1 A2 A3 3 1 B3 B1 B2 1
2 A2 A1 A3 1 2 B2 B1 B3 2
3 A2 A3 A1 2 3 B2 B3 B1 3
4 A3 A1 A2 3 4 B1 B3 B2 2
5 A3 A1 A2 2 5 B3 B2 B1 1
6 A1 A2 A3 1 6 B3 B2 B1 3
7 A3 A2 A1 2 7 B1 B2 B3 2
8 A2 A3 A1 1 8 B1 B3 B2 1
9 A1 A3 A2 3 9 B2 B1 B3 3

Trial L M R Category +/- Ø Sr+ LMR Trial L M R Category +/- Ø Sr+ LMR
1 C2 C1 C3 3 1 C1 A3 B2
2 C1 C2 C3 2 2 B2 A1 C3
3 C3 C2 C1 1 3 C3 B1 A2
4 C2 C3 C1 2 4 B1 C2 A3
5 C2 C3 C1 3 5 B3 C2 A1
6 C1 C2 C3 1 6 B2 C1 A3
7 C3 C1 C2 3 7 A2 C3 B1
8 C3 C1 C2 1 8 B3 A2 C1
9 C1 C3 C2 2 9 A1 B3 C2
Prompting: All blocks should be completed in one session. Instructions should be given without any visual or physical cues.
Reinforcement: Trials are not reinforced. Reinforced maintenance trials may be interspersed (every 2 to 3 trials)
Scoring: + indicates correct responding without prompt. – indicates incorrect or no response after 5 seconds of the delivery of the instruction.
Mastery: N/A
Categories: 1 2 3
Exemplars: A1 A2 C3
B1 B2 C3
C1 C2 C3
3 Listener Testing is conducted under extinction. Teacher should acknowledge response (e.g., “okay”) regardless of accuracy. Reinforced mastered trials can
be interspersed with listener trials.

Chapter 8
Acquiring the Earliest Relational Operants:

Coordination, Distinction, Opposition,
Comparison, and Hierarchy
Carmen Luciano, Miguel Rodríguez, Israel Mañas, and

Francisco Ruiz, University of Almeria, Spain;
Nicholas M. Berens, Center for Advanced Learning, Reno;
Sonsoles Valdivia-Salas, University of Albany, New York
The current chapter describes training of the earliest generalized relational operants,
which involve coordination, distinction, opposition, comparison, and hierarchical rela-
tions. From the outset, we would argue strongly that these verbal repertoires are the roots
of speaking with meaning and listening with understanding. Furthermore, it is our belief
that training these relational operants does much to enhance existing behavior analytic
remediation programs, allowing them to be useful tools in successfully promoting cogni-
tive and linguistic development and sophistication.
A General Overview of Behaviorally Oriented

Language Programs
Early and intensive behavioral intervention is widely recognized as a successful approach
to the remediation of behavioral deficits and excesses in autism spectrum disorder (ASD)
and other developmental disabilities. Although aspects of behavioral training have also
been widely incorporated into other programs that are, for example, more linguistically
or occupationally oriented, they have generally found their home in stringent behavior
analytic regimes, such as those based on Skinner’s (1957) Verbal Behavior (e.g., Greer,
2002). To summarize the basic training elements of these programs, consider the fol-
lowing brief reviews of Lovaas’s (1977) discrete trial training (DTT) and Greer’s (2002)
Comprehensive Application of Behavior Analysis to Schooling (CABAS).
The core elements of DTT may be summarized as follows: training basic rule follow-
ing, verbal imitation, and matching (“first language”); receptive followed by expressive
designation of objects (“intermediate language”); and finally training in the use of prepo-
sitions, pronouns, concepts regarding time, and so on (“advanced language”). CABAS
aims to develop different levels of verbal or instructional hierarchies, arranged in such a
way that the complexity is increased progressively, thereby representing naturally occur-
ring developmental stages (Selinske, Greer, & Lodhi, 1991). These include prelistener;
listener; speaker; speaker-listener; speaker as own listener; reader; writer; writer as own
reader; and verbally governed behavior for problem solving, conversational skills, and
self-talk, math, and writing skills. Recently, CABAS has incorporated additional training
procedures involving multiple-exemplar training (MET) as a means of establishing novel
behavior without direct instruction. Taken together, then, behavior analytic programs
typically promote first vocal imitation and identity matching, then receptive skills, and
finally expressive skills.
Despite their widespread success, it may be surprising to learn that published
empirical support for many aspects of behavioral intervention is not extensive (for
example, Carr & Firth, 2005), and significant limitations have been noted (Luciano,
1984; Sundberg & Michael, 2001). Common criticisms, for example, include concerns
about the implicit nature of the attainment of goals and the sequencing of the language
training. Indeed, follow-up data reveal that one of the main problems associated with
DTT is limited attainment of a range of generalized repertoires (McEachin, Smith,
& Lovaas, 1993). In fact, it is fair to say that little effort on the whole is devoted
to training and testing generalized responding that facilitates the emergence of novel
behaviors. In particular, this would include training in bidirectional stimulus relations
(discussed in greater detail below); using multiple exemplars and learning sets that
include stimuli that do not share formal properties with those used during explicit
training; and programming the transition from nonarbitrary stimulus relations to arbi-
trary relations. Indeed, numerous researchers have argued that these three features in
particular are probably at the root of our basic repertoires of generalized responding.
From this perspective, therefore, programs in which this type of training receives little
attention would, not surprisingly, be associated with limitations in the development of
generalized response patterns.
It is perhaps, then, all the more unfortunate, and even paradoxical, that weaknesses
in generalized responding may in fact be central to autism and developmental disabil-
ity. For instance, individuals in these categories repeat the same names (of objects, body
parts, and people, for example) again and again, but they don’t use them in conversa-
tion, or they do not learn new related names without additional training. In the follow-
ing sections, we argue that training in derived stimulus relations would be beneficial in
establishing these important generalization skills.

New Avenues in the Training of Language Abilities
Research on stimulus equivalence and derived stimulus relations has generated much
behavioral interest in language and cognition. As far back as 1971, Sidman highlighted
the importance of these phenomena for our understanding of language processes and
how they might best be harnessed for the purposes of remediation. Since then, further
advances in understanding and training derived stimulus relations have come from rela-
tional frame theory (RFT; see Hayes, Barnes-Holmes, & Roche, 2001). For example,
RFT has proposed an account of basic language processes that range from simple tacting
to understanding and producing complex analogies. Yet, despite these developments,
language training regimes for developmentally delayed populations have been slow to
incorporate derived stimulus relations into their programs.
Understanding Derived Stimulus Relations

The act of relating is defined as responding to one event in terms of another and is not
inherently complex behavior. That is, nonhumans can be trained to respond relationally
in a manner that focuses on the relations between stimuli, rather than on their physical
properties. For example, a monkey can be trained to respond to the taller of two stimuli,
in which case it is responding to the relative difference in the height of the stimuli, rather
than to their individual height. Nonetheless, this sort of relational responding remains
under the control of the formal or nonarbitrary properties of the stimuli (in other words,
one is actually taller than the other), and as such, from the RFT point of view, the behav-
ior is not considered verbal. For relational responding to be verbal, it has to be controlled
by contextual cues beyond the nonarbitrary stimulus properties that participate in the
relation. For example, a verbally sophisticated individual can be taught that a one-euro
coin is worth more than a fifty-cent coin, and that person will automatically derive that
a fifty-cent coin is worth less than one euro. This derived relation is not based on the
actual difference in the size of the coins, because the fifty-cent coin is actually the larger
(although it is worth less). In a sense then, the symbolic value relations (worth more than
and worth less than) have been arbitrarily applied to the coins. For RFT, in an appropri-
ate context relational responses may be brought to bear on any stimuli, regardless of their
physical properties.
In the language of RFT, it follows that the relational behavior of early learners must
come under appropriate forms of contextual control beginning early on. This enables
the learner to discriminate between the relevant and the irrelevant features of the task
(for example, responding relationally in accordance with the corresponding contextual
cues, rather than physical stimulus properties). For RFT, the history responsible for this
type of contextually controlled relational responding takes place through experience with
multiple exemplars during early natural language interactions. During these, learners are
taught to respond bidirectionally (in other words, bidirectional stimulus relations are
established between objects and words). For instance, when showing a learner a particular
Chapter 8 151

toy (such as a doll), the parent or teacher will say the word “doll” (object-sound rela-
tion), and later the learner’s responses to the doll (such as orienting or pointing) when
the object is named will be reinforced (sound-object). Similar explicit reinforcement of
object-sound and sound-object relations will occur across a multitude of exemplars of
toys, food, people, objects, and so on. Likewise, parents or teachers may explicitly train
the object-name relations. For example, the teacher may ask the learner, “What is it?”
while pointing to a previously named object and provide reinforcement for any approxi-
mated vocal or nonvocal response. Such a history of MET establishes that in certain con-
texts the training of object-name relations reliably predicts the derivation of name-object
relations (and vice versa). That is, this MET history would result in the emergence of a
type of generalized bidirectional responding that could be applied to any new object and
name, and that is controlled by the presence of specific contextual cues (for example, the
question “What’s this?”).
Of course, the bidirectional relations between words and their referents are character-
ized by sameness (in other words, the word means the same as the object and the object
“is” the word) or equivalence. But with RFT, this is only one type of relational frame and
as such describes only certain patterns of relational responding. For instance, patterns of
equivalence, sameness, or identity are referred to as instances of the relational frame of
coordination. But other types of relational responding have also been identified, includ-
ing opposition, distinction, comparison, and hierarchy, as well as spatial and temporal
perspective taking (see Hayes et al., 2001).
For RFT, all relational frames constitute generalized verbal operant classes that com-
prise the properties of mutual and combinatorial entailment and the transfer or trans-
formation of stimulus functions. What distinguishes one relational frame from another,
then, is the nature of the derived response and the specific transformation of functions.
For example, suppose that a stimulus B is described to a learner as ugly (in other words,
“ugly” functions are directly attached to B). If the learner then learns that A is the same
as B and A is also the same as C (a sameness relation is directly established between A
and B and between A and C), then mutual and combinatorial entailments will be derived
between and among the stimuli (in other words, B and C are the same as A, and B and
C are the same as each other). More important, the functions of ugliness established for
B will now transfer to A and C, and the learner will also perceive them to be ugly, and
indeed equally ugly as A (even though the learner was never told directly that either B or
C was ugly). Alternatively, if the learner learns that A is less than B and B less than C, then
she will derive the mutually entailed “B more than A” and “C more than B” relations and
the combinatorially entailed “C more than A” and “A less than C” relations. Naturally, in
this context, there will be a different pattern of transformed stimulus functions in accor-
dance with this comparative but noncoordinated set of relations, with the result that A
will be the least ugly and C will be the ugliest. It is important to note that, although the
function of ugliness was only attached to B directly, the ugliness attributed to C and A is,
respectively, greater and lesser by virtue of the designated arbitrary relations.
In the sections that follow, we focus explicitly on the training of the basic frames in
a manner that enables appropriate derivation and transformation of functions. From the
RFT point of view, these developments make up the basic foundations of language and
cognition.

General Recommendations for Training
What is central from the perspective of RFT is the careful sequencing of training such
that the target derived relations and transformations of function are captured and maxi-
mized. Directly below is a brief description of the general recommendations for learning
relational discriminations. This is followed by a detailed sequencing of the training most
appropriate for establishing and extending the relational frames of coordination, opposi-
tion, distinction, comparison, and hierarchy.
Knowledge of standard procedures. Because the establishment of relational frames

is basically a type of discriminative learning process, teachers should be initially well
equipped with standard procedures for training simple ��
and conditional discrimina-
tions. These include errorless training, extrastimulus or intrastimulus prompting, fading,
shaping, and schedule thinning (e.g., Grant & Evans, 1994; Lancioni & Smeets, 1986;
Luciano, Barnes-Holmes, & Barnes-Holmes, 2002; Pérez-González, 2001).
Discrimination procedures. The selection of the discrimination procedures should be

based on the particular relational frame to be established and should suit the existing
capabilities of the individual learner. Teaching the learner to respond as both listener and
speaker should be promoted throughout. Training should incorporate a range of presentation
formats. These include matching to sample (MTS), observational or respondent-type train-
ing (Leader, Barnes, & Smeets, 1996), and other stimulus pairing observation procedures
(SPOP; see Smyth, Barnes-Holmes, & Forsyth, 2006; Mañas, Barnes-Holmes, & Luciano,
2006). These should be presented as both tabletop procedures and computer programs.
Varying the training and testing formats provides the learner with new learning contexts,
strengthens derived relational responding, and most likely matches natural language.
Exemplar training. The formation of relational frames involves a process of explicit

MET, which will always be followed by testing for derived outcomes with untrained
stimuli. As is the case with simple discriminated operants, MET for a particular type of
relation, as well as the intentional mixing of trials involving different relations, allows the
learner to develop increasing relational complexity. This also pays learning dividends, seen
in improvements in fluency and flexibility in relational responding. It is important to note
that establishing this type of complexity early on does make for a steep learning curve,
but be assured that it will pay off in the long run.
Using relational cues. The learning process requires that specific words or phrases (such
as “same as,” “contrary to,” “part of,” “more than,” and so on) become the relational
cue or relational context (or Crel) that is to be arbitrarily applied. Naturally, therefore,
it is important to ensure that there is a high level of consistency in the exact words or
phrases used as relational cues across the various training trials and contexts. Otherwise,
it would be difficult for the learner to identify the target relational cue. Of course, when
a relational cue is successfully established across stimulus sets, it is highly recommended
that you extend its meaning and function to other words or phrases, so that the relational
repertoire acquires flexibility, fluency, and breadth. For example, when you are training
coordination relations, start off with the word “is” as the main contextual cue and then
use “same as,” “goes with,” or “match.”
Chapter 8 153

Nonarbitrary relations. For specific words to become relational cues, training should
generally start with nonarbitrary stimuli. These may in fact be physically identical in all
respects if you are trying to establish coordination relations, for example. Or you can
choose stimuli that are identical apart from the fact that one is slightly taller than the
other if you are training a “bigger than” comparative relation. After a while, however,
the target relational cue (for example, using “is” for coordination) will have acquired the
intended functions, and its use with a novel set of arbitrarily related stimuli will transform
the functions of those stimuli accordingly. In this case, you will be able to use the same
word (“is”) as a contextual cue for matching stimuli that bear no formal resemblance to
each other whatsoever. Indeed, the generic benefits of using nonarbitrary stimuli in facili-
tating the transition to tasks with arbitrary stimuli are well established (Barnes-Holmes,
Barnes-Holmes, Smeets, Strand, & Friman, 2004; Berens & Hayes, 2007; Zygmont,
Lazar, Dube, & McIlvane, 1992).
Social reinforcement. Reinforcement for the direct act of relating should be social,
generalized, and provided intermittently as soon as possible after the response, so that
responding relationally becomes reinforcing in itself without additional consequences.
The implications of this for the establishment of sense making as a powerful reinforcer
will be described in the section addressing the frame of coordination.
Generalization. The target relational repertoire should be used in daily life interactions
as soon as possible and across as many opportunities as possible. This type of broad train-
ing enables the coherence of the arbitrary relations to become fully embedded in the
learner’s natural language activities.
Training the Basic Relational Operants

The next sections will provide examples that will enable teachers to train each of the target
relational frames. For the sake of synthesis, aspects such as utilizing MET, varying the
contingencies in training and testing, incorporating appropriate corrective strategies, and
the like will not be frequently articulated because they are procedural variables that are
common to establishing all relational frames and are already well established for teachers
and behavioral trainers.
The Relational Frame of Coordination

The frame of coordination is the most basic relational activity that infants learn in
natural language and the one upon which other relational frames are built (Hayes et
al., 2001). Coordination involves behaving in accordance with relations of sameness,
equivalence, identity, and similarity. Put simply, responding in this way involves arbi-
trarily applying the relational cue “is” or an equivalent (such as “is the same as,” “is
like,” “equals,” “corresponds to,” “goes with,” “means,” or “what is”). Figure 8.1 shows the
characteristics of mutual entailment, combinatorial entailment, and transformations of

f unction associated with the relational frame of coordination. Consider a learner who is
told that “A is the same as B and B the same as C.” (These relations are referred to as arbi-
trary because our three imaginary stimuli have no formal similarities—in other words,
they are not actually the same—but by virtue of language we can arbitrarily apply rela-
tions of coordination among them. If, in fact, we had selected three stimuli that were the
same, we would refer to these relations as nonarbitrary because the relationships among
the stimuli are directly based on physical stimulus properties.) As a result of such an
instruction, the learner will derive the mutually entailed but untrained “B same as A”
and “C same as B” relations, as well as the combinatorially entailed “A same as C” and
“C same as A” relations. Notice that this type of relational responding always comprises
bidirectionality within the relational pairs (A to B and B to A, and so on). Put simply, you
need both directions to compose the relational unit.
Formation of Crel “is” or “goes with”
1. MET testing with nonarbitrary stimuli

A
same
2. Testing or MET testing with arbitrary
stimuli same
B C
3. Promoting fluency and flexibility
(across stimuli, contexts, functions,
training/testing formats, and so on)
Crel application A is like B, and B is like C.
Function given to A in particular A is dangerous.

conditions
Resulting transformation B and C are dangerous.
of functions (Cfunc)
Figure 8.1. The characteristics of mutual entailment, combinatorial entailment, and trans-
formations of function associated with the relational frame of coordination. Full arrows
indicate explicit training. Broken arrows indicate derived (in other words, untrained)
responding.
As well as getting these derived or untrained relations for free, you also get free trans-
formations of function that are dictated by the trained and even derived relations. For
example, if the learner is instructed or trained that A is dangerous (see figure 8.1), he will
behave toward B and C the same way that he behaves toward A (in other words, the aver-
sive functions of A are transferred to B and C by virtue of the trained and derived coordi-
nation relations among the three stimuli). So, the emergence of aversive stimulus functions
for B and C are also free. Experimental evidence for the emergence of bidirectionality and
Chapter 8 155

the transfer of functions via coordination relations has come from a range of different
populations and procedures.
For RFT, a history of reinforcement for bidirectional responding across multiple
exemplars promotes generalized relational responding that can be applied to any stimuli
in the context of an appropriate cue. Hence, according to this view, MET is a criti-
cal component in the learning history that gives rise to the target generalized relational
skill. Other accounts, however, place less emphasis on exemplar training and more on
naming, including Horne and Lowe’s naming theory (1996). Although, as we shall see,
naming provides important opportunities for practicing bidirectionality, the former is
not essential for the latter to occur. In fact, recent research has produced evidence of the
establishment of bidirectionality within coordination using multiple exemplars but in the
absence of naming (Luciano, Gómez-Becerra, and Rodríguez-Valverde, 2007). Hence,
the “Establishing Relational Control Using Arbitrary Stimuli” section below primarily
comprises a multiple-exemplar-based training sequence but also incorporates naming.
Formation of Crel of Coordination with Nonarbitrary Relations

First, train and test identity matching using an MTS format. For example, train the
learner to match two identical stimuli under the control of “give me what goes with this”
(teacher points to target object) or “What is the same as this?” When you have explicitly
trained identity matching to criterion, test on the same set (in other words, remove feed-
back), and then test on a new set. If training is necessary on the second set, test with that
set, and then test again with a third set. In all cases, the learner must produce a sound test
performance (a high pass rate, usually in the region of 80 percent correct, with no more than
one error on each trial type) on a set with which direct training has not occurred. This is
basically what we mean when we use the term multiple-exemplar training (MET). Because
our identity matching here involves two identical stimuli, the target relations are nonarbi-
trary, but we are simply using them here to establish coordination functions for the chosen
contextual cues, which we will need later when the coordination relations are arbitrary.
Establishing Relational Control Using Arbitrary Stimuli

Although naming is not necessary for coordination relations (see above), it is one of the
earliest repertoires natural language establishes. It is also, in itself, a frame of coordination
in which the relations apply arbitrarily between objects and their names, most often under
the control of the contextual cue “is.” Consider the following training scenario in which
the teacher points to an object (such as a banana), the learner looks at it, and the teacher
says, “This is a banana.” The teacher then immediately asks the learner to repeat the name
(“banana”). Next, the teacher asks the learner to “point to [or give me] the banana.”
Now let’s work through this example, looking at what is being trained and what is
being derived. Think of the training scenario as involving three target stimuli: seeing or
pointing to the actual banana (let’s call this A); hearing the teacher say, “This is a banana”
(let’s call this B); and the learner saying the name “banana” (let’s call this C). The initially
trained object-sound relation, in which the teacher points to the banana and says, “This is a
banana,” is a type of explicit A-B training. The teacher’s prompting of the name repetition
is explicit B-C training (sound-saying) and when the learner hears herself say the name,

this is like C-B training (saying-sound). When the teacher says, “Point to the banana,”
this is B-A training when the child points to the object (sound-object). Thus, so far what
we have done is trained A-B, B-C, C-B, and B-A relations, leaving only C-A and A-C
relations to be derived. Now, imagine that, later on, the teacher holds up the banana and
asks, “What’s this?” and the learner says, “banana”; this is a type of A-C training, which in
turn involves C-A training if the learner looks at the object while uttering the name, and
more B-A training. So, you can see how much is actually going on in these simple naming
interactions and why with RFT it is not about naming per se but is about the training and
derivation of the various composite relations, which in this case are all coordinated.
Once all of the target relations are present on a trained set, you must then test with
new objects and names to ensure that the learner can derive all of the necessary relational
responses after appropriate training. When you vary what is explicitly trained, you also
vary what has to be derived, so explore as many combinations of training and testing as
you can. Put simply, what is happening across exemplars is that the contextual cue used
during training (such as “This is” and “What is it?”) is being arbitrarily applied to novel
contexts. In other words, the learner is learning the generalized meaning of these words
and phrases and is learning to use them in a generative and untrained way.
Fluency and flexibility should then be promoted across stimuli to highlight the arbi-
trariness of the coordination relation and to strengthen the coordination functions of the
contextual cue. For example, teach the learner the following: point to a car (B1) upon
seeing a doll (A1); point to a banana (B2) upon seeing an orange (A2); point to a ball (C1)
upon seeing a car (B1); and point to grapes (C2) upon seeing a banana (B2). As a result of
this training, the learner should be able (without further training) to point to a ball (C1)
upon seeing a doll (A1) and vice versa, and to point to grapes (C2) upon seeing the orange
(A2), and vice versa. In summary, you directly train the relations A1-B1, B1-C1, A2-B2,
and B2-C2, and the learner derives the coordinated combinatorially entailed relations
A1-C1, C1-A1, A2-C2, and C2-A2. Of course, we assume that the bidirectional mutu-
ally entailed relations that are the reverse of what was trained (B1-A1, C1-B1, B2-A2, and
C2-B2) are also present.
It is not uncommon for learners to struggle to make the transition from nonarbitrary
to arbitrary responding, even with the most proficient of training histories. When this
occurs, training should proceed with the introduction of appropriate shaping and fading
procedures to facilitate the transfer explicitly. That is, the Crel established with nonarbi-
trary stimuli should be presented with arbitrary stimuli. For example, if you were training
with the cue “is,” you could arrange an MTS task in which the correct comparison was
initially identical to the sample (similar to identity matching). Across trials, the target
comparison is then progressively modified until it is eventually physically dissimilar to
the sample, but its selection is still reinforced as the correct comparison, which is like the
sample. Subsequently,
��
a block of trials containing both arbitrary and nonarbitrary rela-
tions should be presented to ensure that the functions of the cue are readily transferred,
and to facilitate greater flexibility in responding.
Crel Application and the Resulting Transformation of Functions

Once appropriate forms of relational responding come under firm control of the des-
ignated contextual cues (for example, “is” controls coordination), you need to focus on the
Chapter 8 157

transfer of functions through the relations. For instance, suppose a frame of coordination is
established among several pairs of stimuli (in other words, A1-B1, B1-C1, and C1-D1). Let’s
say that these pairs of stimuli are cake (A1) and tomato (B1), tomato (B1) and cookie (C1),
and cookie (C1) and fish (D1). Then, the function “things to eat” is trained for cake and
tomato by permitting the learner to eat them upon correctly responding to “What do you
want to eat?” As a result, the functions of eating (perceptual, manding, and so on) should
also transfer indirectly to cookie and fish (via their mutual and combinatorial entailments
to cake and tomato). However, if you were training the function “things to eat with milk”
to cake, then you should expect this function to transfer only to cookie. The following
example describes how to achieve this type of transformation of functions.
Suppose that a frame of coordination has been established under the cue “apples”: a whole
apple (A1), parts of the apple (B1), a picture of an apple (C1), a picture of parts of the apple
(D1), the name “apple” (E1), and the written word apple (F1). And, suppose that training
involving a cookie (A2 to F2) was conducted in the same way. The learner can be trained to
point to “things to eat,” and reinforcement can be provided only when she points to the whole
or portions of the apple or the whole or portions of the cookie—but not for pointing to any
of the pictures or written words. In subsequent trials, the learner may be asked to point to
pictures. In order to strengthen the control of the various contextual cues, you could train the
names of different actions that take place while the learner is eating the apple or the cookie
(for example, tasting, smelling, noticing texture, and manding to eat) and while she is exposed
to the relevant written words or the pictures (of colors, shapes, and so on). This way, the range
of perceptual functions of tasting the apple, for example, are more likely to be present when
the learner is asked, “How does an apple taste?” than it would be if your training simply
involved instructions such as “Point to the picture of the apple.”
Indeed, you can go even further: while the learner is eating the apple you can say,
“It tastes acidic,” and while she is eating the cookie say, “It tastes sweet,” so that new
contextual cues (acidic versus sweet) facilitate additional transformations of function. It is
immediately apparent that this is the way that natural language develops—in a rich and
cumulative manner involving large arrays of features that interconnect and divide stimuli
and events in an increasingly complex fashion. Although it does seem like a lot of training
just to distinguish apples from cookies, the critical fact is that there is an almost limitless
number of dimensions on which these two items differ from one another. And when the
learner learns to abstract these dimensions using appropriate linguistic cues, they can then
be applied to other stimuli without requiring explicit training. Again, you’ll get a lot of
extra miles for your extra effort and creativity early on.
Expanding the Frame of Coordination and the Emergence

of Sense Making
Generating fluency and flexibility with the frame of coordination offers a power-
ful basis for the emergence of sense making. Put simply, the learner learns what it is to
be “right” and feels good when saying or pointing to something that is coherent with
his particular history of reinforcement. The provision of feedback for correct relating
during training and testing (for example, “Yes, this goes with that,” “Correct,” “Well
done,” “Good boy,” and nodding) and the opposite for incorrect responding (for example,

“Incorrect,” “No, it is not,” “Try again,” and shaking the head) exposes the learner to
social contingencies for responding in a manner that is coherent (or not) with the teach-
er’s providing of the contingencies. As a consequence, teachers will soon see the learner
naturally beginning to nod or shake his head, and vocalize yes or no while responding.
Eventually, the act of relating itself becomes reinforcing, and thus responding to new
situations as a derived relational response also becomes reinforcing. This is evident in
subsequent and more verbally sophisticated responses such as “That’s it,” “That’s what I
think,” “It makes sense to me,” “I agree,” and so on.
In our research we have developed a gamelike testing format that allows you to estab-
lish sense making and to enable the learner to detect when an event is nonsensical—
these appear to form the basis of the earliest forms of humor. Suppose two coordination
networks have been established. One contains pictures of a teddy bear, pajamas, and
sunglasses. The other contains pictures of a banana, an orange, and a birthday cake.
Unexpectedly, the teacher tries to dress the banana in the pajamas (thus mixing up the
networks) and asks, for example, “Does this go with that?” Put simply, things to do
with eating aren’t also to do with dressing up. If the learner fails to respond appropri-
ately, exemplars can be used to establish the discrimination of sense making (determining
coordination) and nonsense making (determining a lack of coordination). Indeed, when
other relational frames have been established, these can also be included to create a broad
array of examples of sense making and nonsense making that extend beyond coordina-
tion relations. In any case, in our experience, learners find “silly” games such as these to
be satisfying and funny.
The Relational Frame of Opposition

Behaving in accordance with the frame of opposition means to arbitrarily apply the
relational cue “is the opposite of” (or the like) to novel stimuli or events. The frame of
opposition requires the abstraction of a particular dimension along which stimuli can
be differentiated. According to RFT, opposition should be taught later than coordina-
tion because the combinatorially entailed relations within opposition are coordinated. For
example, if A is the opposite of B and B is the opposite of C, one assumes that A and C
are the same.
The transformations of function via opposition relations work in the same way as
those through coordination relations, but of course the actual types of transformation are
different. For instance, if a learner is told that “A is the opposite of B” and A has an estab-
lished history as funny, then the learner will behave toward B in a manner that is opposite
to A (in other words, B is boring; see figure 8.2). And if B is the opposite of C, C is funny
because it is also the same as A. Although there is only one study to date that has gen-
erated opposite responding in children who did not have it previously (Barnes-Holmes,
Barnes-Holmes, Smeets, Strand, et al., 2004) there is considerable evidence with adults
(e.g., Whelan & Barnes-Holmes, 2004). In order to establish opposition relations and
appropriate transformations of function, we suggest the sequence of training presented
below.
Chapter 8 159

Formation of Crel “is the opposite of”
1. MET testing with nonarbitrary stimuli

A
2. Testing or MET testing with arbitrary
stimuli opposite
opposite
(across stimuli, contexts, functions, B C
Crel application A is opposite to B,
and B is opposite to C.
Function given to A in particular
conditions A is funny.
Resulting transformation B is boring and C is funny.
of functions (Cfunc)
formations of function associated with the relational frame of opposition.
Establishing the Crel of Opposition with Nonarbitrary Stimuli

To begin, make sure that the learner understands the target stimulus dimension, pref-
erably as both listener and speaker. For example, if the training involves the dimension
of temperature, the learner should be able to select a cold item over a hot item and say,
“cold,” when asked, “How does this feel?” The first step for training opposites is to estab-
lish appropriate responding to the words “yes” and “no” as opposite to one another. For
example, take advantage of the learner’s previously established skills in coordination, and
ask about the presence or absence of something. For instance, in the presence of a doll the
teacher might ask, “Is this a doll?” and prompt yes or no as vocal or nonvocal responses
(for example, in this case, yes is coordinated with presence and no is coordinated with
absence; with this in place, you are already building up yes and no as opposites of one
another).
The training of a contextual cue of opposition should start with the teacher present-
ing two stimuli that are identical and have the same salience on the target dimension but
lie at either extreme (such as a hot potato and a cold potato). Encourage the learner to
touch both potatoes and ask, in relation to each, “Is it cold or hot? Are both potatoes the
same cold or hot?” Continue by saying, for example, “This one is cold, but that one is the
opposite of cold—it is hot.” Then hold up both potatoes and ask the learner to point to
the opposite of the cold potato (listener behavior). You can prompt by allowing the learner
to touch the stimulus. You should also request speaker responses by asking, “This one is
cold, and that one is the opposite, so what’s that one like?” Stimulus touching can also be

conducted as feedback, so that when the learner incorrectly says, “cold,” she subsequently
touches the hot stimulus. You can also ask questions such as “What is the temperature of
this one? And is this one the same or opposite?” So, in a sense you are switching between
speaker and listener, between yes and no, between hot and cold, and between opposite
and same—all at the same time. Of course, each of these components may have to be
built up systematically, but eventually you should be able to randomize the questions and
responses in any way you want to enable the learner to readily discriminate opposition.
To help you train, use MET with stimulus sets that highlight different dimensions
(such as wet versus dry, big versus small, full versus empty, and many versus one). You can
also expand the learning by adding other contextual cues, such as “not same.” Fluency
will also be added when the learner can cross dimensions across trials and combine dimen-
sions. Consider a game in which the learner is asked to respond quickly to a question such
as “What is the opposite of cold? Dry? Wet? Hot? Up? Many?” and so on.
When the learner has achieved fluency with several pairs of stimuli, a third stimulus
can be incorporated to establish combinatorially entailed opposite relations. For example,
having learned that a full glass (A) is opposite to an empty glass (B) and that B is the
opposite of a new glass (C), the learner should then derive that A and C are the same.
If this coordination relation is not derived and has to be explicitly trained, then start off
with A and C being physically the same.
Crel Application and the Resulting Transformation of Functions

Now reconsider the example of the coordination relation between the full glass and
the new glass above. In the previous section, we advised that the new glass be physically
identical to the full glass in order to establish the coordination relation. But, of course,
the relation in this case would be nonarbitrary, and our ultimate targets are arbitrary
relations. So, after the previous training is complete, systematically manipulate the dif-
ferences of the derived coordination relation between A and C because, if the relation
is arbitrary, they do not have to be the same. Remember that the reason we suggested
keeping them the same above was simply to enable you to use the nonarbitrary relation
to get you started.
Consider the following scenario in which the teacher presents two similar boxes and
points to one, saying, “This box has big stickers,” and points to the other box, saying,
“This box is the opposite, so what sort of stickers are in this one?” Then the teacher
asks, “Which sticker do you want?” and “Which box would you choose?” Responding
here should be consequated socially and of course by the physical feedback of getting
the stickers. Then, fade out the physical feedback and introduce new sets of stimuli and
new functions. Again, make sure that you are systematically shifting from nonarbitrary
responding to arbitrary responding (in other words, making the learner choose boxes that
contain hypothetical stickers). Then, introduce a third element to highlight the coordina-
tion relations that can be derived from opposition.
To promote flexibility, alter the contextual cues for opposition and coordination. For
instance, you might say, “X is sweet and is the opposite of Z, but Z is the same as P,” and
then ask, “What is Z like? What is P like? If you do or not want to eat something sweet,
what would you choose?”
Chapter 8 161

The Relational Frame of Distinction
The relational frame of distinction involves responding to the differences among
stimuli, also along a particular dimension, by arbitrarily applying the relational cue “is
different from” (“is dissimilar to,” “is not like,” “differs from,” and the like). However,
unlike in opposition relations, in these relations the relevant dimension is often not
implied. For example, if you are told that “bees are different from mammals,” you cannot
know what bees are like, or exactly which features differentiate them from mammals.
Furthermore, the derived combinatorially entailed difference relations are also unspeci-
fied (see figure 8.3). So, if you are told that bees are different from mammals and that
mammals are different from birds, then you cannot know how bees and birds differ, or
whether they are even the same. No study has yet attempted to establish the relational
frame of distinction in a population previously lacking it, although it has been trained
with individuals who likely already had it (see Dixon & Zlomke, 2005). In the section
below we outline the steps you might use to establish arbitrary relational responding in
accordance with distinction.
Formation of Crel “is different from”
1. MET testing with nonarbitrary

stimuli A
2. Testing or MET testing with different ??
arbitrary stimuli unspecified
different
B C
Crel application A is different from B,
and B is different from C.
Function given to A in particular A is green.
conditions
Resulting transformation B is not green.
of functions (Cfunc) Not enough information to know what B
and C look like.
formations of function associated with the relational frame of distinction.
Establishing the Crel of Distinction with Nonarbitrary Stimuli

The stimuli used to train relations of distinction should share a number of dimen-
sions but differ on a number of others. For example, you might start off by using a set of
plastic cups that are identical in size and shape but different in color (for example, two

green cups and two red). Begin training by asking the learner to point to the green cups
and then point to the red cups. Then, while pointing to the green cups, ask, “Are they
the same color or a different color from each other?” and do the same with the red cups.
Then put a green cup and a red cup together and ask the same question. Once appro-
priate responding is established here, introduce a new set of stimuli. Imagine using two
identical yellow cups, but having one loosely covered with a cloth. In this case, ask ques-
tions (while pointing to the covered cup) such as “Is this one the same color as that one?
Is the covered cup yellow? What color is it? Are they different?” Of course, appropriate
responding can be established by direct training, and you have to systematically vary the
dimensions along which a range of stimuli differ (for example, shape, odor, texture, taste,
and so on). Again, once the skill is established, change dimensions from one trial to the
next to establish fluency.
Then, as before, add a third stimulus to test or train the combinatorial relations. For
example, the teacher could present three similar boxes and say, “The bottle in this box
(A) is red and the bottle in that box (B) is a different color. Is the bottle in here (B) red?
Will it be the same color as that one?” Then expand the relations by saying, “This one
(B) is different from that one (C)” and asking about the B-C and A-C relations (mutual
and combinatorial entailments). Again, vary the training across different sets and stimulus
dimensions and train if you have to, but ensure that you end with a test.
Transferring Crel Control to Arbitrary Stimuli and the

Transformation of Functions
As you would expect, the nonarbitrary training of distinction relations relies upon
physical differences among the target stimuli, although, in dimensions other than the
ones targeted, the stimuli are the same. This similarity is useful because it forces the
learner to focus exclusively on the target dimension, rather than on other dimensions and
thus isolates that feature of distinction over other features of coordination. But, of course,
with arbitrary relations, the differences and similarities among stimuli are arbitrary and
can be unspecified, and there are no physical properties to rely upon. This is one of the
reasons why distinction relations are not as straightforward as they sound. Consider the
following example: Place a ball in a box and an unknown object in an identical box so
that the unknown object remains invisible. Pointing to the first box, say, “In here is a
ball,” and pointing to the second box, say, “In here is a dutnyb.” Of course, “dutnyb”
is just a made-up word to emphasize the arbitrariness of the relation. Then establish key
physical features associated with the ball (for example, round and used for playing) and
say, “Dutnyb is a different shape than the shape of a ball” and “Playing with a dutnyb is
different from playing with a ball.” Then ask, “Does a dutnyb have a different shape than
a ball does? If you wanted to play, which would you take? Are they the same size?”
Then, add a third box with yet another invisible item and say, “In this box you will
find a ‘maoyth’ [another made-up word], which is different from a dutnyb.” For the mutu-
ally entailed relations, ask, “Does a dutnyb have a different shape than a maoyth does?
Are dutnyb and maoyth the same shape?” And for the combinatorially entailed relations,
ask, “Are the ball and maoyth the same shape?” and so on. Again, continue with new sets,
new functions, and new dimensions and mix the distinction and coordination relations.
Chapter 8 163

The Relational Frame of Comparison
Comparative relations involve responding to one event in terms of a quantitative or
qualitative relation along a specified dimension with another event. Next to coordina-
tion relations, comparative relations are perhaps the best understood. For example, there
is a well-established literature on transposition in animals, which is really the same as
responding to nonarbitrary comparative relations (Reese, 1968). But to date, there are
only two RFT studies that have demonstrated arbitrary comparative responding in young
children who did not have these responses in their repertoires prior to the experimental
training (Barnes-Holmes, Barnes-Holmes, Smeets, Strand, et al., 2004; Berens & Hayes,
2007), but several studies involving adults (Vitale, Barnes-Holmes, Barnes-Holmes, &
Campbell, in press).
Establishing responding to nonarbitrary comparative relations is relatively easy
because there are so many comparisons that have nonarbitrary features (for example,
more-less, heavier-lighter, brighter-darker, louder-quieter, bigger-smaller, harder-softer,
rougher-smoother). However, some of these have obvious physical counterparts (bigger-
smaller) while others do not (funnier-sadder). In the same way that opposition and dis-
tinction relations entail coordination relations, comparative relations entail opposition.
For example, consider that the comparison between “heavier” and “lighter” implies that
the two are opposite because one is heavy and the other is light. Hence, in our general
sequencing we suggested that you train opposition prior to training comparison.
Formation of Crel “is more/less than”
1. MET testing with nonarbitrary

stimuli A
> >
2. Testing or MET testing with
< <
arbitrary stimuli >
B C
3. Promoting fluency and flexibility <
Crel application A is more than B,
and B is more than C.
Function given to A in particular A is cold.
conditions
Resulting transformation B is less cold than A, and C is less cold
of functions (Cfunc) than B. Hence, C is less cold than A.
formations of function associated with the relational frame of comparison.

Establishing Crel Control Using Nonarbitrary Relations
Start off with two object piles (of blocks or marbles, for example), where each con-
tains a different amount of objects. Ask the learner, “Which one has more?” and “Which
one has less?” Then systematically alter the ratio of difference between the piles, starting
off with extreme differences and making them consistently more similar. Again, intermix
speaker and listener responding. For example, ask, “Does this pile have more or does it
have less?” Then add a third stimulus to incorporate the combinatorially entailed relations
(for example, A>B>C; see figure 8.4). Comparative games are also easy to create in real-
life situations. For example, consider lining up a group of learners from tallest to shortest
on their way out of the classroom for a break.
Transferring Relational Control to Arbitrary Contexts and

Transformation of Functions
Consider the following learning scenario from research by Barnes-Holmes, Barnes-
Holmes, Smeets, Strand, et al., (2004): Construct a problem-solving task that involves
three identically sized paper coins and allows the teacher to describe how the coins
compare to one another in terms of their value. For example, one coin (A on the left) is
worth more than a second coin (B in the middle), which in turn is worth more than a
third coin (C on the right). The learner is asked to pick the coin that “buys as many sweets
as possible, or buys the most.” During this task, the following relations can be targeted:
A>B>C; A<B<C; C>B>A; and C<B<A. Hence, the teacher is always pointing straight
from left to right or from right to left. If you begin to mix up the pointing, you will come
out with unspecified relations to which a correct answer cannot be provided. For example,
in the sequence A>B and B<C, both A and C are worth more than B but we do not know
whether A and C are the same or different. Thus this relation remains unspecified and
a correct answer involves saying, “I cannot know.” In the context of deficits in language
and/or cognition, always use specified comparative relations and avoid the unspecified.
Research has shown that even normally developing adults have difficulties with unspeci-
fied relations (although they can learn them easily; see Vitale et al., in press).
Comparative tasks such as A>B>C, described above, obviously involve arbitrary rela-
tions because the coins are identical in size. So, if explicit training of these relations
fails, then revert to nonarbitrary training as usual. For example, add different numbers
of sweets on top of the coins to create actual comparisons between more than and less
than, and then intermix nonarbitrary and arbitrary trials. Again, always train and test
across exemplars and mix the stimulus sets and relations. In the research available thus
far, learners have also been exposed to four-coin and five-coin sequences with the same
relations (for example, A>B>C>D>E), and this type of expansion of the comparative rela-
tions is useful for extending relational repertoires. Indeed, a good test can be presented if
you are certain the learner has mastered the relations within a four-coin sequence by then
testing immediately on a five-coin sequence to determine the degree to which the target
relational responding can generalize.
In the value-based task above involving three coins, the choosing of the coin that buys
the most is in fact a transformation of functions; combinatorial entailment is also present
when more than two coins are used. Alternative transformations can also be presented
Chapter 8 165

(for example, “If towel A were drier than towel B and B were drier than C, which would
be the best one to choose if you were wet?”)
Expanding Comparisons That Have No Nonarbitrary Counterparts

Tacting emotions is a classic example of comparative relations that have no nonar-
bitrary counterparts. For example, make a pouting face with eyes squeezed shut, and
ask the learner, “How do I feel?” (You might prompt with “You feel sad.”) To establish
appropriate stimulus control, the learner may learn to name emotions by directly experi-
encing them according to the conditions given by the verbal community. When a learner
understands the physical features associated with various emotions (such as tears with
sadness and smiles with happiness), then relative degrees of emotions can be established
in conjunction with comparative relations. For example, present two pictures of faces,
both with a sad mouth, but with tears in the eyes of one and not the other. Ask, “Which
is sadder or happier?” Indeed, this type of application of comparative relations to emo-
tions is a very important skill but a difficult one to learn; thus it requires that all of the
composite relational and other skills (such as perspective taking) be present.
The Relational Frame of Hierarchy

Responding in accordance with contextual cues such as “contains,” “is an attribute of/
member of/part of,” or “belongs to” without being directly instructed to do so is the main
characteristic of hierarchical relational responding. For example, a learner with such a
repertoire will respond correctly to questions like “Are there are more dogs than animals?”
having been told that there are lots of dogs, lots of birds, and lots of cats. Similarly, if the
teacher told the learner his grandmother’s name and then his brother’s name, the learner
would easily derive the name of the brother’s grandmother. The contextual cues that
organize hierarchical frames may be general or specific. For example, the statement “Dogs
are animals” identifies a fairly broad relationship in which dogs are contained within the
animal group, but we are not made aware of the nature or depth of the relationship. In
contrast, if the teacher told the learner something about the teacher’s great-grandmother,
the learner could derive that this person was part of a group of people exactly three gen-
erations before the teacher. Hierarchical relations are also composed of a range of other
relations because the group members can be organized in a whole manner of ways (see
figure 8.5). For example, the teacher might say, “In our family tree, I am coordinated with
my siblings, comparatively lower than my parents, but higher than my children.” This
feature, therefore, highlights the importance of the learner having strong existing capa-
bilities in the other relational repertoires before commencing training in hierarchy. There
appear to be only two studies to date that have been conducted on hierarchical relations
with adults (Griffee & Dougher, 2002; Gil, Luciano, & Ruíz, 2008).

Formation of Crel “is more/less than”
1. MET testing with nonarbitrary A

stimuli
contains Is member of
2. Testing or MET testing with different
arbitrary stimuli
B C
D E
Crel application A contains B and C,
but B and C are different.
Function given to A in particular
B contains D and E.
conditions
A is food.
B is fruit.
C is seafood.
Resulting transformation D (orange) and E (pear) are food,
of functions (Cfunc) but A is not just D and E.
All types of food (A) can be eaten,
so E can be eaten.
If all Ds are juicy, not all As are,
and B is not only juicy. As well,
do not know what C looks like, etc.
formations of function associated with hierarchical relations.
Recommendations for Training

A number of basic recommendations should be followed when attempting to establish
responding in accordance with hierarchical relations:
1. Because hierarchical relations inherently involve a high degree of flexibility

in contextual control, a diverse array of cues should be established from the
outset.
2. The nonarbitrary relational features of hierarchy are so closely intertwined with

socially ascribed (arbitrary) features that the two are difficult to separate. For
example, “My grandmother is comparatively older than me [and there are many
Chapter 8 167

nonarbitrary features to support this comparison] but she is arbitrarily referred
to as my grandmother only because she is my father’s mother.”
3. Starting off with nonarbitrary relations is particularly important here, because

the arbitrary relations are so complex. This will likely mean having to contrive
odd or unusual nonarbitrary categories that allow appropriate forms of contex-
tual control to be readily established.
4. Training hierarchy is typically done from the top down. For example, you
might first establish the category of animals and then train its various members.
But it is very important that the learner readily captures the bidirectionality of
these relations, because the relations contained within are not coordinated. For
example, in top-down training, “My grandmother is older than me,” but in
bottom-up training, “I am younger than her.” Consider a learner who is trained
to point to a bird, a cat, a dog, an elephant, and so on when asked to point to
the “animals,” and to point to car, airplane, truck, boat, and so on as “vehicles.”
Then ask, for example, “Is a bird an animal or a vehicle?”
5. In training, the relational cues for hierarchy are practically inseparable from
Cfuncs. For example, if an item is a member of a food category, by definition
it is something that you can eat. Although this means that transformations of
functions generally do not form a distinct part of training, it does mean that
the derived transformations should be adequately tested.
Training Categories
The real task in teaching hierarchy is to bring already established relational rep-
ertoires under slightly different, but highly specific, forms of contextual control. For
instance, “fruits” and “seafood” may have been related by distinction, yet when “food”
is the contextual cue, they are coordinated (see figure 8.5). For this reason, begin train-
ing by organizing members into common categories. For example, under the cue “food
or things to eat,” establish a number of coordinated members (such as pear, salmon,
pork, and the like), so each member is coordinated with the other and all are contained
within food. Now establish another frame of coordination (such as toys) and a frame of
distinction between food and toys. This is a good pair of examples to start with because
there are practically no instances in which the distinction breaks down (in other words,
you never eat toys and you shouldn’t play with your food). You may even establish the
relational networks by physically placing the two stimulus sets in different cabinets and
physically organize (in other words, coordinate and distinguish) members accordingly.
Then, you could start to create distinction relations within each category. For example,
for foods, put all of the fruit on one shelf and all of the seafood on another, but both still
in the same food cabinet (physical inclusion). Again, make sure that the bidirectionality
between members and between the member and its category is intact. Then establish
combinatorial entailment by introducing a new item (such as a blueberry) and instructing
that a blueberry is “in the same group as a pear.” Then ask, for example, “Which cabinet
does the blueberry belong in? Which shelf should it go on? Can you eat it? Is it more like
a fish or a grapefruit?” If the learner has already learned a number of dimensions of food

tastes (such as sweetness versus sourness), ask her to place the sweet fruits together at one
end and the sour fruits together at the other end of the relevant shelf. It is important that
you subdivide the category members in as many ways as you can think of and encour-
age all of the relevant patterns of derivation and transformations of functions. Also, keep
building out existing frames (for example, “If an apple is sweet and cod is not sweet,
what will the salmon taste like?”). There are literally hundreds of relations and games that
could be constructed when comparing a closet full of food to a closet full of toys. For a
more abstract type of category, such as family trees, you can start with pictures of people
the learner actually knows and build up the network by adding sketches or drawings of
unknown relatives.
Transferring the Relational Control to Arbitrary Contexts

Fading out the physical stimuli (such as the cabinet or the drawings) slowly as train-
ing progresses will be essential, and it will help focus the learner’s attention more closely
on the spoken contextual cues rather than on physical stimulus properties that may have
exerted some influence over responding thus far. One step toward their removal involves
using pictures instead of objects and then fading out the pictures to just words. Be careful
not to fade out the nonarbitrary relations too quickly here or your proposed relational
network will collapse. The almost endless nature of the overlap and distinctions that
constitute our understanding of categories means that you simply have to be creative and
resourceful in your training and try to present every possible relation within a network.
Try to use as many categories as you can from real life, such as cooking implements,
clothes, bedding, games, school items, and so on. Make sure too that you test as many
derived relations and transformations of functions as you can. To do so properly, you may
need to physically draw out the target relational networks for yourself and work through
the component trained and derived relations. When working with sophisticated learners,
you may even present drawings of partial networks and ask them to complete the missing
information. These developments will not only strengthen the target relational networks,
but will also strengthen other existing relational repertoires and create fluency and flex-
ibility within the hierarchies.
Concluding Comments
In the current chapter, we have offered specific guidelines for establishing relational
responding in accordance with contextual cues that control the relations of coordina-
tion, opposition, distinction, comparison, and hierarchy. The training sequences proposed
here have not been tested experimentally, but they are driven by a broad array of empiri-
cal research. For example, existing evidence shows that coordination is the first frame
established in natural language, followed by opposition, comparison, and distinction. As
perhaps the most complex of the frames reviewed here, hierarchical responding requires
much flexibility and fluency in the other frames if it is to facilitate developments in critical
perspective taking.
In our descriptions of the explicit training of the relational operants here, we
strongly emphasized the cautious but systematic transition from nonarbitrary to arbitrary
Chapter 8 169

r esponding and multiple-exemplar training involving testing with new stimuli as key
features of relational training. Again, these guidelines were based on existing empirical
evidence, as well as on some predictions made using RFT that have yet to be empirically
tested. Nonetheless, we believe that the information provided here will help to fill gaps
in existing remediation programs at multiple levels. Indeed, the key aim of the type of
training suggested here is to establish generativity—the holy grail of human language and
cognitive development.
References
Barnes-Holmes, Y., Barnes-Holmes, D., & Smeets, P. M. (2004). Establishing relational
responding in accordance with opposite as generalized operant behavior in young
children. International Journal of Psychology and Psychological Therapy, 4, 559–586.
Barnes-Holmes, Y., Barnes-Holmes, D., Smeets, P. M., Strand, P., & Friman, P. (2004).
Establishing relational responding in accordance with more-than and less-than as
generalized operant behavior in young children. International Journal of Psychology
and Psychological Therapy, 4, 531–558.
Berens, N. M., & Hayes, S. C. (2007). Arbitrarily applicable comparative relations:
Experimental evidence for a relational operant. Journal of Applied Behavior Analysis,
40, 45–71.
Carr, J. E., & Firth, A. M. (2005). The verbal behavior approach to early and intensive
behavioral intervention for autism: A call for additional empirical support. Journal of
Early and Intensive Behavior Intervention, 2, 18–27.
Dixon, M. R., & Zlomke, K. M. (2005). Implementación del precursor del procedimiento
de evaluación relacional en el establecimiento de marcos relacionales de igualdad,
oposición y diferencia [Using the precursor to the relational evaluation procedure
(PREP) to establish the relational frames of sameness, opposition and distinction].
Revista Latinoamericana de Psicología, 37, 305–316.
Gil, E., Luciano, M. C., & Ruíz, F. (2008). Transformación de funciones vía el contexto
relacional de jerarquía [Transformation of functions via the relational context of hier-
archy]. Master’s thesis, University of Almería, Spain.
Grant, L., & Evans, A. (1994). Principles of behavior analysis. New York: Harper-Collins
College Publisher.
Greer, R. D. (2002). Designing teaching strategies: An applied behavior analysis systems
approach. San Diego, CA: Academic Press.
Griffee, K., & Dougher, M. J. (2002). Contextual control of stimulus generalization
and stimulus equivalence in hierarchical categorization. Journal of the Experimental
Hayes, S. C., Barnes-Holmes, D., & Roche, B. (2001). Relational frame theory. A post-
Plenum.

Lancioni, G. E., & Smeets, P. M. (1986). Procedures and parameters of errorless discrimi-
nation training with developmentally impaired individuals. International Review of
Research in Mental Retardation, 14, 135–164.
Leader, G., Barnes, D., & Smeets, P. M. (1996). Establishing equivalence relations using
a respondent-type training procedure. Psychological Record, 46, 685–706.
Lovaas, O. I. (1977). The autistic child: Language development through behavior modifica-
tion. New York: Irvington.
Luciano, M. C. (1984). Implantación de conducta verbal en niños no verbales [Establishing
verbal behavior in nonverbal children]. Revista Española de Pedagogía, 164, 257–282.
Luciano, M. C., Barnes-Holmes, Y., & Barnes-Holmes, D. (2002). Establishing reports of
saying and doing and discriminations of say-do relations. Research in Developmental
Luciano, M. C., Gómez-Becerra, I., & Rodríguez-Valverde, M. (2007). The role of
multiple-exemplar training and naming in establishing derived equivalence in an
infant. Journal of Experimental Analysis of Behavior, 87, 349–365.
Mañas, I., Barnes-Holmes, D., & Luciano, M. C. (2006). Comparing two methods for
the formation of equivalence relations: Stimuli animation by movement and matching to
sample. Paper presented at the second world conference on ACT, RFT, and Contextual
Behavioural Science, London, United Kingdom.
McEachin, J. J., Smith, T., & Lovaas, O. I. (1993). Long-term outcome for children with
autism who received early intensive behavioral treatment. American Journal on Mental
Retardation, 97, 359–372.
Pérez-González, L. A. (2001). Procesos de aprendizaje de discriminaciones condicionales
[Processes of learning conditional discriminations]. Psicothema, 13, 650–658.
Reese, H. W. (1968). The perception of stimulus relations: Discrimination learning and
transposition. New York: Academic Press.
Selinske, J., Greer, R. D., & Lodhi, S. (1991). A functional analysis of the comprenhen-
sive application of behavior analysis to schooling. Journal of Applied Behavior Analysis,
34, 108–118.
Sidman, M. (1971). Reading and auditory‑visual equivalences. Journal of Speech and
Smyth, S., Barnes-Holmes, D., & Forsyth. J. P. (2006). A derived transfer of simple
discrimination and self-reported arousal functions in spider fearful and non-spider
fearful participants. Journal of the Experimental Analysis of Behavior, 85, 223–246.
Sundberg, M. L., & Michael, J. (2001). The benefits of Skinner’s analysis of verbal behav-
ior for children with autism. Behavior Modification, 25, 689–724.
Chapter 8 171

Vitale, A., Barnes-Holmes, Y., Barnes-Holmes, D., & Campbell, C. (in press). Facilitating
responding in accordance with the relational frame of comparison: Systematic empiri-
cal analyses. Psychological Record.
Whelan, R., & Barnes-Holmes, D. (2004). The transformation of consequential func-
tions in accordance with the relational frames of same and opposite. Journal of the
Zygmont, D. M., Lazar, R. M., Dube, W. V., & McIlvane, W. J. (1992). Teaching arbi-
trary matching via sample stimulus-control shaping to young children and mentally
retarded individuals: A methodological note. Journal of the Experimental Analysis of
Behavior, 57, 109–117.

Chapter 9
Applying Relational Operants

to Reading and Spelling
Deisy G. de Souza, Julio C. de Rose, and Camila Domeniconi,

Universidade Federal de São Carlos, Brazil
This chapter outlines a behavioral analysis of reading and spelling. Particular emphasis
is given to the contribution of the research on derived relational responding (or stimulus
equivalence) for the conceptualization and teaching of reading and spelling repertoires to
individuals with developmental disorders. The chapter then presents and discusses practi-
cal questions about teaching the conditional relations and generating the derived relations
involved in reading and spelling. The chapter also discusses how this training can be
programmed so that students can learn to read novel words through recombination of
textual units.
Reading as a Complex Skill with

Multiple Components
Behavior analysts conceive of reading as a special kind of operant behavior. As is the
case with all operant behavior, responses are emitted under the control of discriminative
stimuli and are strengthened and maintained by reinforcement. The discriminative stimuli
for reading are letters, words, and sentences—what we generally call “text.” A person who
reads aloud speaks a series of words that correspond to the words in print. The behavior is
presumably similar when one reads silently, although no one else can hear.
Behavior analysts have distinguished between two components of reading (Skinner,
1957; Staats, 1968): textual behavior and reading comprehension. Textual behavior,
defined as verbal responses under precise, point-by-point control by the text (Skinner,
1957), is often called decoding or word recognition. The text is usually visual (although it
is tactile in Braille). Most people respond vocally to text, but one may respond with signs
as well. The central aspect of textual behavior is the discriminative relation between text
and the verbal response, which share a point-by-point correspondence (Skinner, 1957).
For example, the printed word cat controls the corresponding response, which is differ-
ent from those occasioned by the printed words bat, car, and cart. If a learner gives the
vocal response “cat” to the printed word cat but also to car or bat, his reading is faulty.
The second component of reading is the comprehension of what is read. The authors of
this chapter, as native speakers of Portuguese, can produce textual behavior in response
to a text in Latin because the correspondences between letters and phonemes are similar
in both languages. Therefore, finding a sentence like “Sum venti ingenium breve flos sum,
scilicet, aeris…,”1 we can say the words quite well (that is, we can emit textual behavior),
but that would hardly qualify as reading, because we don’t have the slightest idea of what
the words mean without a translation. Similarly, we would not understand the sentence if
someone else spoke it to us. Therefore, if we say the sentence as textual behavior, we will
not understand it. Staats (1968) tells the story of an intelligence team in Europe during
World War II who intercepted a message written in Russian and were able to understand
it only because one member of the team could respond to the text by pronouncing the
words (but could not understand them) and another could understand the meaning of
spoken words in Russian (but had no textual behavior). The combined behavior consti-
tuted reading. The example is important to demonstrate the two functions involved in
reading; because both functions are usually performed by one person, the distinction
tends to be overlooked.
Therefore, in order to teach reading, we need to establish both components.2 The
student needs to say the words and sentences corresponding to the text and needs to
understand them. The rich dynamics of spontaneous speech are often missing in reading,
even when the reader is already fluent. Beginning readers in particular tend to read word
by word, without the pauses and intonation cues of common speech. This dynamic is
essential for listening comprehension. Responding to a text with dynamics is also a matter
of discrimination learning, but the stimuli to be discriminated are even more subtle than
those involved in word deciphering: they consist of punctuation marks and the seman-
tic and syntactical cues provided by the text. Some researchers and teachers argue that
reading instruction should emphasize comprehension from the beginning and that, conse-
quently, the focus should be on the semantic and syntactical cues that provide the context
from which meaning is extracted, rather than on the discrimination of words themselves.
Although most current researchers point out that there is no opposition between word
1 From the Latin poem “Bulla” (“The Bubble”), by Richard Crashaw (c. 1613–49). An English
translation by Phyllis S. Bowman reads, “I am the brief nature of the wind. To be sure, I am the
flower of the air.”
2 Most reading researchers maintain a cognitive analysis, rather than the behavioral one
expounded in this chapter. A cognitive analysis focuses on mental processes that supposedly
underlie reading. Recent cognitive analyses of reading parallel the behavioral analysis in this
respect. This perspective would consider decoding as the production of the words that correspond
to the text, which must then be understood in a manner that is similar to listening comprehension
(Adams, 1998; Snow, Griffin, & Burns, 2005).

recognition and comprehension, they agree that without word recognition there will be
no comprehension at all (Adams, 1998). For this reason, this chapter will focus on the
discrimination and understanding of words.
In a behavioral analysis, reading and writing are also conceived as different behav-
iors. In reading, the text is the discriminative stimulus for the production of words and
sentences as vocal responses (or sign language). In writing, the text is the product of
responding. The discriminative stimuli may be also textual, as in copying. However,
in order to write with nontextual discriminative stimuli, such as in dictation taking,
where the stimuli are spoken words, one needs to know how to spell the words. Although
reading and spelling typically occur together in the repertoire of a proficient reader, they
often dissociate in beginning readers (Lee & Pegler, 1982) or readers with developmental
disabilities. Nevertheless, under some conditions, which will be discussed later in this
chapter, these behaviors may interact, so that teaching one may lead to the acquisition of
the other.
The interactions between reading and spelling repertoires have been clarified by the
research program on stimulus equivalence (Sidman, 1971, 1994; Sidman & Tailby, 1982),
or more generally, derived relational responding (Barnes-Holmes, Barnes-Holmes, &
Cullinan, 2000; Hayes, 1994). This research has contributed methods that have been
shown to be effective in teaching reading and spelling to both children with typical devel-
opment and children with autism and other developmental disabilities. An important
feature of the stimulus equivalence approach is that it may establish and simultaneously
strengthen the two critical components of reading: word recognition and comprehen-
sion. Another extremely important feature is that teaching a set of behaviors results in
the emergence of other behaviors that are established by derivation, without the need for
direct training.
Matching-to-Sample Procedures and Stimulus

Equivalence to Teach Reading
Matching-to-sample procedures have long been used to teach reading.3 An example is a
program reported by Ribes (1978), which was designed to teach reading in Spanish to
students with intellectual disabilities. The first part of the program taught identity match-
ing of printed words in which students were given a particular printed word as a sample
stimulus and asked to select an identical printed word among comparison stimuli. As they
learned to do this, students also learned to say the word aloud, or emit textual behavior.
In identity matching trials, the instructor presented the printed word to a child and also
said the word; he then asked the child to repeat the word (echoic behavior) and then select
the identical printed word and say its name once again. The correct comparison stimulus
was identical to the sample in two dimensions (color and topography). Along successive
trials the color was faded out, so that at the end selections were based completely on the
3 The behavior analytic literature reports several procedures that have been used to teach
directly textual behavior and other components of reading (e.g., Engelmann & Carnine, 1982).
We selected examples of the matching-to-sample procedure due to its central role in the stimulus
equivalence approach.
Chapter 9 175

identity of the sample and comparison printed words. In this program textual behavior
was taught directly, as the student repeatedly said a word in the presence of the corre-
sponding printed word and not in the presence of other words. Each printed word thus
became a discriminative stimulus for the student to say that word and not other words.
This and other teaching programs based on matching to sample have been quite effec-
tive in teaching reading to persons with developmental disabilities. However, in the early
1970s Murray Sidman discovered that matching procedures might have an even more
valuable outcome: they can generate behaviors that are not explicitly taught and, in so
doing, increase the potential of learning far beyond that which is directly taught (Sidman,
1971, 1994). Sidman first observed this potential to generate untaught behaviors in a
study conducted with a young person with severe intellectual disabilities (Sidman, 1971),
which we describe at length here because many of its features are relevant for effective
teaching. Sidman used twenty words with one syllable, each with three letters: axe, bed,
bee, box, boy, bug, car, cat, cow, dog, ear, hat, hen, hoe, hut, man, pie, pig, saw, and zoo.
The basic procedure, matching to sample, was conducted in an apparatus that had nine
translucent windows arranged in a three-by-three matrix. Samples could be either visual
or auditory. A visual sample was either a picture or a printed word, presented continuously
at the center window. An auditory sample was a word dictated through a loudspeaker
and repeated every two seconds. A touch to the center window brought about eight com-
parison stimuli (printed words or pictures), each appearing on one of the eight peripheral
windows. Only one of the comparison stimuli corresponded to the sample, and the stu-
dent’s task was to select it. Correct selections were reinforced with chimes ringing and the
delivery of a candy and a penny; incorrect selections had no scheduled consequences and
the display remained unchanged until the student made a correct selection.
Sidman first conducted a series of pre-tests, with no consequences for correct responses,
to ensure that the student indeed could not read before the study began. He tested some
matching performances, or, as Sidman and Tailby (1982) identified them later, conditional
discriminations. As shown in figure 9.1, these included the student selecting a printed word
that corresponded to a dictated sample (a performance Sidman called auditory-receptive
reading), selecting a picture that corresponded to a printed word sample, and selecting a
printed word that corresponded to a picture sample. As expected, the student performed
very poorly in the pre-tests involving printed words. He scored highly, however, when the
samples were dictated words and the comparison stimuli were pictures. He could, there-
fore, select the picture that corresponded to each of these twenty words. The youth also
did well when each picture was presented alone and he was asked to name it. However,
he did very poorly, as expected, when each printed word was presented alone and he was
asked to say the word.
Sidman represented these conditional discrimination performances using a triangle
that has become famous as the equivalence triangle (so famous that you probably have seen
it already in other chapters in this volume). In the ABC triangle of figure 9.2 (top), arrows
represent the conditional discriminations. Each arrow points from the set of samples to
the set of comparison stimuli. There are, therefore, two arrows pointing from the dictated
word samples. One of them points to the pictures and the other one points to the printed
words. There are also arrows pointing from the printed words to the pictures and from
the pictures to the printed words. However, there are no arrows pointing to the dictated
words (visual stimuli can be presented simultaneously as choice stimuli but this cannot be
done with auditory stimuli). Dictated words, hence, were presented only as samples and

I II
bed dog
man hut boy
bee cat
dog car hat
III IV
man hut boy
cat bee cat
dog car hat
Figure 9.1. The three-by-three matrix arrangement of stimuli in Sidman’s (1971) study.
Examples of training and test trials. I: Matching pictures to dictated words. II: Matching
printed words to dictated words. III: Matching pictures to printed words. IV: Matching
printed words to pictures.
the arrows point only away from them. The broken lines represent conditional discrimina-
tions that were not initially in the student’s repertoire, as shown by the arrows connecting
the printed words to the pictures and vice versa. The heavier line represents a conditional
discrimination that was directly taught and the thinner line represents one that was already
learned before the study, just to emphasize what was actually taught. Thus, the triangle
of figure 9.2 can be quickly interpreted as representing a study in which the conditional
discrimination AC was taught but AB had been previously learned. The broken arrows
also mean that the researcher or teacher was interested in what would happen to these
conditional discriminations, BC and CB, after AC was taught. Would these conditional
Chapter 9 177

discriminations emerge without being directly taught, derived from the relations taught
in the study and from those learned by the participant before the experiment?
The ABC triangle, however, does not represent all performances that were evaluated.
As documented by the pre-tests, the student had already learned to name the pictures,
represented by the arrow pointing from B (pictures) to D (naming). The student was not
able to read the printed words or, more precisely, to emit textual behavior to them, as rep-
resented by the broken arrow from C to D (naming the printed words). While the ABC
triangle represents matching tasks and stimulus-stimulus relations, the BCD triangle rep-
resents stimulus-response relations (discriminated operants).
If we are familiar with the diagram in figure 9.2, we know, therefore, that this study
taught the AC relations and verified the effect of this training on matching printed words
and pictures and on reading the printed words. For those interested in teaching students
with intellectual disabilities to read, however, the question remains: How do we establish
this behavior with a set of only twenty words?
Sidman (1971) started the teaching procedure with only two sample-choice combi-
nations (for example, “car”-car or “bed”-bed, where the first word indicates the auditory
sample word and the second represents the printed word presented in one of the eight
outer windows). The two samples alternated in an unsystematic order, but on each trial
all of the eight windows presented printed words as comparisons (one correct and seven
incorrect); the window with the correct choice varied across trials. The two combinations
were repeated until the student’s first choices on both were correct (a mastery criterion).
Then a third relation was added (for example, “cat”-cat). When the boy’s first choice on
each of the three pairs was correct, a fourth combination was added, and so on, until the
first choices were correct on the full set of twenty trials.
After the student mastered the performances that had been directly taught, Sidman
could verify whether the other performances would emerge. They did. When the printed
words were presented as samples, the student now selected the corresponding pictures, and
he also selected the appropriate printed word when the pictures were presented as samples.
Also, when each printed word was presented alone, he could say its name (textual behav-
ior). Sidman concluded that dictated words, pictures, and printed words had become
equivalent stimuli and, therefore, the youth could read this set of twenty words with
comprehension. The resulting repertoire is represented in the bottom panel of figure 9.2.
Sidman and colleagues replicated this work with two other youths with severe intel-
lectual disabilities (Sidman & Cresson, 1973). The two participants described in the 1973
paper were even more severely limited than the participant in the first study was and
both had to be taught identity matching with the printed words and to match pictures
to dictated words (the AB relation in figure 9.2) before they could learn to match printed
words to dictated words (the AC relation). At the end of the study, however, both of them
were reading with comprehension.
A notable feature of these studies is the fact that individuals with moderate or severe
intellectual disabilities acquired the ability to read with comprehension. The teaching pro-
cedure assured both comprehension and textual behavior. However, Sidman taught neither
textual behavior nor comprehension directly. Both emerged from matching-to-sample
teaching. The students learned directly only two matching relations: AB and AC. These
two sets of matching relations brought about two sets of untaught matching relations:
matching picture samples to printed words, and matching printed word samples to pictures.
Moreover, teaching the student to match dictated names to printed words produced the

emergence of textual behavior: the students were then able to say the name of the printed
words (the CD relation in figure 9.2). These students with severe disabilities had, there-
fore, learned to read, even if their reading vocabulary comprised only twenty words. They
could say the name of each printed word and they could understand the word’s meaning
by matching the printed word and its referent. This matching was not directly conditioned
but emerged from the equivalence relations formed between dictated words, printed words,
and pictures. ��
In an important reconceptualization of this work, Sidman and his cowork-
ers proposed that stimulus equivalence is the basis of symbolic behavior (Sidman, 1994;
Sidman & Tailby, 1982). Classes comprising printed words, dictated words, and objects
or pictures permit us to infer that the printed words are symbols for the objects, and that
students read the words with comprehension or true understanding.
Expanding the Reading Vocabulary

The reading vocabulary of the students in Sidman’s early studies was limited to the
twenty words directly taught. Although this was a huge accomplishment for these stu-
dents with severe developmental disabilities, how could a more capable student learn to
read a larger number of words, and especially read words not previously taught, as is
typical of a proficient reader?
Skinner (1957) suggested that as students acquire a larger vocabulary of words, stimu-
lus control by textual units smaller than words might develop. This was investigated in
a study by de Rose, de Souza, and Hanna (1996). They used a method similar to that
used by Sidman (1971) to teach children without intellectual disabilities who were doing
poorly in school. The students learned to match fifty-one words to the corresponding
dictated words along a series of teaching units presented in a tabletop format. The first
unit taught students to match three printed words to the respective dictated words. In
the initial teaching trials, a word was dictated as the sample, and the comparison array
presented only the correct comparison stimulus; samples alternated across trials. After
a few trials with the correct comparison only, the next trials presented one incorrect
comparison along with the correct one. The printed word selected in a matching trial
remained visible, and the student had to “copy” it with movable letters. This task is called
constructed-response matching to sample, because the student has to construct a stimulus
that matches the sample (Dube, McDonald, McIlvane, & Mackay, 1991; see also chapter
10 of this volume). Subsequent teaching units used the words taught in the first unit
as a baseline for teaching new words by exclusion. Matching by exclusion is a variety
of the matching-to-sample procedure that has been shown to be particularly useful in
teaching new sample-comparison relations (Dixon, 1977; McIlvane & Stoddard, 1981).
In the reading program an exclusion trial presented two printed words as comparison
stimuli (see figure 9.3 for an example with words used by Sidman [1971]). One of them
was a defined comparison stimulus—that is, a printed word that the student had already
learned to relate to a dictated word. The other was an undefined comparison stimulus,
meaning that the student did not yet know its relation to the corresponding sample.
When the new sample word was dictated, the student could detect the mismatch with the
word she had already learned and thus exclude it, selecting the undefined printed word.
At the end of each unit, a post-test verified whether the student could read the words;
Chapter 9 179

B
PICTURES
A D
DICTATED ORAL NAMING
WORDS BD: picture naming
CD: word naming or
textual behavior
C
PRINTED
WORDS
PICTURES
A D
DICTATED ORAL NAMING
BD: picture naming
WORDS CD: word naming or
textual behavior
C
PRINTED
WORDS
Figure 9.2. Schematic representation of the relations taught and tested by Sidman (1971).
Top: The ABC triangle represents stimulus-stimulus relations; the BCD triangle repre-
sents stimulus-response (operant) relations. Bottom: After equivalence class formation,
members of the class are interchangeable in the control of responding. Consequences (not
represented) maintain the discriminative relation.

if not, the block of training trials was repeated until, in the post-test, the student could
read all of the words taught in the unit. Periodic tests assessed whether students could
match the printed words to the corresponding pictures and vice versa, thus assessing the
formation of equivalence classes.
Exclusion Procedure
Baseline words: car, hut, bee
Novel, training words: bug, pie
Exclusion trial Control trial

bug
car
bug car
car pie
Baseline trial
bee
hut bee
Figure 9.3. Trial types in the exclusion procedure: Exclusion and control trials both
display a baseline and a novel printed word as comparison stimuli. On exclusion trials a
novel word is dictated, and on control trials a baseline word is dictated. On baseline trials
sample and comparisons are baseline words (that is, words previously learned).
At several points in their teaching program, de Rose and colleagues (1996) verified
whether students could read untrained words that were formed by a recombination of
syllables from the trained words. In Portuguese, syllables are repeatable speech units,
pronounced in a relatively similar way in different words. Thus, a two-syllable word such
as bolo (cake) is pronounced with the sound “bo,” followed by the sound “lo,” whereas
vaca (cow) is pronounced “vah-cah.” When faced with an untrained word such as boca
(mouth), could the student assemble this word, pronouncing the sound “bo,” correspond-
ing to the first syllable, followed by the sound “cah,” corresponding to the second syllable?
Chapter 9 181

Five out of seven participants eventually showed this recombination of textual units. The
point in the program at which students began to show recombinative reading varied for
different children. Children also varied in the number of untrained words they read in
two extended tests that were conducted at the middle and at the end of the program.
Scores varied from 20 percent to 86.7 percent at the end of the program (the average
was 40 percent). Students showed also improvement in spelling of the trained and novel
words, both with movable letters and with paper and pencil, although spelling scores were
lower than reading scores.
Although directly teaching a relatively large set of whole words enabled some students
to read untrained words by recombinative generalization, this generalization was much
larger in a new version of the program in which students were directly taught to read the
syllables. After learning a set of words, students learned to match the dictated syllables
of these words to the corresponding printed syllables (see figure 9.4). Each syllable was
dictated as the sample in two matching trials (in a randomized sequence of samples),
with three printed syllables displayed as comparison stimuli. This was repeated until the
student scored 100 percent correct. All twenty students who participated in this research
showed generalized reading, and their scores were much higher than those in the original
study of de Rose and colleagues (1996).
As Sidman (1994) pointed out, while
a whole-word method of teaching reading leaves to chance a pupil’s recognition

of the correspondences between sounds and textual elements [it also assures that]
a pupil can rapidly acquire a preliminary but substantial reading comprehension
vocabulary…that can be used as the basis for teaching correspondences between
the components of sounds and the components of [printed] words that have
already become meaningful to the pupil. (pp. 78–79)
An advantage of the word as a teaching unit is that a typically developing learner

has already acquired spoken language, so she is able to say the words and react to them
as a listener (Greer, Chavez-Brown, Nirgudkar, Stolfi, & Rivera-Valdes, 2005). She can,
therefore, understand the meaning of the words. However, there are many thousands of
words in a language and it is not feasible to directly teach a person to read all of them.
Alphabetic writing systems contribute substantially to solving this problem: a reduced
number of individual letters and letter combinations can be used to write and read any
word in the language (Adams, 1998; Daniels & Bright, 1996). The alphabet does not rep-
resent the words themselves, but rather their constituting sounds, or phonemes, making
it possible for the reader to sound out the word by producing the sounds of the indi-
vidual units and blending them. Thus, readers will exhibit appropriate textual behavior
in response to words such as dereliction or rambunctiousness even when they find these
words for the first time. This is done by assembling responses to the letters and groups of
letters that form the word. It would be possible, therefore, to teach a child to read these
small units, and this would supposedly enable her to read all the words in the language.
This option has at least two major difficulties, however: One is that correspondences
between letters and sounds are far from regular (and even more irregular in English than
in Portuguese), and critical aspects of speech, such as intonation and prosody, are not
easily represented in print. A second difficulty is that children often tend to respond to
words as wholes and do not discriminate the sounds that constitute the words.

B
PICTURES
D
ORAL NAMING
A BD: picture naming
DICTATED CD: word naming
WORDS
A′
DICTATED
SYLLABLES E
SPELLING
C AE: dictation
CE: copy
PRINTED
WORDS
C′
PRINTED
SYLLABLES
Figure 9.4. Schematic representation of the relations taught and tested in the current
version of the teaching program designed by de Rose and colleagues. New training rela-
tions (A'C' and CE) were added to the basic relations in figure 9.2. Dictation taking
probes were also introduced with two response modes: constructed response (AE) and
handwriting (AF, not shown). Solid arrows indicate explicit training. Dotted arrows indi-
cate derived, or untrained, responding.
Researchers have called this ability to recognize the sounds that constitute the words
as repeatable units phonological awareness (in behavioral terms, the ability involves dis-
crimination and abstraction of within-word sound units; cf. Mueller, Olmi, & Saunders,
2000). Usually, children recognize that words can be broken into syllables before they
recognize that syllables can be broken into phonemes (Adams, 1998), and they recognize
onset and rime (respectively, the initial consonant sound, and the vowel and subsequent
consonants in a syllable) units within syllables earlier than they recognize individual pho-
nemes within syllables (Goswami & Bryant, 1990). The relationship between phonological
awareness and reading acquisition has been extensively studied and is somewhat complex
(Bradley & Bryant, 1983). The relationship seems to be bidirectional, with phonological
awareness facilitating reading acquisition, whereas reading acquisition improves phono-
logical awareness, especially phonemic awareness (Burgess & Lonigan, 1998; Lundberg,
1998). It is harder to teach children the correspondences between letters and sounds if
they don’t recognize the sounds within the words (and for this reason, teaching these
prerequisites is of great help in teaching reading). Thus, for the student to read a very
large number of words—particularly words not previously taught—it is necessary to teach
Chapter 9 183

sound-text combinations smaller than words. This makes it possible that “words that are
composed of new combinations of previously learned letters and sounds are named the
first time they are seen” (Mueller et al., 2000, p. 515).
Syllables are convenient sound and textual within-word units in languages considered
“transparent,” such as Portuguese or Spanish. This feature “makes syllable recombina-
tion largely sufficient for novel word decoding” (Mueller et al., 2000, p. 515). In English,
however, syllables are much less convenient as units. For this reason, according to Mueller
and colleagues (2000), when a student is learning to read novel English words, having
the skill of recombining within-syllable units is critical (p. 516). These authors investigated
whether prereading children would learn to read words and demonstrate recombinative gen-
eralization of within-syllable units after they had been taught sets of words with overlap-
ping letters. The words were taught in sets, each set with six words—four words used in
training and two words saved for generalization tests. Within sets the four training words
had overlapping letters, and the novel words were recombinations of the same components
as the trained words. The within-syllable units were onsets and rimes. The authors used a
matrix training strategy to organize the word sets. The distribution of the stimuli (onsets
and rimes) in a matrix shows whether or not all of the units involved in recombinative
reading would be trained, although in different combinations. Displaying the onsets in
the rows and the rimes in the columns, one can ensure that each cell contains a word with
a unique combination of both elements. For example, using s and m as onsets and at, op,
and ug as rimes results in the words sat, sop, sug, mat, mop, and mug. Selecting the words
sat, mat, sop, and mug for training, one can test the words mop and sug for recombinative
reading; note that mop and sug are novel words, but their onsets (m and s) and rimes (op
and ug) would be systematically taught in the training words. Training was conducted
with matching-to-sample procedures: On each trial the sample stimulus was a spoken word
and the comparisons were four printed words, one of which corresponded to the dictated
word. (The procedure was developed in small steps, the description of which the interested
reader can find in the original paper.) The same kind of trials tested whether the children
would select a novel printed word when the corresponding word was dictated. Children
were also tested on reading comprehension (selecting the picture when the sample was a
printed word) and naming the printed word (textual behavior). Three children learned to
select the words that were directly taught and also the novel words after just one (two chil-
dren) or two training sets (one child). The children also obtained high scores in matching
pictures to printed words (comprehension) and named the majority of the printed words
(textual behavior). Two control children, exposed only to the tests, showed low accuracy
across all of the six sets. The same teaching strategy was used with two adult women diag-
nosed with intellectual disabilities (Saunders, O’Donnell, Vaidya, & Williams, 2003). Both
women learned to match dictated words to printed words and demonstrated recombination
of within-syllable units (onset and rimes) with novel words, in the same task. One woman
also showed comprehension of the printed words (selecting the corresponding picture), and
her reading scores increased markedly from pre- to post-tests. The reading gains in these
two studies were expressively higher than those obtained in other studies that manipulated
within-syllable components but not in the systematic way allowed by using the matrix train-
ing. Goswami (1986), for example, showed that after having learned to read a “cue word” a
child would read a novel word “by analogy.” For example, after learning to read words like
hark, coat, and beak, some children were able to read lark, float, and peak. The selection of
words, however, was based only on rimes. On tests a child could be exposed to words for

which the onsets had not been trained, which may explain the low scores in reading new
words and nonwords. Thus, a combination of matrix training and a strategic selection of
words could produce predictable and strong effects in teaching reading.
Delayed Constructed-Response Matching Procedures

to Establish Spelling
If reading with comprehension or understanding requires textual behavior as a basic
component (not the only one, but a fundamental one, without which no reading would be
possible), writing with meaning requires accurate spelling. The production of a sequence
of letters (by typing, handwriting, signing, selecting among a pool of letters on a screen,
and so on) can occur under the control of a variety of types of discriminative stimuli;
these include text, as in copying or transcription, or spoken words, as in dictation taking,
which means that spelling may become a component of different discriminated operants.
As has been shown, spoken words, printed words, objects, and pictures can all become
arbitrarily related by equivalence given the right teaching conditions. For this reason,
even though copying, dictation taking, and textual behavior are operant relations that
can be independently acquired (Skinner, 1957; Lee & Pegler, 1982), they can also become
interrelated or merge into equivalence classes (de Rose et al., 1996; Sidman, 1994) when
their controlling stimuli become members of such classes. A very important implication
of a network like this is that if the student learns one of the relations and also relations
among the stimuli that control the other relations, he may begin to show new skills in the
absence of direct teaching. Studies using the equivalence paradigm show that participants
can produce the stimuli if they have already acquired the necessary response topography
(Sidman, 1971). This could explain, for example, the high accuracy in dictation taking
that emerged as a “by-product” in the study of de Rose and colleagues (1996) and is con-
ceptually represented by the diagram in figure 9.4.
Matching-to-sample methods have been adapted for studying and teaching spelling
as well as reading. Dube and colleagues (1991) taught two young men with intellectual
disabilities to spell by touching a computer touch screen. Stimuli were printed words and
pictures that the participants had previously learned to match (arbitrary matching). On
each trial, a picture and the corresponding printed word were presented as a compound
sample. When the participant touched the sample, an array of letters was displayed in
a choice pool at the bottom of the screen; he could then select the letters to construct
the matching word (constructed-response matching to sample, or CRMTS). Each correct
selection moved the letter to the construction area right below the sample (only selec-
tions of letters that matched the sample in the correct order produced feedback); the trial
was interrupted if the participant selected other letters or a letter that was in the word
sample, but not in the correct order. The task was, therefore, identity matching, but the
participant constructed the stimuli letter by letter as in copying. The printed word was
then gradually faded out, and eventually both participants were constructing the words
in the presence of the picture alone; that is, they learned an arbitrary relation between the
picture and its printed name.
The procedure developed by Dube and colleagues (1991) has been systematically rep-
licated in a series of subsequent studies. The addition of new features to the procedure
Chapter 9 185

proved equally effective in teaching spelling to students diagnosed with intellectual dis-
abilities. Such features include the use of a delayed matching procedure (Stromer &
Mackay, 1992a, 1992b; Stromer, Mackay, & Stoddard, 1992), the use of a tabletop version
in which the CRMTS was conducted with movable letters and the choice pool included
all letters of the alphabet (Hanna, de Souza, de Rose, & Fonseca, 2004), and the use of
computer video models and video rewards (Kinney, Vedora, & Stromer, 2003).
Inaccurate Reading and Spelling as Stimulus Control

by Irrelevant Stimuli
Teachers are especially concerned about errors in reading and spelling. However, it
is important to recognize that what is called an “error” is behavior under the control of
stimuli or aspects of stimuli not intended by the teacher or the researcher (Stoddard &
Sidman, 1967). Instead of just classifying the behavior as an error, it is important to look
more closely at the stimulus control involved in specific types of errors (Lee & Pegler,
1982; Lee & Sanderson, 1987) and verify how procedures and contents of our teaching
practices could prevent or overcome such occurrences.
Dube and colleagues (1991), for example, verified that their participants showed
accurate performances when they selected printed words that corresponded to pictures.
However, when they had to construct the same words letter by letter, their scores were at
chance level, and usually only the selection of the first letter was correct, suggesting that
the choice of a whole word was probably based on this letter. Control by the first letter is
a fairly common finding in the literature on reading acquisition (Adams, 1998).
Misspelling of consonant clusters is also one of the most common types of spelling
errors (Treiman, 1991). Words with consonant clusters have two or more consonants
in sequence, such as SNow, woRD, peNCil. Birnie-Selwyn and Guerin (1997) taught
six elementary-school-aged children to spell words containing initial consonant clusters.
They were taught to match printed words to spoken words and were then tested for spell-
ing dictated words with a constructed-response matching procedure. Two training con-
ditions were compared. On the critical-difference condition, the children were required
to respond to both letters of the consonant cluster. For example, if the target was the
compound stimulus sh in the word show, the incorrect (s-) comparison stimuli would be
slow and snap. On the multiple-difference condition, however, a correct selection could
be based on a single letter in the pair (incorrect comparisons were nice and rest, for
example). Children made fewer errors on the tests after the critical-difference condi-
tion than after the multiple-difference condition (although fewer errors occurred during
the multiple-difference training). Thus, some consonant clusters “errors” might be the
result of selective stimulus control. If participants are taught to respond to multicompo-
nent stimuli and are then tested for responding to the components separately, they may
respond only to some components of the multicomponent stimuli. Because all letters
and their sequence in a word are relevant for accurate reading and spelling, procedures
designed to overcome stimulus selectivity must incorporate stimulus control technology
in order to ensure that discrimination of all relevant elements and matching tasks can
benefit from the critical difference among comparisons (Birnie-Selwyn & Guerin, 1997;
McIlvane & Dube, 2003).

Relevant Aspects in Designing Procedures and a
Curriculum for Teaching Rudimentary Reading
The preceding section presented research on learning processes in reading and spelling
acquisition. These studies have shown many important details in programming contingen-
cies for teaching early reading skills, whether students are typically developing children or
persons with developmental disabilities. In short, teaching reading requires establishment
of textual behavior and also reading with comprehension. Comprehension emerges from
a network of stimulus-stimulus and stimulus-response relations of which text is only one
component, and it builds upon the repertoires construed during language development,
including both speaker and listener skills. Generalized textual behavior—decoding not
only words previously learned but also novel words—requires stimulus control by within-
word and within-syllable textual units. The same seems to be true for generative spell-
ing. Research also shows that textual behavior and spelling can emerge from matching
to sample between dictated words and the respective printed forms, and that generative
reading and spelling is strengthened by focusing upon the correspondences between small
textual units and the respective sounds.
The next section presents some strategies for programming conditions for teaching
early reading skills in applied settings.
Matching-to-Sample Procedures
Matching-to-sample (MTS) procedures are very effective in establishing the stimu-
lus equivalences on which reading and spelling are based. The main characteristics of
matching-to-sample procedures can be summarized as follows: There is a sequence of
trials, each of which displays a sample and two or more comparison stimuli. For each
sample, there is only one correct comparison. The sample varies across trials, and thus the
correct comparison also changes along the sequence. This sequence is unpredictable so
that the student cannot respond correctly by guessing or memorizing the correct alterna-
tive; in addition, there are differential consequences for correct and incorrect responses.
The relation between samples and correct comparisons lies along a continuum from
identity to a purely arbitrary relation. On the identity extreme of this continuum, there-
fore, one of the comparisons is identical to the sample and the others are not. The learner’s
task is to choose, for each sample, the comparison that is identical to it (identity matching
to sample). On the other extreme, the correspondence between samples and comparisons
is based on an entirely arbitrary, conventional relation (arbitrary matching to sample). The
relation between an object and its spoken or written name is an example of an arbitrary,
conventional relation that varies among different verbal communities. When children
learn the vocabulary in a language, they are learning arbitrary relations among names and
objects; each object will correspond to a certain name and each name will correspond to
a given object. (The reference to a one-to-one relation of correspondence between names
and objects is an oversimplification that is useful for our purposes in this chapter.) This
is a symbolic relation (Sidman, 1994). However, there are important differences between
learning naming-object relations in a natural setting and learning the same relations
through matching to sample in educational curricula in school. In natural settings the
Chapter 9 187

child is learning both to recognize the objects (to look at or point to an object when its
name is spoken) and to pronounce the names in the presence of the objects. Typically,
children are also exposed to many words and objects in multiple learning opportunities
that are not explicitly programmed. This absence of programming means that the level
of difficulty does not increase systematically and learning in one step is not necessarily
verified before the child advances to the next. Surely adults, intentionally or not, use
many resources to smooth the child’s progress in learning, and most children exposed
to sufficient stimulation learn to speak without serious difficulties. However, this does
not happen as easily with children with sensory or neurological deficits or children who
are reared in impoverished environments. Hence, the matching-to-sample procedure is
important and useful because it reproduces, in a simplified version, the process of learning
relations that are fundamental for cognitive development. However, the transposition of
matching to sample into structured educational curricula also involves a simplification of
the tasks and requires careful programming of contingencies for teaching them according
to the principles of applied behavior analysis.
General Requirements to Teach Matching to Sample

The necessary conditions for the acquisition of matching to sample are the same as
those required for the acquisition of any operant behavior.
Learner’s active responding. This is a condition inherent to the matching-to-sample

procedure. The learner responds to each trial, selecting one among several comparison
stimuli. The form, or topography, of the selecting response is not relevant: it can be
adapted to the learner’s sensory and motor capabilities. When comparison stimuli are
concrete objects, the most used selection responses are those involving reaching for or
touching an object. When the comparisons are pictures presented on a computer screen,
the responses most frequently used involve selecting a picture by clicking with the mouse
or by touching the picture directly on the screen. When comparison stimuli are printed
on paper, the selection response could be a check mark next to one of the stimuli, a cross
over it, or a circle around it. For individuals with severe motor limitations, the situation
could be adapted to accommodate any differential response that the student could emit
toward the comparison stimuli as a selection response.
Immediate differential reinforcement. To teach matching to sample it is necessary to

program a reinforcing consequence for correct responses (and not for incorrect responses).
It is important to find out which stimuli indeed function as reinforcers for each learner
(see chapter 1 of this volume) and which of those have reinforcing value during a particu-
lar training session, since this value may vary with motivational operations (see chapter 5
of this volume).
Small steps: Subdivision of the task in sequential steps, with gradually increasing
difficulty. Matching to sample is usually a step in a broader learning task. Depending
on the student’s limitations, it may be necessary to also subdivide the matching itself into
smaller units. Some individuals with intellectual disabilities can show marked difficulties
in learning arbitrary matching. Saunders and Spradlin (1990) developed an ingenious
and very effective way to overcome such difficulties. Suppose that the goal is to teach the

student to match two comparison stimuli, denoted for convenience as B1 and B2, to two
sample stimuli, A1 and A2. Therefore, the comparisons B1 and B2 should be presented
on every trial and the sample would be A1 or A2. In the presence of sample A1, selections
of B1 are correct; in the presence of sample A2, selections of B2 are correct. One of the
requirements for matching is that the students distinguish between samples A1 and A2.
This is known as a successive discrimination, since A1 and A2 are presented one at a time
across successive trials. In order to learn the successive discrimination between A1 and
A2, the participants could be taught to name the stimuli (or to present any other dif-
ferential responding to each of them). After the student masters this task, it will then be
possible to teach the other component necessary for matching: the simultaneous discrimi-
nation between B1 and B2. In arbitrary matching this discrimination is continuously
reversed; that is, the correct comparison changes along trials, depending on the sample
that is presented. To achieve this stage, begin by presenting blocks of consecutive trials
with the same sample (for example, in a block of twenty-four trials, the first twelve trials
present sample A1 and the second twelve present sample A2); then, gradually decrease the
number of trials in those blocks (6-6-6-6; 3-3-3-3-3-3-3). In the final phase, the sequence
of sample presentations is randomized. Eventually, as the sequence is repeated with differ-
ent matching tasks, training the components becomes unnecessary, and students become
proficient in learning by trial and error (Saunders & Spradlin, 1990).
Excellence in one level before advancing to the next. In programming for teaching
arbitrary matching, it is necessary to establish a performance criterion to be achieved in
each step of the program. The criterion has to be sufficiently rigorous in order to ensure
proficiency in one step before advancing to next. Even when the previous step has been
learned, performance may deteriorate when a new step is introduced. As soon as any
decline in accuracy is detected, the clinician should immediately go back to previous steps
in the program until an accurate performance is recovered. Sometimes the recovery does
not occur in the immediately preceding step, and it is then necessary to go back many
steps until the performance is completely recovered. Accurate performance may often be
disrupted whenever the program reaches a certain point. This suggests the need to modify
the program at that point. For example, the graduation of levels of difficulty may be inap-
propriate, and the progress from one step to the next may have been too large for a par-
ticular student. In this case it is important to subdivide the program with the inclusion
of intermediary steps. (The reader interested in this topic can find detailed information in
the report by Sidman and Stoddard [1967] on the construction and revision of a program
to teach discriminations to nonverbal persons.)
One example in which it was necessary to break the task of matching printed to dic-
tated words into smaller steps comes from a study in which children enrolled in special
education needed an average of 3.9 training sessions to learn to read each set of three or
four words, whereas preschoolers, first graders (with a history of school failure), and illiter-
ate adults needed, respectively, 1.1, 2.0, and 1.5 sessions per set (Melchiori, de Souza, & de
Rose, 2000). Special education students needed more sessions because they made errors in
reading tests conducted after they achieved criterion in the matching task. The systematic
occurrence of errors suggested that three or four words were large sets for those students.
The authors then used a remedial procedure that consisted of adding only one unlearned
word at a time. For example, the first session was conducted as initially planned, but, if
the set of words for that session had four words and the child read only one of them, then
Chapter 9 189

the next block of training trials would contain two training words distributed among
baseline trials—the one already learned plus only one unlearned word. If the student read
both words in the reading test, then the next block of matching trials would add a third
unlearned word, and so on. Eventually, all of the four words were included in the block
of training trials and the student reached the criterion of 100 percent correct responses
in the reading test. When the number of words per set was reduced, the final achieve-
ment of the special education students was comparable to that of the other groups. This
feature could be employed as a regular practice for teaching students with disabilities—
beginning with only one new word per set and introducing a larger number of words per
set only when the student acquires a more extensive repertoire.
Specific Requirements for Teaching Matching to Sample

In addition to the conditions that are fundamental for the acquisition of any operant
behavior, there are some specific requirements for teaching matching-to-sample perfor-
mances. As mentioned before, learning matching to sample requires successive discrimina-
tions of the sample stimuli and simultaneous discrimination of the comparison stimuli.
Discrimination of sample stimuli. A technique frequently used to increase the prob-

ability of discrimination of the sample stimuli is the displaying, on each trial, of only the
sample stimulus, and waiting until the learner emits an observing response to the sample.
Only then are the comparison stimuli presented, as a consequence for the observing
response. This was done in Sidman’s study (1971): An observing response was required
for the presentation of the comparison stimuli on the peripheral windows. The observ-
ing response was a touch to the central window. Note that the response of touching the
sample ensures the student’s orientation toward the sample but does not ensure that the
student observes and distinguishes the features that differentiate one sample stimulus
from the others. This is because this observing response is nondifferential; in other words,
it (touching) is the same for all samples. A nondifferential observing response ensures
that the student attends to the sample if only to detect its presence. This attention to
the sample may be sufficient to allow the student to detect its distinctive features and
discriminate it from other samples, but it does not ensure this discrimination between
samples. If incorrect responses then occur, this may be an indication that the student does
not discriminate between samples. One alternative is to require a differential response to
each sample stimulus, a solution used by Saunders and Spradlin (1990) to teach match-
ing to sample to individuals who had repeatedly failed in learning this task before. For
example, if the sample is a dictated word, the student may be required to imitate it. If the
sample is a picture, the student may be required to name it. Thus, participants emit a dif-
ferent response in the presence of each sample stimulus. Each trial starts with the display
of the sample stimulus, and if the observing response is the naming of the stimulus, for
example, the learner has to name it, thus producing the presentation of the comparison
stimuli. Accurate naming assures that the learner discriminates each sample. For partici-
pants who have difficulties in producing names, other responses that the participant can
emit could be used (for example, manual signs, different rates in touching the stimuli,
and so on).
It is important to make sure that the participant really discriminates the relevant
features of the sample stimuli and not some other characteristic that could be regularly

present in the stimuli or in the arrangement of trials. Thus, a necessary caution is to
randomize the order of trials, since a nonrandomized order would allow the learner to
respond to the order and not to the samples. For example, suppose that the order of the
samples is always A1, A2, A3, A3, A2, A1, successively. The individual could eventually
select the comparisons B1, B2, B3, B3, B2, B1 under the control of the sequence, paying
no attention to the sample stimuli, and her responding would seem apparently accurate.
Discrimination of comparison stimuli. For the same reason as described above, it is

important to ensure that the student would not select the correct comparison on the basis
of some irrelevant feature. For a simple example, suppose that the instructor is trying
to teach the student to match the comparisons B1 and B2 to the samples A1 and A2,
respectively. On each trial, one of the samples and the two comparisons are presented,
one at the right and the other at the left of the sample stimulus. Suppose, also, that the
instructor was not careful enough in programming the sequence of trials and that in
two consecutive trials the correct comparison stimulus (S+) is located at the right of the
sample and that in the next two it is at the left of the sample, so that the position of
the correct comparison stimulus alternates from left to right at every other trial. In this
case, the student could select the correct comparison on all trials without even looking
at the sample and the comparisons; he could simply learn to alternate between left and
right after every two responses. The learner would be responding under the control of the
position of the comparison stimuli and not under the control of the relation between the
sample and the comparison. It is important to note that the instructor would not be able
to identify the controlling relation, since the overt response would be the same in both
cases. However, if the instructor eliminates the regularity by randomizing the position of
the correct comparison and the learner continues selecting the correct comparison, then
it is possible to ensure that position is not controlling correct choice.
Programming to Minimize Errors in Learning

In any instructional program, the teacher attempts to guide the student’s response
and to establish correct responses to the relevant stimuli, thus maintaining high accuracy.
The procedures for teaching reading described so far include many features to attain
this goal. For instance, the exclusion procedure used by de Rose and colleagues (1996)
has functional similarities to the fading procedure described by Ribes (1978). In Ribes’s
program, the color discrepancy between the correct and incorrect comparisons prompted
the student to select the correct comparison. In exclusion, prompts are based on previ-
ous learning. When a new sample word is dictated, the student can detect the difference
between the sample word and the word she has already learned and thus exclude it,
selecting the undefined printed word. In order to establish responding to the relevant
stimuli, the clinician must also rule out control by irrelevant stimuli. If the procedure
only presented exclusion trials, the student would not need to pay attention to the dic-
tated samples and their relation to the printed words—if she just picked every unfamiliar
printed word, the responses would always be correct. To ensure that students attend to
the samples, the clinician must present control trials, in which the unfamiliar printed
word is displayed in the array of comparison stimuli but a baseline word is dictated as the
Chapter 9 191

sample. These arrangements usually ensure that the student matches the printed to the
dictated words with virtually 100 percent accuracy.
It may seem that if the response is “guided,” the student is not learning but simply
relying on the cues. Quite the contrary—if the student responds correctly, the conse-
quences will affect her responding and then, as the responses and the relation between
responses and the relevant features of the discriminative stimuli are strengthened, the
cues can be gradually removed. Or, depending on the kind of cue, it will simply lose its
function. This is very clear with the exclusion procedure. The first time a novel word is
dictated as the sample and a novel printed word is displayed together with a defined one,
the student selects the novel word. Immediate reinforcement increases the probability
that the student learns the relation between the spoken and the printed word. If the same
spoken and printed word is related on subsequent trials, three to four times in a session,
then the relation between the dictated and the printed word is likely to be established.
The instructor can test whether the relation has indeed been learned using two types of
probes. The first type involves trials with the word just taught dictated as the sample, and
with the corresponding printed word and a novel word as comparisons (if the student
incorrectly chooses the novel word, one can conclude that the previous relation was not
learned and that the response was mostly controlled by the novelty, but if she chooses the
correct word, this can be taken as an indication of learning). The second type is reading
probe trials, with the printed word presented alone and the student being asked to say
the word.
As in other kinds of discrimination learning, stimulus features may influence reading
difficulty for a particular child. Laboratory evidence suggests that for students with no
previous experience with printed stimuli, uppercase letters are usually easier to differ-
entiate from each other than lowercase letters are. For this reason, uppercase might be
preferable in early stages of teaching reading (Adams, 1998, p. 357). However, at some
point students would need to learn the correspondences between upper- and lowercase,
since most reading materials are printed using lowercase. The literature also suggests that
it is better to teach one feature first and add the second only after the student shows
proficiency in the first. The technology of teaching based on the stimulus equivalence
paradigm can be used to teach the arbitrary correspondences between uppercase and
lowercase letters. Finally, the literature suggests that “to minimize confusion between
visually similar letters, especially b and d, it is best to separate their introduction in time
such that one of them is thoroughly familiar before the second is presented” (Adams,
1998, pp. 357–358). A special feature of textual stimuli is that words are composed by
letter sequences. Students need to discriminate not only the letters but also the particular
order in which they are displayed. The constructed-response task (often called anagram
construction), in which students construct a word with movable letters or select letters on a
computer screen, is very useful to teach such sequences. The procedure was developed by
Mackay (1985) to teach relations between colors and their printed names, and it was also
used in the program by de Rose and colleagues (1996) to teach the CE relation (see figure
9.4). It is easily implemented on computers and in a tabletop format as well, as illustrated
in teaching Jay to construct number words (see chapter 10 of this volume).

Programming for Cumulative Learning:
A Curriculum to Teach Early Reading Skills
In a short chapter published in the late 1970s, Murray Sidman (1977) outlined a kind of
training path for teaching rudimentary reading skills to students with severe intellectual
disabilities. In his own words:
When we study reading in retarded people…we are forced to ask questions that
do not arise with normal readers who have already developed the basic skills
through exposure to natural contingencies. Our interest here, therefore, is not the
level of complexity at which you and I operate when we read, but rather in skills
and processes which by themselves may not be considered reading but which are
prerequisite if the retarded student is to take advantage of an existing capacity.
(p. 353)
Figure 9.5 summarizes the main steps in the training sequence for bringing the stu-
dent’s repertoire from simply discriminating the presence or absence of a stimulus to
discriminating and relating printed to spoken words. Sidman was not only suggesting a
sequence of tasks but also pointing out the available behavioral technology for teaching
the skills at each level (much of which had at that time been recently developed by him
and his colleagues). After teaching the student to detect the presence or absence of a visual
stimulus, the teacher proceeded to teach form discriminations. Sidman and Stoddard
(1967), for instance, used stimulus control shaping to teach children with developmental
disabilities to discriminate between circles and ellipses. Children had learned to touch
a window that contained a circle in a three-by-three matrix in which all other windows
were blank. The presence of the circle became a discriminative stimulus (SD or S+), while
a blank window was not touched (S∆ or S-). Faint ellipses were then projected onto the
blank windows and the student had to select the circle, whose position changed from trial
to trial. Over successive trials, the thickness of the ellipses’ lines increased until the lines
of circle and ellipses were identical. Responses were initially controlled by the presence
of any form. Stimulus control by a particular form, the circle, was shaped over successive
trials, as the incorrect stimuli (ellipses) changed progressively to become ever more similar
to the circle. The ellipses were very flat, so that the difference between circle and ellipses
was maximal. In subsequent trials, the vertical axis of the ellipses increased, so that they
became increasingly more similar to the circle. Children continued to respond correctly
until the ellipses were barely distinguishable from the circle. This stimulus control shaping
technique can be used to teach other form-versus-form discriminations (step 3), until a
student learns, for instance, discriminations between letters. Step 4 shifts from simple
discrimination to identity matching to sample. Initially the student may match letters,
followed by letter groups. Auditory stimuli are introduced in step 5, so that matching
to sample is now arbitrary and takes place between dictated words and pictures. This
sequence of steps should establish the prerequisites for auditory-visual matching between
dictated words as samples and printed words as comparison stimuli.
Chapter 9 193

A Possible Path Toward Reading
Step 1: Teach presence versus absence visual discrimination.
Step 2: Teach initial form versus form visual discrimination.
Step 3: Teach many additional visual discriminations (for example, between letters)
via stimulus control shaping methodology.
Step 4: Teach generalized identity matching with letters and letter groupings via
stimulus control shaping.
Step 5: Teach auditory-visual matching to sample with pictures (substantives).
Step 6: Teach auditory-visual matching with corresponding printed words.
Figure 9.5. The training path suggested by Sidman (1977) for teaching the prerequisites
for reading (adapted from McIlvane, 2007).
Later, it turned out that in some steps (especially identity matching and auditory
matching) the available technology was not sufficiently effective to teach some students,
but fortunately several years of research has led to a more advanced technology for teach-
ing such skills in several intermediate steps (Serna, Dube, & McIlvane, 1997). The addi-
tion of intermediate steps for some students does not invalidate the possible path toward
reading proposed by Sidman (1977). On the contrary, it is perfectly consistent with the
behavior analytic approach and points to alternative routes of training that can be used
when the student has difficulty.
For students who are already able to learn new auditory-visual matching relations,
like those in the studies of Sidman (1971) and de Rose and colleagues (1996), teaching
reading and writing can start at step 6 of the sequence proposed by Sidman (1977). In
fact, a combination of steps 5 and 6 in a training package has proven to be a good strat-
egy: matching pictures to spoken words (step 5) and matching printed words to spoken
words (step 6) engenders the formation of equivalence classes comprising pictures, printed
words, and spoken words, as we have already mentioned, allowing for reading with com-
prehension. An additional advantage of mixing steps 5 and 6 is motivational: if matching
pictures to spoken words is an easier task for the students, or if they have already learned
those relations, they receive a high density of reinforcement during a session, which can
maintain their behavior even if, occasionally, they make incorrect choices in selecting the
printed words.
Thus, supposing the students have the prerequisites, how should we teach them new
content? In other words, how should we increase the number of new spoken word–printed
word relations in their repertoires?

Successive Word Sets
Research has clearly shown that learning to read requires a systematic exposure to a
varied amount of textual content (Adams, 1998; Snow, et al., 2005). It is not sufficient
to expose a student many times to a small set of words; rather, increasing the number of
words directly taught increases recombinative reading (de Rose et al., 1996; Mueller et al.,
2000) and spelling (Dube et al., 1991; Kinney et al., 2003). Thus, a systematic curricu-
lum to teach early reading skills at the word level should be planned with a large number
of words. During the program implementation, however, the introduction of periodic
reading probes of words, both directly taught and novel, could monitor the student’s
progress. How many words should be taught is a matter of the individual student’s pace.
Along the way, students may begin to show recombinative reading after having learned
different numbers of word sets (de Rose et al., 1996). Thus, the teaching program can be
prepared in advance, but the student’s performance will be the criterion for considering
whether he is already reading new words and may be prepared for more advanced materi-
als. One way of doing this in a systematic way is to introduce retention probes, as well
as pre-tests for each new word set. This way, students may be spared training with words
they have already read on the pre-tests but will be reexposed to words that they have not
read on the retention probes.
Figure 9.6 shows a possible sequence of tasks and training steps for teaching reading
of successive word sets, based on a study by de Rose and colleagues (1996). As in that
study, measures could be obtained at the beginning and at the end of each teaching unit,
and each unit could be subdivided into several smaller units or teaching steps. The total
number of teaching units should depend on the student’s performance.
Cumulative Baseline
Like any other behavior, reading, especially at this rudimentary level, requires con-
tingencies for its maintenance. Therefore, it is important to include learned words in new
“lessons” for practice—that is, opportunities for the occurrence of the behavior, and thus
for its reinforcement. However, there are other more important reasons for the inclusion
of learned words in new lessons. For example, they may play an important role as a base-
line upon which the new relations are learned. If the teaching procedure is exclusion, as
was the case in the program of de Rose and colleagues (1996), a well-established baseline
is crucial for implementing exclusion trials and control trials (see figure 9.3). Additionally,
a block of training trials can run more smoothly if baseline trials are interspersed among
exclusion trials. On baseline trials, both comparison stimuli are baseline words and the
dictated sample corresponds to one of them. Since this is an established relation, there is
a high probability that the student will make a correct selection, thus leading to further
strengthening of the baseline relation by its immediate consequence. For these reasons,
a block of training trials could comprise two or three times more baseline words as the
sample stimuli (on both control trials and baseline trials) than novel words (on exclusion
trials). Even if the teaching program is not using exclusion procedures, baseline trials
could be included among training trials.
Chapter 9 195

Figure 9.6.

Programming for Recombinative Reading:
Overlapping Word Sets
Not all students with developmental disabilities may be ready for recombinative
reading. For some students, the instructional goal may be to teach sight-reading of func-
tionally important words. For students with more advanced repertoires, however, an
important criterion in the selection of words to be taught is how suitable some word sets
are for favoring recombinative reading, because when this skill is established, the student
becomes able to read words not explicitly trained. As we have already mentioned, when
students are taught to respond to several different words that contain overlapping units,
they may also come to respond to different combinations of the same units (cf. Mueller et
al., 2000). At the word level, within-word and especially within-syllable units, which are
crucial to decoding English words (Mueller et al., 2000; Snow et al., 2005), could be con-
sidered in the selection of word sets for a reading program. Onsets and rimes as within-
syllables units can be sufficient for teaching and generating the reading of novel words in
many students (see Mueller et al., 2000, for examples of word sets). For some students,
however, it may be necessary to break the syllables into still-smaller units, if their errors
involve misnaming or omitting the middle vowel, for example (Saunders et al., 2003). If
that is this case, the teacher could consider selecting words whose rimes differ from one
another by just one letter (the critical difference; cf. Birnie-Selwyn & Guerin, 1997). For
example, word pairs such as hut/hat, pat/pet, rug/rag, and bond/band could be useful for
teaching the discrimination between middle vowels.
As we also mentioned earlier, the training matrix strategy can help in this task, ensur-
ing identification of (and subsequently exposure to) all of the units that will be required
for recombinative reading, and prompting the decision about what words to teach and
what words to save for recombination probes. The examples in figure 9.7 may clarify these
possibilities. In each matrix, the rows show the syllable onset and the columns show the
rimes. The asterisks indicate words saved for recombinative reading probes. Matrices of
different sizes (two by two, two by three, three by two, and three by three) can be used
to generate a list of words that will be used for training and words that will be tested.
Once a word list is produced, it is possible to decide the sequence of words along succes-
sive steps of the teaching program. How many words to teach on each step depends on
the student’s possibilities. For example, the program developed by de Rose and colleagues
(1996) teaches three words per set, and the one developed by Mueller and colleagues
(2000) taught four words per set. However, Melchiori and colleagues (2000) suggested
that for some students it could be more effective to begin by teaching just one new word
at a time and then gradually increasing the number of words per set.
Rimes
Onsets
at op ug
p pat pop pug
m mat mop* mug*
Chapter 9 197

Rimes
Onsets
ilk ink alt ole
s silk sink* salt sole
m milk mink malt* mole
Rimes
Onsets
at int ouse ug
h hat hint* house* hug
m mat* mint mouse mug
Rimes
Onsets
eak ear in it ot
p peak pear* pin pit pot*
b beak bear bin* bit bot
Figure 9.7. Examples of matrices for defining word sets with overlapping within-syllable
units. The first example is from Mueller et al. (2000). Asterisks indicate words for testing
only.
Getting Started
The procedures for teaching reading can be applied with conventional instructional
materials in a tabletop format, or they can run on computers. In both cases, the instruc-
tor will be required to prepare the lessons, that is, to compose blocks with a sequence of
trials for each lesson (and to also construct the corresponding instructional materials if
the option is a tabletop version). Examples from the program of de Rose and colleagues
are shown in table 9.1. Except for the first step, in which a baseline should be established,
all of the subsequent steps begin with a block of trials to assess retention of the words
taught in the previous step. If the student does not pass the retention probes, she returns
to the previous step for retraining; otherwise, she is exposed to a pre-test of the new word
set. The words of each new set can be substituted for the words indicated by numbers in
this table (and other protocols could be prepared with different sequences of trials, differ-
ent positions of the correct choices, and so on). The sequence of trials includes different
trial types, with different functions in this procedure (baseline trials, exclusion trials, and
discrimination trials). If the student shows high accuracy during the matching tasks, then
the post-test verifies the accuracy of a subset of these trials. If a criterion of 100 percent
correct is not reached, the training block is repeated. When the criterion is achieved, the
next subset of trials presents only the new words as comparisons (discrimination post-test),

and the criterion is again 100 percent of correct responses. Finally, the student is exposed
to the reading test. This is just an example of how a teaching session could be planned.
If the session is conducted in a tabletop format, the printed protocol could be used as a
guide for the teacher (suggesting which stimuli to present on each trial, in which position,
and so on), as well a form for recording the student’s responses.
Table 9.1 shows a sample of blocks of trials for a typical session for teaching a new
word set. Some trials are used for tests and others for training. X1, X2, and X3 stand for
words taught in the previous session. Words 1, 2, and 3 are the novel words introduced
in the session. Baseline words are words previously learned. Criterion means that 100
percent accuracy is required for the task; if criterion is not attained, the student repeats
the training. In this example, the training block has thirty-six trials (only thirteen are
shown), twelve exclusion trials (four for each word being taught), nine control trials, and
fifteen baseline trials.
Assessment of Reading and Its Prerequisites

Before beginning to teach reading (or any other behavior), it is fundamental to measure
the student’s repertoire and, based on the obtained measures, make decisions about what
to teach next. Printed-word naming (or textual behavior) is the basic behavior of inter-
est, but measures of other verbal relations can also be obtained, as well as measures of
matching-to-sample skills with different types of stimuli. Table 9.2 lists some tasks that
have been useful in characterizing students’ repertoires during a comprehensive assess-
ment before teaching reading. The pairs of letters are listed to allow for a comparison with
the diagrams presented in figures 9.2 and 9.4. The first letter indicates the set of samples
or controlling stimuli, and the second letter indicates the set of comparison stimuli or the
response classes. The particular relations included in the pre-tests will depend on what
is already known about the student’s repertoire. The assessment of echoic behavior, for
example, determines whether the child responds to spoken words with a point-by-point
correspondence. If the test results are positive, one can conclude that the student has the
repertoire to participate in arbitrary auditory-visual matching (in other words, the listen-
ing component is functioning well) and in naming words and pictures (in other words,
the student can pronounce words). Identity matching tasks are important for determin-
ing whether the student already discriminates differences among pictures and, especially,
among printed words. Matching pictures to spoken words tells whether the student has
already learned arbitrary relations. If not, one should teach those skills (as was done by
Sidman & Cresson, 1973) before teaching the crucial skill of matching printed words to
spoken words. In naming tests, if a student names a picture but not a printed word, this
indicates that the picture-name relation is established and rules out potential problems in
the student’s ability to pronounce the words. For example, at some point the instructor
may need to know whether the student matches uppercase letters to lowercase letters (an
arbitrary matching skill that should not be overlooked, depending on the level of reading
and writing attained by the student). However, if the student has a more advanced rep-
ertoire, then the assessment of textual behavior (CD), dictation taking (selecting letters
[AE], writing with pencil and paper [AF], or both), and matching printed words to
dictated words (AC) will suffice.
Chapter 9 199

Table 9.1. Sample of a Sequence of Blocks of Trials for a
200
Typical Session for Teaching a New Word Set.
Block Number
and Task Trial Correct
Function Trial Types Number Sample Comparison 1 Comparison 2 Comparison 3 Selection Criterion
1. Retention AC(a) 1 word X1 X3 X2 X1 3

AC 2 word X2 X2 X1 X3 1 100%
AC 3 word X3 X1 X3 X2 2

2. Pre-test (3) AC 4 word 1 word 1 word 3 word 2 1
AC 5 word 3 word 2 word 1 word 3 3 -
AC 6 word 2 word 3 word 2 word 1 2
3. Training AC (baseline) 7 baseline 1 word 1 baseline 1 2

(AC + CE) AC (exclusion) 8 word 3 word 3 baseline 4 1
CE(b) 9 word 3 CRMTS
3. Training AC (baseline) 10 baseline 2 baseline 2 word 3 1
(AC + CE) AC (exclusion) 11 word 1 baseline 5 word 1 2 100%
CE 12 word 1 CRMTS
AC (baseline) 13 baseline 3 baseline 3 word 2 1
AC (exclusion) 14 word 2 baseline 6 word 2 2
CE 15 word 2 CRMTS
AC (exclusion) 16 word 1 word 1 baseline 3 1
CE 17 word 1 CRMTS
Table 9.1. Sample of a Sequence of Blocks of Trials for a
Typical Session for Teaching a New Word Set. (continued)
3. Training AC (control) 18 baseline 3 word 3 baseline 3 2
(AC + CE) AC (exclusion) 19 word 3 baseline 1 word 3 2
(Continue…d) ….
4. Exclusion AC 40 word 2 word 2 baseline 6 1 100%
Pos-test AC 41 word 1 word 1 baseline 4 1
AC 42 word 3 baseline 2 word 3 2
5. Discrim- AC 43 word 3 word 2 word 1 word 3 3

ination AC 44 word 2 word 3 word 2 word 1 2 100%
AC 45 word 1 word 1 word 3 word 2 1
…
6. Reading CD(c) 46 word 1 100%
Post-test CD 47 word 3
CD 48 word 1
CD 49 word 2
CD 50 word 3
CD 41 word 2
a – AC - M atching printed to dictated words
Baseline trials – both comparisons are printed words previously learned;
Exclusion trials – the sample is a novel dictated word; the array of comparisons display a novel printed word and a baseline printed word;
Control trials – a novel printed word is displayed as a comparison, but the sample (dictated) and the correct selection are baseline words.
b
– CE = Constructed response: constructing a printed word identical to the sample printed word
Chapter 9
c
– CD = reading (textual behavior)
d
– The total number of training trials varies with the number of exclusion trials with each new word; we have used four exclusion trials and two control trials for each word per session, plus the baseline trials (50%
of the sum of exclusion and control trials).
201
Conclusion
This chapter showed applications of stimulus equivalence or, more generally, derived rela-
tional responding, to teach reading and spelling. The first section showed how reading
and spelling can be thought of in connection with derived relational responding. Since
derived relational responding is usually generated through the matching-to-sample proce-
dure, the second section presented general and specific requirements for teaching match-
ing to sample in order to establish the stimulus equivalence that can give rise to reading
and spelling. The section also presented strategies for how to select contents and organize
teaching unit sequences to promote cumulative learning. Finally, we pointed out that
assessment of reading and its prerequisites, both before and during the curriculum imple-
mentation, is an important practice for providing objective measures of students’ needs
and their advancement along the curriculum.

Table 9.2. Verbal Relations Involved in Reading and Its
Immediate Prerequisites
MEASURES
VERBAL RELATIONS
Echoic behavior AD
Picture naming BD
Printed-word naming (textual behavior) CD
Dictation taking
- selection (constructed response) AE
- handwriting AF
Copying
- selection (constructed response) CE
- handwriting CF
MATCHING TO SAMPLE
Identity matching with pictures BB
Identity matching with printed words CC
(both stimuli in lowercase or both in uppercase) cc
Matching pictures to spoken words AB
Matching printed words to spoken words AC
Matching printed words to pictures BC
Matching pictures to printed words CB
Matching lowercase printed words to uppercase printed Cc
words and vice versa cC
Legend:
Stimuli
A = Spoken or dictated words
B = Pictures
C = Printed words (uppercase)
c = Printed words (lowercase)
Responses
D = Oral naming
E = Writing (constructed response)
F = Writing (handwriting)
Chapter 9 203

References
Adams, M. J. (1998). Beginning to read: Thinking and learning about print. Cambridge,
MA: MIT Press.
Birnie-Selwyn, B., & Guerin, B. (1997). Teaching children to spell: Decreasing consonant
cluster errors by eliminating selective stimulus control. Journal of Applied Behavior
Analysis, 30, 69–91.
Bradley, L., & Bryant, P. E. (1983). Categorising sounds and learning to read: A causal
connection. Nature, 310, 419–421.
Burgess, S. R., & Lonigan, C. J. (1998). Bi-directional relations of phonological sensitivity
and prereading abilities: Evidence from a preschool sample. Journal of Experimental
Child Psychology, 70, 117–141.
Crashaw, R. (1904). Bulla. In Steps to the temple: Delights of the muses, and other poems (p.
173). Cambridge, England: University Press.
Daniels, P. T., & Bright, W. (1996). The world’s writing systems. New York: Oxford
University Press.
De Rose, J. C., de Souza, D. G., & Hanna, E. S. (1996). ��
Teaching reading and spell-
ing: Exclusion and stimulus equivalence. Journal of Applied Behavior Analysis, 29,
451–469.
Dixon, L. S. (1977). The nature of control by spoken words over visual stimulus selection.
Dube, W. V., McDonald, S. J., McIlvane, W. J., & Mackay, H. A. (1991). Constructed-
response matching to sample and spelling instruction. Journal of Applied Behavior
Analysis, 24, 305–317.
Engelmann, S., & Carnine, D. W. (1982). Theory of instruction. New York: Irvington.
Goswami, U. (1986). Children’s use of analogy in learning to read: A developmental
study. Journal of Experimental Child Psychology, 42, 73–83.
Goswami, U., & Bryant, P. E. (1990). Phonological skills and learning to read. Hillsdale,
NJ: Erlbaum.
Greer, R. D., Chavez-Brown, M., Nirgudkar, A. S., Stolfi, L., & Rivera-Valdes, C.
(2005). Acquisition of fluent listener responses and the educational advancement of
young children with autism and severe language delays. European Journal of Behavior
Analysis, 6, 125–144.
Hanna, E. S., de Souza, D. G., de Rose, J. C., &. Fonseca, M. L. (2004). Effects of
delayed constructed-response identity matching on spelling of dictated words. Journal
of Applied Behavior Analysis, 37, 223–227.

Hayes, S. C. (1994). Relational frame theory: A functional approach to verbal events. In
S. C. Hayes, L. J. Hayes, M. Sato, & K. Ono (Eds.), Behavior analysis of language and
cognition. Reno, NV: Context Press.
Kinney, E. M., Vedora, J., & Stromer, R. (2003). Computer-presented video models to
teach generative spelling to a child with an autism spectrum disorder. Journal of
Positive Behavior Interventions, 5, 22–29.
Lee, V. L., & Pegler, A. M. (1982). Effects on spelling of training children to read. Journal
of Experimental Analysis of Behavior, 37, 311–322.
Lee, V. L., & Sanderson, G. M. (1987). Some contingencies of spelling. Analysis of Verbal
Behavior, 5, 1–3.
Lundberg, I. (1998). Why is learning to read a hard task for some children? Scandinavian
Journal of Psychology, 39, 155–157.
Mackay, H. A. (1985). Stimulus equivalence in rudimentary reading and spelling. Analysis
and Intervention in Developmental Disabilities, 5, 373–387.
McIlvane, W. J. (2007). Teaching some basic prerequisites for reading: Thirty years later.
Presented in the Symposium Progress Toward Effective Computerized Instruction of Early
Reading Skills at the 107th American Psychological Association Annual Convention,
San Francisco, CA.
McIlvane, W. J., & Dube, W. V. (2003). Stimulus control topography coherence theory:
Foundations and extensions. Behavior Analyst, 26, 195–213.
McIlvane, W. J., & Stoddard, L. T. (1981). Acquisition of matching-to-sample perfor-
mances in severe retardation: Learning by exclusion. Journal of Mental Deficiency
Research, 25, 33–48.
Melchiori, L. E., de Souza, D. G., & de Rose, J. C. (2000). Reading, equivalence, and
recombination of units: A replication with students with different learning histories.
Journal of Applied Behavior Analysis, 33, 97–100.
Mueller, M. M., Olmi, D. J., & Saunders, K. J. (2000). Recombinative generalization of
within-syllable units in prereading children. Journal of Applied Behavior Analysis, 33,
512–531.
Ribes, E. I. (1978). Tecnicas de modificacion de conducta: Su aplicacion al retardo en el
desarrollo [Behavior modification techniques: Applications to delayed development].
Mexico City: Trillas.
Saunders, K. J., O’Donnell, J. Vaidya, M., & Williams, D. C. (2003). Recombinative
generalization of within-syllable units in nonreading adults with mental retardation.
Journal of Applied Behavior Analysis, 36, 95–99.
Saunders, K. J., & Spradlin, J. E. (1990). Conditional discrimination in mentally retarded
adults: The development of generalized skills. Journal of the Experimental Analysis of
Behavior, 54, 239–250.
Chapter 9 205

Serna, R. W., Dube, W. V., & McIlvane, W. J. (1997). ��
Assessing same/different judg-
ments in individuals with severe intellectual disabilities: A status report. Research in
Sidman, M. (1977). Teaching some basic prerequisites for reading. In P. Mitler (Ed.),
Research to practice in mental retardation: Vol. 2. Education and training. Baltimore:
MD. University Park Press.
Sidman, M. (1994). Equivalence relations and behavior: A research history. Boston, MA:
Authors Cooperative.
Sidman, M., & Cresson, O. (1973). Reading and crossmodal transfer of stimulus equiva-
lence in severe retardation. American Journal of Mental Deficiency, 77, 515–523.
Sidman, M., & Stoddard, L. (1967). Effectiveness of fading in programming simultane-
ous form discrimination for retarded children. Journal of the Experimental Analysis of
Sidman, M., & Tailby, W. (1982). Conditional discrimination vs. matching-to-sample: An
37, 5–22.
Snow, C. E., Griffin, P., & Burns, M. S. (2005). Knowledge to support the teaching of
reading. San Francisco, CA: Jossey-Bass.
Staats, A. W. (1968). Learning, language and cognition: Theory, research and method for study
of human behavior and its development. New York: Holt, Rinehart and Winston.
Stoddard, L. T., & Sidman, M. (1967). Effects of errors on children’s performance on
a circle-ellipse discrimination. Journal of the Experimental Analysis of Behavior, 10,
261–270.
Stromer, R., & Mackay, H. A. (1992a). Delayed constructed-response identity matching
improves the spelling performance of students with mental retardation. Journal of
Behavioral Education, 2, 139–156.
Stromer, R., & Mackay, H. A. (1992b). Spelling and emergent picture-printed word
relations established with delayed identity matching to complex samples. Journal of
Stromer, R., Mackay, H. A., & Stoddard, L. T. (1992). Classroom applications of stimu-
lus equivalence technology. Journal of Behavioral Education, 2, 225–256.
Treiman, R. (1991). Children’s spelling errors on syllable-initial consonant clusters. Journal
of Educational Psychology, 83, 346–360.

Authors’ note
Preparation of this chapter was supported by FAPESP/PRONEX (Grant 2003/09928-
4), and by Conselho Nacional de Desenvolvimento Científico e Tecnológico (Brazilian
Research Council, Grants 550660/2002-2 and 350635/1991-9). We thank Elenice Hanna
for the support and cooperative work on reading acquisition. We also thank Larry
Williams, Harry Mackay, Julia Rocca, Lidia Postalli, and Thaize Reis for their valuable
comments on the text.
Chapter 9 207

Chapter 10
Syntax, Grammatical Transformation,

and Productivity: A Synthesis of Stimulus
Sequences, Equivalence Classes, and
Contextual Control
Harry A. Mackay, Northeastern University and Praxis Inc.;

and Lanny Fields, Queens College and the Graduate School
of the City University of New York
When an individual demonstrates linguistic competence, her utterances are syntacti-

cally correct and productive. Syntactic correctness means that the ordering of words in
an utterance is appropriate to the linguistic community, and productivity means that an
individual will produce syntactically correct utterances with a very large number of new
combinations of words and phrases. It has been said that behavioral principles and proce-
dures cannot account for the induction of behaviors that represent linguistic competence
(Chomsky, 1965; Tomasello, 2003). The material included in this chapter questions the
validity of this assertion, just as previous investigators have done (Dale, 2004; Goldberg,
1995; Guess, Sailor, & Baer, 1974; Mandell & McCabe, 1997; Sidman, 1994; Skinner,
1957). Although a substantial body of literature has grown out of Skinner’s (1957) analy-
sis of verbal behavior (Catania, 1998; Greer and Ross, 2004; Hayes, Barnes-Holmes, &
Roche, 2001; Horne & Lowe, 1996), relatively little empirical research has addressed
methods that induce syntactic repertoires.
In this chapter, we describe a minimal set of behavioral procedures that induce lan-
guagelike performances, which have the orderly properties of syntax and also show pro-
ductivity. As teaching methods, some of the procedures train individuals to respond to
and produce sequences of stimuli or symbols in orders that change appropriately across
ifferent contexts, just as language utterances do. For example, two different video clips
d
would be contextually appropriate for the statements “John pushed Mary” and “Mary
pushed John,” as written or as spoken. Another set of the procedures converts arbitrary
stimuli or symbols into conceptual categories called equivalence classes (for example,
“John” as written and as spoken, and “John” as shown in a picture). In combination, the
procedures that are used and the repertoires that are established can result in the emer-
gence of behaviors that have the essential properties of linguistic competence, syntax, and
productivity. Finally, the procedures that establish contextual control over productions
of different stimulus sequences can be used to account for many of the grammatical
transformations that characterize the flexibility of language usage. The procedures and
the forms of relational learning that they engender will be described fully below, together
with the kinds of productive syntactic repertoires they have been used to establish.
We emphasize that this chapter will focus on early aspects of language learning, par-
ticularly the kinds of simple, linear word sequences that emerge in the early phases of the
development of multiword utterances. These utterances include the adjective-noun and
subject-verb-object structures of English (cf. Ninio, 1998) as captured in phrases such as
“big boy,” “black cat,” “Bill hit Bob,” and “Joe ate corn.” That early repertoire provides
the foundation on which later complex, nested, recursive nonlinear forms of utterance (for
example, sentences that involve several phrases, each of which may include articles and
adjectives) are based. We expect that the procedures that engender the simpler forms of
syntax, productivity, and grammatical transformation will be useful for the teachers and
parents who strive daily to establish and improve language skills in children with autism
and other developmental disabilities.
Characterization of Syntax and Productivity

First, in a given language, the order of words plays a central role in communication and
understanding of information. For example, in English, the subject-verb-object (S-V-O)
construction is common. The subject, actor, or agent is stated as a noun at the beginning
of a sentence. It is followed by the action, which is stated as the verb. Finally comes the
object of the action, which is stated as a noun. Examples are “John hits Mike” and “Henry
drinks milk.” In contrast, if the words occur in a different order, the utterance may have
a different meaning, or it may not make sense in English (for example, “Mike hits John”
and “Henry milk drinks,” respectively). The meanings of utterances, then, depend on the
order of the constituent words. How might current behavior analytic concepts and proce-
dures deal with the repertoires that involve such complex conditional control?
Second, another important aspect of a syntactic performance involves the demonstra-
tion of productivity. This means that, after an individual has learned a few utterances, all
of which have a given syntactic structure, a very large, perhaps infinite, number of new
utterances will emerge without direct training; each will contain a new set of words, and
they will be produced in an appropriate order. When that occurs, one concludes that the
individual has acquired the syntactic frame and can produce any number of novel and
grammatically correct sentences. For example, assume that the names of any individuals
can be substituted for John or Henry (including each other) in the examples given above,
innumerable verbs can be substituted for “hits” or “drinks,” and any number of object

names can be substituted for “Mike” and “milk” (for example, “Henry hits John,” “John
drinks water,” “Billy sings songs,” and “Mary pushes Jane”). All of these syntactically
correct utterances are said to be emergent because they occur in the absence of direct
feedback. And what about the use of “Henry pushes Mary” (or “John” or “Billy”), when
none of the latter three names ever served as an object in an utterance? How does such a
flexible repertoire come about? Later, we will illustrate and discuss procedures that have
produced such flexibility. Note, however, that some substitutions that follow the S-V-O
order can yield combinations that are meaningless (for example, “John hits songs”). It is
beyond the scope of this chapter to consider the conditions that enable an individual to
discriminate between syntactically correct utterances that are and are not meaningful.
Semantic Relations
There is now considerable basic and applied research in behavior analysis that has focused
on methods that give rise to novel performances under appropriate circumstances without
direct training. For example, studies of equivalences between pictures and their printed
and spoken names (heard and said by the student) have begun to clarify the bases of
rudimentary reading, spelling, and other repertoires (Mackay, 1985; Mackay & Sidman,
1984; Stoddard, Brown, Hurlbert, Manoli, & McIlvane, 1989). To illustrate, after a child
has been taught to match pictures and their printed names to their corresponding names
as dictated orally by a teacher, two new visual matching performances (matching pictures
to printed words and printed words to pictures) typically emerge without further training.
The occurrence of these performances without direct training demonstrates the emergent
interchangeability of the stimuli used in the training and testing. It is that interchange-
ability that may give rise to the idea that the words, whether printed, heard, or said, “rep-
resent” or “stand for” the pictured object—in other words, relations among these stimuli
provide a basis for saying that the words have acquired “meaning” or are semantically
related to each other. Equivalence and semantic relations will not be reviewed in detail
here since they are considered elsewhere in this volume (see chapters 7 and 8).
Syntax and Productivity: Novel Sequences from

Elements of Classes
The examples of syntactic repertoires mentioned earlier in the chapter emphasize aspects
of language. The position of events in sequences, however, is also critical for performances
in many other areas such as mathematics, music, and dance. Thus, the establishment of
numerical sequencing is akin to the establishment of a syntax. To illustrate, assume that
the number words “one” through “nine” can be spoken in sequence and that the numbers
can be represented in several ways (for example, as printed numerals and printed words).
Productivity could be demonstrated by the sequencing of the printed words or the printed
numerals without direct training. This possibility was explored by Mackay, Kotlarchyk,
and Stromer (1997). Their study described how novel sequences of numerical representa-
tions were induced by the joint effects of a preexisting repertoire of saying numbers in
Chapter 10 211

order; the formation of equivalence classes that included different representations of the
numbers; and restricted sequence training. The student in the study, Jay, was ten years old
and had cerebral palsy as well as severe intellectual disabilities. He had learned to recite
aloud the oral number names “one” through “nine” in appropriate order before he entered
the experiment. This performance was probably learned as a rote chain. (Note: He never
used “zero.”) He did not write because of his severe motor difficulties. He could, however,
manipulate small cardboard tiles that contained numerals, number words, and printed
letters. The letter tiles were used in a modified version of the “anagram construction”
(spelling) task used by Mackay (1985) to teach relations between colors and their printed
names. With this task, Jay learned to produce number words by constructing sequences
of individual letter tiles but was not required to form the printed letters involved. Tiles
with the numerals or their names printed on them were used in tasks that involved the
sequencing of these stimuli.
Figure 10.1 outlines the study, which involved ten potential classes of numeric
stimuli, each of which included three representations of each of the numbers from zero to
nine—in other words, the printed numerals (for example, 0; box marked B), their printed
names (zero; C), and their teacher-dictated names (“zero”; A). The dictated names were
auditory stimuli that were presented to Jay. (The term dictated numeral names does not
refer to the oral names [D] that were spoken by Jay.)
B
Printed
numerals
A
Dictated
numeral
names
C
Printed
numeral
names D
~~~~~~~~~~
E Construct Oral
numeral
naming
names
Figure 10.1. Schematic outline of the experiment. Boxes A, B, and C represent stimuli.
Arrows indicate a set of conditional relations and point from sample to comparison stimuli
of matching tasks (AB, AC, BC and CB). Box D represents oral names spoken by the
student to numeral samples (BD) and to printed number names (CD). Box E represents
letter-by-letter construction of the number names of numerals (BE) and of the dictated
numeral names (AE). Solid arrows mark relations the student has already learned (AB,
BD) or is explicitly taught (AE). Broken arrows identify relations that emerge after others
have been taught. See text for details.
Jay’s performances were assessed using the eight tasks illustrated schematically in
figure 10.1. Four involved matching-to-sample procedures and might be considered tests

of different “receptive” repertoires. Teachers examined whether Jay could match the
printed numerals (B) and their printed names (C) to the dictated names (A) supplied by
the teacher (ten AB relations and ten AC relations, respectively), the printed names to the
numerals (ten BC relations), and the numerals to the printed names (ten CB relations),
yielding a total of forty relations among stimuli. In each trial of these matching tasks, one
stimulus from a (potential) class served as the sample (for example, the printed numeral
6 in task BC) and other stimuli served as comparisons, one from the same class (six) and
one from a different class (for example, two). The performance considered correct on each
trial was the selection of the comparison belonging to the same class as the sample (the
word six in the preceding example). The remaining four tests examined different “expres-
sive” repertoires. They assessed whether Jay orally named the numerals (BD) and printed
words (CD), and whether he constructed the printed words when presented with the
numerals or the dictated names as samples (BE and AE, respectively).
Initially, Jay only matched numerals to dictation (AB), thus demonstrating one of
the four receptive repertoires tested. However, he did not perform the other three match-
ing tasks, AC, BC, and CB, that involved printed words as samples or comparisons.
Apparently he had learned only ten out of the forty stimulus-stimulus relations included
in the matching tasks. Thus, there was no evidence of classes of equivalent stimuli that
could be substituted for one another. In addition, Jay orally named the numerals (BD),
only one of the four expressive repertoires tested. Thus, he showed only ten of the twenty
naming relations that were possible and none of twenty word constructions.
With regard to the sequencing of the numbers, Jay said the sequence “one” through
“nine,” as noted earlier. However, he did not place the numerals or the printed words in
correct left-to-right sequence. The absence of the latter two performances could have been
attributed to failure to understand instructions for the general sequencing task. However,
that was not the case—he demonstrated that he understood these instructions by placing
tiles side-by-side on the response line provided.
Following the assessments, Jay was trained to use the individual letter tiles to con-
struct the printed words when presented with the dictated names as samples (AE). The
words “zero” through “nine” were trained one at a time in an unsystematic order. Figure
10.2 illustrates the program for the word one. In the first trial (step 1), the dictated sample
presentation was followed by an opportunity to copy the word one—in other words, to
place the letters O, N, and E on the identical letters on the program card. Then, in the
remaining steps, the letters to be matched (or “copied”) were gradually removed from the
program cards one at a time. For the final performance (step 4), the instructor provided
the dictated sample “one”; Jay had to select O, N, and E from a choice pool of letters and
place them in that order on the line provided on the program card. Other programs (for
example, for the word seven) took more steps to complete.
After this training, Jay matched the whole printed words to dictation (task AC) for
the first time and all of the other relations emerged even without direct training. In
these emergent performances, he matched the printed names to the numerals (BC) and
numerals to the printed names (CB). Jay also orally named the printed words (CD),
which was akin to a rudimentary form of oral reading. He also constructed the printed
words when presented with the dictated names (AE). Thus, after Jay learned to con-
struct the ten number words in the task AE, five new performances emerged. Since each
performance involved ten relations between stimuli, fifty new relations were established
without direct training. These emergent performances implied that all ten three-member
Chapter 10 213

N O
F I
ONE
Step 1 V
E
ON _ _
Step 2 ONEN E
O ON E
O
Step 3 O E
ONE
Step 4
Figure 10.2. Illustration of cards and tiles used in successive steps in a program for teach-
ing construction of the word one from its dictated name “one.”
equivalence classes had formed, thus systematically replicating earlier findings (Dube,
McDonald, McIlvane, & Mackay, 1991; Lazar & Mackay, 1982; Mackay & Sidman,
1984; see chapter 8 of this volume).
The preceding section dealt with the procedures that established the discriminative
repertoires critical for the formation of the classes of numeric stimuli. At this point in the
study, then, it became important once again to address questions about the ordering of
the different representations of each number. Would all the members of each equivalence
class substitute for one another and act in the same way in sequence tasks, thus suggest-
ing that each set of representations of the same number was also a functional class? For
example, would the numeral and printed word in each class (for example, 0 and zero, and
1 and one) occupy the same position in tasks that required sequencing of these stimuli?
Jay’s performances on sequence tests answered these questions. He continued to recite
the numeral names “one” through “nine” in order, just as he did before the training.
However, it was surprising to note that he still did not place the corresponding numerals
or words in correct left-to-right sequences even though each printed numeral and each
printed word had become members of the same equivalence classes as the dictated name
and the oral name recited by Jay.
At this point in training and testing, then, the inclusion of the printed numerals
and names in these equivalence classes did not guarantee that the members of each class
would substitute for one another in the same position in different sequences. Thus, the
oral and visual sequencing performances were independent of each other rather than com-
ponents of an integrated repertoire; in other words, the equivalence classes were not also
functional classes. This lack of integration was explored further. Did Jay fail the ten-
stimulus sequencing tests because of the length of the sequences? To answer that question,

the researchers conducted a restricted sequence test with only four numerals: 0, 1, 2, and
3. Because Jay also failed this test, it was concluded that failure on the ten-stimulus tests
was not due to their length.
The sequencing of the ten numerals and ten words was induced with the following
intervention: After he failed to sequence the limited set of numerals, the anagram fading
procedure described above was used to train Jay to construct the appropriate 0-1-2-3
sequence of numeral tiles. On every trial, Jay placed the tiles in the correct sequence and,
it is interesting to note, named the numerals aloud as he placed them, even though he was
not required to do so. Immediately after this limited training, he appropriately sequenced
all ten numerals and also all ten printed words. Notably, after placing each numeral or
word he pointed to and named it. Furthermore, he began each of the visual sequences
with the numeral 0 or the word zero, respectively, placed at the left. To summarize,
membership of all three stimuli in the same class, and brief sequence training with one of
them, yielded generalized performances involving all three. Clearly, the generative power
of the general protocol was large and impressive.
Explanations for Jay’s Performances

Jay’s initial oral sequencing of number names from “one” to “nine” was most likely a
specific rote chain established by an unknown mixture of imitation and differential rein-
forcement. It could also be argued that a chaining account may apply with respect to the
limited training that established the numeral sequence 0-1-2-3; each successive response
(for example, placement of 0 at the left) may have simply produced the specific discrimi-
native stimulus for the next response (selection of the numeral 1, and so on). However,
the emergent performances described above clearly require far more than a traditional
chaining account.
The following alternative provides a much more plausible account of the behavioral
processes responsible for the induction of the extended syntaxlike repertoire that Jay came
to demonstrate. Two different aspects of the emergent performances he displayed are of
interest: the sequencing of the stimuli that were not included in the limited training set
(numerals 4 through 9 and all ten number words), and the location given to two of the
stimuli in the zero class without direct training.
Initially, in addition to reciting number names in order, Jay only matched numerals
to their dictated names and named the numerals. Then, after virtually errorless anagram
(spelling) training, the emergence of five novel matching, spelling, and naming perfor-
mances (depicted in figure 10.1) demonstrated the formation of fifty new relations among
the corresponding numeric stimuli involved. Next, limited training with the printed
numerals 0, 1, 2, and 3 established the left-to-right sequence of these stimuli. The short
sequence thus provided a critical exemplar of the integrated performance that immediately
gave rise to the more generalized sequence performances with the remaining numerals, 4
through 9, and the equivalent printed number words. The emergence of sequential rela-
tions among the members of the numeric classes thus involved the combined effects of the
oral naming and sequencing repertoire taught to Jay prior to the study; the substitutability
of the stimuli in the experimentally induced, numeric equivalence classes; and the brief
Chapter 10 215

training of a short sequence of numeral tiles. These outcomes illustrate the expansion of a
complex repertoire typically ascribed to transfer of training.
A different combination of stimulus class formation and sequencing performances is
involved in the locating of the stimuli in the zero class. Recall that Jay never used “zero” in
reciting number names in order before he received the training that established the prereq-
uisites of the numeric equivalence classes and then the limited 0-1-2-3 sequence training.
After that training, however, he began each sequence with “zero” instead of “one.” The
0-1-2-3 training, which included the numeral 0, sufficed to locate the spoken word “zero”
and the printed word zero as initial members of the sequences to which they belonged. The
training procedures used with Jay then had the beneficial effect of integrating the spoken
numerical sequencing with the sequencing of written numbers and words.
Toward Comprehension: Adding Quantity-Based

Exemplars to Classes
The results of other research with individuals with intellectual disabilities add to the edu-
cational implications of the study with Jay (Gast, VanBiervleit, & Spradlin, 1979; Green,
1993; Maydak, Stromer, Mackay, & Stoddard, 1995). For example, the same conditional
discrimination procedures that were used to establish the classes could also be used to
expand an individual’s repertoire substantially by adding number words of a second lan-
guage, or numbers expressed as Roman numerals or in binary notation. The same pro-
cedures have also been shown useful in teaching the equivalences among fractions and
decimals represented in different ways, such as 1/2, 0.5, and a picture of a square with one
half filled. (See Leader & Barnes-Holmes, 2001, and Lynch & Cuvo, 1995, for relevant
work on equivalences among stimuli such as these.)
Another expansion could involve the addition of sets of dots as members of numeral-
based equivalence classes, such as two dots for the class two, and three dots for the
class three (Maydak et al., 1995). Such stimuli could be readily added to those used
with Jay (numerals and their printed and dictated names). Each equivalence class might
be expanded by establishing a relation between a quantity and a numeral (for example,
matching two dots to the numeral 2). The addition of these stimuli to the repertoire
already displayed by Jay would greatly increase the number of relations (correspondences
among the numeral, name, and quantity stimuli) involved (see Fields, Verhave, & Fath,
1984, and Fields & Verhave, 1987, for relevant discussion). The introduction of these
stimuli also would permit the assessment of performances that involved the ordering
of actual quantities and other stimulus sets of different sizes (for example, pictures of
apples, three actual toy cars; cf. Green, 1993). That kind of assessment would examine
the development of the child’s understanding of ordinal relations between quantities. The
matching-to-sample and sequencing procedures described earlier could then provide the
framework for the development of instructional materials designed specifically to establish
and extend these relational repertoires in individual children.
Yet a third expansion involves the generalization of the newly induced repertoires
across stimuli that resemble those that have become members of the equivalence classes.
Recall that each of the stimuli in an equivalence class remained physically the same across
presentations. Such perceptual invariance, however, is not what individuals experience

in natural environments. For example, although the two-dot stimulus that was added to
the equivalence class of two might have been red, bright, and one-half inch in diameter,
dots encountered in natural settings would vary in size, color, and shape, among other
features. Thus, it is important that the relations that emerge among the stimuli in the
equivalence class also hold in the face of the inevitable variations in stimuli that will be
encountered in the real world (Branch, 1994; Fields, Reeve, Adams, & Verhave, 1991;
Galizio, Stewart, & Pilgrim, 2004).
To obtain this sort of generalization between variants of a member of an equivalence
class and the other members of the equivalence class, it is first necessary to determine
whether the variants of the stimulus that is a member of the basal equivalence class are
interchangeable with that class member. That is, dot pairs that vary in size, relative posi-
tion, color, and/or shape in any combination should result in the selection of the dot pair
that is a member of the equivalence class, while sets of four dots that vary in size, relative
position, color, and/or shape in any combination should result in the selection of the set
of four dots that is a member of the equivalence class. That sort of primary generalization
should not be taken for granted (Stokes & Baer, 1977). Documentation can be obtained
by conducting a test in which all of the variants mentioned above are presented as sample
stimuli, with the two-dot and four-dot members of the equivalence classes as the com-
parison stimuli. Trials with some of these variants are illustrated in figure 10.3. If most
of the two-dot variants resulted in the selection of the two-dot comparison, the variants
would be acting as members of the perceptual class of two dots. Likewise, all of the
four-dot variants would be acting as members of the perceptual class of four dots (Fields,
Reeve, et al., 2002; Reeve & Fields, 2001).
The rationale for using variants such as those illustrated in figure 10.3 is as follows:
It is possible that an individual could have inadvertently learned that the particular sizes,
relative positions, and particular shapes of the two-dot and four-dot stimuli were critical
features of twos and fours. That possibility would be evaluated by looking at errors that
could occur in the above-mentioned primary generalization tests. For example, if a child
had inadvertently concluded that the absolute size of the dots was a component of the
concept of two, she would respond correctly to the first, second, and fourth displays, but
not to the third and fifth displays (and, in all likelihood, not the sixth display.) Likewise,
if another child had inadvertently concluded that the shape of the objects was a defining
component of four, she would respond correctly to all of the displays except the last one.
The same would be true for the other variants used as sample stimuli. In contrast, the
selection of the two-dot comparison in the presence of all of the two-element samples in
figure 10.3 can be obtained only if the individual is attending to the numerosity of the
objects in the sample and disregarding the irrelevant features of the dots, such as size,
shape, color, and relative position.
After the existence of the perceptual classes has been documented, it may be possible
to test for the emergence of new relations among the variants and the remaining members
of the equivalence classes. This would be accomplished by the presentation of the variants
as samples, with comparisons that consisted of the word pair two and four, or the numeral
pair 2 and 4. When done, it is most likely that two or 2 would be selected on trials in
which the sample was any two-dot variant, and four or 4 would be selected on trials in
which the sample was any four-dot variant. These performances would indicate that the
equivalence classes merged with perceptual classes to form generalized equivalence classes
of two and four (Fields et al., 2007; Fields, Fitzer, Tittelbach, & Shamoun, 2005; Fields,
Chapter 10 217

Figure 10.3. Variants of the members of two-dot and four-dot equivalence classes.

Reeve, Adams, Brown, & Verhave, 1997; Rehfeldt & Root, 2004; Lane, Clow, Innis, &
Critchfield, 1998). In addition, the emergence of these performances would support the
view that the relations learned during the formation of equivalence classes should show
appropriate generality across the infinite variation of events that may be encountered in
real-world settings. Similar effects have been reported in a number of laboratory-based
experiments with stimuli that are simpler than those described above (Fields et al., 2007;
Fields et al., 2005; Belanich & Fields, 2003; Fields, Matneja, et al., 2002; Fields, Reeve,
et al., 2002; Fields & Reeve, 2001, 2000; Fields et al., 1997). Given the breadth of stimu-
lus domains used in those experiments, however, it is likely that the outcome of the theo-
retical experiment would occur as predicted.
It is also possible, of course, that very narrow ranges of variants would produce the
selection of the two-dot and four-dot stimuli that were members of their respective equiv-
alence classes. In that case, those restricted ranges of variants that constituted each of the
perceptual classes would not enable a person to demonstrate generalization to the broad
range of variants encountered in natural settings; in other words, the restricted width of
the perceptual class would not be of adaptive utility. Given such an outcome, the range
of variants that functioned as class members could be broadened to a socially appropriate
scope by the use of either forced-choice generalization testing (Reeve & Fields, 2001), or
by multiple-exemplar training (Fields & Reeve, 2000, 2001; Fields, Reeve, et al., 2002;
Mackay, Stromer, & Serna, 1997). Once the width of a perceptual class had been broad-
ened sufficiently by forced-choice generalization testing or by multiple-exemplar training,
the relations between the variants and the other member of an equivalence class could be
measured, as noted above.
“Grammatical” Classes from Materials in Sequences

The behavioral phenomena described with regard to numerical sequencing bears a close
resemblance to the syntactic productivity that could be demonstrated by an individual
who had learned to produce only a few sentences with an S-V-O structure (for example,
“John drinks milk,” “Mary gets water”). An elementary form of productivity would be
demonstrated if the individual substituted the different proper names for one another,
to yield “Mary drinks milk” and “John gets water.” Similarly, interchanging the verbs
“drink” and “gets” would yield “John gets milk” and “Mary drinks water,” and inter-
changing the nouns that are in the third position of the original sentences would yield
“John drinks water” and “Mary gets milk.” More generally then, two or more separate
stimulus sequences (such as sequence A1-A2-A3 and sequence B1-B2-B3) may be trained
directly using one of the procedures described in this chapter. Following the completion
of training, tests can be given to determine whether classes of the stimuli that occupy the
same ordinal positions (for example, first, second, third) in the different sequences have
come to function as members of the classes firsts, seconds, and thirds. That is, after train-
ing, do A1 and B1 form a class, A2 and B2 another class, and so on, similar to “John”,
and “Mary”; “gets” and “drinks”; and “milk” and “water” from the examples provided
above? It is worth noting that the classes may be based on their sequence in the phrase
rather than grammatical classes like proper names, verbs, or nouns (for further consider-
ation, see Meehan & Fields, 1995; cf. also Catania, 1998).
Chapter 10 219

Given these possible outcomes, however, one might ask the following questions: Can
such classes be established via acquisition of sequences of stimuli, including words? That
is, will stimuli that occupy the same position in different sequences come to function as
members of the same class? In addition, can new class members be added by training
with additional sequences, thereby increasing class size? For example, would teaching the
new sequence “Bill likes soda” add “Bill” to the class firsts, “likes” to the class seconds,
and “soda” to the class thirds? Several studies have addressed these questions (Barnes,
Mackay, & Stromer, 2008; Sigurdardottir, Green, & Saunders, 1990). Sigurdardottir
and colleagues (1990) first trained adults to produce four sequences (sequences A, B,
C, and D) that each involved four different stimuli, or forms (in other words, A1-A2-
A3-A4, B1-B2-B3-B4, C1-C2-C3-C4, and D1-D2-D3-D4). In subsequently conducted
matching-to-sample tests, adults matched the forms that had occupied the same positions
in the sequences. For example, A1, the initial form in sequence A, was presented with B1
and B2, B3, or B4 as the comparisons in one test; C1 and C2, C3, or C4 as the com-
parisons in another test; and D1 and D2, D3, or D4 as the comparisons in a third test.
In these tests, the forms B1, C1, and D1 served as the positive comparisons because they
shared the same ordinal position as the form A1. Likewise, A2 was matched with B2, C2,
and D2, and so on. These results demonstrated that the stimuli in a given ordinal position
in multiple sequences can come to function as members of a class of stimuli, consistent
with our previous examples.
It is informative to compare the training given to these adult subjects and that to
which Jay was exposed in the number experiment described earlier. These adults were first
taught to construct sequences and then given matching-to-sample tests to assess whether
the sequence training established classes of stimuli that occupied the same position in dif-
ferent sequences. In contrast, recall that Jay already recited number names in order before
the experiment began and that matching-to-sample tasks were then used to establish the
classes of equivalent numeric stimuli that were critical for the expansion of his number
sequence repertoire. The combined results of these studies thus suggest that the order
in which class formation and sequence training are conducted may be irrelevant to the
induction of the productive, albeit elementary, syntaxlike repertoires described here. In
other words, the same outcomes might be expected if students were first taught stimulus
sequences and then given tests to assess the formation of classes, or students first learned
the stimulus classes and then received sequence training. The relevant research has not
been done. However, research by Maydak and colleagues (1995) suggests that, for some
individuals with intellectual disabilities, sequence training may not yield the immedi-
ate addition of the stimuli to already established equivalence classes of numeric stimuli
(numerals, words, dot quantities, and forms). Research is needed in order to examine the
conditions that may produce independence of such repertoires.
Cross-Sequence Substitutions and New

Syntactic Utterances
Another method for examining whether stimuli from the same ordinal position in differ-
ent sequences substitute for one another involves tests for the emergence of new sequence
productions rather than new matching-to-sample performances. The sequences shown at

the bottom of figure 10.4 illustrate such emergent outcomes after the training of the two
five-stimulus sequences A1–A5 and B1–B5. The untrained sequences involve mixtures of
the stimuli from the A and B sequences and included single substitutions (for example,
A1-B2-A3-A4-A5; A1-A2-A3-B4-A5), and multiple substitutions (for example, A1-B2-
A3-B4-A5). Such performances have been shown in participants with and without devel-
opmental disabilities (Galy, Camps, & Melan, 2003; Lazar, 1977; Stromer & Mackay,
1993; see also Green, Stromer, & Mackay, 1993). The mutual substitution of stimuli in
mixed sequences also confirms that the training established classes of the stimuli that
occupied the same positions in training sequences. These classes may consist of more
than two stimuli (for example, following training with more than two sequences; Stromer
& Mackay, 1993). To illustrate, suppose that an individual who had learned to produce
“John gets milk” and “Mary drinks water,” as described above, was next trained to
produce “Bob likes soda.” The mixed tests described here would then show whether the
training sufficed to yield novel syntactically correct mixtures such as “John likes soda,”
“Bob drinks water,” “Mary likes milk,” and so on.
Two sequences trained
A1 A2 A3 A4 A5
B1 B2 B3 B4 B5
Emergent sequence probes
A1 B2 A3 B4 A5
B1 A2 A3 B4 A5
Figure 10.4. Top: Schematic depicting training of two sequences, sequence A and sequence
B. Alphanumeric notation (for example A1, A2) represents forms displayed as stimuli.
Bottom: Two mixtures of the A and B sets of stimuli presented on probe trials that
examine emergent performances.
Another form of this cross-sequence substitution task does not require training in
construction of the initial sequences and may be used to illustrate both the interchange-
ability of words of a given type or class (such as noun, verb, or adjective) and word order
(cf. Bruner, 1975). For example, suppose that the three initial sentences of the previous
example are presented visually on a blackboard or computer screen as follows:
John gets milk.
Mary drinks water.
Bob likes soda.
Chapter 10 221

Next, the child is trained to construct new utterances by selecting, in sequence, a
word from each column (for example, “Mary gets soda” or “Bob drinks milk”). Would
the child then also construct additional syntactically correct utterances (such as “John
likes water” and “Bob gets soda”)? Would the same behavior continue after addition of a
new exemplar, such as “Tom eats candy”? Longer sentences can be introduced by adding
the article “the” before the final noun in each sentence, although the article “a” cannot be
added appropriately to all the sentences used. Behavioral methods for teaching appropri-
ate use of a particular article remain to be developed.
Overlapping Sequences and Extended

Syntactic Repertoires
Initial development of syntactic aspects of language involves the ordering of two words
in utterances (Brown, 1973). The length of the utterances then increases gradually. One
way in which the gradual extension of utterances may come about involves the blending
of short word sequences as in the following utterances of a child who loses grasp of a ball
rolled toward him by his mother. Suppose that on some occurrences of this situation, the
child says, “Mommy get,” and his mother retrieves and hands him the ball, a likely form
of reinforcement. On other occasions, the child says, “Get ball,” and the same potentially
reinforcing events follow. It seems predictable that the three-word statement “Mommy get
ball” will occur on another, later occasion during which a ball is dropped (cf. Holcomb,
Stromer, & Mackay, 1997). Note that this need not necessarily reflect emergence of the
new utterance. Direct training might have occurred, for example, if the mother made a
statement like “Okay, Mommy will get the ball” (an expansion of the child’s two-word
statements) while retrieving the ball. The child’s three-word statement thus may reflect
what a linguist might call the “privileges of occurrence” for the words. Even if the skill
is trained, however, the history of experiences described above can also quickly develop
extensive generality; a child who has learned the names of other people (such as “Daddy,”
“Peter,” and “Gran”) and other toys (such as “car” and “crayon”) may soon demonstrate
their substitutability in many more utterances with the same structure as “Mommy get
ball” even without direct training. Data that confirm this analysis were presented by
Cullinan, Barnes, Hampson, and Lyddy (1994).
Syntactic Transformations: Embedded

Contextual Control
Syntactic performances frequently involve changes that occur in word order across differ-
ent linguistic contexts. This can be illustrated with the simple grammatical transforma-
tion of a descriptive noun phrase to a descriptive sentence, as suggested by Mackay (1991).
Suppose that a nonverbal child is exposed to the following trial-by-trial procedure that
involves a picture of a red truck, a picture of a green ball, and several cards each contain-
ing one written word (the, a, red, green, ball, and truck). As shown in table 10.2 (train 1),
on some trials the child is presented with the picture of the red truck along with the cards

containing the words the, red, and truck. On other trials (train 2), the child is presented
with the picture of the green ball and the cards containing the words a, green, and ball. In
both cases, the spatial locations of the word cards presented are changed unsystematically
from trial to trial. Further, reinforcement contingencies are designed to teach the child
to place the cards in sequences that correspond to the syntax of a typical English noun
phrase. In the presence of the red truck, reinforcement is provided for placing the, then
red, then truck left to right in that order. Likewise, when the child is shown the picture
of the green ball, reinforcement is provided for placing a, then green, then ball in order.
Placing the word cards, then, is the analogue of uttering the noun phrase that describes
the pictured object.
Table 10.2. Training and Testing for Embedded

Contextual Control of Syntactic Transformations
Stimuli Presented
Picture sample Words displayed Correct construction
Train 1. Red truck truck the red The red truck
2. Green ball green ball a A green ball
3. Red truck truck is red the The truck is red
Test: Green ball is ball a green A ball is green

(Teach if necessary.)
Train 1. Big dog dog big the The big dog

2. Small cat a cat small A small cat
3. Big dog is dog the big The dog is big
Test: Small cat cat is small a A cat is small
(Teach if necessary.)
Next (train 3), the child is presented with the picture of the red truck and four cards
that contain the words is, red, the, and truck and is trained to place the cards in the
sequence The truck is red. This training establishes a sequence that is analogous to utter-
ing a declarative sentence. Notably, in contrast to the earlier training with three cards,
this new sequence requires the reversal of the order of red and truck in the presence of the
fourth card that contains the word is. Now, if appropriate productions of the three trained
sequences are maintained in the presence of three and four cards, the word is clearly pro-
vides trained discriminative stimulus control over occurrence of the different sequences,
yielding the noun phrase “the red truck” in the absence of is and the sentence “The truck
is red” in the presence of is.
Chapter 10 223

Confirming that the word is functions more broadly as a contextual cue for produc-
ing the sentence form instead of the noun phrase requires an additional step with two test
trials. In one, the child will be presented with the picture of the green ball and the words
ball, green, a, and is, a stimulus set that has not been used in training. In the other, the
child will be presented with the same picture and the words ball, green, and a, as in earlier
training. Will the child immediately construct the novel sequence A ball is green without
further training when shown four words, one of which is is? Will the child also continue
to construct the sequence a green ball when shown those three words, as in training, and
the other two previously trained sequences? If so, the presence and absence of the word is
may be said to be contextual cues rather than merely discriminative stimuli that control
the different trained orders of occurrence of the words red and truck. However, if the
appropriate sequences (a green ball and A ball is green) are not constructed on the test
trials, then they should be trained directly. Subsequently, additional training would estab-
lish new exemplars of the same grammatical forms, such as those displayed in the lower
part of table 10.2. Such training would then continue until successful test performances
demonstrated generalized performance with new exemplars.
The following study provides an analogue of the type of contextual control that may
be involved in transformations of word order that emerge during typical language devel-
opment (Holcomb, Barnes, & Mackay, 1991). The participant in this study was a thirty-
nine-year-old man with moderate intellectual disabilities (his Peabody Picture Vocabulary
Test age-equivalent score was six years, three months). He had participated in previous
studies that involved conditional discrimination and sequence learning conducted with a
computer. He was also familiar with the use of tokens and their exchange for money.
The left part of figure 10.5 outlines the experiment, in which the actual stimuli were
tiles on which different forms were printed. The forms are represented in the left part of
figure 10.5 by the alphanumeric designations A1 and A3, O and X, B1 and B3, and C1
and C3. At the right in figure 10.5, English words are listed in the same way as the alpha-
numeric representations, to link the experimental operations and outcomes to potential
syntactic transformations. Thus, the stimuli designated A1 and A3 correspond to Bill and
cake, respectively, the stimuli O and X correspond to ate and its related passive form, was
eaten by, B1 and B3 correspond to Bob and corn, and C1 and C3 correspond to Mary
and egg. The order of the training and testing steps are those actually used in the experi-
mental arrangements. No implications are intended with regard to a typical or necessary
sequence of language development.
The first sequence to be established was A1-O-A3 (English analogue: Bill ate cake).
Initially, only A1 appeared on the screen. After the subject touched it, the second stimu-
lus (designated O) was added to the display. After the second stimulus was touched, the
screen flashed, the computer played a musical jingle, and a token was delivered. In the
next training step, each trial began with the stimuli A1 and O appearing simultaneously.
The positions of the stimuli in the display changed unsystematically from trial to trial.
After the subject touched them reliably in the trained order, the third stimulus, A3, was
added to the display, so that all three stimuli appeared simultaneously at the start of each
trial. Note that if the English words Bill, ate, and cake had been used in the training
rather than the forms, the training would have established the performance of touch-
ing the words in that order (shown at the top right in figure 10.5), regardless of their
positions in the display. The A1-O-A3 sequence then is analogous to a sentence with an
S-V-O structure. The three-term sequence was learned in one session with only one error

------------------------------------------------------------------------------------------------------------------------------------
Train Test Relations
Stage /Test Type Trained and Tested English Equivalents of Relations Trained and Tested
==============================================================================
1. t A1 O A3 BILL ATE CAKE.
2. t A1 O B3 BILL ATE CORN.
B1 O A3 BOB ATE CAKE.
3. * BoB B1 O B3 BOB ATE CORN.
4. t B3 X B1 CORN WAS EATEN BY BOB.
5. * AxA A3 X A1 CAKE WAS EATEN BY BILL.
6. t B3 X C1 CORN WAS EATEN BY MARY.
t C3 X B1 EGG WAS EATEN BY BOB.
7. * CxC C3 X C1 EGG WAS EATEN BY MARY
8. * CoC C1 O C3 MARY ATE EGG.
9. * CoA C1 O A3 TOM ATE CAKE

* AoC A1 O C3 BILL ATE EGG.
* AxC A3 X C1 CAKE WAS EATEN BY MARY.
* CxA C3 X A1 EGG WAS EATEN BY BILL.
------------------------------------------------------------------------------------------------------------------------------------
t = directly trained; B = substituted; * = emergent
Figure 10.5. Schematic outline of successive stages (1 through 9) conducted during train-
ing of three-form sequences directly (t) and by substitution (B). Untrained, emergent
performances are marked by * and identified as new combinations of the A, B, and C
stimuli (AOA, BOB, COC respectively). The list at right illustrates English analogues of
the performances with forms. See text for details.
o ccurring through all the steps in training. That rapid acquisition and high accuracy
can be attributed to the programmed introduction of stimuli in the sequence (Mackay,
Stoddard, & Spencer, 1989).
The next part of training used an exclusion procedure (Mackay et al., 1989; see also
chapter 8) to establish a new sequence, B1-O-B3 (Bob ate corn). Initially (stage 2), B3
replaced A3 in displays that also presented the other two stimuli, A1 and O, of the
sequence already trained. The expectation was that A1 and O would control responding
in the first two sequence positions because of the earlier training with the untrained B3
occurring last in the sequence. Thus, touching the stimuli in the order A1-O-B3 was
considered correct and reinforced. This new sequence was produced immediately and
without error. Trials that contained the newly learned A1-O-B3 were then interspersed
among trials with the previously trained A1-O-A3 sequence to maintain both perfor-
mances. Next, the new stimulus B1 was substituted for the old A1 in displays that also
contained O and A3. This immediately established the new sequence B1-O-A3. During
the session, trials with the two previously trained sequences A1-O-A3 and A1-O-B3 were
interspersed to maintain these performances. Then, displays that contained B1 and B3
with O were presented for the first time, immediately yielding the correct sequence B1-O-
B3, which is analogous to the S-V-O sentence Bob ate corn. Finally, when trials for all
four sequences were presented in a mixed order, each produced the appropriate sequen-
tial selections of the three stimuli. These performances demonstrated the maintenance of
the previously trained sequences A1-O-A3, A1-O-B3, and B1-O-A3, as well as the new
Chapter 10 225

B1-O-B3, a combination of elements from the sequences established earlier. These four
sequences then provided an ongoing cumulative baseline of sequence performances to
which other sequences could be added. As shown at the right in figure 10.5, the same
kind of substitutions in the English sentences (stage 2, corn for cake, in the sequence Bill
ate cake and Bob for Bill in the sequence Bob ate cake) would be expected to yield the
emergent sequence, Bob ate corn, at stage 3.
Establishing Contextual Control of Sequencing

This segment of training (stages 4 through 7 in figure 10.5) was used to establish
contextual control of the order of touching B1 and B3, which was dependent on their
presentation with the old stimulus, designated by O, or the new stimulus, designated by
X (in other words, B1-O-B3 and B3-X-B1, respectively). Whereas O signaled the touching
of B1 first, O second, and B3 third, X signaled the touching of B3 first, X second, and B1
third, the English analogues being Bob ate corn and Corn was eaten by Bob (the third and
fourth combinations in figure 10.5, above).
The B3-X-B1 sequence was trained in the same way as the A1-O-A3 sequence. First,
B3 was displayed alone. After it was touched, X was added to the screen and so on.
Although no errors occurred during the training of the B3-X-B1 sequence (the analogue
of the English Corn was eaten by Bob), it produced errors when it was added to the base-
line of the previously learned sequences that contained O between the A and B stimuli.
Five additional training sessions were conducted with differential reinforcement in order
to establish reliable performance of the B3-X-B1 sequence when it was mixed with the
other sequences.
Once the mixed set of trained sequences was performed perfectly, it became possible
to assess the effect of presenting X with A1 and A3 for the first time. Would that result
in the touching of A3 first, X second, and A1 third, in accordance with the history estab-
lished with respect to X (in other words, controlling reversal of the sequence established
in the presence of O)? Or would A1 be touched first and A3 third, in accordance with the
history of direct reinforcement built up during A1-O-A3 sequence training?
The test mentioned above was designed to assess the emergence of a novel sequence
performance and thus could not be conducted using feedback because training of the
sequence might result. Therefore, the thirteenth session was used to prepare the subject
for the test. Only trials for previously trained sequences A1-O-A3, B1-O-B3, and B3-X-B1
were included. Differential consequences were removed to prepare for introduction of
unreinforced test trials. The purpose of this was to examine whether presentation of the
stimuli A1 and A3 with new contextual stimulus X would control reversal of the order in
which these two stimuli were touched previously. The thirteenth session was introduced
with the teacher’s statement that the computer would keep track of how many trials
were correct so that pennies earned would be delivered at the end of the session. The
session contained forty trials, all presented with no feedback. The sequences produced on
thirty-nine of these trials were consistent with training.
An evaluation of contextual control of sequence reversal was then conducted with ten
probe trials during the test session in which the stimuli A1 and A3 were presented for
the first time with X (stage 5). Although all trials were presented without feedback, the

predicted performances always occurred in the A3-X-A1 trials; A3 was touched first, X
second, and A1 last. These novel performances demonstrated that X and O functioned
as contextual cues for different sequences. The comparable reversal in English would be
production of Cake was eaten by Bill in stage 5, after training of Bill ate cake, Bob ate corn,
and Corn was eaten by Bob. In addition, performance on trials with the previously trained
and emergent sequences A1-O-A3, B1-O-B3, and B3-X-B1 was perfect even though no
feedback was provided on any of these baseline trials. All these sequence performances
were maintained in a subsequent session in which reinforcement was reinstated.
Introducing New Stimuli and Tests for Order Reversal

Training for a new sequence involving new stimuli (stage 7, C3-X-C1; English ana-
logue Egg was eaten by Mary) was conducted by substituting first C1 for B1 and then C3
for B3 into the sequence B3-X-B1. This replicated the earlier training given with the B
stimuli. Performance was virtually perfect throughout the training. The same steps for
yielding the analogous English construction that is illustrated in stage 7 involves the
substitution of Mary for Bob and egg for corn. These substitutions convert the trained
sequence Corn was eaten by Bob to the new sequence Egg was eaten by Mary.
Next (stage 8, session 19), probe trials that involved displays of the stimuli C1 and C3
but with O rather than X (English, Mary, egg, and ate) were given to examine emergence
of the sequence C1-O-C3 (Mary ate egg). The procedure replicated that used in session
14 (no consequence following each trial, and so on) but the baseline was expanded to
include trials with the A3-X-A1 sequence that had emerged on earlier test trials in stage
5. Performance was perfect on the probe trials, and only one error (out-of-sequence selec-
tion) occurred on a baseline trial. Perfect performance then occurred in session 20 when
reinforcement was reinstated.
Tests for Emergent Recombinations of Sequence Elements

In the final part of the study, probe trials were used to assess performance with novel
mixtures of the A and C stimuli with the O and X stimuli, combinations that had never
appeared together. Each of the displays used to test for emergence of the novel sequences,
A1-O-C3, A3-X-C1, C1-O-A3, and C3-X-A1, were presented on ten trials, all with no dif-
ferential feedback. These sequences were the analogues of the English transformations that
are listed in figure 10.5, stage 9. Of the forty test trials with novel sequences, thirty-nine
produced performances that were consistent with the preceding training and testing.
Clearly, the stimuli X and O controlled the order in which other elements included in
the sequence repertoire were produced. The efficiency with which new sequence members
were added and then used in appropriate positions in new sequences is particularly notable.
Only two sequences with three stimuli were trained directly. Simple substitution then pro-
duced four additional sequences. In each instance, one new stimulus was presented in the
context of two others that already occupied particular positions in sequences established
earlier. The remaining eight sequences tested, however, were contextually appropriate, emer-
gent recombinations of the stimuli used in other contexts (marked by * in figure 10.5).
Chapter 10 227

The preceding results suggest that the subject’s capacity to acquire language had been
underestimated and that exposure to the current training methods might yield substan-
tial gains in his ability to communicate with others. In that regard, we note some of the
general advantages stemming from the emphasis we have placed on teaching rather than
testing, and on the analysis of individual subjects rather than groups of subjects. Our
experience has indicated that an individual’s failure to develop a particular discrimina-
tive performance need not mean that the individual cannot acquire that performance
(Mackay & Sidman, 1968). Rather, failure to learn or poor performance may reflect the
use of an inadequate teaching procedure, or it may reflect the individual’s lack of appro-
priate behavior that is a precursor of, or even prerequisite to, that performance. Through
the attempt to build complex performances in a gradual step-by-step progression, the
influence of uncontrolled sources of variation or deficient learning history may be identi-
fied, and corrective measures may be taken. Tests may be given frequently, but they func-
tion as measures of current performance rather than capacity or ability. These measures
are treated as baselines or starting points for additional work. In general, the emphasis
is on generating desired behavior rather than testing current level of functioning. This
approach carries with it important methodological advantages. Sensory, perceptual, and
other deficits that might hamper the measurement of learning can be identified or ruled
out for each individual. Further, deficiencies in attentional or other skills that may be
prerequisite for generalizable, conceptual performance can be corrected with appropriately
augmented or revised training.
Syntactic Transformations: External

Contextual Control
Different languages have different syntaxes. It is commonplace for a bilingual indi-
vidual to utter sentences that have different syntactic structures based on the language
that is appropriate to a given community of listeners (in other words, a given context).
Theoretically, this capacity to produce an infinite number of syntactically correct sentences
in different linguistic contexts could be induced by the formation of equivalence classes,
sequence training, and the establishment of contextual control of those repertoires. In
English, adjectives that modify nouns typically precede the noun being modified, as in
the utterance “red ball,” represented symbolically as adjective-noun. In Spanish, however,
adjectives that modify nouns typically follow the noun being modified, as in the utter-
ance “pelota roja,” represented symbolically as noun-adjective. If the appropriate linguistic
community is indicated with some external cue, individuals should use the appropriate
syntax with an infinite number of combinations of noun and adjective that had not been
used for training.
An approximation of these conditions was described by Lazar and Kotlarchyk
(1986). The subjects in the three-stage experiment were four typically developing young
children. The analogues of the noun and adjective classes were created using two color
patches and eight Greek letters. One class consisted of the red stimulus (red) and the
Greek letters represented symbolically by A1, A2, A3, and A4. The other class consisted
of the green stimulus (green) and four other Greek letters referred to symbolically as B1,
B2, B3, and B4.

In stage 1, the children learned two equivalence classes, each consisting of one color
and four Greek letters. The stimuli in one class were the analogues of nouns; the stimuli
in the other class were the analogues of adjectives. Training was conducted with a sample-
as-node matching-to-sample procedure in which a red stimulus (red) and a green stimulus
(green) served as the sample stimuli for classes 1 and 2, respectively. Subjects were trained
to select the Greek letters represented by A1, A2, A3, and A4 in the presence of the red
sample, and to select the Greek letters represented by B1, B2, B3, and B4 in the presence
of the green sample. For example, when red was presented as the sample with A1 and
B1 as the comparisons, selection of A1 was reinforced. In contrast, when green was pre-
sented as the sample with the same comparisons, selection of B1 was reinforced. Thus, the
trained relations for one class were red-A1, red-A2, red-A3, and red-A4, while the trained
relations for the other class were green-B1, green-B2, green-B3, and green-B4.
The emergence of the two five-member equivalence classes, red-A1-A2-A3-A4 and
green-B1-B2-B3-B4, were evaluated with the presentation of twenty-four equivalence
probes and eight symmetry probes. Each probe consisted of a trial that contained a sample
from one class and two comparisons: one from the same class as the sample and one from
the other class. Each probe was represented in an X-Y-Z format where X was the sample,
while Y and Z were the correct and incorrect comparisons, respectively. Each probe trial
was presented many times and the locations of the correct and incorrect comparisons
were changed unsystematically from trial to trial. No feedback followed performance on
any of the probe trials.
The reversibility of functions for the stimuli in the trained relations in the red class
was evaluated by the performances produced on the symmetry probes A1-red-green,
A2-red-green, A3-red-green, and A4-red-green. Likewise, the reversibility of the stimuli
in the trained relations in the green class was evaluated by the performances produced
on the symmetry probes B1-green-red, B2-green-red, B3-green-red, B4-green-red. The
emergence of equivalence relations among the A stimuli in the red class were evaluated by
the performances on the red class equivalence probes: A1-A2-B2, A1-A3-B3, A1-A4-B4,
A2-A1-B1, A2-A3-B3, A2-A4-B4, A3-A1-B1, A3-A2-B2, A3-B4-A4, A4-A1-B1, A4-A2-
B2, and A4-A3-B3. The emergence of relations among the B stimuli in the green class was
evaluated by the performances produced by the following green class equivalence probes:
B1-B2-A2, B1-B3-A3, B1-B4-A4, B2-B1-A1, B2-B3-A3, B2-B4-A4, B3-B1-A1, B3-B2-
A2, B3-B4-A4, B4-B1-A1, B4-B2-A2, and B4-B3-A3. Note that all of these equivalence
probes involved presentations of novel combinations of Greek letters that had never been
seen together. All of the symmetry and equivalence probes produced the selection of the
correct comparisons, which indicated the emergence of the two equivalence classes red-
A1-A2-A3-A4 and green-B1-B2-B3-B4. In stage 1, then, the red class can be viewed as
the analogue of a set of nouns, while the green class can be viewed as the analogue of a
set of adjectives.
Stage 2 was the analogue of establishing different syntaxes in the presence of two dif-
ferent language communities, which were represented by low- and high-pitched tones. On
all trials, subjects were presented with two stimuli at the same time; one was a particular
stimulus in the noun class and the other was a particular stimulus from the adjective
class (for example, red or green). When the high tone was also present, picking red before
green was reinforced. This was the analogue of learning a noun-adjective syntax. When
the low tone was present, picking the green before red was reinforced, which was the
analogue of learning an adjective-noun syntax.
Chapter 10 229

Stage 3 assessed the generality of the sequencing performances across novel combi-
nations of the stimuli in the red and green classes that had not been used for sequence
training. These included the following sixteen combinations of A and B stimuli in the red
and green classes: A1-B1, A1-B2, A1-B3, A1-B4, A2-B1, A2-B2, A2-B3, A2-B4, A3-B1,
A3-B2, A3-B3, A3-B4, A4-B1, A4-B2, A4-B3, and A4-B4. In addition, all of these sixteen
novel combinations of “nouns” and “adjectives” were presented in the presence of the high
and low tones. Thus, thirty-two tests were presented to evaluate the generalization of
syntax to novel exemplars.
When the sixteen test trials involved the presence of the high tone, subjects always
selected the A stimuli before the B stimuli. In contrast, when the sixteen test trials involved
the presence of the low tone, subjects always selected the B stimuli before the A stimuli.
These performances emerged as soon as these tests were introduced and continued with
test repetition even though the trials were presented in the absence of feedback. These
performances, then, demonstrated the explosive generalization of syntactically appropri-
ate utterances across all possible novel stimulus pairs in a restricted laboratory-contrived
system.
As was the case in the previous section, it should be noted that direct training estab-
lished the different sequences of selecting red and green in the presence of the high and
low tones. Thus, the tones were serving as discriminative stimuli. It would be inappropri-
ate to conclude from the trained performances alone that the high and low stimuli were
exerting contextual control of the response sequences. On the other hand, their contex-
tual functions were demonstrated by the emergence of the sequence performances that
were produced on trials involving the novel combinations of stimuli drawn from the red
and green equivalence classes.
In summary, the emergent performances reported by Lazar and Kotlarchyk (1986)
provide an analogue of the extensive generality with which syntactically appropriate
utterances are produced by multilingual speakers in natural settings. The performances
were induced by three behavioral procedures: the prior formation of equivalence classes;
sequence training with one pair of stimuli, each chosen from one of the equivalence
classes; and training different sequences in the presence of different contextual cues. This
“proof of concept” demonstration suggests the feasibility of using well-known behavioral
procedures to induce the complex contextually controlled syntactic transformations used
by linguistically competent individuals in different settings. One example would be the
transformation of the interrogative to the declarative construction (for example, Is the
truck red? versus The truck is red). In this case, the position of the verb is shifts with the
function of the sentence. In the interrogative construction, the verb is occupies the first
position in the word string. In the declarative construction, is precedes the adjective and
is in the penultimate position in the word string. The use of each construction could be
determined by the presentation of a contextual cue, such as a teacher’s instruction to
construct a statement or question, or a sentence with either a rising intonation or a falling
intonation. Alternatively, visual contextual cues would be supplied by the presence of a
tile showing either a period or a question mark. Training of some sequences would dem-
onstrate discriminative control of declarative and interrogative constructions by the period
and question mark. The appropriate sequencing of new word strings in the presence of
the period and question mark would then demonstrate the contextual control exerted by
these markers, syntactic productivity, and competence in syntactic transformation.

Conclusion
This chapter provided a stimulus control analysis of elementary forms of syntactic con-
structions in which successive elements (words) occur in the linear orders that are common
during early language development. Typical examples include the adjective-noun and
SVO structures of English, but the procedures described may be applied with many other
simple language structures (such as two-word agent-action, and action-object structures)
that provide the basis for more complex structures (for example, agent-action-object such
as Mommy get ball). The critical property of syntax—productivity—requires the extensive
generality of performances as demonstrated throughout the chapter by illustrations of
novel constructions and the recombination of elements without direct training.
Procedures that establish these generalized forms of emergent behavior are firmly
founded on behavioral research on stimulus classes. In one approach, stimuli in a class
become substitutable for each other through equivalence class training. Subsequently, the
training of one sequence using one stimulus from each class in a given serial position
resulted in the emergence of new syntactically appropriate sequences that consisted of
words from appropriate classes. In the reverse approach, many sequences were trained,
and stimuli that occupied the same serial position in the sequences became functionally
substitutable or equivalent to each other. This substitutability permitted the emergence of
recombinations in new syntactically appropriate sequential utterances or constructions.
A third set of procedures was used to establish contextual control over different orders
of the words in utterances. This was important because it induced syntactic-grammatical
transformations (for example, transformation of a construction from active to passive).
The procedures described here can help to make instruction effective and practical,
whether conducted on a tabletop or a computer. The training has frequently resulted in
virtually errorless performances and leads to the coordination of the complex repertoires
of behavior involved. Most of these studies, however, have involved few, or even single,
participants. Thus, the generality of the results is not currently known. Also, this discus-
sion did not deal with the more complex recursive procedures that may require the remote
dependencies involved in, for example, noun-verb number agreement or the semantic
factors involved in the identification of some sequences as syntactically correct and mean-
ingful and others as syntactically correct but meaningless. As Mackay (1991) concluded,
however, the approach used here is valuable, not only because it offers a practical approach
to complex instruction, but also “because it emphasizes variables that directly influence
behavior acquisition. Such an emphasis helps to ensure that the results of experimen-
tal analyses of behavior often have direct implications for applied behavior analysis” (p.
255).
References
Barnes, T., Mackay, H. A., & Stromer, R. (2008). Sequence production and matching to
sample: Transfer of stimulus control. Manuscript submitted for publication.
Belanich, J., & Fields, L. (2003). Generalized equivalence classes as response transfer
networks. Psychological Record, 53, 373–413.
Chapter 10 231

Branch, M. (1994). Stimulus generalization, stimulus equivalence, and response hierar-
chies. In S. C. Hayes, L. J. Hayes, M. Sato, & K. Ono (Eds.), Behavior analysis of
language and cognition. Reno, NV: Context Press.
Brown, R. (1973). A first language. Cambridge, MA: Harvard University Press.
Bruner, J. S. (1975). Toward a theory of instruction. Cambridge, MA: Harvard University
Press.
Catania, A. C. (1998). Learning. Englewood Cliffs, NJ: Prentice Hall.
Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press.
Cullinan, V. A., Barnes, D., Hampson, P. J., & Lyddy, F. (1994). A transfer of explic-
itly and non-explicitly trained sequence responses through equivalence relations:
An experimental demonstration and connectionist model. Psychological Record, 44,
559–586.
Dale, R. (2004). Cognitive and behavioral approaches to language acquisition: Conceptual
and empirical intersections. Behavior Analyst Today, 5, 336–358.
Dube, W. V., McDonald, S. J., McIlvane, W. J., & Mackay, H. A. (1991). Constructed-
response matching to sample and spelling instruction. Journal of Applied Behavior
Analysis, 24, 305–317.
Fields, L., Fitzer, A., Tittelbach, D., & Shamoun, K. (2005). The effect of testing variables
on the formation of linked perceptual classes. Journal of the Experimental Analysis of
Behavior, 84, 243–269.
Fields, L., Matneja, P., Varelas, A., Belanich, J., Fitzer, A., & Shamoun, K. (2002). The
formation of linked perceptual classes. Journal of the Experimental Analysis of Behavior,
78, 271–290.
Fields, L., & Reeve, K. F. (2000). Equivalence classes and natural categories from the gen-
eralization of perceptual and relational classes. In J. C. Leslie & D. Blackman (Eds.),
Issues in experimental and applied analyses of human behavior. Reno, NV: Context
Press.
Fields, L., & Reeve, K. F. (2001). A methodological integration of generalized equiva-
lence classes, natural categories, and cross-modal perception. Psychological Record, 51,
67–88.
Fields, L., Reeve, K. F., Adams, B. J., Brown, J. L., & Verhave, T. (1997). Predicting the
extension of equivalence classes from primary generalization gradients: The merger
of equivalence classes and perceptual classes. Journal of the Experimental Analysis of
Fields, L., Reeve, K. F., Adams, B. J., & Verhave, T. (1991). Stimulus generalization
and equivalence classes: A model for natural categories. Journal of the Experimental
Fields, L., Reeve, K. F., Matneja, P., Varelas, A., Belanich, J., Fitzer, A., et al. (2002).
The formation of generalized categorization repertoires: Effect of multiple domain,

sample, and comparison training. Journal of the Experimental Analysis of Behavior, 78,
291–313.
Fields, L., Tittelbach, D., Shamoun, K., Fitzer, A., Watanabe, M., & Matneja, P. (2007).
The effect of training variables on the formation of linked perceptual classes. Journal
of the Experimental Analysis of Behavior, 87, 97–119.
Fields, L., & Verhave, T. (1987). The structure of equivalence classes. Journal of the
Fields, L., Verhave, T., & Fath, S. (1984). Stimulus equivalence and transitive associa-
tions: A methodological analysis. Journal of the Experimental Analysis of Behavior, 42,
143–157.
Galizio, M., Stewart, K. L., & Pilgrim, C. (2004). Typicality effects in contingency-shaped
generalized equivalence classes. Journal of the Experimental Analysis of Behavior, 82,
253–273.
Galy, E., Camps, J., & Melan, C. (2003). Sequence class formation following learning of
short sequences. Psychological Record, 53, 635–645.
Gast, D. L., VanBiervleit, A., & Spradlin, J. E. (1979). Teaching number-word equiva-
lences: A study of transfer. American Journal of Mental Deficiency, 83, 524–527.
Goldberg, A. E. (1995). Constructions: A construction grammar approach to argument struc-
ture. Chicago: Chicago University Press.
Green, G. (1993). Stimulus control technology for teaching number-quantity equivalences.
Proceedings of the 1992 conference of the National Association for Autism (Australia).
Melbourne, Australia: Victoria Autistic Children’s and Adults’ Association.
Green, G., Stromer, R., & Mackay, H. A. (1993). Relational learning in stimulus sequences.
Psychological Record, 43, 599–616.
Greer, R. D., & Ross, D. E. (2004). Verbal behavior analysis: A program of research in
the induction and expansion of complex verbal behavior. Journal of Early and Intensive
Behavior Intervention, 1, 141–165.
Guess, D., Sailor, W., & Baer, D. M. (1974). To teach language to retarded children. In
R. L. Schiefelbusch & L. L. Lloyd (Eds.), Language perspectives: Acquisition, retarda-
tion, and intervention. Baltimore, MD: University Park Press.
Skinnerian account of human language and cognition (pp. 57, 70, 169–170). New York:
Kluwer Academic/Plenum.
Holcomb, W. L., Barnes, T., & Mackay, H. A. (1991). Contextual control of the production
of three-term sequences. Unpublished manuscript.
Holcomb, W. L., Stromer, R., & Mackay, H. A. (1997). Transitivity and emergent
sequence performances in young children. Journal of Experimental Child Psychology,
65, 96–124.
Chapter 10 233

Lane, S. D., Clow, J. K., Innis, A., & Critchfield, T. S. (1998). Generalization of cross-
modal stimulus equivalence classes: Operant processes as components in human cat-
egory formation. Journal of the Experimental Analysis of Behavior, 70, 267–279.
Lazar, R. (1977). Extending sequence-class membership with matching to sample. Journal
of the Experimental Analysis of Behavior, 27, 381–392.
Lazar, R., & Kotlarchyk, B. J. (1986). Second-order control of sequence-class equivalences
in children. Behavioral Processes, 13, 205–215.
Lazar, R., & Mackay, H. A. (1982). Teaching spelling via stimulus equivalences. Exceptional
Child Education Resources, 14, EC14210.
Leader, G., & Barnes-Holmes, D. (2001). Establishing fraction-decimal equivalence using
a respondent-type training procedure. Psychological Record, 51, 151–166.
Lynch D. C., & Cuvo, A. J. (1995). Stimulus equivalence instruction of fraction-decimal
Mackay, H. A. (1985). Stimulus equivalence in rudimentary reading and spelling. Analysis
and Intervention in Developmental Disabilities, 5, 373–387.
Mackay, H. A. (1991). Stimulus equivalence: Implications for the development of adaptive
behavior. In B. Remington (Ed.), The challenge of severe mental handicap: A behavior
analytic approach. Chichester, England: John Wiley & Sons.
Mackay, H. A., Kotlarchyk, B. J., & Stromer, R. (1997). Stimulus classes, stimu-
lus sequences, and generative behavior. In E. M. Pinkston & D. M. Baer (Eds.),
Environment and behavior. Boulder, CO: Westview Press.
Mackay, H. A., & Sidman, M. (1968). Instructing the mentally retarded in an institu-
tional environment. In G. A. Jervis (Ed.), Expanding concepts in mental retardation: A
symposium. Springfield, IL: Charles C. Thomas.
Mackay, H. A., & Sidman, M. (1984). Teaching new behavior via equivalence relations.
In B. Sperber, C. MacCauley, & P. H. Brooks (Eds.), Learning and cognition in the
mentally retarded. Hillsdale, NJ: L. Erlbaum Associates.
Mackay, H. A., Stoddard, L. T., & Spencer, T. J. (1989). Symbols and meaning classes:
Multiple sequence production and the emergence of ordinal stimulus classes.
Experimental Analysis of Human Behavior Bulletin, 7, 16–17.
Mackay, H. A., Stromer, R., & Serna, R. W. (1997). Emergent behavior and intellec-
tual functioning: Stimulus classes, generalization, and transfer. In S. Soraci and W.
J. McIlvane (Eds.), Perspectives on fundamental processes in intellectual functioning.
Greenwich, CT: Ablex Publishing.
Mandell, C., & McCabe, A. (1997). The problem of meaning: Behavioral and cognitive
perspectives. Amsterdam: Elsevier Science.
Maydak, M., Stromer, R., Mackay, H. A., & Stoddard, L. T. (1995). Stimulus classes in
matching to sample and sequence production: The emergence of numeric relations.
Research in Developmental Disabilities, 16, 179–204.

Meehan, E. F., & Fields, L. (1995). Contextual control of new equivalence classes.
Ninio, A. (1998). Acquiring a dependency grammar: The first three stages in the acquisi-
tion of multiword combinations in Hebrew-speaking children. In G. Makiello-Jarza,
J. Kaiser, & M. Smolczynska (Eds.), Language acquisition and developmental psychol-
ogy. Krakow, Poland: Universitas.
Reeve, K. F., & Fields, L. (2001). Effect of number of forced-choice primary generaliza-
tion test trials on the establishment of perceptual classes along a single dimension.
Rehfeldt, R. A., & Root, S. (2004). The generalization and long-term retention of equiva-
lence relations in adults with mental retardation. Psychological Record, 54, 173–186.
Sidman, M. (1994). Equivalence relations and behavior: A research story. Boston, MA:
Authors Cooperative.
Sigurdardottir, Z. G., Green, G., & Saunders, R. R. (1990). Equivalence classes generated
by sequence training. Journal of the Experimental Analysis of Behavior, 53, 47–63.
Stoddard, L. T., Brown, J., Hurlbert, B., Manoli, C., & McIlvane, W. J. (1989). Teaching
money skills through stimulus class formation, exclusion, and component matching
methods: Three case studies. Research in Developmental Disabilities, 10, 413–439.
Stokes, T. F., & Baer, D. M. (1977). An implicit technology of generalization. Journal of
Stromer, R., & Mackay, H. A. (1993). Human sequential behavior: Relations among
stimuli, class formation, and derived stimulus sequences. Psychological Record, 43,
107–131.
Tomasello, M. (2003). Constructing a language: A usage-based theory of language acquisi-
tion. Cambridge, MA: Harvard University Press.
Chapter 10 235

Chapter 11
Extending Functional Communication

Through Relational Framing
Rocio Rosales and Ruth Anne Rehfeldt,

Southern Illinois University
This chapter describes how to use relational framing to develop and expand communica-
tion repertoires in individuals with severe language deficits. The advantages of program-
ming for the emergence of derived stimulus relations in a mand instructional program
will be highlighted first, followed by step-by-step instructions and problem-solving tips
for the establishment of several targeted skills.
Teaching Using a Technology of Derived

Stimulus Relations
Chapter 5 of this volume discussed several communication systems that may be appro-
priate for individuals with severe disabilities and their caregivers. Regardless of whether
a selection- or topography-based system is employed, a curriculum that focuses upon
the establishment of stimulus relations may permit the emergence of a number of new
manding skills. Specifically, programming for the emergence of a derived mand repertoire
may be advantageous for a number of reasons. Derived manding skills are manding skills
that have not been explicitly reinforced over the course of instruction (Barnes-Holmes,
Barnes-Holmes, & Cullinan, 2000; Murphy & Barnes-Holmes, in press). Rather, the
requested item may participate in a derived stimulus relation with other items of the same
superordinate category. For example, an individual may be explicitly taught to request
a favorite doll by vocally requesting “toy.” The individual may then be taught that a
doll is conditionally related to other items of the same superordinate class, such as a toy
train and a toy car, so that when the individual desires the train or car, he or she may
similarly request “toy” in the absence of direct instruction for doing so (Barnes-Holmes
et al., 2000; Halvey & Rehfeldt, 2005). Such topography-based mands were established
in adults with severe developmental disabilities and rudimentary vocal verbal skills in a
study by Halvey and Rehfeldt (2005). The participants were taught to request three pre-
ferred items from three distinct superordinate categories using their category names. For
example, one individual was taught to vocally request a banana with the vocal request
“fruit” and a Walkman with the vocal request “music.” Participants were then taught
conditional discriminations between pictures of those items and three other pictorially
represented items of the same category. For example, a picture of a banana was condition-
ally related to pictures of two other types of fruit, and a picture of a Walkman was condi-
tionally related to pictures of other music-playing devices. We found that the participants
ultimately used the appropriate category names to mand for the untrained items.
A derived mand may also occur when a stimulus that is used to request a preferred
item enters into a derived stimulus relation with another stimulus. An individual who is
explicitly taught to exchange pictures of preferred items with caregivers to request those
items may then be taught conditional discriminations between the pictures and their
corresponding text, with the result that, when the individual wants access to a particu-
lar item, he or she may exchange text for the item in the absence of direct instruction.
For example, if a learner is taught to relate a picture of candy to its dictated name and
the dictated name with its corresponding text, and the learner is then taught to use the
picture to request the candy, a derived mand may occur in the form of the exchange of
text for the candy, a skill that was never directly taught or reinforced (Rehfeldt & Root,
2005). This was the outcome of such an instructional protocol utilized by Rehfeldt and
Root (2005), who taught adults with severe developmental disabilities to use pictures
to request preferred items using the picture exchange communication system (PECS).
Conditional discriminations were then established between the pictures of preferred items
and their dictated names, and between the dictated names and their corresponding text.
The participants then exchanged text in lieu of pictures to request preferred items. Such
a skill may have important ramifications, since manding with text may be viewed as a
more “adultlike” skill for adults with severe disabilities. In fact, using text rather than
pictures to request items may be viewed as more socially appropriate in some integrated
community settings.
This instructional protocol can also be used to program for the emergence of derived
mands within the context of a chained task (Rosales & Rehfeldt, 2007). This was accom-
plished by employing the interrupted chained task procedure, which is based on con-
triving a transitive conditioned establishing operation (Michael, 1993) by setting up the
training environment so that one stimulus increases the reinforcing value of a second
stimulus (see also chapter 5 of this volume). Learners were taught to complete several
chained tasks to criterion and were then provided with all but one of the items needed to
complete the steps in the task. For example, in the context of making a pitcher of juice
from a powdered mix, the spoon needed to stir the powder was withheld from the array
of items presented. The learner was then taught to request the missing item by exchang-
ing a picture of a spoon for the item itself in order to continue with the next step of the
task. Following a period of instruction in relating dictated names of the items needed for

several tasks to the pictures and the dictated names to the corresponding text, partici-
pants used text to request the items needed to complete the tasks. The participants also
named pictures, read text, and matched pictures to their corresponding text and vice
versa in the absence of training. Moreover, these learners vocally requested the needed
items while completing the chained tasks at one month follow-up. Thus, a number of new
verbal skills may emerge in the absence of direct instruction when a history of reinforced
relational learning is established.
Selection-Based Derived Mands for Preferred Items

This section describes how to program for the emergence of derived mands with individu-
als for whom PECS or a modified picture exchange system is deemed an appropriate com-
munication system. We describe how individuals may come to exchange text for preferred
items following an instructional history of reinforced conditional discriminations.
Instructional stimuli will consist of a three-ring binder in which digitized pictures
of preferred items are secured by Velcro, several objects or activities previously identified
as preferred, and 3-by-5-inch index cards on which the names of the preferred items are
written or printed. The font size and formatting should be the same for all of the textual
stimuli. We recommend that the stimulus cards be labeled for view by the instructor only
(for example, on the back side of the card) to avoid any possible confusion once training
begins. We also recommend that a stimulus placement board be constructed and used
if the conditional discrimination teaching is nonautomated. This may be created out of
poster board or construction paper. An outline of where each comparison stimulus will
be placed during training should be drawn; they should be equidistant apart and horizon-
tally aligned (for example, typically three square or rectangular shapes). This will ensure
that none of the comparison stimuli is placed in such a way that its positioning inadver-
tently cues a correct (or learner-defined) response.
Derived Mand Pre-test and Post-test Probes

Pre-test probes of all the targeted skills must first be conducted in order to determine
if they are in the learner’s repertoire. The derived mands for preferred items will test
whether the learner exchanges the correct written word for a preferred item when this
stimulus is presented among several textual stimuli.
Instructions. Present one preferred item in the learner’s view at a time, with approx-
imately five textual stimuli, including three printed words corresponding to preferred
items and two or more distracter words on the front of the learner’s binder.
Order of Stimuli. We typically present instructional stimuli in nine-trial blocks during

pre-test probes. Only one trial block is conducted during probe trials, with each mand
tested three times per nine-trial block. The positioning of the comparison textual stimuli
on the learner’s binder should change from trial to trial.
Chapter 11 239

Order of Trials. The order of test trials should be randomly determined. Specifically,
each preferred item should be presented three times in a different order per trial block
during derived mand probes for preferred items.
Reinforcement. All baseline pre-test trials are conducted under extinction. Therefore,
it is important to withhold all reinforcement for all trials regardless of the accuracy of
the response. Once a trial is complete, the next trial in the block should be presented.
Intermittent nonspecific reinforcement (such as “You’re working hard!”) may be provided
between trial blocks or during breaks in order to prevent frustration and boredom, but it
is critical that such feedback is not presented contingent upon correct responses. Because
probes are conducted under extinction, we also recommend interspersing test probes
between preferred activities.
Prompts. Instructional prompts are not provided during probe trials.
Mastery Criterion. We often assume that a skill is intact if an individual performs cor-
rectly on eight out of nine trials per block (89 percent correct).
Handling and Preventing Errors. No strategies are used to prevent errors during pre-
test probes.
Stimulus Relations Probes

Below are instructions for conducting the stimulus relations probes:
1. Match dictated name to corresponding picture (A-B). Present the dictated
name of the picture of a preferred item (for example, “Show me book” or
“book”) and three comparison stimuli evenly placed on the stimulus placement
board directly in front of the learner. (This conditional discrimination proce-
dure is an auditory-visual matching task.)
2. Match dictated name to corresponding text (A-C). Present the dictated name
of the text corresponding to the preferred item (for example, “Show me book”
or “book”) and three comparison stimuli evenly spaced on the stimulus place-
ment board directly in front of the learner.
3. Name pictures (B-A). Present one picture at a time, with the instruction “What
is it?”
4. Read text (C-A). Present one word at a time, with the instruction “What does it
say?” This will test for reading comprehension skills (Sidman & Tailby, 1982).
5. Match pictures to text (B-C). Present one picture at a time, with the instruc-
tion “Put with same” or “Match,” and place three comparison stimuli evenly
spaced on the stimulus placement board directly in front of the learner.
6. Match text to pictures (C-B). Present one printed word at a time, with the
instruction “Put with same” or “Match,” and place three comparison stimuli
evenly spaced on the stimulus placement board directly in front of the learner.

Stimulus Relations Probes
Below are instructions for conducting the stimulus relations probes.
Order of Stimuli. As is the case with derived mand pre-test probes, instructional stimuli
are typically presented in nine-trial blocks, with only one trial block conducted per rela-
tion. Each specific relation is tested three times per nine-trial block, and stimuli should
be presented in a different order during each trial block (for example, the positioning of
comparison stimuli should change on the stimulus placement board from trial to trial
during matching tasks).
Order of Trials. The order of test trials should be randomly determined. It may be
helpful to determine the order of trials before the beginning of a session by completing a
data sheet such as the example provided in figure 11.1.
Reinforcement. Test trials for all stimulus relations are conducted under extinc-
tion in order to ensure that these relations are not reinforced during pre-test probes.
Although intermittent nonspecific reinforcement may be provided between trial blocks or
during breaks, it is critical that such feedback is not presented contingent upon correct
responses.
Prompts. Instructional prompts are not provided during probe trials.
Mastery Criterion. We often assume that a skill is intact if an individual performs cor-
rectly on eight out of nine trials per block (89 percent correct).
Handling and Preventing Errors. No strategies are used to prevent errors during pre-
test probes.
Teach the Initial Request

Before a history of reinforced relational responding can be established, the learner
must be directly taught to mand for several preferred items by exchanging the pictures
corresponding to those items. To establish derived mands using PECS, we suggest that
the learner first be trained in the initial three phases of this selection-based system. The
ultimate goal of training is for the individual to discriminate among a variety of pic-
tures presented on a communication book, walk to the communication partner from
a prespecified distance, and exchange the correct picture for a preferred item without
assistance. For detailed information on the phases of PECS, the reader is referred to Frost
and Bondy (1994). These mands should be mastered before conditional discrimination
training begins.
Chapter 11 241

Sample Data Sheet
Probe Tests for Derived Relations
Data Collector: ________________
Date: ________________________
Client Name: _____________________
Name Pictures Read Text Derived Requests

Trial Relation Response Trial Relation Response Trial Relation Response
1 book 1 puzzle 1 Text-puzzle
2 puzzle 2 book 2 Text-book
3 cards 3 cards 3 Text-cards
4 puzzle 4 book 4 Text-book
5 cards 5 cards 5 Text-puzzle
6 book 6 puzzle 6 Text-cards
7 cards 7 cards 7 Text-puzzle
8 book 8 puzzle 8 Text-book
9 puzzle 9 book 9 Text-cards
Percent Correct: / Percent Correct : / Percent Correct: /
Notes: _______________ Notes: _______________ Notes: _______________
_______________________ _______________________ _______________________
Match Words to Pictures

Trial Relation Response Match Pictures to Words
1 Text cards-Pic cards Trial Relation Response
2 Text book-Pic book 1 Pic puzzle-Text puzzle
3 Text puzzle-Pic puzzle 2 Pic cards-Text cards
4 Text book-Pic book 3 Pic book-Text book
5 Text cards-Pic cards 4 Pic puzzle-Text puzzle
6 Text puzzle-Pic puzzle 5 Pic book-Text book
7 Text book-Pic book 6 Pic cards-Text cards
8 Text cards-Pic cards 7 Pic book-Text book
9 Text puzzle-Pic puzzle 8 Pic cards-Text cards
Percent Correct: / 9 Pic puzzle-Text puzzle
Notes: _______________ Percent Correct: /
_______________________ Notes: _______________
Figure 11.1. Example of a data sheet that can be used for pre- and post-test probe trials.
Conditional Discrimination Teaching

During conditional discrimination teaching the learner will be taught to conditionally
relate dictated names of items to their corresponding pictures (which will be referred to
as name-picture matching) and dictated names of items to their corresponding text (which
will be referred to as name-text matching). Prior to each trial of conditional discrimina-
tion teaching, the learner should be given the same instructions that were described for
the respective relations assessed during pre- and post-test probes. The directions for the
ordering of stimuli and trials presented in that section are also relevant to conditional
discrimination teaching. A mastery criterion of eight out of nine (89 percent) correct
responses for three consecutive nine-trial blocks should be established before the clinician
moves on to training the next skill. (We suggest that the reader carefully follow the pro-
cedures outlined in chapter 8 of this volume for establishing name-picture and name-text
conditional discriminations.)

Name-Picture Matching
Below are the components of name-picture matching.
Reinforcement. A correct response will consist of delivering the correct picture to the
instructor as a way of indicating a choice. Independent correct responses should be differ-
entially reinforced with descriptive verbal praise (such as “You’re right! That is the book!”).
If a response is not provided within five seconds or an incorrect response is provided,
the trial is scored as incorrect on the data sheet. Incorrect responses should be repeated
using corrective verbal feedback (such as “Try again”) and the correct response should be
prompted.
Prompts. If errors continue to occur with the delivery of feedback alone and the learner
becomes frustrated, a variety of prompting strategies may be employed. Prompting strate-
gies will vary depending on the individual learner, and they may be combined to obtain
the most effective results (see chapter 6 in this volume). Below are a few prompting strate-
gies that can be used.
Positional prompts. The correct comparison stimulus is placed closer
to the learner than the other two comparison stimuli, and the distance
between the correct comparison stimulus and the other two stimuli is
gradually reduced over the course of subsequent trials.
Gestural prompts. The instructor points to the correct response before
the learner has the opportunity to make an error.
Graduated time delay. The instructor presents a gestural prompt such
as pointing to the correct picture immediately following the presenta-
tion of the dictated name. Once the learner responds correctly on three
consecutive occasions, the delay before the prompt is increased to one
second. This delay is gradually and systematically increased as the learner
begins to respond correctly during the delay interval before each prompt
(Touchette & Howard, 1984).
Any prompts employed during training should be systematically faded (see Sulzer-
Azaroff & Mayer, 1991; Green, 2001).
Handling and preventing errors. At any point during the teaching steps described
above, learners may have trouble attaining mastery criterion for moving to the next step
of training. Alternatively, it may prove difficult for some learners to demonstrate some
or all of the derived stimulus relations even after training has been completed. In such
situations, it might be worthwhile to present each relation to be trained in isolation. For
example, if the stimuli are cards, a puzzle, and a book, the learner would first be taught
to respond to the instruction “Point to cards” by pointing to the picture displayed on the
stimulus placement board (with no other pictures present). Alternatively, an errorless pro-
cedure could be implemented by presenting blank cards in place of the two other sample
stimuli or pictures that are of a different color intensity than those used for the correct
sample stimulus.
Chapter 11 243

Name-Text Matching
The same procedures used to establish name-picture matching can be used to estab-
lish name-text matching. Learning to discriminate between text can be a much more
frustrating and irrelevant task for individuals with significant disabilities who have had
little experience with preacademic tasks. Instructors should be prepared for the fact that
this training is likely to be much more time-consuming and require more specialized
prompting procedures.
Mixed Name-Picture and Name-Text Matching

Once name-picture and name-text matching skills are established, it is a good idea to
train the two sets of conditional discriminations together. We often conduct mixed train-
ing in eighteen-trial blocks, with each relation trained three times per block, as shown in
figure 11.2.
Conduct Post-test Probes

Once all criteria have been met for conditional discrimination training, post-test probes
for the emergence of derived relations and derived mands are conducted in the exact manner
in which pre-test baseline probes were conducted. Probes for the A-B and A-C relations
need not be presented, because these are relations that were directly trained. If the learner
does not meet a criterion of eight out of nine (89 percent) correct per block for any of the
derived relations or derived mands, the instructor may conduct “booster sessions,” which
will provide the learner with repeated opportunities to demonstrate the conditional discrim-
inations that are intact. It may be the case that the learner simply requires more experience
with the conditional discriminations in order for the skills to emerge.
Selection-Based Derived Mands for Items Needed to

Complete Daily Living Skills
The purpose of this section is to describe procedures that are intended to facilitate the
emergence of derived mands within the context of chained tasks for daily living and
leisure activities. The interrupted chain procedure has been shown to be effective for
establishing new mands (see chapter 5 of this volume). Here, our ultimate goal is for the
learner to exchange text for items needed to complete preferred activities when these items
are removed from the array of stimuli presented at the beginning of a learning trial.
Instructional materials will be largely the same as those described in the previous
sections.

Sample Data Sheet
Mixed Training
Data Collector: _________________
Date: ________________________
Client Name: __________________
Trial Stimulus Response Trial Stimulus Response

1 Pic - puzzle 1 Pic - cards
2 Text - cards 2 Text - puzzle
3 Pic - book 3 Pic - puzzle
4 Text - puzzle 4 Text - cards
5 Pic - cards 5 Pic - book
6 Pic - book 6 Pic - puzzle
7 Text - book 7 Text - book
10 Text - book 10 Text - cards
11 Text - puzzle 11 Text - puzzle
12 Text - cards 12 Text - book
13 Pic - book 13 Pic -puzzle
15 Text - book 15 Text - puzzle
17 Text - cards 17 Text - book
18 Text -puzzle 18 Text -cards
Percent Correct: / Percent Correct: /
Notes: _______________ Notes: _______________
_______________________ _______________________
Figure 11.2. Example of data sheet that can be used in mixed name-picture and name-text
conditional discrimination teaching.
Teach the Chained Tasks

If the selected task is one that is not already within the learner’s repertoire, it will need
to be taught before training begins. It is recommended that a task analysis be created and
total task presentation employed (Sulzer-Azaroff & Mayer, 1991) to teach the learner the
chained tasks. For example, if the goal is to teach the learner the chained task of prepar-
ing a bowl of cereal, the following task analysis may be created based upon a number of
observations of an individual who is proficient in performing the task:
1. Open lid of cereal box.
2. Open bag inside cereal box completely.
3. Pour cereal into bowl until it is half full.
Chapter 11 245

4. Unscrew lid from gallon-size milk container.
5. Pour milk to bowl without spilling over.
6. Get spoon and eat cereal.
A multiple opportunity assessment (Cooper, Heron, & Heward, 2006) may be used
to initially assess skills already in the learner’s repertoire. In this assessment, the instruc-
tor completes any step that the learner fails to perform or performs incorrectly, so that
the learner is provided with the opportunity to complete every step of the task analysis
on each trial. For example, if the chained task is to prepare a bowl of cereal, the learner
would be provided with the opportunity to complete all six steps of the task analysis.
If the learner does not complete step 3 (pouring cereal into bowl until half full), the
instructor would complete this step and then give the learner the opportunity to continue
with the rest of the steps in the task analysis. Likewise, if the learner makes an error
that prevents him or her from proceeding with the task, such as pouring the cereal on
the floor, the instructor corrects the error and completes step 3, thus giving the learner
the opportunity to demonstrate any skills that are already within his or her repertoire.
Options for teaching a chained task include forward chaining, backward chaining, and
total task presentation. With forward and backward chaining, only one of the steps in
the chained task is taught to criterion before moving to the next step. In forward chain-
ing, the first step is taught first, whereas in backward chaining, the last step in the task
analysis is taught first. Finally, every step of the task analysis is trained on each teaching
trial when one is implanting a total task presentation technique. (For additional informa-
tion on teaching skill acquisition, the reader is referred to Sulzer-Azaroff & Mayer, 1991.)
Chained tasks may be considered mastered when participants complete 100 percent of
the steps of a task analysis independently on three consecutive trials. Once this criterion
is met, the next step of training may begin.
Teach Mands Within the Context of the Chained Task

Teaching an individual to mand for a missing item within the context of a chained
task consists of interrupting the task by withholding a needed item (keeping it out of
the learner’s sight) and teaching the learner to request the missing item by exchanging
a picture of it for the actual item (see chapter 5 of this volume for further detail on this
procedure). At this point in training, the chained task must be well within the learner’s
repertoire because one or more items will be removed from the array of materials needed
for the task. If the individual does not have all of the steps within his or her repertoire, he
or she might inadvertently skip the step without manding for the missing item.
Instructions
Present a binder with five pictures (two of which are distracter pictures that will not
be used in training). For example, if the selected chained tasks are to prepare a bowl of
cereal, to watch a video on a television set, and to make a cup of hot chocolate, the pic-
tures on the binder may be of a carton of milk, a video, and a cup of hot water to represent

one missing item from each task, respectively. The learner is then instructed to begin the
task with a simple verbal command such as “Make cereal.” Materials needed to perform
all but one or two of the steps in the task should be easily accessible and available.
Order of Stimuli
We suggest presenting mand opportunities in nine-trial blocks. If there are three
items to be trained, the learner should have three opportunities to request each needed
item per trial block. Thus, each of the chained tasks should be completed three times. The
pictures of the items should be attached with Velcro to a binder or board and rearranged
before each trial so that they are not presented in the same manner (for example, putting
the picture of the bowl on top) on consecutive trials.
Order of Trials
If more than one chained task is involved, it’s a good idea to present the tasks in a dif-
ferent order from one instructional session to the next so that the learner does not get in the
habit of exchanging a particular picture first in the session, another picture second, and so
forth. For example, make hot chocolate, watch video, and prepare cereal might be followed
by a different sequence, such as prepare cereal, make hot chocolate, and watch video.
Reinforcement
A correct response consists of the learner exchanging the correct picture for the
actual item that is needed to complete the step in the chained task. Independent correct
responses should be differentially reinforced with descriptive verbal praise (such as “You’re
right! That is the spoon!”) and delivery of the item. If the learner does not respond
within five seconds or responds incorrectly (by exchanging the wrong picture), the trial
is marked incorrect. Incorrect responses should be followed by receipt of the actual item
requested (for example, give the wrong item for five to ten seconds before removing), and
then repeating the trial using corrective verbal feedback (such as “Try again”) and an
appropriate prompting technique as described below.
Prompts
If a learner does not respond on a given trial, an indirect prompt such as “Look
here” (accompanied by pointing in the general direction of the binder) may be pro-
vided. Alternatively, the prompting suggestions provided in the previous sections may be
employed.
Handling and Preventing Errors

If incorrect responses persist, it may be beneficial to conduct repeated trials of phase
3 of PECS (Frost & Bondy, 1994). For example, provide the learner with the opportunity
to request highly preferred items by exchanging pictures selected from an array of at least
five stimuli. If the learner is having difficulty in mastering this step of training, it may
be beneficial to provide maintenance trials of this skill. In addition, the instructor may
Chapter 11 247

require an orienting or observing response (for example, touching the binder) before pre-
sentation of each trial during mand training within the chained tasks; the purpose of this
is to ensure that the learner makes perceptual contact with the stimuli. These responses
should be gradually eliminated. If the learner continues to skip the step in which he
or she is required to mand for the missing item, the instructor should ensure that the
chained task is well mastered.
Pre-test and Post-test Probes

Pre-test probes of all the targeted skills must first be conducted to determine if they
are in the learner’s repertoire; they are again conducted following mastery of the base-
line conditional discriminations. Only the derived mand probe will be described in this
section. This probe will test whether the learner exchanges the correct written word for
the item needed in the next step of the chained task. The same procedures provided in
the section on programming for the emergence of derived mands for preferred items
should be followed for assessing the existence of matching dictated name to correspond-
ing picture, matching dictated name to corresponding text, picture naming, text reading,
and picture-to-text and text-to-picture matching skills prior to instruction.
Instructions
Present the materials needed to complete the chained task (except the item that is
selected for training) and up to five textual stimuli (for example, three words that will be
trained and two distracter words) on a binder with the instruction to begin the task (such
as “Make cereal”).
Order of Stimuli
Instructional stimuli are presented in a nine-trial block, with each mand tested three
times per nine-trial block. Therefore, each selected task will be completed three times to
allow for three opportunities to mand for each missing item. Positioning of the compari-
son textual stimuli should change from one trial to the next on the learner’s binder.
Order of Trials
The order of test trials should be randomly determined. Specifically, trials may be ran-
domized by presenting each task in a different order. For example, make cereal, prepare
hot chocolate, and watch video may be followed by a different sequence, such as watch
video, prepare hot chocolate, and make cereal.
Reinforcement
As is the case with all test probes, trials are conducted under extinction; therefore no
differential reinforcement or feedback is provided for correct, incorrect, or no response.
Once a trial is completed, the next trial in the block is presented. It is important that
the item specified in the mand is not delivered contingent upon correct text exchanges so

that the emergence of the derived mands can be inferred. Instead, the instructor should
perform the step in the chained task once a response is made during the “interruption.”
For example, if the learner exchanges the correct textual stimulus (for example, the printed
word “milk” when the milk is in fact needed in the chain), the instructor performs the
step of opening the container of milk and pouring the milk in the bowl. This is done
if the response is incorrect, or if no response is emitted as well. It is critical that any
nonspecific intermittent reinforcement provided between trials or breaks not be provided
contingent upon correct responses.
Programming for Maintenance and Generalization

It will be important to test whether all of the relations directly trained and those that
emerged after training are present within the learner’s repertoire over an extended period
of time. As with any instructional program, it is recommended that maintenance probes
be conducted once training has been completed. Maintenance probes should be con-
ducted in the same manner as post-test probes. Some studies have demonstrated that
emergent skills may be maintained up to one month following the initial training in par-
ticipants with severe developmental disabilities in the absence of regular opportunities to
display such skills (Rehfeldt & Root, 2005; Rosales & Rehfeldt, 2007). Ideally, however,
the skills acquired will be used on a daily basis.
Stimulus generalization should also be considered throughout all training phases.
Generalization of skills across settings, communication partners, and stimuli is of vital
importance. It is often the case that the materials presented during training will only
slightly resemble those that will be encountered by the learner in the natural environ-
ment. Generalization probes should be conducted before, during, and after training is
completed in order to assess the effects of the training. Generalization will be most likely
if training is conducted in different settings, with different communication partners, and
with task materials that vary slightly. For example, slightly different pictures of preferred
items and textual stimuli created with slightly different font patterns and sizes could be
used, as well as variations of task materials needed to complete chained tasks.
Topography-Based Vocal Derived Mands

The purpose of this section is to provide step-by step instructions on how to establish
derived manding skills for learners with vocal verbal repertoires, in which case the goal
is to program for the emergence of requests for desired items using category names.
Manding for items using their category names may be beneficial for a learner who does
not know or cannot recall the item’s specific name or may be in a situation where an indi-
vidual preferred item is not available but one or more categorically related items are. For
example, if there are several drink items available for which an individual has not learned
specific names, he or she may learn to mand using the category name “drink” after two
or more of these items have been categorically related via a history of derived relational
responding. Hence, in the presence of juice, soda, lemonade, water, milk, or powdered
Chapter 11 249

drink mix, the individual may learn to use the mand “drink” to request any one of these
available items without needing to be directly trained to use each specific mand.
Instructional materials will include at least nine preferred items that are grouped into
three superordinate categories of three stimuli each. For example, a ball, toy train, and toy
car would be categorically related in the category “toy”; popcorn, cookie, and chips would
be categorically related in the category “snack”; and juice, soda, and drink mix could be
categorically related in the category “drink.” A digitized photograph printed on a 3-by-
5-inch index card of each item will also be necessary, as will a stimulus placement board
for use during conditional discrimination training if a nonautomated training procedure
is being used.
Conduct Pre-test and Post-test Probes

Probes conducted before training evaluate whether the learner can relate items
belonging to the same superordinate category and mand for items using their category
names. Similar probes should be conducted again upon the conclusion of all training. The
derived mand probes will first be described, followed by descriptions of all other pre-test
probes. The derived mand probe will test whether the learner vocally requests preferred
items using the corresponding category name. We describe how to prompt, deliver rein-
forcement, and use instructions to facilitate correct responding. The ordering of stimuli
and trials, mastery criterion, and strategies for handling and preventing errors should all
be conducted in a manner identical to that described in previous sections.
Instructions for Derived Mand Probes

Present one preferred item at a time from each of the categories to be trained, along
with the instruction “What do you want?” An additional probe may be conducted where
the item is not in view of the learner. This probe would help to ensure that any mands
emitted by the learner are pure mands, as opposed to mands for items under control of
the visual stimulus (see chapter 5 of this volume).
Order of Stimuli
Instructional stimuli are presented in nine-trial blocks, with each mand tested three
times per nine-trial block. In this case, there will be a total of six derived mands tested (for
example, two stimuli from each category). Therefore, two blocks of nine trials each should
be conducted. For example, if the categories are “toy,” “snack,” and “drink,” two stimuli
from each category will be tested (for example, to see whether the learner will request
both train and car using the corresponding category name “toy” after being trained in
the category name with the ball).
Instructions for Stimulus Relation Probes

Present the sample stimulus (a picture of a preferred item from one superordinate
category) with the verbal instruction “Put with same” or “Match” and three comparison

stimuli, A, B, and C, evenly placed on the stimulus placement board directly in front of
the learner. This conditional discrimination procedure is a visual-visual matching task, in
which sample and comparision stimuli are all visual in modality (i.e., pictures). Below are
instructions for conducting the derived stimulus relations probes.
1. Matches pictures by category A-B. For example, the learner places the picture
of the sample stimulus an apple, on top of the picture of the correct comparison
stimulus, grapes, when the comparison stimuli are pictures of items from three
different categories.
2. Matches pictures by category A-C. For example, the learner places the picture
of the sample stimulus from the same category, an apple, on top of the picture
of the correct comparison stimulus, a banana, when the comparison stimuli are
pictures from three different categories.
3. Matches pictures by category B-A. For example, the learner places the picture
of the grapes on top of the picture of the apple.
4. Matches pictures by category C-A. For example, the learner places the picture
of the banana on top of the picture of the apple.
5. Relates pictures of the same category B-C. For example, the learner places the
picture of the banana with the picture of the grape.
6. Relates pictures of the same category C-B. For example, the learner places the
picture of the grapes with the picture of the banana.
Teaching Vocal Mands Using Category Names

The first step in training is to teach learners to request one preferred item from each
superordinate category using its category name. For example, if the three categories are
“fruit,” “toy,” and “drink,” the learner may be taught to request an apple, a ball, and juice
using the respective category name. If edibles will be used during this part of training,
it is especially important to consider the establishing operations in place at the time of
training. If possible, training sessions should be conducted before or at mealtime in order
to take advantage of the learner’s hunger. The same strategy should also be considered
for tangible items that the learner may have access to throughout the day. When training
sessions begin, it may work to the instructor’s advantage to reserve tangible items that are
used during training sessions for these times alone (in other words, avoid use of the item
at any other time of the day), or to conduct sessions at times when these items are usually
made available.
Instructions
Present each item one at a time in front of the learner but out of his or her reach,
asking the question “What do you want?”
Chapter 11 251

Reinforcement
A correct response consists of the learner using the correct category name to request
the specific item. Independent correct responses should be differentially reinforced with
descriptive verbal praise (such as “You want fruit!”) and access to the item for thirty
seconds (to either consume or manipulate the item). If the learner does not respond
within five seconds or responds incorrectly, the trial should be marked incorrect on the
data sheet. Incorrect responses should result in the delivery of corrective verbal feedback
(such as “Try again”) and an appropriate prompting technique as described below.
Prompts
After the first three trial blocks, a graduated time-delay procedure may be introduced
to promote correct responding if reinforcement alone is not producing correct responses.
With this procedure, a vocal prompt should be presented immediately following the pre-
sentation of the item (for example, “Say ‘fruit.’”). Correct responses should then be rein-
forced. The delay before the presentation of the vocal prompt should be systematically
increased to four seconds following the presentation of the item and the instructions.
Indirect prompts may also be provided to the learner if he or she uses the correct indi-
vidual item name to mand for the item (for example, if the learner says “apple” instead of
“fruit”). In this case, a prompt such as “Okay; what else?” might suffice to evoke a correct
response.
Conditional Discrimination Teaching

In these conditional discrimination sessions, learners are taught to relate pictures of cat-
egorically related preferred items. This conditional discrimination teaching consists of a
visual-visual matching task in which the sample and comparison stimuli are all items
presented in the form of a picture. Figure 11.3 illustrates examples of relations that are
trained and tested.
B C
Figure 11.3. Trained and derived relations for stimuli A, B, and C. Derived relations are
depicted by dashed arrows and trained relations by solid arrows.

Instructions
Present the sample stimulus (such as an apple) and simultaneously present three com-
parison stimuli (such as a train, a grape, and a container of milk) evenly spaced on the
stimulus placement board in front of the learner, with the instruction “Put with same”
or “Match.”
A-B, A-C, and mixed A-B and A-C training should be conducted in that order.
The order of stimuli and training trials, prompts and reinforcement, mastery criterion,
and handling and prevention of errors should be handled in the same manner as that
described in the prior sections on conditional discrimination teaching.
Future Applications
Future applications stem from the work of Murphy and Barnes-Holmes (Murphy,
Barnes-Holmes, & Barnes-Holmes, 2005; Murphy & Barnes-Holmes, in press) in which
high-functioning children with autism demonstrated the emergence of derived mands
by requesting “more” or “less” smiley-face tokens for a board game. Teaching frames of
coordination and opposition in order to increase spontaneous communication may be
beneficial for learners with more sophisticated verbal repertoires. For example, a variety
of names or labels to represent “more” could be established by teaching synonyms (such
as “some” and “extra”) and antonyms (such as “less” and “fewer”) via conditional dis-
crimination training. Learners could be taught first to conditionally relate words with the
same or similar meaning and then to conditionally relate words that have the opposite
meaning. Tests for derived relations would evaluate whether the learner demonstrated
that the words were equivalent and whether he or she used them in the same functional
manner (for example, requests for “more” using the words “some” and extra”) without
further training. Alternatively, a learner could be taught to conditionally relate these same
words within the context of a math lesson. In this situation, the derived relations would
test to see if the learner could perform math problems when different instructions were
presented (such as “Add” or “Sum” and “Take away” or “Subtract”).
This chapter presented an overview of how a curriculum might be implemented in
order to program for the emergence of derived manding skills. Given the immense practi-
cal benefits of using this technology to teach skills to individuals with severe disabilities,
it is hoped that these tools will be put to good use.
References
Cooper, J. O., Heron, T. E., & Heward, W. L. (2006). Applied behavior analysis (2nd ed.).
Columbus, OH: Merrill Prentice Hall.
Chapter 11 253

Frost, L. A., & Bondy, A. S. (1994). The picture exchange communication system training
manual. Cherry Hill, NJ: Pyramid Educational Consultants.
2, 72–85.
Halvey, C., & Rehfeldt, R. A. (2005). Expanding vocal requesting repertoires via rela-
tional responding in adults with severe developmental disabilities. Analysis of Verbal
Michael, J. (1993). Establishing operations. Behavior Analyst, 16, 191–206.
Murphy, C., & Barnes-Holmes, D. (in press). Derived more/less relational mands with
four children diagnosed with autism: Synthesizing Skinner’s Verbal Behavior with
relational frame theory II. Journal of Applied Behavior Analysis.
Murphy, C., Barnes-Holmes, D., & Barnes-Holmes, Y. (2005). Derived manding in
children with autism: Synthesizing Skinner’s Verbal Behavior with relational frame
theory. Journal of Applied Behavior Analysis, 38, 445–462.
Rehfeldt, R. A., & Root, S. L. (2005). Establishing derived requesting skills in adults with
severe developmental disabilities. Journal of Applied Behavior Analysis, 38, 101–105.
Rosales, R., & Rehfeldt, R. A. (2007). Contriving transitive conditioned establishing
operations to establish derived manding skills in adults with severe developmental
disabilities. Journal of Applied Behavior Analysis, 40, 105–121.
Sulzer-Azaroff, B., & Mayer, R. G. (1991). Behavior analysis for lasting change. Belmont,
CA: Wadsworth/Thomson Learning.
Touchette, P. E., & Howard, J. S. (1984). Errorless learning: Reinforcement contingen-
cies and stimulus control transfer in delayed prompting. Journal of Applied Behavior
Analysis, 17, 175–188.

Part 3
Self, Reasoning, Problem Solving,

and Creativity
Chapter 12
Training Analogical Reasoning

as Relational Responding
Ian Stewart, National University of Ireland, Galway; Dermot

Barnes-Holmes, National University of Ireland, Maynooth;
and Tim Weil, University of Nevada, Reno
Analogy is regarded by many psychologists as the core of intelligent behavior and one of
the most important elements of its development (Sternberg, 1985). “Analogy” comes from
the Greek ana logon, “according to a ratio,” and Greek mathematicians used the word
“analogia” to denote a similarity in proportional relationships; it was, of course, Aristotle
who formalized the concept of analogy using the classic proportion schema “A : B :: C :
D” (spoken as “A is to B as C is to D”). Thus, it is not surprising that a critical feature
of analogy is the transfer of existing knowledge from one context to another through an
assessment of the relationship or overlap between them. As a result, analogical reasoning
may underpin much of our untrained behavior in novel contexts because it enables us to
use existing knowledge of familiar contexts to guide such untrained behavior in novel
contexts.
The first part of the current chapter will outline traditional psychological approaches
to understanding analogical reasoning and its development. A modern behavioral account
based on the concept of derived stimulus relations will then be articulated. The main
focus of the chapter thereafter will be to present a comprehensive protocol of steps and
instructions. This should allow practitioners to use principles and practices of stimulus
relations to establish analogical reasoning skills where they are found to be deficient. The
latter sections of the chapter will present a second protocol based on more traditional
analogy tasks that provide useful examples of how to assess the utility of the features of
relational training and testing in the former protocol. Taken together, the latter train-
ing tools should offer useful and important information and guides to the establishment
or enhancement of analogical reasoning abilities in developmentally delayed or devel-
opmentally disabled populations. Whatever one’s academic orientation, there is general
consensus among psychologists that understanding analogy is central to higher cognitive
processing, and that deficits in this regard considerably impair an individual’s cognitive
development.
Traditional Accounts of Analogical Reasoning

Because of the assumed centrality of analogical reasoning in higher cognition, mainstream
cognitive psychologists devoted considerable energies to understanding the basic processes
involved in solving analogies. Again, consistent with understanding the behavior as basi-
cally relational in nature (see etymological roots above), they described an analogy as the
transfer of relational information from a domain that already exists in memory (usually
referred to as the “source,” the “base domain,” or the “vehicle”) to the domain to be
explained (referred to as the “target domain” or the “topic”). That is, in the “A : B :: C :
D” example, you are being asked to transfer information about relations within the source
pair A-B to the target pair C-D. In standard analogy tasks you may be provided with all
four terms and asked to identify the relationship between the two pairs, or the D term
may be missing and you have to specify what it should be. In any case, the relations across
the pairs are coordinated; in other words, the relations between A and B are the same as
the relations between C and D.
Because of the view that the relations contained within analogies are complex (often
referred to as “higher order”), developmental psychologists traditionally believed that chil-
dren under twelve years of age could not solve analogies (e.g., Piaget, Montangero, &
Billeter, 1977). According to Piaget, higher-order relations were present only in formal
operational thinking, which is not fully developed until adolescence. Consider the follow-
ing task: “Bicycle is to handlebars as ship is to ?” In this case the relation-
ship between A and B refers to the steering mechanism, so the correct answer is “rudder.”
With this particular example, it is perhaps not surprising that children under twelve
produced low levels of accuracy.
In attempting to identify the incorrect strategies that the young children in his
studies may have used to try to solve the analogies, Piaget carefully examined their errors.
Specifically, one type of incorrect answer was indicative of simple associations, rather than
higher-order relations. For example, when presented with “Bicycle is to handlebars as ship
is to ?” the children selected “bird” with the justification that “both birds
and ships are found on lakes.” In other words, the children in this case were focusing
solely on the “lower order” relation within the incomplete pair (C and D) and ignoring
completely the first pair (A and B) and, naturally, the higher-order relations among the
pairs.
In subsequent research, Goswami and Brown (1990) closely examined the types of
simple associations that children can use incorrectly in attempts to solve analogies. Consider
the picture sequence from their research: “Bird is to nest as dog is to ”; the
options are doghouse, cat, another dog, and a bone. A correct choice based on analogical
reasoning involves selecting the doghouse, by deriving the target relation of habitat. In

contrast, a simple associative choice would involve selecting the bone (because dogs and
bones frequently go together). Furthermore, the selection of the other dog would indicate
a match on the basis of surface similarity derived primarily from perceptual features,
while the selection of the cat would indicate a category match. The results of the study
indicated that children between four and nine years old performed at levels significantly
above chance, with accuracy increasing systematically with age. Naturally, the researchers
concluded that children as young as four or five can demonstrate analogical reasoning,
although increasing age enables them to rely less upon purely associative strategies and
more on higher-order relations.
One confounding feature, however, that may have influenced the research by both
Piaget and Goswami and Brown has to do with the level of the children’s familiarity
with the target pairs and the relations between them. Specifically, in Piaget’s research,
the children’s analogical reasoning skills may have been underestimated because they did
not have adequate experience with or knowledge of steering mechanisms, for example.
Furthermore, perhaps the age differences recorded by Goswami and Brown resulted
simply from different levels of experience with birds, nests, and so on, rather than differ-
ences in relational competence.
To explore this possibility, Goswami and Brown (1989) designed a series of item
analogies based on causal relations (for example, cutting, wetting, and melting) with
which most three- or four-year-olds are familiar. Consider the following task: “Play dough
is to cut play dough as apple is to .” This series of tasks was presented in
conjunction with additional tasks that determined the children’s level of knowledge of
the target causal relations. For example, they were shown three pictures of items that
had been causally transformed (for example, cut play dough, cut bread, cut apple) and
were asked to select the causal agent responsible for the transformation (a knife, water,
the sun). The results of the study indicated a strong correlation between performances
on the analogies and performances on the control tasks, thus indicating interdependence
between analogical reasoning and familiarity with the target relations. This outcome has
two implications. First, relational knowledge and analogical competence develop natu-
rally together. Second, where relational knowledge is strong, children as young as four
appear likely to be able to derive the target relations involved in some analogies. However,
where relational knowledge is weak, strategies that are associative and likely incorrect will
dominate.
A Behavioral Approach to Analogy

Until the 1990s, behavioral psychologists showed little or no interest in analogical reason-
ing. The advent of relational frame theory (RFT) naturally changed this, given the estab-
lished relational features of analogy and the fact that relational responding is the central
tenet of the theory in question.
Relational frame theory posits coordination (or equivalence) relations as the simplest
type of relation that can be derived and most likely the first to emerge in the natural
developmental sequence (e.g., Lipkens, Hayes, & Hayes, 1993). Put simply, if a learner
is instructed that A is the same as B, then he or she will readily determine that B is
Chapter 12 259

the same as A, even though this has not been instructed directly. In other words, the
coordination relation stated explicitly between A and B enables the learner to derive the
mutually entailed coordination relation between B and A, assuming an appropriate train-
ing history. Furthermore, if the same learner is then trained to match B and C, then
more-complex A-C relations of coordination or equivalence will emerge between these
stimuli, which have never been directly paired and in a direction that was never trained
(in other words, C-A relations also emerge). Of course, what is crucial for RFT is the
fact that derived relations are arbitrary and thus the relations are not based on formal
properties of the stimuli or events in question. For example, our C and A terms don’t go
together simply because they look like each other in some way, just as the word “window”
does not resemble an actual window.
Analogies, however, involve the derivation of relations both within and across pairs.
Hence, in the analogy “A : B :: C : D,” you are asked to derive a relation between A and
B and between C and D, and also to derive the higher-order relation that connects C to
D in the same way that it connects A to B. Consider the analogy presented in figure 12.1
and denoted as “apple : orange :: cat : dog.”
APPLE CAT
Equivalence-
Equivalence
Equivalence Relation Equivalence
Relation Relation
ORANGE DOG
Figure 12.1. The analogy denoted as “apple : orange :: cat : dog” and the relations that
pertain among the elements.
Barnes, Hegarty, and Smeets (1997) provided the first RFT model of analogical
reasoning as the derivation of equivalence relations between equivalence relations or
equivalence-equivalence responding (see figure 12.1). According to this view, apple and
orange participate in a relation of equivalence (because both are fruits), while cat and dog
also participate in a relation of equivalence (because both are animals). A correct analogi-
cal response involves the derivation of these two equivalence relations and the derivation
of a further equivalence relation between the relations (in other words, apple is equivalent
to orange in the same way that cat is equivalent to dog, because each are members of the
same respective class). In the study in question, the researchers reported that both nine-
and twelve-year-olds readily demonstrated the target equivalence-equivalence relations in
tests comprising analogies of this sort.
However, in the RFT model of analogical reasoning, researchers highlighted another
feature of analogy that overlapped with previous suggestions by Goswami and Brown
concerning the role of perceptual features. Specifically, Stewart, Barnes-Holmes, Roche,

and Smeets (2001) argued that the relations within analogies may be readily traced back
to perceptual, formal, or nonarbitrary properties shared by the pairs. Consider again
the analogy from figure 12.1: apple: orange :: cat : dog. In this example, the arbitrary
equivalence relation between the words “apple” and “orange” is based, to some extent,
on the coordination of physical properties shared by the actual stimuli with which the
words are coordinated (both actual fruits are small, spherical, edible, sweet, and so on).
Similarly, the arbitrary equivalence relation between the words “dog” and “cat” is based on
the shared physical properties between the animals (in other words, four legged, mobile,
hairy, and so on). Thus, although the two equivalence relations and the equivalence-
equivalence relation are entirely arbitrary (there are no formal similarities shared by the
words or shared between the words and the actual objects), they can be readily traced
back to shared nonarbitrary features. Put simply, the nonarbitrary features are brought to
bear automatically in the derivation of the coordination relations.
Examining the Development of

Equivalence-Equivalence Relations
If RFT researchers are correct in their belief that analogies comprise primarily
equivalence-equivalence relations, then you would expect that performances based on
these abilities would coincide with those recorded on more traditional tests of analogy.
Across several studies reported by Carpentier, Smeets, and Barnes-Holmes (2002), adults
and nine-year-olds successfully demonstrated equivalence-equivalence relations, but five-
year-olds did not. Indeed, it was only when the five-year-olds were explicitly trained on
the composite equivalence relations prior to exposure to the equivalence-equivalence test
that they finally demonstrated the target relations. Further empirical support for these
outcomes was subsequently reported by Carpentier, Smeets, and Barnes-Holmes (2003).
Although these outcomes of competent analogical reasoning in five-year-olds appear
to be consistent with Goswami and Brown’s research, the extensive relational pretraining
required by the children is not easily reconcilable with the previous research in which no
training appears to have been provided. The most important question then is whether
five-year-olds really solve analogies analogically (in other words, by deriving equivalence-
equivalence relations). According to RFT, derived relational abilities are fairly sophisti-
cated by age five, as is a general relational knowledge base. Given these two facts, there is
no reason to believe that five-year-olds could not solve some analogy tasks, although one
would need to be sure that the derivation of equivalence-equivalence relations, rather than
simple associations, formed the basis of responding. In any case, RFT would predict that
explicit and successful training in equivalence-equivalence responding should generate
improvements in traditional analogy tasks.
Although empirical evidence of this relationship is not available, there is adequate
empirical reason to believe that the claim is viable, and reason to suggest the utility of an
equivalence-equivalence training protocol as a sound means of training analogy. Indeed,
the protocol outlined below has been successfully employed by numerous researchers to
readily establish or facilitate equivalence-equivalence responding.
Chapter 12 261

Protocol 1: From Conditional Discriminations to
Equivalence-Equivalence Relations
Stimuli
Protocol 1 involves nine black and white arbitrary picture stimuli that are identified
by alphanumeric codes (A1, B2, C3, and so on; see figure 12.2). Each picture should
appear individually as a laminated white card (for example, A1 or B1), but, in addition,
each possible pair of stimuli should be presented in a compound picture in which the
items appear side-by-side on the same laminated card (for example, A1C2).
[FIG]
Figure 12.2. The picture stimuli and their alphanumeric codes employed in protocol 1.
Training and Testing Sequence

The basic training and testing sequence in protocol 1 consists of ten phases, presented
in table 12.1, that proceed with increasing complexity from training simple A-B rela-
tions to testing equivalence-equivalence relations. In all of the phases that follow, train-
ing incorporates feedback on all incorrect responses. During testing, responses have no
programmed consequences.

Table 12.1. The Basic Training and
Testing Sequence in Protocol 1
Phase 1: A-B Training & Testing
A1-B1 (18/12) A2-B2 (18/12) A3-B3 (18/12) A-B Mix (36)
Phase 2: A-C Training & Testing

A1-C1 (18/12) A2-C2 (18/12) A3-C3 (18/12) A-C Mix (36)
Phase 3: Mixed A-B and A-C Testing

A-B and A-C Mix (18)
Phase 4: Matched Compound Training & Testing

A-B—AB (18/18)
A-C—AC (18/18)
A-B—AB & A-C—AC Mixed Testing (36)
Phase 5: Matched Compound-Compound Training & Testing

AB—AB (18/18)
AC—AC (18/18)
AB—AB & AC—AC Mixed Testing (36)
Phase 6: Nonmatched Compound-Compound Training & Testing

AB—AB Training (18/18)
AC—AC Training (18/18)
AB—AB & AC—AC Mixed Testing (36)
Phase 7: Symmetry Testing (& Training)

B-A & C-A Mix (36/36)
Phase 8: Equivalence Testing (& Training)

B-C & C-B Mix (36/36)
Phase 9: Derived Matched and Nonmatched Symmetry

Compound-Compound Testing (& Training)
BA—BA & CA—CA Mix (36/36)
Phase 10: Equivalence-Equivalence Testing

BC—BC & CB—CB Mix (36)
Chapter 12 263

Phase 1: A-B training and testing. The set of nine picture stimuli comprises three A-B
relations: A1-B1, A2-B2, and A3-B3. Each of these relations is trained in sequential steps,
working from left to right across table 12.1.
All baseline trials are conducted in a matching-to-sample format that involves the
presentation of one sample stimulus (A1, A2, or A3) and three comparison stimuli (the
three B stimuli). So, when the relation A1-B1 is being trained, A1 will be the sample,
B1 the correct comparison, and B2 and B3 the incorrect comparisons. When A2-B2 is
targeted, A2 will be the sample, B2 the correct comparison, and B1 and B3 incorrect.
Finally, when A3-B3 is trained, A3 will be the sample, B3 the correct comparison, and
B1 and B2 incorrect.
In order to properly counterbalance the locations of the comparison stimuli, espe-
cially the correct one, a teacher might use six possible configurations in the training of
each of the three A-B relations; see figure 12.3. Correct comparisons are identified by the
symbol +. A1-B1 training should involve three exposures to each of these, thus yielding
a total of eighteen A1-B1 training trials (the first number within parentheses in table
12.1). It is better that these are presented from the outset in a randomized block of eigh-
teen trials, but for learners with limited experience, or greater difficulty, in matching to
sample, it may be necessary to group identical trials together (in other words, working
through figure 12.3 systematically, with three exposures to each). Remember, however, all
of the training will eventually have to be randomized before moving on to the next step.
A1 A1 A1
B1 B2 B3 B2 B1 B3 B2 B3 B1
+ + +
A1 A1 A1
B3 B2 B1 B3 B1 B2 B1 B3 B2
+ + +
Figure 12.3. The six possible configurations contained within the training of A-B
relations.
At the beginning of each trial, the teacher places down all four cards, with the sample
on top and the comparisons side-by-side below. The teacher then simply points to the
sample and the learner is asked to match the comparison that goes with the sample. For
some learners, hearing the single word “match” alone will be adequate, while others may
need to hear “Which other picture [pointing to comparisons] goes with this one [point-
ing to sample]?” In a block of eighteen training trials, criterion is reached with eighteen
consecutively correct responses. In all cases, training must continue in blocks of eighteen
until this criterion is reached.

There is often debate about what to do next—test the A1-B1 relations you have just
trained or proceed to training A2-B2. Essentially, it depends on the individual learner, but
presenting a brief A1-B1 test immediately after the A1-B1 training may be safer because, if
the learner passes the test at that point, you can be sure that the relations you just trained
are intact. Because it is very likely that the learner trained easily in eighteen trials and that
the test will be passed, it is often only necessary to present twelve test trials, involving two
exposures to each of the comparison configurations (the second number in parentheses in
table 12.1). Ten correct responses out of the twelve constitutes a pass. If the learner fails
the test, retrain the relations and retest. Continue to do so until the test is passed.
Training the A2-B2 relations is identical to the A1-B1 training, except that A2 is now
the sample and B2 is the correct comparison. In training these relations, work out the
same comparison configurations displayed for A1-B1 in figure 12.3. Passing the A2-B2
test will allow the learner to proceed immediately to training of the A3-B3 relations,
which is identical in format to those before.
Once the three target A-B relations have been trained, tested, and passed individu-
ally, it is wise to have a separate test block that is made up of the three previous tests,
thus mixing A1-B1, A2-B2, and A3-B3 test trials, in a total of thirty-six trials (see table
12.1). Again, these should be presented in a random sequence. If the learner fails the test
by reaching less than thirty-three correct responses, go back and train those relations that
comprise the errors and repeat the test until passed.
Phase 2: A-C training and testing. At this point in the learning sequence, all three
target A-B relations are established and it is time to move on to the A-C relations (A1-C1,
A2-C2, and A3-C3). Naturally, in this case the sample stimuli are the same as before but
C1, C2, and C3 are the comparisons (rather than the B stimuli). These are trained and
tested in exactly the same way as the A-B relations (see table 12.1).
Phase 3: Mixed A-B and A-C testing. It is now time to present a randomized test of all
of the A-B and A-C relations combined. In constructing this test, you can simply select
any three trials from each relation (such as A1-B1) without worrying too much about
comparison configurations. This will yield a total of eighteen test trials that should be
randomly presented, thus mixing up all six types of relations tested thus far. If the test
is failed with fewer than sixteen correct responses, go back and retrain the relations in
which the errors occurred, and retest until the mixed A-B and A-C test block is passed.
Phase 4: Matched (A-B—AB and A-C—AC) compound training and testing.

Phase 4 represents the first point at which compound stimuli appear. The purpose of the
compound stimuli here is to teach the learner to treat the compound stimuli as function-
ally equivalent to the corresponding sample-comparison configurations (for example, the
trained relation A1-B1 as equal to the novel compound A1B1), because this will be impor-
tant later on for the equivalence-equivalence test. Put simply, you take the simple A-B
and A-C relations you have trained thus far and put them together to form compound
stimuli. It is important to emphasize at this point that the compound stimuli employed
in phase 4 are consistent with the A-B and A-C relations previously trained. Although the
compounds presented later will not necessarily be consistent with the previous training,
the use of the matching compounds here is designed to smooth the transition from what
was previously learned; otherwise the previous training may actually interfere with the
Chapter 12 265

new relational targets. So, the basic aim is simply to identify a relation and enable the
learner to match a compound based directly upon it. The six AB and AC compounds
derived directly from the previous training are presented in figure 12.4.
A1B1 A1B2 A1B3 A2B1 A2B2 A2B3 A3B1 A3B2 A3B3

+ +
A1 A2 A3
B1 B2 B3 B1 B2 B3 B1 B2 B3
A1C1 A1C2 A1C3 A2C1 A2C2 A2C3 A3C1 A3C2 A3C3

+ + +
A1 A2 A3
C1 C2 C3 C1 C2 C3 C1 C2 C3
Figure 12.4. The six AB and AC compounds derived directly from the previous training.
Black line indicates the correct relation; + indicates that the target sample-comparison
relation is appropriate in the context of that compound stimulus.
Consider the following A-B—AB compound training trial. As before, the teacher
presents an A stimulus as the sample (such as A1) and three B stimuli as comparisons.
However, now, immediately above the sample, three A-B compounds are presented, one
directly above each comparison. The learner is instructed as in the following example:
The teacher points to the A stimulus and says, “If this [A] goes with this [pointing to the
appropriate B stimulus], which one of these [pointing to the three compounds] would
match?” In essence, the compounds require the learner to attend to the A-B relation pre-
sented and match this with both elements of each compound. In all cases, the compounds
are simply a collection of the sample-comparison pairs.
It is probably easier to train the AB compounds first, excluding AC for now. This type
of training might be presented as a block of eighteen trials, allowing six exposures to each
of the trial types presented in the top row of figure 12.4. Again, this would permit clear
randomization of the correct comparison and the correct compound. Although, as usual,
the trials should be presented in a random sequence, it may be preferable, for a particular
learner, to separate the trials and have six consecutive exposures to each of the three trial
types (as in figure 12.4), but again they must be randomized at some stage. A-B—AB
compound testing should follow the training, and criterion must be reached here on
a randomized block before you proceed. Training and testing the matched A-C—AC
compounds will be identical. Phase 4 should end with a mixed block of randomized
A-B—AB and A-C—AC compound test trials.
Phase 5: Matched (AB—AB and AC—AC) compound-compound training and

testing. Phase 5 involves greater complexity than phase 4 because learners are required

to make compound-compound matches, rather than relation-compound matches, and
this edges them ever closer toward the format of the final equivalence-equivalence test.
However, all of these matches are consistent with the baseline relations established previ-
ously. The tasks here involve matching baseline relational compounds (for example, A1B1)
to other baseline relational compounds (for example, A1B1 with A3B3 and A2C2 with
A1C1). In other words, a coordination relation within one compound is matched to a
coordination relation within another compound. Note also that on this task, an added
element of flexibility appears with the presentation of only two comparison compound
stimuli. Again, this feature increases the task’s similarity with the equivalence-equivalence
test.
If you combine the three types of A-B relations, you get two configurations in which
each one is presented with the two others; see figure 12.5. Consider the compound A1B1
as a sample. On one trial it can be matched with A2B2 (top left) and on another trial
it can be matched with A3B3 (top right). The same occurs for A2B2 as a sample (with
A1B1 and A3B3, middle) and for A3B3 (with A1B1 and A2B2, bottom). In all cases the
incorrect comparison is a compound based on a relation that was not previously trained
(for example, A3B2). Hence, the learner must now match the coordinated relation in
the sample compound with the coordinated relation in the comparison compound and
exclude the difference relation within the incorrect comparison compound. Put simply,
two sames go together, but a same and a different do not. During these trials, the learner
may be simply instructed, “Which one of these [teacher points to the two compound
comparisons] goes with this [pointing to sample compound]?”
The matched AB compound-compound training may be presented in blocks of eigh-
teen trials, thus allowing three exposures to each of the six trial types from figure 12.5
and allowing for adequate use of alternative incorrect compounds and randomization
of stimulus locations. Again, this should be followed by appropriate testing of the same
compounds. Naturally, the AC compounds can be constructed in the same way, and
the training and testing would be the same. Again, this phase would end with a mixed
block of randomized test trials containing the AB—AB compounds and the AC—AC
compounds.
Phase 6: Nonmatched compound-compound training and testing. You may have

noticed that in the previous phase we introduced difference relations within compounds
for the first time. In that phase, these always constituted incorrect comparisons, but in
phase 6 the compound relations do not necessarily match those trained in the baseline
relations, and thus difference-based compounds can now participate in correct matches.
Consider the task of matching A2B3 with A1B1 or A1B2. In this case, the sample stimu-
lus now contains a relation that does not match the baseline relations; hence it is a rela-
tion of difference (or noncoordination). As a result, it should not be matched with A1B1,
because its relation is one of coordination; instead, it should be matched with A1B2,
another difference relation. This is an important feature of the current phase, because the
learner must discriminate that difference relations can also be matched together (as the
same) in the same way as coordinated relations can. Put simply, two difference relations
are the same in just the same way as two same relations are.
The matching of difference relations can be constructed through a range of trial
types. But we would suggest that you keep the AB and AC compounds separate as before
Chapter 12 267

A2B2 A3B2 A2B1 A3B3
A1B1 A1B1
A1B1 A1B2 A3B2 A3B3
A2B2 A2B2
A1B1 A2B3 A1B2 A2B2
A3B3 A3B3
Figure 12.5. The two configurations that emerge when the three types of A-B relations are
combined. Black line indicates the correct relation.
and use two comparisons, one of which has a compound that matches with the baseline
relations and the other that has a nonmatching compound. Of course, the sample will
be unmatched, so the correct comparison will again be the unmatched one. Remember
again to randomize your comparison locations and train, then test (for guidance, see table
12.1).
Phase 7: Symmetry (B-A and C-A) testing and training. It may have been a surprise
to you if you noticed that so far we have incorporated the type of complexity that comes
with matching two compound stimuli, and yet we haven’t even established symmetry or
equivalence. While this is unusual in traditional matching-to-sample training regimes,
we did it this way in order to allow the learner to proceed naturally from those relations
that were trained directly to compounds that were, or were not, consistent with them.
Interjecting symmetry or equivalence training at some point in that process might have
hindered the student more than it helped. However, if we want to target equivalence-

e quivalence relations, we must naturally ensure that our learner first demonstrates symme-
try and equivalence. The present training sequence is advantageous not just in facilitating
compound-compound matching, but also because after working with same and different
compounds, the student will probably find symmetry and equivalence not too difficult.
Phase 7 targets the derived B-A and C-A symmetry relations that should emerge as
a result of training the baseline A-B and A-C relations. These are B1-A1, B2-A2, B3-A3,
C1-A1, C2-A2, and C3-A3. The format for presenting these trials is identical to that of
presenting the baseline relations in phase 1 (one sample and three comparisons). You
should really test the symmetry relations first, rather than training them, because at this
stage they are most likely intact. Hence, we would advise a randomized and mixed thirty-
six-trial test block with six exposures to each of the target B-A and C-A symmetry rela-
tions. This will likely preclude the need for explicit training. However, in the event that
the test is failed, we would advise that you expose the learner to the baseline A-B and A-C
relations again and retest. In contexts such as these, we generally try to avoid explicitly
training relations that should be easily derived, and the symmetry here is a good example
of that. Put simply, the more explicit the training, the less relational derivation. If neces-
sary, of course, the symmetry relations can be explicitly trained, but make sure you test
again afterward.
Phase 8: Equivalence (B-C and C-B) testing and training. In order to target
equivalence-equivalence relations, we must be absolutely sure that the composite equiva-
lence relations are in place. These include the derived B-C and C-B relations that should
emerge as a result of training the baseline A-B and A-C relations. These are B1-C1, B2-C2,
B3-C3, C1-B1, C2-B2, and C3-B3. Again, the format is identical to phase 1. In this case
also, you are advised to test the equivalence relations first, rather than training them,
because they are most likely intact. Again, we suggest a randomized thirty-six-trial test
block with six exposures to each of the target relations. As was the case with symmetry,
the more explicit the training, the less relational derivation.
Phase 9: Derived (BA—BA and CA—CA) matched and nonmatched symmetry

compound-compound testing and training. This phase is similar to phases 5 and
6, except that the learner is now matching the symmetry compounds derived from the
trained baseline relations. You will see from table 12.1 that we advise mixing the BA—BA
and CA—CA trials from the outset, as well as the matched and nonmatched trials. The
more mixing employed here, the greater will be the learner’s flexibility and trial atten-
dance. For illustrative purposes, we have outlined a sample of the BA—BA trials in figure
12.6. Again, test before training and train only if you must. It is important to note
that the incorrect comparison compounds here, in both matched and unmatched trials,
should contain a stimulus that is physically identical to one in the sample. Consider the
following trial: B1A1 is the sample, with B3A3 and B2A1 as comparisons. In this case
A1 within the latter comparison is also in the sample. This is important because, if the
learner is matching by pure association, he or she will incorrectly select B2A1 in the pres-
ence of B1A1 simply because both contain A1, rather than correctly selecting B3A3 with
B1A1 because both contain coordination relations.
Chapter 12 269

BA Compounds (18)
Matched BA Compounds (9) Unmatched BA Compounds (9)
B3A3 B2A1 B1A1 B3A2
B1A1 B2A1
B1A1 B2A3 B3A3 B2A3
B2A2 B3A2
B1A1 B1A3 B2A2 B1A2
B3A3 B1A3
Figure 12.6. A sample of the BA—BA trials.
Phase 10: Equivalence-equivalence (BC—BC and CB—CB) testing. Equivalence-

equivalence testing is almost identical to the symmetry compound testing in the previous
phase, except that the BC and CB relations are targeted. Again, the more mixing the
better, and the learner must have adequate flexibility at this stage to deal with whatever
the test presents. However, if the learning has been arduous, you can keep the matched
and unmatched and BC and CB trials separate in a first exposure, especially if you had

done so in the previous phase. In this case, present that type of test exposure first and a
more randomized test exposure subsequently. Again, remember to include comparisons
that allow incorrect matching based on association. In the unlikely event that the learner
fails the test, reexpose him or her to phase 8 to check whether the equivalence relations
themselves are intact, and work forward from there once more.
Once the trained and tested equivalence-equivalence relations targeted above have
been demonstrated, it is useful to try to walk the learner all the way through the sequence
again with a novel stimulus set and therein reduce the amount of explicit training as
much as possible. This training regime should continue until the teacher is certain that
the learner is readily deriving all of the target relations. Naturally, there may be small
intermediate steps between phases that will be necessary for an individual learner and
these can be readily incorporated. However, remember that the critical features are ran-
domizing, avoiding explicit training of all relations, and using novel stimulus sets.
When such a regimen of training and testing is complete, it is important to provide
numerous examples of analogy in order to test the utility of the protocol, particularly
on traditional analogy tasks (such as those presented on standard IQ tests). As a result,
protocol 2, presented in the next section, allows the teacher to walk through examples of
numerous different types of traditional analogies.
Protocol 2: Assessment and Training of Traditional

Four-Item Analogies
The protocol below describes the steps commonly followed for the training of tradi-
tional four-item analogies.
Stimuli
The stimuli used in training and testing the traditional analogy tasks in protocol 2
are common black and white images (there are seven in each task). As was true with the
previous protocol, these should be presented on white laminated cards, and in this case all
of the pictures are presented on separate cards. In addition, you should also have a blank
white laminated card identical in size to the others, as well as a card that is blank except
for a question mark (see figure 12.7).
Analogy Tasks
The analogy problems provided here are classified into three groups: (1) causal analo-
gies; (2) standard analogies; and (3) proportional analogies. Tables 12.2, 12.3, and 12.4
contain ten sample tasks from each type, respectively. Solutions to all of these analogies
should only be possible when one has considered all three other terms contained within
the two premises (Carpentier, Smeets, Barnes-Holmes, & Stewart, 2004).
Chapter 12 271

Figure 12.7. The stimuli used in training and testing the traditional analogy tasks in
Protocol 2.
The causal analogies are similar to those employed by Goswami and Brown (1989) and
involve simple causal relations (for example, apple : cut apple :: banana : );
see table 12.2). The standard analogies are similar to those employed by Goswami and
Brown (1990) and are more difficult to solve because they are based on more complex
and broader relations (for example, spider : web :: bee : ; see table 12.3). The
third group of problems presented here involves analogies of numerical proportion (for
example, ** : **** :: [[ : ; see table 12.4). Although these might appear easier
to solve because they do not require knowledge of common verbal relations, they may in
fact be more difficult because they are abstract.
All analogy trials involve the presentation of two rows of cards, with the three cards
that make up the analogy stem on the top along with the missing item, and the response
alternatives on the bottom (see figure 12.7). Naturally, the cards within the analogy stem
(often denoted as A, B, and C) are sequenced to match the analogy, the question mark is
at the end (the missing D term), and the response options are randomized.
Training Sequence
The presentation of each analogy involves a number of short phases that include
explicit naming of all items and justification for responding (see figure 12.8). Although
these are perhaps surplus to the analogy solution itself, they do provide useful informa-
tion as regards the types of errors commonly made by learners, and we have found these
to be important in our experimental work. In any case, they don’t make the trials much
longer.

Table 12.2. Sample Tasks
from Causal Analogies
Analogy Stem Response Options
a : b :: c : Analogy
Apple : cut apple :: banana : Cut Hold Cut Eat apple
banana banana orange
Towel : wet towel :: paper : Wet paper Cut paper Wet hat Fold towel
Ball : throw ball :: stick : Throw Break stick Throw Kick ball
stick stone
Banana : eat banana :: nut : Eat nut Hold nut Eat orange Eat
banana
Dog : feed dog :: cat : Feed cat Pet cat Feed Call dog
goldfish
Face : wash face :: hand : Wash Hold hand Wash foot Cover face
hand
Hat : wear hat :: shoe : Wear shoe Polish shoe Wear coat Lift hat
Floor : wash floor :: wall : Wash wall Paint wall Wash car Build wall
Hand : lift hand :: foot : Lift foot Stamp foot Lift head Wave
hand
Cake : cut cake :: bread : Cut bread Eat bread Cut paper Bake cake
Table 12.3. Sample Tasks from

Standard Analogies
Spider : web :: bee : Hive Honey Nest Fly
Dog : dog basket :: baby : Crib Bottle Bed Doll
Train : tracks :: boat : Water Sailor Road Sail
Dress : hanger :: coat : Hook Sleeve Hat stand Jumper
Bird : nest :: dog : Kennel Collar Igloo Cat
Gloves : hands :: shoes : Feet Laces Hat Socks
Chapter 12 273

Cow : milk :: hen : Egg Cluck Honey Chicken
King : crown :: policeman : Police hat Police car Cap Whistle
Bird : beak :: man : Mouth Boy Snout Woman

Window
Wall : cement :: window : Glass Plastic Shutter
cleaner
Table 12.4. Sample Tasks from

Analogies of Proportion
** is to **** as [[ is to [[[[ [ #### ***
>>> is to > as @@@ is to @ @@ $ >>
!! is to !!! as ## is to ### ##### && !!!

^^^^ is to ^ as XXXX is to X XXX ) ^^
$$ is to $ $ as && is to && & %% $
NN is to NNN as LL is to LLL LLLL PPP NNN
RRRR is to RRR as BBBB BBB BB YYY RR

is to
C C C is to CCC as D D D DDD DD SSS CCCC
is to
Z is to ZZZZ as F is to FFFF FF EEEE ZZ
WW is to W as BB is to B BBB TT WWW

Phase 1:
Naming the Analogy Stem
Phase 2:
Guessing the Analogy
Phase 3:
Guess Justification
Phase 4:
Naming the Response Options
Phase 5:
Solving the Analogy
Phase 6:
Analogy Justification
Phase 7:
Remediation
Figure 12.8. Training sequence for traditional analogies.
Phase 1: Naming the analogy stem. At the beginning of phase 1, the learner may be
instructed as follows:
We are now going to play a game about choosing cards with pictures on them.
During the game I am not allowed to tell you whether you are getting the
answers right or wrong.
But at the end of the game, you will get a treat.
Now let’s begin.
The teacher then places the top row of cards on the table, putting the two on the far
left (the A and B cards) close together and the two on the right (the C and D cards) close
together, hence creating a sizable space between A-B and C-D. The learner may then be
instructed as follows:
I have two cards with pictures over on this side. [Teacher points to A and B on
the left-hand side.]
And I have two different cards on this side. [Teacher points to C and D
on the right-hand side.]
These two [pointing right] are a little different because only one has a
picture [pointing to C], while the other has a question mark [pointing to D].
Now [pointing to A], can you tell me what this picture is, please?
Chapter 12 275

What about this picture [pointing to B]?
And finally this picture [pointing to C]?
In order to respond correctly during the picture-naming trials, the learner must say
the correct name of the stimulus or an approximation thereof (such as “bees’ house” for
“hive”). Verbal praise is provided for all correct names (such as, “Yes, you’re right; it’s a
bee hive”), while incorrect responses are prompted (“Is there anything else you think it
might be?”) or a correct answer is provided (“Okay, this is actually a hive. Can you see
that the picture is a hive?”). If the learner still appears uncertain regarding the functions
of the item, then the teacher should describe the picture in detail by drawing his or her
attention to its primary features (for example, saying, “This is where bees live; can you
see any flying in and out in the picture? They are going home.”). Accurate naming and
understanding of the meaning of all target stimuli must be ascertained or established in
this way.
Phase 2: Guessing the analogy. During phase 2, the learner is asked to try to guess the
picture that would complete the analogy, but without being offered the response options.
For example, the learner might be asked, “What picture do you think should go here
[teacher points to the question mark card], in order to finish the pattern?”
Phase 3: Guess justification. In this phase, the teacher simply asks the learner to explain
why he or she provided the previous response. It may indeed be the case that the sophis-
ticated learner can correctly guess the analogy without the response options, in which
case you can move directly to the next trial. However, a simple associative response may
be more likely, and thus the justification should be recorded accurately. Sometimes, the
learner will offer a random guess without paying any attention to the A and B pictures.
In this case, the learner should be prompted to look at the first two pictures (“But what
about these pictures? Can they help?”).
Phase 4: Naming the response options. Phase 3 is identical to phase 1, except that the
learner is required to name the four response options on the bottom row. All other aspects
are identical to those of phase 1.
Phase 5: Solving the analogy. In this phase, the learner is required to derive the ana-
logical relations between the three pictures on the top and the correct picture on the
bottom. The learner may be instructed as follows: “Now, which one of these pictures
[pointing in sequence to each of the four pictures in the bottom row] do you think will
go here [pointing to the question mark] so that we can finish the pattern?” The learner
is required to point to, or clearly indicate, which picture she selects and should also be
encouraged to name the picture. In any case, the teacher must be sure about the learner’s
intended choice.
Phase 6: Analogy justification. All aspects of this phase are identical to those of phase
3.
Phase 7: Remediation. In the remediation phase, the teacher should guide the learner
toward focusing explicitly on the relational elements of the task. So, the teacher first
highlights the relation between A and B and then instructs the learner to try to construct

the same relation between C and D. The teacher then guides the learner in looking for
this relation. Consider the following short dialog for the analogy spider : web :: bee
: (hive/honey/nest/fly).
Teacher: What is in this picture? [Teacher points to A.]
Learner: A spider.
Teacher: Very good. And what is in this picture? [Teacher points to B.]
Learner: A web.
Teacher: Very good. And what do you know about spiders and webs?
Learner: Spiders live in webs.
Teacher: Very good. So this picture is of a spider, and this picture is a web. And
the web is where the spider lives, so it’s his home.
Teacher: Now, let’s look at this picture. [Teacher points to C.] What is this?
Learner: A bee.
Teacher: Very good. Now, over here we have a picture of a spider [pointing to A]
and his home here in the web [pointing to B]. And over here we have a
picture of a bee [pointing to C]. So what do you think the bee is looking
for?
Learner: His home. [If the learner does not produce the correct answer then the
teacher should prompt it by returning to the case of the spider.]
Teacher: Very good. So what is a bee’s home?
Learner: A hive.
Teacher: Very good.
Summary of Protocol 2
Protocol 2 provided a training and testing sequence that is consistent with the tradi-
tional means of assessing and training analogies. With consistent behavioral instruction
and feedback, a verbally sophisticated learner should be able to proceed through with
limited remediation. Put simply, this type of training and testing would function as a
type of exemplar training in which the learner already knows what to do but needs lots
of practice with different examples to enable him or her to focus on the key relational
aspects of the task. A primary difficulty, however, emerges when a teacher cannot know
whether the necessary relational competence is present and whether exemplar training of
this sort alone will be enough to establish it if it is absent. From an RFT perspective, the
answer is hit-or-miss.
Chapter 12 277

Traditional exemplar training may be enough, but it is very possible that it will not.
In this case, under the theory, a more functional and process-oriented training regime
will be necessary in order to adequately train the composite relational skills in a focused
and systematic fashion; protocol 1 would be the necessary tool.
Conclusions
Despite the centrality of analogical reasoning within higher cognition, there are limited
training protocols to help establish the necessary skills. The current chapter contains two
such tools—one of which is derived from a more traditional and cognitively oriented
approach that appears to function as a sort of exemplar training for enhancing analogical
skills. For less verbally sophisticated learners, however, such a training regime may have
limited utility, because the basic ability to derive analogical or equivalence-equivalence
relations is deficient. In this case, an RFT-based protocol that targets equivalence-
equivalence responding and related composite skills may be necessary in order to estab-
lish the basis of analogical abilities. Thereafter, the traditional approach may be used for
skill enhancement. Although the overlap between these two protocols’ targeting of the
same basic processes is an empirical issue, there is considerable reason to believe that
equivalence-equivalence responding is a core feature of analogy and that the training
regime presented here is a useful means of establishing the former.
References
Barnes, D., Hegarty, N., & Smeets, P. (1997). Relating equivalence relations to equiva-
lence relations: A relational framing model of complex human functioning. Analysis
of Verbal Behavior, 14, 57–83.
Carpentier, F., Smeets, P., & Barnes-Holmes, D. (2002). Matching functionally same
relations: Implications for equivalence-equivalence as a model for analogical reason-
ing. Psychological Record, 52, 351–370.
Carpentier, F., Smeets, P., & Barnes-Holmes, D. (2003). Equivalence-equivalence as a
model of analogy: Further analyses. Psychological Record, 53, 349–371.
Carpentier, F., Smeets, P., Barnes-Holmes, D., & Stewart, I. (2004). Matching derived
functionally same stimulus relations: Equivalence-equivalence and classical analogies.
Goswami, U., & Brown, A. (1989). Melting chocolate and melting snowmen: Analogical
reasoning and causal relations. Cognition, 35, 69–95.
Goswami, U., & Brown, A. (1990). Higher-order structure and relational reasoning:
Contrasting analogical and thematic relations. Cognition, 36, 207–226.
Lipkens, G., Hayes, S. C., & Hayes, L. (1993). Longitudinal study of derived stimulus
relations in an infant. Journal of Experimental Child Psychology, 56, 201–239.

Piaget, J., Montangero, J., & Billeter, J. (1977). Les correlates [Correlations]. In J. Piaget
(Ed.), L’abstraction refléchissante. Paris: Presses Universitaires de France.
Sternberg, R. (1985). Beyond IQ: A triarchic theory of human intelligence. New York:
Cambridge University Press.
Stewart, I., Barnes-Holmes, D., Roche, B., & Smeets, P. (2001). Generating derived rela-
tional networks via the abstraction of common physical properties: A possible model
of analogical reasoning. Psychological Record, 51, 381–408.
Chapter 12 279

Chapter 13
Understanding and Training Perspective

Taking as Relational Responding
Louise McHugh, University of Wales, Swansea;

and Yvonne Barnes-Holmes and Dermot Barnes-Holmes,
National University of Ireland, Maynooth
The development of perspective taking has attracted interest from educators and psy-
chologists not least because of its role in developmental disabilities such as autism spec-
trum disorder. Deficits in these abilities (often referred to as “mindblindness”) have been
interpreted by cognitive developmental psychologists in particular as evidence of a more
generic set of skills called “theory of mind” (ToM; Baron-Cohen, 1995). The basic pro-
cesses that underpin ToM capabilities are believed to involve an understanding of the
relationship between one’s own beliefs and actions and the beliefs and actions of others.
These relationships form the focus of ToM and related remediation programs for the
developmentally disabled (Howlin, Baron-Cohen, & Hadwin, 1999).
In contrast, behavioral psychologists have argued for a more functional approach to
understanding and remediating deficits in perspective taking (McHugh, Barnes-Holmes,
O’Hora, & Barnes-Holmes, 2004). Although this alternative approach resembles tradi-
tional applied behavior analysis in some regards, it also incorporates recent research on
derived stimulus relations. According to this view, perspective taking comprises a complex
set of derived relational abilities that are based on our understanding of self, place, and
time. While remediation packages of this type are delivered using traditional behavioral
principles, the focus of the content is primarily relational in nature. In a sense, therefore,
the behavioral approach to perspective taking has become more cognitive in orientation;
thus, the two traditionally opposing views of the subject are more homogeneous than
ever.
The current chapter will review the cognitive and behavioral approaches to perspec-
tive taking. The primary focus of the work, however, will be to articulate the manner in
which perspective taking can be understood and trained as a core set of derived relational
abilities. To this end, the chapter will contain illustrations of how the target stimulus
relations for perspective taking may be tested and trained for the education of young
children, including those with developmental disability.
Understanding Other Minds

Theory of mind is the articulation by cognitive developmental psychologists that takes full
appreciation of the representational nature of the mind and its interaction with behavior
(Suddendorf & Fletcher-Flinn, 2002). Perspective taking forms the core component of
such a representational understanding of our own minds and the minds of others, and
it must be present for the development of subsequent and more complex skills, includ-
ing the understanding of false belief and deception. Indeed, researchers in this area have
argued that, although great apes and very young children appear to have some under-
standing of their own and others’ emotional states (such as desire or hurt), they cannot
attribute informational states (such as knowledge and belief) that provide the context for
these emotions (Whiten & Byrne, 1997).
According to this account, there are five basic stages in the development of perspec-
tive taking, ranging from simple perspective taking to distinguishing between true and
false beliefs (Howlin et al., 1999). At its simplest, the prescribed training sequence aims
to develop perspective-taking skills from physically based beginnings to psychological
events that are dependent on subtle social and verbal cues. For example, the abilities that
make up simple and complex visual perspective taking in levels 1 and 2 are based on the
principle that different people can see the same thing differently, depending on specific
features of the physical environment.
Subsequent levels of perspective taking are believed to be more complex because they
require a shift from purely visual perspective taking to understanding the relationship
between seeing and knowing. That is, in seeing and knowing in level 3, and understand-
ing true belief in level 4, respectively, children learn to appreciate that they and others can
only determine a true belief on the basis of what is actually seen, and that they cannot
know what they have not seen, even when they think they know. Again, however, cues
based on the physical environment dominate the perspective taking, although the child
may be required to remember the layout of a previous environment that differs from one
presented currently.
Level 5 of ToM continues the focus on distinguishing between what is actually seen
and what is believed and, thus, begins the distinction between true and false beliefs. For
example, a child might expect that a candy box contains candies, rather than pencils. In
this case, the belief that there are candies in the box is false, and the true belief that the
box contains pencils only emerges when the child is shown what is inside.
The theoretical and empirical interest in ToM has been considerable, and there is
growing empirical evidence of the applicability of the concepts in remediation packages
for populations where these skills appear deficient (Fisher & Happé, 2005). However,
from a behavioral perspective, questions remain over the basic and functional processes

that make up perspective taking and the most focused means of establishing these. In
the sections that follow we review the modern behavioral approach to perspective taking
and outline in detail a systematic training protocol for establishing the target relational
abilities.
Perspective Taking as Relational Abilities

Behavioral psychologists traditionally showed relatively little interest in perspective taking
and, not surprisingly, did not consider the cognitive aspects of these abilities. The rapid
growth of interest in a modern behavioral account of human language and cognition
known as relational frame theory (RFT) has changed this; as a result, behavioral psychol-
ogy now has a fresh account of perspective taking in view of derived stimulus relations
(Hayes, Barnes-Holmes, & Roche, 2001). In order to explain perspective taking as a set of
relational skills, we must first briefly review the concept of derived stimulus relations.

According to relational frame theory, relating one stimulus or event to another via
their physical properties is a behavioral pattern that characterizes much of nonhuman
conduct and that of very young infants. As humans develop, however, these behaviors
give rise to more complex relational skills that do not depend upon the physical properties
of the stimuli, and it is these latter abilities that are believed to form the core of higher
cognition. Indeed, for RFT, these more abstract relational abilities are synonymous with
language itself and are pivotal in the development of crucial skills, including perspective
taking. In technical terms, RFT refers to these abilities as arbitrarily applicable relational
responding, but they may be loosely defined as the ability to relate events in the world
that have no discernible physical connection. For example, the word “window” bears no
formal or real similarity to an actual window, as is the case with practically all of the
nouns in the English language.
The coordination between words and their referents is one of the earliest examples
of arbitrarily applicable relational responding and accounts for the incidental learning of
thousands of name-object and object-name relations in early childhood. In many of these
interactions, the symmetrical relation between an object’s name and the actual object is
directly trained (in other words, the teacher provides the object name and gives child the
object). This is referred to as a name-object relation of coordination, because the two events
are trained as meaning one and the same thing. On many occasions, the reverse relation
between objects and names (object-name relations) are trained (in other words, the teacher
provides the object and then provides the appropriate name). Again, this is a relation of
coordination between the two events, even though the training in this case occurred in
the opposite direction from that used before. However, both kinds of training occur in
specific contexts with cues that include phrases like “Look at the…” and “What’s this?”
both of which enable the child to readily discriminate that the relations in question are
ones of coordination, irrespective of the actual direction in which the training occurred.
Chapter 13 283

According to RFT, adequate training in name-object and object-name relations gives
rise to generalized response patterns in which training in only one direction will facili-
tate the learning of the opposite direction without explicit training. For example, explicit
training in a new name-object relation may result in a derived object-name relation that
was never explicitly trained (for example, the teacher names the object and hands the
object to the child, then the child is asked “What’s this?” and provides the name). Indeed,
this type of emergent behavior is at the core of RFT and is believed to apply to all types of
stimulus relations, including those that are much more complex than coordination (such
as relations of distinction, opposition, hierarchy, and perspective taking).
Relational Perspective Taking

For RFT, responding in accordance with perspective-taking relations shares the same
qualities of arbitrariness and generalization of other relations, but it is even more complex
because both the physical environment and the person having the perspective are chang-
ing constantly. Although RFT employs the technical term “deictic relations” to describe
the relational abilities involved in perspective taking, the simpler term “perspective-taking
relations” is preferred here.
For RFT, perspective taking involves three key types of relations: I versus you, here
versus there, and now versus then. To categorize these simply, it is possible to think of
I versus you and here versus there as spatial relations, and now versus then as temporal
relations. Furthermore, it is generally believed that the former are simpler than the latter,
and that the most effective sequence of training should follow the same pattern (I/you
first, followed by here/there, and finally now/then).
The easiest way to understand the relational responding at the core of these skills is
to try to determine the target relations involved. For example, if you ask a young child
“Who are you?” the word “you” is often coordinated with the child’s name. Similarly if
you ask “Where are you?” “you” will be coordinated with the actual location at the time
of responding. Of course, children are also learning across these interactions that “you”
refers to themselves (in other words, it is coordinated with the listener) and is in a rela-
tion of distinction with “I,” which is coordinated with the speaker. Hence, in a sense, I
versus you is the easiest perspective-taking relation to learn because of the constancy of
the coordination between “I” and the speaker and between “you” and the listener. These
relations remain the same irrespective of who is speaking and who is listening.
It is difficult to separate the three types of perspective-taking relations from one
another, because they are often learned in combination. For example, in learning to
respond to the question “Where are you?” children are learning not only the coordination
between “you” and themselves but also the coordination between themselves and their
present location. Similarly, if a child is asked, “What are you doing?” the coordination
between “you” and the self as well as between the self and current actions is also being
learned. Hence, across time, the specific cues provided by the question (where or what)
indicate the type of answer necessary and which relational response is most appropriate.
Furthermore, coordination emerges between I, here, and now and you, there, and then,
because “I” is always from this perspective, here, and now, but not from the perspective
of another person, there, and then. As a result, the appropriate relational response can

emerge even when the specific words are not present. For example, you need not ask
“Where are you now?” but simply “Where are you?”
Although in common discourse the perspective-taking relations are frequently interre-
lated, RFT makes functional distinctions that are necessary for establishing the relations
in remediation programs. Specifically, there are two key distinctions based on relation
type (I-you, here-there, or now-then) and level of relational complexity that comprise
simple, reversed, or double reversed relations. This combination generates a range of rela-
tional permutations, illustrated in table 13.1. In the section that follows, we describe the
key relational skills that make up perspective taking and emphasize what appears to be
the most effective means of training them.
Table 13.1. The Eighteen Core Relational

Skills Involved in Perspective Taking
Relation Level of Relational Complexity

Type
Simple Relations Reversed Relations Double Reversed
Relations
I-You Simple I-you Reversed I-you
Here-There Simple I-you Reversed I-you Double reversed
and and I-you and here-there
simple here-there simple here-there
Simple I-you
and
reversed here-there
Now-Then I I
and simple now-then and reversed now-then
You You
and simple now-then and reversed now-then
I I I
and simple here-there and reversed here-there and double reverse
and simple now-then and simple now-then here-there and
I now-then
and simple here-there
and reversed now-then
You You You
and simple here-there and reversed here-there and double reversed
and simple now-then and simple now-then here-there and
You now-then
and simple here-there
and reversed now-then
Chapter 13 285

Training I-You Relations
For RFT, the first step toward establishing perspective taking involves training simple
I-you relations (see table 13.1). At this early stage of training, it is best to focus only on
I-you and omit references to locations or times for the present. Although this type of
interaction is not common, it is possible to present I-you relations without here-there or
now-then relations, but as you will see later, the reverse is not the case. In training, begin
with a simple scenario in which, for example, the teacher has a red brick and the child
has a green brick, and ask the child, “Which brick do I have?” and “Which brick do you
have?”
Training Notes: Getting Started
1. From the outset in training relational perspective taking, you have to be very
careful about which perspective you are speaking from. When a teacher says,
for example, “I” in a question, this is always a reference to the speaker (in other
words, the teacher) and when he or she says “you,” this is always a reference to
the child.
2. Although normal discourse does not include the words “teacher” or “child” and
thus does not explicitly coordinate these with “I” and “you,” respectively, it
may be very helpful for young or language-deficient children that both words
are stated directly. In this case, therefore, the teacher might say the follow-
ing: “Which brick do I, the teacher, have?” and “Which brick do you, [child’s
name], have?”
3. Indeed, further assistance might also be provided by physically pointing to the

appropriate individual referred to in each question, or by pointing to the appro-
priate item.
4. Although these types of prompts will likely ease the learning in the early stages,
all skills must eventually be mastered without prompts.
Simple I-you relations are described as such because the relations are straightforward—
the word “I” is coordinated with the speaker, and “you” is coordinated with the listener.
However, there is also coordination between the brick in front of the teacher or the child
and the bricks referred to in the questions. Hence, there is always more than one coor-
dination relation being learned at any one time. Indeed, that’s why these tasks are, espe-
cially early on, difficult for children to learn. Put simply, there is a lot more going on than
you might realize, and prior to even the first of these tasks, the child must have learned
to alter the perspectives of listener and speaker for conversational purposes.
At this stage, it is also useful to think of these relations another way that will help
you to understand the reversal of relations later. The coordination between I (speaker) and
you (listener) and between the bricks can also be viewed as if-then functions that imply
that “if I am I and you are you, then I have what I have in front of me and you have what
you have in front of you.”

Training Notes: Successful Training
1. Explicit training, especially at this early stage, commonly involves a series of

six trials, each with an I question and a you question. Hence, both questions
must be answered correctly for the trial to be recorded as accurate. Feedback or
reinforcement normally follows the second question of each pair, at which point
any errors are corrected.
2. Across trials, randomize the sequencing of the questions.
3. If explicit training has occurred at any point during a block of trials, the child
must always complete an entire block without error in order to be deemed to
have learned the necessary skills.
4. Once the child responds correctly to all six trials using a familiar set, a novel set
of items is presented to ensure generalization. In the event that errors occur on
the novel block, the appropriate response would be explicitly trained and tested,
and a further novel set would be presented.
5. It is important to remember that neither the teacher nor the child should change
physical location during the presentation of a trial. However, if a child is having
difficulty responding to feedback, it may be beneficial to bring the child over to
the teacher’s side of the table, for example, in order to physically see the brick
from that location. Again, this may be helpful early on, but must eventually be
removed on all test trials.
6. In all cases, a child is deemed to have completed the task when perfect respond-
ing occurs on a block of six trials involving a previously untrained stimulus set
and no feedback or prompting of any kind is given.
7. Successful training and testing of this first relational task must be complete
before the training of a subsequent task is commenced. In order for you to get
an overview of the overall sequence of training the eighteen tasks, the full train-
ing sequence and additional important features are provided in table 13.2.
Chapter 13 287

Table 13.2. An Overview of the
Perspective-Taking Training Sequence
Block Target Relations No. of Important
Number Trials Features
I-You Relations
1 Simple I-you 6 Questions
randomized
2 Reversed I-you 6 Questions
randomized
3 1 and 2 combined 6 3 of each, all
randomized
4 Same as 3 6 Items switching
locations
5 Same as 4 6 Novel stimulus set
Here-There Relations
6 Simple I-you and simple here-there 6 Questions
randomized
7 Simple I-you and simple here-there 6 3 of each, all
Reversed I-you and simple here-there randomized
8 Same as 7 Items switching
locations
9 Simple I-you and simple here-there 6 3 of each, all
Simple I-you and reversed here-there randomized
locations
11 Reversed I-you and simple here-there 6 3 of each, all
locations
13 Reversed I-you and simple here-there 12 4 of each, all
Double reversed I-you and here-there
locations

Now-Then Relations
16 I and simple now-then 6 Questions
randomized
17 You and simple now-then 6 Questions
randomized
18 I and simple now-then 6 3 of each, all
I and reversed now-then randomized
19 You and simple now-then 6 3 of each, all
You and reversed now-then randomized
20 18 and 19 combined 12 3 of each, all
randomized
locations
22 I and simple here-there and simple now-then 12 6 of each, all
You and simple here-there and simple randomized
now-then
locations
24 I and simple here-there and simple now- 12 3 of each, all
then randomized
You and simple here-there and simple
now-then
I and reversed here-there and simple now-
then
You and reversed here-there and simple
now-then
locations
26 I and reversed here-there and simple now- 12 3 of each, all
then randomized
now-then
I and simple here-there and reversed now-
then
You and simple here-there and reversed
now-then
locations
Chapter 13 289

Now-Then Relations
28 I and reversed here-there and simple now- 12 2 of each, all
then randomized
now-then
I and simple here-there and reversed now-
then
You and simple here-there and reversed
now-then
I and double reversed here-there and
now-then
You and double reversed here-there and
now-then
locations
I-you relations can also be made more complex by reversing the relations. As you
would expect, reversed I-you relations are the simplest type of reversed relation, again
because only one relation type is involved. Consider the following task: “If I have a pen
and you have a cup, and if I were you and you were me, what would you have? What
would I have?” The reversing of the I-you relation here alters the if-then functions in
a manner that differs from the I-you relations in simple form. In previous tasks, for
example, the if-then relation was always based on the coordination between “I” and the
speaker and between “you” and the listener, but when the relation is reversed (with the
inclusion of a statement such as “If I were you and you were me”) the functions of the
if-then relations are transformed. In this case, the functions of I are now coordinated with
you and the functions of you are coordinated with I (this is what is meant by the relation
being reversed). Of course, a correct answer here requires that the pen be transferred,
hypothetically and not actually, from I to you and that the cup be transferred from you
to I. Once again, the training and accuracy criteria would be the same as before.
Training Notes: Mixing and Matching
1. Once simple and reversed I-you relations are established, it is wise to take a
sample of trials from each and present them within a mixed randomized block
(see block 3, table 13.2).
2. We usually do this using the same stimuli from the previously novel sets. Once
criterion is reached here, we then begin to switch the items over from person to
person across trials (block 4), and then finally employ a completely novel set (block
5). In this way, we can ensure that there is no spurious source of control over the

child’s responding and that the component relational responses are present. The
use of the novel sets also ensures that the relational responses are derived.
3. Again, of course, children who experience difficulty in learning the reversals
may be assisted by physically swapping items or locations, but this would have
to be removed on test trials.
Training Here-There Relations

Here-there relations inherently contain I-you relations, thus explaining why the latter
must be established first. The simplest form of a here-there relation is referred to as a
simple I-you and simple here-there relation. Consider the following task: “I am standing
at the yellow door, and you are standing at the brown door. Where are you standing?
Where am I standing?” A correct response here requires the child to respond in accor-
dance with a combination of I-you and here-there relations, indicating that I must be
here and you must be there. Again, the if-then relations assume that if I am I and you
are you and here is here and there is there, then I will be here and you will be there—
coordination relations that were not tested in the presentation of I-you alone.
Of course, when two relation types are combined, either one can be reversed, while
the other is kept simple. For example, in reversed I-you and simple here-there relations,
the I-you relation is reversed as before, while the here-there relation remains simple.
Consider the following trial: “I am standing at the yellow door, and you are standing at
the brown door. If I were you and you were me, where would you be standing? Where
would I be standing?”
Training Notes: Adding Complexity

1. You will notice from block 7 (table 13.2) that the first exposure to reversed
I-you and simple here-there relations also contains some trials of simple I-you
and simple here-there relations. This is not essential, and the new trials can be
presented in a separate block. However, it is wise to combine the simple and
reversed tasks as early as possible, because this serves to highlight the relational
distinctions between them. Incidentally, it also ensures that the child must
listen very carefully on each trial.
2. This raises the more generic issue about the degree to which the tasks pre-
sented here can be further subdivided in order to smooth out the learning curve
and ease or preclude difficulties in learning the tasks. Of course, the teacher or
trainer will always know what is best for an individual learner, but the way we
have distinguished the tasks here and the sequence in which they are trained
serve as a reference guide to how the tasks should be systematically ordered and
what needs most emphasis.
Reversed here-there relations (such as simple I-you and reversed here-there relations)
are almost identical to I-you reversals and transform the if-then relations in an almost
identical manner. This is because of the coordination between the two frames, meaning
that “I” remains coordinated with here, and “you” remains coordinated with there. So,
when here moves there, “I” goes too and when there moves here, “you” goes too. So,
functionally, the changing of the relations is identical to an I-you reversal.
Chapter 13 291

Training Double Reversed Relations
For RFT, the greatest level of relational complexity involved in perspective taking
is made up of relations in which two frames are reversed simultaneously—referred to as
double reversed relations or double reversals. The first type of double reversal to be estab-
lished is naturally the double reversed I-you and here-there relation, because only these
two relation types have been established so far. Hence, it is essential that these complex
relations are trained only after all of the simple and reversed components of each relation
have been clearly established and generalized.
In double reversed I-you/here-there relations, both of the target frames are reversed
simultaneously. Consider the following task: “I am standing here at the yellow door and
you are standing there at the brown door. If I were you and you were me, and if here were
there and there were here, where would you be standing? Where would I be standing?” In
this task, the joint reversal of the I-you and the here-there relations increases the complex-
ity of the if-then relations, so that the child must respond in accordance with “if I were
you and if you were me” and “if here were there and if there were here.” This transfers the
yellow door, in this case, from I to you and the brown door from you to me (placing you
hypothetically at the yellow door and me at the brown door) and then transfers the brown
door from here to there and the yellow door from there to here. So, because the double
reversal negates each individual reversal, both you and I end up back where we started.
The same task might be conceptualized as the functions of I-here and you-there being
reversed in the first reversal and then being reversed again in the second reversal. The
reason for this is that “I” is always from the perspective here, and “you” is always from
the perspective there, so that a change in any one part of each relation entails a change
in the other part. That is, as soon as “I” changes, here also changes, and the same can
be said for “you” and there. Consider, for example, after the first I-you reversal, when I
am hypothetically at the brown door and you are at the yellow door. If the coordination
between I and here and between you and there had not been retained, then I would be at
the brown door, but here would still be at the yellow door. Similarly, you would be at the
yellow door but there would still be at the brown door. But this simply does not happen,
because I remains coordinated with here, and you remains coordinated with there, so
that I-here and you-there change together on the first reversal and change together again
on the second reversal. As a result, the task is a good test of perspective taking, because
it implicitly requires that “I” remains coordinated with here, as does “you” with there—
both of which must be the case for one’s perspective to remain stable.
Training Notes: Getting Really Complex
1. By block 13, when the first double reversals emerge, you will notice that the
blocks become longer, with twelve (rather than six) trials. This increase results
from the fact that there are now three critical component tasks, and fewer than
four exposures to each would simply be inadequate.
2. You can add more exposures to each trial type if you think the learner needs it,
but four is the minimum number of each.

Training Now-Then Relations
Like here-there relations, now-then relations must always be combined with I-you
relations. However, temporal relations appear to be unique in that they cannot be com-
bined with both elements of I and you at the same time. Indeed, if one attempts to
combine I-you with now-then, the result is a statement in which not all of the relations
can be specified. For example, imagine if you were told, “Yesterday I was reading; today
you are watching television.” If you were then asked, “What was I doing then? What are
you doing now?” you would be able to answer. But if you were asked, “What am I doing
now?” and “What were you doing then?” you could not answer. This is because in the
former case the relations targeted in the instruction (I-then and you-now) are also tar-
geted in the question. However, in the second set of questions this is not the case; here,
the questions contain I-now and you-then relations that were not in the statement and
cannot be derived from it (hence they are unspecified). In order to present simple now-
then relations, either I or you must be targeted alone, not in combination. Hence, the two
types of now-then tasks are distinguished on this basis.
One of the simplest types of now-then relation involves I (I and simple now-then rela-
tions). Consider the following instruction: “Yesterday I was reading; today I am watch-
ing television. What was I doing then? What am I doing now?” A correct response here
actually requires that the child respond in accordance with you and now-then relations
because of the roles of speaker and listener. This task also requires that for the child now
is coordinated with today and then is coordinated with yesterday. Tasks involving you
relations can be constructed in the same way.
Training Notes: Further Complex Still
1. There is no particular sequence for training I-you and simple now-then relations;
they would appear to be of identical complexity. However, we advise that, while
these relations are being established, you keep I and you in separate blocks until
the component skills are in place (see table 13.2).
2. In developing the training protocol, we used the words “yesterday” and “today,”
although these are not critical. Possible alternatives include using different
times, for example. Ensure, however, that the two points in time are adequately
distinct and that the learner clearly understands the meaning of the terms you
use. We advise this because all aspects of temporal relations are abstract and
complex—because time itself is an abstract concept.
When simple now-then relations are established, reversed now-then relations can be
presented. However, recall that now-then can be combined with either I or you, but
not with both. Hence, the two key skills here are I and reversed now-then and you and
reversed now-then relations. ��
Consider the following task: “Yesterday I was watching tel-
evision; today I am reading. If now was then and then was now, what would I be doing
then? What would I be doing now?” In this case, the now-then reversal transforms the
if-then functions, so that��
now is then and then is now. In this case, the television watch-
ing is transferred from now (today) to then (yesterday), and the reading is transferred
from then to now. Tasks involving you relations are constructed in the same way.
Chapter 13 293

The temporal tasks can also be expanded with the inclusion of here-there relations,
which, at their simplest, comprise simple here-there and simple now-then relations involv-
ing I or you. Consider the following instruction given to a child: “Today I am standing
at the yellow door. Yesterday I was standing at the brown door. Where am I standing
now? Where was I standing then?” Although the level of relational complexity is simple
(because none of the relations are as yet reversed), the combination of all three relation
types would render the task more difficult than it is with only two. So, a correct response
here requires the child to respond in accordance with coordination between here and
now, and between there and then. Again, “you” is coordinated with the listener, here with
the speaker’s location, there with the listener’s location, now with today, and then with
yesterday. Once again, because of the perspective shift between speaker and listener, the
child here is actually coordinating you-here-now and you-there-then (rather than I). You
and simple here-there and simple now-then relations can be constructed in the same way,
but one must remember that the child is actually coordinating I-here-now and I-there-
then (rather than you).
As expected, the combination of the three types of relation can also contain a rever-
sal. In this case, the task can involve either a reversed here-there or a reversed now-then
relation (in other words, I-you and reversed here-there and simple now-then relations or
I-you and simple here-there and reversed now-then relations). Consider the following task:
“Today I am standing at the yellow door. Yesterday I was standing at the brown door. If
here were there and there were here, where would I be standing now? Where would I be
standing then?” A correct response here requires that the functions of the yellow door be
transferred in accordance with the coordination among the three frames (you-there-then
and you-here-now).
Of course, the three relation types can be combined in such a way that both I-you
and here-there remain simple but now-then is reversed (in other words, I-you and simple
here-there and reversed now-then relations). In both cases, the outcome is almost identical
to I-you with reversed here-there and simple now-then relations, because of the coordina-
tion among the relations (for example, I-here-now and I-there-then). ��
Consider the follow-
ing task: “Yesterday I was standing at the yellow door. Today I am standing at the brown
door. If now were then and then were now, where would I be standing then? Where
would I be standing now?”
Because I-you must be separated into I or you when combined with now-then, it is
nonsensical to have I-you/now-then double reversals. But double reversals involving here-
there and now-then relations (involving either I or you) can be constructed. Consider the
following task: “Today I am standing at the yellow door. Yesterday I was standing at the
brown door. If here were there and there were here, and if now were then and then were
now, where would I be standing now? Where would I be standing then?” In this task,
the joint reversal of the here-there and now-then relations again increases the complexity
of the if-then relations, so that the child must respond in accordance with the idea that
here was there and there was here and now was then and then was now. This transfers
the yellow door from here to there and the brown door from there to here (placing me
hypothetically at the brown door now and at the yellow door then) and then transfers the
brown door from now to then and the yellow door from then to now. Hence, a correct
response would place me back at the yellow door here and now and at the brown door
there and then. Once again, then, I end up back where I started. Double reversed here-
there and now-then relations involving you can be constructed in an identical fashion.

Training Notes: Final Training Issues
1. The training protocol for relational perspective taking has been constructed in
a systematic fashion that comprises step-by-step increases in task complexity.
Although the sequence presented is one that we have repeatedly found to be
effective, the length of time children take to complete it or learn the individual
component skills, or the level of prompting or task breakdown necessary for an
individual child, will be highly variable.
2. What is important to emphasize, however, is that progression from one task to

another should not occur without appropriate generalization tests or where there
is any doubt concerning prerequisite skills or competence in preceding tasks.
3. We also know that these are difficult tasks to learn (we have experienced many
adults balking at double reversals at academic conferences), so the learning curve
may be slow and painful for all parties. Despite this difficulty, however, we do
believe that critical relational skills are being targeted.
Factors Affecting the Learning of Relational

Perspective Taking
In the section below, we summarize features of the perspective-taking tasks that may
be relevant to their training and which have been the subject of a number of empirical
investigations. These features include the nature of task presentation, age-based expecta-
tions of competence, training with developmentally delayed populations, and the rela-
tionship between the perspective-taking tasks and emotion. This research has served an
important purpose in determining the validity of the concepts on which the perspective-
taking protocol was based (including relation types and level of complexity) and in high-
lighting the extent to which the suggested training was appropriate to the needs of those
for whom it was developed.
Task Presentation
In the first published empirical article examining relational perspective taking, we
exposed normally developing adult participants to all of the tasks in the test protocol
outlined above (McHugh, Barnes-Holmes, and Barnes-Holmes, 2004). The primary aim
in this research was to determine whether specific ways of presenting the tasks to par-
ticipants would influence their performances. Hence, we divided our sample across four
conditions. In one condition the experimenter read all tasks aloud, participants responded
verbally, and the experimenter recorded each response. In this condition, actual items
(such as bricks, chairs, and pictures) were also presented in conjunction with each task (as
outlined above). We had originally developed the training protocol in this way in order
to reduce the memory demands of the tasks when they were presented to children. In
another condition, the arrangements were the same but the actual items were removed
(making the task completely abstract). In a third condition, the tasks were presented to
participants in written form, and they recorded their own responses in written form; no
Chapter 13 295

items were employed. In the final condition, the protocol was reduced in size, with eigh-
teen types of task continuing to be presented, but with fewer exposures to each.
The results of the study indicated primarily that participants performed best on the
I-you relations, although the here-there and now-then performances were not differenti-
ated. They also performed best on the simple relations, doing better on reversed than on
double reversed relations. These outcomes provided sound initial support for the concep-
tual and practical distinctions we had made among the tasks. The impact of whether the
experimenter read the tasks aloud or whether participants read the tasks to themselves
was negligible, as was the impact of using real objects, and there was no difference in
outcomes between the full-length and shortened protocols. While these latter findings
suggested that the nature of task delivery, number of exposures to each task, and use of
actual objects had little impact on the test performances of adults, there is no reason yet
to believe that this would also be the case in a training context with young children. In
fact, it would seem advisable that the children be provided with as much help as pos-
sible in mastering the tasks (for example, by providing many exposures to each trial type
and using real items). Nonetheless, the data do suggest that it may be useful to present
children who have successfully mastered the protocol with a subsequent test in which the
objects are removed and the protocol is shortened. Indeed, in additional studies, we have
also examined the impact of randomizing all of the tasks within the protocol (rather than
using the sequential format that is appropriate to training), and a competent learner should
also be successful in this context (McHugh, Barnes-Holmes, O’Hora, et al., 2004).
Age-Based Expectations of Competence

The key aim of the second published study on relational perspective taking was to
determine the relationship between competence on the target tasks and age. The purpose
was to determine the approximate age at which we might expect children to normally
master the relevant skills. McHugh, Barnes-Holmes, O’Hora, and colleagues (2004)
studied participants from five age ranges—three to five years (early childhood); six to
eight years (middle childhood); nine to eleven years (late childhood); twelve to four-
teen years (adolescence); and eighteen to thirty years (adulthood)—who were exposed
to the full-length sequential test protocol described previously. As expected, the results
of the study indicated a developmental trend in which performances improved with age.
Specifically, the adult participants produced the best and very strong performances overall,
while participants from the early childhood group produced the worst. On over half of
the trial types, the adolescents performed better than those in late and middle childhood,
and there was some superiority in the performances of those in late childhood relative to
those in middle childhood, particularly on trials involving double reversals. Again, per-
formances in relation type and relational complexity were as expected. In summary, we
would expect adults and adolescents to pass all trials with relative ease; increasing difficul-
ties will be observed in late childhood (nine to eleven years), with greater errors in middle
childhood (six to eight years). Hence, substantive training would likely be necessary with
normally developing children age eight or under. However, as the subsequent research
(described below) attests, it may be beneficial to commence training at a much earlier age
than this, particularly if there is reason to believe that perspective-taking difficulties will
be forthcoming.

In research published in 2001, Barnes-Holmes reported the training of two normally
developing young children on the I-you and here-there parts of the perspective-taking
protocol (now-then relations were not included). The first participant was a seven-year-old
female who immediately demonstrated competence (without training) on the simple rela-
tions; she required some training on the reversed relations, and the most training on the
double reversals. However, the child was exposed to only two sessions of explicit training
overall and subsequent and extensive generalization tests were passed with ease.
The second participant was a three-and-a-half-year-old male. This younger child
required little or no training on the simple relations but required extensive training of the
reversals and double reversals. Indeed, the child required a total of thirty-two training ses-
sions to demonstrate mastery of all relations including double reversals. Nonetheless, even
at this young age, the child subsequently passed extensive generalization tests without
further training. What is perhaps most interesting about these empirical findings is that
they demonstrate the accessibility of the training protocol for use with young children
and the high level of competence that can be established even at a young age.
In related research reported by McHugh (2004), two additional normally develop-
ing children were trained on the same relational tasks, but in this case the protocol also
included all aspects of now-then responding. The first participant in this research was a
four-year-old female who required no initial training on any type of simple relation. The
training required for the three types of reversed relations was greater, but it still amounted
to no more than a total of seven training sessions. As expected, the double reversals posed
greater difficulty, but explicit training of these relations was completed in five sessions.
Once again, generalization tests were passed with perfect responding and no additional
training.
The second participant in this research was a young boy who was almost four at the
beginning of the study. Interestingly, this child was immediately competent on simple
I-you and here-there relations but showed immense difficulty on simple now-then rela-
tions. Indeed, no fewer than sixty-four training sessions were necessary to establish the
target relations. It is important to note, however, that in this particular research, simple
now-then relations were trained immediately after simple I-you and simple here-there rela-
tions, hence the findings may indicate that it is better to train the relations in a sequence
based on relation type than on relational complexity. In other words, the difficulty for
this child on simple now-then relations may not have been as great if all aspects of I-you
and here-there responding had been established previously. Subsequently, training on all
three types of reversed relations was straightforward (requiring a total of seven sessions)
and double reversals required a total of seven sessions of training. Taken together, then,
the initial difficulty in learning the simple now-then relations compared to the subsequent
ease of establishing the more complex relations would suggest that now-then relations
should be established last.
Developmental Disability
There are only two existing studies involving the use of the relational perspective-tak-
ing protocol with children who are not normally developing but who had been diagnosed
as high-functioning autism spectrum disorder or Asperger’s disorder. However, all of the
Chapter 13 297

children here were presented with the protocol as a test or assessment measure, rather than
as a training tool. In the first study, by Rehfeldt, Dillen, Ziomek, and Kowalchuck (2007),
the children with Asperger’s disorder (aged six to thirteen years) were compared with a
normally developing control group and correlations were assessed using the perspective-
taking protocol, the Vineland Adaptive Behavior Scales, and the Social Communication
Questionnaire. Consistent with previous research, errors on reversed relations were sig-
nificantly greater than they were on simple relations, particularly with the children with
Asperger’s disorder. Double reversals also generated more difficulty but were in fact easier
than the reversals for both groups.
In the second relevant piece of research, McGuinness (2005) assessed correlations
between the perspective-taking protocol and the Asperger’s Syndrome Diagnostic Scale
(Myles, Bock, & Simpson, 2001); the Cognitive Abilities Test (Lohman, Hagen, &
Thorndike, 2001); and a traditional ToM assessment. All six participants with Asperger’s
disorder (aged eight to eleven years) showed immediate competence on the three types
of simple relations. Their competence on reversed relations ranged between 56 and 97
percent, and the range on double reversals was 50 to 89 percent. In general, participants
with the lowest performances on the Cognitive Abilities Test produced the most errors
on the relational protocol, but surprisingly there appeared to be no relationship between
performances on the protocol and the ToM tests. Furthermore, there was considerable
discrepancy for participants across ToM tests (some participants passed one test but failed
another), yet there was much overlap with the relational protocol (those who did poorly
on reversals also did poorly on double reversals).
Taken together, these findings did not highlight large differences between the chil-
dren with Asperger’s disorder and normally developing controls, with the exception of
those who produced weaker scores on the Cognitive Abilities Test. However, it may be
the case that one of the critical difficulties with Asperger’s disorder is not that there are
deficiencies in the core relational skills but that there may be weaknesses in the coordina-
tion between these relations and emotions. Although the research reviewed in the follow-
ing section is preliminary, it may shed light on the important relationship between these
two critical aspects of perspective taking.
Emotions
It is clear in all accounts that emotions form a core part of perspective taking that prob-
ably builds upon the basic cognitive skills. For RFT, the relationship between emotion
and perspective taking would be described as the transformation of emotional functions
through the perspective-taking relations. Hence, from this view, competence in the basic
relational responses would need to be established first and this would likely render the
“loading in” or transforming of emotional functions relatively easy.
Only one early and preliminary RFT study has investigated this issue. In part of the
research conducted by Barnes-Holmes (2001), both of the young children were exposed
to an additional block of relational trials that contained emotive terms rather than objects
or locations. For example, the experimenter placed two pictures on the table, one in front
of herself and the other in front of the child. On a simple trial, the child was told, “I have
a happy face and you have a sad face. How do I feel? How do you feel?” On a reversed

trial, the child was instructed, “I have a happy face and you have a sad face. If I were you
and you were me, how would I feel? How would you feel?” Finally, on a double reversal,
the experimenter placed two pictures in front of the child. The picture depicting “today’s
face” was always the bottom picture; “yesterday’s face” was always at the top. The child
was then told, “Yesterday you were standing there at a happy place; today you are stand-
ing here at a sad place. If here was there and there was here, and if now was then and
then was now, how would you feel now? How would you feel then?” Surprisingly, the data
obtained with both children who had been trained and tested on all aspects of I-you and
here-there indicated that they each produced perfect responding on the relational tasks
involving emotion without any explicit training. This outcome provided some preliminary
evidence of the ease with which emotional functions can be transformed through the
perspective-taking relations, once the latter have been firmly established.
Concluding Comments
The current chapter offered a working account of how perspective taking may be under-
stood and trained as a set of core derived relational responses. The chapter reviewed some
existing empirical evidence in support of the various concepts and distinctions contained
within this account, and it presented a training protocol that attempts to highlight the
target relational skills and the most effective training sequence. Training notes were pre-
sented throughout the work to assist teachers and other trainers to focus on the neces-
sary content and to construct an effective and enjoyable learning environment. Existing
empirical evidence using this and related protocols has provided positive support for the
utility of the procedure as an educational tool for use with both normally developing and
developmentally disabled children.
References
Barnes-Holmes, Y. (2001). Analysing relational frames: Studying language and cogni-
tion in young children. Unpublished doctoral thesis, National University of Ireland,
Maynooth.
Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind. Cambridge:
Bradford Books/MIT Press.
Fisher, N., & Happé, F. (2005). A training study of theory of mind and executive func-
tion in children with autistic spectrum disorders. Journal of Autism and Developmental
Disorders, 35, 757–771.
Skinnerian account of human language and cognition. New York: Plenum Kluwer
Academic/.
Howlin, P., Baron-Cohen, S., & Hadwin, J. (1999). Teaching children with autism to
mind-read: A practical guide. Chichester, England: Wiley.
Chapter 13 299

Lohman, D., Hagen, E., & Thorndike, R. (2001). Cognitive abilities test. London: NFER-
Nelson.
McGuinness, R. (2005). Using protocols of relational responding to analyse perspective-tak-
ing and related repertoires in children with Asperger’s syndrome. Unpublished master’s
thesis, National University of Ireland, Maynooth.
McHugh, L. (2004). Integrating relational frame theory and theory of mind: An empiri-
cal investigation. Unpublished doctoral thesis, National University of Ireland,
Maynooth.
McHugh, L., Barnes-Holmes, Y., & Barnes-Holmes, D. (2004). Developmental trends in
perspective taking. Psychological Record, 54, 115–145.
McHugh, L., Barnes-Holmes, Y., O’Hora, D., & Barnes-Holmes, D. (2004). Perspective
taking: A relational frame analysis. Experimental Analysis of Human Behavior Bulletin,
22, 4–10.
Myles, B., Bock, S., & Simpson, R. (2001). Asperger syndrome diagnostic scale. Minneapolis:
Pearson Education.
Rehfeldt, R., Dillen, J. E., Ziomek, M. M., & Kowalchuck, R. (2007). Assessing relational
learning deficits in perspective-taking in children with high-functioning Autism spec-
trum disorder. Psychological Record, 57, 23–47.
Suddendorf, T., & Fletcher-Flinn, C. (2002). Theory of mind and the origin of divergent
thinking. Cognition, 7, 78–113.
Whiten, A., & Byrne, R. (1997). Machiavellian intelligence II: Evaluations and extensions.
Cambridge: Cambridge University Press.

Chapter 14
Establishing Empathy
Sonsoles Valdivia-Salas, University of Albany, New York;

Carmen Luciano, University of Almeria, Spain;
Olga Gutiérrez-Martinez, University of Central Barcelona, Spain;
and Carmelo Visdómine, Justice Administration, Madrid, Spain
According to Goleman (1995), emotional intelligence is the winning combination that

makes us successful in life. Specifically, the skills involved in engaging reciprocally, appre-
ciating another’s perspective, communicating intention, empathizing, and understanding
the complexities of imagination, irony, humor, and other implicit messages contribute
much to the richness of our existence.
Emotional intelligence comprises five principal features that appear to underpin the
core skills identified above: (1) being aware of our own emotions; (2) being able to manage
our emotions; (3) being sensitive to the emotions of others; (4) being able to respond to,
and negotiate with, others emotionally; and (5) being able to use our own emotions to
motivate ourselves (Goleman, 1995; Mayer & Salovey, 1993). As a key component of
emotional intelligence, empathy is primarily encompassed within points 3 and 4, above;
we usually think of empathy as understanding how others feel, and as almost sharing the
feelings, although they are not really ours.
Despite empathy’s centrality to our development and subsequent happiness, normal
learning interactions do not always lead to the expected acquisition of empathic skills.
Naturally, in cases where empathic skills do not develop, the outcome for the self and
others can be very problematic. As remediation, training in empathy is often subsumed
within generic programs for training social skills (including perspective taking), in a
manner that seems to underplay the crucial emotional components of the deficit. In
recent years, however, there has been growing empirical evidence showing that interven-
tions based on the principles and practices of derived relational responding can address
both the cognitive and emotional aspects of deficits in empathy in a highly efficacious
manner.
The current chapter attempts to articulate the principles and practices of derived rela-
tional responding in a manner that will enable teachers to diminish deficits in empathy
when they are identified in individuals with developmental delay or disability. The chapter
is organized into two key parts. First, the concept of empathy is described, along with the
training strategies most frequently employed for the remediation of empathy deficits, and
their common limitations. Then we discuss the conceptual and empirical advantages of
understanding empathy as repertoires of derived relational responding, along with train-
ing strategies for their establishment.
Defining Empathy
The existing literature on empathy is simply vast. Yet it is characterized by a lack of
consensus on definitions of the core concept or agreement on how it develops (Davis,
1996). According to Hoffman (1982), the development of empathy comprises four stages.
Stage 1, global empathy, normally encompasses the first twelve months of a child’s life
and essentially involves generalized imitation of the emotions of others. Stage 2, egocen-
tric empathy, usually emerges around eighteen months and involves the child’s ability to
discriminate the emotions of others while continuing to respond in an egocentric fashion
(i.e., the discriminations of others do not guide one’s own behavior). At stage 3, empathy
for another’s feelings usually emerges and involves the child discriminating more subtle
emotions in others while offering progressively less-egocentric solutions. Finally in stage
4, an understanding of the potential discordance between emotions and actions is discrimi-
nated in late childhood and early adolescence, and the learner begins to respond to the
life conditions that influence the emotions of others, rather than responding to the emo-
tions themselves.
In normal development, the emergence of empathy runs in tandem with increasing
social sophistication. But the developmental trajectory that characterizes autism and other
developmental disabilities is not of this order. In contrast, these individuals often dem-
onstrate core deficits in affective and reciprocal interpersonal relationships with others,
permitting few or no displays of empathy. Indeed, even high-functioning individuals
with autism have greater-than-normal difficulty on tasks that require them to label their
own complex emotions, such as pride and embarrassment (Capps, Yirmiya, & Sigman,
1992), or even to recognize the facial emotional expressions of others (Hobson, 1986).
Interestingly, however, no such deficits are recorded when individuals present with autistic
features but normal levels of verbal abilities (Ozonoff, Pennington, & Rogers, 1990). This
latter finding, therefore, suggests some degree of overlap between empathy and verbal
sophistication.

Conventional Treatments for Empathy Deficits
As one might expect, treatment packages that attempt to remediate deficits in empathy
commonly focus on perspective taking (Nickerson, 1999), and they usually do so
according to the principles and practices of the mainstream cognitive approach known
as theory of mind (ToM; Baron-Cohen, 1995; see also chapter 13 of the current volume).
According to this approach, deficits in empathy come from weaknesses within the five
levels of understanding of informational states of the self and other people. The abili-
ties within these levels are believed to range from simple visual perspective taking to
understanding actions on the basis of true and false belief (Howlin, Baron-Cohen, &
Hadwin, 1999).
According to ToM, a common means of assessing empathy deficits begins with the
Sally-Anne Test (Baron-Cohen, Leslie, & Frith, 1985). Consider the following example
involving this procedure. The teacher holds up two puppets—Sally and Anne. Sally puts
a marble in a basket and leaves the room. Anne then moves the marble over to her box.
When Sally returns, the teacher asks, “Where will Sally look for the marble?” and of
course the correct answer is “in the basket.” Errors on the Sally-Anne Test (suggesting
that Sally will look for the marble in Anne’s box) indicate that the learner is responding
in accordance with Anne’s information, rather than Sally’s, because the former is more
consistent with reality. In other words, the learner is required to appreciate that Sally’s
perspective is different from the learner’s perspective (and different from Anne’s), because
only Sally did not see the marble being moved.
Where ToM adds emotion to perspective taking in an attempt to train empathy is
reflected in the training regime known as “social stories,” used to assess and establish the
recognition and prediction of the emotional responses of others. Critical aspects of this
procedure include (1) consideration of the perspective of the affected individual; (2) the
use of simple visual cues; (3) careful attention to the social context in which the target
episode occurs; (4) extraction of the most relevant aspects of the social situation; (5)
making the most appropriate reaction explicit; (6) inclusion of the possible emotional
reactions of the others involved; (7) the explicit explanation of why a particular reaction
will provide a positive consequence for both perspectives; and (8) an attempt to unravel
the complexities of later social interactions (Marans, Rubin, & Laurent, 2005). This pro-
cedure is often used to help children with autism prepare and rehearse socially appropriate
reactions when confronted with social situations.
To establish even more subtle social and emotional discriminations and appropri-
ate reactions, ToM practitioners often employ the “comic strip conversations” procedure
(Gray 1994). Specifically, this visual device is used to help high-functioning individuals
with autism to understand the literal and nonliteral information that is exchanged within
conversations, by slowing down the social interaction to permit a step-by-step analysis of
what is occurring. Specifically, through the visual representation of different situations,
learners are encouraged to focus on what others might be thinking, and what might
be motivating them to react in certain ways. This procedure is believed to promote the
identification of different thoughts that suggest different alternatives to the same social
situation. In a computerized variation of this procedure called “bubble dialogue,” learners
interact directly with characters in order to experience how thoughts influence overt action
Chapter 14 303

and vice versa (Rajendran & Mitchell, 2000). Other ToM training programs designed to
tackle empathy deficits include the I LAUGH model, the Teach Me Language program,
and the Situations, Options, Choices, Strategies, and Stimulations (SOCSS) program
(Marans et al., 2005).
Our Concerns with ToM

We have a number of difficulties with ToM-based programs used to treat deficits in
empathy. First, we have obvious concerns about the fact that empathy deficits are often
targeted indirectly by programs designed to train perspective taking, or broader social or
conversational skills. In our view, empathy is more than perspective taking, because the
latter involves the attribution of mental states to the self and others, while the former also
incorporates an emotional reaction that is appropriate to the other’s mental state as per-
ceived. Hence, in our view, we have no doubt that empathy involves perspective taking,
but we believe that training in perspective taking would not suffice in the establishment
of empathy, nor would it remediate deficits in this regard. In our opinion, ToM training
regimes simply do not devote adequate attention to understanding and responding to the
emotions of others.
Second, the highly instructional nature of many perspective-taking programs suggests
to us the possibility that, with regard to emotions, learners are simply being trained to
label the emotions expressed by others, which is in fact what individuals with autism and
Asperger’s disorder already do (Lord, 1993). This simple labeling does not mean that the
appropriate discriminations are in place, nor does it mean that they are under appropriate
emotional or social control.
Third, there may be alternative and additional discrimination repertoires that con-
tribute to empathy, the absence of which means that even training in perspective taking
would not yield the emergence of empathy repertoires. For us, a full understanding from
the bottom up is necessary to detect exactly what the problem in empathy is really about
for any individual.
Fourth, despite widespread use of ToM regimes, there is limited empirical evidence
with regard to gains in empathy or even generalization to other ToM tasks (Corbett,
2003; Golan & Baron-Cohen, 2006; Silver & Oakes, 2001). Outcomes associated with
more behaviorally oriented programs for remediating deficits in perspective taking in
empathy do not fare much better. For instance, although Charlop-Christy and Daneshvar
(2003) reported some perspective-taking gains with a combination of video modeling and
multiple-exemplar training, LeBlanc and colleagues (2003) reported weaker outcomes
with video modeling and positive reinforcement. Indeed, even in ToM and behavioral
regimes where there are positive outcomes, there are often no accounts of generalization
to novel tasks, much less changes in social interactions. In other words, the learning of
specific emotional labels does not generalize to other markers of the same emotion, nor
do these labels assist the learner to be more effective in his emotional interactions with
the world.

Empathic Behavior as Relational Responding
The modern behavior analytic concept of relational learning derives primarily from
the account of language and cognition known as relational frame theory (RFT; Hayes,
Barnes-Holmes, & Roche, 2001). The core features of this approach are articulated in
other chapters of the current volume; for example, see our own work in chapter 8, and
chapter 13, on perspective taking. Hence, details of RFT will not be articulated here
above and beyond what is needed to understand the concepts presented within. In the
following section, we take the core features of empathy and translate them behaviorally
and in the language of RFT so that the teaching strategies suggested subsequently will be
more meaningful.
Understanding your own emotions. Behaviorally, we think of this as verbally dis-

criminating one’s own private events. For RFT, this involves coordination relations, for
example, between internal states (feeling sad) and external labels (“I am sad”). However,
it is also more complicated, in that it involves responding in accordance with perspective-
taking relations, because I always discriminate my own experience from the perspective
of I-here-now. There are also coordination relations that enable us, for example, to equate
what is felt now with what was similarly felt some time in the past (I-there-then). There
may even be comparative relations that enable you to discriminate that if X happened
yesterday and that made you feel bad, and X and Y happened today and you feel worse,
then today you are having an even worse day than you did yesterday, and now you feel
still worse. Notice that temporal and spatial relations are essential features of relational
perspective taking because I-there-then is being constantly translated into I-here-now as
I think about it.
Understanding the emotions of others. Naturally, any understanding of others requires

a shift in your perspective from I-here-now to you-here-now (where you is another) or
even you-there-then (although you continue to experience all of this from I-here-now).
In addition to adopting an alternative perspective, you must also be able to respond to
cues that enable discriminations of the likely private events of others (for example, if the
teacher smiles, the learner can guess that she has given the right answer). The cues in this
case will differ to some extent from the cues that allow you to discriminate your own
events, because you cannot actually feel what others feel. However, for RFT, emotional
functions may transform through coordination relations in which both you and another
participate. Imagine, for example, that a young girl (Sonsoles) fell and cut her knee and
began to cry. The next day, her friend (Olga) fell and cut her knee and also began to cry.
If Sonsoles discriminated that Olga’s wound looked similar to hers (coordination relation
between the two knees), then the emotional functions attached to Sonsole’s knee (pain
and distress) would transfer via these relations to Olga and her knee. As a result, Sonsoles
would attribute the same emotional functions she felt yesterday to Olga today. In other
words, once the coordination relations between the knee and the crying are there and the
emotional functions (pain and distress) are there, the expansion of the coordination and
perspective relations to another mean that the emotional functions will transfer in accor-
dance with the relations. So, once you have the relations, you should have the emotions.
Chapter 14 305

Responding empathically. When a young child cries, other very young onlookers also
cry, and of course this only makes the first child cry even more. This phenomenon is
commonly referred to as emotional contagion (Smith, 2006). Thankfully, more appropriate
responding comes later as perspective taking develops and the relational networks become
enlarged, so that an onlooker to a painful event (such as that of Sonsoles in our example
above) learns to act in accordance with what another is experiencing as distinct from what
she is actually experiencing. In other words, the transfer of emotional functions might
make Sonsoles reexperience some of the pain that Olga is having, but she knows that her
own knee is no longer actually sore. As a result, she learns that it is more appropriate to
assist Olga in this instance (perhaps in the same manner in which she was assisted after
her own fall) than to begin to cry about her own remembered pain (although the type of
reaction will be a function of the child’s verbal history). This alternative response pattern
also involves conditional relations along the lines of “If I liked being helped yesterday and
it caused me to feel better, then she will feel the same, so I should offer the same help and
this will result in her feeling better, as I did yesterday.”
Guidelines and Case Study in the Training of Empathy

as Relational Responding
There is no empirical research on the definition or training of empathy from an RFT
perspective—this research simply has not yet been done. But a number of RFT studies
have contributed to our understanding of several aspects of empathy, and they offer impor-
tant starting points for the teaching strategies we have articulated below. Specifically, they
are designed to enable teachers to use RFT-based strategies to teach learners to discrimi-
nate their own private experiences and emotions and the private experiences and emotions
of others. In the case study in the following section, we present four areas of potential
deficits that contribute to a developmentally delayed learner (we have called her Emma)
failing to act empathically toward a peer. At each level, we identify the missing repertoires
that appear to form the basis of Emma’s reaction, and we briefly describe training strate-
gies that might be useful in remediating these deficits.
Case Study
Imagine that it is lunchtime and Emma observes that Tom (one of her classmates)
begins to cry because he forgot his lunch at home and doesn’t have anything to eat. While
observing this, Emma continues to eat her lunch. Assuming that sharing her lunch was
the most appropriate thing for Emma to do in this situation, the following sections each
present a core deficit that might account for Emma’s lack of empathy. (Of course, it may
not always be wise for Emma to share her lunch with Tom. For example, if Tom forgot
his lunch every day, it would not be a good contingency for Emma to share her lunch
every time because Tom would continue to forget his. In this instance, Emma would
have to learn how often it would be acceptable to share her lunch when Tom has forgot-
ten his; hence the empathy here would not be deficient but simply fails to operate under
appropriate contextual control.)

Deficit 1: Failing to discriminate the aversive or reinforcing functions of events.
Emma’s failure to share her lunch with Tom may result from the simple fact that she has
never experienced skipping a meal (an event that is coordinated with forgetting lunch),
in which case she will have no idea about what Tom is experiencing (hunger pangs, for
example). Where this is the deficit, you must establish that the learner has experienced,
and recalls, the private events that allow her to take the perspective on behalf of another.
So you may simply have to start off by providing a range of these experiences in the form
of an explicit training history. Put simply, the learner must be able to make experiential
contact with the aversive or reinforcing functions of events in order to be subsequently
able to name these functions and discriminate them (in terms of causes and consequences)
from other events, as described in Luciano (1996).
Deficit 2: Failing to discriminate and label one’s own emotions. Even if Emma has
direct experience with aversive states such as hunger, she may not respond to Tom in a
similar situation because she cannot appropriately label emotions (for example, she doesn’t
know that crying means sad), even as they apply to herself. Behavioral practitioners often
refer to this as tacting one’s own private experiences. According to Skinner (1945), the
verbal community teaches us to tact our private events and others’ through the transmis-
sion of public correlates. Consider a learner tacting her own current pain by saying, “It
hurts.” The verbal community might use the following markers that allow these personal
experiences to be translated to others: First, there will be visual and public accompani-
ments to pain, such as blood or damaged tissue. Second, the learner will present collateral
responses, like crying or other distorted facial expressions. Third, there will be concor-
dance with previous events experienced by others. For example, the learner might say, “I
feel the same way I did when I broke my wrist.” Fourth, shared analogies may be drawn
when the pain is described as “burning” or “shooting,” for example.
There is considerable empirical evidence of how interoceptive stimuli transfer their
functions to exteroceptive stimuli through derived relations, even when the stimuli were
never directly paired. Consider the following example from DeGrandpre, Bickel, and
Higgins (1992). Imagine the emergence of an equivalence class containing “It aches” and
“It hurts” and the interoceptive stimuli associated with an earache, as experienced by a
child. When the child then experiences similar pain with a headache, new equivalence
relations emerge and she can now say the headache aches, hurts, and is like an earache.
The bidirectionality between a private experience (such as hunger pangs) and a label
(“hunger” or “I am hungry”) can be explicitly taught through MET so that the learner
spontaneously says, “I’m hungry,” when stomach pangs are discriminated (for example,
the learner may be taught to touch his belly while mentally checking to see if anything
can be felt inside). Equally, the absence of such private events should be in a relation of
distinction with statements such as “I’m hungry” (in other words, don’t say you’re hungry
when you’re not). You can teach basic emotions in the same way. If the learner has already
acquired the basic relational operants of coordination, opposition, difference, comparison,
and hierarchy (see chapter 8 of the present volume), whole networks of relations can be
established between labels and emotions and between different emotions (for example,
being in pain and being hungry are both worse than being happy, but being in pain is
worse than being hungry). And remember that when the right relations are in place, the
learner need not actually feel hunger; this can be depicted visually or even stated ver-
bally, and yet the appropriate functions should still transform. Once these networks are
Chapter 14 307

e stablished, it is relatively easy to add in appropriate responses. For example, if the learner
is hungry and you ask whether he would like a snack or juice, the learner should be able
to select a snack, whereas if the deprivation state is thirst, choosing juice would result in
a better consequence for articulating that particular emotion.
Deficit 3: Failing to discriminate and label others’ emotions. Though Emma may
be able to discriminate her own emotions, tact them, and relate them to external events,
she may not be able to do so for another person because the appropriate contextual cues
that guide these discriminations are not established. In colloquial terms, Emma knows
Tom is crying, but she doesn’t know why. Hayes (1994) proposed the term “subtle events”
to describe distinctions between apparent versus unapparent stimulation (for example,
bad news versus a thought), and between apparent versus unapparent responses (such
as crying versus remembering). Hayes thus emphasized that being able to discriminate
another’s subtle events may well be a matter of the shared history between the two indi-
viduals. The training regime described in a recent study provides useful steps that would
enable learners to discriminate the contextual cues for understanding the presence (and
absence) of specific emotions displayed by others (Luciano, Cabello, Molina, Gomez, &
Ortega, 2003).
In this research, forty-two participants were exposed to a series of cards depicting
the same character (Alfredo) in different roles (at work, under stress, with flowers, with
his son, and in his leisure time). Although each role was depicted with more than one
scenario, each role contained a specific contextual cue that denoted a particular thought
and subsequent action. Consider, for example, Alfredo under stress denoted by the pres-
ence of the contextual cue sweating. Participants were first asked to identify a thought
that Alfredo might be having in this situation and what actions he is likely to engage in.
During multiple-exemplar training (MET), participants were presented with four pairs of
cards, each pair depicting Alfredo sweating in a different circumstance. One from each
pair was a thought card (in the presence of drops of sweat, Alfredo always had a thought
about his lack of ability to face a situation (for example, “I’m so nervous, I cannot go to
the party!”), and the other was a related action card (in the presence of sweat, Alfredo
always ended up doing what he thought he was not able to do, such as going to the
party). After explicitly training in the four scenarios, participants were presented with
a card depicting Alfredo sweating in a novel situation, and they were asked, “Based on
what you have learned about Alfredo, what do you think he is thinking now?” and “What
do you think he will do after having such a thought?” The rest of scenarios were trained
and tested in a similar way. As the result of the training, the majority of participants cor-
rectly predicted Alfredo’s thoughts and actions in novel situations. In other words, our
experimental participants readily learned how to abstract the shared cue in situations that
indicated the presence of certain thoughts and actions.
One common outcome of deficits in discriminating the contextual cues that signal
others’ emotional states looks like emotional contagion but is not. Imagine, for example,
that Emma begins to cry when Tom cries, suggesting that the appropriate coordination
relations are present but the perspective taking is weak. In other words, Emma is behaving
as if she is hungry and thus not discriminating between her own emotions and the emo-
tions of others. In this case, perspective-taking skills would be trained by means of MET
involving questions like those included in the protocols reported by McHugh, Barnes-
Holmes, and Barnes-Holmes (2004) and Rehfeldt, Dillen, Ziomek, and Kowalchuk

(2007). Such questions might include “Who forgot or didn’t forget the lunch? Who is
hungry? Are you hungry? If you were Tom, would you be hungry? If Tom were you, what
would your tummy be feeling like?” A whole range of thoughts and feelings from both
perspectives can be established in this way (see chapter 13 in the current volume). Again,
the child needs to have experienced the emotion in question in order for the emotional
transformation of functions to take place (Luciano, 1996). For example, let Emma miss
her lunch and feel hungry, then ask her what she is feeling, whether Tom is feeling the
same, and what Tom would be feeling if he were Emma.
Deficit 4: Failing to respond empathically. It may also be the case that Emma dis-
criminates what Tom is feeling but does not know how to respond appropriately. For
RFT, you would need to establish causal relations between the discrimination of Tom’s
situation and what the onlooker should do next (for example, if Tom cries, then Emma
should assist). Training action-based causal relations in this way is relatively simple as
MET, but it is also wise to train variations of this (for example, if X, then not Y; and if
not X, then Y). Doing so might look like the example involving Alfredo above but with
the addition of onlooker actions, in which the learner matches two cards—one with
Alfredo’s thought cues and another with a helper’s actions.
It might also be the case that Emma fails to respond appropriately to Tom because
she has not learned to care. In other words, she has not yet learned to engage in actions
simply for the “greater good” (in other words, those controlled by long-term or symbolic
contingencies). For RFT, this involves a transition in rule governance from pliance (when
Emma offers help only because of a history of being told to do so, in order to get approval
from others or avoid punishment) to tracking based on short-term and direct contingen-
cies (when Emma offers help because of a history of being subsequently rewarded for
the direct consequences of the actions). This transition may even extend to other types
of rule following in which the controlling contingencies are symbolic. In the latter case,
the contingencies have never been contacted directly but participate in relational frames
with other events that have been contacted. In order to establish this type of higher-
level social responding, you must provide social contingencies upon positive and negative
instances of the learner complying with rules delivered by others (social contingencies
might include “good boy,” “Mum is really happy now,” and the like), as well as descrip-
tions of the natural contingencies of rule following. The social contingencies should then
be progressively faded out so that eventually rule following is controlled by the set of
events that naturally occur when the rule is followed (for example, “If I help Tom, he will
stop crying, and we will all be able to go to the playground after lunch”). (See chapter 16
in the present volume, on the development of self-directed rules.)
Indeed, what Emma is doing here constitutes a type of self-regulation in which a
social perspective potentially ranks higher than Tom’s or even her own.
Concluding Comments
Empathy has been regarded as a key feature of emotional intelligence and a prerequisite
to social success. Yet as practitioners, we have little to be confident about with regard to
its establishment. The guidelines presented here are simply aimed at filling such gaps in
Chapter 14 309

the existing literature and teaching programs, and thereby offering teachers a useful and
informed starting point for assessment and intervention in empathy deficits. Although
the guidelines’ effectiveness has not been empirically demonstrated, they are based on
a strong body of basic and applied research showing the possibilities and benefits of the
relational learning paradigm in the emergence of novel behavior.
References
Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind. Cambridge,
MA: Bradford Books.
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a “theory
of mind”? Cognition, 21, 37–46.
Capps, L., Yirmiya, M., & Sigman, N. (1992). Understanding of simple and complex
emotions in non-retarded children with autism. Journal of Child Psychology and
Psychiatry, 33, 1169–1182.
Charlop-Christy, M. H., & Daneshvar, S. (2003). Using video modeling to teach per-
spective taking to children with autism. Journal of Positive Behavior Interventions, 5,
12–21.
Corbett, B. A. (2003). Video modeling: A window into the world of autism. Behavior
Analyst Today, 4, 367–377.
Davis, M. H. (1996). Empathy: A social psychological approach. New York: Westview.
DeGrandpre, R. J., Bickel, W. K., & Higgins, S. T. (1992). Emergent equivalence rela-
tions between interoceptive (drug) and exteroceptive (visual) stimuli. Journal of the
Gray, C. (1994). Comic strip conversations and social stories. Arlington, TX: Future
Horizons.
Golan, O., & Baron-Cohen, S. (2006). Systemizing empathy: Teaching adults with
Asperger’s syndrome or high-functioning autism to recognize complex emotions using
interactive multimedia. Development and Psychopathology, 18, 591–617.
Goleman, D. (1995). Emotional intelligence. New York: Bantam
Hayes, L. J. (1994). Thinking. In S. C. Hayes, L. J. Hayes, M. Sato, & K. Ono (Eds.),
Behavior analysis of language and cognition. Reno, NV: Context Press.
Plenum.
Hobson, R. P. (1986). The autistic child’s appraisal of expression of emotions. Journal of
Child Psychology and Psychiatry, 27, 321–342.
Hoffman, M. L. (1982). Development of prosocial motivation: Empathy and guilt. In N.
Eisenberg (Ed.), The development of prosocial behaviour. New York: Academic Press.

Howlin, P., Baron-Cohen, S., & Hadwin, J. (1999). Teaching children with autism to
mind-read: A practical guide. Chichester, England: Wiley.
LeBlanc, L. A., Coates, A. M., Daneshvar, S., Charlop-Christy, M. H., Morris, C., &
Lancaster, B. M. (2003). Using video modeling and reinforcement to teach perspec-
tive taking skills to children with autism. Journal of Applied Behavior Analysis, 36,
253–257.
Lord, C. (1993). Early social development in autism. In E. Schopler, M. E. Van Bourgondien,
& M. M. Bristol (Eds.), Preschool issues in autism. New York: Plenum Press.
Luciano, M. C. (1996). Intervención psicológica en retraso en el desarrollo: Una perspec-
tiva funcional [Psychological intervention in developmental disabilities: A functional
approach]. In M. C. Luciano (Ed.), Manual de psicología clínica [Handbook of clinical
psychology]. Valencia, Spain: Promolibro.
Luciano, M. C., Cabello, F., Molina, F., Gomez, I., & Ortega, J. (2003). Enhancing
perspective-taking in the context of thinking. Paper presented at the 29th annual con-
vention of the Association for Behavior Analysis, San Francisco, CA.
Marans, W., Rubin, E., & Laurent, A. (2005). Addressing social communication skills in
individuals with high-functioning autism and Asperger syndrome: Critical priorities
in educational programming. In F. R. Volkmar, R. Paul, A. Klin, & D. Cohen (Eds.),
Handbook of autism and pervasive developmental disorders, Vol. 2. Hoboken, NJ: John
Wiley & Sons.
Mayer, J. D., & Salovey, P. (1993). The intelligence of emotional intelligence. Intelligence,
17, 433–442.
McHugh, L., Barnes-Holmes, Y., & Barnes-Holmes, D. (2004). Perspective-taking as
relational responding: A developmental profile. Psychological Record, 54, 115–144.
Nickerson, R. S. (1999). How we know—and sometimes misjudge—what others know:
Imputing one’s own knowledge to others. Psychological Bulletin, 125, 737–759.
Ozonoff, S., Pennington, B. F., & Rogers, S. J. (1990). Are there emotion perception deficits
in young autistic children? Journal of Child Psychology and Psychiatry, 31, 343–361.
Rajendran, G., & Mitchell, P. (2000). Computer mediated interaction in Asperger’s syn-
drome: The Bubble Dialog program. Computers and Education, 35, 189–207.
Rehfeldt, R. A., Dillen, J. E., Ziomek, M. M., & Kowalchuk, R. K. (2007). Assessing
relational learning deficits in perspective taking in children with high-functioning
autism spectrum disorder. Psychological Record, 57, 23–47.
Silver, M., & Oakes, P. (2001). Evaluation of a new computer intervention to teach people
with autism or Asperger syndrome to recognize and predict emotions in others.
Autism: International Journal of Research and Practice, 5, 299–317.
Skinner, B. F. (1945). The operant analysis of psychological terms. Psychological Review,
52, 270–277.
Smith, A. (2006). Cognitive empathy and emotional empathy in human behaviour and
evolution. Psychological Record, 56, 3–21.
Chapter 14 311

Chapter 15
Mathematical Reasoning
Chris Ninness, James Holland, Glen McCuller, Robin Rumph,

Sharon Ninness, and Jennifer McGinty, Stephen F. Austin State
University; and Mark Dixon, Southern Illinois University
In U.S. public schools, the dominant pedagogical orientation to education is derived from
the constructivist view of learning theory (e.g., Davis, Maher, & Noddings, 1990) and is
often referred to as progressive education (Rumph et al., 2007). Two key pedagogical fea-
tures distinguish this orientation from traditional learning approaches. First, learners are
encouraged to reflect on their own experience and explore learner-centered group activi-
ties through which solutions to problems might be creatively generated. Teaching math
for example, in such a framework, often involves nondirective support while students
construct their own mathematical interpretations or utilize calculating and graphing
technologies—a process referred as a whole math. Second, in keeping with an emphasis
on hands-on learning, constructivist teaching of math relies heavily on physical repre-
sentations of quantitative abstractions rather than the direct acquisition of more abstract
mathematical principles. Indeed, this educational framework is opposed to the extrinsic
reinforcing of academic behavior and has a strong disregard for the memorization or
rote rehearsal of definitions, rules, or theorems. As a result, most U.S. students have
not been subjected to the drill and practice of basic mathematical relations or rules, and
only “a minority of states require knowledge of the standard algorithms of arithmetic for
addition, subtraction, multiplication, and division” (Klein, 2005).
Although an emphasis on the physical or nonarbitrary properties of real stimuli in
the environment through a medium of interactive learning has natural appeal as an edu-
cational regime, particularly with developmentally delayed populations, there is limited
empirical evidence to suggest that it offers any advantages over traditional methods here or
with learners who are normally developing. Indeed, in the context of mathematical attain-
ment, there are mounting concerns about declining standards that might be attributed to
constructivist teaching regimes. For example, U.S. students perform far below students
in the rest of the industrialized world at mathematics (Mayfield & Chase, 2002), and as
many as 50 percent fail to pass fundamental mathematics courses (EnabLearning, 2002).
More specifically, in a 2003 Programme for International Student Assessment (PISA)
survey, U.S. public school students ranked twenty-fourth among students from twenty-
nine nations on math literacy and problem solving, and 25 percent demonstrated math
skills at or below the lowest recordable level of mathematical proficiency. This worrying
trend in core mathematics recorded with students with normally developing potential
has led some researchers to coin the new condition “pervasive mathematics disorder.” It
seems logical to assume, therefore, that even more severe outcomes would be associated
with students of reduced learning potential, such as those with developmental delays. Put
simply, if contemporary teaching strategies fail to foster learning in normally develop-
ing populations, then students with developmental delays have almost no chance of ever
acquiring basic math knowledge.
The primary aim of the current chapter is to select areas of mathematics instruction
that are often observed as difficult to teach explicitly, and to describe a number of behav-
ioral interventions that appear to be highly effective in this regard. Although the aspects
of math selected (such as graphing trigonometric functions) represent a relatively high
level of mathematical ability, these are fundamental features of abstract mathematics and
are ones in which there is a growing body of empirical support for behavioral principles of
instruction in mathematical phenomena. Hence, although concerns with this issue may
have more direct relevance for teachers of normally developing students, the procedures
seem equally applicable to students of lower levels of academic ability and to other areas
of basic math. Put simply, if abstract math instruction is easily learned using these proce-
dures, more basic aspects of math should be the same.
Behavioral Interventions for Teaching Math

Behavioral interventions have a long and established history in effectively targeting aca-
demic and social deficiencies with normally developing, and particularly with develop-
mentally delayed, learners. Practically all of the preceding chapters in the current volume
attest to this fact. The chapters also support the view that, although the basic principles
that guide behavioral instruction have remained the same for more than forty years, the
techniques and content targeted by them has been evolving. For example, behavioral
educators no longer attempt to explicitly train all areas of target skill and content but
increasingly focus instead on the acquisition of novel or emergent behaviors. The study of
stimulus equivalence and other derived stimulus relations has contributed much to our
understanding and teaching of novel behavioral repertoires.
Using Stimulus Equivalence Procedures to

Teach Basic Math
The stimulus equivalence paradigm promotes learning through the development of
stimulus classes, in that when two or more stimuli control common responses, the stimuli

are collectively referred to as a class (Sidman, 1971). Moreover, when similar responses are
brought under the control of two or more stimuli, the process is referred to as stimulus
equivalence because of the functional equivalence between the stimuli in their responses.
Sidman and Tailby (1982) postulated that equivalence between stimuli is demon-
strated by tests of reflexivity, symmetry, and transitivity; matching-to-sample (MTS)
procedures are most commonly employed for the establishment of equivalence. To demon-
strate reflexivity, or identity matching, an individual must match a stimulus to itself (such
as A to A). Consider the following simple mathematical example. After being taught the
printed number 2 the learner must select a 2 from among several choices, without rein-
forcement for that selection. For symmetry, the learner must match two stimuli, regard-
less of which appears as the sample and which as the comparison (this may be denoted
as matching A to B and B to A). For example, if the learner has been taught to select 2
when shown two marks (for example, //; A-B), and then without training selects // when
shown a 2 (B-A), this response is defined as symmetry. To demonstrate transitivity, a third
stimulus must be involved in the class, such as the spoken word “two” (we’ll call this C).
Thus, the learner might be taught to select // when shown 2, and taught to say “two”
when shown // (in other words, A-B and B-C training). Transitivity is demonstrated if
the individual then selects 2 upon hearing “two” (in other words, A-C). This relationship
is also said to be equivalent when the individual simultaneously demonstrates symmetry
and transitivity (in other words, C-A responding). In this example, the learner would
now also say “two” when shown a 2. That is, the word “two” and the numeral 2 have
become interchangeable although they were never directly paired. These simple examples
indicate how readily large classes comprising the basic mathematical relations between
written numbers, written words, and actual numbers of stimuli (real or depicted), can be
established. Where such basic equivalence-based mathematical abilities are deficient, the
current examples can be expanded (for example, 3-///-three) in the construction of a series
of training protocols. Most important, developing the ability to derive equivalence rela-
tions such as these would mean that not all possible examples of number-based relations
would have to be laboriously instructed.
Training Fractions and Decimals

A number of recent studies have demonstrated the applicability of the stimulus equiv-
alence paradigm to the teaching of more advanced mathematical concepts. For example,
Lynch and Cuvo (1995) used equivalence procedures to teach fraction-decimal relations
to normally developing children (eleven to thirteen years old) who had math difficulties.
The stimuli consisted of printed fractions, pictured fractions (using a grid with shaded
slots to represent the fraction), and printed decimals. Training and testing were conducted
via computer using an MTS format with four comparison choices.
The training comprised providing reinforcement to learners for selecting pictured frac-
tions (let’s call these B stimuli) that matched printed fractions (A stimuli), and for select-
ing printed decimals (C stimuli) that matched the pictured fraction (Bs). In other words,
they learned relations between printed fractions and pictured fractions, and between pic-
tured fractions and printed decimals. Naturally, the instructional aim was to establish an
equivalence class among the three types of stimuli that would enable the learners to select
printed decimals in the presence of printed fractions (A-C) and vice versa (C-A). Indeed,
Lynch and Cuvo (1995) reported that all participants achieved the training criterion after
Chapter 15 315

only one or two training sets and subsequently demonstrated the target symmetry, transi-
tivity, and equivalence relations. Again, where such basic equivalence-based mathematical
abilities are deficient, the current examples can be expanded in the construction of a series
of training protocols. And developing the ability to derive equivalence relations such as
these would mean that not all possible examples of relations between printed or pictured
fractions and decimals would have to be instructed directly.
Fostering Generalization
Although equivalence and derived stimulus relations paradigms are built from basic
assumptions regarding generic derivation skills, additional explicit training is sometimes
necessary to foster some types of generalization. In the previous study, for example, Lynch
and Cuvo (1995) assessed the generalization potential of the newly established fraction-
decimal relations by giving participants a generalization test in a paper and pencil format
(rather than via a computer program), as well as two additional tests involving novel frac-
tions and decimals. However, despite their training success, participants failed to general-
ize to the paper and pencil test, with only 63 percent accuracy at best. And accuracies on
the other two generalization tests varied from chance to 100 percent.
Outcomes from subsequent research, however, suggest a number of adjustments to
training that would likely facilitate generalization where it does not emerge, or which
may be better suited to less-able learners, such as those with developmental delays. For
example, Leader and Barnes-Holmes (2001) used equivalence procedures to teach frac-
tion-decimal relations to five-year-old normally developing preschoolers. In this research,
the children were trained (as before) to match basic fractions (1/4 and 2/4) with correspond-
ing decimals and pictured fractions (circles with one or two quarters shaded). However,
the researchers systematically examined three different training techniques. In respondent
training, the trials did not present the target stimuli simultaneously as in the standard
MTS format but operated a brief time delay between the presentation of the sample and
the subsequent presentation of the correct comparison only. Reinforcement was provided
after every five trials if participants had looked at both stimuli. In demonstration trials,
the teacher modeled touching the sample and then touching the correct comparison in
an MTS format. Finally, during no-help trials, learners were shown the sample and com-
parison stimuli (again in MTS format) but no modeling occurred. Reinforcement and
corrective feedback were applied to all demonstration and no-help trials. Across all three
types of trials, all of the children successfully demonstrated equivalence and, most impor-
tant, this generalized to new picture shapes, differing shaded areas, and differing numbers
of shaded areas when they were introduced using progressively novel comparisons across
tests. These alternative training regimes may be incorporated into, or used instead of, the
more standardized MTS training described above in the construction of protocols for
establishing a full range of fraction-decimal relations.
Using Stimulus Equivalence Procedures to Establish

Advanced Math
Equivalence procedures have also been successfully employed to establish mathemati-
cal concepts that are more advanced but which can be taught in exactly the same way

as those described above. A recent study, for example, used equivalence and multiple-
exemplar training procedures (training and testing across an array of stimulus sets) to
establish formula-graph relations (Ninness et al., 2005). In this study, the derivation of
symmetrical graph-to-formula relations successfully emerged across no fewer than thirty-
six novel variations. Furthermore, equivalence performances that involved training rela-
tions of formula to factored formula (as these formulas pertain to reflections and vertical
and horizontal shifts) and deriving formula-to-graph relations were also successfully estab-
lished, and strong generalization performances across novel stimulus sets were recorded
(Ninness et al., 2005). In subsequent work, the same researchers developed Web-based
training protocols based on the original nonautomated procedures, which provided even
more interesting and more accessible training of formula-graph relations.
Web-Based Training of Advanced Mathematical

Relations
Although stimulus equivalence procedures appear to be highly effective in the establish-
ment of both basic and advanced equivalence-based mathematical relations, it is well
known that the learning of mathematical concepts may also be greatly assisted by the
introduction of rules. Findings from our own research provide considerable support for
this view (Ninness et al., 2006; Ninness et al., in revision). Specifically, we established a
multitude of diversified trigonometric formula-to-graph and graph-to-formula relations
using rules (a five-minute PowerPoint presentation on the basics of the rectangular coor-
dinate systems and the definition of mathematical reciprocity) combined with an inter-
locking protocol. The protocol for establishing these high-level mathematical abilities is
presented below.
Protocol
This protocol comprises primarily online multiple-exemplar training and testing with
respect to amplitude and frequency transformation relations. The online mathematical
software can be located on our website at www.faculty.sfasu.edu/ninnessherbe/graph
CalcCN07.html. For illustrative purposes, the current section also includes data from a
normally developing individual.
Pretraining
Pretraining involves training and testing trigonometric formula-to-formula (A-B) and
formula-to-graph (B-C) relations, as they concern positive and negative forms of the sine,
cosine, secant, and cosecant functions. This is followed by assessments of the derived and
untrained symmetry relations B-A and C-B, and the equivalence relations A-C and C-A.
See this chapter’s appendix for a comprehensive presentation of all relations trained.
Chapter 15 317

Online Multiple-Exemplar Training and Testing
Although pretraining will have established trigonometric formula-to-graph and graph-to-
formula relations as they concern positive and negative forms of the sine, cosine, secant,
and cosecant functions, learners will not yet understand trigonometric operations, as
they address amplitude and frequency transformations. This is the main focus of the
online training; the sequence is presented in table 15.1. (Note that * is the software’s
multiplication operator.)
Table 15.1. Training Sequence for Relations Pertaining to

the Graphical Transformations of Amplitude and Frequency
Cosine amplitude transformations with multipliers greater than and less than 1
y = 3*cos(x) y = 0.5*cos(x)
Train/test A1-B1 Test B1-A1 Train/test A2-B2 Test B2-A2
Cosine frequency transformations with multipliers greater than and less than 1
y = cos(2*x) y = cos(0.5*x)
Secant amplitude transformations with multipliers greater than and less than 1
y = 3*sec(x) y = 0.5*sec(x)
Secant frequency transformations with multipliers greater than and less than 1
y = sec(2*x) y = sec(0.5*x)
Train/test B7-C7 Test C7-B7 Train/test C8-A8 Test B8-A8
Note: (Adapted from Ninness et al., in revision)
Figure 15.1 is a representation of the basic format of the online mathematical soft-
ware. Constructing a graph simply involves dragging squares for each of the five (red)
graphing anchors from the top of the screen to locations on the coordinate axis until
the desired shape of the graph is fashioned. (It should be noted that, in order to run this
software, Adobe Acrobat Reader software is required.)

Figure 15.1. An illustration of the online training platform. Note that the text box
containing x = 6.26 and y = 2.98 represents the location of the mouse arrow on the
coordinate axes (adapted from Ninness et al. in revision).
Step 1: Amplitude relations with numbers greater than 1. Explain first that multiply-
ing the cosine function by 2 increases the amplitude of the graph by two. Describe ampli-
tude in terms of high and low points on the graph, and illustrate the way in which these
points “stretch” along the y-axis from ±1 to ±2, respectively. Then change the formula in
the text box from y = cos(x) to y = 2*cos(x), and click “graph.” This causes the amplitude
to increase by two, to ±2. Next, demonstrate how to use the graphing anchors to superim-
pose a graph over the existing sample graph and ask the learner to perform the same task
using the mouse to manipulate the graphing anchors (Peters, 2005). Because the software’s
anchors adjust the graphing line as they are moved via the mouse pointer, remind learners
by saying, for example, “As you move your anchors, be sure to check your x and y coordi-
nates. When you move one anchor, it affects the entire line that you are creating. You will
have to adjust the anchors several times before your graph looks like mine.” This process
is repeated using numbers greater than 1, until the learner is proficient. The learner can
locate any point on a graph by watching the floating text box (Reinhardt & Dowd, 2007)
that moves in conjunction with the mouse arrow (see figure 15.1).
With the graph of the basic cosine function y = cos(x) displayed on screen, type a new
formula in a text box, such as y = 3*cos(x) to test a formula-graph (A-C) relation regard-
ing amplitude expansion of the cosine function. The learner must then construct the
corresponding graph using the graphing anchors. To assess the symmetrical graph-formula
(C-A) relations, type the same formula (multiplier always between 2 and 5) into the lower
right text box. Although this input is not displayed on screen, the graph of the hidden formula
is displayed when you click “graph.” Then ask the learner to type the formula corresponding
Chapter 15 319

to the graph in the far left text box. When the learner clicks “graph,” the graph of the typed
function appears. An accurate response involves exact correspondence with the sample. If the
learner errs on any of the tests of formula-graph (A-C) or graph-formula (C-A) relations, the
training protocol for cosine should be repeated immediately and another test of the relations
should be conducted. Figure 15.2 illustrates a correct response to this test.
Figure 15.2. A correct response to a test of A-C relations pertaining to compression along
the y-axis (adapted from Ninness et al., in revision).
Step 2: Amplitude relations with numbers less than 1. Inform the learner that mul-
tiplying the cosine function by a number less than 1 creates a horizontally compressed
graph. The extent of the compression depends on the particular decimal multiplier
employed. The general rule is as follows: the smaller the decimal multiplier, the greater
the compression on the y-axis. For example, a compression that occurs when one is mul-
tiplying this function by 0.7 (in other words, y = 0.7*cos(x)) will be considerably less than
a compression that occurs when one is multiplying the function by 0.2 (in other words, y
= 0.2*cos(x)). Figure 15.3 illustrates compressions that occur when the cosine function is
multiplied by 0.7 and 0.2, respectively. Clearly, a multiplier of 0.2 (right panel) produces
the greatest compression, as the graph of this function drops to 0.2 along the y-axis.

Figure 15.3. An illustration of vertical compressions that occur when the cosine function
is multiplied by 0.7 and 0.2, respectively (adapted from Ninnesse et al., in revision).
Now explain that multiplying the cosine function by a number less than 1 causes
the amplitude of the graph to compress in proportion with the multiplier (for example,
multiplying by 0.5 reduces the high and low points of the graph from ±1 to ± 0.5). Then
employ the same type of training procedure used in step 1 to train cosine amplitude
reductions. To test a formula-graph (A-C) relation with the graph of y = cos(x), type y
= 0.5*cos(x) into the text box, and ask the learner to construct the corresponding graph
using the graphing anchors. A correct response is illustrated in figure 15.2 above.
To assess graph-formula (C-A) relations, enter the formula y = 0.5*cos(x) into “Input
Hidden” and click “graph.” Again, instruct the learner to enter the formula that corre-
sponds to the graph into one of the vacant text boxes and to click “graph” to obtain a visual
comparison between sample and response formulas. Note that the difference between the
graphs of functions involving fractional (less than 1) multipliers can sometimes be imper-
ceptible; for example, y = 0.6*cos(x) vs. y = 0.5*cos(x). So this requires a finer degree of
discrimination from learners. Thus, learner-generated responses to graph-formula relations
of this type are counted as accurate if they are within 0.1 of the actual formula.
In the event that the learner errs during the assessment of either formula-graph or
graph-formula relations, the formula-graph training protocol for cosine vertical compres-
sion should be repeated, and another assessment of both relations should be conducted.
Retraining is always limited to reexposing the learner to the particular formula-to-graph
relation at hand.
Step 3: Cosine frequency relations with numbers greater than 1. Now explain that
when the argument (the variable inside the parentheses) of the cosine function is multi-
plied by a number greater than 1, the graph compresses along the x-axis, consistent with
the value of that number. For example, multiplying the argument by 2 causes the graph
to compress by a factor of 2 along the x-axis (the curve becomes twice as frequent and
its period becomes half as wide). With the graph of the basic cosine function on screen,
demonstrate this by inserting a 2 into the argument of the formula in the text box: y =
cos(2*x). When the “graph” button is selected, the graph compresses by a factor of 2.
Chapter 15 321

To assess formula-graph relations, enter the formula y = cos(2*x) and ask the learner
to construct a graph using the graphing anchors. As in previous steps, the graph of y =
cos(x) is already displayed, so the learner must use the graphing anchors to arrange the
graph of the new formula. A correct response is illustrated in figure 15.4.
Figure 15.4. A correct response to a test of A-C relations addressing compression along the
x-axis (adapted from Ninness et al., in revision).
To assess graph-formula relations, type the same formula, y = cos(2*x), into “Input
Hidden.” Again, the graph will be displayed but the formula remains hidden. Now ask the
learner to input the formula needed to produce the graph, and click “graph” to compare
the sample and response graphs.

Step 4: Cosine frequency relations with numbers less than 1. To address cosine
transformations, explain that multiplying the argument of this function by a fraction of
1 transforms the graph horizontally to stretch along the x-axis. To wit, if the argument
is multiplied by 0.5, in other words, y = cos(0.5*x), it transforms to half the frequency
and twice the width. To demonstrate this transformation, rearrange the formula by mul-
tiplying the argument by: 0.5 y = cos(0.5*x). When you click “graph,” the graph stretches
horizontally (becoming half as frequent and twice as wide in the transformation of its
period along the y-axis). To assess formula-graph relations, type the formula y = cos(0.5x)
into “Input Hidden” and ask the learner to produce the same graph with the graphing
anchors. A correct response is illustrated in figure 15.5.
Figure 15.5. A correct response to a test of A-C relations addressing a horizontal stretch
(adapted from Ninness et al., in revision).
To assess graph-formula relations, input the formula y = cos(0.5*x) and ask the learner
to produce the formula of the graphed function. Again, clicking “graph” will enable the
learner to verify whether the correspondence is correct.
Chapter 15 323

Step 5: Secant amplitude relations with numbers greater than 1. Because the secant
function is the reciprocal of the cosine function, its graph appears almost like a reflection
of the cosine function. In terms of amplitude and frequency, these two functions (cosine
and secant) transform analogously. Thus, the trigonometric rules you provide here are very
similar to those for training vertical and horizontal transformations of the cosine function.
First explain that multiplying the secant function by a number greater than 1 makes its
graph stretch (move up and down) along the y-axis, consistent with the value of the mul-
tiplier. Similar to the cosine function, multiplying the secant function by 3 stretches each
point vertically to three times its original distance from the x-axis. Note that the easiest
points to watch are the high and low points. The high points will stretch from 1 to 3, and
the low points will stretch from -1 to -3. To demonstrate the operation of this rule, change
the secant formula by multiplying it by 3 in the lower left text field: y = 3*cos(x).
When you click “graph,” the graph transforms by stretching between -3 and +3 on
the y-axis. Using the graphing anchors, construct a superimposed graph of this function
directly over the function and ask the learner to perform the same maneuver. To assess
an A-C relation on the vertical stretch of the secant function, change the formula in the
text box to y = 3*sec(x) and ask the learner to construct the graph of the formula. It is
important to emphasize at this point that the graph of the basic secant function y = sec(x)
should remain on screen. To produce a graph using the anchors, the learner must drag the
anchors to the necessary locations to produce a transformed graph in accordance with the
formula, in other words, y = 3*sec(x).
To assess graph-formula relations, type the same or a very similar formula (multipli-
ers should be confined to a range between 2 and 5) into “Input Hidden.” Again, when
you click “graph” the graph appears, but the learner cannot see the formula responsible
for producing the graph. Now prompt a learner response with a statement such as “Please
type the formula needed to produce the graph in the far left text box.”
When assessing formula-graph and graph-formula relations, the multipliers are kept
as whole numbers ranging between 2 and 5, and a correct response requires a formula
that is an exact match in whole numbers, in other words, y = 4*sec(x); see figure 15.6. If
the learner errs during formula-graph or graph-formula relations, repeat the A-C training
for secant immediately and assess the relations again. Note that the learner should not be
reexposed to the training of previous trigonometric relations.
Step 6: Secant amplitude relations with multipliers less than 1. Advise the learner
that multiplying the secant function by a number less than 1 creates a horizontally com-
pressed graph, where the extent of the compression depends on the particular decimal
multiplier employed. The general rule is that the smaller the decimal multiplier, the greater
the compression along the y-axis. For example, a compression that occurs when multi-
plying this function by 0.7 (in other words, y = 0.7*sec(x)) will be considerably less than
compression when multiplying the function by 0.2 (in other words, y = 0.2*sec(x)). Figure
15.7 illustrates compressions that occur when the secant function is multiplied by 0.7 and
0.2, respectively. Clearly, a multiplier of 0.2 (right panel) produces the greatest compres-
sion, as the graph of this function drops to 0.2 along the y-axis. Note that, depending on
the processing speed of a particular computer, the secant function argument multiplied
by numbers less than 0.5 may generate a graph that does not completely reach the upper
limits of the coordinate axes displayed. However, on most computers, and on most graph-

Figure 15.6. An illustration of amplitude changes that occur when the secant function is
multiplied by 3 and 0.5, respectively (adapted from Ninnesse et al., in revision).
ing procedures, we have not found this to be an impediment. As mentioned earlier, in the
actual training and testing of secant amplitude transformation procedures, the instruc-
tions provided to the learner are the same as training and testing the cosine functions as
they address amplitude transformations with multipliers less than 1.
To test formula-graph relations, enter the formula y = 0.5*sec(x) into the text box and
have the learner to construct a graph. The right panel of figure 15.8 illustrates an accu-
rately constructed graph of y = 0.5 sec(x). To assess graph-to-formula relations, type the
sample or a similar formula into the lower right text box and produce a graph. Point to a
vacant text box (top center) with the mouse arrow, and ask the learner to type the formula
that matches the graph. To assess formula-graph and graph-formula relations, set the mul-
tiplier at 0.5. An accurate response requires that the learner type in a formula containing
a multiplier that is correct within plus or minus .1. Typing the formula into the text box
and clicking “graph” enables the learner to see if the response matches the graph.
Step 7: Secant frequency relations with multipliers greater than 1. Provide direc-
tions that are directly analogous to the training of frequency and amplitude transforma-
tions pertaining to the cosine function. That is, in the same way that the cosine function
changes in frequency when multiplying its argument by a value greater than 1, multi-
plying the argument of a secant function by any number larger than 1 transforms the
graph of secant function such that it compresses along the x-axis. Again, analogous to the
cosine function, the extent to which this function compresses is the same as the multiplier
of function’s argument. To illustrate, consider what happens when the argument of the
secant function is multiplied by the number 2: y = sec(2*x). Just as with the cosine func-
tion (its reciprocal), the graph transforms (by compressing) in such a way that its period
becomes exactly twice as frequent but half as wide. (Note: Prior to transformation, a
period for all functions in these examples is 2π along the x-axis.) The left panel of figure
Chapter 15 325

15.8 illustrates the transformation from y = sec(x) to y = sec(2*x). Similarly, if the argu-
ment of a secant function were multiplied by 3 rather than 2, the graph would transform
so that its period becomes three times as frequent but one-third as wide.
Figure 15.7. An illustration of the compressions that occur when the secant function is
multiplied by numbers less than 1 (adapted from Ninnesse et al., in revision).
Figure 15.8. A graphical representation the transformations that occur when the argument
of the secant function is multiplied by numbers greater than and less than 1 (adapted
from Ninness et al., in revision).

Pilot research in our laboratory suggests that multiplying the argument by 2 appears
to operate particularly well as an exemplar of this phenomenon. Thus, at this point you
should modify the formula from y = sec(x) to y = sec(2*x). When you click “graph,” the
graph of the secant function compresses horizontally and its period becomes twice as
frequent but half as wide as it was in the original equation, y = sec(x).
To assess formula-graph relations, type the formula y = sec(2*x), point to the formula
with the mouse, and ask the learner to construct a graph of the formula using the anchors.
A correct response is illustrated in figure 15.8 above.
To assess graph-formula relations, type the formula into “Input Hidden” and, when
the graph appears, ask the learner to type the matching formula in a vacant text box. The
learner must again click “graph” to reveal whether the formula was accurate.
Step 8: Secant frequency relations with numbers less than 1. In the final step,
provide the learner with another analogy relative to the cosine function. Just as the cosine
function stretches horizontally when its argument is multiplied by a number less than 1,
the secant function stretches in exactly the same way (in other words, the graph of this
function stretches in accordance with the value of the multiplier). Pilot research suggests
that using decimal values of 0.5 are particularly easy for learners during practice sessions.
Thus, you should simply instruct the learner that “if the argument of the secant func-
tion were multiplied by 0.5, in other words, y = sec(0.5*x), its graph would become half
as frequent but its period twice as wide relative to the equation y = sec(x)” (Ninness et
al., in revision). As a demonstration of this operation, retype the formula of the graph by
multiplying the argument by 0.5 in the lower left text box, and click “graph” to illustrate
the transformation of the graph.
To assess formula-graph relations, type the formula y = sec(0.5x) into the text box,
point to the formula with the mouse, and ask the learner to construct a graph of the
formula. The graph of the basic secant function y = sec(x) is displayed on screen. The
learner must now move the graphing anchors to construct a transformation of the graph
in accordance with the newly displayed formula y = sec(0.5*x).
To assess graph-formula relations, enter the formula y = sec(0.5*x) into “Input
Hidden,” point to the vacant top center text box, and prompt the learner to type the
formula required to generate the graph of this function. The learner must then click
“graph” to verify that the formula produced a graph that matches yours. Again, errors
result in reexposure to training, plus another assessment of the formula-to-graph and
graph-to-formula relations.
A Learner’s Data Set

The required number of training exposures to each of the trained A-C relations has
varied across the learners who have been exposed to the online protocol thus far. In the
following sections, we present the data from one learner who required a total of eleven
exposures to attain mastery on the construction of eight cosine and secant amplitude and
frequency functions. On her first attempt, the learner failed to construct the graph of y =
3*cos(x) and was reexposed to training of graph amplitude transformations (A1-C1). She
Chapter 15 327

then passed her next attempt to construct the graph and correctly typed the formula when
given the graph as a sample stimulus. On the next trained formula-to-graph relation, she
successfully constructed the graph of y = 0.5*cos(x) (A2-B2) but failed to type the correct
formula when given its graph as a sample (B2-A2). Hence, she was reexposed to training
of the graph amplitude transformations as they pertain to multipliers less than 1. This
successfully improved her performance. Subsequent to the training of graph constructions
for each function, she accurately constructed graphs and typed the appropriate formulas
for y = cos(2*x), y = cos(0.5*x), y = 3*sec(x), y = 0.5*sec(x), y = sec(2*x), and y = sec(0.5*x),
respectively. Although she constructed the graph of y = sec(0.5x) on her first attempt,
she failed to produce the correct formula (within plus or minus .1). After reexposure to
training of formula-to-graph (A8-C8) relations, she constructed the graph correctly and
entered the correct formula when given the graph of y = sec(0.5x) as a sample.
It is noteworthy that throughout the training and testing of A-C and C-A ampli-
tude and frequency relations, we consistently use the same formula for the assessment of
trained A-C and C-A relations. That is, if we train amplitude transformation using the
formula y = 3*cos(x), we use the same formula to assess the C-A relations. Nevertheless,
the sample learner described above, and several other learners in subsequent studies, occa-
sionally failed to identify these relations. This may have occurred in part because we
explicitly told them that the multipliers within these formulas might change slightly in
the assessment of A-C and C-A relations. This minor deception was deemed necessary to
reduce the likelihood of learners responding to C-A assessments simply on the basis of the
last formula displayed on screen.
Online Data Sheet for Taking Reliability Measures

Although we did not attempt to obtain reliability measures on the performance of
the learner in this protocol, we have developed a simple online tool to allow concur-
rent observers to obtain such measures while using the software at www.faculty.sfasu.edu
/ninnessherbe/graphCalcCN07.html Specifically, in the right panel of the screen, we have
placed a white text box. In the lower right corner of this text box, “Input Hidden” is
displayed. If a second observer (with a second computer) has access to the same website
during an experiment, he or she can collect concurrent data by entering hidden characters
in this text box following each response performed by the learner. For example, entering
the digit 1 might be used to indicate a correct response, while the digit 0 could indicate
an incorrect response. Note that it is very important to mouse-click the bottom right
corner of the data sheet (on the words “Input Hidden”) just prior to entering each 1 or 0.
If you do not, data points will not be recorded in this location, and the loss of data will
not be apparent until the training session is complete.
During a learner’s interactions with the software, both the teacher and the second
observer can enter their respective determinations of the learner’s accuracy. Although
these recordings will not be visible to anyone else during the training, the teacher and
observer can then each drag their respective mouse over their text boxes to highlight and
reveal the data recorded by each other. Agreement measures can then be obtained by
comparing the highlighted digits on each screen (see figure 15.9).

Figure 15.9. Within the far right “Input Hidden” text box, 1 may be used to indicate a
correct response, and 0 an incorrect response (adapted from Ninness et al., in revision).
Concluding Comments
Having established frames of coordination and frames of opposition (in other words, reci-
procity) as they pertain to the training of two four-member relational networks, learners
were tested on a wide range of novel formula-to-graph relations, including complex recip-
rocal functions, in conjunction with diversified amplitude and frequency transformations.
All pilot participants performed extremely well on these tests even when they involved
particularly challenging and novel formula-to-graph relations.
Given the growing number of behavioral and other researchers (including Ensley
& Crawley, 2006) who have contributed to the development of modern computer lan-
guages (such as Visual C#, Visual C++, Visual Basic.Net, and ActionScript), and given
the quickly evolving basic and applied technology in derived stimulus relations, we are
hopeful that our studies may contribute to a technology of teaching (Skinner, 1966).
Derived stimulus relations training is no longer a prototype for innovative training; it
exists as a realized set of behavioral or pedagogical strategies and has been successfully
employed in a multitude of interventions where other types of treatment options have
been found wanting. Increased interest in this branch of behavior analysis has blossomed
into countless publications in peer-reviewed journals and an exponential expansion of
pragmatic implementations. Only recently has derived stimulus relations been associated
with online software engineering technology, but the advantages of blending Web-based
software and derived stimulus relations procedures are too beneficial to ignore. Our goal
has been to provide a foundation from which other protocols might be constructed. In
Chapter 15 329

this chapter, we have attempted to offer instructors of basic mathematics a palette of tools
that may contribute to the development of more sophisticated approaches to computer-
interactive instructional systems.
As a closing note, we would like to make it clear that we are not specifically opposed
to the notion of having learners explore their own conceptions or assumptions regard-
ing mathematics—quite the contrary. We fully support and endorse learner attempts to
discover abstract ideas at every academic enterprise. We advocate any form of flexible
pedagogy that includes a systematic and pragmatic appeal to knowledge. Indeed, much
of the online software we have developed and described herein provides a platform on
which learners might independently investigate the behavior of a wide range of basic and
advanced mathematical functions. We do not want to limit instruction (or any form of
study) to drill and practice, reliance on Web-based technology, or training solely based
on derived stimulus relations. Ours is an appeal to U.S. educators to behave accordingly.
We would ask educators to open themselves to the increasingly obvious fact that there
are points in the development of every learner’s repertoire that must include rehearsal of
foundational and fundamental concepts. We are hopeful that a pedagogy that includes
derived stimulus relations might suggest a more efficient technology for such rehearsal.

Appendix
Pretraining for Math Relations
Note: Adapted from Ninness et al., in revision.
Chapter 15 331

References
Davis, R., Maher, C., & Noddings, N. (1990). Introduction: Constructivist views on
the teaching and learning of mathematics. In R. Davis, C. Maher, & N. Noddings
(Eds.), Constructivist views on the teaching and learning of mathematics. Reston, VA:
National Council of Teachers of Mathematics.
EnabLearning. (2002). EnabLearning, Inc. partners with Mass community colleges in
ground-breaking math initiative. Retrieved April 17, 2003 from www.enablearning
.com/AboutUs/FIPSE.html.
Ensley, D. E., & Crawley, J. W. (2006). Discrete mathematics: Mathematical reasoning and
proof with puzzles, patterns, and games, student solutions manual. New Jersey: Wiley
and Sons.
Klein, D. (2005). The state of state math standards. Thomas B. Fordham Foundation.
Retrieved August 2005 from www.edexcelence.net/foundation/publication/publication
.cfm?id=352.
Leader, G., & Barnes-Holmes, D. (2001). Establishing fractional-decimal equivalence
using a respondent-type training procedure. Psychological Record, 51, 151–165.
Lemke, M., Sen, A., Pahlke, E., Partelow, L., Miller, D., Williams, T., Kastberg, D., &
Jocelyn, L. (2004). International outcomes of learning in mathematics literacy and
problem solving: PISA 2003 results from the US perspective. Washington: NCES.

Lynch, D. C., & Cuvo, A. J. (1995). Stimulus equivalence instruction of fraction-decimal
Mayfield, K. H., & Chase, P. N. (2002). The effect of cumulative practice on mathemat-
ics problem solving. Journal of Applied Behavior Analysis, 35, 105–123.
Ninness, C., Barnes-Holmes, D., Rumph, R., McCuller, G., Ford, A. Payne, R., et al.
(2006). Transformation of mathematical and stimulus functions. Journal of Applied
Ninness, C., Dixon, M., Rumph, R., McCuller, G., Holland, J., Smith, R., et al. (in revi-
sion). Deriving and constructing reciprocal trigonometric relations. Journal of Applied
Behavior Analysis.
Ninness, C., Rumph, R., McCuller, G. Harrison, C., Ford, A., & Ninness, S. K. (2005). A
functional analytic approach to computer-interactive mathematics. Journal of Applied
Peters, K. (2005). Foundation actionscript animation: Making things move! New York:
Springer-Verlag New York.
Programme for International Student Assessment. (2003). Learning for tomorrow’s world:
First results from PISA 2003. Retrieved December 22, 2005 from www.pisa.oecd.org
/dataoecd/1/60/34002216.pdf.
Reinhardt, R., & Dowd, S. (2007). Adobe flash CS3 professional bible. Indianapolis: Wiley
Publishing.
Rumph, R., Ninness, C., McCuller, G., Holland, J., Ward, T., & Wilbourn, T. (2007).
The shame of American education redux. Behavior and Social Issues, 16, 47–49.
Sidman, M., & Tailby, W. (1982). Conditional discrimination vs. matching to sample: an
37, 5–22.
Skinner, B. F. (1966). An operant analysis of problem solving. In B. Klienmuntz (Ed.),
Problem solving: Research, method, and theory. New York: John Wiley.
Chapter 15 333

Chapter 16
Developing Self-Directed Rules
Carmen Luciano, University of Almeria, Spain;

Sonsoles Valdivia-Salas, University of Albany, New York;
Francisco Cabello-Luque, University of Murcia, Spain;
and Monica Hernandez, University of Jaén, Spain
Ryan is an eight-year-old boy who, every night before going to bed, remembers he has to
brush his teeth, then goes into the bathroom and brushes his teeth. Sophia is a thirteen-
year-old girl who, while looking at her photo album, realizes her friend’s birthday is in two
days’ time; she thinks, “I’ll ask Mom to take me to the mall and we’ll buy a nice present
for my friend,” and subsequently she does it. Sally is feeling bad because her grandfather
has had an accident, so she asks her father to drive her to the hospital so she can visit him.
These are just three examples of the many rules that we, as verbally sophisticated human
beings, generate and respond to from morning to night. Many of these events or types of
psychological content function as self-rules that control our behavior in extremely useful
ways (for example, in problem solving, social skills, emotional regulation, and so on). But
sometimes our self-content has a darker side that may or may not control our behavior in
a manner that is favorable to us. Consider the examples below.
Mike is an eight-year-old who can verbalize “I will watch my favorite hockey team
playing in a couple of hours” but he doesn’t do it unless his parents remind him that
the game is about to start. Troy is a thirteen-year-old who says to himself, “He thinks
I am stupid,” every time one of his classmates looks at him. He subsequently reacts by
attacking anyone found to be looking at him, which in turn jeopardizes his academic
achievements and social life. And Maria is a fifteen-year-old who describes herself as a
chocolate lover but is diagnosed with diabetes. She is able to stop eating most candy but
not chocolate (“because I’m a chocolate lover”); her frequent hospital visits indicate that
chocolate is ruining her health.
The current chapter focuses primarily on the relational frame theory (RFT) approach
to understanding rules, particularly self-rules, and their control over behavior (Hayes,
Barnes-Holmes, & Roche, 2001). In the first part of the chapter, we present a general
overview of traditional approaches to the emergence and training of self-rules. In the
second part, the basic principles involved in the emergence of self- and rule-governed
behavior as relational learning will be described. Finally, in the third part, we will provide
a series of guidelines for the training and understanding of self-rules and discriminating
their role in behavior.
Traditional Approaches to Training Self-Rules

In mainstream psychological terms, self-rules or self-instructions are loosely described as
beliefs and expectancies that influence our behavior; they are believed to correlate with
competencies in a range of areas, including academic, cognitive, and emotional. As a
result, numerous training regimes employed with developmentally delayed learners incor-
porate programs for establishing sound rule-following and the generation of, and control
by, self-rules (such as self-management protocols).
Self-Instructional Training
In traditional approaches, self-rules are subsumed within the broader concept of meta-
cognition. According to theory of mind, for example, deficits in generating and following
self-rules are first targeted by the establishment of overt rules that facilitate task comple-
tion, which, it is hoped, will become covert and internalized (Baron-Cohen, 1995). Self-
instructional training is a widely used teaching regime in this area, originally developed
for use with children with impulse-control difficulties and attention-deficit/hyperactivity
disorder (ADHD; Meichenbaum, 1985). Self-instructional training comprises the follow-
ing five procedural steps: (1) the teacher performs a task while talking aloud; (2) the
learner then performs the same task following the teacher’s instructions; (3) the learner
performs the task while instructing himself aloud; (4) the learner whispers the instruc-
tions while performing the task; (5) the learner performs the task while guiding his per-
formance via private speech. Across training sessions, the package of self-instructions is
often enlarged by means of response chaining and successive approximation procedures;
additions such as posters and notes attached to walls are incorporated to prompt self-talk
and generate self-questions (Manning & Payne, 1996). Furthermore, other metacogni-
tive strategies have been used to facilitate generalization, like Socratic dialog (Schleser,
Cohen, Meyers, & Rodick, 1984). On the whole, self-instructional procedures seem to be
generally effective in the short term, but little is known about the conditions under which
they do or do not work, or how generalization to new situations occurs (Ayllon & Milan,
1996; Friedling & O’Leary, 1979; Wicks-Nelson & Israel, 1997).

Contingency-Shaped Behavioral Training
The training of self-rules within the behavioral tradition has focused on contrasts
between immediate versus delayed reinforcers, on the assumption that self-rules frequently
involve reinforcers that are delivered over the medium to long term, rather than the short
term (Grant & Evans, 1994). The concordance between self-rules and subsequent behav-
ior has also been targeted with regimes such as say-do correspondence (Risley & Hart,
1968; Paniagua, 1997; see review in Herruzo & Luciano, 1994).
Behavioral researchers, however, working under the rubric of relational frame theory
have highlighted the verbal nature of rule-governed behavior and have argued that it is
distinct from behavior that is shaped by contingency. From this perspective, therefore,
various core verbal processes regarding the formation of the self and complex repertoires
of derived relational responding must be established in order for self-rules to be gener-
ated and followed in a manner that is favorable to the individual. In the next section, we
describe the RFT account of the development of the self that is necessary for effective
self-rules to guide behavior. In the section thereafter, we articulate the RFT account
of different types of rule following. For didactic purposes, the two aspects will be pre-
sented separately, but they normally develop in a parallel manner, like pieces of the same
puzzle.
Relational Frame Theory and the Self

Self-rules have very humble beginnings in the most basic relationships between
public and private events. Indeed, Skinner (1945) described four key ways in which the
verbal or social community teaches knowledge and naming of private events: (1) the
correlation between a public event (for example, an accidental cut from a knife) and
the assumed private stimulation (pain); (2) the occurrence of a public response (crying,
grimacing); (3) the correlation between a publicly trained verbal discrimination and
the assumed private stimulation; and (4) stimulus induction or metaphorical extension.
However, the bidirectional and relational characteristics of language should be added to
have a complete picture of what is being learned about the formation of private events
(Barnes-Holmes, Hayes, & Dymond, 2001). For RFT, private and public events merge
into coherent relational networks composed of one’s understanding of oneself and one’s
understanding of others that become pivotal in generating one rule over another and in
determining which rules will or will not be followed. According to this view, there are
three levels in the emergence of self that form an essential backdrop to the development
of rule competencies. These are referred to loosely as the three selves: self-as-content,
self-as-process, and self-as-context. Before describing each of these below, it is important
to emphasize that the three selves appear to develop in tandem, rather than sequen-
tially; problems during the formation of any one of the selves limit the emergence of
self-knowledge and self-rules; and these developments and their control over behavior
depend upon the type of verbal regulation dominant in the learner’s history (this will be
addressed later in the chapter).
Chapter 16 337

Self-as-Content
Self-as-content (also known as the conceptualized self ) comprises all the things a person
says about herself to others and to herself. These evaluations, descriptions, and explana-
tions of behavior are first presented to us (instructionally and by contingencies) by others
and are then progressively derived by ourselves. Imagine, for example, a young child who
is accidentally cut by a knife. The coordination relations between the word “pain” and the
painful sensations, as well as her parents’ explicit attempts to connect the knife holding
with the pain and the wound (for example, “Look what that bad knife did to you”),
enable the child to derive the self-rule “If I don’t want to have pain, then I shouldn’t hold
knives.” But such an experience can do much more than teach the child not to play with
knives, because even these simple experiences can result in an almost endless stream of
related derivations, if the child has developed repertoires of derived relational respond-
ing. For example, all new experiences of pain will likely be compared with the knife cut,
and the absence of pain will be in a relation of opposition with it. Coordination between
pain and evaluations of “bad” will be strengthened. Aversive reactions may extend to all
sharp objects. The child’s foolishness may also be coordinated with the mother’s anguish.
In addition, a teacher or parent might provide analogies or fables to alter the functions
of a particular sensation or thought so as to lead the child to behave in a proactive way.
For example, if the child is feeling sad because her friends are playing while she is doing
homework, her mother might provide a fable in which doing a task now might involve
discomfort but is worth it in the long run.
All of this additional psychological content becomes part of the emerging self—one’s
story—as it coordinates with developments in the perspective-taking frames. For example,
imagine a six-year-old boy recently diagnosed with diabetes who hears his older sister say
that “people with diabetes can’t do the same things as others.” From this comment, the
child may derive “I am different from others,” “I am not like my sister,” “I won’t be able
to do the same things as my friends,” and even “I will have no friends.” None of this self-
relevant content could occur if the child were not able to engage in hierarchical relating,
perspective taking (I versus you), comparing, and temporal relations. But when he can,
derivations like these are almost relentless features of his daily life.
Because of the history and ongoing nature of the relationship between public and
private events, self-as-content constantly reflects the output (reasons, evaluations, and so
on) of the social environment and the contingencies operated by it. Hence, coherence
between this and the emerging self increases and continues to be important (for example,
through being right or making sense to oneself). It is worth noting here that to develop
content about oneself also requires additional interactions to prevent attachment to this
self-content. These issues are discussed in the section below.
Self-as-Process
Self-as-process (often called the knowing self ) comprises discriminations of what we are
doing, feeling, thinking, remembering, touching, seeing, and the like. It is not only discrim-
inating between doing and feeling, but also verbally discriminating across a set of actions,
feelings, thoughts, and so on, in order to abstract a sense of self that permits being aware
of private events as they occur in any given moment. Like self-as-content, self-as-process
results from the verbal community establishing relationships and discussing what is being

done, who is doing it, and who is noticing that it is being done, through multiple-exemplar
training across a myriad of experiences and different questions. These experiences include:
different feelings when they are occurring (asking what is felt, who is feeling, and who is
noticing the feeling); thoughts (asking what is thought, who is thinking, and who is notic-
ing thinking); observations (asking what is seen, who is seeing, who is noticing what is seen;
actions (asking what is being done, who is doing it, who is noticing what is being done);
and so on (Hayes, 1984; Luciano, 1996). These multiple discriminations involving multiple
experiences about one’s and another’s behavior are the basis upon which the abstraction
of the only commonality (the “I”) across interactions emerges, such that “I” becomes the
context for many actions, feelings, reasons, and thoughts. In addition, the development of
temporal relations (before-after: “I lost my teddy bear before feeling sad”) and conditional
relations (If-then: “In order not to feel sad, I will take better care of my other teddy bear”)
begins to provide the child with explanations of his and other’s behavior, including his own
private events, and guidelines on how to behave effectively in the future. In other words,
the learner comes to discriminate between “I” and what is being experienced, had, felt,
remembered, thought, or reasoned about, such that “I” becomes experientially located here
and now while the things experienced are located there and then.
Self-as-Context
Self-as-context alters the functions of both self-as-content and self-as-process, because
of the constancy of “I” as separate from all of the person’s experiences, and her ability
to reframe what has happened to her in a manner that is consistent with other public
and private events. In other words, coherence becomes central, and this coherence is itself
what is meant by self-as-context. Self-as-context requires that some developments in self-as-
content and self-as-process have already occurred, but self-as-context also functions as an
ongoing source of interpretation and valued guide to the processes in self-as-content and
self-as-process. In addition to the questions already mentioned (see “Self-as-Process,” above)
are those guiding the interpretation and values—for example, why are you doing this,
what happens if you do this, what do you like, what do you want to do, and so on. Hence,
the three selves are in constant interaction, primarily under the governance and watchful
eye of self-as-context perspective. This perspective does not change once established and
is the abstraction that results from the discrimination of all the experiences occurring in
the person who is named as “me” (that is, establishing a hierarchical relation between the
self-as-context, the self-as-process, and the self-as-content). This parallel formation of the
three levels of the self provides the necessary basis of self-knowledge that facilitates the
emergence of increasingly complex self-rules, as well as guidance regarding whether those
rules should or should not be followed. Such guidance is dependent on the learner’s rule-
following history, establishing, through thousands of interactions, the learner’s values.
Relational Frame Theory and Rule-Governed Behavior

Skinner (1969) defined rule-governed behavior as behavior controlled by rules derived
from contingencies in the form of descriptions specifying occasions, responses, and conse-
quences. Other authors have regarded rules as verbal antecedents (Zettle & Hayes, 1982),
Chapter 16 339

or as contingency-specifying stimuli with function-altering effects (Schlinger & Blakely,
1987). What is common across interpretations is the belief that rules (even those that are
unspoken and said only to the self) exert a significant influence over behavior, and that
this influence can occur even in the absence of apparent contingencies for rule following.
From an applied perspective, what makes it perhaps most difficult to understand is the
fact that, topographically, rule-governed behavior looks identical to contingency-shaped
behavior, and yet they reflect different behavioral processes and each requires training
through different types of intervention. Put simply, the trick is to determine correctly
which it is that you see and in which areas the learner’s competencies and deficits lie. For
example, imagine that Laura’s older brother tells her that the best way to resolve problems
is to ask for the solution. As a result, Laura has some success with resolving her problems,
but she does not make contact with the differential contingencies that would be attached
to the range of possible solutions. In contrast, imagine Miguel, whose brother helps him
to think about the solutions and their contingencies as a process for confronting problems
and finding solutions. For Miguel, although his behavior is rule-governed it will also be
contingency sensitive; for Laura, it will not.
In order to understand rules and self-rules, we first need to explain why a particu-
lar combination of words, facial expression, tone of voice, and so on makes sense or is
understood and then comes to function as a rule that controls behavior. Second, we have
to identify the conditions that give rise to self-rules. And third, we have to identify the
conditions that give rise to the following of self-rules—otherwise known as verbal regula-
tion. Both rule understanding and verbal regulation occur early in a child’s development
and appear to develop in parallel ways.
According to RFT, events are understood because they participate in relational
frames. For example, imagine that Mary is told that when her grandmother says “recreo”
at home it means that ice cream is being served. But when her grandmother says “recreo”
at the school where she teaches, it means that the children can go outside to play. Also,
imagine that Mary is told that her grandmother has another word for “recreo,” which
is “descanso.” Given such a history, it is likely that Mary will go to the kitchen to get
ice cream when she hears either “descanso” or “recreo,” but when she hears the latter at
school her behavior will be very different—she will stop working and go outside to play.
For RFT, Mary is able to have these different (and sometimes similar) reactions to dif-
ferent words only after a history of learning coordination relations, as well as acquiring
the temporal and conditional relations necessary for the appropriate stimulus functions
to transform the behavior (see Barnes-Holmes, O’Hora, et al., 2001; O’Hora & Barnes-
Holmes, 2004). In other words, there is a long history of relational competencies that is
necessary to acquire before there will be any type of sophistication in rule understanding
or rule following, and any attempts to establish rule following without this history (or
with deficits in the three selves) would be largely futile.
But a rule can be understood yet not followed, and RFT stipulates four key reasons
why this occurs. First, the response described in the rule may not be in the person’s rep-
ertoire. For example, you may understand how to produce a particular piece by Mozart,
but you could not do it unless you received appropriate training. Second, the rule may
be provided by someone with little credibility from the perspective of the listener. For
example, a mother’s rule will likely have more impact on a child than a stranger’s (hence
different functions are transformed). Third, reinforcement for rule following may not be

available. For example, a child will follow her mother’s rule but not her sister’s because all
reinforcers are delivered by the former. Fourth, a rule may be poorly elaborated because its
relational network is contradictory or incoherent with the listener’s history; for example,
consider the following rule: “If you want to have loads of friends, go around insulting and
hitting all your peers.” These contextual features provide an important understanding of
the conditions that determine, at least to some extent, why some rules are not followed.
Verbal Regulation: Pliance, Tracking, and Augmenting

Three functional and sequential categories of rule following have been identified by
RFT—pliance (the simplest type of rule following), tracking (a more complex type of rule
following), and augmenting (the most complex type of rule following). Each of these is
articulated in a separate section below (see also Hayes, Zettle, & Rosenfarb, 1989; Hayes
et al., 2001).
Pliance
Early in the socialization process, children are taught to act in accordance with adult
demands, and reinforcement is often contingent upon doing so. For RFT, pliance is the
most fundamental unit of rule-governed behavior and is built up from these early contin-
gencies. It occurs under the control of socially mediated consequences for the correspon-
dence between the rule (often known as the “ply”) and the subsequent behavior. If a child
follows her parent’s rule about going to bed at bedtime because of a history of exemplars
of social reinforcement for compliance (and/or punishment for noncompliance), then we
would call this type of rule following “pliance.” Furthermore, the speaker’s behavior—the
rule—is referred to as a vocal mand that is reinforced by the child’s compliance.
Establishing pliance is important for a number of social reasons. First, it establishes
necessary control over ongoing contingencies (for example, when the instruction “Stop it”
overrides the natural consequences of behavior currently underway). Second, it provides
direct consequences for the speaker’s behavior (for example, if the behavior of a child is
aversive, the mother mands, “Don’t do that!” and the child complies, then the source
of aversive stimulation is immediately removed). Third, it is often socially adjusted. For
example, if a child fails to follow a ply, pliance contingencies are often intensified (for
example, “I told you, no!”). Indeed, pliance abilities probably also result from a history of
multiple-exemplar say-do correspondence because adults are more likely to reinforce this
type of correspondence behavior (Luciano, Herruzo, & Barnes-Holmes, 2001).
Tracking
Tracking is rule-governed behavior under the control of a history of correspondence
between the rule and the way the world is arranged independently of the delivery of the
rule (as well as the agent delivering the rule—that is why it is differentiated from pliance).
In other words, tracking occurs because of a history of natural consequences for engag-
ing in the behavior specified in the rule. For example, if the teacher says, “Wash your
hands,” and the learner washes his hands to get them clean, the behavior on this occasion
Chapter 16 341

is defined as tracking (rather than pliance), because the contingencies controlling the
behavior are those that naturally follow hand washing (texture, scent, feeling of cleanli-
ness, and so on). In this case, the rule provided in tracking is a tact because it describes
the necessary interaction between the environment and the behavior. For tracking skills
to develop, specified environmental contingencies must be sufficiently predictable across
a range of social contexts. Hence, the development of tracking tends to be more subtle
than that of pliance because tracking adds no new consequences to the existing situation
and cannot be automatically socially adjusted until behavior regulation occurs (Hayes,
Gifford, & Hayes, 1998).
It is important, however, to ensure that the learner does not get trapped by track-
ing that is not effective in the long run, and that he learns that tracking is controlled by
increasingly delayed and probabilistic consequences. This is why competencies in tem-
poral relations are so important—because the learner must be able to construct verbal
futures that contain the predicted contingencies controlling current behavior. Without
this, more-complex types of rule-governed behavior would not be possible.
Augmenting
Augmenting is rule following under the control of consequences whose functions
get augmented or diminished by derived means. For example, the aversive functions that
drinking milk may have for a child could be transformed through the rule “Drinking a
glass of milk every day will help you grow big and strong.” This behavioral regulation
allows a better sensitivity to likely possible outcomes, and it facilitates lengthy behavioral
sequences to be performed with regard to those possible consequences. Consequently,
augmenting plays a critical role in the development of some of the most important aspects
of being human, such as self-awareness, moral behavior, and empathic behavior.
There are two types of augmenting. Formative augmenting is behavior under the
control of newly established relational networks providing consequential functions to
novel stimuli. For instance, a child who is very sensitive to others’ feelings and is told,
“Whenever you hear ‘gracias,’ it will mean that the person who is saying it feels grateful
to you,” may start doing things to maximize the chances of hearing “gracias.” Motivative
augmenting, on the other hand, is behavior under the control of relational networks that
temporarily alter the degree to which previously established consequences function as
effective reinforcers or punishers. For example, the rule “If you do not choose now what
you want for lunch, they will bring something you may not like” may augment the rein-
forcing properties of choosing as a way of escaping from having to eat something that is
disliked.
Consider the self-rule “If I do my homework every day, then I will get an A+ at the
end of the semester.” The remote consequences of getting an A+ are effective at present
through the transformation of stimulus functions enabled by the temporal (“every day,”
“end of semester”) and conditional (“if,” “then”) relations included in the rule. In other
words, getting the best grade is brought into the present in each of the activities related
to doing the homework (getting home from school early in order to start homework,
reading, writing, and the like) through a well-developed repertoire of relating events by
if-then and temporal frames. Furthermore, because getting an A+ is in a frame of opposi-
tion with failing exams, for example, then all activities in a conditional and temporal rela-
tion with failing would automatically lose their reinforcing functions. In the next section,

we present guidelines for establishing self-rules and establishing their appropriate control
over behavior in a manner that provides favorable consequences for action.
Guidelines for the Development of Rule-Following

and Self-Directed Behavior
Because functions are attached to particular self-rules by the verbal community, specific
thoughts, memories, and the like may become aversive or reinforcing on this basis. While
these functions are often consistent and coherent with personal established goals, family
dynamics, and other factors, the two sources may become increasingly incompatible over
time, and sound perspective taking and developments in self will be necessary to enable
the learner to decide which to follow in any given instance. For example, social rules
(such as “No one should have to put up with discomfort”) may be inconsistent with a
family’s religious beliefs (that suffering brings one closer to God, for example), or they
may be incompatible with achieving a particular goal that involves discomfort. In these
cases, when discomfort is experienced, the learner will feel conflict and, depending on
her history with discomfort, will be uncertain of how much to tolerate and what should
be done. Or perhaps a learner comes from parents who do not value education, although
the learner is very strong at mathematics and would like to become a math teacher. Now,
if the learner is struggling with math homework one evening and her parent tells her to
forget about it, then the reinforcing functions that had been given to math (as a career)
override the aversive functions of being challenged and disagreeing with her parents, and
the latter loosen their controlling role over escape behavior. So, the functions of self-
content depend upon personal values established by social whim, as they consequently
involve others and stretch across time, and, most important, change the functions of the
immediate experience. Of course, this makes good sense for two reasons: First, the deriva-
tion of much of our self-content and its emotional function is almost unavoidable and is
highly arbitrary; it is also frequently an unreliable guide to action or predictor of positive
consequences. Second, personal histories are always additive, so individuals have a sound
basis of experience across time and content from which to generate reliable self-rules that
have accurate predictive utility and to reach meaningful goals according to what the
person, as self-as-context, values.
Relationally, the skills necessary for effective self-directed behavior may be summa-
rized in the following components:
Fluency and flexibility in coordination: comparative, temporal, and causal
relational framing that allow behavior to be controlled by the specifi-
cation of nonimmediate and probabilistic valued consequences (whose
reinforcing and/or punishing functions have been acquired by relational
means)
A history of rule following that allows contact with verbally established

consequences, as well as diminishing the relative value of immediate,
small, contacted consequences over delayed, larger, and verbally con-
tacted ones
Chapter 16 343

Fluency and flexibility in perspective taking that allow the learner to dis-
criminate herself from others, and to differentiate herself from her own
private contents or self-rules
In summary, behaving in a proactive way, such that there is constant

comparison between what is happening in the here and now and what is
valued across time
In the section that follows, we provide a series of guidelines for establishing the
three basic types of rule following—pliance, tracking, and augmenting. Subsequently, we
present guidelines for training perspective taking with regard to one’s self-content, based
on existing empirical and clinical evidence.
Training Basic Rule Following: Pliance

Early in life, children are taught to act in accordance with adults’ demands, and this
compliance is a core feature of practically all educational regimes. Indeed, in natural
development there are literally hundreds of interactions that focus on pliance (Hayes
et al., 1998). This history usually contains two key elements of training that establish
pliance: establishing an early understanding of the broader nature of compliance itself as
reinforcing (for example, incorporating feedback for compliance such as “Well done; you
are a nice boy because you did what I told you to do”); and if a child fails to follow a
ply, intensifying the pliance contingencies (for example, “I told you, no!”) to ensure that
pliance occurs.
Evidence from the literature on say-do correspondence suggests the following training
steps for establishing pliance as a generalized operant class (Luciano et al., 2001; Luciano,
Barnes-Holmes, and Barnes-Holmes, 2002):
Use a range of options so that the learner must discriminate the conse-
quences associated with each option.
During responding, ask the learner to articulate what he is doing and

why, as he does it.
Social reinforcement should be delivered contingent on the correspon-

dence between the behavior and the rule. For example, you might say,
“You have done what I told you to do” or “You have done what you said
you were going to do.”
Use the same contextual cue across trials targeting different content (such
as “What do you promise to do?” or “I promise I will do…”) so that the
derivation of new rule-following with novel content is more likely.
As the relational repertoires of the learner gain fluency and flexibility,

more-complex rules should be incorporated. For instance, if temporal
relations are well established, use rules like “First tidy up your toys, and
then you can watch TV.”

With a training history such as that described above, it is likely that the learner will
follow rules (including novel ones) where the probability of socially mediated reinforce-
ment (either positive or negative) is perceived as high (see behavioral procedures in Grant
& Evans, 1994). But there are two important caveats. First, if most instances of pliance
are established as escape responses from socially mediated aversive consequences (in other
words, negative reinforcement), then it is likely that the learner will only follow rules pro-
vided in contexts where the aversive consequences are primary, and where the social agent
for mediating the negative reinforcement is functionally present. Second, learners should
generally be taught how to choose to follow one rule over another, to avoid the forma-
tion of overgeneralized pliance. Doing so avoids mediation by others becoming the main
(or even the only) reinforcer in a learner’s life. Indeed, overgeneralized pliance would
limit the learner’s transition to more complex forms of rule following, which will require
constant adjustment to changing contingencies.
The Next Step: Tracking

Unlike pliance, tracking requires sensitivity to naturally changing contingencies,
because these control the tracking behavior. And if tracking is properly established, the
learner’s behavior will be controlled by increasingly delayed, probabilistic, and abstract
consequences, such as care for the natural environment.
To facilitate the transition from pliance to tracking, emphasize the natural conse-
quences of the behavior corresponding to the rule. For example, say, “Now that you have
tidied your toys up, look how much room we have to play on the floor!” or “Now that you
have brushed your teeth, don’t they feel smoother than before?” In this way, the tracks
include frequent and accurate descriptions of the world containing relatively immediate
and highly probable consequences. Also, if the rules (as tracks) are provided by people in
the environment who have been mediating reinforcement up to that point (as plies), they
will probably be followed. In this way, the transition between pliance and tracking can
become almost seamless.
In order to facilitate tracking in novel situations, the teacher (and eventually the
learner herself) should tact the cause-effect relations between doing what the rule specifies
and the changes that it produces in the environment. Consistent with this, reinforcement
should then be contingent not upon action specified in the rule, but upon the relation
between saying and doing. Consider the following example: Give the learner the choice
“You may sit there, or you can go and wash your hands. Now, which are you going to
choose?” Let’s say that the learner indicates that she chooses to go and wash her hands,
so you provide social reinforcement by saying, “Okay, good choice.” And while the action
is being done, you ask, “What are you doing?” and provide reinforcement for the concor-
dance between the stated intention and the subsequent action “Good girl—you did what
you promised to do.” Then, add a description of the natural consequences as follows: “You
washed your hands and look what happened—they look so clean! Smell them—they
smell so much better now! Now you are all set to have your lunch.” This type of training
in tracking should enable the learner to become sensitive to natural changing contingen-
cies, in such a way that her behavior is constantly adaptive.
Chapter 16 345

Gradually incorporate more and more long-term and probabilistic consequences.
Along with the increasing fluency in relational operants (especially comparative, tempo-
ral, and conditional relations), this will allow the learner to understand more-complex
tracks where the consequences are longer term (for example, “You are just five pages from
finding out what happens to the princess at the end,” “If you keep watering the plant
twice a week you will eventually see a big pink flower,” or “Let’s send the invitations
and see how many people will come to the party”). Conditions such as those specified in
the examples promote control by providing increasingly delayed and probabilistic conse-
quences, which will be necessary for the development of the most complex form of rule
following, augmenting.
The Most Complex Type of Rule Following: Augmenting

In augmenting, the reinforcing or punishing functions of consequences are aug-
mented or diminished via the transformation of functions. Although augmenting natu-
rally becomes a very complex type of rule-following behavior, you can set the path for it
early on in the training history, as soon as you see coordination relations, pliance, and
tracking. For instance, imagine that an English-speaking learner who loves sweet food is
told that “dulce” is the same as “sweet” (a word he knows) and is asked, “Do you want to
eat this? It is dulce.” Then the learner will probably try the food. In this case, the learner’s
behavior is under the control of the augmented functions of “dulce” via its coordina-
tion with “sweet” and a history of tracking. Although augmenting can be very complex
behavior, it does not require the presence of all three selves; but when the three selves
are present (especially self-as-context), then augmenting will have a greater impact on the
regulation of the person’s behavior. This is so because the self-as-context is the abstracted
experiential awareness taking place across many actions, feelings, thoughts, memories,
and so on.
Augmenting occurs as pliance and tracking are progressively linked to highly abstract
verbal consequences whose value is altered via the transformation of functions. Thousands
of derived functions are involved in even the earliest relational frames. Imagine, for
example, that a learner who likes race cars is told, “You need to be good at math to
become an engineer who builds race cars.” In this case, the coordination between math
and designing race cars has augmented the reinforcing functions of learning math.
So, training augmenting involves providing the conditions for derived relational
responding so that novel and abstract events can function as effective consequences, which
will be highly effective in the presence of a well-established pliance and tracking history.
The variability of behavior controlled by augmentals is huge, given that the alteration
of reinforcing or aversive functions will depend upon relations established by the verbal
community. For example, a particular culture might provide augmented value to physics
by establishing it in coordination to designing good cars and helping others drive more
safely. One family may relate effort, patience, attending classes, and doing homework to
future success, while another will relate those events to missed opportunities for having
fun with friends. Or one community might relate elderly people to wisdom and helpful
advice, while another relates them to uselessness, cost, and extra duties.

It is important to note the differential impact of augmenting that results from differ-
ent rule-following repertoires. For example, suppose that a relation of cause and effect is
established between recycling and being admired by significant others. For a learner with
a strong history of pliance based on immediate contingencies, this rule may not function
as an augmental because the contingencies it specifies will only be effective in the mid- to
long-term. In contrast, the rule would function as an effective augmental for learners with
a history of pliance that was based on delayed or intermittent contingencies. Imagine now
that recycling was related to a healthier planet (cleaner water, more forests, cleaner air,
and so on) decades from now. This augmental will not have any impact on the previous
children, only on those with well-established tracking based on delayed and probabilistic
consequences.
Toward Effective Self-Directed Behavior

Up to this point, we should have a learner who can adopt the perspective of “I” as the
locus of all her private contents and processes. She can also behave in accordance with the
transformation of functions characteristic of conditional and temporal relations, because
she discriminates the causes and immediate, delayed, and probabilistic consequences of
her behavior and the behavior of others. Now she has to learn whether behaving in accor-
dance with the literal content of her private events is appropriate in terms of the immedi-
ate and mid- to long-term consequences for following self-rules. Put simply, it is time to
teach the learner to discriminate the relative utility of her private events with regard to
attaining meaningful outcomes. For example, if a learner failed a math test and thought,
“I’d better ask the teacher for help,” because of a desire to remain strong at math (and
other delayed consequences such as being organized and responsible), and acted accord-
ingly, this would be an instance of self-rule following that should be promoted. However,
if the child thought, “I’d better not say anything because I’ll look stupid in front of the
class,” and acted accordingly, then this would be under the control of negative reinforce-
ment that offers immediate emotional benefits (avoid feeling bad for not doing well) but
poor long-term consequences in academic achievements. In the latter case, the learner
should first be trained to discriminate the immediate and delayed consequences of her
behavior, and then be trained to react to such thought in a nonliteral manner (in other
words, keep doing now what is going to provide positive reinforcement later). It is then
time to teach the child to notice her own content, to discriminate how useful it is with
regard to what is valued, and to act in accordance with values rather than the literal or
immediate nature of the content.
Establishing Hierarchical Relations Between the Self and Private

Events and Behaving Effectively
Consider the following intervention for sleep disturbance based on protocols reported
by Luciano, Valdivia-Salas, Gutiérrez, Ruíz, and Páez (in press). Imagine six-year-old
Maria, who says, “When I close my eyes at night I am afraid I will die—I need to have
the light on,” after seeing a dead person with closed eyes on TV. To begin with, her
mother may alter the coordination relation between sleep and death by indicating that
Chapter 16 347

there is much more to dying than closing your eyes. For example, she might have Maria
close her eyes for maybe thirty seconds and then discuss the fact that she did not die or
even begin to die and that her mind and body continued to stay active even while her
eyes were closed.
At some point, it is important to teach the learner that not all private events (thoughts,
sensations, memories, and so on) need to have behavior regulation functions (in other
words, they are only thoughts and you need not do anything about them). Ask, for
instance, “Just imagine you did die. As we think about this, tell me what comes into your
mind.” Every time the learner says a thought, pretend it is a colored piece of wood sitting
in the learner’s lap. For example, say, “You are thinking that you won’t see your friends
anymore—that’s a green piece. Hold it in your lap... Now you are feeling sad. That’s a
yellow piece. Hold it in your lap,” and so on.
Such interactions enable the learner, across exemplars of content, to discriminate
herself as the locus for the content, rather than the actual content. For instance, say, “So,
this is Maria [pointing to the learner], and you have been thinking X [pointing to the
green piece], and Y [pointing to the yellow piece], and Z [pointing to a different color]. It
looks like you are very big because you have all these pieces of wood in your lap and all
these thoughts in your mind.” What you are doing here is highlighting the hierarchical
relation (rather than the coordination relation) between Maria and her content, where
Maria contains the content rather than being the same as the content.
In this type of training, start with simple content or self-rules and then progress
toward content that is more distressing or more gripping emotionally (more likely to be
believed and acted upon directly). For example, “Maria, you are here and what you are
thinking is there. Let’s play games with it. Who is more important, Maria or this thought
[pointing to one of the least frightening ones]?” She may say, “me.” You then ask, “And
who is me?” and prompt with “Maria.” So you say, “Here is Maria—you—and your
thoughts are there; some of them you like, and some you do not like. Who do you want
to be in charge, you, Maria, or any of these thoughts?” Then go through all the thoughts
that Maria brought along and establish that she wants to be in charge.
Then coordinate sleeping with the things Maria enjoys doing (watching a movie,
playing with friends, and so on), or things Maria admires in significant others, in such a
way that sleeping is established as a necessary condition for accomplishing those things.
“So, Maria, in order to be the big girl you want to be, what can you do when you go to
bed and the thought ‘I do not want to close my eyes’ comes?” At this point, let Maria
imagine herself playing with the thought, and if necessary, prompt her, for example, with
“Do you think you can put this thought in your doll’s dress pocket and bring the doll
into bed so that she takes care of your thoughts while you sleep and become a big girl?”
When it is time to go to bed, review some of the issues raised during the training,
like “Who needs to sleep? Why is it important? If the thoughts show up, what can you
do with them? You know that by sleeping you will have more energy tomorrow to play,
watch TV, and do your homework.” If Maria wakes up during the night and asks for
help, her parent or practitioner can prompt her to verbalize what she is going to do with
the thought (placing it in her doll’s pocket). If she asks for attention, she will be told
that the issue that is worrying her will now be in her doll’s pocket and will be discussed
tomorrow. What is pivotal, therefore, is to teach the learner to put her thoughts there and
let them be while she does whatever the situation demands here and now (in this case
staying in bed and sleeping).

There are many examples of how learners can be released from the hold of private
events and oriented toward delayed consequences for actions that align with values. For
example, imagine a learner who is having problems at school because he frequently hits
other children. When asked, the boy says that he does it because of the rage he feels. The
immediacy and intensity of the boy’s emotions here have created a high probability that
acting on this basis will be more salient than thinking about the consequences of his
actions. So one way to deal with such a problem is to alter the functions of the emotions
themselves. For instance, ask the learner, “Where do you feel the rage?” expecting him to
point to some specific area of his body (his neck, stomach, head, fist, or other area), and
proceed by saying, “What does it look like? If it were an object, what would that be?”
Let’s say that the learner describes it as like a glove tightening around his throat. Then
ask the learner to close his eyes while you say the following: “We both know that the rage
you feel is not a glove around your throat, but let’s imagine it is. I want you to grab the
glove with your hands, place it in front of you, look at it, and then imagine you could fly
over it. Tell me what you see. Now, fly under it and tell me what you see. Now, let it be
in front of you for a while, and then let it go with the wind.” (You can also ask the learner
to place it in his pocket, or let it be there without doing anything but observing it.) “And
now that that has happened, who is in charge and what might you do instead?”
Follow the same steps with other emotions or thoughts that accompany the rage. For
instance, if the learner says, “They believe I am stupid,” say the following: “Imagine you
could write that thought down on imaginary paper. Do it. Do you have it? What color
is the ink? Is it handwritten or printed? Is it written in capital letters? Now I want you to
crumple the paper up and put it in your pocket (or let it fly away into the wind).” Then
ask, “Who is in charge and what do you want to do now?” It is important to note that in
this exercise the learner is being guided to notice and tact all aspects of his content and
determine whether his next actions need to be based on this.
Concluding Comments
To conclude, the complex repertoires of derived relational responding that make up the
self and our ability to engage in the different types of rule following, including discrimi-
nations of our own private self-rules, are an integral part of the rich fabric of our lives. Of
course, there is no single or easy way to teach these skills to learners who appear to lack
them. Nor is there even an easy way for teachers to determine precisely what is deficient.
To make matters worse, there is little empirical evidence and few sound training protocols
that dictate in a clear and concise fashion what we would do for learners in this regard.
The current chapter offers a first step in this direction. We have attempted to articulate
the basic concepts and processes that make up a relational understanding of the self and
rule-governed behavior, and how these interact in a manner that allows verbally sophisti-
cated human beings to take control of their own private experiences in a manner that is
consistent with what they value.
Chapter 16 349

References
Ayllon, T., & Milan, M. A. (1996). Desorden de déficit atencional e hiperactividad
[Attention-deficit hyperactivity disorder]. In M. C. Luciano (Ed.), Manual de psi-
cología clínica. Infancia y adolescencia [Handbook of clinical psychology. Childhood
and adolescence]. Valencia, Spain: Promolibro.
Barnes-Holmes, D., Hayes, S. C., & Dymond, S. (2001). Self and self-directed rules. In
S. C. Hayes, D. Barnes-Holmes, and B. Roche (Eds.). Relational frame theory. A post-
Plenum.
Barnes-Holmes, D., O’Hora, D., Roche, B., Hayes, S. C., Bissett, R. T., & Lyddy, F.
(2001). Understanding and verbal regulation. In S. C. Hayes, D. Barnes-Holmes, and
B. Roche (Eds.). Relational frame theory. A post-Skinnerian account of human language
and cognition. New York: Kluwer Academic/Plenum.
Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind. Boston:
MIT Press/Brandford Books.
Friedling, C., & O’Leary, S. G. (1979). Effects of self-instructional training on second-
and third-grade hyperactive children: A failure to replicate. Journal of Applied Behavior
Analysis, 12, 211–219.
Grant, L., & Evans, A. (1994). Principles of behavior analysis. New York: Harper Collins.
Hayes, S. C. (1984). Making sense of spirituality. Behaviorism, 12, 99–110.
Hayes, S. C., Barnes-Holmes, D., & Roche, B. (2001). Relational frame theory. A post-
Plenum.
Hayes, S. C., Gifford, E. V., & Hayes, G. J. (1998). Moral behavior and the development
of verbal regulation. Behavior Analyst, 21, 253–279.
Hayes, S. C., Zettle, R., & Rosenfarb, I. (1989). Rule-following. In S. C. Hayes (Ed.),
Rule-governed behavior: Cognition, contingencies, and instructional control. New York:
Plenum Press.
Herruzo, J., & Luciano, M. C. (1994). Procedimientos para establecer la “correspondencia
decir-hacer.” Un análisis de sus elementos y problemas pendientes [Procedures for the
establishment of say-do correspondence. A review of its elements and pending issues].
Acta Comportamentalia, 2, 192–218.
Luciano, M. C. (1996). Intervención psicológica en retraso en el desarrollo: Una perspec-
tiva funcional [Psychological intervention in developmental disabilities: A functional
approach]. In M. C. Luciano (Ed.), Manual de psicología clínica. Infancia y adolescen-
cia. Valencia, Spain: Promolibro.
Luciano, M. C., Barnes-Holmes, Y., & Barnes-Holmes, D. (2002). Establishing reports of
saying and doing and discriminations of say-do relations. Research in Developmental

Luciano, M. C., Herruzo, J., & Barnes-Holmes, D. (2001). Generalization of say-do cor-
respondence. Psychological Record, 51, 111–130.
Luciano, M. C., Valdivia-Salas, S., Gutiérrez, O., Ruíz, F., & Páez, M. (in press). ACT-
based brief protocols in children and adolescents. International Journal of Psychology
and Pyschological Therapy.
Manning, B. H., & Payne, D. B. (1996). Self-talk for teachers and students: Metacognitive
strategies for personal and classroom use. London: Allyn & Bacon.
Meichenbaum, D. (1985). Teaching thinking: A cognitive-behavioral perspective. In S. F.
Chipman, J. W. Segal, & R. Glaser (Eds.), Thinking and learning skills: Research and
open questions (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum.
O’Hora, D., & Barnes-Holmes, D. (2004). Instructional control. Developing a rela-
tional frame analysis. International Journal of Psychology and Psychology Therapy, 4,
263–284.
Paniagua, F. A. (1997). Verbal-nonverbal correspondence training as a case of environ-
mental antecedents. In D. M. Baer & E. M. Pinkston (Eds.), Environment and behav-
ior. Boulder, CO: Westview Press.
Risley, T. R., & Hart, B. (1968). Developing correspondence between the non-verbal
and verbal behavior of preschool children. Journal of Applied Behavior Analysis, 1,
267–281.
Schleser, R., Cohen, R., Meyers, A. W., & Rodick, J. D. (1984). The effects of cogni-
tive level and training procedures on the generalization of self-instructions. Cognitive
Therapy and Research, 8, 187–199.
Schlinger, H., & Blakely, E. (1987). Function-altering effects of contingency-specifying
stimuli. Behavior Analyst, 10, 41–45.
Skinner, B. F. (1945). The operant analysis of psychological terms. Psychological Review,
52, 270–277.
Skinner, B. F. (1969). Contingencies of reinforcement. New York: Appleton-Century-
Crofts.
Wicks-Nelson, R., & Israel, A. C. (1997). Behavior disorders of childhood. Maryland:
Prentice-Hall.
Zettle, R. D., & Hayes, S. C. (1982). Rule-governed behavior: A potential theoretical
framework for cognitive-behavioral therapy. In P. C. Kendall (Ed.), Advances in cogni-
tive behavioral research and therapy. New York: Academic Press.
Chapter 16 351

Chapter 17
Teaching Flexible, Intelligent,

and Creative Behavior
Catriona O’Toole, Carol Murphy, and Dermot Barnes-Holmes,

National University of Ireland, Maynooth;
Jennifer O’Connor, ABACAS, Kilbarrack, Ireland
The understanding of flexible, intelligent, and creative behavior, as well as the use of
training protocols for its establishment, are the two primary aims of the current chapter.
The core rationale of the work rests on the assumption that behavioral flexibility is at the
heart of intelligence and creativity. Hence, the applied training aspects of the chapter
focus primarily on how teachers or practitioners can systematically establish or track flex-
ibility in a range of behavioral domains.
Introduction
Researchers from practically all areas of psychology have, at one time or another, argued
that relational responding is a core component of the skills considered to be markers of
intelligence (including communication, problem solving, and analogical reasoning; Hayes,
Barnes-Holmes, & Roche, 2001). Indeed, a recent series of studies generated by relational
frame theory (RFT)—a functional behavioral approach to language and cognition—have
shown that specific patterns of relational responding predict performances on standard-
ized intelligence tests (O’Hora, Peláez, & Barnes-Holmes, 2005; O’Toole & Barnes-
Holmes, in press). Furthermore, human proficiency in relational responding appears to be
central to distinctions between us and our nonhuman counterparts (Dugdale & Lowe,
2000; Gentner, 2003).
It is likely then, that any attempt to improve intelligent behavior will necessarily
involve establishing or strengthening relational repertoires. The current chapter, however,
will not focus on strengthening relational skills, per se. Several preceding chapters in the
current volume have already done this, with a multitude of demonstrations of how to
build rich and complex relational repertoires. While none of these programs explicitly
claims to teach intelligent behavior, the strengthening of relational networks they instruct
naturally generates increasingly adaptive and intelligent behaviors.
It is the basic assumption of the current chapter that truly intelligent and creative
behaviors require more than the strengthening of relational responding; they also draw
heavily on relational flexibility. This hypothesis is not new, and it is consistent with main-
stream intelligence research from the early 1970s (such as Cattell, 1971; Guilford, 1975).
For example, Kyllonen, Lohman, and Woltz (1984) reported that individuals who adopted
flexible shifting strategies were better able to solve complex cognitive tasks.
According to RFT, intelligent and creative behavior involves flexibility because rela-
tional networks must come under increasingly subtle and flexible forms of contextual
control (Hayes et al., 2001). Consider the very recent empirical study by O’Toole and
Barnes-Holmes (in press). Participants were presented with two relational tasks and were
asked to respond quickly
��
and accurately in ways that were either consistent or inconsist-
ent with previously learned relations. During consistent trials, participants were required to
respond correctly to simple before-after relations (for example, spring comes before summer
[true] and summer comes before spring [false]). However, during inconsistent trials, partici-
pants were basically required to give the “wrong” or counterintuitive response (for example,
summer comes before spring [true] and spring comes before summer [false]). The research-
ers predicted that participants would take longer to respond on inconsistent trials because
these demand a relatively high degree of relational flexibility (in other words, getting the
answer wrong is harder than getting it right when you know what the right answer is), and
indeed this was the case. In conjunction with the relational tasks, participants completed
an intelligence test and the researchers attempted to determine the relationship between
performances on this test and on the relational tasks. Interestingly, inconsistent trials were
better predictors of performance on the intelligence test than consistent trials were, thus
highlighting a correlation between relational flexibility and intelligence.
Flexibility vs. Rigidity

An important corollary of the belief that flexibility is central to intelligence is the view
that negative behavioral outcomes will be associated with behavioral rigidity. And indeed,
there is a wealth of research that indicates that rigidity hinders learning. For example, in
typically developing adults, self-reported rigidity has been found to correlate with insen-
sitivity to operant contingencies (Wulfert, Greenway, Frakas, Hayes, & Dougher, 1994).
In other words, rigidity creates fewer opportunities for natural reinforcers to select adap-
tive and functional behaviors. Similar outcomes were reported by Mullins and Rincover
(1985) in a study involving children with and without a diagnosis of autism. Specifically,
in this research the children were asked to select one of five cards, where each card sig-
naled a different schedule of reinforcement. Although the typically developing children
quickly learned to choose the card associated with the most reinforcement, the children
with autism persistently selected cards associated with less reinforcement, thus rigidly
restricting their own access to reinforcers.

The latter findings may be particularly relevant to the implementation of behavioral
interventions, because one of the most fundamental ways in which we shape and promote
adaptive behaviors is by operating contingencies of reinforcement. These contingencies are
pivotal in strengthening adaptive behavior and weakening maladaptive behavior, and in
highlighting the distinctions between the two types of responding. However, an undesir-
able side effect of contingencies is that the increased frequency of the reinforced behavior
naturally reduces behavioral variability overall. And this in turn increases susceptibility
to rigidity (see Vogel & Annau, 1973). But the problem is not with reinforcement per se.
For example, if reinforcement is delivered contingent upon producing a variety of novel
responses, then the result will be an increase in novelty, not rigidity (for example, Lee,
McComas, & Jawlor, 2002). So reinforcement does not necessarily lead to rigidity.
In order to avoid rigidity as an outcome, training programs need to ensure that, in certain
contexts, reinforcement is delivered contingent upon flexible or nonstandard responses, rather
than only “correct” responses. This becomes all the more important when we consider that
many learners (such as those with autism) exhibit preexistent and oftentimes extreme forms
of rigidity. If flexibility is not incorporated into their interventions, you may inadvertently
increase rigidity in a group for whom rigidity is already considered a core problem. (It is
perhaps noteworthy that, if there is one commonality in the criticism leveled against applied
behavior analysis methods, it is possibly the belief that they create robotic, mechanical, and
rigid behavior, which is in stark contrast with alternative therapies that focus on creativity
and spontaneity [see Lovaas & Wright 2006]. Incorporating flexibility into response reper-
toires would certainly go some way toward addressing this criticism.) One way to begin to
foster flexibility is to require the learner to shift between two different types of responses. In
the following section, we present a training protocol designed to do just that.
Protocol for Establishing Relational Flexibility

The protocol outlined here is based on research conducted with typically developing
children and those with autism (O’Connor, 2004). The program involves training learn-
ers to produce either a symmetrical or asymmetrical relational response, depending on
the presence of subtle contextual cues (symmetry/asymmetry training), and incorporates
explicit flexibility training.
Symmetry relations are a critical early component of verbal abilities (Barnes-Holmes
et al., 2001). Consider a learner exposed to training in word-object relations (for example,
given the words “teddy bear,” select an actual teddy bear) who spontaneously reverses the
relation into a novel object-word relation (in other words, when asked, “What’s this?” and
presented with a teddy bear, the learner says “teddy bear”). For RFT, this latter response is
referred to as a derived symmetry relation because the reverse relation was explicitly trained
but the target sequence was not.
An asymmetry relation is also concerned with reversing a trained relation, but it
involves selecting a stimulus that is not the same as that which was trained. Consider a
learner to whom reinforcement is provided for selecting stimulus B1 in the presence of
A1 and B2 in the presence of A2 (in other words, A1-B1 and A2-B2 relations are directly
trained). When testing the derivation of the target symmetry relations based on the initial
conditional discrimination training, the learner is then presented with B1 or B2 and asked
Chapter 17 355

to select either A1 or A2 as appropriate. In this case, selecting A1 would be the correct
symmetrical response in the presence of B1, and A2 would be correct in the presence of
B2. The opposite selections would be asymmetrical (B1-A2 and B2-A1).
It may at first appear counterintuitive to concern yourself with asymmetry relations,
which effectively require learners to produce the “wrong” answer. But, as noted above, the
ability to provide the “wrong” answer when required—or to shift between two different
types of responding that are correct or wrong depending on the context—is an important
feature of behavioral flexibility, and thus of intelligence. Figure 17.1 provides a schematic
overview of symmetry/asymmetry training and explicit flexibility training in the protocol
for establishing relational flexibility based on the research by O’Connor (2004).
Symmetry and Asymmetry Training Program
Phase 1
Pass Fail:
Repeat phase 1
Phase 2
Explicit Flexibility
Pass Fail Training Program:
Level 1
Level 2
Phase 3 Level 3
Level 4
Flexibility training is designed
Pass Fail:
to facilitate contextually
Repeat Phase 2 controlled symmetry and
asymmetry responding in
Phase 4 phase 2. Thus learners who fail
phase 2 proceed directly to
flexibility training level 1. When
they reach criterion on level 1,
Pass Fail:
they are reexposed to phase 1
Repeat Phase 4 of the symmetry /asymmetry
program. If they again fail to
Phase 5 reach criterion in phase 2, they
proceed to flexibility training
level 2, and so on. Thus,
Pass: Fail: learners are exposed to
repeated cycles of phases 1
Training Return to phase 1 and 2, coupled with levels of
successfully flexibility training until they
completed achieve mastery on phase 2.
Figure 17.1. A schematic overview of symmetry/asymmetry training and explicit flexibility

training in the protocol for establishing relational flexibility.
Prerequisite Skills
Prior to exposure to the protocol, learners must present with basic reading skills.
They must be able to match a spoken word to the corresponding text, and they must have
phonetic awareness (in other words, they must know the sound that goes with each letter).
They must also have listener skills and be able to follow directions.

Feedback and Reinforcement
During all phases of the protocol, a correct response is recorded when the learner
emits the appropriate response within five seconds of stimulus presentation. During all
of the training trials (phases 1, 2, and 4), positive reinforcement (such as a token, small
edible, or praise) is delivered contingent upon correct responding. Following an incorrect
response or no response, you should repeat the antecedent and then model the correct
response. However, during test trials (phases 3 and 5), there are no programmed conse-
quences for responding, whether correct or incorrect. Learners are not required to repeat
incorrect trials at any point.
Stimuli
In phases 1, 2, and 3, two 3-by-5-inch index cards are required, each
with a printed nonsense syllable (such as “vug” and “lup”).
Phases 2, 3, and 5 require two cardboard circles, one red and one blue.
A small ridge should be attached to the back of each to enable the circle
to stand independently on the table. The two cardboard circles should be
identical in form (differing only in color) and approximately 10 inches
in diameter. The circles will function as contextual cues, and in advance
you should designate the color that is to control the targeted responses
(for example, blue for symmetrical responses and red for asymmetrical
responses). To make it easier for yourself, write “symmetry” on the back
of one card as appropriate, and “asymmetry” on the back of the other.
Phases 4 and 5 require four 3-by-5-inch index cards, each with a printed
abstract symbol (see figure 17.2). These cards should be arranged to form
two equivalence classes, each with two members: A1-B1 and A2-B2.
Again, write the appropriate alphanumeric label on the back of each
ˇˇ Ϭ Ϭ
card.
A1 A2
ˇ ˇ Ϭ Ϭ
ϊ ϊ ‫فف‬
B1 B2
ϊ ‫ف‬
ϊ ‫ف‬
Figure 17.2. Examples of abstract stimuli for use in phases 4 and 5 of the symmetry/
asymmetry training.
Chapter 17 357

Instructions
Phase 1: Explicit name training. In this phase the learner is trained to emit the appropriate
textual response to the two written nonsense syllables. Present the training as twenty random
trials—ten for each nonsense syllable. A sample data sheet for use in phase 1 is presented in
figure 17.3. On each trial, hold up the card in front of the learner. On the first presentation of
each, give a vocal model prompt, to which the learner repeats the echoic response (hear “vug,”
say “vug”). For all remaining trials, the learner must emit a textual (see “vug,” say “vug” )
rather than echoic response (in other words, the learner is required to respond independently).
Learners must produce eight out of ten correct responses for each nonsense syllable (a total of
sixteen out of twenty correct responses is required to progress to the next phase of training).
Phase 2: Training symmetry and asymmetry relations under contextual control.

In this phase, the colored circles function as contextual cues, so you should have desig-
nated each as appropriate (for example, blue as symmetry and red as asymmetry). The
two nonsense syllables are placed side-by-side on the table (make sure you randomize the
left and right positions of the nonsense word cards across trials). Training begins with the
following instructions: “I am going to place these two words on the table in front of you
and I’ll say a name. Then you have to choose the card that you think is the right one. If
you get it right I will tell you, and I can also tell you if you get it wrong.”
Then place one of the colored contextual cue cards between the nonsense syllables, but
do not offer the learner information regarding its purpose. This omission is critical for the
current training because behavior that is governed by explicit verbal instructions or rules
may induce rigidity (see Hayes, Zettle, & Rosenfarb, 1989)—and this is precisely what
we wish to avoid. When the contextual cue is in place, name one of the nonsense syllables
(such as “vug”). The learner must point to (or touch) the appropriate nonsense syllable.
The correct response in phase 2 depends on which contextual cue (in other words, which
color) is present. That is, for example, in the presence of the blue circle the learner should
point to the syllable that is the same as the one uttered by the teacher (in other words,
emit the symmetrical response—hear “vug,” touch “vug”), and in the presence of the red
circle the learner should point to the syllable that is different from that uttered (in other
words, emit the asymmetrical response—hear “vug,” touch “lup”).
Phase 2 consists of forty randomized trials—twenty symmetry trials involving the
blue circle, and twenty asymmetry trials involving the red circle. Each of these includes
ten trials in which “vug” is targeted, and ten involving “lup.” Hence, there are four basic
trial types: symmetry/“vug”; symmetry/“lup”; asymmetry/“vug”; and asymmetry/“lup.”
The mastery criterion is thirty-six out of forty, with no more than one error on each trial
type. A sample data sheet for use in phase 2 is presented in figure 17.4.

Symmetry and Asymmetry Training
Phase 1: Explicit Name Training
Client:
Teacher:
Date:
Trial VUG LUP

1*
2
3
4
5
6
7
8
9
10
Mastery
Criterion 8/10 8/10
* Give a vocal model prompt on the first presentation of each stimulus; the
remaining trials require independent responses.
Figure 17.3. A sample data sheet for use in phase 1.
Chapter 17 359

Phase 2: Training Symmetry and Asymmetry
Under Contextual Control
Client:
Teacher:
Date:
Symmetry Asymmetry
Trial
VUG LUP VUG LUP
1
2
3
4
5
6
7
8
9
10
Mastery
Criterion 9/10 9/10 9/10 9/10
Learners will likely have had no difficulty proceeding through the name training
in phase 1 but frequently fail to reach criterion in phase 2. Of course, learners who do
achieve mastery in phase 2 can proceed directly to phase 3 (passing by flexibility training
altogether). But for those who fail, transfer them to level 1 of explicit flexibility training.
In the interests of these latter weaker learners, we have presented the explicit flexibility
training directly below (and in figure 17.1). If specific learners do not require this inter-
vention, ignore the following section and skip directly to phase 3 of symmetry/asymmetry
training.

Explicit Flexibility Training
Stimuli
The red and blue circles used in the symmetry/asymmetry training will also be
employed here, along with the following:
An array of ten familiar pictures, each printed on a 3-by-5-inch card

(such as car, ball, spoon, building brick, hairbrush, pencil, chicken,
tractor, shoe, and cup). The learner should be easily able to name each
picture. These are referred to as set 1.
An array of two hundred familiar pictures (such as horse, sun, pencil,

and cat) also printed on 3-by-5-inch card. These should consist of one
hundred identical pairs of pictures (such as two pictures of a horse, and
so on). Learners may or may not be fluent in naming these items. This
set of pictures is referred to as set 2.
Feedback and Reinforcement

After each training trial, reinforcement is delivered contingent upon a correct response,
and corrective feedback follows an incorrect response.
Mastery Criterion
At all levels, mastery criterion is 100 percent correct responding.
Instructions
Level 1. The aim here is to teach the learner to provide the “right” or “wrong” naming
response. A sample data sheet for use here and in the remaining levels of explicit flexibility
training is presented in figure 17.5. On each trial, place five randomly selected pictures
from set 1 on the table. Point to one of the pictures and provide the learner with one
of two antecedents: “What is it? Give me the right answer” or “What is it? Give me the
wrong answer.” A correct response, of course, involves naming the picture correctly or
incorrectly depending upon your instructions. Where an alternative name is sought (in
other words, when you ask for the wrong name), the learner must provide the name of
any other pictorial item. Level 1 training comprises blocks of twenty randomized trials—
ten requesting the right answer and ten requesting the wrong answer. It is important to
note that neither of the colored contextual cues is present during level 1.
Chapter 17 361

Explicit Flexibility Training Phases 1–4
Client:
Teacher:
Date:
Level of training:
Trial Standard Response Nonstandard Response

1
2
3
4
5
6
7
8
9
10
Mastery
Criterion 10/10 10/10
Figure 17.5. A sample data sheet for use in levels 1 through 4 of the explicit flexibility
training.
Level 2. Level 2 is similar to level 1, except that learners are now required to select, rather
than name, the appropriate picture (including not selecting pictures when asked not to
select them). During each trial, touch one of the pictures and emit one of the following
antecedents: “Give me something. Give me a ” (“spoon,” for example) or
“Give me something. Don’t give me a ” (“spoon,” for example). It is impor-
tant to note that neither of the colored contextual cues is present during level 2.
Level 3. Level 3 is similar to level 2, except that it involves training the learner to make
an appropriate selection in the presence of the red and blue contextual cues from the
symmetry/asymmetry training. So now you are bringing the existing matching and non-
matching responses under contextual control. For example, in the presence of the blue
circle, instruct the learner as follows: “Give me something. Give me a ”

(“spoon,” for example) and, in the presence of the red circle, say, “Give me something.
Don’t give me a ” (“spoon,” for example).
Level 4. The flexibility training in level 4 differs in format from that of the previous three
levels. Specifically, the trials in level 4 should be presented in an MTS format (for a more
detailed description of MTS, see chapter 10). During each trial, place three pictures from
set 2 on the table. Two of the pictures must be identical, with the third being different
(such as two identical cats and a bed). Use one of the identical pictures as a sample, posi-
tioning this farthest away from the learner. The remaining two pictures are used as com-
parisons and placed below the sample (one to the left and one to the right). Then place
one of the contextual cues directly behind the sample. During these trials the learner is
given no verbal antecedent and no explicit instructions regarding the significance of the
contextual cue. In the presence of one contextual cue (such as blue) the learner is required
to select the comparison that is identical to the sample, whereas in the presence of the
other (red), the learner should select the nonmatching comparison. As in the previous
levels, trials are presented in blocks of twenty, ten for each contextual cue. A new array of
stimuli is presented on each trial.
Returning to Symmetry/Asymmetry Training

To mix symmetry/asymmetry training and explicit flexibility training, once learners
reach criterion on level 1 of explicit flexibility training (these had obviously failed phase 2
of symmetry/asymmetry training), reexpose them to phases 1 and 2. Learners who pass
phase 2 at this point proceed immediately to phase 3. Learners who again fail in phase 2
are then exposed to flexibility training at level 2. Once again, when they reach criterion,
reexpose them to phases 1 and 2. Let those who pass proceed to phase 3. Those who fail
should be exposed to flexibility training at level 3 and then reexposed to phases 1 and
2 when they reach criterion. Those who fail phase 2 again are exposed to level 4 of flex-
ibility training before reexposure to phases 1 and 2. In this way, the levels of flexibility
training are employed to facilitate learners’ performances during the symmetry/asymme-
try training under contextual control in phase 2.
Phase 3: Testing symmetry and asymmetry relations under contextual control.

This phase is identical to phase 2, except that the learner is exposed to five (rather than
ten) of the four targeted trial types, and no feedback is given. A sample data sheet for use
in phase 3 is presented in figure 17.6. Prior to commencing this phase, the learner is given
the following instructions: “This time, I am going to place the two words on the table in
front of you as I did before and then I’ll say a name. Once again you have to choose a
card that you think is the right one. The only difference is that this time I can’t tell you
if you got it right or wrong.”
Learners are required to produce sixteen out of twenty correct responses. As in phase
2 only one error is permitted on each of the four trial types. Learners who do not meet
the mastery criterion should be reexposed to phase 2, but those who do so can proceed
immediately to phase 4.
Chapter 17 363

Phase 3: Testing Symmetry and Asymmetry
Under Contextual Control
Client:
Teacher:
Date:
Symmetry Asymmetry
Trial
VUG LUP VUG LUP
1
2
3
4
5
Mastery
Criterion 4/5 4/5 4/5 4/5
Phase 4: MTS training. The purpose of phases 4 and 5 is to ensure that the contextu-
ally controlled symmetry and asymmetry responses can generalize to new sets of stimuli.
Phase 4 employs a standard MTS procedure to explicitly train a pair of conditional dis-
criminations. The stimuli consist of four cards containing abstract symbols, which form
two equivalence classes of two members each, A1-B1 and A2-B2. During each trial, place
a sample card (such as A1) on the table with two comparisons (B1 and B2) below, one
on the left and one on the right (a stimulus placement board may be used to facilitate
placement of these cards). Then instruct the learner as follows: “Now we are going to do
another task, and this is about matching things together. Like the task we did earlier, I
will be able to tell you when you get them right or wrong.”
When the three stimuli are positioned on the table, the teacher touches the sample
stimulus (such as A1) and immediately presents the antecedent “goes with.” The teacher
should then remain looking at the stimuli for five seconds or until the learner emits a
response. During this time it is important that the teacher does not interact or make
eye contact with the learner. This is to ensure that the contingencies remain clear (in
other words, attention or other reinforcement is delivered contingent upon correct
responding).

This training consists of twenty trials, ten in which A1 is the sample and ten with
A2 as the sample. A sample data sheet for use in this phase is presented in figure 17.7.
On all twenty trials, B1 and B2 function as the comparison stimuli, and their left-right
positions should be counterbalanced across trial types. Specifically, on half of the trials
in which A1 is the sample, B1 should be on the right and B2 on the left, with the reverse
on the remaining trials. The ordering of the trials containing either sample should also
be randomized across the twenty training trials. Note that the contextual cues are not
present in this phase.

Phase 4: Matching-to-Sample (MTS) Training
Client:
Teacher:
Date:
Sample = A1 Sample = A2
Trial
Correct comparison = B1 Correct comparison = B2
1
2
3
4
5
6
7
8
9
10
Mastery
Criterion 9/10 9/10
Learners are required to reach a mastery criterion of eighteen out of twenty correct
responses with no more than one error on each of the two trial types (A1-B1 and A2-B2).
Those who meet criterion can proceed to phase 5, but those who do not should be reex-
posed to phase 4.
Chapter 17 365

Phase 5: Contextually controlled MTS testing. In this phase the learner is required
to respond symmetrically and asymmetrically to the conditional relations established in
phase 4, using the contextual cues employed in phases 2 and 3 (in other words, the red
and blue circles). The MTS procedure is again employed, but now the B stimuli function
as samples with the A stimuli as comparisons. Thus, during each trial the teacher places
one of the B stimuli down on the table with the two A stimuli on the left and right below.
On each trial place one of the contextual cues behind the B card when all of the other
stimuli are in position. Then instruct the learner as follows: “This time, we are going to
do the matching task as we did before, but the only difference is that I cannot tell you if
you got it right or wrong.”
The teacher then touches the sample and delivers the antecedent “goes with.” The
teacher should remain focused on the stimuli while awaiting a response. Correct sym-
metrical responses involve the learner selecting the comparison A1 in the presence of the
sample B1, and A2 in the presence of B2. Conversely, correct asymmetrical responses
involve selecting A2 in the presence of B1 and A1 in the presence of B2.
Phase 5 consists of twenty trials, ten in which the contextual cue for symmetrical
responding is present, and ten with the contextual cue for asymmetrical responding. A
sample data sheet is presented in figure 17.8. Of the ten symmetry trials, five involve the
presentation of B1 as the sample and five with B2 as the sample. Again, the locations of
the comparisons should be counterbalanced. Similarly, the ten asymmetry trials contain
five with B1 as the sample and five with B2 is the sample, and the locations of the com-
parisons are counterbalanced. The four trial types are presented randomly across the block
of twenty trials. Learners are required to produce sixteen out of twenty correct responses,
with no more than one error on each of the four trial types (in other words, symmetry,
B1-A1/B2-A2; and asymmetry, B1-A2/B2-A1). Learners who meet this criterion have suc-
cessfully completed the training, and learners who do not can be returned to phase 1.
Performances on the Protocol

In the research by O’Connor (2004), the symmetry/asymmetry training protocol was
presented to ten normally developing children and ten children with autism, all between
six and nine years old. Although both groups had similar levels of verbal ability, they
differed in their performances on the protocol. Table 17.1 presents data sets from four
normally developing children when presented with phases 1 through 5 of the symme-
try/asymmetry training. Table 17.2 presents data sets from three children with autism
exposed to symmetry/asymmetry training and explicit flexibility training, taken from the
research by O’Connor (2004).

Phase 5: Contextually Controlled MTS Testing
Client:
Teacher:
Date:
Symmetry Asymmetry
Sample = B1 Sample = B2 Sample = B1 Sample = B2
Trial
Correct Correct Correct Correct
comparison comparison comparison comparison
= A1 = A2 = A2 = A1
1
2
3
4
5
Mastery
Criterion 4/5 4/5 4/5 4/5
Chapter 17 367

Table 17.1. Accuracy Scores for Four Typically
Developing Children
Experimental Phases
Participant Phase 1 Phase 2 Phase 3 Testing Phase 4 Phase 5 Testing
Training (20) Training (40) (20) Training (20) (20)
1 20 38 20 20 20
2 20 39 20 20 19
3 20 36 20 20 20
4 20 37 20 19 19
Table 17.2. The Total Number of Training Trials

Required to Reach Criterion and Accuracy Scores for
Three Children with Autism
Phase 1 Phase 2 Explicit Flexibility
Phase Phase 4 Phase
Training Training Training 3 Training 5
Participant (20) (40) Testing (20) Testing
Level Level Level Level (20) (20)
1 2 3 4
19 24 22 - - - - - -
1
20 38 - - - - 20 18 18
18 22 31 - - - - - -
19 23 - 20 - - - - -
2
20 32 - - 25 - - - -
20 37 - - - - 18 17 17
19 26 28 - - - - - -
20 27 - 27 - - - - -
3 20 28 - - 20 - - - -
18 26 - - - 61 - - -
19 40 - - - - 19 18 19
Note: Dashed lines (-) indicate absence of exposure to a particular stimulus set.
Each line represents an individual experimental session.

In the original research, all of the normally developing children proceeded rapidly
and competently through the five phases of symmetry/asymmetry training and thus none
required explicit flexibility training. In contrast, six of the ten children with autism had
difficulty producing the contextually controlled symmetry and asymmetry responses or
had difficulty generalizing these to novel sets of stimuli. Exposing these latter children
to explicit flexibility training was found to be extremely beneficial in facilitating their
performances on the contextually controlled symmetry and asymmetry trials (see table
17.2). The success of the latter training demonstrated that relational flexibility can be
taught with considerable ease, and that behavioral rigidity or stereotypy need not be an
immutable characteristic, at least in some respects. As a whole, the combined intervention
protocol constituted a first step in the establishment of flexible relational repertoires, and
thus in training intelligent and creative behavior.
Future Directions in Training Flexible

Relational Responding
Symmetry relations are just one type of relational responding, with others including
responding in accordance with opposition, difference, comparison, and perspective-taking
relations (e.g., Barnes & Hampson, 1993; Barnes & Keenan, 1993; Dymond & Barnes,
1994, 1995, 1996; Roche & Barnes, 1996, 1997; Steele & Hayes, 1991). Although there
are now numerous studies that have reportedly established complex relational repertoires
even with young learners, there remains limited evidence of how flexibility training can
be integrated into these programs (Barnes-Holmes, Barnes-Holmes, & Smeets, 2004;
Barnes-Holmes, Barnes-Holmes, Smeets, Strand, & Friman, 2004).
There is reason to be optimistic that such integration might occur with relative ease
based on empirical evidence from studies conducted with children with autism (Murphy,
Barnes-Holmes, & Barnes-Holmes, 2005; Murphy & Barnes-Holmes, in press). These
studies successfully combined Skinner’s mand with RFT’s derived relational responding
to establish novel untrained manding. Notably, the research involved reversal procedures
that required participants to show patterns of emergent manding that changed in accor-
dance with reversals in contextual contingencies. In effect, participants were required to
provide novel mand responses that were previously “incorrect” across a double reversal
procedure, and they readily learned to do so. The researchers argued that this feature of
training naturally enhanced the flexibility of the children’s target relational repertoires.
Concluding Comments
The current chapter has focused on procedures for establishing flexible relational respond-
ing of a particular type. Specifically, our concern here has been establishing patterns of
responding that might be described as counterintuitive because learners are trained to
provide “wrong” answers in certain specific contexts. Although this is likely to be only
one aspect of behavioral flexibility that underpins intelligent and creative behavior, it is
nonetheless important because proficiency in this type of flexibility may be even more
Chapter 17 369

critical than speed. Consider, for example, areas of complex behavior such as humor,
storytelling, or deception, in which it is important for individuals to derive what others
might perceive as the wrong answer as an important guide to their own behavior. It is
likely that these skills are frequently absent in developmentally delayed populations, at
least in part because of the absence of relational flexibility, even when there are strong
successes in establishing the basic cognitive skills. As a result, it would appear that build-
ing flexibility into relational training programs early on would pay dividends in the long
run, even though it might significantly slow down the initial establishment of the target
relational skills (because the learner must learn to switch back and forth between wrong
and right). In doing so, of course, the right answer is always being derived and, thus, even
generating wrong answers deliberately forces the learner to have yet another exposure to
the right answer. Current evidence also suggests that this integration may be done with
considerable ease. While there is some preliminary evidence of correlations between rela-
tional responding per se and intelligence (particularly verbal), there is also evidence of
a link between relational flexibility and intelligence. However, further research on both
fronts will be needed for greater certainty in either regard. Nonetheless, even the most
cautious teacher or practitioner will experience excitement at the development and poten-
tial utility of systematic protocols that go some way toward establishing intelligent and
creative behavior in cognitively challenged populations.
References
Barnes, D., & Hampson, P. J. (1993). Stimulus equivalence and connectionism: Implications
for behavior analysis and cognitive science. Psychological Record, 43, 617–638.
Barnes, D., & Keenan, M. (1993). A transfer of functions through derived arbitrary and
non-arbitrary stimulus relations. Journal of the Experimental Analysis of Behavior, 59,
61–81.
Barnes-Holmes, Y., Barnes-Homes, D., Roche, B., Healy, O., Lyddy, F., Cullinan, V., et
al. (2001). Psychological development. In S. C. Hayes, D. Barnes-Holmes, and B.
Roche (Eds.), Relational frame theory: A post-Skinnerian account of human language
and cognition. New York: Kluwer Academic/Plenum.
Barnes-Holmes, Y., Barnes-Holmes, D., & Smeets, P. M. (2004). Establishing relational
responding in accordance with opposite as generalized operant behavior in young
children. International Journal of Psychology and Psychological Therapy, 4, 559–586.
Barnes-Holmes, Y., Barnes-Holmes, D., Smeets, P. M., Strand, P., & Friman, P. (2004).
Establishing relational responding in accordance with more-than and less-than as
generalized operant behavior in young children. International Journal of Psychology
and Psychological Therapy, 4, 531–558.
Cattell, R. B. (1971). Abilities: Their structure, growth and action. Boston: Houghton
Mifflin.

Dugdale, N., & Lowe, C. F. (2000). Testing for symmetry in the conditional discrimi-
nations of language-trained chimpanzees. Journal of the Experimental Analysis of
Dymond, S., & Barnes, D. (1994). A transfer of self-discrimination response functions
through equivalence relations. Journal of the Experimental Analysis of Behavior, 62,
251–267.
Dymond, S., & Barnes, D. (1995). A transformation of self-discrimination response func-
tions in accordance with the arbitrarily applicable relations of sameness, more-than,
and less-than. Journal of the Experimental Analysis of Behavior, 64, 163–184.
Dymond, S., & Barnes, D. (1996). A transformation of self-discrimination response func-
tions in accordance with the arbitrarily applicable relations of sameness and opposi-
tion. Psychological Record, 46, 271–300.
Gentner, D. (2003). Why we’re so smart. In D. Gentner and S. Goldin-Meadow (Eds.),
Language in mind: Advances in the study of language and thought. Cambridge, MA:
MIT Press.
Guilford, J. P. (1975). Creativity: A quarter century of progress. In I. A. Taylor & J. W.
Getzels (Eds.), Perspectives in creativity. Chicago: Aldine.
Hayes, S. C., Barnes-Holmes, D., & Roche, B. (Eds.). (2001). Relational frame theory:
A post-Skinnerian account of human language and cognition. New York: Kluwer
Academic/Plenum.
Hayes, S. C., Zettle, R. D., & Rosenfarb, I. (1989). Rule following. In S. C. Hayes (Ed.),
Rule-governed behavior: Cognition, contingencies, and instructional control. New York:
Plenum.
Kyllonen, P. C., Lohman, D. F., & Woltz, D. J. (1984). Computational modeling of
alternative strategies for performing spatial tasks. Journal of Educational Psychology,
76, 1325–1345.
Lee, R., McComas, J., & Jawlor, J. (2002). The effect of differential and lag reinforcement
schedules on varied verbal responding by individuals with autism. Journal of Applied
Lovaas, O., I., & Wright, S. (2006). A reply to recent public critiques… Journal of Early
and Intensive Behavioral Intervention, 3, 234–236.
Mullins, M., & Rincover, A. (1985). Comparing autistic and normal children along the
dimensions of reinforcement maximization, stimulus sampling, and responsiveness to
extinction. Journal of Experimental Child Psychology, 40, 350–374.
Murphy, C., & Barnes-Holmes, D. (in press). Derived more/less relational mands with
four children diagnosed with autism: Synthesizing Skinner’s Verbal Behavior with
relational frame theory II. Journal of Applied Behavior Analysis.
Murphy, C., Barnes-Holmes, D., & Barnes-Holmes, Y. (2005). Derived manding with
seven children diagnosed with autism: Synthesizing Skinner’s Verbal Behavior with
relational frame theory. Journal of Applied Behavior Analysis, 38, 445–462.
Chapter 17 371

O’Connor, J. (2004). Designing procedures to establish repertoires of derived relational
responding. Unpublished doctoral thesis, National University of Ireland, Maynooth.
O’Hora, D., Peláez, M., & Barnes-Holmes, D. (2005). Derived relational responding and
performance on verbal subtests of the WAIS-III. Psychological Record, 55, 155–175.
O’Toole, C., & Barnes-Holmes, D. (in press). Three chronometric indices of relational
responding: The importance of relational flexibility. Psychological Record.
Roche, B., & Barnes, D. (1996). Arbitrarily applicable relational responding and human
sexual categorization: A critical test of the derived difference relation. Psychological
Record, 46, 451–475.
Roche, B., & Barnes, D. (1997). A transformation of respondently conditioned stimulus
function in accordance with arbitrarily applicable relations. Journal of the Experimental
Steele, D., & Hayes S. C. (1991). Stimulus equivalence and arbitrarily applicable rela-
tional responding. Journal of the Experimental Analysis of Behavior, 56, 519–555.
Wulfert, E., Greenway, D. E., Farkas, P., Hayes, S. C., & Dougher, M. J. (1994).
Correlation between self-reported rigidity and rule-governed insensitivity to operant
contingencies. Journal of Applied Behavior Analysis, 27, 659–671.
Vogel, R., & Annau, Z. (1973). An operant discrimination task allowing variability of
reinforced response patterning. Journal of the Experimental Analysis of Behavior, 20,
1–6.

Ruth Anne Rehfeldt, Ph.D., BCBA, is an associate professor in the Rehabilitation
Institute of Southern Illinois University, Carbondale. She holds doctoral and master’s
degrees in psychology from the University of Nevada and a bachelor’s degree in psychol-
ogy from the University of Puget Sound. She is also a board-certified behavior analyst.
Yvonne Barnes-Holmes, Ph.D., is a lecturer in psychology in the department of

Psychology at the National University of Ireland in Maynooth, Ireland.
Foreword writer Steven C. Hayes, Ph.D., is a University of Nevada Foundation Professor

of Psychology at the University of Nevada, Reno. He is author of innumerable books and
scientific articles, including the successful ACT workbook Get Out of Your Mind and Into
Your Life.
Index
A arbitrarily applicable relational

responding, 283
ABLA. See Assessment of Basic Learning arbitrary conditional discriminations, 26
Abilities arbitrary matching, 187, 189
actions: discordance between emotions arbitrary stimuli: comparison and, 165;
and, 302; empathic responding and, coordination and, 156-157; distinction
309 and, 163; hierarchy and, 169
active responding, 188 argument, mathematical, 321
affective social referencing, 73 Aristotle, 257
age factors in perspective taking, 296-297 Asperger’s disorder, 298
alphabetic writing systems, 182 assessment: predictive, 33; reading, 199,
amplitude relations, 319-321 203; stimulus preference, 7-15; verbal
anagram construction, 192, 212 operant, 81
analogical reasoning, 257-278; Assessment of Basic Language and
behavioral approach to, 259-278; Learning Skills (ABLLS), 81
equivalence- equivalence relations Assessment of Basic Learning Abilities
and, 261-271; four-item analogies and, (ABLA), 25-37; common problems
271-278; intelligent behavior and, related to, 32; equivalence relations
257; traditional accounts of, 258-259; and, 35-36; flowchart illustrating, 31;
training and testing sequence for, 262- future research on, 36-37; predictive
271, 272, 275-278 assessment and, 33; procedures and
analogies, 257; causal, 272, 273; four- materials for, 27-30; programming
item, 272-278; proportional, 272, 274; learning using, 34-35
standard, 272, 273-274 asymmetry relations, 355-356, 358
asymmetry training: contextually categorization, 137-140; data sheets for
controlled, 358, 360, 363-364, 366; testing, 146-148; establishing skills in,
explicit name training and, 358, 359; 138-139; general description of, 137;
flexibility training and, 363-369; MTS intraverbal relations and, 140; studies
procedures and, 364-366, 367; sample on teaching, 137-138; teaching vocal
data sets from, 366, 368; schematic mands using, 251-252
overview of, 356 causal analogies, 272, 273
attending behaviors, 18-20 chained tasks, 238-239, 245-248; mands
attention, joint, 52, 69-73 within the context of, 246-248;
attention-deficit/hyperactivity disorder teaching to students, 245-246
(ADHD), 336 classes: adding quantity-based exemplars
auditory-receptive reading, 176 to, 216-219; grammatical, from
augmenting, 342-343, 346-347 materials in sequences, 219-220; novel
aversive stimuli: empathy training and, sequences from elements of, 211-216;
307; mands for removal of, 83, 84, 92 stimulus equivalence paradigm and,
314-315
B classification, 137
Barnes-Holmes, Dermot, 257, 281, 353 Cognitive Abilities Test, 298
Barnes-Holmes, Yvonne, 281 cognitive analysis, 174n
baseline trials, 195 cognitive approach to perspective taking,
behavior: emergent, 284, 314; flexible, 281, 282-283
353-370; manded, 112; rule-governed, combinatorial entailment: comparison
124-125, 339-349; self-directed, 343- and, 164; coordination and, 155;
349; symbolic, 1; textual, 80, 173-174 distinction and, 162; hierarchy and,
behavior analysis, ix-xi 167; opposition and, 160
behavioral approach: to analogical “comic strip conversations” procedure,
reasoning, 259-278; to mathematics 303
instruction, 314-317; to perspective comparison, 164-166; arbitrary contexts
taking, 281, 283-298; to self-rules and, 165; emotion identification and,
training, 337 166; nonarbitrary relations and, 165;
behaviorally oriented language programs, transformations of function and, 164,
149-150 165-166
Berens, Nicholas M., 149 comparison stimuli: commonly used in
bidirectional naming, 130-131 MTS, 115; teaching discrimination of,
bilingual individuals, 228 191
Bowman, Phyllis S., 174n compliance training, 112
“bubble dialog” procedure, 303-304 Comprehensive Application of Behavior
Burrell, Char, 2 Analysis to Schooling (CABAS), 42,
150
concepts, 137
C conceptualized self, 338
CABAS (Comprehensive Application of conditional discriminations, 26;
Behavior Analysis to Schooling), 42, Assessment of Basic Learning Abilities
150 and, 26-32, 36; equivalence triangle
Cabello-Luque, Francisco, 335 and, 176-178, 179; equivalence-
captured establishing operations, 84, equivalence relations and, 262-271;
91-92 matching-to-sample arrangements and,
115, 176; protocol for establishing, skills and, 244-249; explanation
64-69; teaching to students, 242-244, of, 237-239; future applications
252-253 of, 253; generalization probes for,
conditioned reinforcement: listening to 249; maintenance probes for, 249;
adult voices protocol, 46-48; print post-test probes for, 242, 244, 248-
stimuli on a page protocol, 53; sensory 249; preferred items and, 239-244;
matching protocol, 54-56; visual pre-test probes for, 239-240, 242,
tracking protocol, 49-51 248-249; selection-based, 239-249;
conjugate reinforcement, 48-49 stimulus relations probes for, 240-241;
constructed-response matching to sample topography-based vocal, 249-252
(CRMTS), 179, 185 derived relational responding (DRR):
constructivist teaching approach, 313-314 bidirectional naming relations and,
contextual control: embedded, 222- 131; controversy about verbal ability
228; external, 228-230; MTS testing and, 129; importance of joint attention
under, 366, 367; symmetry/asymmetry and social referencing for, 75-76;
training under, 358, 360, 363-364, perspective taking and, 283-284
366 derived stimulus relations: conditional
contingency-shaped behavior training, stimulus teaching and, 252-253;
337 derived mands for daily living skills,
contrived establishing operations, 84-85, 244-249; derived mands for preferred
93, 96-97 items, 239-244; explanation of,
coordination, 154-159; arbitrary stimuli 151-152; future applications of, 253;
and, 156-157; nonarbitrary relations maintenance and generalization
and, 156; sense making and, 158-159; probes, 249; perspective taking and,
transformations of function and, 155, 283-284; research studies on, 129;
157-158 teaching using technology of, 237-239;
cosine frequency relations, 321-323 topography-based vocal derived mands,
Crashaw, Richard, 174n 249-252; Web-based software and, 329
creativity, 353, 354 derived symmetry relations, 355
cross-modal discriminations, 28, 36 differential reinforcement, 16, 188
cross-sequence substitutions, 220-221 differential response, 190
cumulative baseline, 195 Dillon, Courtney M., 79
cusps, 41, 43-44 direct instructional control, 111-126;
generalized relational repertoires and,
D 122; nonrelational, 112-121, 123;
de Rose, Julio C., 173 relational, 123-125
de Souza, Deisy G., 173 discrete trial training (DTT), 150
decimal-fraction relations, 315-316 discrimination learning, 25-26, 68, 153,
decoding, 174 192
deictic relations, 284 distinction, 162-163; arbitrary stimuli
delayed matching procedure, 186 and, 163; nonarbitrary stimuli and,
delayed prompt fading, 118 162-163; transformations of function
Delgado, Jo Ann Pereira, 41 and, 163
demonstration trials, 316 Dixon, Mark, 313
derived mands, 237-253; chained Domeniconi, Camila, 173
tasks and, 245-248; daily living double reversed relations, 292
DRR. See derived relational responding
Index 377
E F
early listeners, 41-42 fading, 68, 118-119
echoics, 80, 86, 89 feelings. See emotions
echolalia, 22, 79 Fields, Lanny, 209
egocentric empathy, 302 Findlay, Katharine, 144
embedded contextual control, 222-228 fixed time delay, 87
emergent behavior, 284, 314 flexibility, 353-370; explicit training in,
emotional contagion, 306 361-363; future directions in training,
emotional intelligence, 301 369; intelligent/creative behavior and,
emotions: comparative relations and, 166; 353, 354; protocol for establishing,
discordance between actions and, 302; 355-360; rigidity vs., 354-355;
discrimination and labeling of, 307- symmetry/asymmetry training in, 356,
309; empathy for another’s feelings 358-360, 363-369
and, 302; perspective taking and, formative augmenting, 342
298-299; understanding in oneself and formula-to-formula relations, 317
others, 305 formula-to-graph relations, 317, 329
empathy, 301-310; development of, 302; four-item analogies, 271-278
emotional intelligence and, 301; as fraction-decimal relations, 315-316
relational responding, 305-309; RFT- frames of coordination, 133-137; diagram
based training strategies in, 306-309; illustrating, 135; matching-to-sample
treatments for deficits in, 303-304 procedures and, 133-134; naming
equivalence classes, 210; adding quantity- and, 130, 134; recommendations for
based exemplars to, 216-219; novel establishing, 135-137
sequences from elements of, 211-216 full physical guidance, 116
equivalence relations, 1; ABLA and,
35-36; analogies and, 261-271 G
equivalence triangle, 176-178, 180 gaze following, 71-72
equivalence-equivalence relations, 261- gaze shifts, 69-70
271; conditional discriminations and, generalization, 154, 249, 316
262-271; examining the development generalization probes, 249
of, 261 generalized identity matching, 68
errorless learning procedures, 85, 119 generalized imitation: mirror protocol
establishing operations (EOs), 80; and, 58-59; observing responses and,
captured, 84, 91-92; contrived, 84-85, 56-59; teaching skills in, 20-22
93, 96-97; mand training and, 82-83, generalized relational repertoires, 122
84-87, 90 generalized response patterns, 284
exclusion procedure, 179, 181 generalized symmetry repertoire, 122
exclusion prompts, 118 gestures: mutual object orienting with,
exemplar training, 153 72; nonrelational MTS prompts as,
explicit flexibility training, 361-363 116-117
explicit name training, 358, 359 global empathy, 302
expressive language, 43 graduated time delay, 87
external contextual control, 228-230 grammatical classes, 219-220
eye contact, 18-20 graph-to-formula relations, 317
Greer, R. Douglas, 41
Gutiérrez-Martinez, Olga, 301

H prerequisite skills for, 63-69; protocol
for establishing, 71-72. See also social
hand-over-hand prompting, 116 referencing
Hayes, Steven C., xi joint object attention, 72
here-there relations, 291
Hernandez, Monica, 335
hierarchy, 166-169; arbitrary contexts
K
and, 169; recommendations for Keohane, Dolleen-Day, 41
teaching, 167-168; training categories knowing self, 338-339
for, 168-169
Higbee, Thomas S., 7 L
Holland, James, 313 language: expressive, 43; functional
abilities in, 81; numerical sequencing
I and, 211-216; observing responses
I LAUGH model, 304 and, 42, 44-48, 59; productivity in,
identity matching, 68, 187, 315 209, 210-211; response topography
if-then relations, 124 for, 81-82; semantic relations in, 211;
imitation: definition of, 56; mirror Skinner’s analysis of, 79-80; syntactic
protocol and, 58-59; observing correctness in, 209, 210-211
responses and, 56-59; teaching to learn units, 42, 57
students, 20-22 learner readiness skills, 16-23; differential
impulse-control disorders, 336 reinforcement and, 16; eye contact and
incidental teaching, 84, 91 attending behaviors, 18-20; generalized
information, mands for, 83 imitation, 20-22; prompting
instructional control. See direct techniques and, 16-17; sitting behavior,
instructional control 17-18
instructions: increasing complexity of, learning: discrimination, 25-26, 68,
113, 125; multiple-step, 113; one-step, 153, 192; errorless, 85, 119; flexibility
112-113; repeating, 120 and, 354-355; minimizing errors in,
instrumental social referencing, 73 191-192; programming for, 34-35;
intelligence: analogical reasoning and, relational, 1
257; emotional, 301; flexibility and, least to most (LTM) prompt fading, 119
353, 354 LeBlanc, Linda A., 79
interrupted chain procedures, 85, 94-95 listener repertoire, 42, 43
intrastimulus prompts, 117 listener training, 131; data sheet for
intraverbal prompts, 89 testing, 148; establishing frames
intraverbal responses, 130 of coordination via, 136-137; tact
intrusiveness, 118 training vs., 131-132
I-you relations, 286-287, 290-291 Lovaas, O. I., 18, 22
Luciano, Carmen, 149, 301, 335
J
Jackson, Marianne L., 25
M
joint attention, 52, 69-73; deficits Mackay, Harry A., 209
in, 70-71; evidence supporting maintenance probes, 249
training in, 72-73; gaze shifts in, Mañas, Israel, 149
69-70; importance for DRR, 75-76;
Index 379
mands, 82-87; data sheet for teaching, most to least (MTL) prompt fading, 119
104; definition of, 79; derived, 237- motivative augmenting, 342
253; eye contact and, 19; importance MTS. See matching-to-sample (MTS)
and types of, 82-83; procedure for procedures
teaching, 84-87; recommendations for multiple stimulus without replacement
teaching, 90; topography-based, 238; (MSWO) preference assessment, 11-14;
verbal operants and, 80 data sheet used for, 13; guidelines for
manual guidance, 116 conducting, 12, 14; instructions for
mass trialing, 120-121 scoring, 14
matching by exclusion, 179 Multiple Tact: Category Mixed data
matching skills, 52 sheet, 147
matching-to-sample (MTS) procedures: multiple-exemplar training (MET):
constructed-response, 179; contextually behavioral interventions and, 150;
controlled, 366, 367; description of, bidirectional responding and, 152;
115; discrimination learning and, coordination relations and, 156;
25-26, 64; exclusion variation of, empathic responding and, 307, 308-
179, 181; frames of coordination and, 309; mathematics instruction and, 317,
133-134; general requirements for, 188- 318-327; naming acquired through,
190; nonrelational, 115-119; reading 131-132; relational instruction
instruction using, 175-179, 187-191, following and, 125
203; specific requirements for, 190- multiple-step instructions, 113
191; spelling instruction using, 185- Murphy, Carol, 353
186; stimulus equivalence and, 315; mutual entailment: comparison and, 164;
symmetry/asymmetry training and, coordination and, 155; distinction and,
364-366, 367 162; hierarchy and, 167; opposition
mathematics instruction, 313-332; and, 160
behavioral interventions for, 314-317; mutual object orienting, 72
constructivist approach to, 313-314;
learner’s data set based on, 327-329; N
pretraining process for, 317, 331-332; name-object relation of coordination, 283
stimulus equivalence procedures for, name-picture matching, 242, 243, 244,
314-317; Web-based protocols for, 317- 245
329 name-text matching, 242, 244, 245
McCuller, Glen, 313 naming, 130-132; as bidirectional
McGinty, Jennifer, 313 relation, 130-131; coordination
McHugh, Louise, 281 relations and, 156-157; importance of
MET. See multiple-exemplar training skills in, 140-141; recommendations
metacognition, 336 for establishing, 132-133; stimulus
Miguel, Caio F., 129 categorization and, 137-140; teaching
milepost skills, 37 to children, 131-132
mindblindness, 281 Natural Language Paradigm (NLP), 89,
mirror protocol, 58-59 102-103
mixed identity matching, 68 Neises, Anna, 2
mixed name-picture/name-text matching, Ninness, Chris, 313
244, 245 Ninness, Robin, 313
modeling: nonrelational MTS prompts, no-help trials, 316
116; teaching distinguished from, 42

nonarbitrary stimuli, 154; comparison sensory matching capacity and, 53-56;
and, 165; coordination and, 156; verbal development and, 59; visual
distinction and, 162-163; hierarchy stimuli as conditioning for, 48-52
and, 167-168; opposition and, 160-161 O’Connor, Jennifer, 353
nondifferential observing response, 190 O’Hora, Denis, 111
nonrelational discriminations, 26 one-step instructions, 112-113
nonrelational instruction following, onsets, 183, 184
112-114; observing responses and, opposition, 159-161; nonarbitrary stimuli
120; recommendations for teaching, and, 160-161; transformations of
114; relevance of, 113-114; repeating function and, 159, 160, 161
instructions in, 120; types of, 112-113 O’Toole, Catriona, 353
nonrelational instructional control, 112- overlapping sequences, 222
122; instruction following in, 112- overlapping word sets, 197-198
114; matching to sample in, 115-119;
relational instructional control vs., P
123; relevance to generalized relational paired stimulus preference assessment,
repertoires, 122 8-11; data sheet used for, 10;
nonrelational matching to sample, 115- instructions for scoring, 9, 11;
119; common comparison stimuli procedures for conducting, 9
in, 115; mass trialing used in, 120- partial physical guidance, 116
121; observing responses and, 120; Pavlovian conditioning procedure, 45
prompt- fading procedures in, 118- PECS (picture exchange communication
119; prompting procedures in, 116- systems), 81, 238, 241
118; recommendations for teaching, Peláez, Martha, 63
121; repeating instructions in, 120; perspective taking, 281-299; age-based
variations on presenting sample stimuli expectations of competence in, 296-
in, 115-116 297; core relational skills involved in,
nonverbal imitation, 20-21 285; derived relational responding and,
now-then relations, 293-294 283-284; developmental disabilities
numerical sequencing, 211-216; and, 297-298; double reversed relations
contextual control of, 226-227; cross- and, 292; emotions and, 298-
sequence substitutions and, 220-221; 299;empathy deficits and, 303-304;
grammatical classes and, 219-220; factors affecting learning of, 295-297;
order reversal in, 227; overlapping, here-there relations and, 291; I- you
222; syntactic transformations relations and, 286-287, 290-291; now-
and, 222-230; tests for emergent then relations and, 293-294; overview
recombinations in, 227-228 of approaches to, 281-282; relational,
284-298; task presentation and, 295-
O 296; theory of mind and, 281, 282-
object-name relations, 283 283, 298, 303-304; training sequence
observing responses, 41-59; adult voices as for, 288-290
conditioning for, 46-48; development pervasive mathematics disorder, 314
of very early, 44-48; discrimination of Petursdottir, Anna I., 129
sample stimuli and, 190; nonrelational phonological awareness, 183-184
instructional control and, 120; print physical guidance, 116
stimuli as conditioning for, 52-53;
Index 381
picture exchange communication systems R
(PECS), 81, 238, 241
play-based training, 102-103 readiness skills. See learner readiness skills
pliance, 309, 341, 344-345 reading, 173-203; assessment of, 199,
position prompts, 17, 117 203; comprehension of text, 174,
positive reinforcement, 7 184; cumulative baseline for, 195;
post-test probes: of derived mands for curriculum for teaching, 193-199,
daily living skills, 248-249; of derived 200; errors based on stimulus
mands for preferred items, 242, 244; control, 186; matching-to-sample
of topography-based vocal derived procedures for, 175-179, 187-191, 203;
mands, 250-251 minimizing errors in, 191-192; multiple
predictive assessments, 33 components of, 173-175; overlapping
preference assessments. See stimulus word sets and, 197-198; phonological
preference assessments awareness and, 183-184; recombinative
prelisteners, 41, 44, 45 generalization and, 182, 184, 197;
prepositions, 100 sample sequence for teaching word sets
pre-test probes: of derived mands for in, 199, 200-201; spelling related to,
daily living skills, 248-249; of derived 175; stimulus equivalence and, 176-
mands for preferred items, 239-240, 179, 180; successive word sets and,
242; of topography-based vocal derived 195, 196; verbal relations involved in,
mands, 250-251 203; vocabulary expansion and, 179-
pretraining for math relations, 317, 331- 185
332 reasoning: analogical, 257-278;
print stimuli, 52-53 mathematical, 313-332
private events, 347-349 receptive body parts, 113
probes: generalization, 249; maintenance, receptive commands, 112
249; post-test, 242, 244, 248-249, 250- receptive discrimination, 87, 131
251; pre-test, 239-240, 242, 248-249, receptive labeling, 87
250-251; stimulus relations, 240-241, receptive vocabulary, 87
250-251 Receptive-Expressive Emergent Language
production responses, 41 Scale (REEL-2), 64
productivity, 209, 231; characterization recombinative generalization, 182, 184,
of, 210-211; numerical sequencing and, 197
211-216 referential behavior, 1
Programme for International Student reflexivity training, 68, 315
Assessment (PISA) survey, 314 Rehfeldt, Ruth Anne, 237
programming for learning, 34-35 reinforcement: conjugate, 48-49;
progressive education, 313 differential, 16, 188; positive, 7
prompt-fading procedures, 118-119 reinforcer identification strategies, 7-15;
prompting techniques: differential answers to common questions about,
reinforcement and, 16-17; nonrelational 14-15; multiple stimulus without
matching to sample and, 116-118 replacement preference assessment,
proportional analogies, 272, 274 11-14; paired stimulus preference
proximity prompts: fading procedures for, assessment, 8-11
118; nonrelational MTS training and, relational cues, 153
117
pure mands, 83

relational discriminations: description repetition: nonrelational instructional
of, 26; recommendations for learning, control and, 120; of vocalizations
153-154 (echolalia), 22
relational flexibility, 354; explicit training respondent training, 316
in, 361-363; future directions in response prompts, 16-17
training, 369; protocol for establishing, reversed here-there relations, 291
355-360; symmetry/asymmetry reversed I-you relations, 290
training in, 356, 358-360, 363-369 reversed now-then relations, 294
relational frame theory (RFT): analogical RFT. See relational frame theory
reasoning and, 259-261; basic relational rigidity, 354-355
operants in, 154-169; communication rimes, 183, 184
repertoires and, 237-253; derived Rodríguez, Miguel, 149
stimulus relations and, 151-152; Rosales, Rocio, 237
empathic behavior and, 305-309; Ruiz, Francisco, 149
flexibility in, 354; generalized class of rule-governed behavior: effectiveness of,
symmetry in, 122; perspective taking 347-349; guidelines for developing,
and, 283-298; recommendations for 343-349; instructional control and,
training based on, 153-154; rule- 124-125; relational frame theory and,
governed behavior and, 124-125, 339- 124-125, 339-343; verbal regulation
343; three selves in, 337-339 and, 340, 341-343
relational frames, 154-169; of comparison, Rumph, Robin, 313
164-166; of coordination, 154-159; of
distinction, 162-163; of hierarchy, 166- S
169; of opposition, 159-161 Sally-Anne Test, 303
relational instructional control, 123-125; sample stimuli: teaching discrimination
establishment of, 123-125; increasing of, 190-191; variations on presenting,
complexity in, 125; nonrelational 115-116
instructional control vs., 123; Sautter, Rachael A., 79
recommendations for teaching, 125 say-do correspondence, 337, 344
relational learning, 1 secant amplitude relations, 324-327
relational operants: recommendations for secant frequency relations, 327
training, 153-154; training based on see-and-do correspondence, 58
RFT, 154-169 selection-based derived mands, 239-249;
relational perspective taking, 284- for daily living skills, 244-249; for
298; core relational skills in, 285; preferred items, 239-244
developmental disabilities and, 297- selection-based systems, 81
298; emotions and, 298-299; factors self-as-content, 338
affecting learning of, 295-297; training self-as-context, 339
sequence for, 286-295. See also self-as-process, 338-339
perspective taking self-awareness, 42-43
relational responding: arbitrarily self-directed behavior: effectiveness of,
applicable, 283; discrimination 347-349; guidelines for developing,
learning and, 26; empathic behavior 343-349; skills required for, 343-344
as, 305-309; flexible, 353-354, 369; self-instructional training, 336
intelligent behavior and, 353-354 self-rules, 336; contingency-shaped
behavior training and, 337; effective
Index 383
self- directed behavior and, 347-349; stimulus control: inaccurate reading and
guidelines for the development of, spelling as, 186; transfer of, 86-87,
343-349; relational frame theory and, 88-90
337-339; rule-governed behavior and, stimulus equivalence procedures: math
339-343; self-instructional training instruction and, 314-317; reading
and, 336; verbal regulation and, 341- instruction and, 176-179, 180
343 stimulus generalization, 249
semantic relations, 211 stimulus preference assessments (SPAs),
sense making, 158-159 7-15; answers to common questions
sensory matching: conditioning protocol, about, 14-15; multiple stimulus
54-56; developing the capacity for, without replacement method, 11-14;
53-56 paired stimulus method, 8-11
sensory matching protocol, 54-56 stimulus prompts, 17
shaping, 16, 86 stimulus relations probes, 240-241, 250-
Sidman, Murray, 1, 176-179, 182, 193 251
sign language, 81-82 successive discrimination, 189
simple here-there relations, 291 successive word sets, 195
simple I-you relations, 286 syllable recombination, 184
simple mands, 82-83 symbolic behavior, 1
simple now-then relations, 293-294 symbolic relations, 187
single item presentation method, 8 symmetry relations, 355, 356, 358
sitting behavior, 17-18 symmetry training: analogical reasoning
Situations, Options, Choices, Strategies, and, 268-269; contextually controlled,
and Stimulations (SOCSS) program, 358, 360, 363-364, 366; explicit
304 name training and, 358, 359;
Skinner, B. F., ix, 41, 43, 79-80, 130 flexibility training and, 363-369; math
social referencing, 73-76; importance instruction and, 315; MTS procedures
for DRR, 75-76; instrumental vs. and, 364-366, 367; sample data sets
affective, 73; paradigm for training in, from, 366, 368; schematic overview of,
75; prerequisite skills for, 63-69. See 356
also joint attention syntactic correctness, 209, 210-211
social reinforcement, 154 syntax: characterization of, 210-211;
social reinforcers, 71 correct use of, 209, 210-211; embedded
“social stories” procedure, 303 contextual control of, 222-228;
Socratic dialog, 336 extended repertoires of, 222; external
SPAs. See stimulus preference assessments contextual control of, 228-230; new
speaker training, 135-136 utterances using, 222; numerical
speaker-as-own-listener repertoire, 42, 44 sequencing and, 211-216, 220-222;
speaker-listener repertoire, 42, 43 productivity of, 209, 210-211, 231;
spelling: errors based on stimulus control, transformations of, 222-230
186; matching-to-sample procedures
for, 185-186; reading related to, 175 T
spontaneous requests, 87 tact training: data sheet for, 105; derived
standard analogies, 272, 273-274 categorization skills and, 138-139;
Stewart, Ian, 257 general recommendations for, 90;
stimulus array, 11 naming skills and, 131-133; play-

based, 102-103; procedures for, 88-90, verbal behavior. See language
98-105, 131-132 Verbal Behavior Milestones Assessment
tacts, 87-88; definitions of, 80, 87; and Placement Program (VB-MAPP),
importance and types of, 87-88; 81
procedures for teaching, 88-90, Verbal Behavior (Skinner), 150
98-105, 131-132; recommendations for verbal developmental cusps, 44
teaching, 90; sensory experiences and, verbal imitation, 20, 21-22
101; verbal operants and, 130 verbal operants, 80, 130
Tarbox, Jonathan, 111 verbal regulation, 340, 341-343
Tarbox, Rachel S. F., 111 Vineland Adaptive Behavior Scale, 33
Teach Me Language program, 304 Visdómine, Carmelo, 301
teaching: constructivist, 313-314; visual stimuli, for conditioning observing
incidental, 84; modeling vs., 42 responses, 48-52
Teaching Individuals with Developmental visual tracking protocol, 49-51
Delays (Lovaas), 18, 22 visual-visual nonidentity match
terminating aversive stimuli, 83, 84, 92 (VVNM), 34
testing: analogical reasoning, 262-271, vocabulary expansion, 179-185
272, 275-278; categorization, 146-148; vocal derived mands, 249-252; pre- and
listener, 148; multiple-exemplar, 318- post-test probes for, 250-251; teaching
327; symmetry, 268-269 using category names, 251-252
textual behavior, 80, 173-174 vocal imitation, 45
theory of mind (ToM): deficits in vocal stimuli, for conditioning observing
empathy and, 303-304; perspective responses, 46-48
taking and, 281, 282-283, 298 voice output communication aids
three-by-three matrix, 176, 177 (VOCA), 81
time-delay procedure, 17, 86-87, 91
topography-based systems, 81 W
topography-based vocal derived mands, Web-based math training, 317-329;
249-252; pre- and post-test probes learner’s data set based on, 327-329;
for, 250-251; teaching using category multiple-exemplar process for, 318-327;
names, 251-252 pretraining process for, 317, 331-332;
touch-screen response training, 67-68 protocol for, 317-327
tracking, 309, 341-342, 345-346 Wechsler Adult Intelligence Scale-Revised
transfer of stimulus control: mand (WAIS-R), 33
training and, 86-87; tact training and, Weil, Tim, 257
88-90 “What’s in the Bag?” game, 96-97
transfer of training, 216 whole math, 313
transformations of function: comparison Williams, W. Larry, 25
and, 164, 165-166; coordination within-stimulus prompts, 117
and, 155, 157-158; distinction and, word recognition, 174
162, 163; hierarchy and, 167, 168; word sets: overlapping, 197-198; sample
opposition and, 159, 160, 161 sequence for teaching, 199, 200-201;
transitivity, 315 successive, 195, 196
V
Valdivia-Salas, Sonsoles, 149, 301, 335
Index 385
more tools for your practice from newharbingerpublications, inc.
THE ABCs OF HUMAN ACT IN PRACTICE AUTISM

BEHAVIOR Case Conceptualization in Behavior Analytic Perspectives
Behavioral Principles for the Acceptance & Commitment US $39.95 / ISBN: 978-1878978325
Practicing Clinician Therapy
CONTEXT PRESS
US $49.95 / ISBN: 978-1572245389 US $59.95 / ISBN: 978-1572244788 WWW.CONTEXTPRESS.COM
ACCEPTANCE & ACCEPTANCE & NEW DIRECTIONS

COMMITMENT THERAPY MINDFULNESS IN BEHAVIOR
FOR THE TREATMENT OF TREATMENTS FOR DEVELOPMENT
POST-TRAUMATIC STRESS CHILDREN & New Directions in Behavior
DISORDER & TRAUMA- ADOLESCENTS Development
RELATED PROBLEMS A Practitioner’s Guide US $29.95 / ISBN: 978-1878978240
A Practitioner’s Guide to US $59.95 / ISBN: 978-1572245419 CONTEXT PRESS
Using Mindfulness & WWW.CONTEXTPRESS.COM
Acceptance Strategies
US $58.95 / ISBN: 978-1572244726
available from
r
n e w h a r b i n g e r p u b l i c a t i o n s, i n c . l e tte
and fine booksellers everywhere ws om
e Ne er.c
r g
To order, call toll free 1-800-748-6273 or visit ou in
our online bookstore at www.newharbinger.com f or arb
p h
(VISA, MC, AMEX / prices subject to change without notice) i g n u t new
S a

Derived Relational Responding Applications For Learners (2009) PDF

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Derived Relational Responding Applications For Learners (2009) PDF

Încărcat de

Drepturi de autor:

Formate disponibile

Derived Relational Responding

Applications for Learners

A Progressive Guide to Change

Distributed in Canada by Raincoast Books

Copyright © 2009 by Ruth Anne Rehfeldt and Yvonne Barnes-Holmes

All Rights Reserved

Acquired by Catharine Sutker; Cover design by Amy Shoup;

A copublication of New Harbinger Publications and Noetic Books

Library of Congress Cataloging-in-Publication Data

10 9 8 7 6 5 4 3 2 1 First printing

Introduction and Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 1

iv Derived Relational Responding

vi Derived Relational Responding

 Have an adequate database, appropriate to the strength of the claims

 Orient the reader toward unresolved empirical issues.

 Do not overlap needlessly with existing volumes.

 Keep the focus always on what is good for the reader.

 Support the further development of the field.

 Provide information in a way that is of practical use to readers.

viii Derived Relational Responding

An Applied Behavioral Psychology of

x Derived Relational Responding

Ruth Anne Rehfeldt, Ph.D., BCBA Yvonne Barnes-Holmes, Ph.D.

2 Derived Relational Responding

Introduction and Acknowledgments 3

Establishing the Prerequisites

Reinforcer Identification Strategies and

Reinforcer Identification Strategies

Paired Stimulus (Forced Choice) Method

8 Derived Relational Responding

Student: Assessed by: Date: Time

Record item with corresponding item number: Circle item selected:

1. ÷ x 100 = %

10 Derived Relational Responding

Multiple Stimulus Without Replacement Method

12 Derived Relational Responding

Student: Assessed by: Date: Time

Answers to Common Questions About

14 Derived Relational Responding

Q: Do I need to use the actual items or activities in the preference assessment? Is it

Summary: Strategies for Reinforcer Identification

Differential Reinforcement and Prompting Techniques

16 Derived Relational Responding

Teaching Appropriate Sitting Behavior

Teaching Eye Contact and Other Attending Behaviors

18 Derived Relational Responding

2. Before presenting an instruction, require the student to look at you or at the

6. Provide reinforcement when the student responds correctly to the direction

7. Also provide reinforcement when eye contact is provided spontaneously during

Teaching Generalized Imitation Skills

20 Derived Relational Responding

3. If the target response involves the manipulation of an object, include distracter

4. Before presenting an instruction, require the student to look at you or at the

8. Fade prompts as quickly as possible.

22 Derived Relational Responding

24 Derived Relational Responding

The Assessment of Basic Learning Abilities

W. Larry Williams and Marianne L. Jackson,

The ABLA and Conditional Discriminations

26 Derived Relational Responding

ABLA Procedures and Materials

28 Derived Relational Responding

Level Materials Instruction Correct Response

Have an adequate database, appropriate to the strength of the claims

Orient the reader toward unresolved empirical issues.

Do not overlap needlessly with existing volumes.

Keep the focus always on what is good for the reader.

Support the further development of the field.

Provide information in a way that is of practical use to readers.

2 Yellow can and red box, “Where does Individual

3 Yellow can and red box, “Where does Individual takes

5 Yellow can and red box “Put it in the Individual places

6 Yellow can and red box “Put it in the Individual places

Try to make training sessions as short as possible. Fifteen minutes is the

Try to ensure that interventions are maximally effective by further

Where there are several comparison stimuli, their locations should be

Reinforcement is provided contingently and immediately on all correct

Trials in which the learner makes an incorrect response are not

Additional interventions (such as shaping and task analyses) are usually

Each training block is generally followed by a single block of five ran-

Prompts can be used throughout all training trials. However, it is impor-

Another important aspect to consider is whether the learner can name