Inteligence Quotiend

NEUROSCIENCE RESEARCH PROGRESS
INTELLIGENCE QUOTIENT
TESTING, ROLE OF GENETICS
AND THE ENVIRONMENT
AND SOCIAL OUTCOMES
No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.

Additional books in this series can be found on Novas website
under the Series tab.
Additional e-books in this series can be found on Novas website
under the e-book tab.
INTELLIGENCE QUOTIENT
TESTING, ROLE OF GENETICS
AND THE ENVIRONMENT
AND SOCIAL OUTCOMES
JOSEPH C. KUSH
EDITOR
New York
Copyright 2013 by Nova Science Publishers, Inc.

All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.
For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com
NOTICE TO THE READER
The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or
implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of
information contained in this book. The Publisher shall not be liable for any special,
consequential, or exemplary damages resulting, in whole or in part, from the readers use of, or
reliance upon, this material. Any parts of this book based on government reports are so indicated
and copyright is claimed for those parts to the extent applicable to compilations of such works.
Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.
This publication is designed to provide accurate and authoritative information with regard to the
subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.
Additional color graphics may be available in the e-book version of this book.
Library of Congress Cataloging-in-Publication Data

ISBN: (eBook)
Library of Congress Control Number: 2013937394
Published by Nova Science Publishers, Inc. New York
CONTENTS
Preface
Chapter 1
Chapter 2
Chapter 3
Chapter 4
vii
The Utilization of Inspection Time as a Supplement
to Standardized Tests of Intelligence:
Arriving Somewhere but Not Here
Joseph C. Kush
IQ and the D2 Dopamine Receptor Gene in Children
of Alcoholics
Ernest P. Noble
17
Human General Intelligence as a Domain General

Psychological Adaptation
Kevin MacDonald
35
Subtests, Factors, and Constructs: What Is Being Measured

by Tests of Intelligence?
Marley W. Watkins and Christina M. Ravert
55
Chapter 5
Race, Genomics and Intelligence: Slight Return

Joseph L. Graves Jr.
Chapter 6
Multi-group Latent Growth Curve Modeling: The Influence

of School Readiness Profiles on Student Achievement Growth
Timothy R. Konold and Monika Townsend
Chapter 7
Chapter 8
Chapter 9
69
87
Error in the Measurement of Low IQ: Implications

for the Diagnosis of Intellectual Disability in Court Cases
Simon Whitaker
111
Using R for the Analysis of Cognitive Ability and Behavior

Genetic Data
A. Alexander Beaujean and Jason Parkin
129
Contribution of Intellectual, Psychological, Developmental

and Socio-economic Data to Highlight Specific Profiles of Highly
Gifted Children
Isabelle Simoes Loureiro, Laurent Lefebvre
and Laurence Vaivre-Douret
169
vi
Chapter 10
Index
Contents
Preschoolers Attachment Representations and Parenting
as Predictors of Intellectual Capacities:
A Person-oriented Approach
Marie Stivenart, Jean Christophe Meunier
and Isabelle Roskam
185
201
PREFACE
The debate regarding the role of genetic and environmental influences on tests of
intelligence has been ongoing for centuries yet particularly relevant today. Modern advances
in genetic mapping and functional magnetic imaging offer researchers the opportunity to
carefully study marker variables with technologies made available only recently. These
techniques, and their implications for IQ testing are carefully discussed in these chapters.
Similarly, sophisticated, advanced, multivariate statistical techniques are also employed by
many of the researchers who contributed to this volume, allowing them to reach conclusions
and make recommendations unavailable to previous generations of researchers. However, in
many ways, this book offers a caution to readers that the use, and sometimes abuse, of IQ
measures is growing at a momentous rate and the use of IQ tests should occur in a more
deliberate, more thoughtful manner.
Each of the contributors to this text recognize the clinical utility and value of IQ tests and
fully support their continue use. The authors are in no way IQ apologists. However, the
authors are conservative in our recommendations and present a critical discussion of possible
misuses of the practices associated with IQ testing. First, and most importantly IQ tests
should not be equated with the construct of intelligence. Current IQ tests do measure
intelligence but often also measure other non-cognitive factors as well. Second, IT test
performance is influenced by social, environmental, cultural, as well as genetic factors and
the authors are only beginning to be able to untangle the unique contributions of these
variables. As a result, the conclusions that are made and the diagnostic decisions that
accompany IQ test administration should be tempered against data-driven theory and bestpractice approaches. Particularly when discussing the role of genetic and environmental
influences on IQ tests it is very important to not step beyond what can be supported by data
and the contributors to this text were mindful of this throughout all stages of their writing.
The book avoids any attempt to make a claim regarding exact estimates of the genetic or
environmental influences on measures of IQ, fully recognizing the complex interplay between
these factors.
The foundation of the scientific method suggests that scientific disciplines should create
knowledge that can be used to explain, predict and control the variables thought valuable by
the discipline. The better an explanation is at making a valid prediction, the more useful the
explanation will be to the scientific discipline. Identifying the variables that influence IQ and
describing how and why they can be used to make predictions is where our chapters have the
most to offer. There is much the authors currently know about genetic and environmental
influences on IQ tests and if the authors are vigilant, they can make very accurate predictions
viii
Joseph C. Kush
about human behavior. However the authors of this text wisely recognize that in most cases,
to go beyond this point to making recommendations for public policy (e.g., controlling
behavior) is a position that cannot yet be fully supported by extant data. Certainly the authors
support the creation of new and rival hypotheses that can be reviewed and critiqued by others
and challenged for replication. This is how science advances. Additionally, many of the
chapters in this text identify explanatory variables related to IQ tests that can assist in
diagnostic decision-making that will continue to incrementally predict academic achievement,
occupational performance and clinical diagnoses.
Public policy implications of IQ tests remains are one of the greatest sources of
controversy in the field of psychology. Given the hostility in the current media against IQ
tests, the authors recognize the critical importance of only making claims that can be
supported by well-reasoned theory and peer-reviewed, replicated research studies. To make
these unfounded claims would be a disservice to the giants in the study of human intelligence
whose shoulders the authors stand upon. However, this book does advance a discussion
beyond the nature nurture debate. The authors know that DNA is both inherited and
influenced by the environment. Genetic differences exist among individuals however
continued research must more carefully examine the complex interplay between genes and the
environment. The scholarly value of this book therefore is that it only addresses issues that
are known and supported by empirical data. The authors believe this will subsequently lead to
a more learned and informed discussion about how to use this knowledge about IQ tests to
benefit the individuals who are assessed by them.
In: Intelligence Quotient

Editor: Joseph C. Kush
ISBN: 978-1-62618-728-3
2013 Nova Science Publishers, Inc.
Chapter 1
THE UTILIZATION OF INSPECTION TIME

AS A SUPPLEMENT TO STANDARDIZED TESTS
OF INTELLIGENCE: ARRIVING SOMEWHERE
BUT NOT HERE
Joseph C. Kush*
Duquesne University, Pittsburgh, PA, US
ABSTRACT
Psychologists continue to debate how best to define human intelligence. There is a
general consensus among theoreticians that intelligence consists of a general, or
overarching ability that underlies all aspects of the construct, as well as smaller,
individual abilities such as verbal ability and abstract reasoning. Although many
commercial tests of intelligence claim to be modeled upon established theories of
intelligence, there is growing evidence that this may not always be the case. The
underlying factor structure of many IQ tests is not always invariant across groups and
does not always align with its theoretical origins. Additionally, IQ tests are increasingly
adding subtests that include non-intellectual aspects such as tasks that are better described
as measures of learning than measures of intelligence. Because many current IQ tests
measure intelligence and other factors, the result is an inflated correlation between IQ
tests and tests of achievement. An alternative approach to traditional IQ testing has
included an examination of brain functioning including blood flow or other physiological
indices as well as the assessment of elementary tests of cognitive abilities. These
measures have much better face validity than commercial IQ tests however their
predictive validity is somewhat lower. This chapter highlights one specific ECT measure,
inspection time, and describes a process where an inspection time task can be combined
with traditional IQ tests to produce a battery that may provide improved diagnostic
utility, as well as a closer match to existing theories of human intelligence.
*
Correspondence concerning this chapter should be addressed to Joseph C. Kush, Department of Instruction and
Leadership in Education, 327 Fisher Hall, Duquesne University, 600 Forbes Avenue, Pittsburgh, PA 15282.
kush@duq.edu.
Joseph C. Kush
Individual tests of intelligence (IQ tests) were originally designed to assist with the
identification of children who were at-risk for school difficulties (Binet, 1905) and the earliest
IQ measures included components that analyzed school examination scores for children
within the Paris school system (Spearman, 1904). Millions of intelligence tests are
administered each year to children in the United States as part of special education eligibility
evaluations (Gregory, 2004; Kamphaus, Petoskey, & Rowe, 2000) and are among the most
commonly used instruments used by psychologists in both clinical and educational settings
(Wilson & Reschly, 1996). IQ tests are typically used as part of a diagnostic battery to predict
occupational success or academic attainment, including school grades or standardized tests of
academic achievement, with typical IQ-academic correlations ranging from 0.40 to 0.70
(Mackintosh, 2012) with a median correlation of approximately 0.50 (Jensen, 1998). In fact,
tests of intelligence often predict academic achievement with more accuracy than other tests
of academic achievement. For example, STEM achievement has been shown to be better
predicted by tests of cognitive ability than by achievement test scores in language or literacy
(Deary, Strand, Smith & Fernandes, 2007; Lu, Weber, Spinath & Shi, 2011). However, not all
IQ tests measure the same thing (Mackintosh, 1998).
While most discussions of intelligence testing often begin around 1900 with the
publication of the Stanford-Binet, in fact the assessment of human abilities can be traced
much earlier in history. As early as 2200 B. C., individuals in China who wanted to work in
public office were required to complete a battery of tests that often took several days to
complete. The process started at the province level and top scoring applicants moved on to
Peking for final examinations. This practice continued until around 1906 and has often served
until present times as the model for identifying civil servants in many Western countries.
Even Plato referenced intelligence, using a metaphor of hard and soft wax to describe rigid
and flexible thinking. Intelligence tests continue to be strongly related to occupational, as well
as educational, economic and social outcomes. Today, a high IQ is seen as necessary in
highly complex, professional or managerial jobs, but is less of an advantage in occupations
that require simple, unskilled problem solving (Gottfredson, 1997). It is important to
remember however, that many other factors beyond intelligence influence performance and it
is the interaction of intelligence and these factors (e.g., motivation, persistence, personality)
that contributes to success in occupational and educational settings (Jensen, 1998).
DEFINING HUMAN INTELLIGENCE

Psychologists have long disputed the definition of intelligence. The British psychologist
Charles Spearman (1904, 1923) introduced the concept of g, or general intelligence, to
explain the common attributes or the overlap of skills he found on many tests of cognitive
ability. In contrast, the American psychologist Louis Thurstone (1924, 1938) argued in favor
of seven independent factors that he termed primary abilities. While most current
psychologists who study human intelligence acknowledge the existence of g, there remains a
great disparity of thought considering the nature and structure of g. The debate focuses on the
disagreement on how many kinds of intelligence exist; is there simply one type of
intelligence a common factor that underlies all of our cognitive processes or are there
multiple, discreet cognitive skills such as verbal ability, numerical ability, and spatial ability?
The Utilization of Inspection Time as a Supplement to Standardized Tests ...
The argument parallels an argument (probably better debated in a bar) concerning the
definition of athletic ability. Is it correct to consider a single unitary construct called athletic
ability, that all humans possess, (although the amount that each of us possesses varies)?
Proponents would argue that athletic ability is typified by the Olympic decathlon champion
who is agile, swift, and strong, or by an athlete like Michael Jordan who was a great
basketball player, scratch golfer, and played minor league, professional baseball. The counterargument emphasizes sport-specific skills and counters that a 330-pound, football, lineman
who is strong but perhaps cant jump high or swim must certainly still be considered to have
some type of athletic ability. Does it make better sense to consider athletic ability as a single
factor, which underlies many unique interdependent skills or should we consider strength,
speed, and endurance as separate entities? Best set-aside for now and reexamined later over a
pint.
g
Classics
French
English
Mathematics
Discrimination
Music
Figure 1. A model of g or general intelligence.
Perhaps the best contemporary definition of human intelligence was offered by a panel of
52 of the most influential researchers in the field of intelligence, coordinated by Linda
Gottfredson (1997), Intelligence is a very general mental capability that, among other things,
involves the ability to reason, plan, solve problems, think abstractly, comprehend complex
ideas, learn quickly and learn from experience. It is not merely book learning, a narrow
academic skill, or test-taking smarts. Rather, it reflects a broader and deeper capability for
comprehending our surroundings-catching on, making sense of things, or figuring out
what to do. Intelligence, so defined, can be measured, and intelligence tests measure it well.
They are among the most accurate (in technical terms, reliable and valid) of all psychological
tests and assessments (p. 13).
There is a general consensus today, among scientists that study human intelligence, that a
hierarchical model provides the most valid and comprehensive framework (Deary, 2001;
Neisser et al., 1996) and there currently exists three major models of intelligence that are
supported by empirical data. Specifically, Horn and Cattells fluid-crystallized (Gf-Gc) model
(Cattell, 1963; Horn & Noll, 1997); the Cattell-Horn-Carroll (CHC) model (Benson, Hulac, &
Kranzler, 2010; Carroll, 1993, 2003; Flanagan, Ortiz, & Alfonso, 2007); and a verbalperceptual-image rotation (VPR) model (Vernon, 1961, 1965), have all been used to describe
the underlying structure of human intelligence.
Each model offers unique strengths and weaknesses in attempting to explain how humans
process information however the biggest distinction centers on the existence (or not) of g, the
general factor of intelligence a debate that has been argued by British and American
psychologists for over a century. This debate continues to be reflected in current-day models
Joseph C. Kush
with the CHC and VPR models both acknowledging the existence of g, while the Gf-Gc
model rejects its existence. It should be noted that Cattells early writings with Horn
supported the notion that g did not exist, however his views evolved to include g with the
introduction of the CHC theory that supplemented his earlier fluid-crystallized (Gf-Gc) model
(Cattell, 1963; Horn & Noll, 1997). In many ways, the CHC model can be seen as an
expansion of the original Gf-Gc model, in which Carroll now proposes a three-stratum model
with over 70 narrow abilities at stratum, three, eight broad second-order abilities at stratum
two, and an overall general or g factor at stratum three (Carroll, 1993).
INTELLIGENCE AND IQ TESTS

Although Carroll (1993) has consistently included g in his hierarchical CHC model, most
research examining his theory has focused on first-order broad abilities. The existence of g, as
a causal variable (Jensen, 1998, Spearman, 1927) is the most popular way that the general
factor has been explained, and as a result, the most common hierarchical model that includes
a g factor is a higher order model in which g is included as a second-order factor.
Alternatively, Gignac (2005, 2006) has described a rival direct hierarchical CFA model also
known and a nested factor, or bifactor model, in which g is included as a first-order factor,
with all of the subtests loading directly on g and their respective broad factors. Although in
the minority, there are some researchers that argue that g is simply a summary index that
offers nothing beyond what is explained by Gf (Gustafsson, 1984; Horn & Blankson, 2005;
Ward et al., 2012). Research has consistently demonstrated that the overwhelming majority of
IQ test variance is associated with higher-order g while substantially less variance is
associated with first-order factors (Bodin et al., 2009; Canivez & Watkins, 2010; Watkins,
2006, 2010). These findings lead to a best practice recommendation: the primary
interpretation of IQ scores should occur at the g level rather than the factor index score level
(Canivez & Kush, in press; Canivez & Watkins, 2010; Watkins, 2006, 2010).
While the Gf-Gc model has been described as the most influential contemporary theory
of human intelligence (Mackintosh, 2012), and the CHC model has the most factor analytic
research to support it (Kaufman, 2009; McGrew, 2009), the VPR model is a modern
extension of a theory originally put forward by Vernon (1961). While Vernons original
theory suggested the presence of a general factor and two broad, second-order verbal and
performance factors, Johnson and Bouchard, (2005) found that the addition of a third, secondorder factor (image-rotation), provided a significantly improved factor analytic fit, for a
battery of 60 cognitive ability measures, over the traditional verbal and performance factors.
Most contemporary intelligence tests have attempted to modify their structure to support CHC
theory (e.g., the Wechsler Intelligence Scale for Children-Fourth Edition (WISC-IV;
Wechsler, 2003), the Stanford-Binet Intelligence Scales-Fifth Edition (SB-5; Roid, 2003),
Kaufman Assessment Battery for Children-Second Edition (KABC-II; Kaufman & Kaufman,
2004), Reynolds Intellectual Assessment Scales (RIAS; Reynolds & Kamphaus, 2003), and
Wide Range Intelligence Test (WRIT; Glutting, Adams, & Sheslow, 2000), and the Wechsler
Adult Intelligence Scale-Fourth Edition (WAIS-IV; Carroll, 1993, 2003; Cattell & Horn,
1978; Horn, 1991; Horn & Cattell, 1966).
Comprehension
Information
Matrix
Reasoning
Picture
Concepts
Similarities
Vocabulary
Digit
Span
Letter-Number
Sequencing
Coding
Symbol
Search
Figure 2. A model of g with the Wechsler scales.

g
Gc
Comprehension
Information
Similarities
Gs
Vocabulary
Figure 3. A higher order model using the Wechsler scales.
Coding
Gf
Symbol
Search
Matrix
Reasoning
Gsm
Picture
Concepts
Digit
Span
Letter-Number
Sequencing
Joseph C. Kush
Increasingly however, commercial, tests of intelligence are being criticized for their lack
of a theoretical foundation or a mismatch between the theory of intelligence and the IQ test
(Canivez & Kush, in press; Kush, 1996). For example, while the Wechsler scales imply CHC
constructs, it remains unclear whether a four-or five-factor CHC model best fits the data
(Canivez, in press; Keith, 2005; Niileksela, Reynolds, & Kaufman, 2012; Watkins, 2010;
Watkins, Wilson, Kotz, Carbone, & Babula, 2006). Relatedly, DiStefano and Dombrowski
(2006) and Canivez (2008), using data from the Stanford-Binet Fifth Edition (SB5)
standardization sample, obtained markedly different results for the SB5 than the CFA results
presented in its technical manual (Roid, 2003) and concluded that the SB5 measured only
one essential dimension (g).
In contrast, more restricted or specialized measures of cognitive ability, such as
nonverbal measures of intelligence, possess not only strong psychometric characteristics
(Braden & Athanasiou, 2005) but are supported by well-articulated models of intelligence
(Kranzler, Flores & Coady, 2010). It is important to remember that IQ tests measure
intelligence as well as other factors including personality (Chamorro-Premuzic, Moutafi, &
Furnham, 2005), academic achievement (Kush, 2005; Sternberg, 1998, 1999), as well as the
context of the learning environment (Ceci, 1991). It is this incremental validity that that
provides IQ tests with the broad explanatory power described by Jensen (1998), yet these
same factors detract from the face validity of these measures, as intellectual factors as well
as non-intellectual factors are being assessed.
AN ALTERNATIVE TO TRADITIONAL IQ TESTS;

THE ASSESSMENT OF BASIC COGNITIVE PROCESSES
For this reason, an alternative approach to the study of human intelligence has moved
away from attempting to describe the underlying structure of intelligence, and has focused
instead on the direct measurement of basic cognitive processes. This distinction is what
Cronbach (1957) referred to as the two disciplines of scientific psychology differential
and experimental. Differential psychology is more applied and focuses on individual
differences in intelligence while experimental psychology considers the manipulation of
variables and the subsequent impact on cognitive processing. The study of mental
chronometry or the selection of more pure measures of cognitive ability, has led a number
of researchers to the elementary cognitive task (ECT) paradigm (e.g., Jensen, 2006; Sheppard
& Vernon, 2007). These tasks are seen as distinct from those found on traditional IQ tests,
however they are thought to assess the same features of cognition that underlie the
performance of more complex intelligence test tasks (Partchev & De Boeck, 2012). To date,
most ECT research has focused on the constructs of working memory and processing speed
(Conway, Cowan, Bunting, Therriault, & Minkoff, 2002; Jensen, 1998). Although the
theoretical rationale underlying mental chronometry is over 50 years old, the sophistication
and accuracy of empirical studies has only recently paralleled advances in technology and has
been the result of cross-disciplinary collaboration.
Examples of commonly used ECTs include those that measure processing speed, or
reaction time (RT), and those that measure speed of information intake, or inspection time
(IT; Jensen, 1998, Kush, Spring, & Barkand, 2012). Current research has shown that RT and
IT correlate with standardized measures of intelligence, typically near a value of

approximately -.50, (Burns & Nettelbeck, 2002; Deary, 2000; Grudnik & Kranzler, 2001;
Kranzler & Jensen, 1989; Nettelbeck, 1987, 1998, 2001). Further, behavioral genetic studies
(Luciano et al., 2001; Posthuma et al., 2001) have demonstrated that ECT measures are
genetically correlated with IQ. It remains unclear however, whether ECT measures relate to a
single g factor or a group factor of intelligence (Mackintosh, 2012). Burns and colleagues
(Burns & Nettelbeck, 2003; OConner & Burns, 2003) have demonstrated that IT correlates
with one of the second-order factors from the Gf-Gc theory: general speed of processing (Gs);
the cognitive factor that refers to the speed at which an individual is able to correctly perform
simple or over-learned problems.
Historically, the speed at which a problem can be solved (strategy free) has commonly
been thought of as an index related to overall cognitive ability (Nettelbeck, 1998, 2001).
Examining intelligence as a property of the brain, Gottfredson (2000) has shown that
individuals with higher levels of intelligence work more quickly and efficiently because they
transmit electrical nerve impulses faster and use less glucose. Processing speed has also been
shown to account for about 30% of the variance in academic performance (Carlson & Jensen,
1982; Luo, Thompson & Detterman, 2003). Among ECTs, IT has produced the highest
correlations with IQ and is thought to measure perceptual speed (Deary & Stough, 1996,
Mackintosh & Bennet, 2002). Additionally, Petrill, Dasen, Thompson and Detterman, (2001)
have shown that the variance IT tasks share with IQ is unique from the variance accounted for
by other ECT measures and IQ.
While Vickers (1970, 1979) is typically credited with the original development of IT
research, one of his doctoral students, Nettelbeck, conducted some of the earliest empirical
studies (Nettelbeck & Lally 1976) and has produced a number of additional studies since that
time (Nettelbeck, 1987, 1998, 2001, 2003). In the majority of IT studies, the task begins with
participants focusing their attention on a simple visual cue, such as a large black dot.
Following a brief delay, subjects are briefly presented with a stimulus that resembles the
Greek letter pi (two vertical lines, one long and one short, connected by a horizontal line at
the top). Two target figures are then randomly exposed, one in which the right leg of the pifigure is longer and one in which the left leg of the pi-figure is longer. Immediately following
the presentation of the pi-figures, a large mask covers the stimuli to prevent any storage in
iconic memory. Because the mask works backwards in time, to disrupt the memory of the
previously presented stimulus figure, the mask is termed a backward mask. The ability to
determine which leg of the stimulus is longer is deemed to so simple that given unlimited
time, most subjects can make the identification with 100% accuracy. The briefest stimulus
duration at which a participant can achieve a given accuracy rate (typically between 70% and
95%) is the participant's threshold IT.
The distinguishing factor of the IT paradigm is that the mental time needed to solve the
task is recorded separately from the time needed to physically indicate the correct answer.
Processing speed has been shown to be comprised of distinct factors including movement
time, decision time, and visualization speed (O'Connor & Burns, 2003). Anderson and Miller
(1998), have defined IT as the stimulus exposure duration required by a subject to make a
simple perceptual judgment, for example, the relative length of two lines (p. 239). Because
of the elemental nature and simplicity associated with IT measures, they reflect a stronger
connection to biological processes than do psychometric test scores (Jensen, 2006). The first
interval, IT, reflects the amount of time that the subject needs to cognitively solve the
Joseph C. Kush
problem and is considered more important than is the second interval, which merely reflects
the amount of time the subject needs to motorically indicate their answer. Many theorists
(e.g., Burns & Nettelbeck, 2003; Deary, 2000; Jensen, 2006) consider IT to be a more pure
measure of intelligence than standardized IQ batteries, as the influences of verbal skills,
memory, and socialization are minimized, making the results more culture-fair.
It remains unclear however, whether measures of IT are better characterized as an index
of higher-level cognitive processes such as the speed and efficiency of cognitive processing
(Anderson, 1992; Jensen, 2006; Nettelbeck, 1987, 1998, 2001; Nettelbeck, 1987; Nettelbeck
& Wilson, 2005), or strategy use (Mackintosh, 1986), or if IT tasks reflect more specific
cognitive processes such as attentional control (Fox, Roring, & Mitchum, 2009) or task
consistency (Bors et al., 1999).
CRITERION VALIDITY: USING IQ TESTS TO MAKE PREDICTIONS

While clearly not yet resolved, an ongoing debate about the underlying nature of human
intelligence must continue to gather data from both differential and experimental
methodologies. However as additional factor analytic studies continue to examine the
construct validity of IQ tests it is critical to remember that discussions about the structure of
IQ tests should not be confused with a discussion on the structure of intelligence. Factor
analytic studies are valuable in determining how many types of intelligence are being
measured by an IQ test, but they provide no information about the diagnostic utility of the test
(Canivez, Konold, Collins, & Wilson, 2009; Carroll, 1997). The factor structure of IQ tests
must therefore be examined with relations to external criteria to see how well they perform
(Canivez, in press). Relatedly, it is important to remember Jensens caution from over 30
years ago (1979) that intelligence must be distinguished from both learning and memory.
The prediction of academic achievement is perhaps the primary use of intelligence tests
(Brown, Reynolds, & Whitaker, 1999; Weiss & Prifitera, 1995). Individual tests of
intelligence were originally designed to assist with the identification from children who were
at-risk for school failure (Binet, 1905; Binet & Simon, 1905) and Binets original IQ test
included components that analyzed the school performance of children within the Paris school
system (Spearman, 1904).
Research has consistently demonstrated that intelligence tests are strong predictors of
academic achievement (Brody, 2002; Carroll, 1993; Gottfredson, 2008; Naglieri & Bornstein,
2003) with median individual IQ-achievement correlations typically near .50 (Brody, 2002;
Neisser et al., 1996). In instances where the IQ and achievement tests are co-normed on the
same population, the correlation rises to near .70 (Elliott, 2007; Glutting, Adams, & Sheslow,
2000; Kaufman & Kaufman, 2004; Reynolds & Kamphaus, 2003; Roid, 2003; Wechsler,
2003, 2008; Woodcock, McGrew, & Mather, 2001). Finally, when individual and group IQ
measures are included together, 85% to 90% of predictable criterion variable variance is
accounted for by g (Thorndike, 1986).
As Kush has argued previously (2005), with correlations in the .80 range, approximately
two-thirds of the information contained on IQ test and tests of academic achievement reflects
shared variance or overlapping content, a figure that is simply too high for instruments
thought to be measuring related yet discrete constructs. This view is supported by a number of
researchers (e.g., Gardner, 1983; Gould, 1996; Murdoch, 2007) and by Sternberg (1998,
1999, 2003) who has argued that IQ tests are largely measures of achievement, as they
measure vocabulary, information and arithmetic problem solving. These bloated correlations
can be reduced substantially if only the g portion of the test is used to predict achievement.
For example, Glutting, Watkins, Konold, and McDermott (2006) demonstrated that the
WISCIV FSIQ predicted substantial portions of variance in reading and mathematics scores
although the four factor index scores did not contribute additional meaningful predictive
power.
A POSSIBLE ALTERNATIVE
As a result, Kush, Spring, and Barkand (2012) have concluded that psychologists often
find themselves facing a Catch-22. Commercially published IQ tests are the most frequently
used source for predicting academic achievement (Freberg, Vandiver, Watkins, & Canivez,
2008; Parker & Benedict, 2002; Sattler, 2001) however as mentioned previously, these tests
include non-intellectual factors including reading skills, test-taking strategies and cultural
familiarity. Therefore, the inclusion of these components improves the predictive power of
the instruments because intelligence and other factors are being assessed (Watkins et al.,
2007).
Clearly intelligence is related to, but not identical with, academic achievement, and as
Naglieri has convincingly demonstrated (Naglieri & Das, 1997; Naglieri & Rojahn, 2004),
most current IQ tests are contaminated with achievement content that confounds their
interpretability. Unlike traditional indices of psychometric validity, (e.g., face, content,
construct, predictive), the continued practice by publishers of commercial IQ tests of adding
content that is overly laden with achievement content or adding spurious IQ subscales that fail
to align with intellectual theory has been termed cash validity (Kush, Spring & Barkand,
2012), that is, an attempt simply to increase the marketability of the IQ test.
In contrast, while measures of inspection time clearly align more obviously to
psychometric g than many commercial IQ tests, their narrow focus which minimizes memory
and learned knowledge also restricts their predictive validity or their ability to predict
occupational or academic performance (Kush, Spring & Barkand, 2012). Hence the tradeoff:
the focus on a single cognitive process provides a more pure measure of g as well as greater
theoretical tractability but at the same time offers less predictive power. For theoreticians, the
IQ/IT relationship has already been well established. IT measures have been shown to make
independent contributions to the prediction of IQ (Kranzler & Jensen, 1991; Nettelbeck &
Rabbitt, 1992). Additionally, as Nettelbeck has shown (1987), measures of inspection time
account for approximately 25% of the variance in IQ tests, a finding that was confirmed in a
meta-analysis of 31 studies conducted by Kranzler and Jensen (1989). However, the
supplemental inclusion of an IT measure with IQ test administration offers a promising
approach for practitioners. For example, the inclusion of an IT measure within a standard
assessment battery may improve diagnostic utility. For example, IT has been shown to be an
important biological marker in the diagnosis of anxiety disorders (In-Albon, Dubi, Rapee, &
Schneider, 2009), Korsakoffs and Alzheimers diseases (Deary et al., 1991), Parkinsons
disease (Johnson, et al., 2004), or major depression (Chase, Michael, Bullmore, Sahakian &
10
Joseph C. Kush
Robbins, 2010) as these populations demonstrate longer IT response times as a result of their
impairment. Additionally, because IT has been shown to be stable across the lifespan, and
free of gender differences, IT may serve as a biomarker to determine cognitive changes
associated with aging (Burns & Nettelbeck, 2005). Finally, by using IT as a supplemental,
individual measure rather than the less informative composite IQ index, a multivariate genetic
analysis with IQ subtest scores will establish whether the IT/IQ relationship is influenced by
the same or separate genetic factors (Luciano et al., 2004) Further, it appears that most of the
genetic variance in IT is better explained by genes influencing general ability rather than the
perceptual speed/organization factor (Luciano et al., 2004). For example, IT has demonstrated
a strong phenotypic relationship with Wechslers Digit Symbol (Crawford et al., 1998; Deary,
1993), although the relationship may be mediated by a general fluid ability factor rather than
a factor characterizing perceptual speed (Luciano et al., 2004). Perceptual speed, as
demonstrated by an inspection time task, has been shown to be an intermediate phenotype in
studies attempting to determine genetic loci that determine variance in intelligence
(Posthuma, de Geus, & Boomsma, 2001), accounting for approximately 10% of the genetic
variance in Verbal IQ and 20% of the genetic variance in Performance IQ. Based on fMRI
studies, there is preliminary research to suggest that the patterns of brain activation during IT
performance are surprisingly associated with complex cognitive tasks (Deary, et al., 2001;
Luciano, et al., 2005) with the areas of activation appearing to be the cingulate gyrus of the
limbic lobe and the inferior, medial and superior frontal gyri of the frontal lobes. And because
IT is less cognitively complex than IQ, it may prove simpler to model biologically, and as a
result present an easier model for the selection of other candidate genes influencing cognitive
ability (Luciano, et al., 2005).
REFERENCES
Anderson, M. (1992). Intelligence and development: A cognitive theory. Oxford: Blackwell.
Anderson, M., & Miller, K. L. (1998). Modularity, mental retardation, and speed of
processing. Developmental Science, 1, 239245. doi:10.1111/1467-7687.00037.
Benson, N., Hulac, D. M., & Kranzler, J. H. (2010). Independent examination of the
Wechsler Adult Intelligence Scale Fourth Edition (WAIS-IV): What does the WAIS-IV
measure? Psychological Assessment, 22, 121-130.
Binet, A. (1905). New methods for the diagnosis of the intellectual level of subnormals.
LAnnee Psychologique, 12, 191-244.
Binet, A., & Simon, T. (1905). Methodes nouvelle pour le diagnostic du niveau intellectual
des anormaux. LAnnee Psychologique, 11, 191-244.
Bodin, D., Pardini, D. A., Burns, T. G., & Stevens, A. B. (2009). Higher order factor structure
of the WISC-IV in a clinical neuropsychological sample. Child Neuropsychology, 15,
417-424.
Bors, D. A., Stokes, T. L., Forrin, B., & Hodder, S. L. (1999). Inspection time and
intelligence: Practice, strategies, and attention. Intelligence, 27, 111129.
Braden, J. P., & Athanasiou, M. S. (2005). A comparative review of nonverbal measures of
intelligence. In D. P. Flanagan & P. L. Harrison (Eds.), Contemporary intellectual
assessment: Theories, tests, and issues (pp. 557-577). New York: Guilford Press.
11
Brody, N. (2002). g and the one-many problem: Is one enough? In The nature of intelligence
(Novartis Foundation Symposium 233) (pp. 122-135). NY: Wiley.
Brown, R. T., Reynolds, C. R., & Whitaker, J. S. (1999).Bias in mental testing since bias in
mental testing. School Psychology Quarterly, 14, 208-238.
Burns, N. R., & Nettelbeck, T. (2003). Inspection time in the structure of cognitive abilities:
Where does IT fit? Intelligence, 31, 237-255. doi:10.1016/S0160-2896(02)00120-4.
Burns, N. R., & Nettelbeck, T. (2005). Inspection time and speed of processing: Sex
differences on perceptual speed but not IT. Personality and Individual Differences, 39,
439-446.
Burns, N. R., Nettelbeck, T. & Cooper, C. J. (1999). Inspection time correlates with general
speed of processing but not with fluid ability. Intelligence, 27, 37-44.
Canivez, G. L. (in press). Incremental criterion validity of WAIS-IV Factor Index Scores:
Relationships with WIAT-II and WIAT-III subtest and composite scores. Psychological
Assessment.
Canivez, G. L. (2008). Hierarchical factor structure of the Stanford-Binet Intelligence
ScalesFifth Edition. School Psychology Quarterly, 23, 533541.
Canivez, G. L., Konold, T. R., Collins, J. M., & Wilson, G. (2009).Construct validity of the
Wechsler Abbreviated Scale of Intelligence and Wide Range Intelligence Test:
Convergent and structural validity. School Psychology Quarterly, 24, 252-265.
Canivez, G. L., & Kush, J. C. (in press). WISCIV and WAISIV structural validity:
Alternate methods, alternate results: Commentary on Weiss et al. (2013a) and Weiss et al.
(2013b). Journal of Psycho educational Assessment.
Canivez, G. L., & Watkins, M. W. (2010). Investigation of the factor structure of the
Wechsler Adult Intelligence Scale Fourth Edition (WAIS-IV): Exploratory and higher
order factor analyses. Psychological Assessment, 22, 827-836.
Carlson, J. S., & Jensen, C. M. (1982). Reaction time, movement time, and intelligence: A
replication and extension. Intelligence, 6, 265-274.
Carroll, J. B. (1993).Human cognitive abilities. Cambridge, United Kingdom: Cambridge
University Press.
Carroll, J. B. (2003). The higher-stratum structure of cognitive abilities: Current evidence
supports g and about ten broad factors. In H. Nyborg (Ed.), The scientific study of general
intelligence: Tribute to Arthur R. Jensen (pp. 521). New York: Pergamon Press.
Carroll, J. B. (1997).Theoretical and technical issues in identifying a factor of general
intelligence.In B. Devlin, S. E. Fienberg, D. P. Resnick, & K. Roeder (Eds.), Intelligence,
genes, and success: Scientists respond to the bell curve (pp. 125-156).New York:
Springer-Verlag.
Cattell, R. B. (1963). The theory of fluid and crystalized intelligence: A critical experiment.
Journal of Educational Psychology, 54, 1-22.
Ceci, S. J. (1991). How much does schooling influence general intelligence and its cognitive
components? A reassessment of the evidence. Developmental Psychology, 27, 703-722.
Chamorro-Premuzic, T., Moutafi, J., & Furnham, A. (2005). The relationship between
personality traits, subjectively assessed and fluid intelligence. Personality and Individual
Differences, 38, 1517-1528.
Chase, H. W., Michael, A., Bullmore, E. T., Sahakian, B. J.,& Robbins, T. W.(2010).
Paradoxical enhancement of choice reaction time performance in patients with major
depression.Journal of Psychopharmacology, 24, 471-479.
12
Joseph C. Kush
Conway, A., Cowan, N., Bunting, M., Therriault, D., & Minkoff, S. (2002). A latent variable
analysis of working memory capacity, short-term memory capacity, processing speed,
and general fluid intelligence. Intelligence, 30, 163-183.
Crawford, J. R., Deary, I. J., Allan, K. M., and Gustafsson, J. (1998). Evaluating competing
models of the relationship between inspection time and psychometric intelligence.
Intelligence, 26, 2742.
Cronbach, L. J. (1957). The two disciplines of scientific psychology. American Psychologist,
12, 671-684.
Deary, I. J. (1993). Inspection time and WAIS-R IQ subtypes: A confirmatory factor analysis
study. Intelligence, 17, 223236.
Deary, I. J. (2000) Looking down on human intelligence. Cambridge University Press,
Cambridge.
Deary, I. J. (2001). Human intelligence differences: Toward a recent history. Trends in
Cognitive Science, 5(3), 127-130.
Deary I. J., Hunter, R., Langan, S. J., & Goodwin, G. M. (1991). Inspection time,
psychometric intelligence and clinical estimates of cognitive ability in pre-senile
Alzheimers disease and Korsakoffs patients. Brain, 114, 25432554.
Deary, I. J., Simonotto, E., Marshall, A., Marshall, I., Goddard, N., Wardlaw, J. M., (2001).
The functional anatomy of inspection time: a pilot fMRI study. Intelligence, 29, 497510.
Deary, I. J., & Stough, C. (1996). Intelligence and inspection time: Achievements, prospects
and problems. American Psychologist, 51, 599608.
Deary, I. J., Strand, S., Smith, P., & Fernandes, C. (2007). Intelligence and educational
achievement. Intelligence, 35, 13-21.
DiStefano, C., & Dombrowski, S. C. (2006). Investigating the theoretical structure of the
Stanford-BinetFifth Edition. Journal of Psychoeducational Assessment, 24, 123136.
Elliott, C. D. (2007). Differential Ability Scales-Second Edition: Introductory and technical
handbook. San Antonio, TX: The Psychological Corporation.
Flanagan, D. P., Ortiz, S. O., & Alfonso, V. C. (2007). Essentials of cross-battery assessment
(2nd ed.). New York: Wiley.
Fox, M. C., Roring, R. W., & Mitchum, A. L. (2009). Reversing the speedIQ correlation:
Intra-individual variability and attentional control in the inspection time paradigm.
Intelligence, 37, 76-80.
Freberg, M. E., Vandiver, B. J., Watkins, M. W., & Canivez, G. L. (2008). Significant factor
score variability and the validity of the WISCIII Full Scale IQ in predicting later
academic achievement. Applied Neuropsychology, 15, 131139.
Gardner, H. (1983). Frames of mind: The theory of multiple intelligences. New York: Basic
Books.
Gignac, G. E. (2005). Revisiting the factor structure of the WAIS-R: Insights through nested
factor modeling. Assessment, 12, 320-329.
Gignac, G. E. (2006). The WAIS-III as a nested factors model: A useful alternative to the
more conventional oblique and higher-order models. Journal of Individual Differences,
27, 73-86.
Glutting, J. J., Adams, W., & Sheslow, D. (2000). Wide Range Intelligence Test: Manual.
Wilmington, DE: Wide Range, Inc.
Glutting, J. J., Watkins, M. W., Konold, T. R., & McDermott, P. A. (2006). Distinctions
without a difference: The utility of observed versus latent factors from the WISC-IV in
13
estimating reading and math achievement on the WIAT-II. Journal of Special Education,
40, 103-114.
Gottfredson, L. S. (1997). Mainstream science on intelligence: An editorial with 52
signatories, history, and bibliography. Intelligence, 24, 1323.
Gottfredson, L. S. (2000). Pretending that intelligence doesnt matter. Cerebrum, 2, 7596.
Gottfredson, L. S. (2008). Of what value is intelligence? In A. Prifitera, D. Saklofske, & L. G.
Weiss (Eds.), WISC-IV clinical assessment and intervention (2nd ed., pp. 545-564).
Amsterdam: Elsevier.
Gould, S. J. (1996). The mismeasure of man (2nd ed.). New York: Norton.
Gregory, R. J. (2004). Psychological testing: History, principles, and applications (4th ed.)
Boston: Allyn & Bacon.
Grudnik, J. L., & Kranzler, J. H. (2001). Meta-analysis of the relationship between
intelligence and inspection time. Intelligence, 29, 523-535. doi:10.1016/S01602896(01)00078-2
Gustafsson, J. E. (1984). A unifying model for the structure of intellectual abilities.
Intelligence, 8, 179203. doi:10.1016/0160-2896(84)90008-4.
Horn, J. L., & Blankson, N. (2005). Foundations for better understanding of cognitive
abilities. In D. P. Flanagan & P. L. Harrison (Eds.), Contemporary intellectual
assessment: Theories, tests, and issues (2nd ed., pp. 4168). New York, NY: Guilford
Press.
Horn, J. & Noll, J. (1994). A system for understanding cognitive capabilities: A theory and
the evidence on which it is based. In D. K. Detterman (Ed.), Current Topics in Human
Intelligence, Vol. 4: Theories of Intelligence (pp. 93-132). New York, NY: SpringerVerlag.
Horn, J. L., & Noll, J. (1997). Human cognitive abilities: Gf-Gc theory.In D. P. Flanagan, J.
L. Genshaft, P. L. Harrison (Eds.). Contemporary intellectual assessment (pp. 53-91).
New York: Guilford Press.
In-Albon, T., Dubi, K., Rapee, R. M., & Schneider, S.(2009).Forced choice reaction time
paradigm in children with separation anxiety disorder, social phobia, and nonanxious
controls.Behaviour Research and Therapy, 47, 10581065.
Jensen, A. R. (1979). The nature of intelligence and its relation to learning. Journal of
Research and Development in Education, 12, 7985.
Jensen, A. R. (1998). The g factor: The science of mental ability. Westport, CT: Praeger
Publishers.
Jensen, A. R. (2006). Clocking the mind: Mental chronometry and individual differences.
Oxford, UK: Elsevier.
Johnson, A. M., Almeida, Q. J., Stough, C., Thompson, J. C., Singarayer, R., & Jog, M. S.
(2004). Visual inspection time in Parkinsons disease: deficits in early stages of cognitive
processing. Neuropsychologia, 42, 577-583.
Johnson, W., & Bouchard, T. J., Jr. (2005). Constructive replication of the visual-perceptualimage rotation model in Thurstones (1941) battery of 60 tests of mental ability.
Intelligence, 33(4), 417-430.
Kamphaus, R. W., Petoskey, M. D., & Rowe, E. W.(2000).Current trends in psychological
testing of children. Professional Psychology: Research and Practice, 31, 155-164.
Kaufman, A. S. (2009). IQ testing 101. New York: Springer Publishing.
14
Joseph C. Kush
Kaufman, A. S., & Kaufman, N. L. (2004). Kaufman Assessment Battery for Children-Second
Edition. Circle Pines, MN: AGS Publishing.
Keith, T. Z. (2005). Using confirmatory factor analysis to aid in understanding the constructs
measured by intelligence tests. In D. P. Flanagan & P. L. Harrison (Eds.), Contemporary
intellectual assessment: Theories, tests, and issues (2nd ed., pp. 581614). New York,
NY: Guilford Press.
Kranzler, J. H., Flores, C. G. & Coady, M. (2010). Examination of the cross-battery approach
for the cognitive assessment of children and youth from diverse linguistic and cultural
backgrounds. School Psychology Review, 39, 431-446.
Kranzler, J. H., & Jensen, A. R. (1989). Inspection time and intelligence: A meta-analysis.
Intelligence, 13, 329-347. doi: 10.1016/S0160-2896(89)80006-6.
Kush, J. C. (1996). Factor structure of the WISC-III for students with learning disabilities.
Journal of Psycho-educational Assessment, 14, 32-40.
Kush, J. C. (2005). Review of the Stanford-Binet Intelligence Scales: Fifth Edition. In R. S.
Spies & B. S. Plake (Eds.), The Sixteenth Mental Measurements Yearbook (pp. 979984).
Lincoln, NE: Buros Institute of Mental Measurements.
Kush, J. C., Spring, M. B., and Barkand, J. (2012). Advances in the assessment of cognitive
skills using computer-based measurement. Behavior Research Methods, 44, 125-134.
DOI 10.3758/s13428-011-0136-2
Lu, L., Weber, H., Spinath, F., & Shi, J. (2011). Predicting school achievement from
cognitive and non-cognitive variables in a Chinese sample of elementary school children.
Intelligence, 39, 130-140.
Luciano, M., Posthuma, D., Wright, M. J., de Geus, E. J. C., Smith, G. A., Geffen, G. M.,
Boomsma, D. I., & Martin, N. G. (2005). Perceptual speed does not cause intelligence,
and intelligence does not cause perceptual speed. Biological Psychology, 70, 18.
Luciano, M., Smith, G. A., Wright, M. J., Geffen, G. M., Geffen, L. B., & Martin, N. G.
(2001). On the heritability of inspection time and its covariance with IQ: A twin study.
Intelligence, 29, 443457.
Luciano, M., Wright, M. J., Geffen, G. M., Geffen, L. B., Smith, G. A., & Martin, N. G.
(2004). A genetic investigation of the covariation among inspection time, choice reaction
time, and IQ subtest scores. Behavior Genetics, 34, 41-50.
Luo, D., Thompson, L. A.,& Detterman, D. K. (2003). The phenotypic and behavioral genetic
covariation between elemental cognitive components and scholastic performance.
Behavior Genetics, 33, 221-246.
Mackintosh, N. J. (1986). The biology of intelligence? British Journal of Psychology, 77,
118.
Mackintosh, N. J. (1998). IQ and human intelligence. Oxford, England: Oxford University
Press.
Mackintosh, N. J. (2012). IQ and human intelligence (2nd ed.). New York: Oxford University
Press.
Mackintosh, N. J., & Bennet, E. S. (2002). IT, IQ, and perceptual speed. Personality and
Individual Differences, 32, 685693.
McGrew, K. S. (2009). CHC theory and the human cognitive abilities project: Standing on
the shoulders of the giants of psychometric intelligence research. Intelligence, 37(1), 110.
Murdoch, S. (2007). IQ: A smart history of a failed idea. Hoboken, NJ: Wiley.
15
Naglieri, J. A., & Bornstein, B. T. (2003). Intelligence and achievement: Just how correlated
are they? Journal of Psychoeducational Assessment, 21, 244260.
Naglieri, J. A., & Das, J. P. (1997). Cognitive Assessment System: Interpretive handbook.
Itasca, IL: Riverside.
Naglieri, J. A., & Rojahn, J. (2004). Construct validity of the PASS theory and CAS:
Correlations with achievement. Journal of Educational Psychology, 96, 174181.
Neisser, U., Boodoo, G., Bouchard, Jr. T. J., Boykin, A. W., Brody, N., Ceci, S. J., Halpern,
D. F., Loehlin, J. C., Perloff, R., Sternberg, R. J., & Urbina, S. (1996). Intelligence:
Knowns and unknowns. American Psychologist, 51, 77-101.
Nettelbeck, T. (1987). Inspection time and intelligence. In P. A. Vernon (Ed.), Speed of
information processing and intelligence (pp. 295346). Norwood, NJ: Ablex.
Nettelbeck, T. (1998). Jensens chronometric research: Neither simple for sufficient but a
good place to start. Intelligence, 26, 233241. doi:10.1016/S0160-2896(99)80006-3
Nettelbeck, T. (2001). Correlation between inspection time and psychometric abilities: A
personal interpretation. Intelligence, 29, 459474. doi:10.1016/S0160-2896(01)
00072-1.
Nettelbeck, T. (2003). Inspection time and g. In H. Nyborg (Ed.), The scientific study of
general intelligence: Tribute to Arthur R. Jensen, pp. 77-91. New York: Pergamon.
Nettelbeck, T., & Lally, M. (1976). Inspection time and measured intelligence. British
Journal of Psychology, 67, 1722.
Nettelbeck, T., and Rabbitt, P. M. A. (1992). Aging, cognitive performance, and mental
speed. Intelligence, 16,189205.
Nettelbeck, T., & Wilson, C. (2005). Uncertainty about the biology of intelligence: A role for
a marker test. Cortex, 41, 234235.
Niileksela, C. R., Reynolds, M. R., & Kaufman, A. S. (2012, December 17). An alternative
CattellHornCarroll (CHC) factor structure of the WAIS-IV: Age invariance of an
alternative model for ages 7090. Psychological Assessment. Advance online publication.
doi:10.1037/a0031175.
OConnor, T. A. & Burns, N. R. (2003). Inspection time and general speed of processing.
Personality and Individual Differences, 35, 713-724.
Parker, D. R., & Benedict, K. B. (2002). Assessment and intervention: Promoting successful
transitions for college students with ADHD. Assessment for Effective Intervention, 27,
324.
Partchev, I., & De Boeck, P. (2012). Can fast and slow intelligence be differentiated?
Intelligence, 40, 23-32. doi:10.1016/j.intell.2011.11.022.
Petrill, S. A., Dasen, L., Thompson, L. A., & Detterman, D. K. (2001). Inspection time and
the relationship among elementary cognitive tasks, general intelligence and specific
cognitive abilities. Intelligence, 29, 487496.
Posthuma, D., de Geus, E. J. C., & Boomsma, D. I. (2001). Perceptual speed and IQ are
associated through common genetic factors. Behavior Genetics, 31, 593602.
Reynolds, C. R., & Kamphaus, R. W. (2003). Reynolds Intellectual Assessment Scales. Lutz,
FL: Psychological Assessment Resources, Inc.
Roid, G. H. (2003). Stanford-Binet Intelligence Scales-Fifth Edition: Technical manual.
Itasca, IL: Riverside Publishing.
Sattler, J. M. (2001). Assessment of children (4th ed.). San Diego, CA: Jerome Sattler.
16
Joseph C. Kush
Sheppard, L. D., & Vernon, P. A. (2007). Intelligence and speed of information processing: A
review of 50 years of research. Personality and individual Differences, 44, 247-259.
Spearman, C. (1904). General intelligence objectively determined and measured. The
American Journal of Psychology, 15, 201-293.
Spearman, C. (1923). The nature of intelligence and the principles of cognition (2nd ed.).
London: Macmillan.
Spearman, C. (1927). The abilities of man: Their nature and measurement. New York, NY:
Macmillan.
Sternberg, R. J. (1998).Abilities are forms of developing expertise. Educational Researcher,
27, 11-20.
Sternberg, R. J. (1999).Intelligence as developing expertise. Contemporary Educational
Psychology, 24, 359-375.
Sternberg, R. J. (2003).What is an expert student? Educational Researcher, 32, 5-9.
Thorndike, R. L. (1986). The role of general ability in prediction. Journal of Vocational
Behavior, 29, 332-339.
Thurstone, L. L. (1924). The nature of intelligence. London: Routledge.
Thurstone, L. L. (1938). Primary mental abilities. Chicago: University of Chicago Press.
Vernon, P. E. (1961). The structure of human abilities. London: Methuen.
Vernon, P. E. (1965). Ability factors and environmental influences. American Psychologist,
20, 723-733.
Vickers, D. (1970). Evidence for an accumulator model of psycho-physical discrimination.
Ergonomics, 13, 3758. doi:10.1080/ 00140137008931117.
Vickers, D. (1979). Decision processes in visual perception. London: Academic Press.
Ward, L. C., Bergman, M. A., & Hebert, K. R. (2012). WAIS-IV subtest covariance structure:
Conceptual and statistical considerations. Psychological Assessment, 24, 328340.
doi:10.1037/a0025614.
Watkins, M. W. (2006). Orthogonal higher order structure of the Wechsler Intelligence Scale
for Children Fourth Edition. Psychological Assessment, 18, 123-125.
Watkins, M. W. (2010). Structure of the Wechsler Intelligence Scale for ChildrenFourth
Edition among a national sample of referred students. Psychological Assessment, 22, 782787.
Watkins, M. W., Lei, P.-W., & Canivez, G. L. (2007). Psychometric intelligence and
achievement: A cross-lagged panel analysis. Intelligence, 35, 5968.
Watkins, M. W., Wilson, S. M., Kotz, K. M., Carbone, M. C., & Babula, T. (2006). Factor
structure of the Wechsler Intelligence Scale for ChildrenFourth Edition among referred
students. Educational and Psychological Measurement, 66, 975983.
Wechsler, D. (2003). Wechsler Intelligence Scale for Children-Fourth Edition technical and
interpretive manual. San Antonio, TX: Psychological Corporation.
Wechsler, D. (2008). Wechsler Adult Intelligence Scale-Fourth Edition: Technical and
interpretive manual. San Antonio, TX: Pearson.
Wilson, M. S., & Reschly, D. J. (1996). Assessment in school psychology training and
practice.School Psychology Review, 25, 9-23.
Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). Woodcock-Johnson III Tests of
Cognitive Abilities. Itasca, IL: Riverside Publishing.

ISBN: 978-1-62618-728-3
Chapter 2
IQ AND THE D2 DOPAMINE RECEPTOR GENE

IN CHILDREN OF ALCOHOLICS
Ernest P. Noble*
Alcohol Research Center, Department of Psychiatry & Biobehavioral Sciences,
Semel Institute for Neuroscience & Human Behavior,
David Geffen School of Medicine at UCLA,
University of California at Los Angeles,
Los Angeles, California, US
ABSTRACT
Since the dopaminergic system has been implicated in intelligence, a highly heritable
trait, the association of the D2 dopamine receptor (DRD2)/Ankyrin repeat kinase
(ANKK1) genotypes and the DRD2 C957T genotypes with Intelligence Quotient (IQ)
was studied. Participants consisted of children of alcoholics of Caucasian,
Hispanic/Latino, and African-American descent. They were administered the Wechsler
Intelligence Scale for Children, Third Edition, to determine their IQ. Genotypes of the
DRD2/ANKK1 (the TaqI A [A1A1, A1A2, and A2A2]) and genotypes of the DRD2
C957T (TT, TC, and CC) were obtained. Carriers of the A1+ genotype (A1A1, A1A2) in
the Caucasian and Hispanic/Latino groups and carriers of the T + genotype (TT, TC) in the
African-American group, genotypes which are linked to reduce numbers of brain D2
dopamine receptors, had significantly higher IQ scores when compared to carriers of the
A1- genotype (A2A2) and the T- genotype (CC), respectively. The study suggests that
variants of the DRD2 gene may be a significant contributor to IQ.
Keywords: Addiction, dopamine, DRD2, intelligence, learning
INTRODUCTION
General intelligence, or cognitive ability, is composed of visuospatial and verbal abilities,
processing speed, and working memory (Ward et al., 2000). Cognitive ability, as measured by
18
Ernest P. Noble
IQ, is predicted by both genetic and socioenvironmental factors (Turkheimer et al., 2003;
Harden et al., 2007). The high heritability of cognitive ability has been demonstrated in twin
and adoption studies (Bouchard et al., 1990; Plomin & Neiderhiser, 1991; Plomin, 1999) with
an increased genetic effect with age during childhood and adolescence (Hoekstra, Bartels &
Boomsma, 2007). As a result of these heritability findings, there has been an ongoing search
for candidate genes that are associated with cognitive ability. Promising genetic candidates
for cognition are related to variants of dopamine receptor genes, including the DRD1, DRD2
and DRD4, and variants of catechol-O-methyltransferase (COMT) and the brain-derived
neurotrophic factor (BDNF) (Savitz et al., 2006).
DOPAMINE AND COGNITION

There is impressive evidence for the involvement of the dopamine (DA) system in human
cognition (Previc, 1999 for review). Previc (1999) has hypothesized that DA is the key
neurotransmitter regulating six predominately left-hemisphere cognitive skills critical to
human language and thought: motor planning, working memory, cognitive ability, abstract
reasoning, temporal analysis/sequencing, and generativity. Although this hypothesis may be
speculative, the evidence linking DA to the prefrontal and cognitive functioning is now well
established (Cools & Robbins, 2004 for review). Thus, dopaminergic genes are plausible
candidates in the study of the heritability of cognition. In this regard, one dopaminergic gene
that has garnered some attention is the D2 dopamine receptor (DRD2) gene. The TaqI A
single nucleotide polymorphism of the DRD2 gene (rs 1800497) is expressed as two alleles:
the A1 and the A2 (Grandy et al., 1989). This polymorphism was originally located in the 3
region of the DRD2. However, subsequently it was found to be located within exon 8 of an
adjacent gene, the ankyrin repeat and kinase containing 1 (ANKK1) (Neville et al., 2004). The
A1+ genotype (A1A1, A1A2), compared to the A1- genotype (A2A2), is associated with
reduced brain D2 dopamine receptors (Noble et al., 1991; Thompson et al., 1997; Pohjalainen
et al., 1998; Jnsson et al., 1999) and diminished glucose metabolism (Noble et al., 1997).
Another polymorphism of the DRD2 gene that has received recent attention is the C957T (rs
6277), a single nucleotide polymorphism (SNP) at exon 7 (Duan et al., 2003). This SNP
results in the presence of the 957T allele, and the C957 allele. The 957T+ genotype (TT, TC),
compared to the T- genotype (CC), is associated with decreased translation of D2 dopamine
receptor mRNA, as well as decreased D2 dopamine receptor mRNA stability (Duan et al.,
2003), leading to reduced D2 dopamine receptor density in the brain (Hirvonen et al., 2004).
Since individuals with the A1+ genotype have fewer D2 dopamine receptors than those
with the A1- genotype and because individuals with the T+ genotype have fewer D2 dopamine
receptors than those with the T- genotype, does cognition differ between individuals with
these variants?
There are several studies that suggest the involvement of the DRD2 gene in IQ/cognition.
The relationship of the Taq1A variants to caudate volume and cognitive performance was
studied in memory-impaired Caucasian subjects (Bartres-Faz et al., 2002). Compared with
carriers of the A1- (A2A2) genotype, A1+ (A1A1, A1A2) genotype carriers performed better
on measures of general cognitive functioning and long-term verbal memory and had larger
caudate nucleus volume. A study of Chinese female subjects (Tsai et al., 2002) examined the
IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics
19
association of DRD2 Taq1A and N-methyl-D-aspartate receptor 2B subunit (GRIN2B) gene

variants with IQ. Individuals with the A1+ genotype had higher mean performance IQ than
individuals with the A1- genotype. However, no significant differences with IQ scores were
found for the GRIN2B variants.
Birth weight was associated with educational attainment in Finnish men with the A1+
genotype, but not in men carrying the A1- genotype (Keltingkangas-Janvinen et al., 2007). In
a neurocognitive study of Caucasian children, carriers of the A1+ genotype had significantly
higher Full Scale IQ scores than carriers of the A1- genotype (Antolin et al., 2009).
Furthermore, the DRD2 A1+ genotype was associated with the highest level of educational
attainment in American adolescents (Beaver et al., 2012). However, in studies that have
compared high g group with average g group, no significant difference in DRD2 Taq1A
genotypes was found between these groups (Ball et al. 1988; Petrill et al., 1997; Moises et al.,
2001).
A variant of the DRD2 gene that has been associated with cognitive functioning is the
C957T, but IQ does not appear to have been examined to date. The relationship of the C957T
to performance on the Wisconsin Card Sorting Test, a task which assesses executive
functioning, was studied in healthy Caucasian volunteers (Rodriguez-Jimenez et al., 2006).
Carriers of the T+ (TT, TC) genotype demonstrated better cognitive performance by achieving
a greater number of categories and having a lesser number of perseverative errors and
response than those with the T- (CC) genotype. A study of healthy Caucasian adults (Xu et al.,
2007) found that those with the T+ genotype, when compared to those with the T- genotype
showed significantly better performance on the Word Serial Position Test (WSPT), a task of
working memory. General cognitive ability was also studied in 55-80 year olds living in
Scotland (Bolton et al., 2010). The DRD2 T- genotype was associated with decreased
cognitive ability in these individuals.
SOCIOECONOMIC STATUS (SES) AND IQ

Besides variants of the DRD2 gene that may affect cognition, SES may have an
important impact on cognitive ability (Kaufman & Doppelt, 1976; Sattler, 1992; Wahlsten,
1997). SES is a proxy of the quality of a childs environment based on social and economic
factors such as parental age, income and education. Adjustments for SES reduced the
differences in IQ scores between African-American and Caucasian children (Brooks-Gunn et
al., 1996). However, phenotypic variability in cognitive ability may be the result of separate
contributions of genetic and environmental factors (Block & Dworkin, 1974; Plomin et al.,
1977; Harden et al., 2007). One way to test the importance of gene/environment interactions
is to examine whether the proportion of variance in phenotypes due to genetic differences
varies across different environments. Among a large sample of twins, siblings, half-siblings,
cousins and unrelated siblings reared together, the heritability of cognitive ability varied
across social class levels (Rowe et al., 1999). The more educated the parents, the higher the
heritability of verbal IQ (h 2 = 0.74 versus 0.26). Shared environmental effects were
pronounced when the parents had less than a high school education.
In a study of 12-year-old Swedish twin pairs, measures of verbal ability and inductive
reasoning were determined and SES was estimated from parental education and occupation
20
Ernest P. Noble
(Fischbein, 1980). Heritabilities were highest in the high-SES group and lowest in the lowSES group. Parental education moderated both genetic and shared environmental influences
on verbal ability. As parental education increased, heritability increased and the effects of
shared environment decreased. In families with low SES, 60% of the variance in cognitive
ability came from shared environment, whereas in high SES families, genetic influences were
stronger (Turkheimer et al., 2003). In support of this finding, another study found that in
high-SES families (SES was measured by parental income and education), genetic influences
accounted for about 55% of the variance in the childrens cognitive ability and shared
environmental influences accounted for approximately 35% (Harden et al., 2007). In the lowSES families, 39% of the variance was accounted for by genetic influences and 45% was
accounted for by shared environment.
CURRENT STUDY
Children of alcoholics (COAs) are considered to be a more homogeneous group than
children in the general population because of high shared genetic and environmental
influences. For that reason, this is a worthy population in which to study the effects of
genetics on the development of complex human traits such as cognitive ability. The primary
purpose of the present study was to assess, within a sample of ethnically and racially diverse
COAs, the role the DRD2 gene plays in the heritability of cognitive ability as measured by
IQ, while statistically controlling for age, sex and SES.
METHODS
Participants
The sample consisted of 194 (97 male, 97 female) healthy 12 to 17 year-old (M = 14.4,
SD = 1.3) adolescent COAs. For children to be included in this study, either one or both
parents had to meet DSM-IV diagnostic criteria for alcohol dependence. Any mothers who
were alcohol dependent and reported continuing to drink during the pregnancy of their child
were excluded. All adolescent participants were drug and alcohol free at the time of testing.
Participants' parents reported their own and their parents' racial/ethnic background. Based on
this information, the sample of adolescents was divided into three racial/ethnic groups:
Caucasian (N = 109), Hispanic/Latino (N = 38), and African-American (N = 47). The
research protocol had the approval of the UCLA Human Subject Protection Committee. All
participants gave informed consent and were monetarily compensated for their participation.
Intelligence Quotient
Participants completed the Wechsler Intelligence Scale for Children-Third Edition
(WISC-III) to provide a measure of their IQ (Wechsler, 1991). It also provides Index scores
21
on the subcategories of Verbal Comprehension (VC), Freedom from Distractibility (FD),

Processing Speed (PS), and Perceptual Organization (PO).
Genotyping
A 10 ml blood sample was drawn from each participant. DNA was extracted using
standard techniques and subsequently used as a template for determination of alleles of the
polymorphisms outlined below.
DRD2/ANKK1 TaqI A Genotypes

The DRD2 TaqI A polymorphism was determined by a PCR procedure (Grandy et al.,
1993). Two alleles were obtained: the A1 allele (the uncleaved 310 bp fragment) and the A2
allele (the cleaved 180 bp and 130 bp fragments). In this study, participants with the A1A1 or
A1A2 genotypes were labeled as the A1+ genotype group, whereas participants with the
A2A2 genotype were labeled as the A1- genotype group.
DRD2 C957T Genotypes
The DRD2 C957T polymorphism was determined by a PCR procedure (Duan et al.,
2003). Two alleles were obtained: the T allele (uncleaved 196 bp fragment) and the C allele
(the cleaved 174 bp and 22 bp fragments). In this study, participants with the TT or TC
genotypes were labeled as the T+ genotype group, whereas participants with the CC genotype
were labeled as the T- genotype group.
Socioeconomic Status (SES)

SES was operationally defined by regressing parental income onto parental age, then
taking the standardized residual and multiplying it by the mean of the standardized years of
parental education (Fischbein, 1980). If both the mothers and fathers years of education and
income were reported, then the mean from both parents was used. If only one parents years
of education and/ or income were reported, only that one parents data were used in the
analysis. Independent sample t-tests were used to compare SES scores and age between
race/ethnicity and DRD2 genotype groups separately. Since previous research has shown that
parental SES significantly predicts childrens IQ scores (McLoyd, 1998 for review), SES was
entered into all ANCOVA analyses as a covariate.
Age and Gender

Some studies have shown a neuroanatomical basis for differences in age (Casey et al.,
2000) and between males and females (Lynn et al., 2000) in cognitive ability. Due to these
age effects (Reiss et al., 1996; Schmithorst & Holland, 2006) and gender effects (Halpern,
1997; Lynn, 1998; Schmithorst & Holland, 2006; Strand et al., 2006; Waber et al., 2007) on
IQ in children, both of these variables were controlled in all analyses.
22
Ernest P. Noble
Data Analysis
Linkage Disequilibrium (LD) was conducted within each of the three racial/ethnic
groups to determine if the TaqI A and the C957T polymorphisms were in LD.
One-way ANOVAs were conducted to determine whether there were significant
differences among parental demographic variables (age, education, income, and SES) within
the three racial/ethnic groups. Additionally, one-way ANOVAs were conducted to test for
significant age differences in children among the three groups. The influence of the
racial/ethnic grouping variable on IQ was tested using a one-way ANCOVA with
racial/ethnic group as the independent variable, and age with SES as covariates.
One-way ANCOVAs tested for significant differences in IQ in the total sample using the
children's TaqI A and C957T genotypes as predictor variables and age, gender and SES as
covariates. The influence of the genotype and racial/ethnic group on IQ was tested using a 3 x
2 x 2 x 2 (race x TaqI A x C957T x gender) between subjects ANCOVA, with children's age
and parental SES as covariates. Based on the result of the initial two ANCOVAs, three
additional analyses were conducted after dividing the total sample into three subsamples
using the racial/ethnic variable as the grouping variable. These analyses were 2 x 2 x 2 (TaqI
A x C957T x gender) between subjects ANCOVA, with children's age and parental SES as
covariates. Independent samples t-tests were used to conduct pair-wise comparisons to
determine significant differences in directionality when necessary.
An = 0.05 was used for all tests. To give an estimation of the size of the differences,
effects sizes (2) were reported for all tests.
RESULTS
Table 1. Parental Age, Education, Income and Socioeconomic Status (SES)
Parental Background
Data
Caucasian Group
(n=109)
Age (years)a
Education (years)b
Income(dollars)c
SESd
44.6 5.0
14.3 2.0
4.23 2.58 (x 104)
0.23 0.66
Hispanic/Latino
Group
(n=38)
43.6 5.8
12.2 2.4
2.88 2.04 (x 104)
-0.40 0.63
African-American
Group
(n=47)
39.9 5.6
13.0 1.5
2.94 2.08 (x 104)
-0.21 0.54
All values are mean standard deviation (SD).

SES score is defined by regressing parental income onto parental age, taking the standardized residual
and multiplying it by the mean of the standardized years of parental education.
a
Significant differences in parental age were found between Caucasians and African-American groups
and Hispanic/Latino and African-American groups.
b
Significant differences in parental education were found between Caucasian and Hispanic/Latino
groups and Caucasian and African-American groups.
c
Significant differences in parental income were found between Caucasian and Hispanic and Caucasian
and African-American groups.
d
Significant differences in SES were found between Caucasian and Hispanic/Latino groups and
between Caucasian and African-American groups.
See Results section for significance tests.
23
Linkage Disequilibrium (LD)

Significant LD was detected between the DRD2 TaqI A and C957T in the Caucasian
group (p < 0.001) and in the Hispanic/Latino group (p < 0.001).
However, significant LD was not detected between the DRD2 TaqI A and C957T in the
African-American group. These findings are in accord with a previous study reporting on the
LD of these polymorphisms in different ethnic/racial groups (Duan et al., 2003).
Demographics
Parental age, education, income and SES are shown in Table 1. A One-way ANOVA
indicated significant differences in parental age among the three racial/ethnic groups (F =
13.1, df =2,191, MSE = 28.0, p < 0.001; 2 = 0.12). Caucasian parents were significantly
older than African-American parents (t = 5.2, df = 154, p < 0.001) and Hispanic/Latino
parents were significantly older than African-American parents (t = 3.0, df = 83, p < 0.01).
There was no significant difference in parental age between the Caucasian and
Hispanic/Latino parents.
A one-way ANOVA indicated that there was a significant difference in parental level of
education among the three racial/ethnic groups (F = 18.5, df =2,190, MSE = 4.0, p < 0.001; 2
= 0.16). The Caucasian parents reported a higher level of education than the Hispanic/Latino
parents (t = 5.3, df = 144, p < 0.001). Similarly, the Caucasian parents reported a higher level
of education than African-American parents (t = 3.9, df = 153, p < 0.001). There was no
significant difference in parental level of education between the Hispanic/Latino and AfricanAmerican parents.
Using a one-way ANOVA, a significant difference was found in parental income among
the three racial/ethnic groups (F = 10.8, df =2,191, MSE = 3.7, p < 0.001; 2 = 0.10). The
Caucasian parents reported significantly greater income than Hispanic/Latino parents (t = 3.5,
df = 145, p < 0.01) and African-American parents (t = 3.8, df = 154, p < 0.01). There was no
difference in parental income between the Hispanic/Latino and African-American parents.
When SES was considered, a one-way ANOVA indicated a significant difference among
the three racial/ethnic groups (F = 18.7, df = 2,191, MSE = 0.6, p < 0.001; 2 = 0.16). The
Caucasian group had a significantly higher standardized SES score when compared to both
the Hispanic/Latino group (t = 5.1, df = 145, p < 0.001) and the African-American group (t =
4.0, df = 154, p < 0.001). There was no significant difference in SES scores between the
Hispanic/Latino and African-American parents.
Table 2. Full Scale IQ Scores in the Total Sample of the Three Racial/Ethnic Groups*
Total Sample (n = 194)
IQ
Caucasian (n= 109)
102.8 14.8
Hispanic/Latino (n=38)
93.6 15.4
African-American (n= 47)
85.4 12.5
All values are mean SD.
*
Controlled for age, gender, and SES.
See Results section for results of post hoc analyses.
< 0.001
0.24
24
Ernest P. Noble
Age and IQ Scores in the Total Sample by Racial/Ethnic Group

The one-way ANOVA results indicated that there was a significant difference in
children's age among the three racial/ethnic groups. The Caucasian group was significantly
older than the African-American group (t = 2.2, df = 154, p < 0.05).
The relationship of IQ scores in the total sample by the three racial/ethnic groups is
shown in Table 2. After controlling for age, gender, and SES, the ANCOVA results revealed
a significant main effect of race on IQ scores (F =26.9, df =2,168, MSE = 143.0, p < 0.001, 2
= 0.24). T-tests showed that the Caucasian group had significantly higher IQ scores when
compared to both the Hispanic/Latino group (t = 4.9, df = 145, p < 0.001); and the AfricanAmerican group (t = 8.6, df = 154, p < 0.001). The Hispanic/Latino group also had
significantly higher IQ scores than the African-American group (t = 3.1, df = 83, p < 0.01).
Table 3. Full Scale IQ Scores in the Total Sample by DRD2 Genotypes*
TaqI A
IQ
A1+ genotype (n= 89)
96.5 14.2
A1- genotype (n= 105)
91.3 17.0
C957T
IQ
T+ genotype (n= 119)
96.4 16.0
T- genotype (n= 75)
91.5 14.5
*
< 0.05
0.03
< 0.05
0.03
IQ Scores in the Total Sample by DRD2 Genotypes

The relationship of IQ scores to DRD2 genotypes in the total sample is shown in Table 3.
The ANCOVA results revealed that after controlling for age, gender, and SES, there was a
main effect of the TaqI A genotype on IQ scores, with the A1+ genotype having higher IQ
scores than the A1- genotype (F = 5.4, df =1,168, MSE = 143.0, p < 0.05, 2 = 0.03).
Similarly, after controlling for age, gender, and SES, there was a significant main effect on IQ
scores by C957T genotype, with the T+ genotype having higher IQ scores than the Tgenotype (F = 4.8, df = 1,168, MSE = 143.0, p < 0.05, 2 = 0.03).
The ANCOVA results also indicated that there was a significant racial/ethnic group x
C957T genotype interaction (F = 3.1, df =2,168, MSE = 143.0, p < 0.05, 2 = 0.04), with
African-Americans with the T+ genotype having significantly higher IQ scores than AfricanAmericans with the T- genotype (t = 3.1, df = 45, p < 0.01). There were no significant
differences in IQ scores by the C957T genotypes in the Caucasian or Hispanic/Latino groups.
Based on the foregoing significant race x C957T genotype interaction, in order to
determine genotype differences, the total sample was separated into three separate subsamples
and additional between-subjects ANCOVAs were conducted, with age, gender, and SES
entered as covariates.
25
Table 4. Verbal Comprehension (VC), Freedom from Distractibility (FD), Perceptual

Organization (PO), and Processing Speed (PS) Scores by DRD2 Genotypes in Children
of Three Racial/Ethnic Groups
VC
Caucasian Group
A1+
(n = 43)
Taq 1A
A1(n = 66)
T+
( n = 81)
C957T
T(n = 28)
Hispanic/Latino Group
A1+
(n = 24)
Taq 1A
A1(n = 14)
T+
( n = 18)
C957T
T(n = 20)
African-American Group
A1+
(n = 22)
Taq 1A
A1(n = 25)
T+
( n = 20)
C957T
T(n = 27)
107.7 15.3*
FD
PO
106.7 14.6*
106.3 18.6
0.06
PS
106.8 16.6
0.06
103.3 12.2
104.1 14.6
101.4 13.8
102.8 14.3
104.8 13.5
103.6 15.9
103.9 14.4
103.7 17.0
105.6 14.1
109.0 17.0
102.5 14.2
106.4 11.2
95.1 11.2
101.5 12.7
94.1 13.3**
94.0 12.0***
0.20
0.34
90.0 11.0
89.1 13.7
93.1 9.4
98.4 12.8
93.5 12.2
94.5 14.6
93.8 10.5
99.8 11.8
91.8 13.1
90.2 10.6
94.9 10.8
100.8 13.7
83.9 14.4
89.2 15.8
85.8 14.1
96.2 12.9
84.3 17.4
91.7 15.4
84.2 10.8
95.4 14.1
93.7 15.1
89.0 14.0*
92.2 17.4***
0.28
78.2 11.8
100.5 10.7*
0.11
88.2 15.6
82.1 10.2
0.11
92.2 14.3
All values are mean standard deviation (SD).

*
p < 0.05, ** p < 0.01, *** p < 0.001.
VC (verbal comprehension) index score includes comprehension, vocabulary, similarities and
information subtests.
FD (freedom from distractability) index score includes digit span and arithmetic subtests.
PO (Perceptual Organization) index score includes object assembly, picture arrangement, block design
and picture completion.
Childrens Index Scores by DRD2 Genotypes in Three Racial/Ethnic Groups

Table 4 shows that in the Caucasian children group, those with the A1+ genotype had
significantly higher mean Verbal Comprehension (VC) Index scores than those with the A1genotype, F (1, 102) = 6.1, MSE = 152.3, p < 0.05, 2 = 0.06. Those carrying the DRD2 A1+
genotype had significantly higher Perceptual Organization (PO) Index scores than those
carrying the A1- genotype, F (1, 102) = 6.1, MSE = 191.8, p < 0.05, 2 = 0.06.
In the Hispanic/Latino group, those carrying the DRD2 A1+ genotype had significantly
higher mean Verbal Comprehension (VC) Index scores than those with the A1- genotype, F
(1, 31) = 7.7, MSE = 118.2, p < 0.01, 2 = 0.20. Those with the DRD2 A1+ genotype had
significantly higher mean Freedom from Distractibility (FD) Index scores than those carrying
the A1- genotype, F (1, 31) = 16.2, MSE = 104.3, p < 0.001, 2 = 0.34.
26
Ernest P. Noble
There were no significant differences in any of the Index Scores when the C957T
genotypes were compared in children of the Caucasian or in the Hispanic/Latino groups.
In the African-American group, those with the T+ genotype had significantly higher mean
Verbal Comprehension (VC) Index scores than those with the T- genotype, F (1, 40) = 15.4,
MSE = 166.2, p < 0.001, 2 = 0.28. Those with the T+ genotype had significantly higher mean
Perceptual Organization (PO) Index scores, F (1, 40) = 4.7, MSE = 144.2, p < 0.05, 2 = 0.11,
and PS scores, F (1, 40) = 4.8, MSE = 174.9, p < 0.05, 2 = 0.11, than those with the Tgenotype.
There were no significant differences in any of the Index Scores when the Taq1 A
genotypes were compared in the African-American group.
Table 5. Full Scale IQ Scores by DRD2 Genotypes in each of the
Three Racial/Ethnic Groups*
Caucasian Group (n = 109)
A1+ (n=43)
TaqI A
A1- (n=66)
T+ (n=81)
C957T
T- (n=28)
Hispanic/Latino Group (n = 38)
A1+ (n=24)
TaqI A
A1- (n=14)
T+ (n=18)
C957T
T- (n=20)
African-American Group (n = 47)
A1+ (n=22)
TaqI A
A1- (n=25)
T+ (n=20)
C957T
T- (n=27)
n. s. = not significant.
*
IQ
107.5 13.4
100.8 14.9
104.5 13.0
103.7 12.5
97.4 13.3
83.3 14.1
92.0 10.8
88.7 14.3
84.0 12.1
85.2 11.9
91.2 11.9
78.1 11.9
< 0.05
0.06
n. s.
0.00
< 0.01
0.21
n. s.
0.02
n. s.
0.00
< 0.001
0.27
IQ Scores by DRD2 Genotypes within the Three Racial/Ethnic Groups

The relationship of IQ scores to DRD2 genotypes in each of the three racial/ethnic groups
is shown in Table 5. After controlling for age, gender, and SES, the ANCOVA for the
Caucasian group indicated that IQ scores were significantly different by TaqI A genotype,
with those with the A1+genotype having higher IQ scores than those with the A1- genotype (F
= 5.9, df = 1,99, MSE = 146.4, p < 0.05, 2 = . There were no significant IQ score
differences between the C957T genotypes, nor were there any significant interactions
between the two DRD2 genotypes.
After controlling for age, gender and SES, the ANCOVA for the Hispanic/Latino group
indicated that IQ scores were significantly different by TaqI A genotype, with those with the
A1+ genotype having significantly higher scores than those with the A1- genotype (F = 8.6, df
27
= 1,28, MSE = 91.1, p < 0.01, 2 = 0. There were no significant IQ score differences
between the C957T genotypes, nor were there any significant interactions between the two
DRD2 genotypes.
There were also significant differences in IQ scores between DRD2 genotype groups in
the African-American group; however, for this group the C957T, not the TaqI A, was the
significant predictor. Those with the T+ genotype had significantly higher IQ scores than
those with the T- genotype (F = 13.9, df = 1,37, MSE = 137.9, p < 0.001, 2 = 0. There
were no significant IQ score differences between the TaqI A genotypes, nor were there any
significant interactions between the two DRD2 genotypes.
DISCUSSION
The present study showed differences in IQ scores among the three racial/ethnic groups
studied. IQ scores were significantly and progressively higher in the order of first the AfricanAmerican group, then the Hispanic/Latino group, followed by the Caucasian group. This
finding is in concert with studies that showed similar IQ differences among these racial/ethnic
groups (Vincent, 1991; Prifitera & Saklofske, 1998; Rushton & Jensen, 2006). Additionally,
the present study found IQ score differences among these racial/ethnic groups based on two
different polymorphisms of the DRD2 gene. Specifically, in both the Caucasian and
Hispanic/Latino groups, the A1+ genotype (A1A1, A1A2) compared to the A1- genotype
(A2A2) had higher IQ scores, whereas in the African-American group, the T+ genotype (TT,
TC) compared to the T- genotype (CC) had higher IQ scores. These findings were significant
even after controlling for age, gender and SES. The involvement of the DRD2 gene in IQ is
supported by a large number of studies, which suggest a role for the dopaminergic system in
human intelligence (Cools & Robins, 2004 for review).
DA is considered to be the main neurotransmitter modulating the activation of the reward
system of the brain. When alcohol and other drugs of abuse are consumed, the released DA in
the brain (Di Chiara & Imperato, 1988) activates the nucleus accumbens, inducing feelings of
euphoria and pleasure (Wise & Rompr, 1989). The activation of this DA reward pathway is
considered to be the key element that governs the development of addictive behaviors,
including alcohol and other drug dependence (Koob & Le Moal, 2001; Bowirrat & OscarBerman, 2005; Bressan & Crippa, 2005). However, the release of DA in several brain
structures, in addition to the nucleus accumbens, suggests that beyond its involvement in
reward behaviors, DA is also involved in reinforcement learning (vide infra).
Learning, like the administration of alcohol and other abused drugs, releases DA in the
brain. In a landmark PET study, striatal DA release was observed during performance on a
video game (Koepp et al., 1998). This association of striatal DA release during cognitive
functioning was followed by other PET studies. A Finnish study (Aalto et al., 2005), using a
high affinity D2 DA receptor ligand ([11C] FLB 457), found frontal and temporal DA release
during working memory and attention tasks. Another PET study (Christian et al., 2006), using
another D2 DA receptor ligand ([18F] fallypride), observed brain DA release during learning a
spatial attention task. Additional molecular imaging studies have similarly shown an
association between DA release and cognition in human subjects (Cropley et al., 2006 for
review).
28
Ernest P. Noble
In a functional magnetic imaging (fMRI) study (Cohen et al,. 2007), A1+ subjects, with
reduced dopamine receptors, when administered a D2 dopamine receptor agonist, showed
increased neural reward responses in the medial orbitofrontal cortex, cingulated cortex and
striatum, but decreased reward responses in these regions for A1- subjects. In another fMRI
study (Klein et al., 2007), A1+ subjects were found to avoid actions with negative
consequences less efficiently than A1- subjects. Their medial frontal cortex, involved in
feedback monitoring, responded less to negative feedback than A1- subjects. The authors of
these two studies suggest that decreased sensitivity to negative action consequences in A1+
subjects may explain their increased risk for developing addictive disorders (Noble, 2003).
Functional magnetic imaging (fMRI) studies have found that positive feedback learning
was associated with activation in the striatum (nucleus accumbens) as well as in the anterior
and posterior cingulate cortex and orbitofrontal cortex. However, no activation was observed
for the comparison negative feedback learning (Knutson et al., 2000; Breiter et al., 2001;
Elliott et al., 2003; Nieuwenhuis et al., 2005a,b; Marco-Pallares et al., 2007). A1+ allelic
individuals have shown a decreased capacity to learn from negative characteristics of stimuli
via feedback and manifest less nucleus accumbens and rostral cingulate zone activity over
repeated learning trials (Klein et al., 2007). A1+ deficits in reversal learning, with deficient
feedback integration that stimuli are no longer being reinforced, are also evidenced by
maintaining a previously reinforced response (Jocham et al., 2009).
Positive reinforcement learning has been facilitated amongst A1+ allelic individuals using
bromocriptine, a DA agonist, whereas performance decrements were evident amongst A1individuals. Furthermore, these enhancements in A1+ individuals were accompanied by
nucleus accumbens activation (Kirsch et al., 2006). Consistent with the notion that A1+
individuals show improvement in reinforcement sensitivity over repeated learning trials,
along with poorer inhibitory control than do A1- individuals, suggesting that A1+ individuals
may be sensitive to initial signals of reinforcement but less sensitive to negative feedback
(White et al., 2008). Another genetic study (Frank et al., 2007) also found an association of
variants of two dopaminergic genes, the DRD2 C957T and the COMT Val/Met, with
reinforcement learning. These studies underscore the important role of the dopaminergic
system in learning and reinforcement effects (Wise, 2004; Salamone et al., 2005 for reviews).
The reduced D2 dopamine receptors found in subjects with the A1+ and T+ genotypes has
led to the suggestion that it is the hypodopaminergic state of the DRD2 gene that is a key
element in the development of alcohol and other drug dependence (Bowirrat & OscarBerman, 2005; Melis et al., 2005, La Foll et al., 2009 for reviews). The commonality of
diminished D2 dopamine receptors in drug addictive disorders and in higher learning is
intriguing. Given the misuse of drugs which enhance DA release, the influence of DA on
learning may create particularly powerful conditioned associations regarding substances
themselves, reflected in stronger and more difficult to shift alcohol expectancies (Young et
al., 2004) or in affect regulation expectancies related to addictive eating (Stice et al., 2008).
Although commonality does not infer causality, it does raise the issue that the higher learning
found in individuals with reduced D2 dopamine receptors may have a positive reinforcement
of learning and is a contributor to addictive disorders.
There are some limitations to the present study. Although it identified two
polymorphisms in the DRD2 gene that were associated with higher IQ scores, there are other
dopaminergic genes with hypodopaminergic variants (e.g., DAT, COMT, DRD4) that could
associate with higher IQ scores. Future studies should assess alternative dopaminergic genes
29
to determine what association they have with cognitive ability. Although this is one of the
first studies that identified genetic differences based on different racial/ethnic groups, sample
size precluded dividing the sample by ethnicity and sex simultaneously; however, sex was
added to the analyses to control for its effects. Future research should investigate across
genders the association of the DRD2 polymorphism with IQ. The participants in the present
study were a relatively homogeneous group of COAs, which allows for stronger predictive
statements about their genes and environment, but the results do not necessarily generalize to
children of non-alcoholics. Studies should be conducted utilizing other populations to
determine if the current findings can be supported in the general population.
In summary, the present study sought to examine the association of the DRD2 gene with
cognitive ability. When SES, age and sex were controlled, and genetic factors were taken into
consideration, medium to large effect sizes were found between variants of the DRD2 gene
and IQ. In the Caucasian and Hispanic/Latino groups, the TaqIA polymorphism was the
significant predictor of IQ, whereas for the African-American group, the C957T
polymorphism was the significant predictor of IQ. Moreover, the genotype associated with
higher learning was the genotype associated with reduced brain D2 dopamine receptors.
Overall, the present study implicates the DRD2, a hypodopaminergic gene in the positive
reinforcement effect of learning and that this gene is a significant contributor to IQ.
REFERENCES
Aalto, S., Brck, A., Laine, M., Ngren, K. & Rinne, J. O. (2005) Frontal and temporal
dopamine release during working memory and attention tasks in healthy humans: A
positron emission tomography study using the high-affinity dopamine D2 receptor ligand
[11C] FLB 457. The Journal of Neuroscience: the official journal of the society for
Neuroscience 25, 2471-2477.
Antolin, T., Berman, S.M., Conner, B.T., Ozkaragoz, T., Sheen, C., Ritchie, T., & Noble, E.P.
(2009) D2 dopamine receptor (DRD2) gene, P300, and personality in children of
alcoholics. Psychiatry Res. 166, 91-101.
Ball, D., Hill, L., Thalia, E.C., Chorney, M.J., Chorney, K., Thompson, L.A., Detterman,
D.K., Benbow, C., Lubinski, D., Owen, M., McGuffin, P., & Plomin, R. (1998)
Dopamine markers and general cognitive ability. NeuroReport. 9, 347-349.
Bartres-Faz, D., Junque, C., Serra-Grabulosa, J. M., Lopez-Alomar, A., Moya, A., Bargallo,
N., Mercader, J. M., Moral, P., & Clemente, I. C. (2002) Dopamine DRD2 Taq 1
polymorphism associates with caudate nucleus volume and cognitive performance in
memory impaired subjects. Neuroreport 13,1121-1125.
Beaver, K.M. Write, J.P., DeLisi, M., & Vaughn, M.G. (2012) Dopaminergic polymorhpisms
and educational: Results from a longitudinal sample of Americans. Development
Psychology. 48, 932-938.
Block, N.J. & Dworkin, G. (1974) IQ, heritability and inequality, Part 2. Philosophy and
Public Affairs 4, 40-99.
Bolton, J.L., Marioni, R.E., Deary, I.J., Harris, S.E., Stewart, M.C., Murray, G.D., Gerry,
F.R., Fowkes, R., & Price, J.F. (2010) Association between polymorphisms of the
30
Ernest P. Noble
dopamine receptor D2 and catechol-O-methyl transferase genes and cognitive function.

Behvaioral Genetics. 40, 630-638.
Bouchard, T. J. Jr, Lykken, D. T., McGue, M., Segal, N. L. & Tellegen, A. (1990) Sources of
human psychological differences: The Minnesota Study of Twins Reared Apart. Science
2250, 223-228.
Bowirrat, A. & Oscar-Berman, M. (2005) Relationship between dopaminergic
neurotransmission, alcoholism, and reward deficiency syndrome. American Journal of
Medical Genetics. Part B, Neuropsychiatric Genetic] 132, 29-37.
Breiter, H. C., Aharon, I., Kahneman, D., Dale, A. & Shizgal, P. (2001) Functional imaging
of neural responses to expectancy and experience of monetary gains and losses. Neuron
20, 619-639.
Bressan, R. A. & Crippa, J. A. (2005) The role of dopamine in reward and pleasure behavior
Review of data from preclinical research. Acta Psychiatrica Scandinavica 111 (Suppl
427):14-21.
Brooks-Gunn, J., Klebanov, P. K. & Duncan, G. J. (1996) Ethnic differences in children's
intelligence test scores: role of economic deprivation, home environment and maternal
characteristics. Child Development 67, 396-408.
Casey, B. J., Giedd, J. N, & Thomas, K. M. (2000) Structural and functional brain
development and its relation to cognitive development. Biological Psychology 54, 241257.
Christian, B. T., Lehrer, D. S., Shi, B., Narayanan, T. K., Strohmeyer, P. S., Buchsbaum, M.
S. & Mantil, J. C. (2006) Measuring dopamine neuromodulation in the thalamus: Using
[F-18] fallypride PET to study dopamine release during a spatial learning task.
NeuroImage 31, 139-152.
Cohen, M.X., Krohn-Grimberghe, A., Elger, C.E., & Weber, B. (2007) Dopamine gene
predicts the brains response to dopaminergic drug. Eur J Neurosci. 26, 3652-3660.
Cools, R. & Robbins, T. W. (2004) Chemistry of the adaptive mind. Philosophical
Transactions. Series A: Mathematical, physical, and engineering sciences 362, 28712888.
Cropley, V. L., Fujita, M., Innis, R. B, & Nathan, P. J. (2006) Molecular imaging of the
dopaminergic system and its association with human cognitive function. Biological
Psychiatry 59, 898-907.
Di Chiara, G. & Imperato, A. (1988) Drugs abused by humans preferentially increase
synaptic dopamine concentrations in the mesolimbic system of freely moving rats.
Proceedings of the National Academy of Sciences of the United States of America 85,
5274-5278.
Duan, J., Wainwright, M. S., Comeron, J. M., Saitou, N., Sanders, A. R., Gelernter, J. &
Gejman, P. V. (2003) Synonymous mutations in the human dopamine receptor D2
(DRD2) affect mRNA stability and synthesis of the receptor. Human Molecular Genetics
12, 205-216.
Elliott, R., Newman, J. L., Longe, O. A. & Deakin, J.F. (2003) Differential response patterns
in the striatum and orbitofrontal cortex to financial reward in humans: A parametric
functional magnetic imaging study. The Journal of Neuroscience: the official journal of
the society for Neuroscience 23, 303-307.
Fischbein, S. (1980) IQ and social class. Intelligence 4, 51-63.
31
Frank, M. J., Moustafa, A. A., Haughey, H. M., Curran, T. & Hutchison, K. E. (2007) Genetic
triple dissociation reveals multiple roles for dopamine in reinforcement learning.
Proceedings of the National Academy of Sciences of the United States of America 104,
16311-16316.
Grandy, D. K., Litt, M., Allen, L., Bunzow, J. R., Marchionni, M., Makam, H., Reed, L.,
Magenis, R. E. & Civelli, O. (1989) The human dopamine D2 deceptor gene is located on
chromosome 11 at q22-q23 and identifies a Taq1 RFLP. American Journal of Human
Genetics 45, 778-785.
Grandy, D. K., Zhang, Y. & Civelli ,O. (1993) PCR detection of the Taq1 A RFLP at the
DRD2 locus. Human Molecular Genetics 2(12), 2197.
Halpern, D. F. (1997) Sex differences in intelligence: Implications for education. The
American Psychologist 52, 1091-1102.
Harden, K. P., Turkheimer, E. & Loehlin, J. C. (2007) Genotype by environment interaction
in adolescents cognitive aptitude. Behavior Genetics 37, 273-283.
Hirvonen, M., Laakso, A., Ngren, K., Rinne, J. O., Pohjalainen, T. & Hietala, J. (2004)
C957T polymorphism of the dopamine D2 receptor (DRD2) gene affects striatal DRD2
availability in vivo. Molecular Psychiatry 9, 1060-1061.
Hoekstra, R. A., Bartels, M., & Boomsma, D. I. (2007) Longitudinal genetic study of verbal
and nonverbal IQ from early childhood to young adulthood. Learning and Individual
Differences 17, 97-114.
Jocham, G., Klein, T. A., Neumann, J., von Cramon, D. Y., Reuter, M., & Ullsperger, M.
(2009) Dopamine DRD2 polymorphism alters reversal learning and associated neural
activity. The Journal of Neuroscience, 29, 3695-3704.
Jnsson, E. G., Nothen, M. M., Grunhage, F., Farde, L., Nakashima, Y., Propping, P. &
Sedvall, G. C. (1999) Polymorphisms in the dopamine D2 receptor gene and their
relationships to striatal dopamine receptor density of health volunteers. Molecular
Psychiatry 4, 290-296.
Kaufman, A. S. & Doppelt, J. E. (1976) Analysis of WISC-R standardization data in terms of
the stratification variables. Child Development 47, 165-171.
Kelikangas-Jarvinen, L., Elovanio, M., Kivimaki, M., Raitakari, O.T. Viikari, J.S.A., &
Legtimaki, T. (2007) Dopamine receptor D2 gene Taq1A (C32806T) polymorphism
modifies the relationship between birth weight and educational attainment in adulthood:
21- year follow up of the cardiovascular risk in young Finns study. Pediatrics. 120,
756-761.
Kirsch, P., Reuter, M., Mier, D., Lonsdorf, T., Stark, R., Gallhofer, B., Vaitl, D., & Hennig, J.
(2006). Imaging gene-substance interactions: The effect of the DRD2 Taq1A
polymorphism and the dopamine agonist bromocriptine on the brain activation during the
anticipation of reward. Neuroscience Letters, 405, 196-201.
Klein, T. A., Neumann, J., Reuter, M., Hennig, J., von Cramon, D. Y., & Ullsperger, M.
(2007). Genetically determined differences in learning from errors. Science, 318, 16421645.
Knutson, B., Westdorp, A., Kaiser, E. & Hommer, D. ( 2000) FMRI visualization of brain
activity during a monetary incentive delay task. NeuroImage 12, 20-27.
Koepp, M. J., Gunn, R. N., Lawrence, A. D., Cunningham, V. J., Dagher, A., Jones, T.,
Brooks, D. J., Bench, C. J. & Grasby, P. M. (1998) Evidence for striatal dopamine release
during a video game. Nature 393, 266-268.
32
Ernest P. Noble
Koob, G. F. & Le Moal, M. (2001) Drug addiction, dysregulation of reward, and allostasis.
Neuropsychopharmacology 24(2), 97-129.
Le Foll, B., Gallo, A., Le Strat, Y., Lu, L. & Garabod, P. (2009) Genetics of dopamine
receptors and drug addiction: A Comprehensive Review. Behavioural Pharmacology 20,
1-17.
Lynn, R. (1998) Sex differences in intelligence: a rejoinder to Mackintosh. Journal of
Biosocial Science 30, 529-532.
Lynn, R., Allik, J. & Must, O. (2000) Sex differences in brain size, stature and intelligence in
children and adolescents: some evidence from Estonia. Personality and Individual
Differences 29, 555-560.
Marco-Pallares, J., Muller, S. V., & Munte, T. F. (2007) Learning by doing. an fMRI study of
feedback related brain activations. Neuroreport 18, 1423-1426.
McLoyd, V. C. (1998) Socioeconomic disadvantage and child development. The American
Psychologist 53, 185-204.
Melis, M., Spiga, S. & Diana, M. (2005) The dopamine hypothesis of drug addiction:
hypodopaminergic state. International Review of Neurobiology 63, 101-154.
Moises, H. W., Frieboes, R. M., Spelzhaus, P., Yang, L., Kohnke, M., Herden-Kirchhoff, O.,
Vetter, P., Neppert, J., & Gottesman, I. I. (2001) No association between dopamine D2
receptor gene (DRD2) and human intelligence. Journal of Neural Transmission, 108,
115-121.
Neville, M. J., Johnstone, E. C., & Walton, R. T. (2004) Identification and characterization of
ANKK1: A Novel kinase gene closely linked to DRD2 on chromosome band 11q23.1
Human Mutation 23, 540-545.
Nieuwenhuis, S., Heslenfeld, D. J., Alting von Gensau, N. J., Mars, R. B., Holroyd, C. B. &
Young, N. (2005a) Activity in human reward-sensitive brain areas is strongly cortex
dependent. NeuroImage 25, 1302-1309.
Nieuwenhuis, S., Slagter, H. A., Alting von Geusau, N. J., Heslenfeld, D. J. & Holroyd, C. B.
(2005b) Knowing good from bad: Differential activation of human cortical areas by
positive and negative outcomes. European Journal of Neuroscience 21, 3161-3168.
Noble, E.P. D2 dopamine receptor in psychiatric and neurological disorders and its
phenotypes. (2003) American Journal of Medical Genetics. Part B Neuropsychiatric
Genetics. 21, 103-125.
Noble, E. P., Blum, K., Ritchie, T., Montgomery, A., & Sheridan, P. J. (1991) Allelic
association of the D2 dopamine receptor gene with receptor-binding characteristics in
alcoholism. Archives of General Psychiatry 48, 648-654.
Noble, E. P., Gottschalk, L. A., Fallon, J. H., Ritchie, T. L., & Wu, J. C. (1997) D2 dopamine
receptor polymorphism and brain regional glucose metabolism. American Journal of
Medical Genetics. Part B, Neuropsychiatric Genetics 74(2), 162-166.
Petrill, S. A., Plomin, R., McClearn, G. E., Smith, D. L., Vignetti , S., Chorney, M. J.,
Chorney, K., Thompson, L. A., Detterman, D. K., Benbow, C., Lubinski, D., Daniels, J.,
Owen, M. & McGuffin, P. (1997) No association between general cognitive ability and
the A1 allele of the D2 dopamine receptor gene. Behavior Genetics 27, 29-31.
Plomin, R., DeFries, J. C. & Loehlin, J. C. (1977) Genotype-environmental interaction and
correlation in the analysis of human behavior. Psychological Bulletin 84, 309-322.
Plomin, R. & Neiderhiser J. (1991) Quantitative genetics, molecular genetics, and
intelligence. Intelligence 15, 369-387.
33
Plomin, R. (1999) Genetics and general cognitive ability. Nature 402(Suppl.):C25-C29

Pohjalainen, T., Rinne, J. O., Nagren, K., Lehikoinen, P., Anttila, K., Syvalahti, E. K. G. &
Hietala, J. (1998) The A1 allele of the human D2 dopamine receptor gene predicts low
D2 receptor availability in healthy volunteers. Molecular Psychiatry 3, 256-260.
Previc, F. H. (1999) Dopamine and the origins of human intelligence. Brain and Cognition
41, 299-350.
Prifitera, A. & Saklofske, D. (eds) (1998) WISC-III. Clinical use and interpretations.:
Academic Press, San Diego, p 12.
Reiss, A. L., Abrams, M. T., Singer, H. S., Ross, J. L. & Denckla, M. B. (1996) Brain
development, gender and IQ in children. A volumetric imaging study. Brain 119, 17631774.
Reuter, M., Roth, S., Holve, K., & Kennig, J. (2006) Identification of first candidate genes for
creativity: A pilot study. Brain Research, 1069, 190-197.
Rodriguez-Jimenez, R., Hoenicka, J., Jiminez-Arriero, M. A., Ponce, G., Bagney, A., Aragues,
M. & Palomo, T. (2006) Performance in the Wisconsin Card Sorting Test and the C957T
polymorphism of the DRD2 gene in healthy volunteers. Neuropsychobiology 54, 166-170.
Rowe, D. C., Jacobson, K. C. & van den Oord, E. J. C. G. (1999) Genetic and environmental
influences on vocabulary IQ: parental education level as moderator. Child Development
70, 1151-1162.
Rushton, J. P. & Jensen, A. R. (2006) The totality of available evidence shows the race IQ
gap still remains. Psychological Science: A Journal of the American Psychological
Society/ APS 17, 921-922.
Salamone, J. D., Correa, M., Mingote, S. M. & Weber, S. M. (2005) Beyond the reward
hypothesis: alternative functions of nucleus accumbens dopamine. Current Opinion in
Pharmacology 5, 34-41.
Sattler, J. M. (1992) Assessment Of Children: Revised and Updated Third Edition. Jerome M.
Sattler, Publisher, Inc, San Diego, CA.
Savitz, J., Solms, M., & Ramesar, R. (2006) The molecular genetics of cognition: Dopamine,
COMT and BDNF. Genes, Brain and Behavior, 5, 311-328.
Schmithorst, V. J. & Holland, S. K. (2006) Functional MRI evidence for disparate
developmental processes underlying intelligence in boys and girls. NeuroImage 31, 13661379.
Stice, E., Spoor, S., Bohon, C., & Small, D. M. (2008). Relation between obesity and blunted
striatal response to food is moderated by TaqIA A1 allele. Science, 322, 449-452.Strand,
A., Deary, I. J. & Smith, P. (2006) Sex differences in cognitive abilities test scores: a UK
national picture. British Journal of Educational Psychology 76, 463-480.
Strand, A., Deary, I.J., Smith, P. (2006) Sex differences in cognitive abilities test scores: a
UK national picture. Br J Educ Psych. 76, 463-480.
Thompson, J., Thomas, N., Singleton, A., Piggott, M., Lloyd, S., Perry, E. K., Morris, C. M.,
Perry, R. H., Ferrier, I. N. & Court, J. A. (1997) D2 dopamine receptor gene (DRD2)
Taq1 A polymorphism: Reduced dopamine D2 receptor binding in the human striatum
associated with the A1 allele. Pharmacogenetics 7, 479-484.
Tsai, S. J., Yu, Y. W., Lin, C. H., Chen, T. J., Chen, S. P. & Hong, C. J. (2002) Dopamine D2
receptor and N-methyl-D-aspartate receptor 2B subunit genetic variants and intelligence.
Neuropsychobiology 45, 128-130.
34
Ernest P. Noble
Tucker, D. M. & Williamson, P. A. (1984) Asymmetric neural control systems in human selfregulation. Psychological Review 91, 185-215.
Turkheimer, E., Haley, A., Waldron, M., DOnofrio, B., & Gottesman, II. (2003)
Socioeconomic status modifies heritability of IQ in young children. Psychological
Science: A Journal of the American Psychological Society/ APS 14, 623- 628.
Vincent, K. R. (1991) Black/White IQ differences: Does age make the difference? Journal of
Clinical Psychology 47, 266-270.
Waber, D. P., De Moor, C., Forbes, P. W., Almli, C. R., Botteron, K. N., Leonard, G.,
Milovan, D., Paus, T., & Rumsey, J., The Brain Development Cooperative Group. (2007)
The NIH MRI study of normal brain development: Performance of a population based
sample of healthy children aged 6 to 18 years on a neuropsychological battery. Journal of
the International Neuropsychological Society 13, 1-18.
Wahlsten, D. (1997) The malleability of intelligence is not constrained by heritability. In:
Devlin, B., Feinberg, S. E., Resnick, D. & Roeder K (eds). Intelligence, genes, and
success: Scientists respond to the Bell Curve. Springer-Verlag, New York, p. 71-87.
Ward, L. C., Ryan, J. J. & Axelrod, B. N. (2000) Confirmatory factor analyses of the WAISIII standardization data. Psychological Assessment 12, 341-345.
Wechsler, D. (1991) Wechsler Intelligence Scale for Children, Third edition. San Antonio,
TX: The Psychological Corporation.
White, M. J., Morris, C. P., Lawford, B. R., & Young, R. M. (2008) Behavioral phenotypes of
impulsivity related to the ANKK1 gene are independent of an acute stressor. Behavioral
and Brain Functions, Nov. 24;4:54.
Wise, R. A. (2004) Dopamine, learning and motivation. Nature Reviews. Neuroscience 5,
483-484.
Wise, R. A. & Rompr, P-P. (1989) Brain dopamine and reward. Annual Review of
Psychology 40, 191-225.
Xu, H., Kellendonk, C. B., Simpson, E. H., Keilp, J. G., Bruder, G. E., Polan, H. J., Kandel,
E. R. & Gilliam, T. C. (2007) DRD2 957T polymorphism interacts with the COMT
Val158Met polymorphism in human working memory ability. Schizophrenia Research
90, 104-107.
Young, R. M., Lawford, B. R., Feeney G. F., Ritchie, T., & Noble, E. P. (2004) Alcoholrelated expectancies are associated with the D-2 dopamine receptor and GABA(A)
receptor beta 3 subunit genes. Psychiatry Research, 127, 171-183.

ISBN: 978-1-62618-728-3
Chapter 3
HUMAN GENERAL INTELLIGENCE AS A DOMAIN

GENERAL PSYCHOLOGICAL ADAPTATION
Kevin MacDonald
Department of Psychology, California State University, Long Beach
Long Beach, CA, US
ABSTRACT
The concept of general intelligence as measured by standard IQ tests has always
been a difficult fit for evolutionary psychology. This paper argues that intelligence is a
set of domain general abilities which was not designed to solve any specific problem
from the human evolutionary past. Rather, general intelligence equips humans to make
mental models of the environment and to develop action plans based on these models. It
is thus ideally suited to solve evolutionarily ancient problems of survival and
reproduction, but also to solve novel problems and to create ideologies (e.g., Marxism)
that guide and rationalize behavior. In the human Environment of Evolutionary
Adaptedness (EEA), these action plans evolved as means of achieving affective states,
such as assuaging hunger, achieving social status, or other evolved goal states. Moreover,
it is argued that the most important mechanism underlying general intelligence, the
executive processes of working memory, is not tied to regularities in the EEA.
THE MODULARITY DEBATE IN EVOLUTIONARY PSYCHOLOGY

Early on, the field of evolutionary psychology coalesced around the work of Leda
Cosmides and John Tooby (1992) whose views conflicted with two aspects central to the
century-long research tradition that has grown up around intelligence. First, the whole point
of intelligence quotient (IQ) testing was to provide a measure of individual differences in
cognitive ability. Evolutionary psychology, on the other hand, concentrated on human
universals. For example, regarding personality research, Tooby and Cosmides (1990)
proposed that variation in personality was non-adaptive noise (but see MacDonald, 1995,
2012; Penke, Denissen & Miller, 2007).
36
Kevin MacDonald
In the area of intelligence, it is difficult to conceptualize individual differences as noise,

since individual differences have a very large number of real world correlates that have been
linked to reproductive success. Thus, in contemporary societies, IQ is linked with higher
social status, and greater income and education, but negatively with fertility (Gottfredson,
2007). There is also a long tradition linking increasing hominid brain weight (corrected for
body size), increasing encephalization, longevity, and a relatively K-style reproductive
pattern (e.g., later age of reproduction) (see Rushton, 2004). The linkage between IQ and
variation in life history patterns indicates that variation in IQ is an aspect of a suite of life
history traits and thus unlikely to be simply non-adaptive noise.
Even more problematicand the focus of the present essay has been that research on
intelligence has centered around the concept of intelligence as a general purpose problem
solver, whereas the emphasis within evolutionary psychology has been to conceive the mind
as a set of adaptations designed to solve specific problems encountered in the EEA. The basic
logic of evolutionary psychology is that when the environment presents long-standing
problems and recurrent cues relevant to solving them, the best solution is to evolve domainspecific mechanisms, or modules, specialized to handle specific inputs and generate particular
solutions (Geary, 2005).
However, while all parties to the discussion agree that modules designed to solve specific
problems have evolved, controversy surrounds the proposal that some evolved psychological
mechanisms do not fit standard conceptualizations of modules. For example, Chiappe and
Gardner (2012) emphasize the distinction between systems that utilize implicit processing and
systems characterized by explicit processing.
Implicit and explicit mechanisms may be contrasted on a number of dimensions (e.g.,
Geary, 2005; Lieberman, 2007; MacDonald, 2008; Satpute & Lieberman, 2006; Stanovich,
1999, 2004; see Table 1). Implicit processing is automatic, effortless, relatively fast, and
involves parallel processing of large amounts of information. Implicit processing is
characteristic of what Stanovich (2004) terms the autonomous set of systems, which responds
automatically to domain-relevant information. For example, the visual systems of monkeys
and humans contain numerous areas specialized for different aspects of vision (e.g., Zeki,
1993).
Table 1. Characteristics of Implicit and Explicit Cognitive Systems
Implicit System
Not Reflectively Conscious
Automatic
Fast
Evolved Early
Parallel Processing
High Capacity
Effortless
Evolutionary Adaptation or acquired
by practice
Explicit System
Conscious
Controllable
Relatively Slow
Evolved Late
Sequential Processing
Limited by Attentional and Working Memory
Resources.
Effortful
Acquisition by Culture and Formal Tuition
Human General Intelligence as a Domain General Psychological Adaptation
37
Areas specialized for color and for motion are sensitive to different aspects of visual
stimulation; processing in these different areas occurs in parallel and results in a unitary
image. Other modules proposed in the cognitive literature include modules for social
exchange (Cosmides, 1989), theory of mind (Baron-Cohen, 1995), fear (LeDoux, 2000), folk
physics (Povinelli, 2000), and grammar acquisition (Pinker, 1994).
Although implicit processing is characteristic of evolved modules, it is not restricted to
evolved modules. It occurs in a wide range of circumstances, including skills and appraisals
that have become automatic with practice or repetition, perceptual interpretations of behavior
(e.g., stereotypes), and priming effects (Bargh & Chartrand, 1999). Modules, as defined here,
therefore need not be domain specific; they may also result from domain general processes of
associative and implicit learning (Stanovich, 2004, p. 39; see below).
On the other hand, explicit processing is conscious, controllable, effortful, relatively
slow, and involves serial processing of relatively small amounts of information. Such
processing is characteristic of what Stanovich (2004) terms the analytic system characterized
by context-free mechanisms of logical thought, planning, and cognitive control. The analytic
system is sensitive to linguistic input that allows for explicit representations of the context,
including hypothetical representations of the possible consequences of actions. Explicit
processing is typically experienced as an internal linguistic monologue emerging in a freely
chosen way from oneself and [is] associated with the experience of agency or will (Satpute
& Lieberman, 2006, p. 88).
The view that general intelligence is domain general has been defended by Chiappe and
MacDonald (2005) and attacked by Barrett and Kurzban (2006). Chiappe and Gardner (2012)
responded to Barrett and Kurzban who in turn wrote a rejoinder (Barrett & Kurzban, 2012). I
summarize the main points of this discussion, and the following section develops a proposal
for circumventing a particular argument against domain generality proposed by Barrett and
Kurzban (2012).
Barrett and Kurzban cheerfully acknowledge that their position represents a substantial
retreat from original formulations of evolutionary psychologists. As originally formulated,
evolutionary psychologists proposed that the mind must consist solely of a suite of
mechanisms designed to solve specific problems (Tooby and Cosmides, 1992). That is, these
mechanisms are activated by particular content domains for which they are specifically
designed.
Advocates of evolutionary views do not deny that humans learn, reason, develop, or
acquire a culture; however, they do argue that these functions are accomplished at least in
part through the operation of cognitive mechanisms that are content-specialized
mechanisms that are activated by particular content domains and that are designed to
process information from those domains. (Tooby & Cosmides, 1992, p. 166)
Barrett and Kurzban (2006) retreat from supposing that content specificity is the hallmark
of a module. Instead, domains are construed as content-free rules that make information able
to be processed by the brain: Domains should be construed in terms of the formal properties
of information that render it processable by some computational procedure. In this sense,
even the rules of so-called content-independent logicsfor example, modus ponens are
domain specific, in that modus ponens operates only on propositional representations of a
particular form (Barrett & Kurzban, 2006, p. 634). Thus, for example, working memory or
38
Kevin MacDonald
modus ponens becomes modular simply because the input to working memory must be
encoded in a specialized manner.
Since the brain necessarily has formatting requirements for all possible inputs, all
psychological mechanisms are necessarily modular by this definition. It should be apparent
that this is a radical departure from the original thrust of evolutionary psychology
emphasizing content-specialized mechanisms designed to solve specific problems. Modus
ponens and working memory are not designed to solve any particular problem; rather they
may be used to solve a very wide and undefined range of problemssolving mathematical
problems, making analogies, or figuring out the most efficient way to manufacture pianos, all
of which involve explicit processing.
As Chiappe and Gardner (2012) note, if something is ruled out by the Barrett and
Kurzban (2006) approach, it certainly isnt anything that has been taken seriously in
psychology. Since all psychological mechanism are by definition modular, Barrett and
Kurzbans argument would vitiate Cosmides & Tooby (2002) claim that domain-general
mechanisms are inherently weak because jacks of all trades are masters of none. They
achieve generality only at the price of broad ineptitude (p. 170). Chiappe and MacDonald
(2005) showed that jacks of all trades could evolve, only to have it argued that even jacks
of all trades like working memory and modus ponens turn out to be modular and domain
specific because they necessarily have formatting requirements for their inputs. One wonders
what mechanisms Tooby and Cosmides were attempting to exclude in their rejection of
domain generality, since now it turns out that all conceivable psychological mechanisms are
necessarily modular and domain specific. Clearly, the view of Tooby and Cosmides (1992, p.
77) that organisms are integrated collections of problem-solving mechanisms has been lost
in this reformulation. No problem need be specified; since the emphasis has been shifted to
the very weak claim that all inputs to the brain must be formatted in specific ways in order to
be processed by the brain.
Barrett and Kurzban (2012, pp. 684685) adopt another definition of module, which, like
the definition in terms of formatting requirements, necessarily includes all psychological
mechanisms:
Our view of modularity defines modules in precisely this way: if X is a mechanism,
and if it has a design (i.e., has been shaped by the process of natural selection acting over
evolutionary time), then it is what we are calling a module.
Thus anything that evolved must necessarily be modular, including mechanisms such as
working memory and modus ponens. Since evolution is the only reasonable way that any
mechanism could have come into existence, this definition necessarily includes all
psychological mechanisms.
The issue then becomes, are there limitations on what types of mechanisms can evolve?
Traditionally, evolutionary psychologists have argued that the concept of evolution by natural
selection does indeed place limits on what types of psychological mechanisms can evolve.
These limits derive from the standard definition of an adaptation as necessarily being tied to
environmental regularities:
An adaptation is (1) a system of inherited and reliably developing properties that
recurs among members of a species that (2) became incorporated into the species
39
standard design because during the period of their incorporation, (3) they were
coordinated with a set of statistically recurrent structural properties outside the
adaptation (within in the environment or in other parts of the organism), (4) in such a
way that the causal interaction of the two (in the context of the rest of the properties of
the organism) produced functional outcomes (Tooby & Cosmides, 1992, pp. 6162;
my emphasis)
It is only those conditions that recur, statistically accumulating across many
generations, that lead to the construction of complex adaptations . . . For this reason, a
major part of adaptationist analysis involves sifting for these environmental or
organismic regularities or invariances. (Tooby & Cosmides, 1992, p. 69; my emphasis)
The proposal then is that adaptations can only evolve by tracking regularities. Cues to
these regularities are the proper domain of the adaptation; hence all adaptations are domain
specific and therefore modular. Barrett and Kurzban (2012) clearly agree that recurrences are
essential to the construction of adaptations and they attempt to analyze the ability to solve
novel problems within this framework:
Here, we think, there is a definitional issue: at a certain level, the terms design and
novelty are incompatible with each other, because adaptation is impossible without
some environmental signal, even if statistical and fuzzy, to adapt to. If novel means
bears no resemblance to anything in the past, then design to deal with novelty is a
priori impossible. To be clear, we dont think adaptations designed for novelty are
impossible, but only if we redefine novelty so as to not make adaptation to it
impossible. (Barrett and Kurzban, 2012, p. 686; my emphasis)
Thus, according to Barrett and Kurzban (2012), adaptation to novelty would be

impossible without recurrences to adapt to. They show that some adaptations may respond to
novelty as a byproduct of past selection. Novel tokens of types that recurred over
evolutionary time are a paradigmatic example. For example, a novel food item (say
genetically modified food, or a novel creation of a chef) would be processed by the digestive
system because it has enough similarity to the sorts of food for which the digestive system
was designed. Similarly, a novel three-dimensional object will be processed in a functional
manner by the visual system because the novel item does not depart substantially from the
regularities that resulted in the evolution of the system.
The problem here is that, although such examples are compelling accounts of particular
cases, they do not provide an analysis of the actual mechanisms that form the basis of human
general intelligence, such as the executive processes of working memory discussed in the
following. That is, it would have to be shown that (1) the executive processes of working
memory evolved to track a certain environmental regularity and (2) that these processes are
then able to solve specific types of novel problems because the novel problems have
sufficient similarity to the recurrent features of the EEA that originally resulted in the
evolution of the executive processes of working memory. So far as I am aware, this argument
has not been made in the literature, and, given the extremely wide range inputs to working
memory, it would appear to be a daunting problem to find what environmental regularities the
executive processes of working memory were originally designed to respond to.
40
Kevin MacDonald
HOW ADAPTATIONS CAN EVOLVE IN THE ABSENCE

OF ENVIRONMENTAL RECURRENCES
In the following, I elaborate an argument for how psychological adaptations can evolve in
the absence of environmental regularities based on MacDonald, 1991 (see also Chiappe &
MacDonald 2005; MacDonald & Hershberger, 2008). The claim by Barrett and Kurzban
(2012) that there must be some environmental signal in order for adaptations to evolve is an
attempt to solve the frame problem discussed by cognitive scientists (e.g., Dennett, 1987;
Fodor, 1983). The frame problem is the problem of determining which problems are relevant
and what actions are relevant for solving them. Environmental regularities effectively frame a
problem to be solved and enable the evolution of mechanisms able to respond to the
regularity. The regularity provides a built-in sense of relevancea built-in sense of what the
problem is. Input stemming from an environmental regularity is automatically framed by the
relevant modules because they are designed to be attuned to a particular environmental
regularity.
The above is a compelling argument for the existence of at least some modular, domainspecific mechanisms. Nevertheless, an important aspect of evolution has been to solve the
frame problem in a manner that does not rely on environmental regularities for the evolution
of psychological adaptations. The proposal is that humans and other animals have evolved
motivational systems that solve the frame problem by equipping them with systems that
provide signals when their evolved goals are being met. For example, the hunger mechanism
provides a signal telling the child to look for food and begin feeding. How the child goes
about getting a familiar food item is unspecified and does not depend on environmental
regularities. However, the motivational system effectively frames the problem: It tells the
child what the problem is (the feeling of hunger), and it tells the child when the problem has
been solved (satiation). This signal is not a response to an environmental regularity, but rather
signals that an internal goal has been met; further, achievement of this internal goal (e.g.,
satisfying hunger) must have been linked to reproductive success in the EEA; but there is no
need for reproductive success to be linked to any environmental regularity. As described in
the following, such a system enables the evolution of mechanisms able to take advantage of
ephemeral environmental regularities (classical and operant conditioning) or imitate
successful others (social learning). Ultimately, via the elaboration of the domain general
mechanisms of general intelligence, affectively grounded systems enable organisms to solve
novel problems and (more commonly) to solve ancient evolutionary problems by novel means
means that are more efficient than any possible architecture that is linked to environmental
regularities.
From this perspective, a watershed event in evolutionary history was the evolution of
psychological signalspositive or negative feelingsthat inform the animal when its goals
of survival and reproduction are being met or unmet. Imagine a primitive organism equipped
only with if p, then q devices, where p represents recurrent environmental events and q
represents an evolved response to the event: If a certain environmental situation p occurs
(e.g., presence of food), then respond with behavior q (eating). Such an organism would
completely satisfy the requirements for a psychological adaptation as described by Barrett and
Kurzban (2012): The mind is constructed with mechanisms designed to respond adaptively to
recurrent environmental events (the presence of ps). The mechanism is entirely modular,
41
designed to deal exclusively with a particular kind of input (domain-relevant information) by

encoding the input in a manner that can be processed by the animals nervous system; and it
produces a particular kind of output (e.g., behavior such as eating p), thereby solving a very
specific problem. Its disadvantage would be that there would be no way to take advantage of
nonrecurring information in order to find foodfor example, the information that a certain
ephemerally available stimulus is a cue for food (classical conditioning), the chance discovery
that a certain behavior is a good way to obtain food (operant conditioning), or observing
another animal successfully obtaining food (social learning).
Examples of if p, then q systems are the fixed signaling systems of nonhuman primates
and other animals discussed by Oller and Griebel (2005). Such signals occur in particular
recurrent contexts (e.g., threat, danger, alarm, greeting) and are coupled to the specific
circumstances surrounding their use and the functions they serve. Their meaning is therefore
fixed. The breakthrough in human language was the evolution of contextual freedom, in
which each sound can be produced voluntarily and can be coupled, via learning, to an endless
variety of social functions that are not dependent on environmental recurrences. These
functions can change quickly over time, making them ideal for dealing with uncertain, novel
situations. As in the case of social learning (see below), there is undoubtedly a great deal of
specialized neural machinery underlying human language ability. However, like social
learning, it functions as a domain-general system, with no evolutionarily fixed inputs or
outputs and no fixed relationship to particular environmental regularities. Even infants 3 to 6
months of age are capable of many-to-many mappings between signal and function; there are
a wide variety of signals, many with no social function at all (Oller & Griebel, 2005).
The evolution of motivating systems goes a long way toward solving the frame problem.
(It is also, quite probably, the evolutionary origin of consciousness, because by definition, the
animal must be aware of these motivational cues. In the absence of consciously felt
motivating systems, plants can respond to environmental regularities but are incapable of
learning, much less general intelligence.) A hungry child may indeed be confronted with an
infinite number of behavioral choices, but such a child easily narrows down this infinite array
by choosing behaviors likely to satisfy his or her hunger. The motive of hunger, and the fact
that certain behaviors reliably result in satiating hunger, give structure to the childs behavior
and enable him or her to choose adaptively among the infinite number of possible behaviors.
The childs behavior is not random because it is motivated by the desire to assuage the feeling
of hunger.
Motivational mechanisms can thus be thought of as a set of adaptive problems to be
solved but whose solution is underspecified. Learning mechanisms are examples of the
evolution of hyperplastic mechanisms, mechanisms such as the immune system, which are
unspecialized because they are not responsive to recurrent environmental events and because
there is no selection for a particular phenotypic result (West-Eberhard, 2003, p. 178). Such
systems enable the evolution of any cognitive mechanism, no matter how opportunistic,
flexible, or domain-general, that is able to solve the problem. The child could solve his or her
hunger problem by successfully getting the attention of the caregiver. The problem could be
solved if the child stumbled onto a novel contingency (how to open the refrigerator door); or
it could be solved by imitating others eating a novel food; or the child could develop a
sophisticated plan based on imagining possible outcomes and relying on mechanisms of
general intelligencethe g factor of intelligence research. None of these ways of solving the
42
Kevin MacDonald
problem need result in solutions that were successful in our evolutionary past. This is
illustrated in Figure 1.
Level 1 EVOLVED MOTIVE DISPOSITIONS
Level 2 PERSONAL STRIVINGS
Level 3 CONCERNS, PROJECTS, TASKS (Utilize Domain-General Mechanisms)
Level 4 SPECIFIC ACTION UNITS (Utilize Domain-General Mechanisms)
EXAMPLE:
Evolved Motive Disposition: INTIMACY
Personal Striving: INTIMATE RELATIONSHIP WITH A GIVEN PERSON
Concern, Project, Task: Arrange Meeting, Improve appearance, Get promotion
Action Units: Find phone number, Begin dieting, Work weekends
Figure 1. Hierarchical model of motivation showing relationships between domain-specific and
domain-general mechanisms. Adapted from Emmons (1989).
Motivation represents a major point of contact between evolutionary approaches and

approaches based on learning theory. Learning theories generally suppose that some
motivational systems are biological in origin, but traditionally they have tended toward
biological minimalism. They posit only a bare minimum of evolved motivational systems. For
example, traditional drive theory proposed that rats and people have drives to consume food,
satisfy thirst, have sex, and escape pain. For an evolutionist, this leaves out a great many other
things that organisms desire innately. Personality theory provides a basis for supposing there
are several evolved motive dispositions (EMDs), including evolved motives for seeking out
social status, sexual gratification, felt security (safety), love, and a sense of accomplishment
(MacDonald, 1991, 1995, 2012). Glenn Weisfeld (1997) has expanded on this list by
specifying 16 affects that provide positive or negative signals of adaptive significance: tactile
pleasure and pain, thirst, tasting and smelling, disgust or nausea, fatigue, drowsiness, sexual
feelings, loneliness and affection receiving, interest and boredom, beauty appreciation, music
appreciation and noise annoyance, humor appreciation, pride and shame, anger, and fear. One
can quarrel with the details of such a list, but there is little doubt that there is a wide range of
positive feelings that humans are innately designed to experience and a wide range of
negative feelings that humans are innately designed to avoid.
One type of novelty that organisms must adapt to is ephemeral regularities. Whereas
longstanding regularities give rise to adaptations as traditionally understood, ephemeral
regularities are novel situations with huge benefits for organisms able to exploit them. The
quite ancient solution to the problem of exploiting novel regularities has been the evolution of
domain general learning mechanisms that are not tied to environmental regularities.
The pursuit of evolved motives allows for flexible strategizing and the evolution of
domain-general cognitive mechanismslearning mechanisms and the mechanisms of general
intelligence useful for attaining evolved desires. This fits well with research showing that
problem solving is opportunistic: People satisfy their goals, including evolved goals such as
satisfying hunger, by using any and all available mechanisms. For example, children typically
experiment with a variety of strategies and then select the ones that are effective. Children are
bricoleurs, tinkerers who constantly experiment with a wide range of processes to find
solutions to problems as they occur. Children bring to bear varied processes and strategies,
43
gradually coming through experience to select those that are most effective. . . . Young
bricoleurs . . . make do with whatever cognitive tools are at hand (Deloache, Miller, &
Pierroutsakos, 1998, p. 803).
Although this paper is concerned mainly with general intelligence, mention should be
made of mechanisms that channel learning in adaptive directions and allow for the
transmission of novel cultural variants in the absence of recurrent cues over evolutionary
time. For example, prestige (dominance, power), warmth (prototypically from parents,
enabling parents to be effective models for children) and similarity of model to self (e.g.,
children are biased toward learning from same-sex models) bias social learning; all of these
biasing mechanisms allow for the transmission of novel cultural variants that are themselves
not tied to recurrences in the EEA (MacDonald, 1991). Barrett and Kurzban (2012) accept
this, but argue that biasing mechanisms for social learning (their example relates to
prestigious models) allow novel cultural variants to spread only if accepting cultural variants
on the basis of the prestige of the model is associated with reproductive success. Thus, instead
of learning being attuned to environmental recurrences acting over evolutionary time, they
propose that the regularity ultimately is between reproductive success and acceptance of a
cultural variant.
Nevertheless, there is no theoretical necessity for biasing mechanisms that are tied to
reproductive success in order for domain general learning to evolve. As noted above, it is
sufficient if the satisfaction of internal affective states is linked to reproductive success. Thus,
humans are able to learn from models that do not have prestige or warmth or any of the other
biasing mechanisms uncovered by social learning research if the learner sees the modeled
behavior as successful in achieving a desired goal and regards it as useful in some way. Quite
often, the utility of socially learned behaviors is measured in quite practical termsthe
probable outcome of learning on mundane proximal goals far removed from reproductive
success, such as repairing a TV set, baking a cake. Models are attended to if they are
successful in producing the desired behavior and if the learner has a motive for learning the
behavior. Evolved biasing mechanisms need not enter the picture.
Similar considerations apply to classical and operant conditioning. As an example of a
contrary point of view, Tooby and Cosmides (1992; p. 95) claim that support for domaingenerality in operant conditioning relies on data from experimenter-invented, laboratory
limited, arbitrary tasks. They criticize traditional learning experiments for not focusing
exclusively on ecologically valid, natural taskstasks that deal with problems that were
recurrent in the animals EEA.
Such a stance obviously begs the question of whether there are problems that were not
recurrent in the evolutionary past that can be solved by learning in the absence of biasing
mechanisms. It is certainly true that investigations of learning tasks, especially in animals,
have sometimes revealed specialized learning mechanisms (e.g., rats predisposition to link
nausea with recent food intake [Garcia & Koelling, 1966]). However, an equally remarkable
aspect of learning is that, for example, pigeons can learn to peck keys to satisfy their evolved
goals of staving off hunger even in experimenter-contrived, arbitrary, novel situations without
environmental regularities stemming from the animals EEA. Although pecking for food is
undoubtedly a species-typical behavior for pigeons, pigeons, like rats learning to push levers,
are also able to learn a variety of arbitrary, experimenter-contrived behaviors that are not
components of the animals species-typical foraging behavior. In other words, they are able to
solve a fundamental problem of adaptation (getting food) in a novel and even arbitrary
44
Kevin MacDonald
environment that presents none of the recurrent associations between the animals behavior
and obtaining food experienced in the animals EEA. Similarly, humans are able to learn lists
of nonsense syllablesanother example highlighted by Tooby and Cosmides (1992), despite
the fact that learning such lists was not a recurrent problem in the EEA. People can learn such
lists because their learning mechanisms can be harnessed to new affectively tinged goals,
such as getting course credit as a subject in a psychology study.
In general, operant conditioning, classical conditioning and social learning did not evolve
to link specific recurrent aspects of the EEA with reproductive success. The mechanisms
underlying these abilities imply a great deal of evolved machinery, and there are important
cases where evolution has shaped learning in ways that depart from domain-generality.
However, in general, there are no specified inputs to these systems linked to environmental
regularities in the EEA.
The input to associational mechanisms of rats and humans verges on whatever is
detectable by the sense organs, and operant behaviors span virtually the entire range of
physically possible motor behaviors. Because of their domain-generality, these mechanisms
allow humans and animals to solve problems with features not recurrent in the EEA.
Finally, there are examples in nature where animals, like humans, are able to go beyond
associative learning mechanisms in solving novel problems that are not connected to
regularities in the EEA. Thus New Caledonia crows are able to develop a causal rule that
enables them to solve novel problems (Taylor et al., 2010). Further, they quickly process
causal information and use it to solve novel problems utilizing new tool types (dropping
stones into water so that food items are accessible) that are not utilized in their natural
environments and have no relationship to their established behavioral repertoire (Taylor et al.,
2011).
These results do not appear explicable by the use of associative learning, nor is the causal
rule linked to regularities utilized by the crows in their EEA. Rather, the results indicate an
ability to develop and utilize an abstract causal rule in order to solve an affective goal
assuaging hungerby novel means.
There has been considerable debate on what evolutionary pressures resulted in human
general intelligence: rapid climactic change, ecological maximization, social competition or
some combination of these (see Geary, 2005 for a review). Note that the above argument is
compatible with all of these scenarios. That is, the solution of novel problems is not central to
the argument. Rather, the affective basis of domain generality is evolutionarily ancient,
resulting primitively in simple associative learning mechanisms (classical and operant
conditioning), social learning, the ability to form abstract causal rules (in New Caledonia
crows), and finally human general intelligence as a suite of mechanisms, particularly the
executive processes of working memory (see below) underlying the ability to manipulate
information from a variety of sources in order to achieve goals that may or may not be linked
with affective motivational systems derived from the evolutionary past (i.e., EMDs;
MacDonald, 1991]).
Besides solving novel problems, human general intelligence is adept at finding novel,
more efficient ways to solve old problems of survival and reproduction (e.g., finding better
ways to extract resources from the environment or developing more effective military tactics).
45
DOMAIN-GENERAL MECHANISMS UNDERLYING GENERAL

INTELLIGENCE: WORKING MEMORY CAPACITY AS A PARADIGM
Perhaps the most obvious way that general intelligence is domain general is that the g
factor of general intelligence correlates with a very wide range of mental abilities. Individuals
who are good at mathematical reasoning also tend to be relatively good at verbal
comprehension and rotating figures in space. It seems unlikely then, that general intelligence
evolved to solve particular problems tied to regularities in past environments. Indeed,
common models of general intelligence propose that the g factor is at the top of a hierarchy of
a set of modular processes, such as verbal and spatial reasoning, that have unique inputs with
very specific formatting requirements and unique outputs (e.g., MacDonald & Hershberger,
2005).
For example, Case, Demetriou, Platsidou and Kazi (2001; see also Demetriou, Elklides,
and Platsidou, 1993) proposed a model of intelligence in which general processescore
capacities, including working memory and processing speedare able to process
information from a variety of more specialized, domain-specific content areas as inputs. The
general processes are proposed to constitute the fundamental processes underlying stage
changes in cognitive development. That is, relatively rapid changes in the general processes
constitute transition zones between stages of cognitive development. Similarly, Case (1998)
describes domain-general central conceptual structures (CCS) that serve to integrate and
organize information from modular systems of number, space, and theory of mind: Although
the content that they serve to organize is modular, the structures themselves reflect a set of
principles and constraints that are system wide in their nature, and that change with age in a
predictable fashion (p. 770).
A conceptually similar model of general intelligence is provided by Geary (2005; see
Figure 2). A central executive is able to direct attention and manipulate information that it
receives from inputs from highly specialized, domain-specific mechanisms. These inputs, in
addition to information typically associated with intelligence tests (e.g., spatial and verbal
information), include a very wide range of information (visual object recognition, face
recognition, auditory, olfactory, kinesthetic and gustatory). The central executive can amplify
attention to particular areas as needed and is able to manipulate the information to create
mental models and other goal-relevant representations as well as inhibit information that is
irrelevant to the goal.
General intelligence is the result of the fundamental revolution in brain design by which
the primate brain moved away from massively parallel implicit processing with widely
converging and diverging connections between individual neurons to a more serial,
hierarchical design (Striedter, 2005, p. 340). The result of this revolution was that information
came to localized in one central area which could then be appraised and acted on by a central
executive via explicit processing as opposed to a multitude of reflexive, implicitly processed
connections acting subcortically and in parallel. Thus, for many animals, conflicts between
approach and withdrawal are resolved simply by the summed strength of the competing
implicitly processed action tendencies (thirst versus pain avoidance)a standard ethological
account (e.g., Goetz & Walters, 1997). On the other hand, for humans and some other
animals, information is routed to the prefrontal cortex where it may be held in working
memory and acted on by the executive processes of working memory. (Humans also retain
46
Kevin MacDonald
implicitly operating modular connections, resulting in two pathways for many stimuli, such as
loud noises that activate the fear systemthe ancient evolutionary route directly to the
amygdala which operates very quickly and results in reflexive fear, and a more recent, slower
pathway to the cortex where it is evaluated and acted upon by the central executive via
explicit processing [LeDoux, 1996, 2000].)
Figure 2. An illustration of the proposed relationship between the central executive and a partial list of
inputs: auditory, spatial, visual, gustatory, kinesthetic, and olfactory. The central executive directs
attention to particular domains (e.g., spatial), resulting in a conscious representation of the information
from that domain. At Time 1 there is no specific focus of attention, but at Time 2 attention is focused
on a face, resulting in conscious awareness of the face. (From Figure 7.1 of Geary, 2005).
The processing of this disparate information by the central executive is a prime example
of explicit processing. As noted above, a fundamental divide among psychological
mechanisms is the distinction between implicit and explicit processing, with implicit
processing characteristic of modules as traditionally understood in evolutionary psychology.
Explicit processing is intimately related to general intelligence. And, although necessarily
having formatting requirements for inputs, explicit processing is domain general in the sense
that its inputs can include the ability to make mental models taking in a wide range of
information that is not tied to recurrences in the EEA, including all aspects of human culture
(MacDonald, 2008, 2009). These explicit mental models can then be used to achieve the
EMDs described above, including an immense variety of sub-goals that may be involved in
plans to achieve EMDs but which require navigation through the complexities of modern life
far removed from the EEA (e.g., figuring out the best way to finagle a job promotion or
develop a winning strategy for getting elected to political office). Whereas dedicated modules
47
evaluate costs and benefits implicitly, such sub-goals require explicit analysis of costs and
benefits of a wide array of cultural information (e.g., likelihood of legal consequences,
financial considerations).
Explicit analyses of costs and benefits are able to override implicit, prepotent responses
that have evolved in response to evolutionary regularities. The control processes associated
with explicit processing are centered in the prefrontal cortex (PFC). The PFC is involved in
top-down processing utilized during attempts to match behavior to intentions or internal
states. It is especially important when previous connections between inputs, thoughts, and
actions are not well established, as in confronting novel problems, rather than either innate or
well-established learned connections (Miller & Cohen, 2001). Executive control permits
goal-directed override of primitive and inflexible reactions to environmental stimuli
(Gazzaley & DEsposito, 2008, p. 188).
Explicit processing is called into play when confronting nonroutine tasks that require
flexible responses, retention of information over time, and planning future courses of action
(Dehaene & Naccache, 2001; Miller & Cohen, 2001)all of which are central to intelligent
behavior. Further, explicit processing implies conscious awareness (Stanovich, 2004), and
theories of conscious awareness have converged on the proposition that they are adaptive
because they allow consideration of different kinds of information from systems with
different functions and phylogenetic origins (Morsella, 2005). Because the PFC is widely
connected to sensory, cognitive, affective, and motor modalities, it is well suited to integrate
information useful for making plans and for the production of skilled, intentionally controlled
movements (Gazzaley & DEsposito, 2008; Striedter, 2005).
The control function of explicit processing over implicit processing has become well
established in the area of intelligence research. Unlike the vast majority of animals, humans
can control automatic, heuristic processing and make decisions that depend on explicit
processing. Controlling heuristic processing requires effortful, controlled problem solving and
makes demands on attention and working memory resources. Stanovich (1999) provides
evidence that people with higher general intelligence are better able to selectively control
heuristic, automatic, socially contextualized processing. An example is evaluating a valid
syllogism with a false premise. Consider the following:
All blue people live in red houses.
John is a blue person.
John lives in a red house.
Drawing the correct inference requires decoupling from experience in which there are no
blue people and forming a mental model of a hypothetical situation in which there are blue
people, all of whom live in red houses. The mental models involved in explicit problem
solving include explicitly represented information involving language or images (JohnsonLaird, 1983).
Constructing mental models utilizes working memory. As noted above, models of
intelligence feature central executives with access to input from a wide range of information,
much of it the output of implicit processing. The information is held in working memory and
may be utilized to create plans and evaluate possible outcomes prior to enacting the plan.
Research on human general intelligence has implicated working memory capacity as a
key domain general intellectual ability (Engle, Tuholski, Laughlin, & Conway, 1999;
48
Kevin MacDonald
Kyllonen & Christal, 1990). Thus Kane, Hambrick and Conway (2004) reanalyzed 10 studies
with over 3000 subjects, finding a mean correlation of 0.72 between working memory
capacity and fluid intelligence. Engle et al. (1999) showed that the executive functions of
working memory (assessed by tasks involving attentional control) predicted general
intelligence (i.e., g), but that short-term memory capacity (assessed by tasks such as memory
for sets of words) did not. Working memory capacity thus consists of capacity (e.g., how
many items can be held in memory) and also executive control (the ability to direct attention
and keep it focused on some input while not being distracted by other input).
The various storage buffers of working memory are indeed domain specific (e.g.,
phonological loop, visual/spatial). Nevertheless, the available evidence suggests that,
although performance on complex [memory] span tasks may be influenced by domainspecific processing competencies, they have a commonality in their measurement of a
domain-free ability to control attention (Feldman Barrett, Tugade, & Engle, 2004, p. 556).
This conclusion is based on evidence that even though the various tasks do indeed require the
use of domain specific memory buffers, performance strongly covaries across domains of
performanceexactly as expected on the hierarchical model of general intelligence described
above in which a general factor is at the highest level over a variety of specific ability factors,
and where performance on the specific ability factors is correlated with performance on other
specific ability factors. Further, Feldman Barrett et al. note that capacity in domain specific
processing spheres is unrelated to the speed and accuracy of computation using information
from the specific capacity spheresalso highly compatible with the importance of a domain
general processor responsible for speed and accuracy across domains.
Similarly, Engle (2010, p. S17) notes that The domain-general aspect of working
memoryattention control has established reliability and validity of measurement.
Individual differences in domain-general working memory capacity have been shown to be
important to a wide range of both speech-based and visual/spatial-based tasks. As noted, it is
the domain general aspect of working memory capacity rather than domain specific capacity
spheres that is highly correlated with general intelligence.
Further illustrating the domain generality of WMC, Engle (2010, S20) reviews data
indicating that individual differences in WMC are correlated with a wide range of higherorder cognitive tasks, including tasks related to reading and listening comprehension,
reasoning, bridge playing, and learning to perform a complex task such as computer
programming. People with higher WMC are also better able to block out a very wide range of
intrusive, irrelevant thoughts and representationsanother indication of domain generality.
The attentional processes of WMC are critical to goal management, which involves
constructing, executing, and maintaining a mental plan of action during the solution of a
novel problem (Carpenter et al., 1990). For example, the Ravens Progressive Matrices fluid
intelligence test and the Tower of Hanoi problem (in which participants must develop a longterm plan with multiple subgoals) require one to be able to activate multiple goals and keep
track of the satisfaction of each of them (Carpenter et al., p. 413). Performance on these tasks
in the study by Carpenter et al. was highly correlated (r = .77), which suggests that substantial
goal management was necessary in both tasks. Executive functions underlying general
intelligence are thus involved when problems call for substantial planning and keeping track
of various subgoals without distraction.
49
IDEOLOGY AND THE CREATION OF EVOLUTIONARILY

NOVEL GOALS
Although general intelligence is clearly useful for creating novel means of achieving
ancient evolutionary goals in any environment, it is worth pointing out that humans are also
able to create novel goals that are unrelated to human evolved motive dispositions. That is,
there are many goals in addition to the evolved feeling states described above (e.g., love,
social status, safety) whose satisfaction or avoidance are naturally desired by humans.
I have noted above that the input to the attentional processes of WMC may span a very
wide range of inputs and that a characteristic outcome of such processes is the construction of
mental models useful in solving problems in evolutionarily ancient and modern
environments. A particularly interesting input (or output in the case of creators of ideologies)
consists of what one might term big picture mental models of how the world works or
ought to workwhat is usually termed ideology (MacDonald, 1991, 2009, 2010). Ideologies
are the result of explicit processing; they are explicit belief systems, and they may motivate
behavior in a top-down manner. Examples are Marxism, liberal democracy, capitalism,
religious views of creation and an afterlife, and moral idealism (i.e., moral principles that
apply independently of human interests and thus may act to channel behavior in a non-selfinterested manner; see MacDonald, 2010).
A paradigmatic mental model of this type are the various utopian models of ideal human
behavior, such as Marxism. Marxism not only proposed an explanation for human behavior
and the patterns of history, it also envisioned a post-revolutionary world without social
classes or other hierarchical relationships between human groups, and it energized and
rationalized extreme violence against perceived class enemies.
It goes without saying that throughout history there have been many ideologies that have
been highly motivating. The basis for this claim is the control function of explicit processing
described above. That is, explicitly held beliefs are able to exert a control function over
behavior and over evolved predispositions (including EMDs), and they are able to exert this
control function independently of external processes of social control (e.g., punishment). For
example, a person may refrain from engaging in a particular behavior to which he is
predisposed as a result of evolved modules (e.g., various forms of aggression [MacDonald,
2008]), and he may do so because he believes that he would be sent to prison, or because of
he believes that he will be punished for it in an afterlife, or because he believes that it violates
an important moral principle or Gods law. The success of Calvinism in 16th-century Geneva
depended not only on the threat of externally applied sanctions, but also on the persuasiveness
of the explicit beliefs that constituted Calvinist religious ideology (Wilson, 2002).
CONCLUSION
Human general intelligence is a highly limited but extremely powerful domain general
ability. It is able to take in a very wide range of inputs, including cultural and linguistic input,
focus attention on these inputs and create plans for action. At the apex of this evolutionary
process, human general intelligence is able to create powerful scientific mental models able to
provide explanations and predictions of processes in the natural world. The mental models
50
Kevin MacDonald
created by general intelligence are fallible. In particular, ideological mental models produced
by explicit processing may often result in maladaptive behavior, as when ideologies without
rational foundation motivate behavior that is not in the interests those who accept the
ideology (see MacDonald, 2009, 2010).
Human general intelligence is the evolutionary outcome not of responding to regularities
in past environments, but rather of natural selection for ever more sophisticated ways of
achieving ancient evolutionary goals of survival and reproduction. As it has become
elaborated in humans, it is able to create and accept ideologies that go beyond evolutionary
goals by create goals that are far removed from evolutionary regularities and only tenuously
related to evolutionary fitness.
ACKNOWLEDGMENT
I thank Dan Chiappe for helpful comments on an earlier version of this paper.
REFERENCES
Bargh, J. A., & Chartrand, T. L. (1999). The unbearable automaticity of being. American
Psychologist 54, 462479.
Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind. Cambridge,
MA: MIT Press/Bradford Books.
Barrett, H. C., & Kurzban, R. (2006). Modularity in cognition: Framing the debate.
Psychological Review, 113, 628647.
Barrett, H. C., & Kurzban, R. (2012). What are the functions of System 2 modules? A reply to
Chiappe and Gardner (2012). Theory and Psychology, 22, 683688.
Carpenter, P., Just, M., & Shell, P. (1990). What one intelligence test measures: A theoretical
account of the processing in the Raven Progressive Matrices Test. Psychological Review,
97, 404431.
Case, R. (1998). The development of conceptual structure. In W. Damon (Series Ed.) & D.
Kuhn & R. S. Siegler (Vol. Eds.), Handbook of child psychology: Vol. 2. Cognition,
perception, and language (5th ed., pp. 745800). New York: Wiley.
Case, R., Demetriou, A., Platsidou, M., & Kazi, S. (2001). Integrating concepts and tests of
intelligence from the differential and developmental traditions. Intelligence, 29, 307336.
Chiappe, D., & Gardner, R. (2012). The modularity debate in evolutionary psychology.
Theory & Psychology, 22, 669682.
Chiappe, D., & MacDonald, K. B. (2005). The evolution of domain-general mechanisms in
intelligence and learning. Journal of General Psychology 132(1), 540.
Conway, A. R. A., Kane, M. J., & Engle, R. W. (2003). Working memory capacity and its
relation to general intelligence. Trends in Cognitive Sciences 7(12), 547552.
Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how humans
reason? Cognition, 31, 187276.
Dehaene, S., & Naccache, L. (2001). Towards a cognitive neuroscience of consciousness:
basic evidence and a workspace framework. Cognition, 79, 137.
51
Demetriou, A., Elklides, A, & Platsidou, M., (1993). The architecture and dynamics of the
developing mind. Monographs of the Society for Research in Child Development, 58(5/6,
Serial No. 234).
Dennett, D. (1987). Cognitive wheels: The frame problem of AI. In Z. Pylyshyn (Ed.), The
robots dilemma (pp. 4164). Norwood, NJ: Ablex.
Emmons, R. A. (1989). The personal striving approach to personality. In L. A. Pervin (Ed.),
Goal concepts in personality and social psychology (pp. 87126). Hillsdale, NJ: Erlbaum.
Engle, R. W. (2010). Role of working memory capacity in cognitive control. Current
Anthropology, 51(S1), S17S26.
Engle, R., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. (1999). Working memory,
short-term memory, and general fluid intelligence: A latent-variable approach. Journal of
Experimental Psychology: General, 128, 309331.
Feldman Barrett, L., Tugade, M. M., Engle, R. W. (2004). Individual differences in working
memory capacity and dual process theories of mind. Psychological Bulletin, 130. 553
573.
Fodor, J. A. (1983). The modularity of mind. Cambridge, MA: MIT Press.
Garcia, J., & Koelling, R. (1966). Relation of cue to consequence in avoidance learning.
Psychonomic Science, 4, 123124.
Gazzaley, A., & DEsposito, M. (2008). Unifying prefrontal cortex function. In B. Miller & J.
Cummings (Eds.) The Human Frontal Lobes, pp. 187206. New York: Guilford
Publications.
Geary, D. C. (2005). The origin of mind: Evolution of brain, cognition, and general
intelligence. Washington, DC: American Psychological Association.
Goetz, P., & D. Walters (1997) The dynamics of recurrent behaviour networks. Adaptive
Behaviour, 6(2), 245-282.
Gottfredson, L. (2007). Why g matters: The complexity of everyday life. Intelligence 24, 79
132.
Johnson-Laird, P. N. (1983). Mental models. Cambridge, UK: Cambridge University Press.
Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (little more than) working
memory capacity?! Intelligence, 14, 389433.
LeDoux, J. (1996). The emotional brain: The mysterious underpinnings of emotional life.
New York: Simon & Schuster.
LeDoux, J. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23, 155
184.
Kane, M. J., Hambrick, D. Z., & Conway, A. R. A. (2005). Working Memory Capacity and
Fluid Intelligence Are Strongly Related Constructs: Comment on Ackerman, Beier, and
Boyle (2005). Psychological Bulletin. 131(1), 6671.
Lieberman, M. D. (2007). Social cognitive neuroscience: A review of core processes. Annual
Review of Psychology, 58, 259289.
MacDonald, K. B. (1991). A perspective on Darwinian psychology: The importance of
domain-general mechanisms, plasticity, and individual differences. Ethology and
Sociobiology, 12, 449480.
MacDonald, K. B. (1995). Evolution, the five-factor model, and levels of personality. Journal
of Personality 63, 525567.
52
Kevin MacDonald
MacDonald, K., & Hershberger, S. (2005). Theoretical Issues in the Study of Evolution and
Development. In R. Burgess and K. MacDonald (Eds.), Evolutionary Perspectives on
Human Development, 2nd edition, pp. 2172. Thousand Oaks, CA: Sage.
MacDonald, K. B. (2008). Effortful control, explicit processing and the regulation of human
evolved predispositions. Psychological Review, 115(4), 10121031.
MacDonald, K. (2009). Evolution, psychology, and a conflict theory of culture. Evolutionary
Psychology, 7(2). 208233.
MacDonald, K. (April, 2010). Evolution and a Dual Processing Theory of Culture:
Applications to Moral Idealism and Political Philosophy. Politics and Culture
(unpaginated).
MacDonald, K. B. (2012). Temperament and evolution. In M. Zentner and R. L. Shiner
(Eds.), Handbook of Temperament. New York: Guilford Press, 273296.
Miller, E. K., & Cohen, J. (2001). An integrative theory of prefrontal cortex function. Annual
Review of Neuroscience, 24, 167202.
Morsella, E. (2005). The function of phenomenal states: Supramodular interaction theory.
Psychological Review, 112, 10001021.
Oller, K., & Griebel, U. (2005). Contextual freedom in human infant vocalization and the
evolution of language. In R. Burgess and K. MacDonald (Eds.), Evolutionary
Perspectives on Human Development, 2nd edition, pp. 135165. Thousand Oaks, CA:
Sage.
Penke, L., Denissen, J. J. A., & Miller, G. (2007). The evolutionary genetics of personality.
European Journal of Personality, 21, 549587.
Pinker, S. J. (1994). The language instinct. New York: William Morrow & Co.
Povinelli, D. J. (2000). Folk physics for apes: The chimpanzees theory of how the world
works. New York: Oxford University Press.
Rushton, J. P. (2004). Placing intelligence into an evolutionary framework or how g fits into
the rK matrix of life-history traits including longevity. Intelligence 32, 321328.
Satpute, A. B., & Lieberman, M. D. (2006). Integrating automatic and controlled processes
into neurocognitive models of social cognition. Brain Research, 1079, 8697.
Stanovich, K. E. (1999). Who is rational? Studies of individual differences in reasoning.
Hillsdale, NJ: Erlbaum.
Stanovich, K. E. (2004). The Robots Rebellion: Finding Meaning in the Age of Darwin.
Chicago: The University of Chicago Press.
Striedter, G. F. (2005). Principles of brain evolution. Sunderland, MA: Sinauer Associates.
Taylor, A. H., Elliffe, D. M., Hunt, G. R., & Gray, R. D. (2010). Complex cognition and
behavioural innovation in New Caledonian crows. Proceedings of the Royal Society,
Biological Sciences 277(1694): 26372643.
Taylor, A. H., Elliffe, D. M., Hunt, G. R., Emery, N. J., Clayton, N. S., & Gray, R. D. (2011).
New Caledonian Crows Learn the Functional Properties of Novel Tool Types. PLoS
ONE 6(12): e26887. doi:10.1371/journal.pone.0026887.
Tooby, J., & Cosmides, L. (1992). The psychological foundations of culture. In J. Barkow, L.
Cosmides, & J. Tooby (Eds.), The adapted mind: Evolutionary psychology and the
generation of culture (pp. 19136). New York: Cambridge University Press.
Weisfeld, G. (1997). Research on emotions and future developments in human ethology. In
A. Schmitt, K. Atzwanger, K. Grammer, & K. Schfer (Eds.), New aspects of human
ethology (pp. 2546). New York: Plenum.
53
West-Eberhard, M. J. (2003). Developmental plasticity and evolution. New York: Oxford

University Press.
Wilson, D. S. (2002). Darwins cathedral: Evolution, religion, and the nature of society.
Chicago: University of Chicago Press.
Zeki, S. (1993). A vision of the brain. London: Blackwell.

ISBN: 978-1-62618-728-3
Chapter 4
SUBTESTS, FACTORS, AND CONSTRUCTS:

WHAT IS BEING MEASURED BY TESTS
OF INTELLIGENCE?
Marley W. Watkins1,* and Christina M. Ravert2,
1
Department of Educational Psychology, Baylor University, Waco, Texas, US

2
Manheim Township School District, Lancaster, Pennsylvania, US
ABSTRACT
Confirmatory factor analyses were conducted to investigate hypotheses regarding the
Arithmetic and Digit Span subtests that form the third factor of the Wechsler Intelligence
Scale for Children. Previous research suggested that these subtests may be measures of
working memory, quantitative ability/reasoning, or some other ability. When the
Wechsler Intelligence Scale for Children-Third Edition was administered to a referral
sample of adolescents in conjunction with marker tests for memory and quantitative
reasoning, a five factor solution was optimal. The resulting factors were verbal
comprehension, perceptual organization, processing speed, quantitative reasoning, and
memory. Based on these results, the WISC Arithmetic and Digit Span subtests are
measures of different abilities with Arithmetic being a measure of quantitative reasoning
and Digit Span a measure of memory. Given these results, interpretation of the
Arithmetic subtest as a measure of memory may not be accurate.
Keywords: Intelligence, CFA, WISC-III, WISC-IV, Arithmetic, Digit Span
This report is based on the dissertation of the first author that was supervised by the second author. The
contributions of professors Barbara Schaefer, Hoi Suen, and Lisa Conyers are gratefully acknowledged. An
abbreviated version was presented at the International Test Commission meeting in Amsterdam, Netherlands.
Correspondence concerning this chapter should be addressed to Marley W. Watkins, Department of Educational
Psychology, Baylor University, One Bear Place #97301, Waco, TX 76798-7301. E-mail:
marley_watkins@baylor.edu.
christina_ravert@mtwp.net..
56
There appears to be a general consensus that child versions of the Wechsler Intelligence
Scale measure verbal comprehension and perceptual organization/reasoning abilities
(Zachary, 1990), but there has been less agreement as to the existence and nature of additional
abilities measured by the Wechsler child scale. For example, the first factor analysis of the
Wechsler Intelligence Scale for Children (WISC; Wechsler, 1949) found verbal
comprehension and perceptual organization dimensions as well as a third factor that was
loaded by the Arithmetic and Digit Span subtests (Cohen, 1959). This third factor was labeled
Freedom from Distractibility (FD) and its interpretation as a memory factor was explicitly
disclaimed.
Factor analyses of the Wechsler Intelligence Scale for Children-Revised (WISC-R;
Wechsler, 1974) also produced three factors labeled Verbal Comprehension (VC), Perceptual
Organization (PO), and Freedom from Distractibility (Kaufman, 1975). The first two factors
were consistent with those found with the WISC, but the FD factor was formed by the
Arithmetic, Digit Span, and Coding subtests. After endorsing the FD label, Kaufman (1975)
suggested that this factor might be a measure of numerical ability.
Wielkiewicz (1990) reported that a wide range of hypotheses had been advanced to
account for low scores on the FD factor, including problems with concentration/distractibility,
motivation, and memory. In fact, a joint factor analysis of the WISC-R and the Differential
Ability Scales (Elliott, 1990) revealed that the FD factor was not an intact entity (Stone,
1992). A summary of the results of eight studies that jointly factor analyzed the WISC-R and
the Woodcock Johnson-Revised Tests of Cognitive Abilities and Achievement (Woodcock &
Johnson, 1989) also concluded that the WISC-R FD subtests were actually measures of
separate abilities rather than common indicators of attention or distractibility (Woodcock,
1990). Specifically, Woodcock reported that Arithmetic was a measure of quantitative ability,
Coding was a measure of processing speed, and Digit Span was a measure of short-term
memory.
To help strengthen the FD factor, a new Performance subtest, Symbol Search, was added
to the WISC when it was next revised. It was assumed that the structure of the new Wechsler
Intelligence Scale for Children-Third Edition (WISC-III; Wechsler, 1991) would consist of
VC, PO, and FD factors. In actuality, the Coding and Symbol Search subtests split away to
form a new fourth factor, named Processing Speed (PS), while the Arithmetic and Digit Span
subtests remained as dual measures of the FD factor.
Over time, evidence has accumulated to suggest that the FD factor might not be a
measure of attention/distractibility (Cohen, Becker, & Campbell, 1990; Oakland, Broom, &
Glutting, 2000; Reinecke, Beebe, & Stein, 1999; Riccio, Cohen, Hall, & Ross, 1997). For
example, Riccio et al. (1997) correlated WISC-III factor index scores with behavioral and
neuropsychological measures and found that the FD factor did not significantly correlate with
any of the behavioral scales that measured attention. Similar findings were reported by
Lowman, Schwanz, and Kamphaus (1996), who found a non-significant relationship between
FD factor scores and hyperactivity and attention problem scale scores from the teacher-report
form of the Behavior Assessment System for Children (BASC-TRS; Reynolds & Kamphaus,
1992).
Given the weakness of the evidence supporting an attention/distractibility explanation of
the WISC third factor, Keith and Witta (1997) argued that Quantitative Reasoning would be a
better name because of its very high loading on g (.90), the obvious numerical content of
Arithmetic and Digit Span subtests, and the higher loading of Arithmetic (.82) than Digit
Subtests, Factors, and Constructs
57
Span (.52) on the FD factor. Carroll (1997) also opined that Quantitative Reasoning might be
a better name for the FD factor. In contrast, Kranzler (1997) posited that Working Memory
Efficiency provided a more appropriate label for the FD factor. Prifitera, Weiss, and
Soklofske (1998) agreed that Working Memory was a better name for the FD factor given the
tasks required by the FD subtests. These opinions were investigated by a joint confirmatory
factor analysis of the WISC-III and Woodcock Johnson-III cognitive and achievement tests
(Woodcock, McGrew, & Mather, 2001), which supported a model in which Arithmetic
loaded with math achievement measures on a quantitative knowledge factor and the Digit
Span subtest loaded on a short-term memory factor (Phelps, McGrew, Knopik & Ford, 2005).
In recognition of the debate surrounding the FD factor, the Wechsler Intelligence Scale
for Children-Fourth Edition (WISC-IV; Wechsler, 2003) third factor was renamed Working
Memory (WM). Additionally, a new subtest, Letter-Number Sequencing, was added to that
factor, the Digit Span subtest was retained as a core subtest, and the Arithmetic subtest was
lengthened and made an optional rather than mandatory WM subtest. Subsequent factor
analyses of the WISC-IV normative sample found problems with placement of Arithmetic
on the Working Memory factor (Keith, 2005, p. 595). Flanagan and Kaufman (2004) also
expressed concern about the Arithmetic subtest, suggesting that it is a measure of quantitative
knowledge rather than working memory.
After more than 50 years and four versions of the WISC, there remain four major
hypotheses about what is measured by the Arithmetic and Digit Span subtests. Specifically,
both subtests measure memory, both subtests measure quantitative reasoning, one subtest
measures memory and the other measures quantitative reasoning, or both subtests measure
some other ability. One way to test these hypotheses and identify the construct(s) that
underlie the WISC third factor is to administer tests that correspond with the quantitative
knowledge and short-term memory hypotheses in conjunction with the WISC. This is similar
to the confirmatory cross battery technique used to develop the integrated intelligence theory
(Flanagan, McGrew, & Ortiz, 2000). However, the tests used in the cross battery approach
were subtests selected from other intelligence measures, and many of those subtests lacked
sufficient reliability and validity evidence to verify that they are indeed psychometrically
sound measures of the hypothesized constructs (Phelps et al., 2005). Therefore, it is important
to use tests (marker variables) that have been well researched and found to be
psychometrically adequate measures of the respective constructs hypothesized to be measured
by the third factor. As noted by Child (2006), marker variables are those about which we
have reliable knowledge of their properties and are deliberately placed in factor analyses to
form a nucleus around which less well-defined variables can be assessed (p. 56).
The Educational Testing Service (ETS) developed a kit of brief factor-referenced
cognitive tests (Ekstrom, French, Harman, & Dermen, 1976) that Carroll (1985)
recommended for use as cognitive marker variables. Only factors that were deemed
sufficiently established were included. To select the factors to consider for the kit, a panel of
20 prominent persons working in the field of factor analysis and human assessment set the
necessary criteria (Ekstrom, 1973). A total of 72 marker variables for 23 cognitive factors
were included in the 1976 edition of the Kit of Factor-Referenced Cognitive Tests (Ekstrom
et al., 1976). Marker variables for two of these cognitive factors (memory span and general
reasoning) were utilized in the present study. It should be noted that the general reasoning
factor consists solely of marker variables that are quantitative in nature and can thus be
considered a quantitative reasoning factor (Carroll, 1993).
58
If the WISC third factor is really a measure of either short-term memory or quantitative
reasoning ability, then the data obtained from this study should fit a model where the Digit
Span and Arithmetic subtests align with that construct as represented by its marker variables.
It is also possible that the data will best fit a model where the two third factor subtests align
with measures of separate constructs as suggested by Woodcock (1990) and Flanagan et al.
(2000). In that case, the data would best fit a model where the Arithmetic subtest loads on a
factor consisting of the quantitative reasoning marker variables and the Digit Span subtest
loads on a factor comprised of the memory marker variables. If the third factor truly measures
attention or some other ability, then the data should best fit a model where the third factor
subtests remain a separate factor.
Therefore, the purpose of this study is to investigate the nature of the third factor of the
WISC. The WISC-III was chosen because its administration routinely included both Digit
Span and Arithmetic subtests whereas Arithmetic is an optional subtest on the WISC-IV and
optional subtests are rarely administered in clinical practice (Canivez & Watkins, 1998).
Marker variables reflecting the constructs of short-term memory and quantitative reasoning
were also administered and the resulting data were analyzed using confirmatory factor
analysis to identify the construct(s) being measured by the third factor.
METHOD
Participants
A total of 87 (51 male and 36 female) students in grades 6-12 from the mid-Atlantic
region of the country participated in this study. Ethnicity of the participating students was
reported to be 85% White, 9% Black, 4% Hispanic, and 2% Other. The majority of the
students were in grades six through eight (92%) with a mean age of 13.4 years (SD = 1.3).
Both special education identified (21% with learning disabilities, 13% with mental
retardation, 8% with serious emotional disabilities, 8% gifted, and 5% with other heath
impairments) and referred but nonidentified students (45%) were included in the sample.
Instruments
The WISC-III is appropriate for children between the ages of 6 years, 0 months through
16 years, 11 months. It contains 13 subtests, but only 10 are mandatory. If both Digit Span
and Symbol Search are administered in conjunction with the ten required subtests, four factor
indices can be computed: VC, PO, FD, and PS. Each of the four factor indices has a mean of
100 and a standard deviation of 15. These factor indices were derived based on the results of
exploratory and confirmatory factor analyses of the standardization sample data using 12 of
the WISC-III subtests. Full details of the WISC-III and its standardization are presented in
Wechsler (1991). Additional reliability and validity data are provided by Sattler (2001) as
well as Zimmerman and Woo-Sam (1997).
A Kit of Factor-Referenced Cognitive Tests was developed by the Educational Testing
Service (Ekstrom et al., 1976) to provide researchers with a means of identifying cognitive
59
factors in factor analytic studies. The use of common variables (marker variables) that assess
identified constructs can facilitate the interpretation of other factors that contain those
variables (Carroll, 1997; Ekstrom et al., 1976). The specific tests or marker variables that
were included in the 1976 edition of the kit were selected after reviewing the 1963 version of
the kit and the relevant literature published between editions of the kit. In addition, the
variables were subjected to field experiments to verify that they were useful markers for their
respective factors.
This study utilized marker variables of memory span and general reasoning factors.
Visual Number Span and Auditory Letter Span, which are appropriate for students in grades
6-16, were selected from the memory span factor. With the Visual Number Span test, the
student is visually presented with a series of digits of varying lengths at a rate of 1 digit per
second. Immediately after the visual presentation of the series, the student is asked to write
the series of numbers from memory. The Auditory Letter Span test requires the student to
listen to a series of letters and then write the series from memory after it has been presented.
Two quantitative reasoning marker variables appropriate for students in grades 6-12 were
also selected. For the Arithmetic Aptitude Test, the student is asked to read word problems
that require only arithmetic to solve, and instructed to solve the problem and choose the
correct answer from among five choices that are presented within a ten-minute time limit. In
contrast, for the Necessary Arithmetic Operations test the student is asked to read a word
problem and select the arithmetic operation(s) needed to solve the problem. Four answer
choices are provided for each problem, and the student is given five minutes to complete the
test.
Psychometric evidence on these marker variables was provided by Ekstrom et al. (1976).
For example, reliability estimates for Visual Number Span and Auditory Letter Span were .63
and .65, respectively, and the memory span factor and/or its marker variables had been
referenced in 21 studies prior to publication of the kit. A reliability estimate of .82 was
reported for the Arithmetic Aptitude Test and .72 for the Necessary Arithmetic Operations
test. The general reasoning factor and its marker variables had been referenced in 67 studies
prior to publication of the kit. The Arithmetic Aptitude Test and Necessary Arithmetic
Operations were timed tasks making them inappropriate for calculating an internal
consistency coefficient in this sample. The remaining marker tests, Auditory Letter Span and
Visual Number Span, had alpha coefficients of .74 and .80, respectively, in this sample.
Procedures
Data were collected by alumni and graduate students of a mid-Atlantic doctoral school
psychology program following approval by the universitys IRB. A total of 235 non-retired
alumni were asked to obtain parental consent for additional testing on students referred for a
psycho-educational evaluation that contained the WISC-III. Alumni could either collect all
WISC-III and marker variable data themselves or allow doctoral school psychology students
to collect part or all of the data in their school districts. A total of 36 cases were collected by
22 alumni whereas the remaining 51 cases were collected by graduate students in four school
districts. In all cases, the WISC-III was administered first, followed by the marker variables.
60
Analysis
Confirmatory factor analysis (CFA) using maximum likelihood estimation was applied to
the covariance matrix using Mplus version 7 for the Macintosh. The obtained solutions were
checked for convergence and the adequacy of the parameter estimates and their associated
standard errors were examined prior to considering the reported fit indices. According to Hu
and Bentler (1998), values .95 for CFI, .08 for SRMR, and .06 for RMSEA indicate that
there is a relatively good fit between the hypothesized model and the sample data.
Consequently, these cutoff values were applied in the current study. Models were also
evaluated with the Bayesian Information Criterion (BIC; Schwarz, 1978), which is a
parsimony corrected measure where the smallest value indicates the relatively best fit among
the tested models.
Various experts have proposed inconsistent rules of thumb to determine the minimum
sample size needed for factor analysis, but recent research has revealed that when
communalities are high, good recovery of population factors can be achieved with relatively
small samples (MacCallum, Widaman, Preacher, & Hong, 2001). In most studies,
communalities .60 were considered high. Given a median communality of .66 for the
current variables, good recovery of population factors is feasible with a smaller sample size.
Based on model parameters, a sample size of 78 would be needed to detect poor fit (Loehlin,
2003). Therefore, the current sample size of 87 should be sufficient.
Each of the four models selected for this study was designed to test a specific hypothesis
about the third factor or the subtests loading on that factor. Given that there is evidence to
suggest that the Arithmetic and Digit Span subtests are measures of memory and/or
quantitative reasoning, models were developed to reflect each of these hypotheses.
Model I
The first model was designed to test the hypothesis that the WISC third factor is really a
measure of memory span (MS). In this model, five factors were specified: VC, PO, PS, Shortterm Memory Span (MS), and Quantitative Reasoning (QR). The VC factor was hypothesized
to have significant loadings from four WISC-III subtests including Information, Vocabulary,
Similarities, and Comprehension. Likewise, four WISC-III subtests were hypothesized to load
on the PO factor including Picture Completion, Picture Arrangement, Block Design, and
Object Assembly. The PS factor was hypothesized to consist of the WISC-III Coding and
Symbol Search subtests, and the MS factor was hypothesized to include the WISC-III
Arithmetic and Digit Span subtests as well as the memory span marker variables including the
Visual Number Span Test and the Auditory Letter Span Test (Ekstrom et al., 1976). The
quantitative reasoning (QR) factor was hypothesized to be made up of the marker variables
for the general reasoning factor including the Arithmetic Aptitude Test Part 1 and the
Necessary Arithmetic Operations Test (Ekstrom et al., 1976).
Model II
The second model was developed to test the hypothesis that the WISC third factor really
reflects a quantitative reasoning construct. As with the first model, this model also consisted
of five factors: VC, PO, PS, MS, and QR. The VC, PO, and PS factors were specified to have
the same subtest composition as in the first model; however, the MS factor was hypothesized
61
to contain only the two marker variables for memory span (Ekstrom et al., 1976). In this
model, the WISC-III Arithmetic and Digit Span subtests were specified to load on the QR
factor along with the two marker variables for quantitative reasoning (Ekstrom et al., 1976).
Model III
The third model was constructed to test the hypothesis that the WISC Arithmetic subtest
is a measure of quantitative reasoning and the Digit Span subtest is a measure of memory.
Five factors were also specified for this model including VC, PO, PS, MS, and QR. The first
three factors were specified to have the same subtest composition that they had in Models I
and II. However, in this third model the MS factor was hypothesized to include the two
memory span marker variables as well as the WISC-III Digit Span subtest, and the QR factor
was specified to have significant loadings from the WISC-III Arithmetic subtest and the two
quantitative reasoning marker variables.
Table 1. Fit Statistics for Plausible Models Employing Twelve Subtests from the
Wechsler Intelligence Score for Children-Third Edition and Four Marker Variables
from Memory and Quantitative Reasoning Factors
Model
I
II
III
IV
III-Hierarchical
df
94
94
94
89
99
2
148.75
138.63
111.34*
132.51
119.14*
CFI
.930
.943
.978
.944
.974
SRMR
.077
.062
.050
.060
.055
RMSEA
.082
.074
.046
.075
.048
RMSEA
90% CI
.056-.106
.046-.099
.000-.077
.046-.101
.000-.078
BIC
7947.01
7936.89
7909.60
7953.10
7895.07
* p .05.
Model IV
The fourth model was specified to consider the possibility that none of the evaluated
hypotheses were correct, and the WISC third factor is a measure of some other construct such
as attention or executive processes. Six factors were specified in this model including VC,
PO, PS, MS, QR, and FD. The first three factors were specified to have the same composition
that they had in Models I, II, and III. The MS and QR factors were hypothesized to be formed
by their respective marker variables (Ekstrom et al. 1976), and the FD factor was specified to
include the WISC-III Arithmetic and Digit Span subtests.
RESULTS
Means (with standard deviations in parentheses) for the WISC-III Full Scale IQ, Verbal
IQ, and Performance IQ were 96.8 (17.9), 97.5 (17.9), and 96.7 (17.5), respectively. An
examination of the univariate skewness and kurtosis values for the 16 variables under
consideration indicated that all were within accepted limits for normality (Fabrigar, Wegener,
MacCallum, & Strahan, 1999).
62

IN
.83
VOC
.97
SIM
.84
VC
.72
COM
.81
PC
.67
PA
.79
BD
.83
OA
CD
SS
.55
PO
.64
.63
.72
.84
.84
.52
.76
PS
.83
.41
DS
VNS
.77
.81
.65
MS
.81
ALS
.64
AR
AAT
NAO
.85
.55
QR
.64
Note. IN = WISC-III Information subtest, VOC = WISC-III Vocabulary subtest, SIM = WISC-III
Similarities subtest, COM = WISC-III Comprehension subtest, PC = WISC-III Picture Completion
subtest, PA = WISC-III Picture Completion subtest, BD = WISC-III Block Design subtest, OA =
WISC-III Object Assembly subtest, CD = WISC-III Coding subtest, SS = WISC-III Symbol
Search subtest, DS = WISC-III Digit Span subtest, AR = WISC-III Arithmetic subtest, VNS =
Visual Number Span marker variable, ALS = Auditory Letter Span marker variable, AAT =
Arithmetic Aptitude Test marker variable, NAO = Necessary Arithmetic Operations marker
variable, VC = Verbal Comprehension factor, PO = Perceptual Organization factor, PS =
Processing Speed factor, MS = Memory factor, and QR = Quantitative Reasoning factor.
Figure 1. Most Plausible Model Employing Twelve Subtests from the Wechsler Intelligence Score for
Children-Third Edition (WISC-III) and Four Marker Variables from Memory and Quantitative
Reasoning Factors.
63
Two other conditions for multivariate normality are that all linear combinations of
variables follow a normal distribution and all subsets of variables in the data set are normally
distributed (Stevens, 2009). This was verified by examining the scatterplots of all possible
variable pairs. All scatterplots had an elliptical shape. In addition, multivariate kurtosis was
examined with the Normalized Estimate (Mardia, 1974) value calculated with EQS 6.1.
Normalized Estimate values between -3 and +3 indicate that the data is multivariate normal
(Bentler & Wu, 2002). The Normalized Estimate value for the sample data was -.51,
indicating that that the data was multivariate normal.
A review of model fit statistics (see Table 1) indicates that only Model III met a priori
guidelines for good fit (i.e., CFI .95, SRMR .08, and RMSEA .06) and was the only
model with a nonsignificant chi-square value. Model III was also relatively superior to the
other models based on CFI (Chen, 2007) and BIC (Raftery, 1995) comparisons. Model III
specified the WISC-III Arithmetic subtest as a measure of quantitative reasoning and the
Digit Span subtest as a measure of memory.
Model III (with standardized coefficients) is illustrated in Figure 1. The five factors were
substantially correlated, with coefficients ranging from .41 between the PS and MS factors to
.84 between the VC and QR factors. The correlations among first order factors might be
explained by a second order factor that would be more parsimonious than the first order
structure of Model III (Thompson, 2004). In recognition of the importance of a higher order
factor in investigations of cognitive ability (Jensen, 2002), a hierarchical structure was added
to Model III. As anticipated, this higher order variant of Model III was an excellent fit to the
data (see Table 1). The standardized coefficients displayed in Figure 2 indicate that all five
first order factors were strongly related to the higher order general intelligence factor.
DISCUSSION
Confirmatory factor analyses were conducted to investigate several hypotheses regarding
the Arithmetic and Digit Span subtests that formed the third factor of the WISC. Previous
research suggested that the subtests may be measures of working memory, quantitative
ability/reasoning, or some other ability. Alternatively, the two WISC subtests may be
measures of different abilities. When the WISC-III was administered in conjunction with
marker tests for memory and quantitative reasoning, a five factor solution was optimal.
Results from the current study suggested that the WISC-III Arithmetic and Digit Span
subtests are measures of different abilities with Arithmetic being a measure of quantitative
reasoning and Digit Span a measure of memory.
These results are in line with theory and research developed from cross battery
assessment of the Wechsler and Woodcock Johnson scales (Flanagan et al., 2000; McGrew &
Woodcock, 2001; Phelps et al., 2005; Woodcock, 1990; Woodcock et al., 2001) and are
consistent with the theory and research results reported by Carroll (1993, 1997), who
indicated that the Arithmetic subtest may be a poor representation of fluid intelligence and
Digit Span may represent general memory. Additional support for the current results was
provided by Keith (2005) who compared results from hierarchical and nested-factors CFA
models in the WISC-IV standardization sample and found that the loading of Arithmetic on
64
the Working Memory factor varied considerably between the two models (hierarchical model
= .80 and nested-factors model = .11).
IN
.83
VOC
.97
SIM
.84
VC
.72
COM
PC
.67
PA
.77
BD
.85
OA
PO
.90
.63
.90
CD
.80
SS
.80
PS
.69
.66
DS
VNS
.77
.82
.93
MS
.81
ALS
AR
AAT
NAO
.85
.54
QR
.64
Note. IN = WISC-III Information subtest, VOC = WISC-III Vocabulary subtest, SIM = WISC-III
Similarities subtest, COM = WISC-III Comprehension subtest, PC = WISC-III Picture Completion
subtest, PA = WISC-III Picture Completion subtest, BD = WISC-III Block Design subtest, OA =
WISC-III Object Assembly subtest, CD = WISC-III Coding subtest, SS = WISC-III Symbol
Search subtest, DS = WISC-III Digit Span subtest, AR = WISC-III Arithmetic subtest, VNS =
Visual Number Span marker variable, ALS = Auditory Letter Span marker variable, AAT =
Arithmetic Aptitude Test marker variable, NAO = Necessary Arithmetic Operations marker
variable, VC = Verbal Comprehension factor, PO = Perceptual Organization factor, PS =
Processing Speed factor, MS = Memory factor, QR = Quantitative Reasoning factor, and g =
General Intelligence factor.
Figure 2. An Hierarchical Version of the Most Plausible Model Employing Twelve Subtests from the
Wechsler Intelligence Score for Children-Third Edition (WISC-III) and Four Marker Variables from
Memory and Quantitative Reasoning Factors.
65
The low loading of Arithmetic on the Working Memory factor in the nested-factors
model suggests that it may not be a measure of working memory. In contrast, exploratory and
confirmatory factor analyses in the WISC-IV manual suggested that the Arithmetic subtest
does associate with the working memory subtests. However, no other quantitative tasks were
included in those analyses. In addition, an inspection of the WISC-IV exploratory factor
analytic loadings broken down by age group reveals a decline in Arithmetics loading on the
Working Memory factor as age increases (.73 for 6-7 year-olds, .57 for 8-10 year-olds, .39 for
11-13 year-olds, and .40 for 14-16 year-olds).
In agreement with the results from the current study, Wechsler (2003) placed the Digit
Span subtest of the WISC-IV on a Working Memory factor. Contrary to these results,
Wechsler (2003) specified the Arithmetic subtest of the WISC-IV as an optional measure of
Working Memory. Even after modifying the Arithmetic subtest to increase demands on
working memory, researchers have questioned the placement of Arithmetic on the Working
Memory factor (Keith, 2005, p. 595). Flanagan and Kaufman (2004) also expressed concern
about the WISC-IV Arithmetic subtest, suggesting that it is a measure of quantitative
knowledge rather than working memory. Although the Arithmetic subtest is no longer a
required subtest on the WISC-IV, it can be substituted for Digit Span or Letter-Number
Sequencing. Given the current results as well as the concerns of Keith (2005) and Flanagan
and Kaufman (2004), it may not be prudent to compute a WISC-IV working memory index
score when the Arithmetic subtest is substituted for either WM core subtest, nor to interpret
the Arithmetic subtest as a measure of working memory until additional research has
specifically investigated its factorial validity. After substitution, the underlying construct
intended to be measured by the index may change (Kaufman, Flanagan, Alfonso, &
Mascolo, 2006, p. 291) and the validity of the resulting Indexes and Full Scale IQ is
unknown (Ryan & Glass, 2006, p. 190).
This study was not without its limitations, which must be considered when examining the
results and their implications. First, the size of the sample was small for CFA methods. CFA
is generally regarded as a large sample methodology; however, more recent research suggests
that it can be used with smaller samples under favorable conditions such as high variable
communalities (MacCallum et al., 2001). Fortunately, communalities were relatively high for
the current sample.
A second limitation of the current sample is that it was far less representative of the
population of exceptional students in the United States than desired. Almost the entire sample
was from the Mid-Atlantic section of the United States and some exceptionalities were not
represented, including visual impairment, autism, and multiple disabilities. In addition, most
of the students were White and of middle school age. Therefore, it is unclear how these
results would generalize to the population of students referred for psychoeducational testing
and to the general population.
Finally, using volunteer professionals and graduate students to collect data may have
impacted the results. Some participants were tested by professional school psychologists and
others by graduate students while still others may have been tested by more than one
examiner across time. The impact of these disparate procedures is unknown, but examiner
familiarity effects have been demonstrated (Fuchs & Fuchs, 1986). Given these limitations,
additional research in this area is needed to validate findings with other Wechsler scales.
Until then, clinicians should be circumspect when using the Wechsler Arithmetic subtest as a
measure of working memory.
66
REFERENCES
Bartlett, M. S. (1954). A note on the multiplying factors for various chi square
approximations. Journal of the Royal Statistical Society, 16 (Series B), 296-298.
Bentler, P. M., & Wu, E. J. C. (2002). EQS 6 for Windows users guide. Encino, CA:
Multivariate Software.
Canivez, G. L., & Watkins, M. W. (1998). Long-term stability of the Wechsler Intelligence
Scale for Children-Third Edition. Psychological Assessment, 10, 285-291.
Carroll, J. B. (1985). Exploratory factor analysis: A tutorial. In D. K. Detterman (Ed.),
Current topics in human intelligence, Vol. 1: Research methodology (pp. 25-58).
Norwood, NJ: Ablex.
Carroll, J. B. (1993). What abilities are measured by the WISC-III? In B. A. Bracken (Ed.),
Journal of Psychoeducational Assessment Monograph Series Advances in
Psychoeducational Assessment Wechsler Intelligence Scale for Children: Third Edition
(pp. 134-143). Germantown, TN: Psychological Corporation.
Carroll, J. B. (1997). Commentary on Keith and Wittas hierarchical cross-age confirmatory
factor analysis of the WISC-III. School Psychology Quarterly, 12, 108-109.
Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance.
Structural Equation Modeling, 14, 464-504.
Child, D. (2006). The essentials of factor analysis (3rd ed.). New York: Continuum.
Cohen, J. (1959). The factorial structure of the WISC at ages 7-6, 10-6, and 13-6. Journal of
Consulting Psychology, 23, 285-299.
Cohen, M., Becker, M. G., & Campbell, R. (1990). Relationships among four methods of
assessment of children with Attention Deficit-Hyperactivity Disorder. Journal of School
Ekstrom, R. B. (1973). Cognitive factors: Some recent literature. (Technical Report No. 2).
Princeton, NJ: Educational Testing Service.
Ekstrom, R. B., French, J. W., Harman, H. H., & Dermen, D. (1976). Kit of factor referenced
cognitive tests. Princeton, NJ: Educational Testing Service.
Elliott, C. D. (1990). Introductory and technical handbook for the Differential Ability Scales.
San Antonio, TX: Psychological Corporation.
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the
use of exploratory factor analysis in psychological research. Psychological Methods, 4,
272-299.
Flanagan, D. P., & Kaufman, A. S. (2004). Essential of WISC-IV assessment. Hoboken, NJ:
Wiley.
Flanagan, D. P., McGrew, K. S., & Ortiz, S. O. (2000). The Wechsler Intelligence Scales and
Gf-Gc theory: A contemporary approach to interpretation. Needham Heights, MA: Allyn
& Bacon.
Fuchs, D., & Fuchs, L. S. (1986). Test procedure bias: A meta-analysis of examiner
familiarity effects. Review of Educational Research, 56, 243-262.
Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis.
Psychometrika, 30, 179-185.
Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity to
underparameterized model misspecification. Psychological Methods, 3, 424-453.
67
Jensen, A. R. (2002). Psychometric g: Definition and substantiation. In R. J. Sternberg & E.

L. Grigorenko (Eds.), The general factor of intelligence: How general is it? (pp. 39-53).
Mahwah, NJ: Erlbaum.
Kaiser, H. F. (1960). The application of electronic computers to factor analysis. Educational
& Psychological Measurement, 20, 141-151.
Kamphaus, R. W. (1993). Clinical Assessment of childrens intelligence. Boston, MA: Allyn
and Bacon.
Kaufman, A. S. (1975). Factor analysis of the WISC-R at 11 age levels between 6 and 16
years. Journal of Consulting and Clinical Psychology, 43, 135-147.
Kaufman, A. S., Flanagan, D. P., Alfonso, V. C., & Mascolo, J. T. (2006). [Review of the
Wechsler Intelligence Scale for Children, Fourth Edition (WISC-IV)]. Journal of
Psychoeducational Assessment, 24, 278-295. 56.
Keith, T. Z. (2005). Using confirmatory factor analysis to aid in understanding the constructs
measured by intelligence tests. In D. P. Flanagan & P. L. Harrison (Eds.), Contemporary
intellectual assessment: Theories, tests, and issues (2nd ed.; pp. 581-614). New York:
Guilford.
Keith, T. Z., Fine, J. G., Taub, G. E., Reynolds, M. R., & Kranzler, J. H. (2006). Higher
order, multisample, confirmatory factor analysis of the Wechsler Intelligence Scale for
Children-Fourth Edition: What does it measure? School Psychology Review, 35, 108-127.
Keith, T. Z., & Witta, E. L. (1997). Hierarchical and cross-age confirmatory factor analysis of
the WISC-III: What does it measure? School Psychology Quarterly, 12, 89-107.
Kranzler, J. H. (1997). What does the WISC-III measure? Comments on the relationship
between intelligence, working memory capacity, and information processing speed and
efficiency. School Psychology Quarterly, 12, 110-116.
Loehlin, J. C. (2003). Latent variable models: An introduction to factor, path, and structural
analysis (4th ed.). Mahwah, NJ: Erlbaum.
Lowman, M. G., Schwanz, K. A., & Kamphaus, R. W. (1996). WISC-III third factor: Critical
measurement issues. Canadian Journal of School Psychology, 12, 15-22.
MacCallum, R. C., Widaman, K. F., Preacher, K. J., & Hong, S. (2001). Sample size in factor
analysis: The role of model error. Multivariate Behavioral Research, 36, 611-637.
Mardia, K. V. (1974). Applications of some measures of multivariate skewness and kurtosis
in testing normality and robustness studies. Sankhya, B36, 115-128.
McGrew, K. S., & Woodcock, R. W. (2001). Technical manual. Woodcock Johnson III.
Itasca, IL: Riverside Publishing.
Oakland, T., Broom, J., & Glutting, J. (2000). Use of freedom from distractibility and
processing speed to assess children's test-taking behaviors. Journal of School Psychology,
38, 469-475.
Phelps, L., McGrew, K. S., Knopik, S. N., & Ford, L. (2005). The general (g), broad, and
narrow CHC stratum characteristics of the WJ-III and WISC-III tests: A confirmatory
cross-battery investigation. School Psychology Quarterly, 20, 66-88.
Prifitera, A., Weiss, L. G., & Soklofske, D. H. (1998). The WISC-III in context. In A.
Prifitera & D. H. Soklofske (Eds.), WISC-III clinical use and interpretation: Scientistpractitioner perspectives (pp. 1-38). San Diego, CA: Academic Press.
Raftery, A. E. (1995). Bayesian model selection in social research. Sociological Methodology,
25, 111-163.
68
Reinecke, M. A., Beebe, D. W., & Stein, M. A. (1999). The third factor of the WISC-III: Its
(probably) not Freedom from Distractibility. Journal of the American Academy of Child
& Adolescent Psychiatry, 38, 322-328.
Reynolds, C. R., & Kamphaus, R. W. (1992). Behavior Assessment System for Children
Teacher Rating Scales (BASC-TRS). Circle Pines, MN: American Guidance Service.
Riccio, C. A., Cohen, M. J., Hall, J., & Ross, C. M. (1997). The third and fourth factors of the
WISC-III: What they dont measure. Journal of Psychoeducational Assessment, 15, 2739.
Ryan, J. J., & Glass, L. A. (2006). Substituting supplementary subtests for core subtests on
reliability of WISC-IV indexes and full scale IQ. Psychological Reports, 98, 187-190.
Sattler, J. M. (2001). Assessment of children: Cognitive applications (4th ed.). San Diego,
CA: Author.
Schwarz, G. (1978). Estimating the dimension of a model. Annals of Statistics, 6, 461464.
Stevens, J. P. (2009). Applied multivariate statistics for the social sciences (5th ed.). New
York: Routledge.
Stone, B. J. (1992). Joint confirmatory factor analyses of the DAS and WISC-R. Journal of
School Psychology, 30, 185-195.
Thompson, B. (2004). Exploratory and confirmatory factor analysis: Understanding concepts
and applications. Washington, DC: American Psychological Association.
Wechsler, D. (1949). Wechsler Intelligence Scale for Children. New York: Psychological
Corporation.
Wechsler, D. (1974). Wechsler Intelligence Scale for Children-Revised Edition manual. New
York: Psychological Corporation.
Wechsler, D. (1991). Wechsler Intelligence Scale for Children-Third Edition manual. San
Antonio, TX: Psychological Corporation.
Wechsler, D. (2003). Wechsler Intelligence Scale for Children-Fourth Edition technical and
interpretive manual. San Antonio, TX: Psychological Corporation.
Wielkiewicz, R. M. (1990). Interpreting low scores on the WISC-R third factor: Its more
than distractibility. Psychological Assessment, 2, 91-97.
Woodcock, R. W. (1990). Theoretical foundations of the WJ-R measures of cognitive ability.
Journal of Psychoeducational Assessment, 8, 231-258.
Woodcock, R. W., & Johnson, M. B. (1989). Woodcock Johnson Psycho-Educational
Battery-Revised. Allen, TX: DLM Teaching Resources.
Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). Woodcock Johnson III tests of
cognitive abilities. Itasca, IL: Riverside Publishing.
Zachary, R. A. (1990). Wechslers intelligence scales: Theoretical and practical
considerations. Journal of Psychoeducational Assessment, 8, 276-289.
Zimmerman, I. L., & Woo-Sam, J. M. (1997). Review of the criterion-related validity of the
WISC-III: The first five years. Perceptual and Motor Skills, 85, 531-546.
Zwick, W. R., & Velicer, W. F. (1986). Comparison of five rules for determining the number
of components to retain. Psychological Bulletin, 99, 432-442.

ISBN: 978-1-62618-728-3
Chapter 5
RACE, GENOMICS AND INTELLIGENCE:

SLIGHT RETURN
Joint School for Nanosciences & Nanoengineering,
North Carolina A and T State University and UNC Greensboro,
Greensboro, NC, US
ABSTRACT
The notion that genetic/genomic differences among racial groups are responsible for
observed differences in measured intelligence is recurring. Certainly, genetics, race, and
intelligence arguments have been advanced with each increase in the sophistication of
genetics (Graves 2005a). Not surprisingly this has continued in the modern genomics era.
The reader must be aware that there are significant definitional issues that impact theseis
arguments. The first refers to the notion of race. The second definition that must be
considered is the notion of intelligence. I refer the reader to Sternberg (2012) for a good
review of the intelligence question. Surprisingly (not too surprisingly once we familiarize
ourselves with the history of the race and intelligence debate; see Graves 2005b),
measurers of intelligence have not effectively addressed racial definitions. In the main,
they have focused their attention on differences between groups of people that are clearly
socially defined, but not so clearly biologically differentiated in ways that would support
their claims. Thus this chapter will first discuss and describe human biological/ genetic
differentiation, specifically with how this variation impacts genetic causal factors that
purportedly impact intelligence, and finally modern genomic approaches to the genetic
influences on intelligence.
WHAT DO THEY MEAN BY RACE? DOES PSYCHOMETRIC

RESEARCH ADHERE TO BIOLOGICAL/GENETIC DEFINITIONS?
None of the proponents of racial psychometry have ever fully outlined what they mean by
race. Historically their work has focused on measured differentials between what are socially
70
defined but not biological races. In the fall of 2012 (last few weeks at the time of this
writing), two of the giants of racial psychometry, J. Phillipe Rushton and Arthur Jensen,
passed away. Rushton spent a great deal of time attempting to demonstrate the existence of
biological races in the human species (Rushton, 2000). His views on this issue were widely
accepted by psychometricians. He identified three major racial divisions of mankind, blacks,
whites, and Orientals. In Race, Evolution, and Behavior, Rushton claimed that these three
racial divisions differed in a series of traits such as brain size, intelligence, reproduction,
personality, maturation, and social organization. His argument was that differences in lifehistory evolution (specifically via the r- and K-selection continuum) accounted for these
racial features. Rushtons argument failed on a number of levels (Graves, 2002a, b). Not the
least of which is the general failure of r-/K-selection models to predict anything of value in
actual real-world organisms; the fact that he applied the r-/K-model backward, and finally the
quality of his data was of suspect with regard to comparing the groups he described (Graves,
2002 a, b).
All of these failures are made even worse when one realizes that he was applying them to
an essentially biologically arbitrary grouping of people. Blacks, Whites, and Orientals are
categories that have social but not biological reality. Rushton was actually relying on a 17th
century conception of biological race. Specifically he was selectively measuring
morphological traits and utilizing those to define membership in races. There have been four
ways in which the race concept has been conceived: essentialist, taxonomic, population, and
lineage. Essentialist notions of race are ancient, and are not necessarily related to any concept
of common descent. Essentialism claims that there is an essence of traits that can be
understood as characterizing a species (and thus any races that might exist within it). The
taxonomic race concept suggests that there are aggregate populations of a species possessing
phenotypic similarities and inhabiting geographic subdivisions of the range of the species.
The taxonomic race conception fails due to the principle of discordance. Discordance means
that physical traits, which result from differential portions of an organisms genome, are
being influenced by different and often unrelated selection regimes. So for example, we do
not expect that selection regimes which influence skin color (solar intensity) to be associated
with those that influence height. The Kenyan Watusi are tall people, displaying tropical body
proportions and have dark skin. The Aleut people are short people displaying arctic body
proportions with light skin. Yet, there are Biaka pygmies who are short and have dark skin.
Thus, evolutionary biologists long ago rejected the essentialist and taxonomic definitions of
species (and thereby those same definitions applied to race as well). This means that only the
population and lineage definitions of race retain any utility.
Population definitions of race revolve around how much genetic variation exists within
and between supposed racial groups. If there is more genetic variation within a group than
between them, we really cannot support the notion that the groups have diverged sufficiently
to describe biological races. For evolutionary biology, biological races occur as part of the
speciation process. Through the species history, local adaptation and genetic drift may cause
sufficient changes in some populations such that they eventually form new species from the
original founders. This process is dynamic, and populations may diverge or converge
genetically over the lifetime of the species without ever giving rise to new ones.
Obviously, given this process, the amount of variation cannot be evaluated at one or a
small number of genetic loci. Or so it would seem if geographical races result from the
71
gradual process of speciation. Ernst Mayr discussed this in his classic work, Population
Species, and Evolution.
Biological Races
A necessary corollary of any theory of gradual speciation is that there should exist in
nature forms or varieties or populations that are incipient species. Kinds of animals that
show no (or only slight) structural differences, although clearly separable by biological
characters, are called biological races.(p. 258).
Mayr did not explain what he meant by clearly separable, but he did state that of all the
phenomena listed as races in the biological literature, the best case for the existence of true
biological races in nature are the host races formed by various insect species on different
plant species. It is notable that he did not think that humans fit this definition well. Yet it does
not take much genetic variation at all to account for host-race formation in insects. In the
most well-studied case of this, Rhagoletis pomonella (the apple maggot fly) has formed two
host races throughout its geographic range. The fly ancestral host is the hawthorn tree
(Crataegus spp). About 150 years ago, this fly species was recorded as being a pest of
cultivated apple trees (Malus). It is now known that allele frequencies at about four loci are
differentiated between the two host races. In this case, the loci deal with simple traits that
impact fitness on the host plants, feeding and development time, and mate choice. Thus, in
the case of insect host races (which are biological races), one does not need many loci to
create races. No one, however, would claim that racial formation in organisms with more
complex behavior is this simple.
POPULATION SUBDIVISION
Population subdivision is a means to test the amount of genetic variation amongst
subpopulations within a species. This concept was developed by American evolutionary
geneticist Sewall Wright (1978). Populations, which have undergone significant adaptation to
local conditions, differ in population dynamic history, and limited gene flow between them
should differ in allele frequencies at a number of loci. The population subdivision statistic
(FST) compares the allelic diversity of each of the subpopulations against a pooled total
population. Since Wrights invention of F coefficients, which examine the proportioning of
genetic variation between different levels within a species, population geneticists have
utilized a minimum value of differentiation between subpopulations and the total species as
the threshold for identifying the existence of biological races (FST> 0.250). Wright chose this
value to maximize the probability that the subgroups were actually fixed for alternative
different alleles at various loci.
Four nucleotides can be found in DNA, adenosine (A), Thymine (T), Gua- nine (G), and
Cytosine (C). In coding regions of the genetic code three nucleotides in succession
determine which amino acid should be placed in the resultant protein. The code is redundant,
but a change in a position can result in a different amino acid being specified. When we
examine the coding and non-coding regions of DNA in a population, most people will have
72
the same nucleotide at the vast majority of the positions within the code. However, at some
positions, a variant will be found in some individuals. Such a variant is called a SNP. One
study examined 4,833 single nucleotide polymorphisms (SNPs) in 538 clusters across the
human genome in Europeans (N = 30), African Americans (N = 30), and Asians (N = 40). In
the study the mean frequency for FST at each locus was 0.083, with only 10 percent of the loci
exceeding FST of 0.18 and about 6.5 percent exceeding FST of 0.250. This is consistent with
the general finding that, averaged across the genome, FST in humans does not approach
Wrights threshold (and is generally FST = 0.110). Utilizing eleven genes that have been
reputedly associated with general intelligence (ASPM, OXTR, CCKAR, ADRB2, DTNBP1,
ALDH5A1, IGF2R, CHRM2, MCPH1, DRD4, and CTSD; Deary, Johnson, & Houlihan,
2009) I calculated FST from the SNPs currently reported within these genes. The data on
SNPs FST values was retrieved from the Allele Frequency Database (ALFRED, maintained
by the Kidd Laboratory at Yale University). I calculated FST values for SNPs found within
genes and for all SNPs found within these eleven genes. Table 1 reports the mean and
standard deviation for FST within each gene. Of the eleven, nine have mean values well below
Wrights threshold, OXTR barely exceeds it (0.251) and ASPM is well differentiated at
0.322. The FST values for each SNP were calculated from populations worldwide. The range
of populations sampled varied between 4 and 87. However the vast majority of the SNPs
frequencies in these genes were sampled from around 50 populations varying from regions
identified as Africa (sub-Saharan Africa), Europe, Asia (Middle East and Eurasia), East Asia,
Oceania, and the Americas. Generally, there were more populations sampled in Europe and
East Asia, compared to Africa. This discrepancy in sampling makes all world-wide
calculations of genetic variability suspect, simply because the data we have at present is not
representative of the entire spectrum of human populations. The mean FST for all SNPs from
these seven genes is 0.150, with a S. D. of 0.075. Only 12.5% exceeded Wrights threshold of
0.250, see Figure 1. This is to be expected in this sample, since nine of the eleven genes
examined had mean FST values for all SNPs within them below Wrights threshold. Despite
the limitations of sampling across world populations, the analysis presented above does not
support the notion that there should be racially differentiated genetic variation for genetic
variants associated for intelligence. There are a number of caveats to this conclusion. First,
this analysis did not determine the distribution of SNP changes, which were deleterious (acted
on by purifying selection), neutral (acted on by genetic drift), and beneficial (acted upon by
positive selection).
Neutral substitutions should not impact the function of the gene product, and therefore
frequency variation should be solely determined by genetic drift. Such variation should also
be completely irrelevant to any claim of the variants effect on cognitive function. Given that
the majority of SNP changes are neutral, this means that even the modest FST results
discussed above are an over estimate of how different populations are in the genetic variation
that actually impacts cognitive function. For example, the oxytocin receptor gene (OXTR) has
at least 14 SNPs, which have been evaluated for some impact on cognitive or emotional
function (see Table 2, drawn from ALFRED, 2012). The gene is found at chromosome 3p25
and spans about 19,207 bases (19 Kb). It has four exons and three introns. Some of these
variants fall within introns and others exons. Some are tag SNPs. Tag SNPs are
representative SNPs which fall in genomic regions of high linkage disequilibrium.
73

Table 1. Mean SNP FST within eleven genes reputedly associated
with general intelligence
Gene Name
Abnormal spindle homolog (ASPM)
Oxytocin Receptor (OXTR)
Cholecystokinin A Receptor (CCKAR)
2-Adrenergic Receptor
Dysbindin-1
Aldehyde dehydrogenase 5 family, member A1
(ALDH5A1)
Insulin-like Growth Factor 2 (IGF2R)
Cholinergic muscarinic 2 receptor (CHRM2)
Microcephalin 1 (MCPH1)
Dopamine D4 receptor (DRD4)
Cathepsin D (CTSD)
Location
1q31
3p25
4p15
5q33
6p22
6p22
# SNPs
4
14
7
12
37
14
Mean FST
0.322
0.251
0.117
0.116
0.114
0.148
S. D.
0.097
0.097
0.063
0.060
0.037
0.068
6p26
7q33
8p23
11p15
11p15
48
34
169
2
2
0.132
0.180
0.157
0.133
0.070
0.053
0.071
0.081
0.041
0.008
These SNPs may not be causative themselves, but instead linked to genetic variants
within the region which may play a causative role. The SNP rs7632287 is mapped at
chromosome position 8,791,446. According to the latest build of the human genome, OXTR
receptor resides between chromosome position 8,792,094 and 8,811,300. Thus in actuality
this SNP falls outside of OXTR. Yet at least one study (Wu, Li & Su, 2012) evaluated this
SNP for impact on emotional and cognitive empathy. In this study, no relationship between
rs7632287 was found for cognitive or emotional empathy. However, rs237887 and rs4686302
were associated with emotional and rs2268491 was associated with cognitive empathy.
Table 2. Common SNPs found within the Oxytocin Receptor (OXTR) locus
SNP ref. number
rs7632287
rs1042778
rs11706648
rs237887
rs2268490
rs237888
rs2268491
rs2268492
rs237889
rs11131149
rs237897
rs4686302
rs2301261
rs6777726
Chr. Pos.
8,791,446
8,794,545
8,796,547
8,797,042
8,797,085
8,797,095
8,800,398
8,800,672
8,802,483
8,802,851
8,808,285
8,809,222
8,810,896
8,813,494
FST
0.304
0.240
0.117
0.087
0.136
0.266
0.077
0.117
0.107
0.147
0.170
0.097
0.079
0.079
Poly.
A/G
G/T
A/C
A/G
C/T
C/T
C/T
C/T
C/T
A/G
A/G
C/T
A/G
A/G
Africa
0.750 0.380
0.880 -- 0.480
0.880 0.662
0.850 0.500
0.850 0.560
0.880 0.560
0.800 0.600
0.960 0.830
1.000 0.730
0.730 0.430
0.910 0.720
1.000 0.950
0.880 0.750
0.880 0.770
East Asia
0.140--0.000
0.400 0.050
0.850 0.600
0.850 0.330
0.900 0.440
1.000 0.940
0.900 0.500
0.900 0.600
0.900 0.350
0.300 0.150
0.500 0.060
0.900 0.640
1.000 0.720
1.000 0.720
In the case of the SNPs associated with emotional empathy, rs237887, showed that the A
allele was better. FST at this locus is very low (0.087) and A is the major allele. If one were
74
making a case for the relationship between this locus, emotional empathy, and evolutionary
fitness, the results follow the expected pattern. If emotional empathy increases fitness we
would expect genetic variants associated with it to sweep to fixation, thus low FST (all
populations should be similar in A frequencies). This is not seen however with rs4686302. In
this case, T is better for the phenotype, but C is the major allele (FST again low at 0.097).
Finally, in the case of cognitive empathy (more the concern of this discussion), we again see
FST low (0.077) however heterozygotes CT > TT > CC and C is the major allele. This pattern
is hard to explain if cognitive empathy had any relationship to fitness. These sorts of
complications underscore the difficulty of assigning simple genetic explanations to complex
phenotypes, and the relationship of the latter to evolutionary fitness.
Figure 1 illustrates that some SNPs associated with variation in intelligence have high
FST. (from these genes, 12%). With regard to the entire genome, greater than 95% of genetic
markers between major socially-defined ethnic/racial groups have delta (values less than
0.500 (Dean et al., 1994). Delta () is defined as px py where these represent the frequency
for a specific allele at a locus in two different populations (x and y). Shriver et al., (1997)
coined the term population-specific allele (PSA) to describe genetic markers with large
allele-frequency differences between populations. Shriver and colleagues defined a PSA as
those loci with > 0.500 for a specific allele. In their 1997 analysis, Shriver and colleagues
identified six such loci with large values between Africans, African Americans, European
Americans, Hispanics, and Amerindians. These loci were FY-null, RB2300, LPL, CKMM,
PV92, and DRD2 (which has been linked in some studies to cognitive performance, (Moises
et al., 2001). The FY-null allele is an adaptation that helps to resist malaria infection.
Therefore we should expect it to be in high frequency in regions with high malaria
prevalence. It results from having a nucleotide C as opposed to T at position within the
GATA-1 binding site within the DARC gene. For 22 Sub-Saharan African populations the
mean frequency of the C allele is 0.900 with a variance of 0.05; similarly for 33 European
populations the C allele frequency is 0.02, with variance of 0.006. Yet, the frequency within
the Mandenka and Somalis the C allele is reported as 0.000 and 0.364 and both are SubSaharan African populations (ALFRED, 2012). Conversely, the while the mean for the C
allele in Europe is very low, Andalusians have a C allele frequency of 0.222. Given the
proximity of this region of Spain with North Africa, this higher C allele frequency might be
evidence of gene flow (admixture) between Africa and Europe. This is also supported by the
historical record of Moorish domination of the Andalusian region of Spain for several
centuries (Glick 2005; Watts 1894). In addition, we observe that there is not always high FST
in loci that influence visually observable traits, such as skin, eye, and hair color. Global FST
calculated from 108 populations for the rs12896399 SNP within the SLC24A4 locus that
influences brown versus blue eye color is only 0.133 (data from ALFRED 2012). Figure 2
shows the distribution of values for SNPs found in seven genes reputedly associated with
cognitive performance between sub-Saharan Africa and East Asia. These values were
calculated by utilizing the median value for populations listed in the ALFRED database. SubSaharan Africa and East Asia were compared to the long-standing psychometrician claim of
greater genetically based intelligence of East Asians over Europeans, and of both over
Africans (Rushton, 1999). The same caveats apply with regards to the samples available
which produced this Figure.
75

FST Distribution for SNP's within Eleven Genes
Reputedly Associated With Intelligence
120
Mean = 0.150
S.D. = 0.075
N = 352
100
Frequency
80
60
40
20
0
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
FST
FST values were calculated from data available in ALFRED 2012.

Figure 1. Distribution of FST in Eleven Genes Reputedly Associated with Intelligence.
Distribution for SNP's within Seven Genes
Reputedly Associated With Intelligence
(Sub-Saharan Africa versus East Asia)
100
90
Mean = 0.137
S.D. = 0.216
N = 138
80
Frequency
70
60
50
40
30
20
10
0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Delta ( )
Delta () values were calculated from median values of populations listed in ALFRED database, 2012.
Figure 2. Delta Distribution for SNPs with Seven Genes Reputedly Associated with Intelligence (SubSaharan Africa versus East Asia).
76
Once again we see that the mean values for these SNPs (0.137 +/- 0.216) fall well
below the 0.500 threshold required to be considered population-specific by Shriver et al.
(1997). Only 10 of the 134 in this sample exceeded that threshold. To illustrate the character
of SNP variation in a specific gene, values for DTNBP1 was examined. Burdick et al. 2006
reported that a risk haplotype for cognitive deficit (CTCTAC) existed at SNPs rs909706;
rs1018381; rs2619522; rs760761; rs2619528; and rs1011313 in the BTNP1 gene. ALFRED
has data for four of these six SNPs and values between Sub-Saharan Africa and East Asia
were 0.015; 0.029; 0.056; and 0.163 for rs909706; rs2619522; rs760761; and rs1011313
respectively. Again, these results do not make a strong case for this locus contributing to a
genetic intelligence differential between these groups.
Woodley disputes the notion that population subdivision is a sufficient reason to
invalidate the existence of race in anatomically modern humans. This is first attempted by
reference to Lewontins fallacy. Lewontins fallacy was first coined by the statistician A.
W. F. Edwards in an essay that appeared in the journal Bioessays in 2003. This piece claimed
that population geneticist Richard Lewontin was in error when he argued that the amount of
genetic variation within populations was far greater than that between populations, which
therefore invalidated the ability to assign individuals to racial groups. Edwards argued that if
enough genetic loci are used, individuals can be distinguished into clusters, and that such
clusters could represent biological races. Graves 2005c points out the many problems with
this analysis. The largest problem is how population genetic data has been collected thus far.
Generally, samples are taken from groups who are descended from populations whose
geographicrange is very far apart. Thus, we expect clustering of allele frequencies if one
compares sub-Saharan Africans to Northern Europeans to East Asians. However, if the full
continuum of human populations were examined, when a specific cluster began and when
another ended would not be so clear. For example, in 2009 54,794 SNPs were examined in
1,928 individuals from 73 Asian populations. These data were compared with data from subSaharan Africans (Yorubans) and European Americans. Running Structure with K = 14
showed that linguistic groups tended to cluster together; however, there were populations that
fell in clusters that did not belong to their linguistic or geographic affinity.
CLADISTIC RACES
The other criterion by which biological races might be identified is their unique genetic
lineages. Andreasen (1998, 2004) argues that is precisely how human races can and should be
identified (by cladistic races). Thus if the human species really has unique genetic lineages,
then it should be possible to represent various human races on an evolutionary tree.
Templeton pointed out that if this were so, then the genetic distances between all non-African
and African populations should be the same. If on the other hand, genetic distances between
populations reflect the amount of gene flow, then there should be a strong relationship
between geographic and genetic distance, which is referred to as isolation by distance
(IBD). But we do not find that the genetic distances between all non-African and African
populations are the same, and thus the cladistic race argument fails immediately. On the other
hand, we do find that genetic distance (measured by population subdivision, FST) and pairwise geographic distance between populations shows a highly statistically significant
77
correlation (R2 = 0.7679). Despite the overwhelming use of cluster algorithms (Structure) to
examine human diversity (which is consistent with the notion of independent evolutionary
lineages), the data strongly supports the IBD explanation.
GENOME-WIDE APPROACHES
The preceding section looked at genetic variation within a number of candidate genes
reputedly associated with intelligence and their distribution. The data available did not
support a genetic explanation for IQ score, cognitive performance differentials between
supposed racial groups. It can be argued that this analysis was superficial, and particularly
that it suffered from the non-representative data sampled on these candidate genes from the
groups in question. This could be true, but I will argue that this is not the real problem. The
deeper issue results from the genetic character of complex traits themselves and thus the
inadequacy of any candidate gene approach to tell us anything useful about the genetic
controls of such traits (vid. Rose, Mueller, & Burke, 2011). A further complication results
from the inadequacy of the psychometrician conception of race, which is deeply flawed and
essentialist (Graves, 2011). The use of such a conception will by necessity result in hit-andmiss predictions; missing more than hitting.
At least 1/3 of the 20,000 genes of the human exome are actively expressed in the
human brain (Institute of Neurological and Brain Disorders, 2012). Given that brain function
undoubtedly has something to do with cognitive performance, the notion that we can identify
a small number of candidate genes which will capture the majority of the variation in the
normal range of human intelligence is absurd. Clearly there are a number of mutations which
occur in specific genes that reduce intelligence. These mutations have high penetrance and
have pleiotropic impacts on a variety of physiological systems. Such pathological mutations
have been well-studied and we know a great deal about how natural selection acts to reduce
the frequency of such alleles. In this regard, the brain is no different from any other organ.
However, comparing high penetrance mutations that are pathologic in character to genetic
variants which account for normal variation in cognitive function is like comparing apples
and watermelons. The analysis of such mutations requires entirely different intellectual
programs. The latter program is part and parcel of the general analysis of complex
phenotypes. In the first decade of the 21st century the genetic analysis of complex traits has
moved to genome wide association studies (GWAS). Association is a statistical relationship
about the co-occurrence of alleles or phenotypes (Strachan & Read 2011). Association can
have many causes, such as direct causation (allele A causes phenotype B); epistatic effect
(people who display phenotype B may be more likely to survive and reproductive if they have
allele A); population stratification (the population contains several subpopulations, and
phenotype B and allele A happen to be in high frequency in some of the subpopulations);
Type 1 error (association studies test large numbers of markers, such that 5% will be
significant at the p = 0.05 and 1% will be significant at the p = 0.01 level without any true
effect. For this reason p < 1 x 10-8 are often used for significance); and linkage disequilibrium
(the allele A marks a chromosome segment that contains a sequence variant causing
phenotype B). Probably the best example of this approach is provided by GWAS studies of
height. This is a trait whose high heritability has been well-established, yet studies conducted
78
with hundreds of thousands of individuals have only explained a small fraction of the
heritable variation in the trait. This results from small effect sizes, low population
frequencies, population size, marker density, and the rate at which LD diminishes with map
distance (Allen et al., 2010; Hill, 2012; Park et al., 2010; Yang et. al., 2010). To date, the best
GWAS study that exists is that associated with human height (Hill, 2012; Lango et al., 2010;
Yang, 2012). This study utilized a metapopulation analysis of data from studies comprising
over 180,000 individuals and 180 quantitative trait loci (QTL) have been identified with the
high statistical stringency of p < 5 x 10-8. None of the SNPs individually accounts for more
than 0.11% of the variance height and estimated homozygote differences are typically under 1
cm, compared to the phenotypic variance of 7 cm. The heritability of height estimated from
relatives is about 80% (h2 = 0.80). Thus together the identified SNPs account for only about
10% of the variance in h2. Even if all unidentified common variants of similar effect size were
found, they would only add an additional 16% to the known cause of variation in height
(Lango et al., 2010; Park, 2010).
The situation with regard to understanding the genetic variance associated with cognitive
performance is nowhere near this advanced. This, of course, is a major obstacle for the
proponents of genetic differentials in racial intelligence, given we know so little concerning
the genetic determinants of intelligence. Butcher et al. (2008) utilized a genome-wide
quantitative trait locus association scan to detect SNPs associated with variation in general
intelligence (g). They examined DNA and intelligence scores from 4039 twin pairs derived
from families participating in the Twins and Early Development Study (in Wales and
England, and only children from families who self-identified as white). The children were
tested and placed in low versus high general intelligence (g) groups and SNPs identified
within each group. The study examined over 500,000 SNPs. The mean value was small
between the groups ( = 0.025). However, 47 SNPs were chosen as the highest quality from
a pool of SNPs whose values were > 0.15 0.30. Six SNPs were eventually chosen as
most informative, rs11691504, 2q31.3, intergenic; rs1378810, 3q22.1, intron 55 (DNAJC13);
rs2496143, intron 5 (TBC1D7); rs11761076, 7q32.1, intergenic; rs174455, 11q12.3, intron 1
(FADS3); and rs7195954, 16p13.3, intergenic. It is notable that none of these SNPs occurs
within a coding segment, although it is possible that these are tag-SNPs since in Europeans
one-third to one-half of chromosomes are structured in LD blocks that can extend from a few
to 300 KBs (Hill, 2012; Mueller et al. 2005). Davis et al. (2010) also conducted a genomewide analysis for g-related SNPs in this same population. They found nine SNPs which
were suggestive. These nine accounted for in total, 1.22% of the variation in g. None of the
SNPs identified in Butcher et al. (2008) were replicated in this latter study. This is a real
problem, given that the studies were conducted in the same populations and assayed the same
phenotype (see Table 3). It also possible that these studies results are completely specious due
to the relatively small number of individuals used and the small effect of the variants in
question. There is little data on these SNPs in ALFRED, but that which exists suggests that
they could not contribute to any racial differential in intelligence (see FST and values in
Table 3). More recently, Gail Davies and colleagues (2011), in a genome-wide scan of
549,692 single-nucleotide polymorphisms (SNPs) from 3,511 unrelated adults, concluded that
40 percent of the crystallized-intelligence variation and 51 percent of the fluid-intelligence
variation was accounted for by linkage disequilibrium between known SNP markers and
unknown causal variants.
79

Table 3. SNPs associated with general intelligence (g) from TED studies
Study
Butcher et al.
2008
SNP
rs11691504
rs1378810
rs2496143
rs11761076
rs174455
rs7195954
Davis et al. 2010 rs4917360
rs451041
rs10997145
rs17806457
rs4131535
rs9294506
rs10223126
rs12522702
rs17792746
Position
2q31.3
3q22.1
6p24.1
7q32.1
11q12.3
16p13.3
9, 136277066
11, 3017301
10, 67891313
9, 111931456
12, 93621919
6, 92302221
5, 171512206
5, 92517721
20, 6307041
Gene
intergenic
intron 55 (DNAJC13)
intron 5 (TBC1D7)
intergenic
intron 1 (FADS3)
intergenic
RXRA, WDR5
CARS
CTNNA3
PALM2-AKAP2
TMCC3
MAP3K7
STK10
NR2F1, AARDC3
FERMT1, BMP2
FST/
?
?
?
?
?
?
= 0.02
0.095
?
?
0.041
?
?
?
?
Using SNP data alone, similar to the studies above, they could only account for 1% of the
variation in IQ scores. These studies indicate that individual difference in human intelligence
is determined by many genes of small effect. Thus the evidence summarized above suggests
that we can state that cognitive performance is definitely influenced by genetic variation, that
cognitive performance is heritable (and therefore varies by families), we cannot reasonably
state that differentials measured in socially-defined racial groups results from differences in
underlying genetic potential for cognitive performance.
Finally, new information is coming to light about the significance of rare copy number
variations and complex phenotypes (Jarick et al., 2012; Swaminathan et al., 2012). Yeo et al.
(2011) argues for the importance of rare CNVs in predicting individual intelligence.
Examining CNVs in 74 individuals they found a significant negative correlation on score on
the Wechsler Abbreviated Scale of Intelligence (WASI). If this is so, then this again bodes
poorly for the proponents of racial variation in intelligence, particularly those who adhere to
the superiority of Europeans over Sub-Saharan Africans. Recent whole genome analyses
suggest that Europeans and other non-Africans have far more of these sorts of deleterious
CNVs than Sub-Saharan Africans (Fu et al., 2012; Tenessen et al., 2012).
CONCLUSION: RACE AND INTELLIGENCE

IN HISTORICAL PERSPECTIVE
The Bell Curve
Intelligence and Class Structure in American Life authored by psychometrician Richard
J. Herrnstein (who died in 1994 shortly before its publication) and conservative sociologist
Charles Murray, was published in September 1994. The book followed in a long line of
80
psychometric and eugenic thinking concerning the relationship between intelligence and the
class structure of society (e. g., Francis Galton, Charles Brigham, Charles Daven- port, Cyril
Burt, Aubrey Shuey, Arthur Jensen, and others). In short, The Bell Curve argued that the
primary determinant of an individuals intelligence was genetic (nature), as opposed to
environmental (nurture). In addition, Herrnstein and Murray argued that since the major
human races differed in their genetic composition, they must also differ in their intelligence.
The result of this difference structured social life; hence the social supremacy of Europeanderived Americans was mainly due to a genetic composition which predisposed them to
superior cognitive performance. Sir Francis Galton (18221911) was the modern originator of
this line of reasoning, which stretches back in the Western world to Aristotle (Gould, 1981;
Graves, 2005a).
Psychologist Arthur Jensen paved the way for Herrnsteins psychometric analysis with
his landmark 1969 article in Harvard Educational Review, titled How Much Can We Boost
IQ and Scholastic Achievement. Jensen claimed that the differential Negro birthrate as a
function of socioeconomic status was higher than that of the white population. He argued that
the danger here was that because Negroes had lower genetically determined IQs than whites,
there must be an overall lowering of intelligence in the American population. He blamed
unguided welfare policies for increasing the rate of birth among Negroes: Is there a danger
that current welfare policies, unaided by eugenic foresight, could lead to the genetic
enslavement of a substantial segment of our population (Jensen 1969, p. 95).
Jensen presented evidence that Negro intelligence was one standard deviation (fifteen IQ
points) below that of whites and also argued that early intervention programs, such as Head
Start, were a waste of time because they were unlikely to alter genetically determined limits
of intelligence. Just as later with The Bell Curve, the Jensen article was siphoned up by the
popular media. It was discussed in Newsweek under the title Born Dumb? and in U. S. News
and World Report in an article titledCan Negroes Learn the Way Whites Do? Predictably,
as reports about Jensens research flooded the media, little attention was paid to a story by
Olive Walker that showed that the highest IQ scores in the city of Los Angeles were being
posted by students at an African American elementary school (Chase, 1977; Walker, 1970).
Two years after Jensens article appeared, Herrnstein would publish a piece for popular
consumption in The Atlantic Monthly titled simplyIQ. Following Jensens lead, Herrnstein
argued that America was fundamentally a meritocracy, and that each individuals position
within it was best explained by innate characteristics, such as IQ.
The publication of The Bell Curve in 1994 did not represent any new theoretical
conceptions concerning race, social status, and IQ (Graves & Johnson, 1995). The book
included some sociological analysis by Murray relating AFQT (Armed Forces Qualifying
Test) scores to a range of primarily dysfunctional social characteristics. Although the army
did not use AFQT scores as IQ scores, Herrnstein and Murray did (Darity & Mason, 1998).
Finally, there was no serious peer review of The Bell Curve before it was published by the
Free Press. Rather, it was circulated among a closed circle of psychometricians who already
agreed with its conclusions. The book had four parts. The first part concerned the emergence
of an cognitive elite in modern societies, the second part dealt with how cognitive
performance influenced social behavior, the third part evaluated how these cognitive
differentials were influencing the nation, and the final part attempted to provide proscriptions
for living in a society with innate racial differentials of intelligence.
81
Despite the hype, Herrnstein and Murray made a very weak case for the relationship of
general intelligence to important social outcomes. The main problem with their argument is
that it contained flawed statistical techniques, made weak inferences, and relied on tainted
sources (Graves & Place, 1995). T he statistical support for the multiple regression models
they used to relate social outcomes, such as poverty, unemployment, divorce rates, and outof-wedlock births, was weak. For example, 50/60 correlation coefficients in their models
were between 0.00 and 0.10. This meant that, at most, 10 percent of the variance in the
outcome of these models was explained by general intelligence scores. Explained in another
way, 90 percent of these social outcomes are explained by other factors. Furthermore,
Herrnstein and Murray drew inferences from their statistical models using methods that were
never before published in the peer-reviewed literature, so these methods hadnot been
validated by professional statisticians (Graves & Place, 1995).
Finally, many of the studies they used to make their case for the relationship between
general intelligence and social status were written by and funded by individuals and
organizations with demonstrated racial biases. For example, seventeen authors who were
frequently cited in The Bell Curve had contributed articles to The Mankind Quarterly, an
avowedly fascist publication (Lane, 1994). In addition, many of the studies cited were funded
by the Pioneer Fund, an organization founded by the fascist Wickliffe Draper (18911972)
and whose charter called for race betterment in the United States. In 1989, the Pioneer Fund
proclaimed that the United States should abandon its efforts at integration because raising
the intelligence of blacks or others was beyond our capacities.
Despite the fact that the core genetic and sociological claims of The Bell Curve were
discredited, racialist thinking in psychometry is alive and well. For example, in October of
2007, Nobelist James Watson gave an interview in which he claimed: he was "inherently
gloomy about the prospect of Africa" because "all our social policies are based on the fact
that their intelligence is the same as ours whereas all the testing says not really". He further
went on to say that there was a natural desire that all human beings should be equal but
"people who have to deal with black employees find this not true" (Milmo, 2007). Given that
at the time of this statement, James Watson was the Chancellor of Cold Spring Harbor
Laboratories and previous claims he had made about genetics and complex traits, it was
assumed that Watson meant that genetic predispositions were responsible for the intelligence
differences he was referring to in this quote. Shortly, after the statement appeared in the press,
Watson resigned as chancellor (Bhattacharjee, 2007).
Despite the overwhelming evidence that human genetic variation does not match our
socially defined categories, individuals seem to continue their desire to maintain racial
thinking in sociology, biomedicine, and psychometrics (Ceci & Williams, 2009; Shaio et al.,
2012). However, new genomic research seems to make all of these claims even more absurd.
For example, genomic claims of the superiority of Europeans and East Asians with regards to
intelligence, run afoul of the greater deleterious variation in these groups compared to SubSaharan Africans (Fu et al., 2012; Loehmueller et al. 2008; Tennessen et al. 2012).
Furthermore, all genetic hypotheses for phenotypic variance require that genotypes are
assayed in equivalent environments, not just at time of assay, but for at least two previous
generations to equalize maternal effects. This never happens in racialized and socially
stratified societies. Thus it can be argued that if one wishes to see the cognitive performance
of the socially subordinated improve, then these people should be provided with equalize
deducation, nutrition, health care, and freedom from differential exposure to toxic materials
82
like lead. We know that psychosocial stress influences brain development (McEwen &
Gianaros 2011) and that environmental factors influence a variety of complex phenotypes in
quite complex ways (Zhang et al., 2012). Therefore it matters a great deal how one addresses
the question of the existence and origin of intelligence disparity. The more complex the trait,
the more likely it is to be disrupted by poor environment. This is crucial since racialized
societies do not provide members of socially defined races equal environmental
circumstances. Thus we will never really be able to launch a credible research program on the
genetic differentials of intelligence associated with genomic variation associated with
ancestry, until such social justice is achieved. Unfortunately, this has never really been the
aim of those most invested in this research.
REFERENCES
Allele Frequency Database (ALFRED), A resource supported by the National Science
Foundation, maintained by the Kenneth Kidd Laboratory, http:// alfred.med.yale.edu/.
Allen H. L., Estrada K., Lettre G., Berndt S. I., Weedon M. N. et al., (2010). Hundreds of
variants clustered in genomic loci and biological pathways affect human height, Nature,
467, 832838.
Andreasen, R., O. (1998). A New Perspective on the Race Debate. British Journal for the
Philosophy of Science, 49, 199225.
Andreasen, R. O. (2004). The Cladistic Race Concept: A Defense, Biology and Philosophy,
19, 42542.
Barnett, J. H., Scoriels, L., & Munaf, M. R. (2008). Meta-analysis of the Cognitive Effects
of the Catechol-O-Transferase Gene Val 158/108 Met Polymorphism. Biological
Psychiatry, 64, 137144.
Bartels, M., Rietveld, M. J. H., van Baal, G. C. M., & Boomsma, D. I. (2002). Genetic and
Environmental Influenceson the Development of Intelligence. Behavioral Genetics, 32
(4), 237249.
Bhattacharjee, Y., (2007).Watson Retires From Cold Spring Harbor Lab, Scie- nce Now,
October 25, http://news.sciencemag.org/sciencenow/2007/10/ 25-02.html.
Burdick, K., Lencz, T. Funke, B., et al. (2006). Genetic variation in DTNBP1 influences
general cognitive ability, Human Molecular Genetics, 15(10), 1563-1568.
Butcher, L. M., Davis, O. S. P., Craig, I. W., & Plomin, R. (2008). Genome-wide quantitative
trait locus association scan of general cognitive ability using pooled DNA and 500K
singe nucleotide polymorphism microarrays, Genes, Brain and Behavior, 7, 435-446.
Born Dumb? (1969). Newsweek (March 31), 84.
Can Negroes Learn the Way Whites Do? Findings of a Top Authority. (1969). U. S. News
&World Report (March 10), 4851.
Ceci S., & Williams W. M. (2009). Darwin 200: Should scientists study race and IQ? YES:
The scientific truth must be pursued, Nature, 457, 788-789.
Chase, A. (1977). The Legacy of Malthus: The Social Costs of the New Scientific Racism. New
York: Knopf.
Darity, W. A., Jr., & Mason, P. L. (1998). Evidence on Discrimination in Employment:
Codes of Color, Codes of Gender. Journal of Economic Perspectives, 12(2), 6390.
83
Davies, G., Tenesa, A., Payton, A., et al. (2011). Genome-wide Association Studies Establish
that Human Intelligence Is Highly Heritable and Poly- genic. Molecular Psychiatry 16,
9961005.
Dean, M., Stephens, J. C., Winkler, C. et al. (1994). Polymorphic admixture typing in human
ethnic populations, Am. J. Human Genet., 55, 788-808.
Deary, I. J., Penke, L., & Johnson, W. (2010). The Neuroscience of Human Intelligence
Differences. Nature Reviews: Neuroscience, 11, 201211.
Deary, I. J., Strand, S., Smith, P., & Fernandes, C. (2007). Intelligence and Educational
Achievement. Intelligence, 35, 1321.
Fish, J. M., (Ed). (2002). Race and Intelligence: Separating Science from Myth. Mahwah, NJ:
Erlbaum.
Fox, S. & Spector, P. E. (2000). Relations of Emotional Intelligence, Practical Intelligence,
General Intelligence, and Trait Affectivity with Interview Outcomes: Its Not All Just g.
Journal of Organizational Behavior, 21 (2), 203220.
Fu, W., OConnor, T., Jun, G. et al. (2012). Analysis of 6,515 exomes reveals the recent
origin of most human proteincoding variants, Nature, doi: 10. 1038/nature11690.
Glick, T. F. (2005). Islamic and Christian Spain in the Early Middle Ages, Leiden, The
Netherlands: Koninklijke Brill NV.
Gottfredson, L. S. et al. (1997a). Mainstream Science on Intelligence: An Editorial with 52
Signatories, History, and Bibliography. Intelligence, 24 (1), 1323; originally published
in the Wall Street Journal, December 13, 1994.
Gottfredson, L. S. (1997b). Why g Matters: The Complexity of Everyday Life. Intelligence,
24(1), 79132.
Graves, J. L. (2002). What a Tangled Web He Weaves: Race, Reproductive Strategies, and
Rushtons Life History Theory. Anthropological Theory, 2 (2), 131154.
Graves, J. L. (2005a). The Emperors New Clothes: Biological Theories of Race at the
Millennium. New Brunswick, NJ: Rutgers University Press.
Graves, J. L. (2005b). The Race Myth: Why We Pretend Race Exists in Ame- rica. New York:
Dutton.
Graves, J. L. (2005c). What we know and what we dont know: Human Genetic Variation and
the social construction of race, for Is Race Real?, A collection of essays solicited by The
Social Science Research Council, edited by Craig Calhoun, President Social Science
Research Council, April 6th 2005.
Graves, J. L. & Johnson, A. (1995). The Pseudoscience of Psychometry and the Bell Curve.
Special Issue: Myth and Realities: African Americans and the Measurement of Human
Difference. Journal of Negro Education, 64 (3), 277294.
Graves, J. L. & Place, T. (1995). Race and IQ Revisited: Figures Never Lie, But Often Liars
Figure. Sage Race Relations Abstracts, 20(2), 450.
Herrnstein, R. J. (1971). I. Q. Atlantic Monthly (September), 4364.
Herrnstein, R. J., & Murray, C. (1994). The Bell Curve: Intelligence and Class Structure in
American Life. New York: Free Press.
Inoue, T., Kimura, T., Azuma, C., et al. (1994). Structural organization of the human oxytocin
receptor gene, The Journal of Biological Chemistry, 269 (51), 32451-32456.
Jarick I., Volckmar A. L., Ptter C. et al. (2012). Genome-wide analysis of rare copy number
variations reveals PARK2 as a candidate gene for attention-deficit/hyperactivity disorder,
Mol Psychiatry, doi: 0. 1038/ mp. 2012.161.
84
Jensen, A. R. (1969). How Much Can We Boost IQ and Scholastic Achieve- ment? Harvard
Educational Review, 38(1), 1123.
Johnson, W., McGue, M., & Iacono, W. G. (2006).Genetic and Environmental Influences on
Academic Achievement Trajectories during Adolescence. Developmental Psychology, 42
(3), 513542.
Lane, C. (1994). The Tainted Sources of The Bell Curve. New York Review of Books
(December 1), 419.
Mayr, E. (1974). Population, Species, and Evolution, Cambridge, MA: Belk- nap Press.
McCartney, K., Harris, M. J., & Bernieri, F. (1990). Growing Up and Growing Apart: A
Developmental Meta-analysis of Twin Studies. Psychological Bulletin, 107, 226237.
McEwen, B. S. & Gianaros, P. J. (2002). Stress- and allostasis-induced brain plasticity, Annu.
Rev. Med., 62, 431-445.
Milmo, C., (2007). Fury at DNA pioneer's theory: Africans are less intelligent than
Westerners, The Independent, October 17, http:// www. independent.
co.uk/news/science/fury-at-dna-pioneers-theory-africans-are-less-intelli-gent-thanwesterners-394898.html.
Moffitt, T. E., Caspi, A., Harkness, A. R., & Silva, P. A. (1993). The Natural History of
Change in Intellectual Performance: Who Changes? How Much? Is it Meaningful?
Journal of Child Psychology and Psychiatry, 34 (4), 455506.
Mueller, J. C., Lohmussaar, E., Reedik, M., et al. (2005). Linkage disequilibrium patterns
and tagSNP transferability among European popu lations, Am. J. Human Genetics 76,
387-398.
National Institute of Neurological Disorders and Stroke, Brain Basics, (2012).
http://www.ninds.nih.gov/disorders/brain_basics/genes_at_work.htm.
Park J. H., Wacholder S., Gail M. H., Peters, U., Jacobs K. B., et al. (2010). Estimation of
effect size distribution from genome-wide association studies and implications for
future discoveries, Nature Genetics, 42, 570575.
Rose, M. R, Mueller, L. D., & Burke, M. K. (2011). New Experiments for an Undivided
Genetics, Genetics, 188(1), 1-10.
Schiao, J. L. Bode, T., Beyer, A., & Selvig, D. (2012). The genomic challenge to the social
construction of race, Sociological Theory, 30, 67 88.
Shimokata, H., Ando, F., Niino, N., Miyasaka, K., & Funakoshi, A. (2005). Cholecystokinin
A receptor gene promoter polymorphism and intelligence, AEP, 15(3), 196-201.
Shriver, M. D. & Kittles, R. A, (2004). Genetic ancestry and the search for personalized
genetic histories, Nature Reviews: Genetics 5: 611-618. Stra- chen, T. and Read, A.,
(2011). Human Molecular Genetics 4th Ed., New York, NY: Garland Science.
Strenze, T. (2007). Intelligence and Socioeconomic Success: AMeta-analytic Review of
Longitudinal Research. Intelligence, 35 (5), 401426.
Sternberg, R., J. (2012). Intelligence: State of the art, Dialogues in Clinical Neuroscience,
14(1), 19-27.
Swaminathan G. J., Bragin E., Chatzimichali E. A., et al. (2012). DECIPHER: web-based,
community resource for clinical interpretation of rarevariants in developmental disorders,
Hum Mol Genet., 21(R1), R37-44.
Tennessen, J. A., Bigham, A. W., OConnor, T. D. (2012). Evolution and functional impact
of rare coding variation from deep sequencing of human exomes, Science, 337, 64-69.
85
Walker, O. (1970). The Windsor Hills School Story. Equity and Excellence in Education
(Integrated Education: Race and Schools), 8 (3), 49.
Watts, H. E., (1894). The Christian Recovery of Spain: Being the Story of Spain from the
Moorish Conquest to the Fall of Granada (711-1492 A. D). New York, NY: G. P.
Putnams Sons.
Wellcome Trust Case Control Consortium, (2007). Genome-wide association study of 14,000
cases of seven common diseases and 3,000 shared controls, Nature, 447, 661-678. PMID:
17554300 DOI: 10. 1038/nature 05911.
Wilson, R. S. (1978). Synchronies in Mental Development: An Epigenetic Perspective.
Science, 202, 939948.
Woodley, M. A. (2010). Is Homo sapiens Polytypic? Human Taxonomic Diversity and Its
Implications, Medical Hypotheses, 74, 195201.
Wright, S. (1978). Evolution and the Genetics of Populations, Vol. 4: Variabi- lity within and
among Natural Populations, Chicago, IL: University of Chicago Press.
Wu, N., Li, Z., & Su Y. (2012). The association between oxytocin receptor gene
polymorphism (OXTR) and trait empathy, J. Affective Disorders, 138, 468-472.
Yang J., Benyamin B., McEvoy B. P., Gordon S., Henders A. K., et al. (2010). Common
SNPs explain a large proportion of the heritability for human height, Nature Genetics, 42,
565569.
Zhang, D., Li, S., Tan, Q., & Pang, Z. (2012). Twin-based DNA methylation analysis takes
center stage of studies of human complex diseases, J. Gene- tics and Genomics, 39, 581586.

ISBN: 978-1-62618-728-3
Chapter 6
MULTI-GROUP LATENT GROWTH CURVE

MODELING: THE INFLUENCE OF SCHOOL
READINESS PROFILES ON STUDENT
ACHIEVEMENT GROWTH
University of Virginia, Charlottesville, VA, US
ABSTRACT
School readiness is the product of many interacting influences on children and is an
important element of early childhood education. The multidimensional nature of school
readiness was operationalized in terms of six indicators that included social skills,
interactions with parents, problem behaviors, and performance on tests of cognition and
attention to identify six homogeneous profiles of readiness in a sample of 964 typically
developing children. The influence of these profile groupings on students achievement
growth in reading and mathematics achievement was evaluated through multi-group
latent growth curve analysis. Results indicated that student growth on these two
achievements from 54 month of age through grade five was influenced by the patterns of
strengths and weaknesses reflected in their readiness profile; and support the notion that
school readiness is multi-dimensional, that strengths in some areas may compensate for
weaknesses in others, and that there may be more than one route to successful
educational outcomes.
Definitions and descriptions of school readiness have been informed by a variety of stake
holders. While research into the proximal and distal effects of preschool intervention
programs for disadvantaged children has a long history (Ramey, et al., 1974, Weikart, Bond
& McNeil, 1978), a renewed emphasis on school readiness was fueled in 1991 with the
presidents release of the National Education Goals report that was endorsed by governors in
all 50 states. The first of these goals stated that all children should enter school ready to learn
(National Education Goals Panel, 1991). Shortly thereafter, extensive efforts were underway
to articulate and unite differing perspectives from researchers, policy makers, teachers and
88
parents, regarding what it means to be ready for school (Kagan & Kauerz, 2007; RimmKaufman, Pianta & Cox, 2000; Saluja, Scott-Little, & Clifford, 2000; Zigler, Gordic, &
Styfco, 2007). From a policy perspective, the National Education Goals Panel published
detailed descriptions for five domains of child development that, together, comprised
readiness (Kagan, Moore, & Bredekamp, 1995). These domains consisted of: (1) cognition
and general knowledge; (2) social and emotional development; (3) approaches toward
learning; (4) language development; and (5) physical well-being and motor development,
with emphasis on the interrelatedness of these domains.
The multidimensional conception of school readiness is supported by research (see
Rimm-Kaufman & Pianta, 2000 for a summary) in both typically developing and at-risk
populations (Luthar, Cicchettiiggetti, & Becker, 2000). Conse- quently, there exists a variety
of pursuant paths to assessments and descript -tions of what school readiness entails. We
focus broadly on the areas of cognition and social-emotional and behavioral development that
have been found to have material associations with early academic achievement (Curby,
Rimm-Kaurman, & Ponitiz, 2009; Konold & Pianta, 2005; LaParo & Pianta, 2001). We
examine how childrens patterns in these areas of school readiness are related to childrens
growth trajectories in reading skills across elementary school grades.
COGNITIVE SKILLS
Results from a meta-analysis of 70 published studies revealed an average correlation
between preschool cognitive performance and elementary school grades of .43 (Lla Paro &
Pianta, 2001). Several contemporary theories of cognitive ability have been proposed in the
past several decades that hold promise for informing instruction. The majority of these
theories are strongly rooted in the empirically researched paradigm of information processing
that is concerned with the cognitive processes involved in the successful execution of various
tasks. The theories differ from one another largely in terms of the number of processes
believed to be involved, how the processes are related to one another, and the level of detail
required for a proper assessment of childrens strengths and weaknesses that are useful for
informing instructional interventions. Examples of operational models of cognitive
functioning that derive roots within the information processing paradigm include the
Planning, Attention, Simultaneous, and Successive (PASS) theory (Naglieri & Das, 1990);
the Gf-Gc theory (Horn & Cattell, 1966); Carrolls (1993) three-stratum theory; and the
Cattell-Horn-Carroll (CHC) theory of cognitive abilities.
Among the multitude of cognitive processes articulated within these theoretical frameworks, memory and attention are often included in descriptions of executive functioning (e.g.,
Blair, 2002), along with inhibitory control skills, similar to those assessed on the continuous
performance tasks of sustained attention and response inhibition (Dowsett & Livesey, 2000).
Executive functioning has been shown to have concurrent, predictive and parallel growth
links to academic achievement in emergent literacy (McClelland et al., 2007). Some attribute
particular importance to attentional self-regulation in readiness, as significant not only for the
cognitive processes relevant to learning but also for social competence necessary in
classrooms (Thompson & Raikes, 2007). Similarly, Blair (2002) has argued that executive
Multi-group Latent Growth Curve Modeling
89
functioning provides support for self-regulation which is viewed as a core element of school
readiness.
Research also supports the memory (e.g., working memory) and attentional (e.g.,
response inhibition during attention tasks) components of executive functioning as being
associated with achievement in reading and math, and adjustment in school (Blair, 2002;
Gathercole & Pickering, 2000; Kurdek & Sinclair, 2000; Torgesen et al., 1999). In a
comprehensive project involving analysis and meta-analysis of six large scale datasets,
Duncan et al. (2007) consistently identified attention as one of the top three predictors for
later academic achievement.
SOCIAL AND BEHAVIORAL SKILLS

While cognitive skills are clearly important for school readiness, many states now
encourage or require observations of school readiness that extend beyond measures of
cognitive development (Children Now, 2009). Primary among these are skills related to
social, emotional and behavioral development that are often key concerns for caregivers and
teachers. At the same time, parents and teachers often differ with respect to the emphasis they
place on the importance of cognitive and social/emotional readiness. Although kindergarten
teachers and parents consistently agree that social and emotional development is a key
component of readiness (Heaviside & Farris, 1993; Rimm-Kaufman & Pianta, 2000), parents
tend to find it to be less important than teachers. For example, in a large survey more than
50% of kindergarten teachers considered empathy, turn-taking, sharing, and not being
disruptive as important elements of readiness, while fewer than 10% considered knowledge of
letters of the alphabet crucial to readiness (Boyer, 1991). At the same time, 58% of parents
thought knowing letters and numbers was important. Unfortunately, over 20% of kindergarten
teachers reported that their incoming classes had social skills deficiencies and 36% reported
that children lacked academic skills (Rimm-Kaufman & Pianta, 2000).
Although social and emotional skills are interdependent, emotional development focuses
on self-concept, self-efficacy, and the ability to express emotions and empathy; whereas
positive social development can be characterized by behavioral skills that provide for
successful interactions with others (Gresham, 1987). Because learning can be considered a
social process (Zins, Bloodworth, Weissberg & Walberg, 2004), childrens academic progress
can become impaired by skill limitations in these areas. For example, children lacking social
skills and emotional regulation have reduced opportunities to learn academic content
(Bracken & Fischel, 2007; Denham, 2006; Ladd, Birch & Buhs, 1999). Similarly, preschool
externalizing behavior problems are consistently associated with adjustment and learning
problems in elementary classrooms (Broidy et al., 2003; Rimm-Kaufman, Pianta, & Cox,
2000; Shonkoff & Phillips, 2000).
By contrast, children with more positive skill development in these areas are better
positioned to manage important tasks (e.g., following directions) that are linked to both entry
into school (Ladd, Herald & Kochel, 2006) and long term success (Hamre & Pianta, 2001;
Ladd & Burgess, 1999; Ladd, Herald & Kochel, 2006) that translate into academic
achievement gains that have been documented both concurrently and longitudinally (e.g.,
Bursuck & Asher, 1986; Downer & Pianta, 2006; Kupersmidt, Coie & Dodge, 1990; Malecki
90
& Elliott, 2002; Margalit, 1998; Parker & Asher, 1993; Vaughn, Hogan, Kouzekanai &
Shapiro, 1990; Wentzel, 1993). For example, Lla Paro and Piantas (2001) meta-analysis of
70 published studies revealed an average correlation of .32 between preschool socialbehavioral assessments and elementary school grades, and Teo et al. (1996) demonstrated that
positive social behavior was associated with standardized test scores after controlling for a
variety of factors (e.g., cognitive ability, ethnicity, teacher preferences). Children who
demonstrate positive affect, who are cooperative and play well with peers, and who have less
trouble managing and expressing emotions are better positioned for academic success. These
children receive more encouragement and instruction from teachers and their opportunities to
learn from peers are also increased. A notable exception to the well- documented links
between the importance of social skills on student achievement in upper elementary and
middle school grades can be found in Duncan et al. (2007).
PRESENT STUDY
Early education classrooms vary widely in terms of the experiences and skills that
children arrive with upon school entry (i.e., readiness). In addition, readiness itself is a
multidimensional construct that captures a variety of different functions in which individual
students can demonstrate varying degrees of strengths and weaknesses. Several studies have
modeled student level school readiness heterogeneity through person centered data analytic
approaches (Colvig-Amir, Liu, & Amilio et al., 2005; Hair, Halle, Terry-Humen, Lavelle, &
Calkins et al., 2006; Konold & Pianta, 2005), with the resulting profiles being found to be
associated with educational outcomes. We continue the work of Konold & Pianta (2005) in
which profiles of school readiness were derived for children at 54 months of age in the areas
of social skills, interactions with parents, problem behaviors, and cognition. Six normative
profiles were identified that were found to be predictive of first grade reading achievement.
Childrens readiness patterns were derived through cluster analysis across three measures
of executive functioning and three social-behavioral measures (see Konold & Pianta, 2005).
All measures were administered to children at 54 months of age. Memory and attention
served as proxies for executive functioning (Blair, 2002), and were measured through
Memory for Sentences and Incomplete Words subtests of the Woodcock Johnson tests of
Cognitive Ability (Woodcock & Johnson, 1989). In addition, Omission Errors from the
Continuous Performance Task (CPT; Mirsky et al., 1991) was used as an indicator of
impulsivity; both working memory and impulsivity are core aspects of neural theories of executive functioning (Barkley, 1997).
Memory for Sentences measures short-term memory and some aspects of verbal
comprehension. Internal consistency for this task is reported to be .92. The Incomplete Words
test focuses on auditory processing and its internal consistency is reported to be .79. Studies
of concurrent validity for the Woodcock Johnson are presented in the technical manual
(McGrew, Werder & Woodcock, 1991), and others have reported favorable psychometric
characteristics of these measures in relation to concurrent and predictive validity (Evans,
Floyd, McGrew, & Leforgee, 1993; Floyd, Shaver & McGrew, 2003; Reilly, 1985;
Vanderwood, McGrew, Flanagan & Keith, 2002). Omission Errors on the CPT was the most
widely used measure for sustained attention at the time of data collection. Test-retest
91
reliability for measures of inattention and impulsivity derived from the CPT ranged from .65.74 (Halperin, Sharma, Greenblatt & Schwartz, 1991). This method for measuring attention
has excellent face validity and studies have demonstrated acceptable level of validity for use
as a diagnostic tool of attention and impulsivity in clinical settings (Beale, Matthew, Oliver &
Corballis, 1987; Klee & Garfinkel, 1983; Swanson, 1981).
The social-behavioral measures included Social Skills, Externalizing Behavior, and
Positive Engagement. Social Skills was measured with the Social Skill Rating System (SSQ;
Gresham & Elliott, 1990), Externalizing Behavior was measured with the Child Behavior
Checklist (CBCL; Achenbach, 1991), and a mother-child semi-structured observation of
Positive Engagement was used to assess social behavior. The SSQ requires informants to
reflect on the frequency of behaviors in the areas of cooperation, assertion, responsibility, and
self-control.
The total score is derived as the sum of all 38 items, with higher scores reflecting higher
levels of perceived social competence. The instruments technical manual reports evidence to
support the content, criterion, and construct validity of the SSRS (Gresham & Elliott, 1990).
In addition, independent reviews rate the SSRS psychometric properties favorably (Demaray
et al., 1995) with empirical support to substantiate the SSRSs ability to discriminate among
groups of students (e.g., Stinnett, Oehler-Stinnett & Stout, 1989), and reveal moderate to high
correlations with developmental change and school adjustment (Walker & McConnell, 1988).
Externalizing Problems on the CBCL were obtained through mother reports. Higher
scores on this subscale indicate parent perceptions of more severe behavior problems. The
CBCL manual (Achenbach, 1991) reports internal consistency for this measure to be .93. In
addition, favorable evidence of convergent, discriminant, and predictive validity of this
measure is reported. For example, the measure has been shown to be effective in
discriminating between clinical and non-clinical levels of externalizing behaviors (Achenbach
& Rescorla, 2000), maintain stable measurement properties overtime (Kerr, Lunkenheimer &
Olson, 2007; Konold, Walthall, & Pianta, 2004), and for generating diagnostic hypotheses
that are consistent with the Diagnostic and Statistical Manual of Mental Disorders (DSM-III;
American Psychiatric Association) (Lowe, 1998).
Child Positive Engagement was measured using a semi-structured laboratory observation
of mother-child engagement. Pairs were given tasks to complete, some of which were too
difficult for children to complete without support, and 15 minute videotapes were made of the
ensuing interactions. Teams of five or six trained coders then rated the tapes for quality of
mother-child interactions on scales ranging from 1=very low to 7=very high. Each data
collector passed certification procedures based on a common certifiers review of videotapes
of the data collector administering the procedures. The certification procedures were designed
to ensure that standard data collection procedures were used across sites. Reliability estimates
exceeded .80.
Descriptions of the cluster- derived (see Konold & Pianta, 2005) patterns of readiness
across the aforementioned measures of executive functioning and social-behavioral skills are
provided below, along with prevalence rates in this sample. Children within groups represent
more homogeneous patterns of functioning.
These profile types served as a grouping variable in the multi-group latent growth curve
models for determining whether different patterns of readiness had differential associations
with childrens reading growth.
92
Group 1: Attention Problems (Prevalence = 10%). Children with this profile

demonstrated a relatively flat pattern of average-ranged scores with the exception of
elevated scores on Omission Errors (the highest among the 6 groups) reflecting likely
problems with attention.
Group 2: Low Cognitive Ability (Prevalence = 7%). The fewest number of children
were identified with this pattern that was unique for lower levels of cognitive
functioning in relation to other groups, with slightly elevated Omission Error scores.
By contrast, childrens socio-emotional skills were generally within the average
range of functioning.
Group 3: Low/Average Social and Cognitive Skills (Prevalence = 20%). Children in
this group generally were within the average range of functioning, with somewhat
depressed social skills and low average levels of cognitive ability.
Group 4: Social and Externalizing Problems (Prevalence = 17%). This profile was
characterized by lower mid-range social skills, with the lowest scores for Positive
Engagement and the highest scores for Externalizing Behaviors. Executive functioning scores were generally in the average range with a slight elevation on Omission
Errors, reflecting the possibility of a relative weakness in attention.
Group 5: High Social Competence (Prevalence = 24%). This group demonstrated the
highest levels of social skills on average and the fewest externalizing behavior
problems. Measures of executive functioning were in the average range.
Group 6: High Cognitive Ability/Mild Externalizing (Prevalence = 22%). Children in
this group demonstrated somewhat higher levels of externalizing behavior problems
with above average (the highest levels of the six groups) performance on the
measures of executive functioning.
The purpose of the current chapter is to extend this work by examining the potential
moderating effects of school readiness, as operationalized through childrens readiness
profiles, on childrens reading and mathematics development from 54 months age through
grade five. Latent variable growth curve analysis was employed to model childrens growth in
Letter-Word Identification and Applied Problems across four time points. Thereafter, multigroup growth curve models were estimated across groups of children identified to have
different profiles of school readiness. This allowed for examination of the extent to which
children with different patterns of school readiness, when measured at 54 months of age,
differed with respect to their concurrent (i.e., intercept) and developmental growth trajectories
(i.e., slope), as well as the association that these readiness grouping had on the variation of
these estimates.
METHODS
Participants
Data were obtained from the longitudinal National Institute of Child Health and Human
Development (NICHD) Study of Early Child Care and Youth Development (SECCYD) that
followed over 1,000 children from birth to 15 years of age. Participants were recruited by ten
93
research sites during the first eleven months of 1991 from hospitals located near the sites.
During selected 24-hour intervals, all women giving birth in a hospital were screened. This
screening consisted of an evaluation of eligibility with exclusions for mothers less than 18
years of age, multiple births, medical complications, expected moves during the course of the
study, adoption plans, lack of maternal English fluency, and location of the family outside the
area or in an unsafe neighborhood. From those remaining, a conditionally random sample of
mothers was selected who met the eligibility criteria and agreed to be telephoned in two
weeks. Conditioning assured adequate representation of single mothers, mothers without a
high school degree, and ethnic minorities. At the two-week call, families were excluded if the
baby had been hospitalized for more than seven days, the family expected to move in the next
three years, or the family could not be reached within three attempts. Additional details
regarding recruitment and selection procedures can be obtained at: http://secc. rti.org.
Data for 964 typically-developing children were available on measures that were of direct
interest in this investigation. There were approximately equal numbers of males (50.1%) and
females, and children were classified as Anglo (83.3%), African American (10.8%), Hispanic
(3.5%), and other (4%). Approximately 25% of the sample was below poverty level as
indicated by an income-to-needs ratio (U. S. Department of Labor, 1994) below 1.85.
Measures
Accurate and automatic recognition of words is necessary for reading, and in contrast to
other ways in which decoding occurs (e.g. sound-letter matching, context cues), sight
recognition places the fewest cognitive processing demands on the reader (Metsala & Ehri,
1998). While the ultimate goal of reading is comprehension, comprehension cannot occur in
the absence of decoding ability (Stanovich, 1985), thus both are necessary and important to
the process (Adams, 1990; NRP, 2000; Hoover & Gough, 1990). Word recognition therefore
serves as a good proxy diagnostic for reading, especially at younger ages (Stanovich, 2000).
While comprehension becomes more important over time (Jackson & McClelland, 1979;
Stanovich, Cunningham & Feeman, 1984; Vellutino, Tunmer, Jaccard, & Chen et al., 2007),
word recognition still accounts for substantial portions of variance in reading achievement.
This remains the case even after comprehension performance is partialed out (Bell & Perfetti,
1994, Scarborough, 1984). Accordingly, this variable represents a useful outcome measure
for reading over the course of reading development from preschool to fifth grade.
The Letter-Word Identification subtest was used as the outcome variable to measure
growth in reading achievement and the Applied Problems subtest served as a measure of
childrens analytic and practical problem solving skills in mathematics. Both were from the
Woodcock-Johnson Tests of Achieve- ment-Revised (WJ-R; Woodcock & Johnson, 1989;
Woodcock & Mather, 1989), and both have yielded reliability coefficients in the .90s and
concurrent validity correlations with other achievement tests ranging from .60 to .70
(Woodcock & Johnson, 1990). Childrens achievements on these measures were assessed at
four time points that aligned with their development at 4-5 years of age and grades 1, 3, and
5. Rasch-based W score scales were used to allow for assessments of growth over time by
placing all age groups on a common scales. W scores on the WJ-R were centered to a value of
500 in order to approximate the average performance of students in grade five (Woodcock,
1997).
94
Data Analyses
Latent growth curve analysis (LGCA; McArdle & Anderson, 1990; Meredith & Tisak,
1990) is a form of structural equation modeling that is useful for measuring change over time.
LGCA allows for the measurement of both group and individual variation in growth (Fan &
Konold, 2009) through estimation of: 1) linear and non-linear slopes that model the rate of
change across time, 2) mean intercept and slope values reflecting the group average at initial
status (or time 1) and the average rate of growth across individuals, respectively, 3) individual
variation in intercepts and slopes, and 4) the correlation between the intercept and slope.
In the current investigation, student achievement trajectories in Letter-Word Identification and Applied Problems were separately modeled across four time points. Measurements
were obtained at roughly 2 year intervals beginning at 4.5 years of age (Time 1), and grades
one (Time 2), three (Time 3), and five (Time 4). Thereafter, multi-group latent growth curve
models were examined across the six readiness groups to evaluate whether these subgroups
demonstrated different aspects of change in the two achievements under consideration.
Table 1. WJ-R Achievement Outcome Descriptive Statistics
54M
Letter-Word
Identification
54 Months (54M)
Grade 1 (G1)
Grade 3 (G3)
Grade 5 (G5)
Applied Problems
54 Months
Grade 1
Grade 3
Grade 5
M
SD
Letter-Word Identification
G1
G3
G5
.56
.51
.49
.75
.66
.85
.54
.48
.43
.42
371.12
21.60
.44
.56
.52
.48
453.16
23.55
.48
.51
.56
.52
494.52
17.85
510.98
16.40
54M
Applied Problems
G1
G3
G5
.58
.55
.53
426.39
18.32
.66
.67
470.90
15.16
.73
498.26
11.90
510.71
11.08
LWI = Letter-Word Identification, AP = Applied Problems, 54M = 54 months of age, G1 = Grade 1,

G3 = Grade 3, G5 = Grade 5.
Prominent features of the latent growth curve model are illustrated in Figure 1 and depict
the manner in which student achievements were modeled. LGCA allows for the estimation of
both latent intercept and latent slope terms, as illustrated by the ellipses. The specification of
a latent intercept provides for measurement of initial status (i.e., achievement status at 54
months of age), and the latent slope yields a measure of growth across the time. The measured
WJ-R achievements (i.e., Applied Problems and Letter-Word Identification) are distinguished
from the directly unobserved latent variables by their enclosure in boxes. Unconditional
growth-curves for both achievements were evaluated separately to estimate the initial status
(i.e., intercept) of children on these variables as well as their growth (i.e., slope) across the
four, approximately equally spaced two year, time points. All paths linking the intercept to
the observed achievements were fixed to 1, and growth parameters were empirically
95
estimated by freeing the middle two parameters (i.e., slope weights for Time 2 and Time 3)
and fixing the first (= 0 for initial measurement) and fourth (= 6 for fourth measurement after
six years) to provide a scale that reflects the total number years for the repeated
measurements. If the empirically estimated values for the 2nd and the 3rd coefficients are
different from theoretical expectation (i.e., 2 and 4, respectively given the approximate equal
spacing of measurements in two year intervals) for the middle two coefficients under the
linearity condition, evidence for a non-linear growth pattern would be indicated. This
approach is often referred to as latent basis (McArdle & Bell, 2000) or spline modeling
(Bollen & Curran, 19992006), and provides a more flexible approach to estimating patterns of
growth than is afforded by linear constraints.
The models described above were further investigated for invariance through multi-group
analyses that focused on evaluating the potential of differential growth patterns on the two
achievements across the six readiness groups. Nested model comparisons contrasted freely
estimated relationships across the six groups with equality constraints imposed across various
parameter estimates within the latent growth curve model. All nested model comparisons
were evaluated with chi-square difference (X2D) tests.
Figure 1. Latent Growth Model of Student Achievement.
Table 2. Letter-word identification and applied problems latent growth curve model estimates
Letter-Word Identification
Multi-Group Readiness Models
Total
Pattern
Coefficients
1
1,0
1,3.5
1,5.3
1,6.0
366.57
24.13
2
1,0
1,3.5
1,5.3
1,6.0
355.42
24.32
3
1,0
1,3.5
1,5.3
1,6.0
372.20
22.79
4
1,0
1,3.5
1,5.3
1,6.0
361.58
23.82
5
1,0
1,3.5
1,5.3
1,6.0
375.09
23.29
6
1,0
1,3.5
1,5.3
1,6.0
380.13
22.85
280.07
16.11
5.19
4.95
280.07
280.07
5.19
5.19
-.63
-.58
Intercept
Slope
1,0
1,3.5
1,5.39
1,6.0
371.11
23.33
Intercept V
% Change
Slope V
% Change
333.85
-5.46
--
280.07
280.07
280.07
5.19
5.19
5.19
R I,S1
-.52
-.50
-.34
-.48
Fit Statistics
Chi-Square
DF
RMSEA
NFI
TLI
CFI
108.42
3
.19
.95
.83
.95
-.32
159.15
38
.06
.92
.90
.94
Applied Problems
Multi-Group Readiness Models
Total
1
1,0
1,3.1
1,5.2
1,6.0
422.87
14.50
2
1,0
1,3.2
1,5.0
1,6.0
406.51
16.10
3
1,0
1,3.1
1,5.1
1,6.0
425.21
14.28
4
1,0
1,3.0
1,5.2
1,6.0
419.15
14.32
5
1,0
1,3.1
1,5.1
1,6.0
431.36
13.68
6
1,0
1,3.3
1,5.0
1,6.0
435.65
13.22
206.05
-1.95
--
157.07
23.77
1.36
192.85
6.41
1.36
134.93
34.52
1.36
160.31
22.20
1.36
30.26
107.44
47.86
1.36
105.99
48.56
1.36
-.75
-.57
-.52
-.62
-.69
-.70
1,0
1,3.2
1,5.1
1,6.0
426.40
14.06
9.27
3
.05
.99
.99
1.0
-.57
66.82
28
.04
.95
.94
.97
Note. V = Variance.
Underlined values denote statistically significant cross-groups differences and freely estimated values, non-underlined values statistically equivalent.
97
All models were estimated with the Analysis of Moment Structures 19.0 (AMOS)
program, using full information maximum likelihood estimation to accommodate missing
data. Numerous measures of model fit exist for evaluating the quality of measurement
models, most developed under a somewhat different theoretical framework focusing on
different components of fit (Browne & Cudeck, 1993; Hu & Bentler, 1995). For this reason, it
is generally recommended that multiple measures be considered to highlight different aspects
of fit (Tanaka, 1993). Four measures of fit were considered in evaluating model quality.
These included the Bentler-Bonett normed fit index (NFI), Tucker-Lewis index (TLI),
comparative fit index (CFI), and root mean square error of approximation (RMSEA). These
four measures generally range between 0 and 1.0. Traditionally, values of .90 or greater on
the first three measures are often taken as evidence of good fitting models (Bentler & Bonett,
1980). However, more recent research suggests that better fitting models produce values
around .95 (Hu & Bentler, 1999). Alternatively, smaller RMSEA values support better fitting
models, with values of .10 or less indicating good fit (Fan, Thompson, & Wang, 1999).
RESULTS
Means, standard deviations, and correlations among the two achievements across four
time points are presented in Table 1. For ease of presentation, results from the total sample
LGCA and multi-group LGCA are presented separately by achievement type below. Model
fit statistics and estimated model parameters are presented in Table 2.
Letter-word Identification
Total sample model summary statistics are presented on the left side of Table 2. Most
measures of fit were favorable. The NFI and CFI were within expectation for good fitting
models (i.e., > .95), and the TLI and RMSEA left room for improvement. Inspection of the
empirically estimated pattern coefficients for the middle two time points (3.5 and 5.3,
respectively) revealed deviations from linearity (i.e., the theoretical expectation under the
linearity assumption for equally spaced observations separated by two years would be 2 and
4, respectively). The starting point for Letter-Word Identification (intercept = 371.11) and
rate of growth (i.e., slope = 23.33) were both statistically significant. When the rate of growth
is applied to the empirically estimated parameterization of the growth factor (i.e., 0, 3.5, 5.3,
6.00), childrens growth from 54 months of age to grade one was [3.5 x 23.33] = 81.66; from
grade one to grade three was [(5.3 -3.5) x 23.33] = 41.99; from grade three to grade five was
[(6-5.3) x 23.33] = 16.33; and across all six years was [6 x 23.33] = 139.98. Consequently,
average overall growth was greatest from time 1 to time 2 (81.66 / 139.98 = 58%), and
slowed somewhat from time 2 to time 3 (41.99 / 139.98 = 30%), and from time 3 to time 4
(16.33/139.98 = 12%). At the same time, there was statistically significant variation in LetterWord Identification scores among children both at 4.5 years of age, S2Intercept = 333.85, p< .05;
and across time, S2Slope = 5.46, p< .05. The intercept-slope correlation of -.52 indicates that
children with lower initial Letter-Word Identification scores had faster growth rates across
future time points.
98
Multi-group latent variable growth models examined differences in Letter-Word

Identification change across the six readiness groups. Results revealed a statistically significant decline in model fit when an unconstrained multi-group general form growth model
(X2(18) = 141.84) was compared to a fully constrained model (X2(53) = 318.74; X2D(35) =
176.90, p < .05) in which cross-group equality constraints were imposed on the pattern
coefficients, intercepts, slopes, intercept and slope variances, and the intercept-slope
correlation; indicating that at least some of the growth parameters were different across
readiness groups. An investigation of partial invariance of the estimated growth parameters
was conducted to specifically determine whether a subset of these growth estimates could be
considered invariant (i.e., statistically indistinguishable across groups). Here, growth model
parameters were constrained to be equal across groups in turn (e.g., pattern coefficients
followed by intercepts, etc.). If a growth estimate was identified as being invariant across
groups, the constraint was left in place as the remaining parameters were examined. In
instances where a parameter was not found to be invariant across groups, it was allowed to be
freely estimated across groups throughout examination of the remaining parameters. This
iterative process revealed that the pattern coefficients, intercept variances, and slope variances
were statistically equivalent across groups.
By contrast, growth model readiness group differences were found to exist for intercepts,
slopes, and intercept-slope correlations. A comparison of the unconstrained multi-group
growth model with this partially constrained (i.e., cross-group equality constraints on pattern
coefficients, and intercept and slope variances; with intercepts, slopes, and intercept-slope
correlations freely estimated across groups) model (X2(38) = 159.15) revealed the two to be
statistically indistinguishable (X2D (20) = 17.31, p > .05). Moreover, model fit estimates for
the partially constrained model can be considered acceptable with NFI, TLI, and CFI values
at or above .90, and the RMSEA at .06.
Figure 2. Readiness Group Growth Patterns: Letter Word Identification.
99
Separate estimates for the six groups are shown on the left side of Table 2, with
underlined values reflecting the statistically significant cross-group differences. In terms of
overall effect sizes, accounting for different readiness groups in the model explained 16.11%
of the variance in childrens Letter-Word Identification scores at 54 months of age, and
4.95% of the variance in their growth on this outcome. The intercept values indicate average
Letter-Word Identification scores at 54 months. These values ranged from a low of 355.42
(group 2) to a high of 380.13 (group 6). While the empirically estimated pattern coefficients
indicated the same level of departure from linearity across all groups; statistically significant
differences in slope values across groups indicated that different rates of growth were present.
For example, group 3 had the slowest growth rate (slope = 22.79) across the four time points
with an overall change from 54 months of age to grade 5 of (6 x 22.79 =) 136.74 points. By
contrast, group 2 had the fastest rate of growth (slope = 24.32) with an overall change of (6 x
24.32 =) 145.92 points. Calculated growth rates for each of the six groups are presented in
Figure 2. Finally, intercept-slope correlations were also found to be statistically significant
across groups as shown in Table 2. All values were negative indicating a general tendency for
children with lower initial scores to demonstrate faster rates of growth over time. However,
the absolute values of these relationships varied across groups, ranging from a low of .32 to a
high of .63.
Applied Problems
Model summary statistics for the total sample are presented on the right side of Table 2.
All measures of fit were favorable with NFI, CFI, and TLI values greater than .95; and an
RMSEA of .05. Inspection of the empirically estimated pattern coefficients for the middle
two time points (3.2 and 5.1, respectively) revealed deviations from linearity. The starting
point for Applied Problems (intercept = 426.40) and rate of growth (i.e., slope = 14.06) were
both statistically significant. When the rate of growth is applied to the empirically estimated
parameterization of the growth factor (i.e., 0, 3.2, 5.1, 6.00), childrens growth from 54
months of age to grade one was [3.2 x 14.06] = 44.99; from grade one to grade three was
[(5.1 - 3.2) x 14.06] = 26.71; from grade three to grade five was [(6-5.1) x 14.06] = 12.65;
and across all six years was [6 x 14.06] = 84.36. Consequently, average overall growth was
greatest from time 1 to time 2 (44.99 / 84.36 = 53%), and slowed somewhat from time 2 to
time 3 (26.71 / 84.36 = 32%), and from time 3 to time 4 (12.65 / 84.36 = 15%). Here again,
there was statistically significant variation in Applied Problem scores among children both at
4.5 years of age, S2Intercept = 206.05, p< .05; and across time, S2Slope = 1.95, p< .05. The intercept-slope correlation of -.75 indicates that children with lower initial Applied Problem scores
had faster growth rates across future time points.
Multi-group latent variable growth models examined differences in Applied Problem
score change across the six readiness groups. Results revealed a statistically significant
decline in model fit when an unconstrained multi-group general form growth model (X2(18) =
49.99) was compared to a fully constrained model similar to that investigated for Letter-Word
Identification (X2(53) = 368.41; X2D (35) = 318.42, p < .05); indicating that at least some of
the growth parameters were different across readiness groups. An investigation of partial
invariance of the estimated growth parameters was conducted to determine whether a subset
of these growth estimates could be considered invariant. The same design strategy for
100
investigating partial invariance for Letter-Word Identification was used here. This iterative
process revealed that only the slope variances and intercept-slope correlations were
statistically equivalent across groups. By contrast, growth model readiness group differences
were found to exist for the pattern coefficients, intercepts, slopes, and intercept variances. A
comparison of the unconstrained multi-group growth model with this partially constrained
model (X2(28) = 66.82) revealed the two to be statistically indistinguishable (X2D (10) = 16.83,
p > .05). Moreover, model fit estimates for the partially constrained model were acceptable
with NFI, TLI, and CFI values at or above .94, and the RMSEA at .04.
Separate estimates for the six groups are shown on the right side of Table 2, with
underlined values reflecting the statistically significant cross-group differences. In terms of
overall effect sizes, accounting for different readiness groups in the model explained 30.26%
of the variance in childrens growth in Applied Problems, and varying amounts of variance in
their scores on this variable at 54 months of age. These estimates ranged from a low of 6.41%
for group 2 to a high of 48.56% for group 6. The intercept values indicate average Applied
Problem scores at 54 months. These values ranged from a low of 406.51 (group 2) to a high
of 435.65 (group 6). The empirically estimated pattern coefficients indicated different levels
of departure from linearity across all groups; statistically significant differences in slope
values across groups further indicated that different rates of growth were present. For
example, group 6 had the slowest growth rate (slope = 13.22) across the four time points with
an overall change from 54 months of age to grade 5 of (6 x 13.22 =) 79.32 points. By
contrast, group 2 had the fastest rate of growth (slope = 16.10) with an overall change of (6 x
16.10 =) 96.60 points. Calculated growth rates for each of the six groups are presented in
Figure 3.
DISCUSSION
School readiness is the product of many interacting influences on children that can
include the perspectives and goals of a community, classroom, and teacher; as well as
childrens skills, experiences, and learning opportunities. Early childhood research is often
multidisciplinary in nature because developmental domains that eventually differentiate into
separate abilities are not functionally independent when children are very young (Burchinal,
Peis- ner-Feinberg, & Pianta, 2002). Multidisciplinary approaches seek to understand the
dynamic interplay among multiple influences related to childrens development across a suite
of developmental domains associated with school readiness (Bronfenbrenner & Morris, 1998;
Ford & Lerner, 1992; Luthar, et al., 2000). This conceptualization maintains that readiness
includes both child maturation factors as well as environmental supports Meisels, 1999).
Research that combines social aspects of child development with other domains of
readiness addresses a call from those who believe strongly in the interdependence of different
domains for child readiness (Bredekamp & Copple 1997; Hyson, Copple, & Jones, 2006;
NAEYC, 1995; Rutter & Maughan, 2002). These interdisciplinary efforts often attempt to
link social and emotional developmental specialists with neurobiologists, who describe
cognition, emotion, and social engagement in physical terms (Marshall & Fox, 2006). The
early work of Bronfenbrenner and Morris (1998) was influential in the landmark document
From Neurons to Neighborhoods: The Science of Early Childhood Development (Shonkoff &
101
Phillips, 2000) that linked biological and nurturing factors to child maturation, privileging
neither as uniformly dominant. Blair et al. point to reciprocally-related developments in areas
of the prefrontal cortex that are associated with both cognition and emotion, occurring around
the age of school entry (Blair, 2002; Blair, et al., 2007). Others have sought to investigate
relationships between social cognition and language at earlier ages, which might impact later
literacy (Pruden, Hirsh-Pasek, & Golinkoff, 2006).
Neurobiologists have also begun investigating combinations of cognitive and social-emotional variables in constructs described as executive function or learning related skills.
Executive function (EF) refers to the combining of working memory, attention, and inhibitory
control applied toward goal-oriented activity. This construct is sometimes further broken
down into the more emotional hot EF and the more abstract cool EF, paralleling the
distinction made above between cognitive self-regulation and emotional regulation. Physical
support for this distinction comes by way of differential prefrontal cortex activity associated
with affective vs. abstract problem-solving (Zelazo, Qu & Mller, 2005).
Figure 3. Group Growth Patterns: Applied Problems.
Learning opportunities or affordances refer to the environmental resources available to

children (Kainz & Vernon-Feagans, 2007). Affordances in the ecologies of young children
can exist in the form of material resources (Hamre & Pianta, 2005) and social processes
(Ladd, Herald & Kochel, 2006), and can potentially serve to constrain or enhance childrens
development (Kainz & Vernon-Feagans, 2007). Current policy efforts recognize the importance of a multidimensional perspective of school readiness. Thirteen states currently encourage assessments of kindergarten readiness that include and extend beyond the assessment of
cognitive skills (Children Now, 2009).
Children are already on developmental trajectories by the time they enter formal schooling, demonstrating varying degrees of readiness across different domains of importance (Kagan et al., 1995). These differences tend to persist over time (Hulslander, Olson, Willcutt &
Wadsworth, 2010), leading to substantial disparities in academic achievement across social
102
and demographic groups (Lee & Burkham, 2002). Fortunately, the initial years of schooling
hold promise for improving achievement growth (Pianta, Belsky, Vandergrift, Houts, &
Morrison, 2008; Sonnenschein, Stapleton & Benson, 2010; Snow, Burns, & Griffin et al.,
1998). Although within person developmental trajectories may become increasingly stable
throughout the early years of elementary school (Alexander & Entwisle, 1988), early
childhood is also a time of considerable change. Development may be uneven, with
unpredictable periods of growth and plateaus (Benner & Beckmann, 1990). Childrens
profiles of school readiness in the current chapter represent a snap-shot of their proficiencies,
across six measures of cognitive executive functioning and social-behavior skills, at the
formidable time just prior to entering kindergarten. This chapter provides additional evidence
of the relationship between profiles of pre-kindergarten readiness and achievement growth
across the elementary school years.
Three of the six emergent readiness profiles examined in this chapter were characterized
by deficits or problems in one or more areas (i.e., Attention Problems, Low Cognitive Ability,
and Social and Externalizing Problems), one was primarily defined by strengths in Social
Competence, and the remaining two profiles reflected average scores or a pattern of strengths
and weaknesses, respectively (i.e., Low/Average Social and Cognitive Skills and High
Cognitive Ability/Mild Externalizing Behavior Problems). While not exhaustive of all
possible patterns of strengths and weaknesses that may present within young children, they
are characteristic of the profiles we would expect to observe in similar samples of typically
developing youth (see Konold & Pianta, 2005).
It was hypothesized that these readiness profile configurations would be associated with
childrens concurrent (i.e., intercepts) and longitudinal (i.e., slopes) achievements in reading
and mathematics, and that these associations would be moderated by profile type. Results of
our analyses clearly revealed the importance of different readiness profiles in explaining
variation in childrens achievements. Multi-group analyses that modeled subgroups of
children with different readiness profiles revealed invariance across a number of important
growth parameters. Similar to concurrent results reported elsewhere (Konold & Pianta, 2005),
starting points (i.e., intercepts) on both Letter-Word Identification (LWI) and Applied
Problems (AP) varied as a function of group membership. That is, differences were found to
exist between children with different readiness profiles on both LWI and AP when
achievements in these domains were assessed at 54 months of age. In both instances, children
with the profiles characterized by deficits in one or more area (i.e., Attention Problems, Low
Cognitive Ability, and Social and Externalizing Problems) had lower intercept values than
children in with average readiness scores or with strengths in one or more areas (i.e.,
Low/Average Social and Cognitive Skills, and High Cognitive Ability/Mild Externalizing
Behavior Problems).
Similarly, although children with different readiness profiles were found to have the same
LWI achievement growth shapes across the four time point as evidenced by the empirically
estimated pattern coefficients that were found to be invariant, differences between readiness
profile groupings for rates of growth were observed on both LWI and AP. Here, children with
the three profiles defined by deficits in one or more areas, that were found to have the lower
achievement intercept values, demonstrated faster rates of growth than children with profiles
that were characterized by average readiness profiles or strengths in one or more area. That is,
regardless of the achievement type under consideration, the three profiles with the lowest
initial status scores on LWI and AP demonstrated the faster rates of growth from 54 months
103
of age through grade 5, and the three profiles with the highest initial status demonstrated
slower rates of growth over the same time period. Thus, while children entering school less
equipped to face the challenges of formal schooling may begin (and remain) behind their
peers that are better prepared for school, their rates of growth over the early years of formal
schooling are more rapid. These patterns are also reflected in the negative intercept-slope
correlations that were observed for both LWI and AP.
The associations between readiness groups and achievements revealed in this chapter are
well aligned with previous variable-centered research documenting linkages between cognitive functioning and social-behavioral skills with educational outcomes. For example,
childrens levels of attention have consistently predicted reading ability (Duncan et al., 2007,
Thompson & Raikes, 2007), and has been particularly important at earlier ages (Hinshaw,
1992). Likewise, young childrens regulatory abilities, both behavioral and emotional, have
been shown to predict student achievement (Howse, Calkins, Anastopoulos, Keane & Shelton, 2003), even after controlling for cognitive ability (Graziano, et al., 2007).
These results are reflective of a compensatory process that recognizes that there may be
more than one avenue to success (as measured here by performance on standardized tests) in
educational outcomes among typically developing children (McClelland et al., 20070). The
non-linear patterns that constitute profiles of readiness in this chapter provide a mechanism to
capture weaknesses in some areas (e.g., social behaviors) and strengths in other areas (e.g.,
cognitive ability) that can balance out to some degree when evaluated relative to educational
outcomes. As a result, the current chapter provides additional support for continued efforts to
consider a variety of inputs when assessing childrens school readiness (Children Now, 2009).
REFERENCES
Achenbach, T. M. (1991). Manual for the Child Behavior Checklist/4-18 and Profiles.
Burlington, VT: University of Vermont, Department of Psychiatry.
Achenbach, T. M. & Rescorla, L. A. (2000). Manual for the ASEBA preschool forms & profiles. Burlington, VT: University of Vermont, Research Center for Children, Youth &
Families.
Adams, M. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA:
MIT Press.
Alexander, K. L. & Entwisle, D. R. (1988). Achievement in the first 2 years of school:
Patterns and processes. Monographs of the Society for Research in Child Development,
serial no. 218, vol. 53, No 2. Malden, MA: Black- well.
Barkley, R. A. (1997). Behavioral inhibition, sustained attention, and executive functions:
Constructing a unifying theory of ADHD. Psychological Bulletin, 121, 6594.
Beale, I. L., Matthew, P. J., Oliver, S. & Cornballis, M. C. (1987). Perfor- mance of disabled
and normal readers on the continuous performance test. Journal of Abnormal Child
Psychology, 15, 229-23 8.
Bell, L. C. & Perfetti, C. A. (1994). Reading skill: Some adult comparisons. Journal of
Educational Psychology. 86(2), 244-255.
104
Benner, S. M. & Beckmann, B. W. (1990). Is the child making expected progress? Measuring
change in rate of development with young children. Early Education and Development,
1, 424-437.
Bentler, P. M. & Bonett, D. G. (1980). Significance tests and goodness-of-fit in the analysis
of covariance structures. Psychological Bulletin, 88, 58 8-606.
Blair, C. (2002). School readiness: Integrating cognition and emotion in a neurobiological
conceptualization of child functioning at school entry. American Psychologist, 57, 111127.
Blair, C., Knipe, H., Cummings, E., Baker, D. P., Gamson, D., Eslinger, P. et al. (2007). A
developmental neuroscience approach to the study of school readiness. In R. Pianta, M.
Cox & K. Snow (Eds.). School Readiness & the Transition to Kindergarten in the Era of
Accountability (pp. 149-174). Baltimore: Brookes.
Bollen, K. A. & Curran, P. J. (2006). Latent curve models: A structural equation perspective.
Hoboken, N. J.: John Wiley & Sons, Inc.
Boyer, E. L. (1991). Ready to learn: A mandate for the nation. Princeton, NY: Carnegie
foundation for the advancement of teaching.
Bracken, S. S. & Fischel, J. E. (2007). Relationships between social skills, behavioral
problems, and school readiness for head start children. NHSA Dialog, 10, 109-126.
Bredekamp, S. & Copple, C. (1997). Developmentally appropriate practice in early childhood programs (revised edition). Washington, DC: National Association for the Education of Young Children (NAEYC).
Broidy, L., Nagin, D., Tremblay, R., Bates, J., Brame, B., Dodge, K., Fergusson, D.,
Horwood, J., Loeber, R., Laird, R., Lynam, D., Moffitt, T., Pettit, G., & Vitaro, F.
(2003). Developmental trajectories of childhood disruptive behaviors and adolescent
delinquency: A six-site, cross-national study. Developmental Psychology, 39(2), 222245.
Bronfenbrenner, U. & Morris, P. A. (1998). The ecology of developmental processes. In W.
Damon & R. M. Lerner (Eds.), Handbook of child psychology: Theoretical models of
human development. (5th ed., pp. 993-1028). New York: Wiley.
Browne, M. & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. Bollen & J.
Long (Eds.), Testing structural equation models (pp. 136-162). Newbury Park, CA: Sage.
Burchinal, M. R., Peisner-Feinberg, E., Pianta, R. C. (2002). Development of academic skills
from preschool through second grade: Family and classroom predictors of developmental
trajectories. Journal of School Psychology (40), 415-436.
Bursuck, W. D. & Asher, S. R. (1986). The relationship between social competence and
achievement in elementary school children. Journal of Clinical Psychology, 15, 4149.
Carroll, J. B. (1993). Human cognitive abilities. Cambridge, United Kingdom: Cambridge
University Press.
Children Now (2009). Kindergarten readiness data: Improving childrens success in school.
Education Policy Brief. Oakland, CA: Children Now. Retrieved 06/16/2009 from http://
publications.childrennow.org/assets/ pdf/preschool/prek09_policybrief.pdf.
Colvig-Amir, L., Liu, D. & Mobilio, L. (2005). Assessing school readiness in Santa Clara
County: Results from the 2004 school readiness assessment project. San Jose, CA: Applied Survey Research. Retrieved January 30, 2008 from http://www. Applied survey
research.org/www/products/2005_ Santa_Clara_KSRA_Report.pdf.
105
Curby, T. W., Rimm-Kaufman, S. E. & Ponitz, C. C. (2009). Teacher-child interactions and

childrens achievement trajectories across kinder garten and first grade. Journal of Educational Psychology, 101, 912-925.
Demaray, E., Maray, M. K., Ruffalo, S. L., Carlson, J., Busse, R. T., Olson, A. E. et al.
(1995). Social skills assessment: a comparative evaluation of six published rating scales.
School Psychology Review, 4, 648-671.
Denham, S. A. (2006). Social-emotional competence as support for school readiness: What is
it and how do we assess it? Early Education and Development, 17, 57-89.
Downer, J. T. & Pianta, R. C. (2006). Academic and cognitive functioning in first grade:
Associations with earlier home and child care predictors and with concurrent home and
classroom experiences. School Psychology Review, 35(1), 11-30.
Dowsett, S. M. & Livesey, D. J. (2000). The development of inhibitory control in preschool
children: Effects of executive skills training. Developmental Psychobiology, 36(2),
161-174.
Duncan, G. J., Dowsett, C. J, Claessens, A., Magnuson, K., Huston, A., Klebanov, P. et al.
(2007). School readiness and later achievement. Developmental Psychology, 43, 14281446.
Evans, J. J., Floyd, R. G., McGrew, K. S. & Leforgee, M. H. (1993). The relations between
measures of Cattell-Horn-Carroll (CHC) cognitive abilities and reading achievement
during childhood and adolescence. School Psychology Review, 31(2), 246-262.
Fan, X., Thompson, B. & Wang, L. (1999). Effects of sample size, estimation methods, and
model specifications on structural equation modeling fit indexes. Structural Equation
Modeling, 6, 56-83.
Fan, X. & Konold, T. R. (2009). Latent growth curve analysis in structural equation
modeling: Concepts and implementations. In Teo, T. & Khine, M. S. (eds.), Structural
Equation Modeling: Concepts and Applications in Educational Research (pp. 29-58).
Rotterdam, Netherlands: Sense Publi- shers.
Floyd, R. G., Shaver, R. B. & McGrew, K. S. (2003) Interpretation of the Woodcock-Johnson
III tests of cognitive abilities: Acting on evidence. In F. Schrank & D. Flanagan (Eds.)
WJ III clinical use and interpretation: Scientist-practitioner perspectives. San Diego,
CA: Academic Press, 2003.
Ford, D. H. & Lerner, R. M. (1992). Developmental systems theory: An integrative approach.
Newbury Park, CA: Sage.
Gathercole, S. E. & Pickering, S. J. (2000). Working memory deficits in children with low
achievements in the national curriculum at 7 years of age. British Journal of Educational
Graziano, P. A., Reavis, R. D., Keane, S. P. & Calkins, S. D. (2007). The role of emotion
regulation in childrens early academic success. Journal of School Psychology, 45, 3-19.
Gresham, F. M. & Elliot, S. N. (1990). Social Skills Rating System: Manual. Circle Pines,
MN: American Guidance Service.
Hair, E., Halle, T., Terry-Humen, E., Lavelle, B. & Calkins, J. (2006). Childrens school
readiness in the ECLS-K: Predictions to academic, health, and social outcomes in first
grade. Early Childhood Research Quarterly, 21, 431-454.
Halperin, J. M, Sharma, V., Greenblatt, E. & Schwartz, S. T. (1991). Assessment of the
continuous performance test: Reliability and validity in a non-referred sample.
Psychological Assessment: Journal of Consulting and Clinical Psychology, 3, 603-608.
106
Hamre, B. K. & Pianta, R. C. (2005). Can instructional and emotional support in the firstgrade classroom make a difference for children at risk of school failure? Child Development, 76, 949-967.
Hamre, B. K. & Pianta, R. C. (2001). Early teacherchild relationships and the trajectory of
childrens school outcomes through eighth grade. Child Development, 72, 625638.
Heaviside, S. & Farris, S. (1993). Public school kindergarten teachers views on childrens
readiness for school. Washington, DC: National Center for Education Statistics.
Hinshaw, S. P. (1992). Externalizing Behavior Problems and academic under- achievement in
childhood and adolescence: Causal relationships and underlying mechanisms. Psychological Bulletin, 111, 127-155.
Hoover, W. A. & Gough, P. B. (1990). The simple view of reading. Reading and Writing: An
Interdisciplinary Journal, 2, 127-160.
Horn, J. L. & Cattell, R. B. (1966). Refinement and test of the theory of fluid and crystallized
general intelligences. Journal of Educational Psychology, 57, 253-170.
Howse, R. B., Calkins, S. D., Anastopoulos, A. D., Keane, S. P. & Shelton, T. L. (2003).
Regulatory contributors to childrens kindergarten achievement. Early Education and
Development, 14, 101-119.
Hu, L. & Bentler, P. (1999). Cutoff criteria for fit indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.
Hulslander, J., Olson, R. K., Willcutt, E. G. & Wadsworth, S. J. (2010). Longi tudinal
Stability of Reading-Related Skills and their Prediction of Reading Development.
Scientific Studies of Reading, 14, 111-136.
Hyson, M., Copple, C. & Jones, J. (2006). Early childhood development and education. In A.
Reinninger & I. E. Sigel (Eds.), Child psychology in practice (6th ed. volume 4, pp. 3056). Hoboken, NJ: John Wiley.
Jackson, M. D. & McClelland, J. L. (1979). Processing determinants of reading speed.
Journal of Experimental Psychology, 108(2), 151-181.
Kagan, S. L. & Kauerz, K. (2007). Reaching for the whole: Integration and alignment in early
education policy. In R. Pianta, M. Cox & K. Snow (Eds.) School Readiness & the
Transition to Kindergarten in the Era of Accountability (pp. 11-30). Baltimore: Brookes.
Kagan, S. L., Moore, E. & Bredekamp, S. (1995). Reconsidering childrens early development and learning: Toward common views and vocabulary. Washington, DC: National
Education Goals Panel Goal 1 Technical Planning Group.
Kainz, K. & Vernon-Feagans, L. (2007). The ecology of early reading development for
children in poverty. The Elementary School Journal, 107, 407-427.
Kerr, D. C., Lunkenheimer, E. S. & Olson, S. L. (2007). Assessment of child problem
behaviors by multiple informants: A longitudinal study from preschool to school entry.
Journal of Child Psychology and Psychiatry, 48 (10), 967-975.
Klee, S. H. & Garfinkel, B. D. (1983). A computerized continuous performance task: A new
measure of inattention. Journal of Abnormal Child Psychology, 11, 487-496.
Konold, T. R. & Pianta, R. C. (2005). Empirically-derived, person-oriented patterns of school
readiness in typically-developing children: Description and prediction to first-grade
achievement. Applied Developmental Science, 9, 174-187.
Konold, T. R., Walthall, J. C. & Pianta, R. C. (2004). The behavior of child behavior ratings:
Measurement structure of the Child Behavior Checklist across times, informants, and
child gender. Behavioral Disorders, 29(4), 372-383.
107
Kupersmidt, J. B., Coie, J. D., & Dodge, K. (1990). The role of poor peer relationships in the
development of disorder. In S. R. Asher & J. D. Coie (Eds.), Peer rejection in childhood.
Cambridge studies in social and emotional development (pp. 274305). New York:
Cambridge University Press.
Kurdek, L. A. & Sinclair, R. J. (2000). Psychological, family, and peer predictors of academic
outcomes in first- through fifth-grade children. Journal of Educational Psychology, 92,
449457.
Ladd, G. W., Birch, S. H. & Buhs, E. S. (1999) Childrens social and scholastic lives in
kindergarten: Related spheres of influence? Child Development, 70, 1373-1400.
Ladd, G. W. & Burgess, K. B. (1999). Charting the relationship trajectories of aggressive,
withdrawn, and aggressive/withdrawn children during early grade school. Child Development, 70 (4), 910-929.
Ladd, G. W., Herald, S. L. & Kochel, K. P. (2006). School Readiness: Are there social
prerequisites? Early Education and Development, 17, 115-150.
Lla Paro, K. M. & Pianta, R. C. (2001). Predicting childrens competence in the early school
years: A meta-analytic review. Review of Educational Research, 70, 443-484.
Lla Paro, K. M., Pianta, R. C. & Stuhlman, M. (2004). Classroom assessment scoring system
(CLASS). Findings from the pre-k year. Elementary School Journal, 104, 409-426.
Lee, V. E. & Burkam, D. T. (2002). Inequality at the Starting Gate: Social Background
Differences in Achievement as Children Begin School. Washington, DC: Economic Policy Institute.
Lowe, L. A. (1998). Using the Child Behavior Checklist in assessing conduct disorder: Issues
of reliability and validity. Research on Social Work Practice, 8(3), 286-3011.
Luthar, S. S., Cicchetti, D. & Becker, B. (2000). The construct of resilience: A critical
evaluation and guidelines for future work. Child Development, 71 (3), 543-562.
Malecki, C. K. & Elliot, S. N. (2002). Childrens social behaviors as predict- tors of academic
achievement: A longitudinal analysis. School Psychology Quarterly, 17(1), 1-23.
Margalit, M. (1998). Sense of coherence and loneliness experience among preschool children
with learning disabilities. Journal of Learning Disabilities, 31, 173-180.
Marshall, P. J. & Fox, N. A. (2006). The development of social engagement: Neurobiological
perspectives. New York: Oxford University Press.
McArdle, J. J. & Anderson, E. (1990). Latent variable growth models for research on aging.
In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (3rd edition,
pp. 21-44). San Diego, CA: Academic Press.
McArdle, J. J. & Bell, R. Q. (2000). An Introduction to latent growth models for
developmental data analysis. In T. D. Little, K. U. Schnabel, and J. Baumert (Eds.),
Modeling longitudinal and multilevel data (pp. 69-108). Mahwah, New Jersey: LEA.
McClelland, M. M., Cameron, C. E., Connor, C. M., Farris, C. L., Jewkes, A. M., & Morrison, F. J. (2007). Links between behavioral regulation and preschoolers literacy, vocabulary, and math skills. Developmental Psychology, 43, 947-959.
McGrew, K. S., Werder, J. K. &Woodcock, R. W. (1991). Woodcock-Johnson technical
manual. Allen, TX: DLM Teaching Resources.
Meisels, S. J. (1999). Assessing readiness. In R. C. Pianta & M. J. Cox (Eds.), The transition
to kindergarten (pp. 3966). Baltimore: Brookes.
Meredith, W. & Tisak, J. (1990). Latent curve analysis. Psychometrika, 55, 107-122.
108
Metsala, J. L., Ehri, L. C. (1998). Word recognition in early literacy. Mahwah, NJ: Lawrence
Erlbaum.
Mirsky, A. F., Anthrony, B. J., Duncan, C. C., Ahearn, M. B. & Kellam, S. G. (1991).
Analysis of the elements of attention: A neuropsychological approach. Neuropsychology
Review, 2, 109-145.
Naglieri, J. A. & Das, J. P. (1990). Planning, attention, simultaneous, and successive
(PASS) cognitive processes as a model for intelligence. Journal of Psychoeducational
Assessment, 8, 303-337.
National Association for the education of Young Children (1995). NAEYC position statement
on school readiness. Retrieved on January 9, 2008 from http://www.naeyc.org/about/
positions/pdf/PSREADY98.PDF.
National Education Goals Panel. (1991). The national education goals report: Building a
nation of learners.Washington, DC: U. S. Government printing office.
Parker, J. P. & Asher, S. R. (1993). Friendship and friendship quality in middle childhood:
Links with peer group acceptance and feeling of loneliness and social dissatisfaction.
Developmental Psychology, 29, 611-621.
Pianta, R. C., Belsky, J., Vandergrift, N., Houts, R. & Morrison, F. J. (2008). Classroom
effects on childrens achievement trajectories in elementary school. American Educational Research Journal, 45, 365-397.
Pruden, S. M., Hirsh-Pasek, K., Golinkoff, R. M. (2006). The social dimension in language
development: A rich history and a new frontier. In P. J. Marshall & N. A. Fox, (Eds.),
The development of social engagement: Neurobiological perspectives (pp. 118-152).
New York: Oxford University Press.
Ramey, C. T., Collier, A. M., Sparling, J. J., Loda, F. A., Campbell, F. A., Ingram, D. L. et
al., (1974). The Carolina Abecedarian Project: A longitudinal and multidisciplinary approach to the prevention of developmental retardation. Chapel Hill, NC: Frank Porter
Graham Center. Retrieved April 12, 2009 from http://eric.ed.gov/ERICDocs/data/eric
docs2sql/content_storage_01/0000019b/80/35/44/0d.pdf.
Reilly, T. P. (1985). Concurrent and predictive validity of the WISC-R, Mc Carthy Scales,
Woodcock-Johnson and academic achievement. Psychology in the Schools, 22(4), 380382.
Rimm-Kaufman, S. E. & Pianta, R. C. (2000). An ecological perspective on the transition to
kindergarten: A theoretical framework to guide empirical research. Journal of Applied
Developmental Psychology, 21, 491511.
Rimm-Kaufman, S. E., Pianta, R. C. & Cox, M. J. (2000) Teachers judgments of problems in
the transition to kindergarten. Early Childhood Research Quarterly, 15, 147166.
Rutter, M. & Maughan, B. (2002). School effectiveness findings 1979-2002. Journal of
School Psychology, 40(6), 451-475.
Saluja, G., Scott-Little, C. & Clifford, R. (2000). Readiness for school: A survey of state
policies and definitions. Early Childhood Research and Practice, 2. Retrieved January
19, 2008 from http://ecrp.uiuc.edu/ v2n2/ saluja.Html.
Scarborough, H. S. (1984). Continuity between childhood dyslexia and adult reading. British
Journal of Psychology, 75(3), 329-348.
Shonkoff, J. P. & Phillips, D. A. (Eds.). (2000). From neurons to neighborhoods: The science
of early childhood development. Washington, DC: National Academy Press.
109
Snow, C. E., Burns, M. S. & Griffin, P. (Eds.). (1998). Preventing reading difficulties in
young children. Washington, D. C.: National Academy Press.
Sonnenschein, S., Stapleton, L. & Benson, A. (2010). The relation between the type and
amount of instruction and growth in children's reading competencies. American
Educational Research Journal, 47, 358-389.
Stanovich, K. E. (2000). The importance of word recognition in models of reading. In K. E.
Stanovich, Progress in understanding reading: Scientific foundations and new frontiers,
pp. 205-242. New York: Guilford Press.
Stanovich, K. E. (19852000). Explaining the variance in reading ability in terms of
psychological processes: What have we learned? Annals of Dyslexia, 35, 67-96.
Stanovich, K. E., Cunnningham, A. E. & Feeman, D. J. (1984). Relation between early
reading acquisition and word decoding with and without context: A longitudinal study of
first-grade children. Journal of Educational Psychology, 76 (4), 668-677.
Stinnett, T. A., Oehler-Stinnett, J. J. & Stout, L. J. (1989). Ability of the Social Skills Rating
System-Teacher Version to discriminate behavior disordered, emotionally disturbed and
non-handicapped students. School Psychology Review, 18, 526-535.
Swanson, L. (1981). Vigilance deficit in learning disabled children: A signal detection
analysis. Journal of Child Psychology and Psychiatry, 22, 393-399.
Tanaka, J. S. (1993). Multifaceted conceptions of fit in structural equation models. In K. S.
Bollen & J. S. Long (Eds.), Testing Structural Equation Models (pp. 10-39). Newbury
Park, CA: Sage.
Teo, A., Carlson, E., Mathieu, P. J., Egeland, B. & Sroufe, L. A. (1996). A prospective
longitudinal study of psychosocial predictors of achievement. Journal of School Psychology, 34(3), 285-306.
Thompson, R. A. & Raikes, H. A. (2007). The social and emotional foundations of school
readiness.In D. F. Perry, R. K. Kaufmann, & J. Knitzer (Eds.), Social and emotional
health in early childhood: Building bridges between services and systems, (pp. 13-35).
Baltimore: Brookes.
Torgesen, J. K., Wagner, R. K., Rashotte, C. A., Rose, E., Lindamood, P. & Conway, T.
(1999). Preventing reading failure in young children with phonological processing disabilities: Group and individual responses to instruction. Journal of Educational Psychology, 91, 579 593.
Vanderwood, M. L., McGrew, K. S., Flanagan, D. P , & Keith, T. Z. (2002). The contribution
of general and specific cognitive abilities to reading achievement. Learning and Individual Differences, 13(2), 159-188.
Vaughn, S., Hogan, A., Kouzekanani, K. & Shapiro, S. (1990). Peer acceptance, selfperception and social skills of learning disabled students before identification. Journal of
Educational Psychology, 82, 101-106.
Vellutino, F. R., Tunmer, W. E. & Jaccard, J. J., Chen, R. (2007). Components of reading
ability: Multivariate evidence for a convergent skills model of reading development.
Scientific Studies of Reading, 11, 3-32.
Walker, H. M. & McConnell, S. R. (1988). The Walker-McConnell Scale of Social
Competence and School Adjustment. Austin, TX: Pro-Ed.
Weikart, D. P., Bond, J. T. & McNeil, J. T. (1978). The Ypsilanti Perry Preschool Project:
Preschool years and longitudinal results through fourth grade. Ypsilanti, MI: High/
Scope Press.
110
Wentzel, K. R. (1993). Does being good make the grade? Social behavior and academic
competence in middle school. Journal of Educational Psychology, 85, 357364.
Woodcock, R.W. (1997). Woodcock Diagnostic Reading Battery: Examiners Manual.
Chicago: Riverside.
Woodcock. R. W. & Johnson. M. B., (1990). Tests of achievement. WJ-R Examiner's manual.
Allen, TX: DLM Teaching Resources.
Woodcock, R. W. & Johnson, M. B. (1989). Woodcock-Johnson Psycho-Educational BatteryRevised. Chicago: Riverside.
Woodcock, R. W. & Mather, N. (1989). WJ-R tests of achievement: Examiners manual. In
R. W. Woodcock & M. B. Johnson, Woodcock-Johnson psycho-educational batteryrevised. Allen, TX: DLM Teaching Resources.
Zelazo, P. D., Qu, L. & Mller, U. (2005). Hot and cool aspects of executive function:
Relations in early development. In W. Schneider, R. Schumann-Hengsteler, & B. Sodian
(Eds.), Young childrens cognitive development: Interrelationships among executive function, working memory, verbal ability, and theory of mind. Mahwah, NJ: Lawrence Erlbaum.
Zigler, E., Gordic, B. & Styfco, S. J. (2007). What is the goal of Hhead Sstart? Four decades
of confusion and debate. NHSA Dialog, 10, 83-97.
Zins, J., Bloodworth, M., Weissberg, R., & Walberg, H. (2004). The scientific base linking
social and emotional learning to school success. In J. Zins, R. Weissberg, M. Wang., &
H. J. Walberg (Eds.), Building academic success on social and emotional learning: What
does the research say? (pp. 1-22). New York: Teachers Press, Columbia University.

ISBN: 978-1-62618-728-3
Chapter 7
ERROR IN THE MEASUREMENT OF LOW IQ:

IMPLICATIONS FOR THE DIAGNOSIS
OF INTELLECTUAL DISABILITY IN COURT CASES
Simon Whitaker
Mental Health and Learning Disability Division
University of Huddersfield, West Yorkshire, UK
ABSTRACT
Since the ruling of the Supreme Court in the case of Atkins vs. Virginia prohibited
the execution of individuals with mental retardation, obtaining a diagnosis of intellectual
disability can have life and death significance. A key part of all definitions is having a
significantly low intellectual ability. Some states require that the individual being
assessed has a measured IQ below a specified figure, usually 70. The accuracy to which
low IQ can be assessed is therefore key to how reliably a diagnosis of mental retardation
can be made. Recent evidence on the assessment of low intellectual ability is reviewed
and it is argued that a measured IQ is only a rough indicator of an individuals true
intellectual ability. The implication of this for cranial cases is considered and a number of
recommendations made, including that an explicit distinction needs to be drawn between
measured IQ and true intellectual ability.
INTRODUCTION
A diagnosis of intellectual disability (ID) or what used to be called mental retardation
(MR), could always have a major effect on peoples lives. On the positive side it could
provide services, finance, and help in schools. On the negative side it can be a stigmatizing
label that an individual may seek to avoid (Baroff 1999). However, since the Supreme Court,
in the case of Atkins vs. Virginia, prohibited the execution of individuals with MR, it can
have life and death implications (Flynn 2006; Flynn 2007; Schalock et al. 2007).
The Supreme Court did not provide a definition of MR, leaving it up to individual states
to develop their own. This has resulted in a variety of definitions, which require different
112
Simon Whitaker
information to establish whether an individual has ID (c.f. Duvall and Morris 2006). A lack of
intellectual ability has been explicitly part of most definitions since at least 1959, when the
American Association on Mental Retardation (AAMR) (Heber 1959) defined it as a
Subaverage general intellectual functioning which originates during the developmental
period and is associated with impairment in one or more of the following: (1) maturation, (2)
learning, (3) social adjustment. (Cited by AAMR 2002, page 21). However, the different
definitions used by individual states (Duvall and Morris 2006, Death Penalty Information
Center (DPIC) http://www.deathpenaltyinfo.org/state-statutes-prohibiting-death-penaltypeople-mental-retardation) differ in whether an IQ cut-off point is specified. For example,
according to DPICs website, Maryland defines MR as: An individual who has significantly
subaverage intellectual functioning as evidenced by an IQ of 70 or below on an individually
administered IQ test, and impairment in adaptive behavior. The age of onset is before the age
of 22.
As this definition states an IQ cutoff point of 70, it implies that unless an individual has a
measured IQ of 70 or below he/she cannot be considered to have MR. On the other hand, the
definition used in California is: Significantly subaverage general intellectual functioning
existing concurrently with deficits in adaptive behavior and manifested before the age of 18.
Here no IQ cutoff point is specified so a measured IQ above 70 would not automatically rule
out a diagnosis of MR.
It is the aim of this chapter to outline recent research findings on the accuracy to which
low IQ can be measured and to consider the implications of this for definitions of ID/MR and
its diagnosis in capital cases.
ERROR IN THE MEASUREMENT OF LOW IQ

It has always been accepted that IQ tests are subject to some error due to non-intellectual
variables affecting the IQ score. The manuals of most modern intellectual assessment, such as
the Wechsler Adult Intelligence Scale fourth edition (WAIS-IV) (Wechsler 2008) and
Wechsler Intelligence Scale for Children (WISC-IV) (Wechsler 2003a), provide information
as to the accuracy of the assessments and examiners are encouraged to include this
information in their reports. In the case of the WISC-IV and WAIS-IV it is claimed in the
manuals that the measured IQ will be within five points of the true IQ 95% of the time.
However, this claimed five point accuracy for modern tests may be very optimistic and
misleading to those who are not familiar with how it is calculated. I therefore intend to give a
brief description of how error can affect test scores and how test accuracy is calculated.
These errors are of two broad types, chance and systematic (Anastasi and Urbina 1997).
Chance errors are due to a large number of relatively small factors that may or may not
occur during an assessment. These errors have the effect of diminishing the accuracy of an
individual assessment but have a much smaller effect on the mean of several assessments.
According to Anastasi and Urbina (1997) there are three types of chance error in the
measurement of IQ. First, a lack of internal consistency in the test due to test items measuring
factors other than the psychological trait being assessed cause this. Secondly, temporal error,
which is due to variation in the conditions under which assessments are administered, for
example, the level of distraction in the room or the level of motivation of the client. Thirdly,
Error in the Measurement of Low IQ
113
there is scorer error, which is due to inconsistency in scoring the assessment and can be
assessed by correlating the different scorers.
The degree to which each of these errors affect measured IQ can be represented
statistically by the 95% confidence interval, which is the range of scores, either side of the
measured IQ, in which the notional true IQ has a 95% chance of falling. Anastasi and Urbina
(1997) provide the following formula for calculating it:
95% Confidence interval = 1.96 SD (1-r)
where SD is the standard deviation of the test, usually set at 15 and r is the reliability
coefficient of the test. This confidence interval can be calculated separately for each of the
above sources of chance error.
In the past the reliability score for lack of internal consistency have been calculated by
the split-half reliability method, where the items on a subtest are split in two halves and a
correlation is found between them. However, it is now more commonly specified in terms of
coefficient alpha, which gives a more sophisticated measure of this error. Temporal error
reliability is calculated by correlating the scores when the same test is administered on two
occasions to the same people. Scorer error is indicated by the correlation between scores
when two separate scorers score the same test. It has been argued by Whitaker (2008, 2010)
that an estimate of error due to a lack of internal consistency does not take into account
temporal error or scorer error, and that an estimate of temporal error does not take into
account error due to a lack of internal consistency, though it may take into account scorer
error. Therefore to gain an estimate of the total chance error affecting a test score, one must
take into account both error due to a lack of internal consistency, temporal error and scorer
error. It is notable that this is not done when the 95% confidence interval is calculated for the
most commonly used IQ tests such as the WAIS-III, WISC-IV, and WAIS-IV (Wechsler
1997a, Wechsler 2003a, Wechsler 2008), where it is calculated taking into account only one
source of error per subtest, usually the lack of internal consistency. This failure to use all the
chance error affecting IQ scores in calculating the 95% confidence interval will have the
effect of producing a much smaller confidence interval than if all sources of error were taken
into account. If all sources of chance error were used then it would result in a 95% confidence
interval much greater than the 5 points suggested in the test manuals.
Systematic error. These errors cause one IQ test to systematically score either higher or
lower than other IQ tests, so that one test will on average score a fixed number of points
higher or lower than another IQ test. Some of these errors are now well understood, such as
the Flynn effect: the tendency for the intellectual capacity of the population as a whole to go
up from one generation to the next, causing tests to overestimate IQ as they go out of date.
Other sources of systematic error are not yet clear.
ERROR IN THE LOW IQ RANGE

When one considers how tests are developed it seems likely that both chance and
systematic error will be greater in the low IQ range than in the average range. IQ tests are
standardized using a representative sample of the population as a whole; most people in the
114
Simon Whitaker
sample will therefore be in the average intellectual range and relatively few will be at the high
and low extremes. The standardization samples for both the WISC-IV and WAIS-IV
(Wechsler 2003a, Wechsler 2008) were split into groups of 200 people at different age levels
and the test standardization was essentially done on these sub-samples. Therefore, as IQ is set
to have a mean of 100 and an SD of 15 and is normally distributed, one would expect these
sub-samples to have only five people with IQ of less than 70, and none with IQs less than 55
in any sample of 200 people. Having only a small number of people with low intellectual
ability in the sample could have a number of influences on the accuracy of the test in the low
ability range. First, sampling error would have a much greater effect. For example, if, rather
than having five individuals with true IQs less than 70, the sample had only two, which is
perfectly possible when taking a random sample of 200, then the bottom 2% of the sample
would perform better than if it was a truly representative sample. This would mean that test
criteria for getting an IQ of about 70 would be set too high. Second, it would mean that the
relationship between performance on the test and obtained IQ would have to be based on an
extrapolation from the relationship found in the mid range, rather than being empirically
derived. This is because the IQ an individual obtains is based on how well they perform
compared to the standardization sample. If, however, an individual performs less well than
anybody in the standardization sample, the only way an IQ can be allocated is to assume that
the relationship between performance on the test, found in the average range, continues into
the low range and this may not be valid. Thirdly, as the vast majority of subjects in the
standardization sample pass the test items that effectively measure low IQ, the psychometric
properties of items may not have been properly assessed. Recent evidence seems to support
the notion that both chance and systematic error may be greater in the low range.
Chance Error
Temporal error. The current evidence suggests that temporal error may be greater in the
low range than it is in the average range. A meta-analysis (Whitaker 2008) of the test re-test
reliability for Full Scale IQ (FSIQ) for assessments in the low IQ range (IQ<80) found a
weighted mean test re-test reliability of .82. To explain what this means in terms of
quantifying the error: All the studies used in the meta analysis gave the same test to the same
individuals twice. As one would expect the individuals true IQ to remain the same between
assessments, it would be expected that if the tests were not subject to error the same score
would be obtained on both assessments so the correlation would be 1.00. Therefore, the
difference between the actual obtained correlation and 1.00 represents the amount of error due
to a lack of stability of the test. The temporal error is therefore one minus .82, which is .18
and the corresponding 95% confidence interval is 12.5 points. As several of the studies in the
meta analysis reported the proportion of IQs that changed by specific amounts, Whitaker was
able to check how accurate this 95% confidence interval was in predicting IQ change when it
was re assessed. It would be expected that, for a 95% confidence interval 12.5 points for
stability, 61% of IQs would change by less than 6 points and 13% would change by 10 points
or more. In the studies, 57% of IQs changed by less than six points and 14% changed by 10
points or more, which is a good estimate of what was predicted, suggesting that the 95%
confidence interval of 12.5 points is accurate for clients in the low ability range. However, it
is considerably greater than the corresponding 95% confidence intervals for temporal error of
115
9.75 and 5.88 for the WISC-IV and WAIS-IV respectively for people in the average ability
range (Wechsler 2003a, Wechsler 1997a). There is therefore more fluctuation in individuals
tested IQs at the low level than in the average range. This may be due to greater error in
measurement, actual change in intellectual ability between tests or both. Whitaker (2008)
found that there was no significant relationship between the test re-test reliability and the
interval between testing. This suggests the change was not due to a systematic change in
intellectual ability over time and the most likely explanation for the change in scores is
measurement error, however, there is a possibility that intellectual ability fluctuates from day
to day.
Internal consistency. There is now an up to date estimate of error due to a lack of internal
consistency in the low range. As part of the validation of the WAIS-IV (Wechsler 2008), it
was given to a group of 75 adults with mild ID and 35 with moderate ID and the internal
consistency reliability calculated for those sub-tests where this could be done. On average this
reliability was approximately the same as the internal consistency reliability found for the
standardization sample, which means that the overall reliability for internal consistency in the
low range is approximately .98, the reliability found in the average range. This would mean
the error due to a lack of internal consistency in the low range is about .02 and the 95%
confidence interval about 4.2 points.
Total chance error. If the error due to lack of internal consistency of .02 is added to the
error due to a lack of temporal stability, found by Whitaker (2008), to be .18, it gives an
estimate of the total chance error in the low range of .20. It was argued by Whitaker (2010)
that an estimate of the effective total reliability could be obtained by subtracting the estimate
of total error from one. If this is done for the total chance error of .20, it gives an effective
total reliability figure of .80, which corresponds to a 95% confidence interval of 13 points.
Systematic Error
A floor effect. All the Wechsler intellectual assessments measure IQ by giving the client a
number of subtests measuring different aspects of intellectual ability. The maximum possible
raw score on the different subtests is different on each of them. As part of the calculation of
FS IQ, the raw scores on these subtests are converted to scaled scores with a mean of 10, an
SD of three, and a range between one and 19. Whitaker (2005) has suggested that allocating a
scaled score of one to low raw scores or a raw score of zero could result in an overestimate of
intellectual ability. This can be illustrated by looking at the relationship between raw scores
and scaled scores for the Digit Span subtest for age groups 15:8 to 15:11 taken from WISCIV Administrative Manual (Wechsler 2003b):
Raw Score:
Scaled Score:
18 17 16 15 14 13 12 11 10 0-9
10 9 8 7 6 5 4 3 2 1
The relationship is linear between raw score 18 and raw score 10, a reduction in a raw
score by one corresponding to a reduction in a scaled score by one. However, all raw scores
from nine down to zero are then given a scaled score of one. There is no empirical reason to
suppose that all raw scores below nine are equivalent to a scaled score of one, and logic
suggests that the linear relationship between scaled scores and raw scores should continue for
116
Simon Whitaker
some way below raw score nine. Therefore a scaled score of one given for a raw score of less
than nine is likely to be an overestimate of the clients ability. So generally when a scaled
score of one is given there is a distinct possibility that the clients ability is being
overestimated. This will clearly affect IQs in the 40s where scaled scores of one are
inevitable. The degree to which it also affects IQs in the 50s, 60s and 70s was investigated by
Whitaker and Wood (2008), who plotted the distribution of scaled scores from WISC-III
(UK) (Wechsler 1992), and WAIS-III (UK) (Wechsler 1997b), assessments that had been
given as part of clinical practice for people with ID. The distribution of scaled scores for the
WAIS-III (UK) was approximately normal with very few scaled scores of one, suggesting
that the floor effect would only be a potential problem for IQs in the 40s and 50s. However,
with the WISC-III (UK) there was a skewed distribution with more scaled scores of one than
any other scaled score. Scaled scores of one were found at all IQ levels up to and including
those in the 70s, where they accounted for 10% of scaled scores. A similar distribution of
scaled scores to that found on the WISC-III (UK) has now been found on the WISC-IV (UK)
(Whitaker and Gordon 2012). There is therefore a distinct possibility that, particularly on the
WISC-III and WISC-IV, IQ scores are increased at low ability level due to this floor effect.
The Flynn Effect. There is good evidence that the intellectual ability of the population as
a whole has increased over at least the last 100 years at about 0.3 of an IQ point per year on
average (Flynn 1984; 1987; 2007). This is known as the Flynn effect, after James Flynn, who
initially researched it. The evidence also shows that the effect occurs at the low IQ levels
(Flynn 1985; 2006) and is still doing so today in the US (Flynn 2009). The implication of the
Flynn effect for the assessment of IQ is that tests will become less accurate and overestimate
intellectual ability as they become more out of date. On average a test will overestimate an
individual IQ by about 0.3 IQ for each year since it was standardized. It has therefore been
argued by Flynn (2007; 2009) that it is possible to compensate for this error by subtracting .3
of an IQ point from the measured IQ for each year between the test being standardized and
given.
Error apparent from the differences between IQ scales. It is accepted that different IQ
tests will give slightly different results (Floyd et al. 2008). In the absence of a test that is
clearly an accurate measure of true intellectual ability the best that can be done is to decide
which of the many IQ tests is likely to be the most accurate and take that as the gold
standard assessment against which other assessments should be compared.
The Wechsler assessments should have a good claim to be regarded as the gold standard
assessments. They have evolved over 70 years since the Wechsler Bellevue was first
published in 1939 (Wechsler 1939), are apparently well standardized and are probably the
most widely used tests of child and adult intelligence.
However, it has been reported that early versions of the WISC scored systematically
lower than the equivalent WAIS when used in the low intellectual range (Flynn 1985, Spitz
1986, 1989). There is also a recent study by Gordon et al. (2010) that compared the WISC-IV
(UK) and the WAIS-III (UK) in the low range. Both assessments were given in
counterbalanced order to a group of 16 year-olds receiving special education. In each case the
FS-IQ on the WISC-IV (UK) was less than that on the WAIS-III (UK); the mean FS-IQ on
the WISC-IV (UK) was 53.00, which compared to a mean of 64.82 on the WAIS-III (UK), a
difference of just less than 12 points. The correlation between the two assessments was
relatively high (r=.93), suggesting that the tests were both measuring the same trait and were
given consistently.
117
Of this 12 point difference between the two tests, it is likely that two points are due to the
Flynn effect as the WAIS-III was standardized six years before the WISC-IV, while the
remaining 10 points are due to yet unknown factors.
As the degree to which either assessment is in error is not known, it is clearly possible
that either the WISC-IV (UK) is systematically underestimating true IQ by up to 10 points, or
the WAIS-III (UK) is systematically overestimating true IQ by 10 points or both assessments
are making systematic errors of less than 10 points. Although this work was done with the
UK versions of the WISC-IV and the WAIS-III, there is unpublished evidence (abstract is
available Bresnahan 2008) that the same effect occurs on the US versions of the tests.
Combined error. Whitaker (2010) combined the error in the low range from lack of
internal consistency, temporal error, the floor effect, the Flynn effect and the error apparent
from the difference between the WISC-IV and WAIS-III and estimated the 95% confidence
intervals for both the WISC-IV and WAIS-III. For the WISC-IV there was an effective
confidence interval, which extends 16 points below the measured IQ and 25 points above it.
For the WAIS-III the effective confidence interval extended 18 points above the measured IQ
and 28 points below. This analysis is based on a number of assumptions with regard to
combining measurement error and makes use of data from the UK versions of the WISC and
WAIS. It also based the error due to a lack of internal consistency on a study by Davis (1966)
in which a mean reliability figure of .92 was found for the low range on the WISC, which is
significantly lower than was found in the standardization of the WAIS-IV referred to above.
Because of this, the effective 95% confidence for total chance error was 15 points rather
than the 13 points suggested above, using the internal consistency of .98 taken from the
WAIS-IV standardization. It therefore will be argued that this very large margin of error will
not apply in the US today. However, it would be very difficult to escape the conclusion that
the degree of error in the measurement of low IQ is much greater and the tests far less
accurate than had previously been supposed.
OTHER ERROR SPECIFIC TO LEGAL ASSESSMENTS

In addition to the above sources of error that Whitaker (2010) combined, the literature on
the measurement of IQ in forensic cases highlights two additional sources of error:
malingering and the practice effect.
Malingering. This is deliberately underperforming on an assessment in order to get a low
score. An individual appealing the death penalty on the grounds of having an ID may be
motivated to do this. It is clear from the literature (c.f. Salekin and Doane 2009) that courts
are well aware of the possibility of this error in assessment, however, currently there does not
seem to be a reliable way of detecting when it is occurring when assessing an individual with
low intellectual ability.
Practice effect. This occurs when an individual does better on a test the second time
he/she is given it due to having had the opportunity to practice when they were first given it.
The degree to which it affects the second test score can be estimated from studies that have
given the same test to the same clients on two occasions, such as test re-test reliability studies.
Both the WISC-IV and WAIS-IV manuals (Wechsler 2003, Wechsler 2008) describe such
studies, with mean FSIQ increasing by 5.6 points on the WISC-IV over an average test re-test
118
Simon Whitaker
interval of 32 days and by 4.3 points on the WAIS-IV over a mean interval of 22 days.
However, practice effect is likely to decrease as the interval between testing increases and
may well also be a function of ability level. In this respect it is notable that Whitaker s
(2008) meta analysis of test re-test reliability in the low range found a mean increase in FSIQ
of only .41 of an IQ point for a mean test re-test interval of 2.3 years.
CORRECTING FOR ERRORS IN THE ASSESSMENT

OF LOW IQ
It may be possible to reduce and even eliminate some of these errors in the assessment of
an individuals intellectual ability. However, it is the contention of the current author that it is
not currently possible to reduce the errors to such an extent that the accuracy of the tests
approaches the five points suggested in the manuals, nor is it possible to quantify how
accurate the tests are after most corrections have been applied.
The error due to lack of internal consistency reported in the WAIS-IV manual, even in
the low IQ range, is very small. The error is due to inconsistent test items and the experience
of the client, possibly over many years. Neither of these can be changed for an individual
assessment and so there is no score of reducing this error for an individual assessment. If,
however, it is assumed that these errors will be the same when tests are given in a clinical or
forensic setting, and there is no reason to suppose they would not be, then the accuracy of the
test can be quantified with regard to lack of internal consistency by the 95% confidence
interval.
Temporal error is due to changes in the state of the individual, such as alertness and
motivation, changes in the environment in which the test is given, such as level of distraction,
or changes in the way the test is given, such as how the accurately the test instructions are
given, between assessments. Here being very strict about only giving the test in optimum
conditions may well reduce error. It is likely that many of the assessments that are done in
clinical and criminal justice settings are not done under optimal conditions and this may be
one reason why the test re-test reliability found by Whitaker (2008) is lower than that found
when the tests were standardized. For example in criminal justice settings such as prisons
there may be distractions due to noise; the client may be depressed or anxious; he/she may not
be motivated to do well, the examiner may not be used to giving assessments in such settings
and/or to giving assessments in the low IQ range. However, even though these errors could be
reduced, it will not be possible to tell by how much they have been reduced in numerical
terms. Without further studies in which the test re-test is assessed under various conditions,
one must rely on the studies that have been done up to now which, according to Whitaker
(2008), suggest a 95% confidence interval of 12.5 points for the stability of IQ over time.
The Flynn effect could be minimized by always using the latest standardization of a test,
because if a test has only just been standardized there will be no Flynn effect. If a test is a few
years old, Flynn (2009) has argued that the effect can be corrected by subtracting 0.3 of an IQ
point for each year since the test was standardized. However, whereas it may have been the
case for the US in 2009, it may not always be the case and does not seem to be the case in
other areas of the world at the moment. There is evidence from Scandinavia that the effect
may have gone into reverse in the low range (Teasdale and Owen 2005), resulting in tests
119
underestimating IQ as the go out of date. It therefore cannot be assumed that this method of
correction for the Flynn effect will continue to be valid, nor can we tell, without extensive
studies, how valid it is at any one time.
Whitaker and Gordon (2012) have suggested that the floor effect could be corrected for
by extrapolating the relationship between raw scores and scaled scores down below scaled
score one, so that scaled scores of zero and less can be allocated to low raw scores. However,
although this method seems logical and may be valid for small correction where a scaled
score of zero or minus one is given instead of a scaled score of one, there is no empirical
evidence to support the procedure. Therefore, although applying some correction for the floor
effect may result in a more accurate assessment, it is not possible to say how much more
accurate an IQ score corrected for the floor effect is than one that is not corrected.
The error apparent from the differences between tests may be very difficult to correct for.
It could only be done if there was firm evidence that one test was accurate and it was known
by how much other tests scores systematically differed from this gold standard test. Adding
or subtracting the appropriate number of points from their obtained scores could then correct
inaccurate tests. However, this is currently not possible for a number of reasons: First,
although Gordon et al. (2010) have produced evidence that the WISC-IV systematically
measured lower than the WAIS-III, the study was based on a small sample of 16-year-olds
and so it cannot be assumed that the exact difference between the tests due to factors other
than the Flynn effect is 10 points. Secondly, the WAIS-III is now no longer used and there are
no studies comparing the WAIS-IV and WISC-IV in the low intellectual range. Thirdly
although it could be argued that the WISC-IV is more likely to be accurate than the WAISIV, as getting a representative sample of children with low IQs would seem to be less subject
to error than getting one of adults (c.f. Flynn and Weiss 2007), this is only speculation and in
reality we do not know to what degree either of these gold standard tests is accurate or
inaccurate or indeed if other less well known tests are more accurate in the low range.
The practice effect could be eliminated by not assessing an individual twice on the same
assessment. However, if this has to be done, then subtracting an appropriate number of IQ
points from the second assessment could reduce the error. The problem then is deciding how
many points to reduce the second assessment by. The practice effect will vary depending on
which test is being given, the time interval between assessments as well as factors within the
individual such as his/her level of intellectual ability. Therefore one cannot be sure how
accurate a corrected score is.
Clearly one way to eliminate error due to malingering is not to assess individuals who
may well be motivated to do poorly in the assessment, which may very well be the case when
an assessment is being done in a death penalty case. However, if this had to be done, there
does not seem to be any reliable way of detecting malingering (Salekin and Doane 2009),
which means there also is no way of telling to what degree a score has been affected by it.
It is clear therefore that IQ tests are subject to more error, particularly in the low range,
than has previously been accepted. It is also likely that, although some of these errors can be
reduced or even eliminated, it is not possible to say how accurate the assessments are in terms
of a 95% confidence interval without making assumptions for which there in a lack of
evidence. One is therefore left with the relatively wide margins of error of the order of that
suggested by Whitaker (2010).
120
Simon Whitaker
CURRENT APPRECIATION OF THESE ERRORS

Judging by the current literature on the measurement of low IQ there seems to be only a
partial awareness of the errors outlined in this paper. There is a clear understanding of the
Flynn effect (Ceci et al. 2003; Flynn 2009; Olley 2009a). There is some awareness of the poor
stability in the low range (Olley 2009a, b). There is little questioning of the practice of
calculating the 95% confidence interval on the basis of only one source of error, though
Gresham (2009) does discuss it. Apart from Whitaker and Wood (2008) there is no
acknowledgement of the floor effect, which could increase scores by the same order of
magnitude as the Flynn effect. There seems to be no realization that the WISC and WAIS
may differ by the order of 10 points. Apart from Whitaker (2010), there has been no attempt
to combine all these errors and get an overall degree of confidence. There seems to be an over
reliance on information on reliability given in the test manuals, which is based on data
obtained using a non-clinical and non-forensic sample who, on the whole, had average
intellectual abilities. If this lack of awareness extends to those advising courts and doing
assessment in criminal cases then it is likely that courts are being misled.
IMPLICATIONS AND RECOMMENDATIONS

The greater error in the measurement of low IQ has a number of implications for the
assessment of intellectual ability in criminal cases and particularly death penalty cases. These
are outlined below, together with some recommendations as to how best to deal to address
them.
CONFUSION BETWEEN TRUE INTELLECTUAL ABILITY

AND MEASURED IQ
It was noted above that some state definitions of ID specify a specific IQ figure below
which a defendant must score to be considered to have MR, while others simply indicate that
the individual must have a significantly low intellectual ability. It is likely that the lack of
precision in the measurement of low IQ will have a greater negative impact in states where
the definition of MR specifies an IQ figure. If the individual being assessed has a true
intellectual ability in the mild to borderline range, whether or not he/she obtains a measured
IQ below the specified figure will be to some extent a matter of luck. But more than this, it
gives a score for the assessor to produce the IQ score that may suit the case of those engaging
them to do the assessment. Although clearly the responsibility of an expert witness is to give
impartial advice to the court, it is possible that they may unconsciously swayed to produce a
particular result, or that the attorneys employing them may have chosen them for their
apparent views or track record in obtaining certain results. Therefore, if the assessor is
commissioned by the defense it is possible that they may want the client to have a measured
IQ less than the specified figure. This could be made more likely if they do the following:
Put emphasis on WISC-IV rather than WAIS-IV scores. Although in a death penalty case
that the defendant would be older than 16, so it would not be possible to reduce his/her score
121
by assessing him/her on the WISC-IV rather than the WAIS-IV, the assessor could still argue
that greater weight should be given to WISC-IV and WISC-III assessments done when the
defendant was a child rather than WAIS-IV assessments done following committing the
offence. In part this is because of problem with malingering but also as it has been suggested
that the WISC-IV may be a more accurate assessment at low levels as it is easier to get a
representative population sample of children with low intellectual ability than it is of adults to
standardize the test (c.f. Flynn and Weiss 2007).
Allow the assessment to be done under sub-optimal clinical or forensic conditions. It is
likely that if an IQ assessment is done under sub-optimal conditions, such as may occur in
some prison or clinic settings, the score will be reduced. In order to obtain a lower score the
assessor may therefore not insist on a distraction free environment or that the defendant was
in a state to give of their best.
Ensure the defendant is aware of the consequences of a high score. For example in death
penalty cases the assessor could make it clear to the defendant that a high score will increase
the likelihood of being executed.
Not using a test that the client has been assessed on recently. By not re-assessing the
defendant on a test he/she has previously been assessed on, there will be no practice effect.
Correct for the Flynn and floor effects. It was argued above that scores can be artificially
increased by both the Flynn effect and the floor effect but that these scores could be corrected
to some extent.
If the assessor was working for the prosecution he/she may wish to ensure a high score,
in which case he/she could:
Put emphasis on the WAIS-IV rather than the WISC-IV. The assessor could argue that
emphasis should be put on more recent WAIS-IV score than older WISC-IV or WISC-III
scores as the WAIS-IV results reflect how the individual is now and that higher scores are
less likely to be subject to error.
Ensure that assessments are done under optimal conditions. Rather than accepting poor
conditions such as may be offered in a prison setting the assessor could insist on as near to
optimal conditions as possible. They could also put off an assessment if the defendant was not
in an optimal condition himself or herself to be assessed.
Ensure that the individual is motivated to do well on the assessment. Rather than
emphasize the negative effects of doing well on the assessment emphasis could be put on the
positive consequence, such as showing that you are smart.
Re-assess on a test that has been used recently in the past. By using an assessment that
the individual has only recently been assessed on, possibly by the defense assessor, the score
should be increased due to the practice effect.
Not correcting for the Flynn or floor effects. The obtained scores could simply be
presented without any correction. Also, if possible, the assessor could use an older version of
the test to maximize the Flynn effect.
Although it is not be possible to put an exact figure on the degree to which it would affect
scores, if an individual were assessed under these two extreme sets of conditions, it is likely
that the two obtained IQ scores would differ by the order of 25 points. Twelve points would
be attributable to differences between the tests used due to the Flynn effect and other factors,
10 points to temporal error due to difference in assessment conditions, one point for the floor
effect, two points for the practice effect, as well as some effect due to malingering. However,
both assessments would produce a measured IQ score. If a court was unaware that these
Simon Whitaker
122
manipulations could have such a large effect on measured IQ and regarded measured IQ as a
good indicator of true intellectual ability, then it would be likely to accept an IQ score as a
good indicator of the defendants actual intellectual ability. The likelihood of this happening
could be reduced if there was an explicit distinction drawn between measured IQ and true
intellectual ability. An individuals true intellectual ability, at any one time, is the IQ score
that he/she would obtain if they were assessed with a perfectly standardized, valid and
reliable IQ test without any error in administration. An individuals measured IQ is the IQ
score he/she would obtain if he/she were a given a current IQ test, under a particular set of
conditions. Currently measured IQ is only a loose indicator of true intellectual ability.
IMPROVING THE ACCURACY OF MEASURED IQ

As noted above, an individuals measured IQ and his/her true intellectual ability may
differ by the order of 25 points. If this is acknowledged and the assessors and the courts are
motivated to get the best estimate of true intellectual ability, there are a number of things that
can be done:
It should be made explicit in any reports and evidence given to the court that
measured IQ and true intellectual ability are not the same thing and that measured IQ
is only a rough estimate of true intellectual ability.
The assessor should correct for the Flynn effect and the floor effect. As these
corrections may be contentious, the fact that they have been made should be
indicated, together with a justification for having made them, and a statement as to
how much difference it made to the IQ scores.
If it is known that the test being used systematically measures lower or higher than
another gold standard test, then this should be made very clear in the report. So, for
example, if reporting on the results of a WAIS-III or WAIS-IV, it should be noted
that there is evidence that it measures higher than the equivalent the WISC-IV, and
that it is not clear which test is producing the best estimate of true intellectual ability.
Although this will not eliminate the systematic error it will make courts aware of it.
Although IQ tests done prior to the defendant being charged may not be adversely
affected by motivation to do poorly, the information that comes with them may
reduce their value. If one knows the test that was used and when it was given one can
adjust for the Flynn effect and make it comparable with other tests. If the raw scores
are available the score can be corrected for any floor effect. If there is information as
to the conditions under which the assessment was given one can speculate as to
whether the score was higher or lower than would have occurred if the assessment
was done in optimal conditions. However, simply giving an IQ score without any
information about how the IQ was obtained is almost worthless. If an old IQ
assessment is used in evidence the assessor should obtain as much information as
possible as to how it was done and indicate in his/her report how this information
would affect the score and how reliable the score is felt to be.
If there have been several intellectual assessments given the assessor should look for
consistency in the different IQ scores. In order to do this all the scores should be
123
corrected in the same way for the Flynn effect, floor effect and the difference
between tests. A key question is then: do the scores give more or less the same
result? If so, it provides stronger evidence that there is consistency in measured IQ,
suggesting that chance error has had only a relatively small effect on the scores. It
will also provide evidence as to whether the individuals true intellectual ability is
above or below the criterion IQ for a diagnosis of ID. However, some caution would
have to be exercised before drawing such a conclusion as systematic error is not
corrected for. In correcting for the systematic error apparent between tests we can
adjust a score on one test so that it is comparable to the score on another test. For
example, if an individual scored 62 on the WISC-IV, with no floor effect, and 72 on
the WAIS-III, when corrected for the Flynn effect and with no floor effect, then this
10 points, which evidence suggests is the expected difference between the two tests
(Gordon et al. 2010), could be subtracted from the WISC-III score, making the
adjusted score on the WAIS-III 62, the same as the score on the WISC-IV. One may
therefore be tempted to say that the true intellectual ability of the defendant was 62.
However, this would be to assume that it is the WAIS-III that is systematically in
error in measuring true intellectual ability and not the WISC-IV, whereas all we can
say is that either test may be in error by up to 10 points, so the individuals true IQ
could be 72. The other practical problem with this is that we do not know the degree
to which most tests differ from each other in the low range so in reality adjusting test
scores so they are consistent with other scores is not something that can be done.
It may well be the case that, even when test scores have been adjusted, there are still
significant differences between the results of assessments done at different times and/or with
different tests. In this situation it is not entirely clear what should be done, though there are a
number of options, each with advantages and disadvantages. One possibility would be to take
the average score. This would assume that some assessments were subject to errors that
increased the scores above true intellectual ability and some to errors that decreased scores
below true intellectual ability, so that averaging the scores cancelled out the errors to give a
good estimate of true intellectual ability. However, this would also assume that the
individuals true intellectual ability was the same on each occasions he/she was assessed and
that the effect of errors was evenly distributed between those that increased scores and those
that decreased scores, which are assumptions that one would not normally have evidence for.
A second approach would be to argue that most error will decrease score, so that lower
scores are therefore likely to be subject to more error and therefore the highest scores are
more accurate. However, whereas it does seem that high scores are more likely to be correct,
this is not inevitable and the high scores could be due to systematic error or change in
intellectual ability over time.
Real world consistency, that is how the defendant has coped with intellectually loaded
everyday tasks, could give support to a particular score. In part this could be indicated by an
assessment of the individual level of adaptive behavior on a scale such as the Vineland II
(Sparrow et al. 2005). However, the assessor should be aware that there are often only low
correlations between adaptive behavior and IQ, for example the correlation between the
Vineland II composite score and FS IQ on the WISC-III is only .09. So ability to do some
tasks such as socialize with others or have daily living skills are not indicative of high IQ. On
the other hand IQ does correlate well with academic ability. As part of the standardization of
124
Simon Whitaker
the WAIS-IV, it was compared with the Wechsler Individual Achievement Test (WIAT-II)
(Wechsler 2001). FS IQ on the WISC-IV correlated with reading .78, with Math .78, with
written language .76, with oral language .75 and with total achievement .87. These relatively
high correlations suggest that if an individual has a genuine IQ less than 70 it is unlikely that
they would be performing in the average range academically.
It is likely that even if the above suggestions are followed it still will not be possible to
give an exact figure on an individuals IQ or an exact confidence interval indicating how
accurate the test result is. It should therefore be made clear in any information given to a court
that any score is not exact and the confidence interval is many more points than that stated in
the manual.
If the IQ score is required for a diagnosis, the issues may not be so much what the exact
IQ figure is but rather whether it is below the critical figure required for a diagnosis of ID. To
take an extreme example, if an individual had several measured IQs in the 90s, then, even
though it would not be possible to put an exact figure on his/her IQ, one could say with a high
degree of certainty that the individuals true intellectual ability was above 70. However, it is
likely that the majority of defendants arguing in court that they have ID will have measured
IQs in the 60s and 70s, in which case it will not be possible to say with any certainty that their
true intellectual ability falls above or below 70. The best approach in such cases may be to
give an estimate of the probability of the individuals true intellectual ability falling above or
below the critical figure in non exact term, using such works as: possible, likely or unlikely.
COURTS UNDERSTANDING OF THE INFORMATION

It has been pointed out that courts may have some difficulty understanding complex
information about psychometrics and so fall back on their own preconceived ideas of what
constitutes MR/ID (Greenspan and Switzky 2006). Often these lay ideas as to what MR/ID is
are somewhat different from what is intended by the formal definitions of MR/ID and may
require an individual to show very obvious signs of disability. The problems in measurement
of low IQ outlined here are likely to add to the confusion that courts have and there may be a
tendency to fall back on a lay interpretation of what IQ is. The way the courts think about the
accuracy to which IQ can be measured may be influenced by the way IQ is spoken about by
the general public, usually referred to as a single whole number, very much in the same way
as a persons height may be reported in feet and inches. It may well be that the way measured
quantities are reported implies a degree of accuracy to which they have been measured. For
example, an individuals height is usually be reported as being in feet and inches implying
that height can be measured to an accuracy of one inch, which it can. An individuals IQ is
usually spoken about amongst lay people and reported in the media as a whole number,
implying that it can be measured to the level of accuracy of one IQ point. It is therefore likely
that most lay people consider that IQ can be measured to an accuracy of one point. This then
may lead courts to regard the concept of error in its measurement with suspicion. It is
therefore important that it is clearly explained to courts that measured IQ is subject to error
and it is only a rough estimate of true intellectual ability. Possibly a good analogy to help then
grasp the point is an individuals weight which will fluctuate from day to day by a pound or
125
so, and will vary between scales, though the overall accuracy of measuring an individuals
weight is far greater than that of his/her IQ.
BURDEN OF PROOF
When the Supreme Court made its ruling prohibiting the execution of people with MR
and when the states drew up their definition of MR, the prevailing opinion was that IQ could
be measured to an accuracy of 5 points. It is the core argument of this paper that the margin of
error is much greater than 5 points. Therefore, on the one hand it is much more difficult to
prove absolutely that somebody has a true intellectual ability of less than 70, and on the other
hand it raises the possibility that an individual with measured IQs in the 80s may have a true
intellectual ability in the 60s.
The burden of proof for establishing that a defendant has ID usually lies with the defense.
The defense will therefore have to present evidence and argument to the court to demonstrate
that the individual does have ID, which may well be challenged by the prosecution. As a great
deal of what was considered to be established fact with regard to the measurement of low IQ
has been brought into question, the likelihood is that there will be considerable argument in
court as to what a defendants IQ is. However, in reality, with our current state of knowledge,
in many cases it will not be possible to say with any degree of certainty that an individual has
a true intellectual ability above or below 70. It is therefore a question of whether the benefit
of the doubt is given to the defendant or the prosecution. If it is given to the prosecution then
it is likely that people who genuinely do have MR will not be able to establish this and will be
executed, which was not the intention of the Supreme Court.
REFERENCES
American Association on Mental Retardation (2002). Mental Retardation: Definition,
Classification, and System of Supports (10th Edition). Washington DC: American
Association on Mental Retardation.
Anastasi, A. and Urbina,S. (1997). Psychological Testing (seventh edition). Upper Saddler
River: Prentice-Hall Inc.
Baroff, G.S. (1999). General learning disorder: A new designation for mental retardation.
Mental Retardation, 37, 68-70.
Bresnahan, J. A. (2008). A preliminary study of WISC-iv and WAIS-III IQ scores for
students with extremely low cognitive functioning. Abstract of unpublished
dissertation. Dissertation Abstracts International: Section B: The Science and
Engineering Vol 68(9-B), 6383.
Ceci, S.J., Scullin, M. and Kanaya, T. (2003). The difficulty of basing death penalty
eligibility on IQ cutoff scores for mentally retarded. Ethics and Behavior, 13, 11-17.
Davis, L.J. (1966).The internal consistency of the WISC with the mentally retarded.
American Journal of Mental Deficiency, 70, 714-716.
126
Simon Whitaker
Duvall, J.C. and Morris, R.J. (2006). Assessing mental retardation in death penalty cases:
Critical issues for psychologists and psychological practice. Professional Psychology:
Research and Practice, 37, 658-665.
Floyd, R.G., Clark, M.H. and Shadish, W.R. (2008). The exchangeability of IQs: Implications
for professional psychology. Professional Psychology: Research and Practice, 39, 414423.
Flynn, J.R. (1984). The mean IQ of Americans: Massive gains 1932 to 1978. Psychological
Bulletin, 95, 29-51.
Flynn, J.R. (1985). Wechsler intelligence tests: Do we really have a criterion of mental
retardation? American Journal of Mental Deficiency, 90, 236-244.
Flynn, J.R. (1987). Massive gains in 14 nations: What IQ tests really measure. Psychological
Bulletin, 101, 171-191.
Flynn, J.R. (2006).Tethering the elephant capital cases, IQ and the Flynn Effect. Psychology,
Public Policy and Law, 12, 170-189.
Flynn, J.R. (2007). What is Intelligence: Beyond the Flynn Effect. Cambridge: Cambridge
University Press.
Flynn, J.R. (2009). The WAIS-III and WAIS-IV: Daubert motions favor the certainly false
over the approximately true. Applied Neurology, 16, 98-104.
Flynn, R.J. and Weiss, L.G. (2007). American IQ gains from 1932 to 2002: The WISC
subtests and educational progress. International Journal of Testing, 7, 209-224.
Gordon, S., Duff, S. Davison, T and Whitaker, S. (2010). Comparison of the WAIS-III and
WISC-IV in 16 year old special education students. Journal of Applied Research in
Intellectual Disability. 23, 197-200.
Greenspan, S. and Switzky, H.N. (2006). Lessons from the Atkins decision for the next
AAMR manual. In H.N. Switzky, and S. Greenspan (Eds) What is Mental Retardation?
Ideas for an Evolving Disability in the 21st Century: Revised and Updated Edition.
Washington DC: American Association on Mental Retardation.
Gresham, F.M. (2009). Interpretation of intelligence test scores in Atkins cases: Conceptual
and psychometric issues. Applied Neuropsychology, 16, 91-97.
Heber, R. (1959). A manual on terminology and classification in mental retardation (rev.
ed.), Washington DC: American Association of Mental Deficiency.
Olley, J.G. (2009a). Challenges in implementing the Atkins decision. American Journal of
Forensic Psychology, 27, 63-73.
Olley, J.G. (2009b). Knowledge and experience required for experts in Atkins cases. Applied
Neuropsychology, 16, 135-140.
Salekin K.L. and Doane, B.M. (2009). Malingering intellectual disability: The value of
available measurements and methods. Applied Neuropsychology, 16, 105-113.
Schalock, R.L., Luckasson, R.A., Shogren, K.A., Borthwick-Duffy, S., Bradley, V., Buntinx,
W.H.E., Coulter, D.L., Craig, E. P. M., Gomez, S.C., Lachapelle, Y., Reeve, A., Snell,
M.E., Speat, S., Tasse M.J., Thompson, J.R., Verdugo, M.A., Wehmeyer, M.L. and
Yeager, M.H. (2007). The renaming of mental retardation: Understanding the change to
the term intellectual disabilities. Intellectual and Developmental Disabilities, 45, 116124.
Sparrow, S.S., Cicchetti, D.V. and Balla, D.A. (2005). Vineland-II Vineland Adaptive
Behavior Scale Second edition. Circle Pines: AGS Publishing.
127
Spitz, H.H. (1986). Disparity in mental retarded persons IQs derived from different
intelligence tests. American Journal of Mental Deficiency, 90, 588-591.
Spitz, H.H. (1989). Variations in the Wechsler interscale IQ disparities at different levels of
IQ. Intelligence, 13, 157-167.
Teasdale, T.W. and Owen, D.R. (2005). A long-term rise and recent decline in intellectual
test performance: the Flynn Effect in reverse. Personality and Individual Differences,
39, 837-843.
Wechsler, D. (1939). Wechsler-Bellevue Intelligence Scale. New York: The Psychological
Corporation.
Wechsler, D. (1992). WISC-III UK Administrative and scoring Manual. London: The
Psychological Corporation.
Wechsler, D. (1997a). WAIS-III, WMS-III: Technical and Interactive Manual. San Antonio,
Wechsler, D. (1997b). WAIS-III UK Administrative and scoring Manual. London: The
Psychological Corporation.
Wechsler, D. (2003a). Wechsler Intelligence Scale for Children Fourth Edition: Technical
and Interactive Manual. San Antonio, TX: The Psychological Corporation.
Wechsler, D. (2001). Wechsler Individual Achievement Test Second Edition. San Antonio,
Wechsler, D. (2003b). Wechsler Intelligence Scale for Children Fourth Edition:
Administrative and scoring Manual. San Antonio, TX: The Psychological Corporation.
Wechsler, D. (2008). Wechsler Adult Intelligence Scale Fourth Edition: Technical and
Interactive Manual. San Antonio, TX: The Psychological Corporation.
Whitaker, S. (2005). The use of the WISC-III and the WAIS-III with people with a learning
disability: Three concerns. Clinical Psychology, 50, 39-40.
Whitaker, S. (2008). The stability of IQ in people with low intellectual ability: An analysis of
the literature. Intellectual and Developmental Disabilities, 46, 120-128.
Whitaker, S. (2010). Error in the estimation of intellectual ability in the low range using the
WISC-IV and WAIS-III. Personality and Individual Differences 48, 517521.
Whitaker, S. and Gordon, S. (2012). Floor effects on the WISC-IV. Submitted to Journal of
Applied Research in Intellectual Journal of Developmental Disabilities, 57, 40-47.
Whitaker, S. and Wood, C. (2008). The distribution of scale score and possible floor effects
on the WISC-III and WAIS-III. Journal of Applied Research in Intellectual Disabilities.
21, 136-141.

ISBN: 978-1-62618-728-3
Chapter 8
U SING R FOR THE A NALYSIS OF C OGNITIVE

A BILITY AND B EHAVIOR G ENETIC D ATA
A. Alexander Beaujean*,1 and Jason Parkin,2
1
Baylor Psychometric Laboratory, Baylor University

Waco, Texas, US
2
Lake Washington School District
Redmond, Washington, US
Keywords: R, Cognitive Ability, Behavior Genetics, lavaan, Latent Variable Models
1.
The R Program
1.1.
Description
R (R Development Core Team, 2011) is a free, very flexible programming language, available for most operating systems (e.g., Windows, Mac, Linux). Thus, many professional
statisticians, university students and faculty, as well as businesses are turning to it for their
data analysis needs. The New York Times even published an article about Rs growing popularity (Vance, 2009).
R is currently maintained by the R core-development team (an international team of volunteer developers), and the R Projects web page is http://www.r-project.org (also known
as Comprehensive R Archive Network [CRAN]). This is the main site for R information,
directions for obtaining the software, accompanying packages, and some user documentation.
1.2.
Installing R
You can download R by going to http://www.r-project.org/. Then, on the left-hand side,

click on CRAN and then select a mirror from the USA (unless you want to use R it a
different language). This will bring you to a page with links to download R for Linux, Mac,
E-mail address: Alex Beaujean@baylor.edu

E-mail address: jrpzpf@mail.missouri.edu
130
and Windows. Be sure to download the latest version.

There are GUIs for R , developed by third parties.
A partial list can be
found at R Wiki (http://rwiki.sciviews.org/doku.php?id=guis:projects) and the R site (
http://www.sciviews.org/ rgui/). There are also many text editors that are either designed
to interact with R, or can be modified to do so. Typing in R text editor (or something
similar) into your favorite search engine will bring up many different text editor options, as
well as peoples opinons of the editors.
1.3.
Starting R
If you use the interactive mode for R (as opposed to batch mode), when you initially start
it, something similar to the following syntax will automatically appear.
1 R version 2.15.0 (2012-03-30)
2 Copyright (C) 2012 The R Foundation for Statistical Computing
3 ISBN 3-900051-07-0
4 Platform: x86_64-apple-darwin9.8.0/x86_64 (64-bit)
5
6 R is free software and comes with ABSOLUTELY NO WARRANTY.
7 You are welcome to redistribute it under certain conditions.
8 Type license() or licence() for distribution details.
9
10 Natural language support but running in an English locale
11
12 R is a collaborative project with many contributors.
13 Type contributors() for more information and
14 citation() on how to cite R or R packages in publications.
15
16 Type demo() for some demos, help() for on-line help, or
17 help.start() for an HTML browser interface to help.
18 Type q() to quit R.
19
20 >
The > is called the prompt. It is not typed (if you type it, R will assume you mean
greater than). Instead, it is used to indicate where you are to type. For interactive mode
uses of R , you will type in all your commands at the > prompt. If a command is too long
to fit on a single line, a + is used for the continuation prompt. Another symbol that you
will use frequently in R is the left arrow, <-, which is Rs standard assignment operator
(another option is using =, but it is better to reserve using = when your are defining values
for arguments). The <- is Rs way of assigning whatever is on the right of the arrow to the
object on the left of the arrow.
The easiest way to begin to use R is as a calculator. For example:
1 > 2+2
2 [1] 4
Using R for the Analysis of Cognitive Ability and Behavior Genetic Data
1.4.
131
Functions
R stores variables, data, functions, results, etc, in the computers active memory in the form
of named objects. The user can then do actions on these objects with operators (arithmetic,
logical, comparison) and functions (which are themselves objects). Much of R use is done
through functions that you can apply to data or other objects. R functions are a set of
instructions that take your input, compute the desired value(s), and return the result. R
comes pre-loaded with a set of commonly used functions, but you can add additional ones
by either loading packages with the desired functions, or by writing your own function.
To use functions: (a) give the functions name followed by parentheses; and (b) in the
parentheses, give the necessary values for the functions argument(s). Here is an example
of a function to calculate the arithmetic mean, named ArithMean().
1 > #Gives the arithmetic mean
2 > ArithMean<-function(x) {
3 + Sx<-sum(x) #the sum() function is native to R
4 + Mean<- Sx/length(x) #the length() function is native to R
5 + return(Mean)
6+ }
7 > ArithMean(c(5,10,15)) #the c() function is is defined later
8 [1] 10
First, we told R to define the function named ArithMean(), which only takes one
argument, x (see line 2). The left brace, {, indicates where the text of the function is going
to start and the right brace, }, is placed at the end of the function (line 6). After defining the
function, line 7 gives the syntax to evaluate one call to it. Since the sum of the numbers in
the vector c(5,10,15) is 30, the length of the vector (i.e., the number of elements) is 3,
and the call to the function returned the value 10.
In the ArithMean() function, x is the formal argument, whereas in the call to function c(5,10,15) is the actual argument. The formal argument is a placeholder, but
c(5,10,15) is the value used in the computation. Sometimes R functions have default
arguments, which are values that a functions arguments will automatically initialize unless
you specify a different value.
1.4.1.
Useful R Functions
There are some very helpful R functions that you will use repeatedly.
Comment. This is not really a function, but in R anything after the # sign is assumed
to be a comment and R will ignore it. Line 1 of the ArithMean() function was
a comment, which R ignored when reading that syntax. Comments are extremely
helpful functions, as annotating your R code can save you a lot of time and effort.
Concatenate. The concatenate function, c(), will concatenate (i.e, join) the arguments you include in the function. If you use the c() function in conjunction with
<-, you can assign whatever you are concatenating into a new object. For example, to
make a data set of 5 observations with the values 4, 5 3, 6, 9, and name it newData
you would use the following syntax.
132
1 > newData<-c(4,5,3,6,9)
Help. The help() function will obtain information about a function (or certain
special words or characters). You can also use a question mark, ?, as a shortcut for
the help function. For example, the following two lines of syntax will return the same
results.
1 > help(mean)
2
3 > ?mean
The help() function returns a page that (at least) describes the function (or word/character), its arguments, and gives some examples of how to use the function. Some pages
have much more detail than others. If you just want to examine/run the example syntax for
a function, you can use the example() function.
1 > example(mean)
2
3 mean> x <- c(0:10, 50)
4
5 mean> xm <- mean(x)
6
7 mean> c(xm, mean(x, trim = 0.10))
8 [1] 8.75 5.50
If you want help on an entire R package, then use the package argument in the help()
function.
1 help(package=lavaan) #you need to install the lavaan package first
If you do not know exactly you need to do within R, then you can search Rs documentation via the help.search() function. The argument you use in the function needs to
be enclosed in quotation marks. For example, if you are interested in testing to see if a
variable follows a normal distribution, you could use the following syntax.
1 > help.search("normality")
This produces a response that contains functions from packages that might be of interest. An
example of such output is shown in Figure 1, which indicates that in the stats package
(a package R installs by default) there is a function called shapiro.test() that will
perform the Shapiro-Wilk test for normality.
Topic
shapiro.test
Package
stats
Description
Shapiro-Wilk Normality Test
Figure 1. Example Results from help.search() Function.

Another useful way to get help is to use the Rseek web site (http://www.rseek.org/),
which is a site that uses Google to find R functions, lists, code, etc. If you are totally lost
on where to start asking for help, then typing help.start() into R will return much of
the important documentation needed to navigate R, as well as providing yet another search
engine for R helping materials.
1.5.
133
R Packages
Functions are a very important part of using R. When you first downloaded R, it came
with some base packages that supply functions for many statistical analysis [e.g., mean(),
var()]. These functions, however, may not do the specific analysis you specifically need
to do. Thus, you can look to see if a contributed R package can do what you need to do.
These R packages usually consist of data and functions that were written in the R language
(although sometimes they are written in FORTRAN or C and then linked back into R). This
user-contribution ability is extremely powerful, as there are many experts in various fields
using R, some of whom have contributed packages that can make your data analysis projects
much easier.
The list of R packages, with a short description of what they do, can be found in
CRAN, but it is very long and hard to navigate unless you know the specific name of
the package for which you are looking. An alternative is to examine the CRAN task
views http://cran.r-project.org/web/views/, which is designed to help users find packages
associated with specific types of work. For example, the Psychometrics http://cran.rproject.org/web/views/Psychometrics.html view has many packages dealing with item and
test analysis.
To install a package from the command line, use the install.packages() function, naming the package to install in quotation marks. For example, to install the BaylorEdPsych (Beaujean, 2012) package, use the following syntax.
1 install.packages("BaylorEdPsych", dep = TRUE)
The dep = TRUE argument tells R that in addition to the package of interest, also download any other package upon which the package of interest is dependent. (This saves you
from having to download each required package separately.)
You only need to install a package to the hard disk one time, but you will need to load
it into memory every time you start a new R session and need to use one of the packages
functions using the library() function, e.g.,
1 library(BaylorEdPsych)
(Notice there are no quotation marks around the package name.)

R is case sensitive, so Install.packages("BaylorEdPsych", dep
= TRUE),
install.packages("BaylorEdpsych", dep = TRUE),
or
install.Packages("BaylorEdPsych", dep = TRUE) (or any other permutation) will result in R returning an error message.
1.6.
1.6.1.
Inputting Data
Concatenate
The easiest way to enter data into R is to type it in directly using the concatenate, c()
function and assign a name to it. To verify that that your data is in your object (i.e., a
vector), just type the objects name.
1 > newData<-c(4,5,3,6,9)
2 > newData
134
3 [1] 4 5 3 6 9
Then you can immediately apply functions to the new object, newData
1 > mean(newData)
2 [1] 5.4
3 > sum(newData)
4 [1] 27
1.6.2.
Reading In Data from an External Source
Unless the data set has one variable with a few observations (e.g., data from a textbook
example), you will usually want to store your data in an external file and have R load the
data into its working memory.
read.table() One way to read externally-stored data is the read.table function.
Before doing this, however, it will be beneficial to do three things to your data. First, change
all missing values to NA, which is the default missing value indicator in R. Second, make
sure all the variable names are only one word (i.e, there are no spaces), but you can use
the . in lieu of a space (e.g., first.name). Third, either save the file as a tab-delimited text
(.txt) file or a comma-delimited .txt or .csv file. Most spreadsheet and database programs
can save data either way.
In order to read the file, you will need to point R to the directory where it is located.
In R you have to use a forward slash (how Mac and other UNIX-type systems store files)
or double backslash when giving a file location, e.g., C:\\Regression\\Data.csv.
Lets say you have a .csv file called data.csv that has your data (with a label on the
first row of each variable) located in a folder called name, which is in a folder called file.
You can type either of the following syntaxes into R to read the file.
1 # Windows
2 new.data<-read.table("C:\\file\\name\\data.csv", header=TRUE, sep=",")
3 new.data<-read.csv("C:\\file\\name\\data.csv", header=TRUE)
4 new.data<-read.csv("C:/file/name/data.csv")
5
6 # Unix-type systems
7 new.data<-read.table("/Users/first_last/file/name/data.csv", header=TRUE,
sep=",")
8 new.data<-read.csv("/Users/first_last/file/name/data.csv", header=TRUE)
Notice that the read.csv() function is the just like the read.table() function, only
it assumes you have comma-delimited values, so you do not have to use the sep=","
argument with it.
If you store your data under many sublayers of folders, or you forget the exact location,
you can search for your data file using the file.choose() function.
1 new.data<-read.table(file.choose(),header=T,sep=",")
1.6.3.
135
Inputting a Correlation/Covariance Matrix
In some situations you will not have access to raw data, but will have access to a covariance matrix. You could type the entire matrix into R using the matrix() function, but
since such matrices are symmetric, you can make use of the diag(), upper.tri()
and lower.tri() functions. By default, R will assume you are entering the data for the
matrix by columns. The following code will create a correlation matrix titled CorM that
consists of the correlations among four variables. In addition, give names to the variables
using the rownames() and colnames() functions.
1>
2>
3>
4>
CorM<-diag(4) #4 x 4 Diagonal matrix

CorMNames<-c("Var1", "Var2", "Var3", "Var4") #Names of the variables
rownames(CorM)<-colnames(CorM)<-CorMNames #Gives row and column names
CorM[lower.tri(CorM, diag=FALSE)]<-c(.85, .84, .68, .61, .59, .41) #
lower triangle of matrix, order is by columns
5 > CorM[upper.tri(CorM, diag=FALSE)] <- CorM[lower.tri(CorM)] #make matrix
full
6 > CorM
7
8
Var1 Var2 Var3 Var4
9 Var1 1.00 0.85 0.84 0.61
10 Var2 0.85 1.00 0.68 0.59
11 Var3 0.84 0.61 1.00 0.41
12 Var4 0.68 0.59 0.41 1.00
1.7.
The lavaan Package
lavaan (Rosseel, 2012) is an R package designed to do structural equation modeling. Information and documentation about it can be found on the web page: http://www.lavaan.org
which redirects to http://lavaan.ugent.be.
You can install lavaan from CRAN via the following R syntax
1 install.packages("lavaan", dependencies=TRUE)
To compute a model in lavaan, first you have to specify the model as text, using a few
pre-defined commands that are shown in Table 1.
Table 1. lavaan commands
Syntax
Command
Example
Regress onto
(Co)varaince
Constant/mean
Define latent variable
:=
Define non-model parameters
<
Define formative variables
Regress B onto A: B A
Variance of A: A A
Covariance of A and B: A B
Regress B onto A, and and include the intercept in the model:
B 1 + A
Define Factor 1 by A-D:
F1 = A+B+C+D
Define parameter u2 to be 2 times the square of u:
u2 := 2*(u 2)
Define Variable A by X1 X4 :
A < X1 + X2 + X3 + X4
Within the model, you can label parameters by premultiplying the label onto one of the
136
A
y
w
C
1
z
eC
CResid
eD
DResid
Figure 2. Path Model Example.

variables on the left hand side of an equation. For example, the syntax for specifying the
model in Figure 2 in lavaan would be as follows.
1 example.model<-
2 C y*B + w*A
3 D z*C + x*A
4 CC_Resid *C #Optional, included just to label residual variance
5 DD_Resid *D #Optional, included just to label residual variance
6
Line 1, consists of the name of the model (example.model), the assignment operator
(<-) and single apostrophe, which indicates that the subsequent syntax will be passed as
text. Line 2 defines the structural model for C, with labels for each parameter, designed
to map onto the labels in Figure 2; likewise, line 3 defines the structural model for D with
labels. Lines 4 and 5 are optional. By default, lavaan creates an error term for each
endogenous variable and estimates the variance while constraining the path to one. So, the
only thing lines 4 and 5 contribute is the labeling of these error variances. Likewise, by
default lavaan creates a covariance term for each exogenous variable (represented by the
two-headed arrow from A to B in Figure 2). Line 6 is another single apostrophe, indicating
the end of the syntax-as-text option.
You can fit the specified model by using either the cfa() (for factor analytic models),
or sem() (for structural equation models) functions. The cfa() and sem() functions are
really just the calls to a more general lavaan() function with some of the default values
specified differently. To obtain the default values for the cfa() or sem() functions, as
137
well as their various arguments, type ?cfa or ?sem, respectively, in R.

lavaan accepts either a covariance matrix (with sample size) or raw data as input,
using the sample.cov= (with sample.nobs=) or data= arguments, respectively. If
using the covariance matrix as input, you can optionally also input a mean vector with
the sample.mean= argument. By default, lavaan uses normal-theory maximum likelihood as the parameter estimation technique for continuous-variable indicators, but you can
change this to generalized least squares, weighted least squares (ADF), unweighted least
squares, or diagonally weighted least squares (for categorical indictors) by specifying them
in the estimator= argument.
For example, say data was collected from 500 individuals on variables A-D that
were specified in example.model, and the data has been inputted into R, naming it example.data. To estimate the parameters in example.model using
example.data, use the following lavaan syntax.
1 example.fit<-sem(example.model, data=example.data)
To input the correlation matrix of the data (named, say, example.cov) instead of the raw
data, use the following syntax
1 example.fit<-sem(example.model, sample.cov=example.cov, sample.nobs=500)
Both calls will return nothing on-screen because the the results are stored in the
example.fit object. To obtain the results, use the summary() function.
1 summary(example.fit)
By default, the summary() function for lavaan objects will produce (a) a note
indicating if the minimization algorithm converged; (b) the sample size; (c) estimator; (d) 2 ; (e) 2 degrees of freedom; (f) p-value of the 2 ; (g) the unstandardized parameter estimates; (h) their parameter estimates standard errors; (i) the ratio of the parameter estimates and their standard errors (i.e., Z); and (j) the ratios p-value. The summary() function has these default specifications for its arguments: standardized=FALSE, fit.measures=FALSE, rsquare=FALSE,
modindices=FALSE. Setting standardized=TRUE will include standardized estimates in the results, setting fit.measures=TRUE will include various fit indices in the
results, setting rsquare=TRUE will include the R2 for each exogenous variable, and setting modindices=TRUE will include fit indices.
For the many other functions the package includes, see the documentation for the
lavaan package here http://cran.r-project.org/web/packages/lavaan/lavaan.pdf
2.
2.1.
Latent Variable Models of Cognitive Ability

Single Factor Model
Much of the cognitive ability literature indicates that there is a single general (common)
intelligence factor (g) that influences most measures of cognitive ability (Jensen, 1998;
Spearman, 1927). Such a model, using subtests from the Wechsler Intelligence Scale for
Children-Fourth Edition (WISC-IV (Wechsler, 2003a)) is shown in Figure 3. This model
138
specifies that the only reason the subtests are correlated is because of g, and that once you
account for g, the subtests are no longer correlated. g does not explain all of the variance in
the subtests, however. Those residual/error functions accounting for non-g related variance
are represented as single-headed arrows not attached to the common factor.
g
Comprehension
Information
Similarities
Vocabulary
Matrix
Reasoning
Picture
Concepts
Digit
Span
Letter-number
Sequencing
Symbol
Search
Coding
Figure 3. Single factor model of cognitive ability from the WISC=IV.

Some sample correlations and standard deviations (SD) for the following WISC-IV
subtests are shown in Table 2 taken from (Parkin & Beaujean, 2012), p. 118: Comprehension, Information, Matrix Reasoning, Picture Concepts, Similarities, Vocabulary, Digit
Span, Letter Number Sequencing, Coding and Symbol Search.
Table 2. Correlations and standard deviations for WISC-IV subtests
1
2
3
4
5
6
7
8
9
10
Subtest
Comprehension
Information
Matrix Reasoning
Picture Concepts
Similarities
Vocabulary
Digit Span
Letter Number
Coding
Symbol Search
SD
1
1.00
0.62
0.34
0.33
0.63
0.71
0.41
0.44
0.27
0.34
2.88
10
1.00
0.51
0.37
0.73
0.74
0.42
0.50
0.30
0.43
3.01
1.00
0.39
0.48
0.46
0.37
0.43
0.28
0.41
2.89
1.00
0.38
0.37
0.31
0.38
0.27
0.30
2.98
1.00
0.74
0.43
0.50
0.23
0.38
3.03
1.00
0.45
0.52
0.29
0.39
3.02
1.00
0.51
0.30
0.37
2.98
1.00
0.32
0.45
2.99
1.00
0.49
2.96
1.00
3.12
Note. Correlations and Standard Deviations taken from (Parkin & Beaujean, 2012), p. 118.
To analyze this model in lavaan, we need to (a) input the correlation matrix and SD
vector; (b) convert the correlations to covariances (Cudeck, 1989); (c) specify the factor
model; (d) estimate the model parameters; and (e) assess the model to see how well it fits
the data.
First, enter the data (the correlation matrix and standard deviation vector). One option
when entering covariance/correlation matrices is to type the entire matrix into R using the
matrix() function, but a simpler option is to make use of the fact that such matrices
are symmetric and use the diagonal, upper and lower triangle, and transpose functions,
diag(), upper.tri(), lower.tri(), and t(), respectively. By default, R assumes
the data being entered for a matrix is by columns.
1 WiscIV.cor<-matrix(NA, nrow=10, ncol=10) #Create empty 10 x 10 matrix
2 diag(WiscIV.cor)<-1 #Place 1s on the diagonal of each matrix
3 #Create the lower triangle of the matrix
139
4 WiscIV.cor[lower.tri(WiscIV.cor)]<-c(0.62, 0.34, 0.33, 0.63, 0.71, 0.41,

0.44, 0.27, 0.34, 0.51, 0.37, 0.73, 0.74, 0.42, 0.50, 0.30, 0.43,
0.39, 0.48, 0.46, 0.37, 0.43, 0.28, 0.41, 0.38, 0.37, 0.31, 0.38,
0.27, 0.30, 0.74, 0.43, 0.50, 0.23, 0.38, 0.45, 0.52, 0.29, 0.39,
0.51, 0.30, 0.37, 0.32, 0.45, 0.49)
5 #Transpose the lower triangle to input the upper triangle of the matrix
6 WiscIV.cor[upper.tri(WiscIV.cor)]<-t(WiscIV.cor)[upper.tri(WiscIV.cor)]
7 #input the SDs
8 WiscIV.sd<-c(2.88,3.01,2.89,2.98,3.03,3.02,2.98,2.99,2.96,3.12)
Next, convert the correlations to covariances.

1 #make covaraice out of correlations
2 library(lavaan)
3 WiscIV.cov<-cor2cov(WiscIV.cor, WiscIV.sd)
4 #name the rows and columns of the covariance matrix
5 rownames(WiscIV.cov)<-colnames(WiscIV.cov)<-c("Comprehension", "
Information", "Matrix.Reasoning", "Picture.Concepts", "Similarities",
"Vocabulary", "Digit.Span", "Letter.Number", "Coding", "Symbol.
Search")
6 #To see the actual WiscIV.cov object
7 WiscIV.cov
The cor2cov() function is part of the lavaan package, so you need to load that, via
the library() function, before using it. The covariance matrices used for analysis in
lavaan need to have both row and column names or lavaan will not be able to match
the variable in the data with the variable in the model.
Next, specify the single factor model in lavaan.
1 WISC.oneFactor.model<-
2 g= Comprehension + Information + Matrix.Reasoning + Picture.Concepts +
Similarities + Vocabulary + Digit.Span + Letter.Number + Coding +
Symbol.Search
3
The model itself is enclosed in apostrophes (i.e., it is a text object). g is defined (=) using
all 10 WISC-IV subtests, which is specified by using the + sign before each additional
variable after the =.
To obtain the parameter estimates, first, estimate the parameters using the cfa()
function.
1 WISC.oneFactor.fit<-cfa(model=WISC.oneFactor.model, sample.cov=WiscIV.cov
, sample.nobs=550, std.lv=TRUE)
The first two arguments to the cfa() function are the model and the data (i.e., covariance
matrix), which we have already defined. The third argument, sample.nobs(), tells
lavaan the sample size from which the covariance matrix was calculated and needs to be
used anytime a covariance matrix is used for input. The last argument, std.lv=TRUE,
tells lavaan to estimate all the factor loadings (pattern coefficients) and constrain the
latent variables variance to unity. If the std.lv=TRUE option is not used, then the factor
variance would be estimated and the first indicator variables loading would be set to unity.
To print the parameter estimates, use the summary() function. Specifying the
standardized=TRUE argument in the summary() function produces the standardized
factor loadings (standardized regression weights) as well as the unstandardized loadings.
140
1 lavaan (0.5-9) converged normally after 28 iterations

2
3 Number of observations
550
4
5 Estimator
ML
6 Minimum Function Chi-square
276.092
7 Degrees of freedom
35
8 P-value
0.000
9
10 Parameter estimates:
11
12 Information
Expected
13 Standard Errors
Standard
14
15
Estimate Std.err Z-value P(>|z|)
16 Latent variables:
17 g =
18
Comprehension
2.159
0.107
20.118
0.000
19
Information
2.534
0.106
23.911
0.000
20
Matrix.Resnng
1.680
0.116
14.439
0.000
21
Pictur.Cncpts
1.408
0.124
11.328
0.000
22
Similarities
2.525
0.107
23.537
0.000
23
Vocabulary
2.615
0.105
25.003
0.000
24
Digit.Span
1.649
0.121
13.613
0.000
25
Letter.Number
1.896
0.118
16.093
0.000
26
Coding
1.127
0.126
8.917
0.000
27
Symbol.Search
1.616
0.128
12.590
0.000
28
29 Variances:
30
Comprehension
3.619
0.246
31
Information
2.624
0.202
32
Matrix.Resnng
5.515
0.348
33
Pictur.Cncpts
6.882
0.426
34
Similarities
2.789
0.211
35
Vocabulary
2.268
0.187
36
Digit.Span
6.144
0.385
37
Letter.Number
5.327
0.341
38
Coding
7.477
0.458
39
Symbol.Search
7.105
0.443
40
g
1.000
Std.lv
Std.all
2.159
2.534
1.680
1.408
2.525
2.615
1.649
1.896
1.127
1.616
0.750
0.843
0.582
0.473
0.834
0.867
0.554
0.635
0.381
0.518
3.619
2.624
5.515
6.882
2.789
2.268
6.144
5.327
7.477
7.105
1.000
0.437
0.290
0.662
0.776
0.304
0.249
0.693
0.597
0.855
0.731
1.000
The top part of the output can be used as a double-check to make sure the model was
specified correctly. For this model, n = 550 (correct!) and the df = 1011/2(10+10) =
5520 = 35 (correct!). The factor loadings in the Estimate column give the unstandardized loadings, while the loadings in the Std.all column give the standardized loadings.
Both indicate that each indicator is a relatively strong measure of g, with Vocabulary being
the most saturated (i.e., having the strongest loading) and Coding being the least.
To estimate the communality of each indicator (i.e., the amount of variance g explains),
either calculate the values manually by squaring each loading, or let R do the work by
extracting the loadings by using the inspect() function with the what="rsquare"
argument.
1 > inspect(object=WISC.oneFactor.fit, what="rsquare")
2 Comprehension
3 0.5628758
4
5 Similarities
6 0.6956403
7
8 Letter.Number
9 0.4030142
141
Information Matrix.Reasoning Picture.Concepts

0.7098617
0.3384281
0.2236483
Vocabulary
0.7508676
Digit.Span
0.3069146
Coding
0.1451074
Symbol.Search
0.2688097
From the output, you can see that g explains from 15-75% of the variance in the subtests.
Last, to obtain some measures of model fit, use the fitMeasures() function.
1 > fitMeasures(WISC.oneFactor.fit)
2 chisq
df
pvalue
baseline.chisq
3 276.092
35.000
0.000
2552.014
4
5 baseline.df
baseline.pvalue
6 45.000
0.000
7
8 cfi
tli
logl unrestricted.logl
9 0.904
0.876
-12676.518
-12538.472
10 npar
aic
11 20.000
25393.037
12
13 bic
ntotal
bic2
rmsea
14 25479.235
550.000
25415.747
0.112
15
16 rmsea.ci.lower
rmsea.ci.upper
17 0.100
0.124
18
19 rmsea.pvalue
srmr
srmr_nomean
20 0.000
0.067
0.067
While lavaan does not print out every fit statistic values, it does print enough information
to estimate any other fit statistics. For example, McDonalds (McDonald, 1989) noncentrality index (Mc), is calculated as:
M c = exp(0.5dT )
(1)
where
exp() is the exponentiation function, and
dT is the scaled NCP (noncentrality index) for the target model, i.e, dT =
This is calculated in R using the following syntax:
1 > NCP<-as.numeric(fitMeasures(WISC.oneFactor.fit,
- as.numeric(fitMeasures(WISC.oneFactor.fit,
Non-centrality Parameter
2 > d<- NCP/550 #scaled NCP
3 > Mc<-exp(-0.5 *d)
4 > Mc
5 [1] 0.8031815
2 df
n1 .
fit.measures="chisq"))
fit.measures="df")) #
where the as.numeric() function strips away everything in the object except the number.
142
2.2.
Multifactor Model
Many current scholars in cognitive ability research argue that there is more than a single
factor (general) factor influencing performance on cognitive ability tests (e.g. (Newton &
McGrew, 2010)). Instead, they posit that there are multiple factors, although these factors are related to each other. Such a multifactor model of the WISC-IV data is shown
in Figure 4, which posits that there are four factors: (a) Comprehension-Knowledge (Gc),
(b) Fluid Intelligence (Gf ), (c) Short-term Retrieval (Gsm), and (d) Processing Speed (Gs)
(cf. (Wechsler, 2003b)).
Gc
Comprehension
Gsm
Gf
Information
Similarities
Vocabulary
Matrix
Reasoning
Picture
Concepts
Digit
Span
Gs
Letter-number
Sequencing
Coding
Symbol
Search
Figure 4. Four Factor Model of Cognitive Ability from the WISC-IV.

Since this model uses the same data as the single factor model, the data entry component
is the same. The model specification in lavaan is as follow.
1 > WISC.fourFactor.model<-
2 Gc = Comprehension + Information + Similarities + Vocabulary
3 Gf = Matrix.Reasoning + Picture.Concepts
4 Gsm = Digit.Span + Letter.Number
5 Gs = Coding + Symbol.Search
6
To estimate the parameters and obtain the fit statistics, use the following syntax.
1 > WISC.fourFactor.fit<-cfa(model=WISC.fourFactor.model, sample.cov=WiscIV
.cov, sample.nobs=550, std.lv=TRUE)
2 > summary(WISC.fourFactor.fit, fit.measure=TRUE, standardized=TRUE)
3
5
550
7
8 Estimator
ML
51.634
29
11 P-value
0.006
12
13 Chi-square test baseline model:
14
2552.014
45
17 P-value
0.000
18
19 Full model versus baseline model:
20
21 Comparative Fit Index (CFI)
0.991
22 Tucker-Lewis Index (TLI)
0.986
23
24 Loglikelihood and Information Criteria:
25
26 Loglikelihood user model (H0)
-12564.289
27 Loglikelihood unrestricted model (H1)
-12538.472
28
29 Number of free parameters
26
30 Akaike (AIC)
25180.578
31 Bayesian (BIC)
25292.636
32 Sample-size adjusted Bayesian (BIC)
25210.101
33
34 Root Mean Square Error of Approximation:
35
36 RMSEA
0.038
37 90 Percent Confidence Interval
0.020 0.054
38 P-value RMSEA <= 0.05
0.885
39
40 Standardized Root Mean Square Residual:
41
42 SRMR
0.020
43
45
46 Information
Expected
47 Standard Errors
Standard
48
49
51 Gc =
52
Comprehension
2.192
0.107
20.526
0.000
53
Information
2.544
0.106
24.002
0.000
54
Similarities
2.558
0.107
23.963
0.000
55
Vocabulary
2.670
0.104
25.780
0.000
56 Gf =
57
Matrix.Resnng
1.999
0.136
14.698
0.000
58
Pictur.Cncpts
1.677
0.136
12.340
0.000
59 Gsm =
60
Digit.Span
1.968
0.126
15.589
0.000
61
Letter.Number
2.305
0.126
18.234
0.000
62 Gs =
63
Coding
1.778
0.136
13.060
0.000
64
Symbol.Search
2.541
0.152
16.740
0.000
65
66 Covariances:
67 Gc
68
Gf
0.779
0.041
18.896
0.000
69
Gsm
0.765
0.033
23.395
0.000
70
Gs
0.560
0.042
13.171
0.000
71 Gf
72
Gsm
0.824
0.050
16.405
0.000
73
Gs
0.705
0.054
12.940
0.000
143
Std.lv
Std.all
2.192
2.544
2.558
2.670
0.762
0.846
0.845
0.885
1.999
1.677
0.692
0.563
1.968
2.305
0.661
0.772
1.778
2.541
0.601
0.815
0.779
0.765
0.560
0.779
0.765
0.560
0.824
0.705
0.824
0.705
144
74 Gsm
75
Gs
76
77 Variances:
78
Comprehension
79
Information
80
Similarities
81
Vocabulary
82
Matrix.Resnng
83
Pictur.Cncpts
84
Digit.Span
85
Letter.Number
86
Coding
87
Symbol.Search
88
Gc
89
Gf
90
Gsm
91
Gs
0.707
0.047
3.474
2.573
2.621
1.977
4.340
6.052
4.991
3.611
5.585
3.261
1.000
1.000
1.000
1.000
0.240
0.204
0.207
0.181
0.424
0.434
0.377
0.381
0.428
0.574
15.199
0.000
0.707
0.707
3.474
2.573
2.621
1.977
4.340
6.052
4.991
3.611
5.585
3.261
1.000
1.000
1.000
1.000
0.420
0.284
0.286
0.217
0.521
0.683
0.563
0.405
0.639
0.336
1.000
1.000
1.000
1.000
Notice the fit.measure=TRUE argument in the summary() function. Specifying this

gives the fit indices along with the model parameter estimates, which is an alternative to
using the fitMeasures() function.
As the single factor model is nested in the four-factor model (i.e., the one-factor model
is a more restrictive version of the four factor model (Brunner, Nagy, & Wilhelm, 2012)),
the models can be compared directly using the anova() function.
1 Chi Square Difference Test
2
3
Df
AIC
BIC
Chisq Chisq diff Df diff Pr(>Chisq)
4 WISC.fourFactor.fit 29 25181 25293 51.634
5 WISC.oneFactor.fit 35 25393 25479 276.092
224.46
6 < 2.2e-16
***
6 --7 Signif. codes: 0 "***" 0.001 "**" 0.01 "*" 0.05 "." 0.1 " " 1
It appears that from both the change in 2 values and the other fit measures that the fourfactor model fits the data somewhat better than the single factor model.
2.3.
Higher-Order Model
The models considered thus far have only estimated factors that directly influence the subtests. An alternative to this kind of model is a higher-order model, which specifies that there
are factors that directly influence the subtests (i.e, first-order factors) as well as factors that
directly influence the factors (i.e., second-order factors) (Rindskopf & Rose, 1988). A typical higher-order model for cognitive abilities data is one that posits that g is the sole reason
why the the first-order factors are correlated with each other (Carroll, 1993). Such a model
is shown in Figure 5.
The specification of this model is similar to that from the single- and multi-factor models, except this model also requires specifying a factor made up of the four other factors.
1 WISC.higherOrder.model<-
2 gc = Comprehension + Information +
Similarities + Vocabulary
145
Gc
Comprehension
Gf
Information
Similarities
Vocabulary
Matrix
Reasoning
Gsm
Picture
Concepts
Digit
Span
Gs
Letter-number
Sequencing
Coding
Symbol
Search
Figure 5. Higher Order Factor Model of Cognitive Ability from the WISC-IV.
3 gf = Matrix.Reasoning + Picture.Concepts
4 gsm = Digit.Span + Letter.Number
5 gs = Coding + Symbol.Search
6
7 g= NA*gf + gc + gsm + gs
8 g1*g
9
Notice the NA* in front of the gf term in line 7. This tells lavaan to estimate this loading
instead of constraining it to unity (i.e., 1), which is the default. The trade-off for doing this
is that gs variance must be constrained to unity, which is done with the g1*g syntax in
line 8. This is done because it is better to estimate the (residual) variances of the first-order
factors instead of constrain them to be unity.
Next, estimate the parameters and obtain the results and fit statistics.
1 > WISC.higherOrder.fit<-cfa(model=WISC.higherOrder.model, sample.cov=
WiscIV.cov, sample.nobs=550)
2 > summary(WISC.higherOrder.fit, fit.measure=TRUE, standardized=TRUE)
4
550
6
7 Estimator
ML
57.592
31
10 P-value
0.003
11
13
2552.014
45
16 P-value
0.000
17
19
0.989
0.985
22
24
-12567.268
-12538.472
27
24
146
29 Akaike (AIC)
25182.537
30 Bayesian (BIC)
25285.975
25209.788
32
34
35 RMSEA
0.039
0.023 0.055
0.856
38
40
41 SRMR
0.023
42
44
45 Information
Expected
46 Standard Errors
Standard
47
48
50 gc =
51
Comprehension
1.000
52
Information
1.160
0.056
20.813
0.000
53
Similarities
1.165
0.056
20.753
0.000
54
Vocabulary
1.217
0.056
21.857
0.000
55 gf =
56
Matrix.Resnng
1.000
57
Pictur.Cncpts
0.847
0.079
10.754
0.000
58 gsm =
59
Digit.Span
1.000
60
Letter.Number
1.172
0.087
13.505
0.000
61 gs =
62
Coding
1.000
63
Symbol.Search
1.441
0.142
10.141
0.000
64 g =
65
gf
1.851
0.122
15.173
0.000
66
gc
1.794
0.112
16.023
0.000
67
gsm
1.822
0.130
14.070
0.000
68
gs
1.293
0.133
9.709
0.000
69
70 Variances:
71
g
1.000
72
Comprehension
3.467
0.240
73
Information
2.567
0.203
74
Similarities
2.634
0.208
75
Vocabulary
1.976
0.181
76
Matrix.Resnng
4.377
0.424
77
Pictur.Cncpts
6.026
0.435
78
Digit.Span
4.996
0.378
79
Letter.Number
3.606
0.382
80
Coding
5.610
0.430
81
Symbol.Search
3.210
0.584
82
gc
1.596
0.226
83
gf
0.533
0.340
Std.lv
Std.all
2.194
2.545
2.555
2.670
0.762
0.846
0.844
0.885
1.990
1.685
0.689
0.566
1.967
2.306
0.661
0.772
1.771
2.551
0.599
0.818
0.930
0.818
0.927
0.730
0.930
0.818
0.927
0.730
1.000
3.467
2.567
2.634
1.976
4.377
6.026
4.996
3.606
5.610
3.210
0.332
0.135
1.000
0.419
0.284
0.287
0.217
0.525
0.680
0.564
0.404
0.641
0.330
0.332
0.135
84
85
gsm
gs
0.547
1.464
0.241
0.263
0.141
0.467
147
0.141
0.467
Notice the lack of std.lv=TRUE argument in the cfa() function, which is equivalent to
including the argument std.lv=FALSE because that is the default value for the function.
To estimate the direct impact of g and the first-order factors on the WISC-IV subtests,
use Wrights (Wright, 1934, 1968) rules (cf. (Loehlin, 2004)). Specifically, the factor
loadings of the subtests on g are computed by multiplying the factor loading of each subtest
on the corresponding first-order factor by the factor loading of this first-order factor on g.
For example, the standardized loading of the Comprehension subtest score on g is computed
as .762 .818 = .623. Further, the loadings of the subtests on a specific factor can be
computed by multiplying the factor loading of each subtest on the corresponding first-order
factor by the standard deviation of the corresponding specific factor. For example, the
standardized loading of the Comprehension
subtest score on Gc is .762 .576 = .439 (the
[standardized] variance of Gc is .332 and .332 = .576).
This model fits the data very similarly to the four-factor model, although since the
higher-order model is more parsimonious (it estimates 24 parameters instead of the 26 parameters the four-factor model estimates), it is probably a better model for this data. One
other way to see how well this model fit the data is to inspect the residual correlations of
the first-order factors (i.e., the difference between the model-implied correlations among
the first-order constructs and the corresponding correlations in the first-order factor model
(McDonald, 2011)). The actual and implied correlations are given in Table 3. The residuals are given in Table 4, and range from -.04 to .03, which are likely not of much concern.
Another way to see how well the model fits is to examine the amount of variance in the firstorder factors explained by the second order factor (see Table 5). With this data, g explains
between 54 and 87% of the first order factors variances.
Table 3. Correlations among first-order factors (Actual correlations are in upper
triangle and implied correlations are in the lower triangle)
height
Gf
Gc
Gsm
Gs
Gf
1.00
0.76
0.86
0.68
Gc
0.78
1.00
0.76
0.60
Gsm
0.82
0.77
1.00
0.68
Gs
0.71
0.56
0.71
1.00
Table 4. Residual correlations of first-order factors
Gf
Gc
Gsm
Gs
Gf
Gc
Gsm
0.02
-0.04
0.03
0.01
-0.04
0.03
Gs
148

Table 5. Variances of first-order factors
Gc
Gf
Gsm
Gs
Observed
Variance
4.81
4.00
3.87
3.16
Residual
Variance
1.60
0.53
0.55
1.46
R2
0.67
0.87
0.86
0.54
Schmid-Leiman Transformation (Schmid & Leiman, 1957) developed a transformation of the higher-order factor model to yield uncorrelated first-order factors that represent
both the second-order and the first-order factors. This transformation of the factor loadings
makes them reflect the incremental influence of both general and specific abilities on the
indicator variable. As this procedure just transforms the higher order factor model, the fit
of both models will be identical (Yung, Thissen, & McLeod, 1999).
The Schmid-Leiman (S-L) transformation can be used to estimate the direct impact of
g and the first-order constructs on the subtests. You can calculate the factor loadings of the
manifest subtest scores on g and the first order factors using Wrights (Wright, 1934, 1968)
rules as was done in the previous section. This would be tedious to do by hand for each
indicator, but can be carried out efficiently using matrices following the steps outlined in
(Gorsuch, 1983) (see Figure 6 ).
Figure 6. Steps for Conducting the Schmid-Leiman Transformation.
149
The psych package (Revelle, 2012) has the schmid() function that will conduct the
S-L transformation in a EFA context (see also the bi-factor rotation in the GPArotation
package (Bernaards & Jennrich, 2005; Jennrich & Bentler, 2011)). The syntax below will
complete the S-L transformation using the output from the second-order CFA model.
1 #Residual variances
2 > U2<-diag(c(.332, .135, .141, .467))
3 > U<-sqrt(U2)
4 > rownames(U)<-colnames(U)<-c("gc","gf", "gsm", "gs")
5 > #First Order loadings
6 > Loadings<-matrix(0,ncol=4,nrow=10)
7 > colnames(Loadings)<-c("gc","gf", "gsm", "gs")
8 > rownames(Loadings)<-c("Comprehension", "Information", "Similarities", "
Vocabulary", "Matrix.Reasoning", "Picture.Concepts", "Digit.Span", "
Letter.Number", "Coding", "Symbol.Search")
9 > Loadings[1:4,1]<-c(0.762, 0.846, 0.844, 0.885) #Gc loadings
10 > Loadings[5:6,2]<-c(0.689, 0.566) #Gf loadings
11 > Loadings[7:8,3]<-c(0.661, 0.772) #Gsm loadings
12 > Loadings[9:10,4]<-c(0.599, 0.818) #Gs loadings
13 > #Second-order loadings
14 > Higher.Loadings<-matrix(c(.818,.930, .927, .730),ncol=1)
15 > colnames(Higher.Loadings)<-"g"
16 > rownames(Higher.Loadings)<-c("gc","gf", "gsm", "gs")
17 > A<-cbind(Higher.Loadings,U)
18 >
19 > #S-L transformed matrix
20 > Loadings%*%A
21
g
gc
gf
gsm
gs
22 Comprehension
0.623316 0.4390601 0.0000000 0.0000000 0.0000000
23 Information
0.692028 0.4874605 0.0000000 0.0000000 0.0000000
24 Similarities
0.690392 0.4863081 0.0000000 0.0000000 0.0000000
25 Vocabulary
0.723930 0.5099321 0.0000000 0.0000000 0.0000000
26 Matrix.Reasoning 0.640770 0.0000000 0.2531548 0.0000000 0.0000000
27 Picture.Concepts 0.526380 0.0000000 0.2079617 0.0000000 0.0000000
28 Digit.Span
0.612747 0.0000000 0.0000000 0.2482053 0.0000000
29 Letter.Number
0.715644 0.0000000 0.0000000 0.2898857 0.0000000
30 Coding
0.437270 0.0000000 0.0000000 0.0000000 0.4093410
31 Symbol.Search
0.597140 0.0000000 0.0000000 0.0000000 0.5589999
The S-L transformation orthogonalizes the relationship between the higher-order and
lower-order factors. That is, first the highest order factor solution is determined, then the
next highest order is determined based on the variance orthogonal to the highest order.
Moreover, the S-L factors are proportionality constrained. This constraint affects the proportion of variance in the subtest scores explained by general and specific ability constructs.
Specifically, for a given set of subtests, the ratios of variance attributable to the respective
first-order ability to variance attributable to g are constrained to be the same. For example, the standardized factor loadings on the first order factor for Comprehension is .762
and for Information is .846. The standardized factor loadings of these subtests on g are
.762 .818 = .623 and .846 .818 = .692, respectively. The variance ratio for the
2
.8462
= 1.50 and for Information is .692
Comprehension subtest is .762
2 = 1.50
.6232
150
2.4.
Hierarchical Model
An alternative to the higher-order model is a hierarchal model, which specifies that all the
factors are first-order factors, only some of these first order factors are more general than
others (Rindskopf & Rose, 1988). Sometimes hierarchical models are called a bi-factor
models or nested-factor models (Chen, West, & Sousa, 2006; Brunner et al., 2012; Mulaik
& Quartetti, 1997).
(Chen et al., 2006) write that hierarchal models should be investigated when (a) there is
a general factor that is hypothesized to account for the commonality of the items; (b) there
are multiple domain specific factors, each of which is hypothesized to account for the unique
influence of the specific domain over and above the general factor; or (c) interest is in the
domain specific factors as well as the common factor. They further argue that a hierarchal
model is often better than a higher-order model as hierarchical models (a) can be used to
study the role of domain specific factors that are independent of the general factor; (b) allow
for the direct examination of the strength of the relationship between the first-order factors
and their associated indicator variables via the factor loadings, whereas such relationships
cannot be directly examined in the second-order factor models. (c) can be useful in testing
whether a first-order factor predict external variables, over and above the general factor, as
the domain specific factors are directly represented as independent factors; and (d) allow
for the testing of measurement invariance of the domain specific factors, in addition to the
general factor, whereas the second-order model only allows for invariance in the secondorder factor to be directly tested for invariance because the first-order factors are represented
by disturbances.
An example of a hierarchical model is shown in Figure 7. Notice that g, Gc, Gf, Gsm,
and Gs are all first-order factors, but that g is uncorrelated with the other factors. Thus, this
model is similar to the S-L transformation, but here there are no proportionality constraints.
The syntax given below is for a hierarchical model with uncorrelated domain specific factors, but a more general model could allow them to be correlated (Rindskopf & Rose, 1988).
The specification of the hierarchical model in lavaan is
1 ## Hierarchical model
2 WISC.hierarchical.model<-
3 gc = Comprehension + Information + Similarities + Vocabulary
4 gf = a*Matrix.Reasoning + a*Picture.Concepts
5 gsm = b*Digit.Span + b*Letter.Number
6 gs = c*Coding + c*Symbol.Search
7 g= Comprehension + Information + Matrix.Reasoning + Picture.Concepts +
Symbol.Search
8 g0*gc + 0*gf + 0*gsm + 0*gs
9 gc 0*gf + 0*gsm + 0*gs
10 gf 0*gsm + 0*gs
11 gsm 0*gs
12
Lines 1-3 and 7 are very similar to the other models specified in this chapter. Lines
4-6 and lines 8-11 are different, though. Lines 4-6 specify that for Gs, Gsm, and Gs, both
indicator variables are pre-multiplied by the same label, which constrains the parameter
estimate to be the same (see Section 3. for more information on constriants). This has to be
151
Comprehension
Gc
Information
Similarities
Vocabulary
Matrix
Reasoning
Picture
Concepts
Gf
Digit
Span
Gsm
Letter-number
Sequencing
Coding
Symbol
Search
Gs
Figure 7. Hierarchical Model of the WISC-IV Subtests.

done for these latent variables or the model would be empirically underidentified (Kenny,
1979; Rindskopf, 1984).
The default in lavaan is for all exogenous variables to be correlated with each, so line
8 tells lavaan that g needs to be uncorrelated with Gc, Gf, Gsm and Gs, line 9 specifies
that Gc need to be uncorrelated with Gf, Gsm and Gs, and so on. Notice that, for each
variable, the parameter constraints are done through one statement instead of separate ones.
The left hand side of the double tilde specifies one of the variables and the multiple + signs
on the right hand side of the double tilde means that the the constraints are all going to
apply to a variables relationship with the variable of the left hand side.
The syntax to estimate the models parameters and calculate the fit indices is the same
as with the previous models, only replacing the model argument.
1 > WISC.hierarchical.fit<-cfa(model=WISC.hierarchical.model, sample.cov=
WiscIV.cov, sample.nobs=550, std.lv=TRUE)
2 > summary(WISC.hierarchical.fit, fit.measure=TRUE, standardized=TRUE)
4
550
6
7 Estimator
ML
50.345
28
10 P-value
0.006
11
13
2552.014
45
16 P-value
0.000
17
19
0.991
0.986
22
24
-12563.645
152

-12538.472
27
27
29 Akaike (AIC)
25181.290
30 Bayesian (BIC)
25297.657
25211.948
32
34
35 RMSEA
0.038
0.020 0.055
0.873
38
40
41 SRMR
0.022
42
44
45 Information
Expected
46 Standard Errors
Standard
47
48
50 gc =
51
Comprhnsn
1.424
0.131
10.873
0.000
52
Informatn
1.259
0.127
9.917
0.000
53
Similarts
1.374
0.128
10.724
0.000
54
Vocabulry
1.662
0.122
13.652
0.000
55 gf =
56
Mtrx.Rsnn (a)
0.635
0.221
2.874
0.004
57
Pctr.Cncp (a)
0.635
0.221
2.874
0.004
58 gsm =
59
Digit.Spn (b)
0.867
0.163
5.321
0.000
60
Lttr.Nmbr (b)
0.867
0.163
5.321
0.000
61 gs =
62
Coding
(c)
1.460
0.114
12.794
0.000
63
Symbl.Src (c)
1.460
0.114
12.794
0.000
64 g =
65
Comprhnsn
1.709
0.127
13.496
0.000
66
Informatn
2.190
0.124
17.674
0.000
67
Mtrx.Rsnn
1.882
0.121
15.502
0.000
68
Pctr.Cncp
1.568
0.132
11.898
0.000
69
Similarts
2.131
0.126
16.862
0.000
70
Vocabulry
2.135
0.126
16.961
0.000
71
Digit.Spn
1.792
0.129
13.844
0.000
72
Lttr.Nmbr
2.112
0.123
17.105
0.000
73
Coding
1.280
0.133
9.614
0.000
74
Symbl.Src
1.864
0.133
14.056
0.000
75
76 Covariances:
77 gc
78
g
0.000
79 gf
80
g
0.000
Std.lv
Std.all
1.424
1.259
1.374
1.662
0.495
0.419
0.454
0.551
0.635
0.635
0.220
0.213
0.867
0.867
0.291
0.290
1.460
1.460
0.494
0.468
1.709
2.190
1.882
1.568
2.131
2.135
1.792
2.112
1.280
1.864
0.594
0.728
0.652
0.527
0.704
0.708
0.602
0.707
0.433
0.598
0.000
0.000
0.000
0.000
81 gsm
82
g
83 gs
84
g
85 gc
86
gf
87
gsm
88
gs
89 gf
90
gsm
91
gs
92 gsm
93
gs
94
95 Variances:
96
Comprehension
97
Information
98
Similarities
99
Vocabulary
100
Matrix.Resnng
101
Pictur.Cncpts
102
Digit.Span
103
Letter.Number
104
Coding
105
Symbol.Search
106
gc
107
gf
108
gsm
109
gs
110
g
153
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
3.330
2.661
2.735
1.783
4.393
6.004
4.902
3.711
4.977
4.110
1.000
1.000
1.000
1.000
1.000
0.402
0.294
0.298
0.196
0.527
0.677
0.553
0.416
0.569
0.423
1.000
1.000
1.000
1.000
1.000
3.330
2.661
2.735
1.783
4.393
6.004
4.902
3.711
4.977
4.110
1.000
1.000
1.000
1.000
1.000
0.251
0.202
0.212
0.208
0.384
0.444
0.381
0.345
0.409
0.392
All of the fit indices indicate that this model fits the data fairly well, and arguably better
than the other models of this data.
One nice feature of the hierarchical model that the higher-order model does not have is
that it can be used to test if the domain specific factors (i.e., Gf, Gc, Gsm , Gs) explain any
additional variance above and beyond the general factor, (i.e, g) by removing the domain
specific factors from the model and examining how it fits the data. For example, the results
from the hierarchical model indicate that factor loadings for the Gf and Gsm are relatively
small (standardized loadings <0.30), so we can fit a model without these factors.
1>
2+
3+
4+
5+
WISC.hierarchical2.model<-
gc = Comprehension + Information +
gs = c*Coding + c*Symbol.Search
Similarities + Vocabulary
g= Comprehension + Information + Matrix.Reasoning + Picture.Concepts +

Symbol.Search
6 + g0*gc + 0*gs
7 + gc 0*gs
8+
9>
10 > WISC.hierarchical2.fit<-cfa(model=WISC.hierarchical2.model, sample.cov=
WiscIV.cov, sample.nobs=550, std.lv=TRUE)
11 > summary(WISC.hierarchical2.fit, fit.measure=TRUE, standardized=TRUE)
154

13
550
15
16 Estimator
ML
60.949
30
19 P-value
0.001
20
22
2552.014
45
25 P-value
0.000
26
28
0.988
0.981
31
33
-12568.947
-12538.472
36
25
38 Akaike (AIC)
25187.894
39 Bayesian (BIC)
25295.642
25216.281
41
43
44 RMSEA
0.043
0.027 0.059
0.744
47
49
50 SRMR
0.023
51
53
54 Information
Expected
55 Standard Errors
Standard
56
57
59 gc =
60
Comprhnsn
1.435
0.125
11.516
0.000
61
Informatn
1.337
0.119
11.228
0.000
62
Similarts
1.436
0.121
11.896
0.000
63
Vocabulry
1.694
0.115
14.734
0.000
64 gs =
65
Coding
(c)
1.460
0.113
12.905
0.000
66
Symbl.Src (c)
1.460
0.113
12.905
0.000
Std.lv
Std.all
1.435
1.337
1.436
1.694
0.499
0.445
0.474
0.561
1.460
1.460
0.494
0.468
67 g =
68
Comprhnsn
69
Informatn
70
Mtrx.Rsnn
71
Pctr.Cncp
72
Similarts
73
Vocabulry
74
Digit.Spn
75
Lttr.Nmbr
76
Coding
77
Symbl.Src
78
79 Covariances:
80 gc
81
g
82 gs
83
g
84 gc
85
gs
86
87 Variances:
88
Comprehension
89
Information
90
Similarities
91
Vocabulary
92
Coding
93
Symbol.Search
94
Matrix.Resnng
95
Pictur.Cncpts
96
Digit.Span
97
Letter.Number
98
gc
99
gs
100
g
1.695
2.140
1.873
1.592
2.092
2.106
1.878
2.178
1.285
1.857
1.695
2.140
1.873
1.592
2.092
2.106
1.878
2.178
1.285
1.857
0.589
0.712
0.649
0.535
0.691
0.698
0.631
0.729
0.435
0.596
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
3.346
2.677
2.727
1.799
4.963
4.139
4.828
6.331
5.336
4.179
1.000
1.000
1.000
0.404
0.296
0.298
0.198
0.567
0.426
0.579
0.714
0.602
0.468
1.000
1.000
1.000
3.346
2.677
2.727
1.799
4.963
4.139
4.828
6.331
5.336
4.179
1.000
1.000
1.000
0.124
0.122
0.119
0.128
0.125
0.124
0.124
0.119
0.132
0.132
0.248
0.204
0.213
0.204
0.408
0.391
0.341
0.415
0.371
0.328
13.625
17.472
15.699
12.414
16.780
17.009
15.161
18.248
9.738
14.106
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
155
While difference in square is moderately large (2df =1 = 10.60, p = .01), none of the
other measures of model fit are appreciably different between the two models. Moreover,
the change in R2 values for the four subtests affected by the removal of the two factors
(Matrix Reasoning, Picture Concepts, Digit Span, and Letter-Number Sequencing) only
change by .04-.05 units. Consequently, it appears that the covariance among the Gf and
Gsm subtests are mostly explained by g, alone.
3.
Behavior Genetic Analysis
Behavior genetics is the study of the genetic and environmental influences on psychological
traits (Plomin, DeFries, McClearn, & McGuffin, 2008). It can answer many questions that
traditional research designs cannot, such as the interaction of genotype with sex, age and
lifestyle factors, cultural transmission, genetic and environmental stability over time, and
the causes of co-morbidity between traits and diseases (Boomsma, Busjahn, & Peltonen,
2002). Traditionally, these analysis have used natural experiments to study these relationships (Rutter, 2007), such as twins reared in differing environments (Bouchard, 1991)
156
and adoption studies (Jensen, 1997), as well as the general study of sibling similarities and
differences (Jensen, 1980; Dunn & Plomin, 1992).
Initially, studies in this field used variations of regression models in conjunction with
the kinship of individuals within the sample to partition the phenotypic variance into genetic
and environmental components (DeFries & Fulker, 1985; Falconer, 1960), and such models
are still used (Lynch & Walsh, 1998; Falconer & Mackay, 1996). With the advent of more
powerful computers and data analysis programs, the field has moved to using latent variable
models for much of its analyses (Neale & Maes, 1992). That trend is especially evident
by the development of the the Mx program (Neale, Boker, Xie, & Maes, 2003) and its
translation into the R language via OpenMx (Boker et al., 2011).
The use of latent variable models for behavioral genetic data requires two extensions
from the models this chapter has covered thus far. First, the ability to estimate parameters
for different groups concurrently, and, second, the ability to constrain parameter estimates
within and between groups.
To input the data from multiple groups in lavaan, the data from the different groups
needs to be combined into a single list. In R, a list is a single objected made up of an
ordered (and possibly named) collection of objects (technically called components). As an
example, say we obtained Full Scale IQ (FSIQ) scores in a set of monozygotic (MZ) and
dizygotic (DZ) twins. The correlation of the scores between the MZ twins was .86, but
for the DZ the correlation was .60 ((Bouchard & McGue, 1981) p. 1056). To enter both
datasets in R as a list, first we need to enter the correlation matrices for the MZ and DZ
twins, separately.
1 > MZ.cor<-matrix(c(1,.86,.86,1),nrow=2)
2 > DZ.cor<-matrix(c(1,.60,.60,1),nrow=2)
Then, we need to name the rows and columns in each matrix

1>
2>
3>
4>
rownames(MZ.cor)<-c("P1",
colnames(MZ.cor)<-c("P1",
rownames(DZ.cor)<-c("P1",
colnames(DZ.cor)<-c("P1",
"P2")
"P2")
"P2")
"P2")
where P1 stands for the measured phenotype from twin 1, and P2 stands for the measured phenotype from twin 2.
Assuming the standard deviation is 15 and does not differ between groups (KoeppenSchomerus, Spinath, & Plomin, 2003), lavaans cor2cov() function can convert each
correlation matrix to a covariance matrix.
1 > DZ.cov<-cor2cov(DZ.cor, c(15,15))
2 > MZ.cov<-cor2cov(MZ.cor, c(15,15))
To combine the two matrices into a list, use the list() function.
1 sib.cov<-list(MZ=MZ.cov,DZ=DZ.cov)
In the previous syntax, the list components were entered in the form of name=object. Giving the list elements names will help in parameter interpretation, as lavaan will use the
names in the summary() output.
To examine a list in the R console, type the lists name.
157
1 > sib.cov
2 $MZ
3
P1
P2
4 P1 225.0 193.5
5 P2 193.5 225.0
6
7 $DZ
8
P1 P2
9 P1 225 135
10 P2 135 225
The same syntax can be used to enter the twin groups sample sizes, although the data can
be entered as scalars (i.e., one-dimensional matrices) instead of a matrix.
1 > sib.n<-list(MZ=4672,DZ=5546)
2
3 > sib.n
4 $MZ
5 [1] 4672
6
7 $DZ
8 [1] 5546
Because the covariances from both groups are combined into a single list object, the
lavaan syntax to estimate the parameters the will look the same as with the other CFA
models, e.g., cfa(model, sample.cov=sib.cor, sample.nobs=sib.n).
The only thing left to do is to specify the model. In order to do so, though, certain parameters need constrained.
Constraining parameters in lavaan can take one of two forms. First, to constrain
a parameter to a specific value, just premultiply the parameter by that value, e.g., g
=3*Comprehension. To constrain a parameter estimate to be the same for m > 1
groups, then the constrained parameter needs to have the same label in all the groups. This
can be done in one of two ways. The first way is by specifying m separate models and
giving the parameters to be constrained the same label. The second way is by specifying
one model, but using Rs concatenate function, c(), to apply multiple labels, e.g., for two
groups: g = c(a,a)*Comprehension.
3.1.
3.1.1.
Behavior Genetic Models

Twin Models
Behavior genetic models can be quite complex, depending the degree(s) of kinship present
in the data, as well as the number of phenotypes (i.e., observable behavior) measured. For
the classic model though, we will assume there is just one phenotype measured in both
MZ and DZ twins. (For some multivariate extensions of this model, see, e.g. (Loehlin,
2004; Neale et al., 2003)). The genetic influence on the phenotype can be decomposed
into (a) additive effects of alleles at various loci, (b) dominance effects of alleles at various
loci, and (c) epistatic interactions between loci (Plomin et al., 2008). Often with human
samples, epistatic and dominance effects are confounded, so are lumped into a single nonadditive genetic effects category. The environmental influence on the phenotype can be
158
decomposed into (a) effects due to a shared environment, such as being raised by the same
parents in the same house (aka between-family effects); and (b) effects due to an unshared
environment, such as having different peers or attending different schools (aka withinfamily effects). These unshared effects also include random environmental events, such as
getting into automobile accident, as well as random measurement events (i.e, measurement
error).
For relatives i and j, their phenotypes, Pi and Pj , are assumed to be a linear function
of the additive genetic influence (Ai and Aj ), non-additive influence (Di and Dj ), shared
environmental influence (Ci and Cj ) and unshared environmental variance (Ei and Ej ).
Thus,
P1 = a1 A1 + d1 D1 + c1 C1 + e1 E1
P2 = a2 A2 + d2 D2 + c2 C2 + e2 E2
(2)
For a given set of twins, the influences of A, C, D, and E are not expected to differ.
That is, we would not expect, say, the additive genetic influence estimates (a1 and a2 ) or
the unshared environmental influence estimates (e1 and e2 ) to differ. Thus, Equation 2 can
be simplified to
P1 = aA1 + dD1 + cC1 + eE1
P2 = aA2 + dD2 + cC2 + eE2
(3)
If twins (no matter the zygosity) are reared together, the shared environment influence
(C) is going to be the same. Likewise, by its definition, the non-shared environment (E) is
going to be completely different. From quantitative genetic theory, we know that for MZ
twins, the genetic influence (i.e., A and D) on a trait will be the same for both twins. For
DZ twins, on average, the additive genetic influence will only be 1/2 the same and the nonadditive genetic influence will be be 1/4 the same.1 From this information, we can build a
path model showing the relationship on a given phenotype between two twin siblings, as is
shown in Figure 8.
Using the labels in Figure 8, the expected correlations on the measured phenotype for
MZ and DZ twins reared together are
= a2 + d2 + c2
(4)
= 0.5a2 + 0.25d2 + c2
(5)
M Z rP1 ,P2
and
DZ rP1 ,P2
respectively. The variance for the trait (for either sibling) is

P2 = a2 + d2 + c2 + e2
(6)
Together, equations 4, 5, and 6 represent four unknown parameters (a, b, c and d), but
use input from only three known statistics (M Z rP1 ,P2 , DZ rP1 ,P2 , and P2 ). Thus, a classic twin design can only estimate three of the four parameters (Martin, Eaves, Kearsey, &
1
This assumes random mating and that the additive genetic influence is uncorrelated with shared environmental or dominance influence.
159
MZ=1/
DZ=.5
A1
E1
A2
P1
P2
E2
D1
D2
MZ=1/
DZ=.25
Figure 8. ACDE Model.

Davies, 1978). To estimate all the parameters within the same model would require additional data (e.g., twins separated at birth, relatives of twins (Neale & Maes, 1992)). Not all
the parameters need to be estimated, however, if, say, the hypothesis is that only additive
genetic components and random environmental events are influencing the phenotype. In
fact, with twin designs it is typical to test the the following series of models that postulate
different genetic and environmental components are influencing behavior: ACE, ADE, AE,
CE and E (Maes, 2005).
Below is the lavaan syntax for the full ACDE model for a single measured variable,
using MZ and DZ twins raised together. The model assumes the MZ twins are the first set
of data entered into the R list. To estimate other models, just comment out the unnecessary
parameters.
1 ACDE.model<-
2 #Latent Variables
3 A1= NA*P1 + c(a,a)*P1
4 A2= NA*P2 + c(a,a)*P2
5 C= NA*P1 + c(c,c)*P1 + c(c,c)*P2
6 D1 = NA*P1 + c(d,d)*P1
7 D2 = NA*P2 + c(d,d)*P2
8
9 #Variances
10 A1 1*A1
11 A2 1*A2
12 C 1*C
13 D1 1*D1
160
14 D2 1*D2
15 P1c(e2,e2)*P1
16 P2c(e2,e2)*P2
17
18 #covariances
19 A1 c(1,.5) *A2 + 0*C + 0*D1 + 0*D2
20 A2 0*C + 0*D1 + 0*D2
21 C 0*D1 + 0*D2
22 D1 c(1,.25) *D2
23
Lines 3-6 define the latent variables A1, A2, C, D1, and D2 respectively. The c(a,a)*P1
syntax tells lavaan to use the label a for the parameter estimating the influence from A1
to P1 in both the MZ and DZ groups. As the parameter estimating the influence from A2
to P2 has the same label, lavaan will constrain it to be the same value as well. A similar
interpretation follows for the other c()* terms.
This model does not create a E1 or E2 latent variable because, at least for a design with
MZ and DZ twins raised together, these terms represent error/residual variance. Consequently, this model constrains the error variances to be the same for for the MZ and DZ
twins (meaninge2 is estimated instead of e lines 15-16). The estimate for e can be obtained via e = e2 .
Lines 10-14 constrain the variances of A1, A2, and C to unity, while lines 19-22 constrain the correlation of both A variables with C and both D variables to zero. The correlation of all residual terms in lavaan are already constrained to zero; thus, there is no need
to specify anything about E1 and E2s covariance. Lines 19 and 22 serve a dual purpose.
In addition to making correlations equal zero, line 19 also constrains the between-twin A
correlations to be 1 in the first group (MZ twins) and .5 in the second group (DZ twins).
Likewise, line 22 constrains the between-twin D correlations to be 1 for MZ twins and .25
in DZ twins.
The results for the AE model are given below.
1 > AE.fit<-cfa(AE.model, sample.cov=sib.cov, sample.nobs=sib.n)
2 > summary(AE.fit, fit.measures=TRUE, standardized=TRUE)
4
5 Number of observations per group
6 MZ
4672
7 DZ
5546
8
9 Estimator
ML
321.458
4
12 P-value
0.000
13
14 Chi-square for each group:
15
16 MZ
11.971
17 DZ
309.487
18
20
8761.454
2
23 P-value
0.000
24
26
0.964
0.982
29
31
-80117.143
-79956.414
34
2
36 Akaike (AIC)
160238.287
37 Bayesian (BIC)
160252.751
160246.395
39
41
42 RMSEA
0.125
0.113 0.136
0.000
45
47
48 SRMR
0.085
49
51
52 Information
Expected
53 Standard Errors
Standard
54
55 Group 1 [MZ]:
56
57
59 A1 =
60
P1
(a)
13.570
0.091 148.344
0.000
61 A2 =
62
P2
(a)
13.570
0.091 148.344
0.000
63
64 Covariances:
65 A1
66
A2
1.000
67
68 Variances:
69
A1
1.000
70
A2
1.000
71
P1
(e2)
29.930
0.611
72
P2
(e2)
29.930
0.611
73
74
75
76 Group 2 [DZ]:
161
Std.lv
Std.all
13.570
0.927
13.570
0.927
1.000
1.000
1.000
1.000
29.930
29.930
1.000
1.000
0.140
0.140
162
77
78
80 A1 =
81
P1
(a)
82 A2 =
83
P2
(a)
84
85 Covariances:
86 A1
87
A2
88
89 Variances:
90
A1
91
A2
92
P1
(e2)
93
P2
(e2)
Estimate
Std.err
Z-value
P(>|z|)
Std.lv
Std.all
13.570
0.091
148.344
0.000
13.570
0.927
13.570
0.091
148.344
0.000
13.570
0.927
0.500
0.500
0.500
1.000
1.000
29.930
29.930
1.000
1.000
29.930
29.930
1.000
1.000
0.140
0.140
0.611
0.611
The results look a little different than previous output. First, the models parameter estimates are given in two separate sections, one for Group 1 and the other for Group 2. These
groups are labeled MZ and DZ, respectively, because that is what we named them in the
list() function. To get percent of variance in cognitive ability explained by additive
genetic influences from AE model (i.e., the narrow heritability), square the standardized a
term (i.e., the term in the Std.all column): a2 = .9272 .86. Likewise, the amount of
variance explained by the unshared environment is e2 = .14. These values should sum up
to one, as they do in this model.
Table 6 has the fit values for all the models. The AE and CE models did not appear
to fit the data very well, and the E model (i.e., the null model), fit the data horribly. The
ADE model shows the exact same fit as the AE model, which normally would be odd, but
can be explained by examining the estimated value for d: 0.00. The fact that there was no
dominance influence is not surprising, given that 1/2 of the MZ correlation ( .86
2 = .43) is
less than the DZ twin correlation (.60) (Evans, Gillespie, & Martin, 2002). Thus, it appears
that the ACE model is the best fitting model of the five, and in fact the ACE model fit
the data extremely well, as its estimated values for a, c and e were able to re-create the
covariances perfectly.
Table 6. Goodness-of-fit statistics and parameter estimates for MZ and DZ twin
models of cognitive ability
Model
ACE
AE
CE
E
ADE
df
SRMR
RMSEA
a2
c2
e2
d2
< 0.00a
321.46
1329.95
8761.45
321.46
3
4
4
5
3
0.00
0.09
0.07
0.42
0.09
0.00
0.13
0.26
0.59
0.14
0.52
0.86
0.86
0.34
0.72
0.14
0.14
0.28
1.00
0.14
0.00
Note. SRMR: standardized root mean square error; RMSEA: root mean square error of approximation.
a 2
= 0.000001843065.
3.1.2.
163
Non-Twin Sibling Models
While twin data can be very powerful, it can be difficult to collect such data. Collecting
sibling data, however, is much easier, and can also give interesting information about environmental influences (Plomin, 1994; Jensen, 1980), specifically the power of the shared
and non-shared environment for a given trait (Turkheimer & Waldron, 2000).
A non-twin sibling model for a single phenotype is shown in Figure 9. In such analyses,
the goal is to examine the influence of shared versus non-shared environmental influences,
so there are only two parameters estimated: e and c. As with the twin models, e2 is the
amount of variance explained by the non-shared environment, while c2 is the amount of
variance explained by the shared environment. For an example of a multivariate non-twin
sibling analysis, see, e.g., (Kretschmer & Pike, 2010).
c
P1
e
E1
c
P2
e
E2
Figure 9. Univariate Sibling Analysis.

Some example data come from (Foong, Beaujean, Frisby, & Knoop, 2012), who gathered cognitive ability information on a group of siblings (n = 17) in middle school using
the Woodcock Johnson-Third Edition Tests of Cognitive Abilities (Woodcock, McGrew, &
Mather, 2001). For this data, the siblings were spilt by age into into older (P1) and younger
(P2) sibling categories. The lavaan syntax to input and name the covariance matrix is
given below.
1 #Sibling data
2 BIA.cov<-matrix(c(293.4706, 119.1287, 119.1287, 220.7574),ncol=2)
3 colnames(BIA.cov)<-rownames(BIA.cov)<-c("P1", "P2")
Specifying the model is very similar to the ACDE type models, only here we are just
specifying the C and E components of that model.
1 sibling.model<-
2 Shared = NA*P1 + c*P1 + c*P2
3 P1e2*P1
4 P2e2*P2
5 Shared1*Shared
6
The results from this model are given below.
164
1 sibling.fit<-cfa(sibling.model, sample.cov=BIA.cov, sample.nobs=17)

2 summary(sibling.fit, standardized=TRUE)
4
17
6
7 Estimator
ML
0.438
1
10 P-value
0.508
11
13
14 Information
Expected
15 Standard Errors
Standard
16
17
Std.lv Std.all
19 Shared =
20
P1
(c)
10.589
3.054
3.467
0.001
10.589
0.681
21
P2
(c)
10.589
3.054
3.467
0.001
10.589
0.681
22
23 Variances:
24
P1
(e2)
129.869
44.545
129.869
0.537
25
P2
(e2)
129.869
44.545
129.869
0.537
26
Shared
1.000
1.000
1.000
The results indicate that the model fit the data relatively well (2 = .438). According
to this model, the shared environment explains approximately 46% of the variance (i.e.,
.6812 = .46), while the unique environment explains approximately 54% of the variance.
4.
Conclusion
This chapter has shown how R (R Development Core Team, 2011) can be a powerful tool
in the analysis of latent variable models of cognitive ability data, specifically focussing on
the lavaan (Rosseel, 2012) package, which is designed especially for structural equation
modeling with latent variables. With such tools, the analysis of latent variables models
can be accessible to any who are interested in such models. When added to fact that the
only cost involved in using R is the time it takes to learn the language, it is easy to see why
(Kelley, Lai, & Wu, 2008) write that there is no time like the present to begin incorporating
R into ones set of statistical tools (p. 569).
References
Beaujean, AA. 2012. BaylorEdPsych: R package for Baylor University Educational
Psychology quantitative courses (Version 0.5) BaylorEdPsych: R package for Baylor University Educational Psychology quantitative courses (Version 0.5) [Computer
software]. Waco, TXBaylor University.
165
Bernaards, CA., & Jennrich, RI. 2005. Gradient projection algorithms and software for
arbitrary rotation criteria in factor analysis Gradient projection algorithms and software for arbitrary rotation criteria in factor analysis. Educational and Psychological
Measurement655676-696. 10.1177/0013164404272507
Boker, S., Neale, M., Maes, H., Wilde, M., Spiegel, M., Brick, T., Spies, J., Estabrook, R.,
Kenny, S., Bates, T., Mehta, P., & Fox, J. 2011. OpenMx: An open source extended
structural equation modeling framework OpenMx: An open source extended structural equation modeling framework. Psychometrika762306-317. 10.1007/s11336010-9200-6
Boomsma, D., Busjahn, A., & Peltonen, L. 2002. Classical twin studies and beyond Classical twin studies and beyond. Nature Reviews Genetics311872-882. 10.1038/nrg932
Bouchard, TJ. 1991. A twice-told tale: Twins reared apart A twice-told tale: Twins reared
apart. In D. Cicchetti & WM. Grove (Eds.), Thinking clearly about psychology:
Essays in honor of Paul E. Meehl, Vol. 1: Matters of public interest Vol. 2: Personality
and psychopathology. Thinking clearly about psychology: Essays in honor of Paul E.
Meehl, Vol. 1: Matters of public interest Vol. 2: Personality and psychopathology.
(p. 188-215). Minneapolis, MNUniversity of Minnesota Press.
Bouchard, TJ., & McGue, M.
1981.
Familial studies of intelligence: A review Familial studies of intelligence: A review. Science21244981055-1059.
10.1126/science.7195071
Brunner, M., Nagy, G., & Wilhelm, O. 2012. A tutorial on hierarchically structured constructs A tutorial on hierarchically structured constructs. Journal of
Personality804796-846. 10.1111/j.1467-6494.2011.00749.x
Carroll, JB. 1993. Human cognitive abilities: A survey of factor-analytic studies Human
cognitive abilities: A survey of factor-analytic studies. New YorkCambridge University Press.
Chen, FF., West, SG., & Sousa, KH. 2006. A comparison of bifactor and second-order
models of quality of life A comparison of bifactor and second-order models of quality
of life. Multivariate Behavioral Research412189-225. 10.1207/s15327906mbr41025
Cudeck, R. 1989. Analysis of correlation matrices using covariance structure models
Analysis of correlation matrices using covariance structure models. Psychological
Bulletin1052317-327. 10.1037/0033-2909.105.2.317
DeFries, JC., & Fulker, DW. 1985. Multiple regression analysis of twin data Multiple regression analysis of twin data. Behavior Genetics155467-473. 10.1007/bf01066239
Dunn, J., & Plomin, R. 1992. Separate lives: Why siblings sre so different Separate lives:
Why siblings sre so different. New YorkWiley.
Evans, DM., Gillespie, NA., & Martin, NG. 2002. Biometrical genetics Biometrical genetics. Biological Psychology611-233-51. 10.1016/s0301-0511(02)00051-0
Falconer, DS. 1960. Introduction to quantitative genetics Introduction to quantitative genetics. LondonOliver and Boyd.
Falconer, DS., & Mackay, TFC. 1996. Introduction to quantitative genetics Introduction to
quantitative genetics (4th ed.). Harlow, Essex, UKAddison Wesley Longman.
Foong, C., Beaujean, AA., Frisby, CL., & Knoop, A. 2012. Environmental influence on
mathematical reaction time: A sibling study. Manuscript submitted for publication
Environmental influence on mathematical reaction time: A sibling study. Manuscript
166
submitted for publication.

Gorsuch, RL. 1983. Factor analysis Factor analysis. Hillsdale, NJLawrence Erlbaum.
Jennrich, R., & Bentler, P. 2011. Exploratory bi-factor analysis Exploratory bi-factor analysis. Psychometrika764537-549. 10.1007/s11336-011-9218-4
Jensen, AR. 1980. Uses of sibling data in educational and psychological research Uses
of sibling data in educational and psychological research. American Educational
Research Journal172153-170. 10.2307/1162480
Jensen, AR. 1997. Adoption data and two g-related hypotheses Adoption data and two
g-related hypotheses. Intelligence2511-6. 10.1016/s0160-2896(97)90003-9
Jensen, AR. 1998. The g factor: The science of mental ability The g factor: The science of
mental ability. Westport, CTPraeger Publishers/Greenwood Publishing Group. Print
Kelley, K., Lai, K., & Wu, PJ. 2008. Using R for data analysis: A best practice for research
Using R for data analysis: A best practice for research. In JW. Osborne (Ed.), Best
practices in quantitative methods Best practices in quantitative methods (p. 535-572).
Thousand Oaks, CASage.
Kenny, DA. 1979. Correlation and causality Correlation and causality. New YorkJohn
Wiley.
Koeppen-Schomerus, G., Spinath, FM., & Plomin, R. 2003. Twins and non-twin
siblings: Different estimates of shared environmental influence in early childhood Twins and non-twin siblings: Different estimates of shared environmental influence in early childhood. Twin Research and Human Genetics60297-105.
doi:10.1375/twin.6.2.97
Kretschmer, T., & Pike, A. 2010. Links between nonshared friendship experiences
and adolescent siblings differences in aspirations Links between nonshared friendship experiences and adolescent siblings differences in aspirations. Journal of
Adolescence331101-110. 10.1016/j.adolescence.2009.05.001
Loehlin, JC. 2004. Latent variable models: An introduction to factor, path, and structural equation analysis Latent variable models: An introduction to factor, path, and
structural equation analysis (4th ed.). Mahwah, NJErlbaum.
Lynch, M., & Walsh, B. 1998. Genetics and analysis of quantitative traits Genetics and
analysis of quantitative traits. Sunderland, MASinauer.
Maes, HH. 2005. ACE Model ACE Model. In BS. Everitt & DC. Howell (Eds.), Encyclopedia of Statistics in Behavioral Science. Encyclopedia of Statistics in Behavioral
Science. John Wiley and Sons, Ltd.
Martin, NG., Eaves, LJ., Kearsey, MJ., & Davies, P. 1978. The power of the classical twin study The power of the classical twin study. Heredity40197-116.
10.1038/hdy.1978.10
McDonald, RP. 1989. An index of goodness-of-fit based on noncentrality An index of goodness-of-fit based on noncentrality. Journal of Classification6197-103.
10.1007/bf01908590
McDonald, RP.
2011.
Measuring latent quantities Measuring latent quantities.
Psychometrika764511-536. 10.1007/s11336-011-9223-7
Mulaik, SA., & Quartetti, DA. 1997. First order or higher order general factor? First order
or higher order general factor? Structural Equation Modeling: A Multidisciplinary
Journal43193-211. 10.1080/10705519709540071
167
Neale, MC., Boker, SM., Xie, G., & Maes, HH. 2003. Mx: Statistical modeling Mx:
Statistical modeling (6th ed.). Richmond, VAVirginia Commonwealth University.
http://www.vipbg.vcu.edu/ vipbg/software/mxmanual.pdf
Neale, MC., & Maes, HHM. 1992. Methodology for genetic studies of twins and families
Methodology for genetic studies of twins and families. Dordrecht, The NetherlandsKluwer Academic Publishers.
Newton, JH., & McGrew, KS. 2010. Introduction to the special issue: Current research in
Cattell-Horn-Carroll-based assessment Introduction to the special issue: Current research in Cattell-Horn-Carroll-based assessment. Psychology in the Schools477621634. 10.1002/pits.20495
Parkin, J., & Beaujean, AA. 2012. The effects of Wechsler Intelligence Scale for ChildrenFourth Edition cognitive abilities on math achievement The effects of Wechsler Intelligence Scale for Children-Fourth Edition cognitive abilities on math achievement.
Journal of School Psychology50113-128. 10.1016/j.jsp.2011.08.003
Plomin, R. 1994. The Emanuel Miller Memorial Lecture 1993: Genetic research and identification of environmental influences The Emanuel Miller Memorial Lecture 1993:
Genetic research and identification of environmental influences. Journal of Child
Psychology and Psychiatry355817-834. 10.1111/j.1469-7610.1994.tb02297.x
Plomin, R., DeFries, JC., McClearn, GE., & McGuffin, P. 2008. Behavioral genetics
Behavioral genetics (5th ed.). New YorkWorth.
R Development Core Team. 2011. R: A language and environment for statistical computing R: A language and environment for statistical computing. Vienna, AustriaR
Foundation for Statistical Computing. Available from http://www.R-project.org.
Revelle, W. 2012. psych: Procedures for psychological, psychometric, and personality
research (Version 1.2.4) [Computer software]. psych: Procedures for psychological, psychometric, and personality research (Version 1.2.4) [Computer software].
Evanston, ILNorthwestern University.
Rindskopf, D. 1984. Structural equation models: Empirical identification, Heywood
cases, and related problems Structural equation models: Empirical identification,
Heywood cases, and related problems. Sociological Methods and Research131109119. 10.1177/0049124184013001004
Rindskopf, D., & Rose, T.
1988.
Some theory and applications of confirmatory second-order factor analysis Some theory and applications of confirmatory second-order factor analysis. Multivariate Behavioral Research23151-67.
10.1207/s15327906mbr23013
Rosseel, Y. 2012. lavaan: An R package for structural equation modeling lavaan: An R
package for structural equation modeling. Journal of Statistical Software4821-36.
Rutter, M. 2007. Proceeding from observed correlation to causal inference: The use
of natural experiments Proceeding from observed correlation to causal inference:
The use of natural experiments. Perspectives on Psychological Science24377-395.
10.1111/j.1745-6916.2007.00050.x
Schmid, J., & Leiman, J. 1957. The development of hierarchical factor solutions The development of hierarchical factor solutions. Psychometrika22153-61.
10.1007/bf02289209
Spearman, CE. 1927. The abilities of man: Their nature and measurement The abilities of
168
man: Their nature and measurement. New YorkBlackburn Press.

Turkheimer, E., & Waldron, MC. 2000. Nonshared environment: A theoretical, methodological and quantitative review Nonshared environment: A theoretical, methodological and quantitative review. Psychological Bulletin126178-108. 10.1037/00332909.126.1.78
Vance, A. 2009Jan 6. Data analysts captivated by Rs power Data analysts captivated by
Rs power. New York Times.
Wechsler, D. 2003a. Wechsler intelligence scale for children Wechsler intelligence scale
for children (4th ed.). San Antonio, TXThe Psychological Corporation.
Wechsler, D. 2003b. WISC-IV technical and interpretive manual WISC-IV technical and
interpretive manual. San Antonio, TXThe Psychological Corporation.
Woodcock, RW., McGrew, KS., & Mather, N. 2001. Woodcock-Johnson tests of cognitive
abilities Woodcock-Johnson tests of cognitive abilities (3rd ed.). Itasca, ILRiverside.
Wright, S. 1934. The method of path coefficients The method of path coefficients. The
Annals of Mathematical Statistics53161-215.
Wright, S. 1968. Evolution and the genetics of populations: Genetics and biometric foundations Evolution and the genetics of populations: Genetics and biometric foundations
(Vol. 1). ChicagoUniversity of Chicago Press.
Yung, YF., Thissen, D., & McLeod, LD. 1999. On the relationship between the higherorder factor model and the hierarchical factor model On the relationship between the
higher-order factor model and the hierarchical factor model. Psychometrika642113128. 10.1007/bf02294531

ISBN: 978-1-62618-728-3
Chapter 9
CONTRIBUTION OF INTELLECTUAL,
PSYCHOLOGICAL, DEVELOPMENTAL
AND SOCIO-ECONOMIC DATA TO HIGHLIGHT
SPECIFIC PROFILES OF HIGHLY GIFTED CHILDREN
Isabelle Simoes Loureiro1,, Laurent Lefebvre1
and Laurence Vaivre-Douret2
1
Cognitive Sciences, University of Mons, Mons, Belgium

2
Paris Descartes University, France
2
Sorbonne Paris Cit; Inserm, Paris
2
University of Paris-Sud and Paris Descartes University, France
2
Necker Hospital, Department of Child Psychiatry, Paris, France
2
Cochin-Port Royal Hospital, Department of Pediatrics, Paris, France
ABSTRACT
The aim of this paper is to analyze previously recorded data on intellectual,
cognitive, social, psychological, developmental and physiological profiles of highly
gifted children. This paper also deliberates socio-economic measures.
The criterion currently used to determine high potentialities is the intelligence
quotient (IQ) defined by the Wechsler Intelligence Scale for Children Fourth Edition
(WISC-IV). The analysis of specific intellectual profiles enables singularities in specific
highly gifted groups of children to become apparent. There is often a significant
difference between Verbal Intelligence Quotient (VIQ) and Performance Intelli- gence
Quotient (PIQ) in the third version of Wechslers intelligence test (WISC-III), and also
between the factorial indices in WISC-IV in highly gifted children with some authors
subsequently noticing that the size of the difference between these quotients is positively
correlated to the appearance of neuropsychological impairments.
Correspondence concerning this chapter should be addressed to Isabelle Simoes Loureiro, Cognitive Sciences,
University of Mons, Place du Parc, 20, B-7000 Mons, Belgium. E-mail: isabelle.simoesloureiro@umons.ac.be.
Tel.: (+32) 65 37 31 61. Fax: (+32) 65 37 31 32.
170
I. S. Loureiro, L. Lefebvre and L. Vaivre-Douret

Highly gifted children are more frequently concerned with psycholo-gical and social
troubles whatever the IQ profile. Many children with high potentialities suffer from
anxiety, depression, sleep disorders and social integration disorders.
In addition, developmental data and psychophysiological characte-ristics frequently
associated with highly gifted abilities are also examined, including socio-economic
measures. Findings show that a favorable environment is correlated to a high IQ.
We propose that the specific profiles observed in highly gifted children can be
interpreted like the products of a specific cerebral structure that results from the
combination of innate and environmental factors.
Keywords: Highly gifted children, cognitive profile, psychological profile, developmental

data, socio-economic data
INTRODUCTION
Highly gifted children represent 2 5% of the population, which corresponds to one to
two children per classroom. This phenomenon is thus not isolated, and contrary to what one
might think, it is not an unquestionable advantage. High abilities can also underline cognitive,
psychological and social difficulties. Indeed, some children will overcome these problems,
while others will evidence problems later in their school career. Peers may find them difficult
to cope with, which is why some gifted children will abstain from fully utilizing their talents.
This standardization may be the preliminary step to the atrophy of their capacities a real
deterioration of unexercised cognitive functions.
The objective of this chapter is to discuss data on certain specificities in highly gifted
children. First of all, the intellectual and cognitive profile will be examined. Indeed, an
examination of IQ test profiles shows some particularities in highly gifted children, including
inconsistent subtest profiles. To follow, we will also consider psychological and social levels
of children with high potentialities, with characteristics observed by some authors, for
example common disharmonies, anxiety, isolation and sleep disorders. We will then analyze
developmental and physiological aspects, followed by an examination of some socioeconomic measures.
INTELLECTUAL AND COGNITIVE PROFILES OF HIGHLY

GIFTED CHILDREN
Psychometrical Profile
The notion of Intelligence Quotient (IQ) is largely accepted in our society to represent
intellectual abilities of a subject. Wechsler tests are among the most commonly used test to
assess the cognitive skills of children (Grgoire, 2000). Evaluating high abilities is often
based on this sole criterion however educators must also consider that multidimensional
models of intelligence are becoming increasingly accepted (Perreira-Fradin et al., 2010) and
high abilities should not be defined by a single number but rather by a set of special
characteristics. Wechsler himself thought that intelligence tests cannot evaluate all aspects of
171
intelligence and insisted on non-intellectual factors of intelligence, such as personality

(Wechsler, 1950 quoted by Grgoire, 2000). Furthermore, he thought that exterior factors of
intelligence, including education, attention and perseverance, should be considered when
analyzing intelligence tests, as these factors are inseparable from daily life.
The Wechsler Intelligence Scale for Children third version (WISC-III) is used to identify
cognitive strengths and weaknesses. The WISC-III divides the scale in two measures: the
Verbal Intelligence Scale (VIQ) and the Performance Intelligence Scale (PIQ). Canivez and
Watkins (1998, 1999), have demonstrated the long-term stability of WISCIII IQ scores (over
a period of 3 years). Grgoire (2000) points out that these two measures are impaired by
certain errors of assessment, and that the difference between VIQ and PIQ is inevitably
affected by this error. At the 5% threshold (known as sufficiently careful), the difference is
worth 12 points (Grgoire, 2000). Lower than this, the difference is considered to be the result
of aleatory errors, while an interpretation is necessary for any difference above 12 points.
Interpretative strategies have been developed, for example.
If subtest scores do not vary significantly, the strategy is to analyze the top of WISC-III
scores hierarchy: Full Intellectual Quotient (FIQ), Verbal Intellectual Quotient (VIQ) and
Performance Intellectual Quotient (PIQ). In other cases, IQ score represents a summary of
diverse abilities and does not represent a unitary entity (Kaufman & Lichtenberger, 2000, p.
424, quoted by Watkins and Canivez, 2004). Subtest scores are the focus of the analysis and
when a subtest significantly differs from the childs own average, a specific interpretation in
terms of strengths and weaknesses is proposed. The practice of profile analysis is
controversial. While authors of several prominent texts (Kaufman, 1994; Sattler, 2008) have
recommended that the meaningfulness, interpretability, of the IQs of students be
questioned when there is substantial variability among subtest profiles, others have argued
that interpretation of profiles provides no additional meaningful information beyond IQ (e.g.,
Freberg, Vandiver, Watkins, & Canivez, 2008; Watkins, Glutting, & Lei, 2007). According to
Watkins and Canivez (2004), the practice of subtest interpretation is unreliable because the
long-term (3 years) stability is poor. For Watkins and Kush (1994), subtest profile analyses
lacks empirical research. For example, in a sample of 1200 special educational needs (SEN)
students, they found that only 3.6 % of the students demonstrated profiles considered to be
unique. However, in highly gifted children, there may be a large percentage of unique
profiles, specifically, Fishkin et al. (1996, quoted by Ricci, 1998) investigated WISC-III
subtest scores of highly gifted children, and determined that subtest scatter scores occurred
more frequently in highly gifted children than in a sample of average children.
When gifted children are tested with the WISC-III, a significant difference between VIQ
and PIQ is often found (Silver and Clampit, 1990, Simoes Loureiro, 2012). Silver and
Clampit (1990) have shown that a 12-point difference between VIQ and PIQ reaches a 40%
proportion in a general population and that a 21-point gap is evident in 5% of a general
population. However, the 21-point difference frequency increases by 20% in children with a
Full IQ (FIQ) of 130 and by 25% in children with an FIQ higher than 140. This is why this
difference is no longer considered to be rare in highly gifted children. Additionally, as
Vaivre-Douret (2004) has affirmed, differences between VIQ and PIQ may be a sign of
psychological or neuropsychological difficulty.
Psychometric profiles of highly gifted children are often described as heterogeneous with
VIQ superior to PIQ (Terrassier, 2002). Thirty-four percent of the children (FIQ > 130)
showed a significant difference between VIQ and PIQ in Hollinger and Koseks research
172
(1986, quoted by Pereira-Fradin, 2004), while the proportion reaches 48% in Bessou et al.s
research (2005).
The psychometrical profiles of these children have a particular subtest pattern:
Similarities and Comprehension subtests have higher scores (Bessou et al., 2005; Waldon and
Saphire, 1990) while Coding is amongst the weakest (Bessou et al., 2005; Waldon and
Saphire, 1990, Brown and Hwang, 1991, Simoes Loureiro, 2012; Vaivre-Douret, 2004).
Coding has been shown to produce a weak correlation with giftedness (Kaufman, 1992). For
Silverman, Gilman and Falks (n.d), Coding diminishes IQ scores of gifted children as Coding
failure can be interpreted as graphomotor troubles. Pencil tasks are rarely well-performed by
gifted children, which is why some authors believe that Coding should not be incorporated
into WISC tests (Silverman, Gilman and Falks (n.d); Adda, 1999). However, Simoes Loureiro
(2012) has discovered a correlation between a low score in Coding and the presence of a
processing disorder. Coding failure may not only come from an attention disorder but also
from a calligraphy disorder (Vaivre-Douret, 2004). Coding can thus be used in
neuropsychological examinations and can be viewed as an indicator of cognitive difficulties.
Digit Span, Arithmetic and Object Assembly have also been seen as a low score for gifted
students (Bessou et al., 2005; Brown and Hwang, 1991; Simoes Loureiro, 2012). Current
research, however, indicates that the interpretation of subtest profiles are unlikely to inform
meaningful treatment (Gustafsson & Snow, 1997; Watkins, Glutting, & Youngstrom, 2005).
The WISC-IV highlights four factorial indices: verbal comprehension, perceptual
reasoning, working memory and processing speed. According to Silverman, Gilman and Falk
(n.d.), verbal comprehension and perceptual reasoning are good indices of high potentialities
while working memory and processing speed are less linked to giftedness. Intellectual
profiles of highly gifted children are also characterized by discrepancies between composite
scores (Gilman et al., 2008). Pereira-Fradin, Caroff and Jacquet (2010) found different kinds
of problems linked to identifying highly gifted children with the WISC-IV. First of all,
validity and fidelity of the evaluation of high potentialities with the WISC-IV seem to be
insufficient. Another disadvantage is the lack of sensibility of the norms for a high level of
intellectual abilities. There is a ceiling effect in three subtests of the perceptual reasoning
index (Matrix Reasoning, Picture Concepts and Block Design). These authors still consider
that the introduction of four indices in the fourth edition can make identifying highly gifted
children more complicated. Some psychologists continue to refer to the only FIQ while others
worry about the indices. Flanagan and Kaufman (2004) deem FIQ as inexplicable if the
difference between two Composite scores is 23 points or more. Yet Silverman, Gilman and
Falk (n.d), cite a mean discrepancy of 13 points in a group of highly gifted children, while
according to Falk, Silverman and Moran (2004, quoted by Silverman, Gilman and Falk, n.d),
the discrepancies are even larger. If a discrepancy of 23 points or more exists, Flanagan and
Kaufman (2004) suggest combining the Verbal Comprehension Index and Perceptual
Reasoning Index in a General Ability index. While some researchers have identified potential
subtest abnormalities in gifted children, their opinions are based primarily on anecdotal
commentary and to date no research has compared the profiles of this clinical sample against
the profiles found in a normal population. Until this research is conducted, these claims must
continue to be considered speculative.
In conclusion, we think that intelligence tests like the WISC (including both the third and
fourth editions) act as tools to identify high abilities, however clinicians must take into
account not only the scores obtained but also all the other characteristics of the individual
173
child. We must also stress that high abilities can also be expressed in other contexts, for
example art, creativity, personality and the psychological sphere. Besanon, Zesnani and
Lubart (2010) remind us that conative factors like curiosity, perfectionism, a thirst for
knowledge and novelty-seeking are correlated to high creative potentials. To our knowledge,
intelligence tests do not evaluate all these factors. We thus insist on non-intellectual
evaluations when identifying gifted children.
Neuropsychological Profiles
Researchers examining highly gifted children have focused primarily on the mechanisms
of two particular cognitive functions, namely memory and executive functioning underlying
exceptional cognitive abilities. An examination of these skills (Simoes Loureiro et al., 2009:
Vaivre-Douret, 2004; Jambaqu, 2004) may be used to assist the identification of highly
gifted children with neuropsychological weaknesses or deficits.
Evoked potentials have demonstrated that highly gifted children manage to transmit more
information to their cerebral cortex (Grubar, 1997). Specifically, high information-processing
abilities (detection, perceptive discrimination, storage, and recall) have been observed.
Analytical processes, such as comparisons, matching features, and mental configuration are
powerful. Greater cognitive mobility and an analogical reasoning characterize their cognitive
activity (Planche, 2000). Cerebral functioning is also more intuitive and analogical and
originates from a greater right hemisphere activity (Siaud-Facchin, 2002). It has been
demonstrated that a quick nervous impulse transmission is correlated to a high level of
intelligence (Grubar, 1997, Reed & Jensen, 1992). The right hemisphere is more implicated
(Oboyle and Benbow, 1990) and interhemispheric transmission is more developed (Singh
and Oboyle, 2004). These abilities lead to ease and speed in understanding. Grubar (1997)
described more efficient memory in highly gifted children, attributed to an increase of REM
sleep phases and duration. Petre-Quadrens (1969) demonstrates that oculomotor activity is
higher during REM sleep phases of highly gifted children. Proportionally speaking, with
regard to short-term memory and duration, the short-term memory of people with an IQ of
140 is 2.5 times that of a normal subject (De Groot, 1974, quoted by Grubar, 1997). Excellent
verbal long-term memory and superior learning abilities in highly gifted children are
subsequently prominent (Jane et al., 1996).
Regarding language, precocity is noticed with babbling at around 4 months old, and the
first sentence at around 18 months (the association of two words). Later, children like to
play with language-clarifying words by employing synonyms or opposites, creating
neologisms or playing on a concept. Later, they seek to identify letters and to reproduce them
early even if the letters are unknown. However, highly gifted children can suffer from
language troubles, such as articulation disorders, stuttering or speech disorders (VaivreDouret, 2004). These language deficits can be explained by a higher testosterone level in
utero that inhibit the development of some left hemi- spheric areas that are linked to language
and which favor, as compen- sation, some right hemisphere areas (Chiron et al., 1997).
As far as attention is concerned, Houd (2005) suggests that highly gifted children have
exceptional abilities in focusing their concentration to avoid irrelevant information. In
contrast, however, other researchers have found that gifted childrens attention can also be a
weakness (Simoes Loureiro et al., 2009; Tordjman, 2007). Vaivre-Douret (2004) puts forth
174
anomalous failures in the subtests Coding and Symbol Search in the WISC-III, suggesting an
attention deficit. This lack of attention can also be the result of boredom or lack of interest for
a task, and could be observed at school for these children. When children show an interest in
a subject however, they are able to focus on it. This state is described by Csikszentmihalyi
(1990) as a super motivation that allows these children to succeed beyond their abilities.
Despite this, a gifted child can present an attention deficit disorder with hyperactivity (known
as ADHD) or without it (known as ADD) (Simoes Loureiro et al., 2009). As for diagnosis, we
must tread carefully since it could be the result of boredom at school, or in fact a real
attention problem. According to Albaret and Soppelsa (2003), the diagnosis of ADHD
(Attention Deficit-Hyperactivity Disorder) must be based on copious information: the
neuropsychological tests of the child as well as the comments and explanations of the parents
about their childs behavior.
For some authors, executive functions, such as inhibition and shifting, are not correlated
to intelligence, as acknowledged by Friedman, Miyake, Corley, Young, Defries and Hewitt,
2007; Welsh, Pennington and Groisser, 1991). Nevertheless, extensive research supports the
idea that executive function is related to intelligence (Arffa, 2007; Baron, 2003; Mahone,
Hagelthorn, Cutting, Schuerholz, Pelletier and Rawlins, 2002). Some studies have shown that
gifted children have results superior to those of normal children in mental flexibility (Dover
& Shore, 1991), inhibition (Davidson & Sternberger, 1984; Sternberg & Davidson, 1983),
planning (Planche, 2000) and categorization (Arffa, Lovell, Podell & Goldberg, 1998). In
problem solving, they use generalization capacities and inhibition processing and show a
particularly mobile cognitive functioning (Planche, 2005). For Siaud-Facchin (2004), their
specific thinking organization can explain that children with high potentialities can also
display difficulties at school. A lack of planning and executive slowness in the case of these
children can sometimes be identified; they have no working system and have multiple
difficulties managing their material. Thus, certain data in scientific literature and empirical
facts indicate some contradictions.
In the psychomotor domain, we notice different observations depending on authors.
Terrassier (1999) observed that there are no advanced concerning psychomotor aspects. On
the contrary, he talks about a gap between thinking and the accomplishment of concrete tasks
(Terrassier, 1999), perhaps because at school, these children do not like manual tasks.
Terrassier believes that a highly gifted child who cannot follow the normal mental planning
can react by becoming anxious, which can lead to hypertonia (Terrassier, 1999). It is for this
reason that manual tasks are avoided, but then writing and spelling can consequently
deteriorate. These children are described as being clumsy and can be affected by frustration
followed by a sense of shame if this continues. The children could subsequently abandon
these tasks resulting in a deterioration of psychomotor functioning. Nevertheless, this kind of
dyssyn- chrony only appears at school age, and contrary to the aforementioned, there is a real
harmony with regard to psychomotor development before the age of two (Vaivre-Douret,
2002). Vaivre-Douret (2004) continues by emphasizing the fact that psychomotor
disharmony, described by Terrassier (1999), appears rather like a deterioration of abilities
than a real characteristic of highly gifted children (Vaivre-Douret, 2002).
Grubar and Duyme (1997) showed that the cerebral plasticity of highly gifted children is
maintained for longer in gifted children than in normal children, which indicates complex and
quick learning abilities. However, this cerebral plasticity is ambivalent because it can
deteriorate the potential of children if the environment is not sufficiently stimulating for the
175
reason that highly gifted children are more vulnerable to environment (Vaivre-Douret, 2003).
Indeed, although cognitive functioning of highly gifted children is often described as more
efficient than the cognitive functioning of normal children, if children are neglected by the
environment, they may not develop all their potential and their abilities could also weaken
(Bessou, Tyrrell & Yziquel, 2004). Terrassier (1999) adds that if we do not recognize the
abilities of highly gifted children, they could give up and renounce their exceptional abilities
(Terrassier, 1999).
In conclusion, the analysis of the cognitive profile of highly gifted children in addition to
the basic intellectual test can reveal inconsistencies. A significant discrepancy between VIQ
and PIQ in WISC-III and between factorial indices in WISC- IV is the first indicator for the
clinician who should investigate the neuropsychological system.
PSYCHOLOGICAL, PSYCHOPATHOLOGICAL
AND SOCIAL FUNCTIONING
The first stages of psychoaffective development defined by the Freudian landmarks come
in rapid succession in children of high-level potentialities, with a certain progression
depending on the responses of the environment (Adda, 1999; Vaivre-Douret, 2002). The
awareness of the difference in sex and generation appears early. The nature of the cognitive
functioning allows children of high potentialities to develop a well-focused critical and selfcritical sense as early as 3 years old. Their sense of humor is sometimes impressive.
Moreover, they show generosity and empathy towards others (Vaivre-Douret, 2004).
Highly gifted children are more sensitive than children of average ability (Gibello, 1997).
Emotional intensity seems to be a trait of their particular functio ning (Robert & Lovett,
1994). For some authors, such as Cupertino (1993), these ultra-sensitive children present
heterogeneity regarding emotional abilities and there is no specific emotional profile in highly
gifted children because each subject is unique. Other authors, such as Guignard and Zenasni
(2004), suggest that some emotional characteristics, such as anxiety, affective intensity,
hyperstimulability and emotional intelligence, are redundant. Terrassier (1999) observed a
difference between intellectual development and affectivity in highly gifted children.
Immaturity is often employed to talk about affectivity development associated with high
abilities. These children cannot endure frustration, disagreements or contrariety in daily
living. Adda (2003) posits the hypothesis of a normal affectivity development. She has
claimed that the contrast between the intellect and the affectivity is so important that it can
give the impression of a gap. Some children interpret their sensitivity as the results of a
hypermaturity (Vaivre-Douret, 2002). Others do not notice any gap between the affectivity
field and the intellect (Siaud Facchin, 2002). According to Revol et al. (2003), these
emotional characteristics predispose to depression. Indeed, intelligence can be the source of
stress when it gives access to existential questioning that the child cannot assume, like
questions about death, diseases or disasters (Revol, 2004). If children are left alone without
any response, they can feel confused or distressed, which can lead to anxiety, depression or
denial of their own feelings. According to Simoes Loureiro et al. (2010), anxiety is more
frequent in highly gifted children compared to a sample of average-ability children. Parents
often described anxiety signs precociously, like separation difficulties, psychosomatic
176
symptoms (e.g. stomach ache, nausea and vomiting), verbal complaints and sleep disorders.
Revol et al. (2004) also underlines sleeping disorders, described as difficulty sleeping, waking
up in the night, nightmares and/or disturbed sleep. Oppositional disorders are the most
common symptoms in highly gifted children, according to Revol (2004). Simoes Loureiro et
al. (2010) have also discovered that parents of highly gifted children complain more often
about the oppositional behavior of their children. Terrassier (1979, 1999) also described the
opposition of children with high abilities to the parents rules.
Concerning social development of highly gifted child, while Andreani and Pagnin (1993)
propose the idea of a social equilibrium, the notion of social integration disorder is very
widely held. Terrassier (1996) even talks about social disharmony. Within a family
environment, parents can ignore childrens high abilities; they are considered like any other
child and their potentialities cannot be sufficiently stimulated. Moreover, highly gifted
children can directly ask many questions to those adults who do not have systematic
responses. Families can perceive their children as impertinent, the consequence of which
gives them a feeling of disbelief and insignificance. At school, these feelings can increase.
Highly gifted children want to share their interests with other schoolchildren but often feel
frustrated by others behavior and can become isolated and/or adopt social hyper-adaptability
measures. In class, the teacher frequently ignores the presence of highly gifted children, who
then expect normal performances from them. Terrassier (1999) has called this phenomenon
the negative Pygmalion effect. Unrecognized, children do not use their abilities to their full
potential. This can also occur within the family setting or elsewhere. It is therefore in this
context that development is compromised and can alter memory or attention (Terrassier,
1996). The albatross complex has been described by Gauvrit (2002) as: children who are
disadvantaged by their potentialities and are sometimes tempted to renounce to their
exceptional aptitudes to be accepted by their peers. This is otherwise known as an intellectual
inhibition.
The knowledge of the particular psychological, emotional and social functioning in
children of high potentialities contributes to the identification of highly gifted children.
Indeed, IQ alone can sometimes hide the potentialities when there are psychological or
neuropsychological troubles that can diminish IQ score. In addition to intelligence
evaluations, clinicians must refer to the analysis of the psycho-affective state to understand
the strengths and the weaknesses of a specific child.
DEVELOPMENTAL AND PSYCHOPHYSIOLOGICAL

CHARACTERISTICS
Terrassier (1979, 1999) and Gibello (1997) describe psychomotor disharmony
associated with high potentialities. This disharmony appears at school age, which is
considered quite late. Vaivre-Douret (2004) highlights that psychomotor and psychological
development harmoniously occur before the age of two. She emphasizes that psychomotor
development of highly gifted children from 0 to 3 years old is above average. In a recent
French study of Vaivre-Douret (2004), longitudinal data on psychomotor development of
highly gifted children has been obtained. It appears that newborns postural and neuromotor
responses seem to be more mature while some archaic reflexes disappear precociously. There
177
is significant development in postuto-motor and locomotor maturation; babies sit up unaided

at around 7 months and walk independently at around 1 year. Nevertheless, Louis et al.s
study (2005) did not find precocious acquisitions for walking-age between normal children
and highly gifted children. However, Simoes et al. (2010) did not find signifi-cant differences
for these development stages either, but when they reiterate their statistical treatment with age
corrections needed for premature children, they do observe significant advances. We interpret
this psychomotor precocity like the result from the childrens own motivation, their desire to
make body movements and acquire autonomy and power over their environment. The
dysharmony observed later in highly gifted children is probably the consequence of a
degradation of the unexercised functions (Vaivre-Douret, 2002). However, this characteristic
is not systematic. Indeed, it does not appear in highly gifted children who have received an
adequate stimulation. However, when the environment is not sufficiently stimulating, children
can develop neuropsychological disorders, which can have consequences on intellectual
quotient (Vaivre Douret, 2002). According to Terrassier (1999), psychomotor precocity does
not exist in highly gifted children. On the contrary, this author also observes a gap between
the thinking and the execution of concrete tasks. At school, these children have the most
difficulty when they undergo motor hand activities. On account of these difficulties, children
can avoid other manual activities and withdraw completely. Writing and spelling can also be
neglected. Children can be perceived as clumsy and they can feel frustrated and ashamed if
this persists. Motor difficulties in highly gifted children can also derive from dyspraxia.
Gerard and Brun (2005) said that prematurity is associated with the risk of cerebral injuries in
the parieto-occipital area. These lesions can evolve in dyspraxia. We must stress that Louis et
al. (2004) illustrate that highly gifted children are more frequently born prematurely. The
hypotheses of prematurity in the development of future motor problems can be contemplated.
In 1982, Geshwind and Behan wrote a paper that has marked the neuropsychological
society. They proposed that cerebral lateralization was linked to fetal testosterone level: an
increase intra-uterine testosterone level leads to a more developed right non-dominant
hemisphere. The GBG model (GeshwindBehan-Galaburda Model) claims that left-handed
people have suffered from intra-uterine complications that impacts the cerebral localization
and leads to the development of anomalous cerebral dominance. They explain a link between
left-handedness, immune deficits, learning disabilities and giftedness. In 1983, Benbow
showed that highly gifted children are more frequently boys with myopia, who are lefthanded or ambidextrous. More recently, Louis et al. (2004) showed particularities in highly
gifted children aged 8 to 11 years old, including such as abnormal pregnancy, prematurity and
perinatal stress. Highly gifted children more often develop headaches, allergies, myopia and
asthma. As stated by Simoes Loureiro et al. (2010), a larger proportion of children with
myopia, allergies and headaches is found in their sample of highly gifted children. According
to them, these particularities are the consequence of cerebral plasticity. Pathological events
during pregnancy can disrupt the neural organization and can cause particular cerebral
plasticity as described by Grubar (1997) and Vaivre-Douret (2011). Camos (2004) believes
that these characteristics are due to a high level of testosterone present in utero. The
consequence is a more symmetric development of the cerebral hemispheres. Indeed,
according to Ostatnikova et al. (2006), sexual hormones influence cerebral structures during
the prenatal period. It can be hypothesized that the process of a specific developmental
cerebral maturing is the result of distress in pregnancy, which can subsequently lead to a
specific cognitive functioning (Louis et al., 2005) together with a bigger cerebral plasticity
178
(Simoes Loureiro et al., 2010). Environment intervenes later; when encouraged, children can
enhance their abilities and performance level. However, their cerebral vulnerability can also
reduce the cognitive potentialities if the environment is not favorable (Simoes Loureiro et al.,
2010). Identifying highly gifted children is a real challenge as the preservation of their high
cognitive functioning depends on what the environment can offer. It is important to take
environmental factors into account, with regard to achieving the potentialities of
exceptionally gifted children. The hurdle is the child being able to accept achievement.
SOCIO-ECONOMIC DATA
It is important to remember that these are socio-economic factors that are positively
linked to high potentialities, and that favorable conditions in the childs environment (e.g.
sociocultural, educational and socio-affective, etc.) foster the realization of the childs highlevel potentialities. High-parental socio-professional status is associated with high IQ
populations (Turkheimer et al., 2003), with talents and potentialities depending on
socioeconomic environments (Vallot, 1973). Children of high potentialities more frequently
live with both parents (married or not) with a good level of education (Louis et al., 2004), and
more than 79% of fathers of the highly gifted reach a superior educational level (Louis et al.,
2003). It seems that a good socio-economic environment can increase IQ and that a weak
socio-economical level can reduce it (Duyme and Capron, 1997). Moreover, Vaivre-Douret
(2010) found that when exposed to a favorable perinatal environment (i.e. with few
pregnancy complications) and a favorable, postnatal, parental and socio-professional
environment, preterm infants, identified as gifted at school age, showed a significant
relationship between homogeneity in anthropometric variables and future motor and
intellectual development, compared to those born full term. Belgian interuniversity research
(Ministre de la Communaut franaise de Belgique, 2001) underlines that the socioeconomic and cultural level influences school results and childrens behavior when learning.
The innate or hereditary question regarding intelligence has not yet been resolved.
Parents who reach a superior educational level and a high-level socio-professional occupation
probably have some potentialities, which contribute to stimulating their child/children. We
propose that there is an interaction between innate cerebral maturing processes and the
individual experience within a given environment and the educational pattern. The notion of
cerebral plasticity in highly gifted children here is central. High potentialities can be fostered
by the environment but can also deteriorate if the people surrounding the child/children do not
deliver favorable stimulation.
CONCLUSION
The goal of this chapter is to suggest that a more thorough analysis of the heterogeneous
profiles of children of high-level potentialities is crucial. Although IQ is the common
reference to diagnose high intellectual levels, some neuropsychological and/or
psychological disorders can sometimes mask it. The knowledge of the psychological, social
and developmental characteristics in highly gifted children enriches the identification of these
179
gifted children. Children who endure this difference without any explanation may
sometimes feel confused and/or stressed, and understanding this is the first step toward selfacceptance and realization.
REFERENCES
Adda, A. (2003). Le livre de lenfant dou, le dcouvrir, le comprendre, laccompagner sur la
voie du plein panouissement. [The book of the gifted child]. Solal.
Albaret, J.-M., & Soppelsa, R. (2003). Evaluation psychomotrice et neuro psychologique du
Trouble Dficit de lAttention/ Hyperactivit, Examens et interventions. [Psychomotor
and neuropsychological evaluation of attention deficit/hyperactivity disorder, exams and
interventions]. Rdu- cation Orthophonique, 127, 53-68.
Arffa, S., Lovell, M., Podell, K., & Goldberg, E. (1998). Wisconsin Card Sort- ing Test
performance in above and superior school children: Relationship to intelligence and age.
Archives of Clinical Neuropsychology, 13, 713720.
Arffa, S. (2007). The relationship of intelligence to executive function and non-executive
function measures in a sample of average, above average, and gifted youth. Archives of
Clinical Neuropsychology, 22, 969978.
Benbow, C. (1986). Physiological correlates of extreme intellectual precocity.
Neuropsychologia, 24, 719-725.
Benbow, C., Stanley, J., Marshall, K., & Zonderman, A. (1983). Structure of intelligence in
intellectually precocious children and in their parents. Intelligence, 7(2), 129-152.
Besanon, M., Zenasni, F. & Lubart, T. (2010). Le haut potentiel cratif. En- fance, 1, 77-84.
Bessou, A., Tyrrell, C., & Yziquel M. (2004). Educational and career pathways of 28 gifted
adults. Neuropsychiatrie de lenfance et de ladolescence, 52, 154159.
Bessou, A., Montlahuc, C., Louis, J., Fourneret, P., & Revol, O. (2005). Profils
psychomtriques de 245 enfants intellectuellement prcoces au WISC-III. [Psychometric
profile of 245 intellectually precocious children (WISC III)]. ANAE, 81, 23-28.
Canivez, G. L., & Watkins, M. W. (1998). Long term stability of the Wechsler Intelligence
Scale for Children-Third Edition. Psychological Assessment, 10, 285-291.
Canivez, G. L., & Watkins, M. W. (1999). Long term stability of the Wechsler Intelligence
Scale for Children-Third Edition among demographic subgroups: Gender, race, and age.
Journal of Psychoeducational Assess ment, 17, 300-313.
Camos, V. (2004). Comptences exceptionnelles en mathmatiques. Psychologie Franaise,
49 (3), 321-336.
Chiron, C., Jambaqu, I., Nabbout, R., Lounes, R., Syrota, A., & Dulac O. (1997). The rightbrain is dominant in human infants. Brain, 120, pp. 10 57-1065.
Csikszentmihalyi, M. (1990). Flow: The Psychology of Optimal Experience, New York:
Harper and Row.
Dark, V. J., & Benbow, C. (1984). Type of stimulus mediates the relationship between
performance and type of precocity. Intelligence, 19, 337-357.
Davidson, J. E., & Sternberg, R. J. (1984). The role of insight in intellectual giftedness. Gifted
Child Quarterly, 28, 58-64.
Dickinson, R. M. (1970). Caring for the gifted. North Quincy, MA: Christop- her.
180
Dover, A., & Shore, B. M. (1991). Giftedness and flexibility on a mathematic- cal setbreaking task. Gifted Child Quarterly, 35, 99-105.
Flanagan, D. P., & Kaufman, A. S. (2004). Essentials of WISC-IV assessment. Hoboken, NJ:
John Wiley & Sons.
Friedman, N. P., Miyake, A., Corley, R. P., Young, S. E., Defries, J. C., Hewitt, J. K. (2006).
Not all executive functions are related to intelligence. Psychological Science, 17, 172-9.
Freberg, M. E., Vandiver, B. J., Watkins, M. W., & Canivez, G. L. (2008). Significant factor
score variability and the validity of the WISCIII Full Scale IQ in predicting later
academic achievement. Applied Neuropsycho- logy, 15, 131139.
Gauvrit, A. (2002). Le complexe de lalbatros, linhibition intellectuelle chez lenfant
intellectuellement prcoce, se dfendre ou sinterdire? ANAE, 67, 141-149.
Gilman, B., Rimm, S., Silverman, L. (2008). A Quick Guide to Using the WISC-IV to
Identify Gifted Children From Optimizing WISC-IV Asses- sment of Primary-Aged
Gifted Children, a panel discussion with Barbara Gilman, Sylvia Rimm and Linda
Silverman. NAGC 55th, Annual Conven- tion, Tampa, FL, November 1, 2008 from
http://www.ksde.org/ Portals/0/Special%20Education%20Services/gifted/QuickGuideto
UsingtheWISC-IV.pdf).
Grgoire, J. (2000). L'valuation clinique de l'intelligence de l'enfant: Thorie et pratique du
WISC-III, Edition Mardaga.
Grubar, J.-C., Duyme, M., & Cte S. (1997). La prcocit intellectuelle. De la mythologie
la gntique. Edition Mardaga.
Guignard, J. H, & Zenasni, F. F. (2004). Les caractristiques motionnelles des enfants haut
potentiel. Psychologie Franaise, 49 (3), 305-319.
Gustafsson, J. E., & Snow, R. E. (1997). Ability profiles. In R. F. Dillon (Ed.) Handbook on
testing (pp. 107135). Westport, CT: Greenwood Press.
Jambaqu, I. (2004). Contribution de la neuropsychologie dveloppementale ltude des
sujets haut potentiel, une revue de questions. [Contribution of developmental
neuropsychology to the study of gifted individuals]. Psychologie franaise, 49, 267-276.
Kaufman, A. S. (1992). Evaluation of the WISC-III and WPPSI-R for gifted children. Roeper
Review, 14, 154-158.
Kaufman, A. S. (1994). Intelligent testing with the WISCIII. New York: John Wiley &
Sons.
Louis, J., Revol, O., Nemoz, C., Dulac, R. M., & Fourneret, P. (2005). Les facteurs
psychophysiologiques de la prcocit intellectuelle: Rsultats dune enqute comparative
chez lenfant entre 8 et 11 ans. Archives de pdiatrie, 12, 520-525.
Mahone, E. M., Hagelthorn, K. M., Cutting, L. E., Schuerholz, L. J, Pelletier, S. F., Rawlins,
C. et al. (2002). Effects of IQ on executive function measures in children with ADHD.
Psychology Press, 8, 5265.
Ministre de la Communaut franaise de Belgique (2001). Les enfants et les adolescents
haut potentiel, Rapport final de la recherche-action inter universitaire (FUNDP, U. C. L.,
U. L. B., ULg, U. M. H.), 31 aot 2001. Polycopi. Bruxelles, Ministre de la
Communaut franaise de Belgique.
Mouchiroud, C. H., (2004). Haut potentiel intellectuel et dveloppement social. Psychologie
Franaise, 49 (3), 293-304.
O'boyle, M. W., & Benbow, C. (1990). Enhanced right hemisphere during cognitive
processing may relate to intellectual precocity. Neuropsycholo- gia, 28, 211-216.
181
OstatnIkova, D., Celec, P., Putz, Z., Hodosy, J., Schmidt, F., Laznibatova, J., & Kudela,
M. (2006). Intelligence and salivary testosterone levels in prepubertal children.
Neuropsychologia, 45, 1378-1385.
Petre-Quadrens, O., (1969). Contribution ltude de la phase dite paradoxale du sommeil,
Acta Neurologica Psychiatrica Belgica, 69, pp. 769-898.
Pereira-Fradin, M., La variabilit intra-individuelle chez les enfants haut potentiel
intellectuel, Psychologie Franaise, 49 (3), 263-266, 2004.
Pereira-Fradin, M., Caroff, X. & Jacquet, A.-Y. (2010). Le Wisc-IV permet-il damliorer
lidentification des enfants haut potentiel? Enfance, 1, 11-26.
Planche, P. (2000). Le fonctionnement et le dveloppement cognitifs de lenfant
intellectuellement prcoce. [The cognitive functioning and development of highly gifted
child]. Lanne psychologique, 100, 503-52 5.
Planche, P. (2005). Le fonctionnement et le dveloppement cognitif de lenfant
intellectuellement prcoce: Quelques spcificits. [The cognitive functioning of gifted
children] A. N. A. E. Approche Neuropsychologique des Apprentissages chez lEnfant,
17, 16-22.
Ricci, N. (1998). Profile analysis of WISC-III with Gifted Canadian Children. University of
Victoria
from
https://circle.ubc.ca/bitstream/handle/2429/12206/ubc_20020226.pdf?sequence=1.
Reed, T. E., & Jensen, A. R. (1992). Conduction velocity in a brain nerve pathway of normal
adults correlates with intelligence level. Intelligence, 16, 259-272.
Revol, O., Louis, J., & Fourneret, P. (2004). Lenfant prcoce, signes particu- liers.
Neuropsychiatrie de lenfance et de ladolescence, 52, 148-153.
Roberts, S. M., & Lovett, S. B. (1994). Examining the F in gifted: Academically gifted
adolescents physiological and affective responses to scholastic failure. Journal for the
Education of the Gifted, 17(3), 241-259.
Sattler, J. M. (2008). Assessment of Children: Cognitive foundations (5th. ed.). San Diego:
Author.
Siaud-Facchin, J. (2002). Lenfant surdou, laider grandir, laider russir, dition Odile
Jacob.
Siaud-Facchin, J. (2004). Comprendre les difficults dapprentissage de lenfant surdou: Un
fonctionnement intellectuel singulier? [How under stand the difficulties of learning of the
gifted child: is a particular intellectual functioning?]. Neuropsychiatrie de lEnfance et de
lAdolesce- nce, 52, 142-147.
Silver, S. & Clampit, M. (1990). Frequency of discrepancies between deviation quotients in
profiles of children with above-average IQs on the Wechsler intelligence scale for
childrenRevised, Journal of School Psychology, 28 (3), pp. 251-256.
Silverman, Gilman and Falk (n.d). Who Are the Gifted Using the New WISC-IV? from
http://www.gifteddevelopment.com/PDF_files/NewWISC.pdf).
Simoes Loureiro I., Lowenthal L., Lefebvre L., Vaivre-Douret L. (2009), Le trouble
dficitaire de lattention chez lenfant haut potentiel, tude exploratoire. Approche
Neuropsychologique des Apprentissages de l'En fant, 101, 317-322.
Simoes Loureiro I., Lowenthal L., Lefebvre L., Vaivre-Douret L. (2010), Etude des
caractristiques psychologiques, physiologiques et sociales des enfants haut potentiel.
Enfance, 1, 27-44.
182
Simoes Loureiro, I. (2012). Profils neuropsychologiques des enfants haut potentiel.

ditions Universitaires Europennes.
Singh, H., & Oboyle, M. W. (2004). Interhemispheric Interaction During GlobalLocal
Processing in Mathematically Gifted Adolescents, average-Ability Youth, and College
Students. Neuropsychology, 18 (2), 371377.
Sternberg, R. J., & Davidson, J. E. (1983). Insight in the gifted. Educational Psychologist, 18,
51-57.
Terrassier, J.-C. (1979). Le syndrome de dyssynchronie. Neuropsychologie de lenfance et de
ladolescence, 27, 445-50.
Terrassier, J.-C. (1996). Le dveloppement psychologique des enfants intellectuel lement
prcoces, psychiatrie et psychobiologie. Journal de pdiatrie et de puriculture, 9, 221226.
Terrassier, J.-C.(1999). Les enfants surdous ou la prcocit embarrassante, 4 me dition
augmente, ESF Editeur, Paris.
Terrassier, J.-C., & Gouillou, P. (1998). Guide pratique de lenfant surdou, comment russir
en tant surdou? [Guidebook of highly gifted child, how to succed whith high
potentialities]. ESF diteur, Paris.
Tordjman, S. (2007). la rencontre des difficults prsentes par les enfants surdous.
[Focus on difficulties encountered by gifted children]. Archives de Pdiatrie, 14, 685
687.
Turkheimer, E., Haley, A., Waldron, M., DOnofrio, B. & Gottesman, I (2003).
Socioeconomic status modifies heritability of IQ in young children. Psychological
Science, 14, (6), 623628.
Vaivre-Douret, L. (2002). Le dveloppement de lenfant aux aptitudes hautement
performantes (surdous), importance des fonctions neuro-psychomotrices. ANAE, 67,
95-110.
Vaivre-Douret, L. (2003). Les caractristiques prcoces des enfants hautes potentialits.
Journal franais de psychiatrie, 18, 33-35.
Vaivre-Douret, L. (2004a). Les caractristiques dveloppementales dun chantillon denfants
tout venant hautes potentialits (surdous): Suivi prophylactique. Neuropsychiatrie de
lenfance et de ladolescence, 52, 129-41.
Vaivre-Douret, L. (2004b). Point de vue dveloppemental sur lenfant hautes potentialits
(surdous). Journal de pdiatrie et de puriculture, 17, 254-261.
Vaivre-Douet, L. (2011). Developmental and Cognitive Characteristics of High-Level
Potentialities (Highly Gifted) Children. International Journal of Pediatrics.
doi:10.1155/2011/420297.
Vallot, F. (1973). Enqute Nationale sur le Niveau Intellectuel des Enfants dge Scolaire,
Travaux et Documents, Cahier no. 64. UF, Paris, France, 1973.
Waldron, K. A., & Saphire, D. G. (1990). An analysis of WISC-R factors for gifted students
with learning disabilities. Learning disabilities, 23 (8), 491-498.
Watkins, M. W., & Canivez, G. L. (2004). Temporal stability of WISC-III subtest composite
strengths and weaknesses. Psychological Assessment, 16, 133-138.
Watkins, M. W., Glutting, J. J., & Lei, P. W. (2007). Validity of the full-scale IQ when there
is significant variability among WISCIII and WISCIV factor scores. Applied
Neuropsychology, 14, 1320.
183
Watkins, M. W., Glutting, J. J., & Youngstrom, E. A. (2005) Issues in subtest profile analysis.
In D. P. Flanagan & P. L. Harrison (Eds.) Contemporary intellectual assessment (2nd ed.,
pp. 251268). New York: Guilford Press.
Watkins, M. W. & Kush, J. (1994). WISC-R.subtest analysis: The right way, the wrong way.
or no way? School Psychology Review, 23, 638-649.
Wechsler, D. (1996). Manuel du WISC-III, Echelle dintelligence de Wechsler pour enfants.
3 d., Paris, ECPA.
Wechsler, D., (1996). WISC-III (Wechsler Intelligence Scale For Children. ECPA, version
franaise.
Welsh, M. C., Pennington, B. F., & Groisser, D. B. (1991). A normative-developmental study
of executive function: A window on prefrontal function in children. Developmental
Neuropsychology, 7, 131149.
Wiley, J., & Goldstein, D. (1991). Sex, handedness and allergy: are they related to academic
giftedness? Journal for the Education of the Gifted, 14, 412-422.

ISBN: 978-1-62618-728-3
Chapter 10
PRESCHOOLERS ATTACHMENT REPRESENTATIONS

AND PARENTING AS PREDICTORS OF INTELLECTUAL
CAPACITIES: A PERSON-ORIENTED APPROACH
Marie Stivenart, Jean Christophe Meunier
and Isabelle Roskam
Psychological Sciences Research Institute,
Louvain la Neuve, Belgium
ABSTRACT
The present study was designed as a person-oriented approach in the context of a
two-year follow-up. First, different subgroups were identified by a cluster analysis
approach based on preschoolers attachment representations and parenting, focusing on
two key dimensions (support and control). Second, these different subgroups were
hypothesized to differ in their level of intellectual capacities two years later. Cluster
analysis and comparison of means were used with a sample of 305 French-speaking
preschoolers. The mean age of the children was 54.7 months (SD = 11.35) at the time of
recruitment. The preschoolers attachment representations were assessed with the French
version of the Attachment Story Completion Task while their intellectual capacities were
assessed using two subtests (Block Design and Information) of the WPPSI-III. Parenting
was assessed, separately for the mothers and the fathers, using the Evaluation des
pratiques ducatives parentales (EPEP). A three-cluster solution emerged from our
analysis. Preschool children in cluster 1 were insecure and disorganized, with parenting at
a moderate level of both support and control. In cluster 2, preschool children were
moderately secure and disorganized, while parenting was particularly high in control and
low in support, both for the mothers and the fathers. In cluster 3, preschool children were
secure and organized, with positive parenting mainly focused on support (instead of low
on control). Preschoolers in cluster 3 were less at risk of displaying lower levels of verbal
capacities (Information) than children in clusters 1 and 2, while no distinction between
clusters appeared for reasoning capacities (Block Design). These results highlighted the
relevance of considering the environmental outcomes to predicting intellectual, especially
E-mail: marie.stievenart@uclouvain.be.
186
Marie Stivenart, Jean Christophe Meunier and Isabelle Roskam

verbal, abilities. The implications of the results for both theoretical and clinical purposes
are discussed.
Keywords: Attachment representations, childrearing practices, intellectual abilities, cluster

analysis, person-oriented approach
INTRODUCTION
Intellectual abilities in preschoolers have been separately associated with various
predictors within the child or within the family environment. However, a multiple-predictor
perspective has recently indicated that a single predictor is not enough to derail development,
whereas the accumulation of multiple predictors is often deleterious (DeKlyen & Greenberg,
2008). For this reason, studies combining predictors from different ecological levels, both the
child and the family environment, are needed to provide accurate accounts of intellectual
abilities in young children. The current study pursues this important objective by relating later
intellectual abilities to both preschoolers individual attachment representations and
parenting, something that is rarely done in the current literature. For the first time, to our
knowledge, the predictive power of the combination of attachment representations and
parenting for later intellectual abilities (IQ) is tested in this paper.
INTERRELATIONS BETWEEN ATTACHMENT AND IQ

Most previous studies relating attachment and IQ have considered childrens attachment
behavior, measured with the Strange Situation Procedure (SSP, Ainsworth, Blehar, Waters,
& Wall, 1978). This is based on observations of the childs behavior during separation and
reunion with the caregiver (principally the mother), and the childs attachment patterns are
inferred from this according to the balance between exploration of the environment and
attachment behavior. Insecure children tend to minimize or maximize their attachment
behavior in favor or at the expense of exploration. Secure children optimally balance their
attachment and exploratory behavior. Children with insecure early attachment experiences
were described by Lewis and Feiring (1991) as vulnerable, since these early experiences
were thought to lead children to be at risk of later development disorders. More recently, a
disorganized attachment pattern has been defined (Main & Solomon, 1986). These children
seem to lack the organized attachment strategies which would enable them to cope with stress
in difficult situations by using the caregiver as a source of comfort. Many researchers have
concluded that disorganization is a major risk factor for development disorders, and perhaps
even a greater risk than insecurity (for a meta-analysis, see van Ijzendoorn, Schuengel, &
Bakermans-Kranenburg, 1999).
Research results on the relations between attachment behavior and IQ in children are
somewhat inconsistent. In a meta-analysis, van Ijzendoorn, Dijkstra and Bus (1995)
concluded that attachment was only weakly associated with IQ. Conversely, other studies
have detected a significant link between attachment behavior and IQ: attachment patterns
Preschoolers Attachment Representations and Parenting as Predictors
187
have been found to predict IQ, especially verbal IQ, one year (van Ijzendoorn & Van VlietVisser, 1988) and three years (O'Connor & McCartney, 2007) later.
While childrens attachment behavior has been related to attachment representations (e.g.
Bretherton, Ridgeway, & Cassidy, 1990; Gloger-Tippelt, Gomille, Koenig, & Vetter, 2002;
Solomon, George, & De Jong, 1995), far fewer studies have considered the relations between
childrens attachment representations and IQ than the relations between their attachment
behaviors and IQ. This is somewhat surprising since current attachment representations - a
key concept in the attachment theory developed by Bowlby - are cognitive concepts.
Consequently, attachment representations and IQ could be considered as related to each other
in the cognition area. In line with this conceptual point, the reciprocal relations between
childrens attachment representations and IQ were recently studied using the Attachment
Story Completion Task (ASCT) (Bretherton, et al., 1990) to assess childrens attachment
representations through the completion of stories presented using figurines and dealing with
several attachment-related themes. The results confirmed the existence of reciprocal relations
between attachment and IQ, with, for example, IQ impacting on the development of
attachment representations and conversely, secure and disorganized attachment
representations impacting on verbal IQ (Stievenart, Roskam, Meunier, & Van de Moortele,
2011).
These results tend to confirm that attachment relationships could have an impact on the
development of IQ. As mentioned by OConnor and McCartney (2007), secure children are
more likely to explore their environment, and so to learn something, since they have already
acquired security with their attachment figures. Conversely, insecure and/or disorganized
children are less able to profit from environmental resources because they are more concerned
about the reactions and/or the attitudes of their attachment figures. Consequently, secure
attachment would favor high cognitive abilities. Conversely, insecure and/or disorganized
attachment representations appear to favor low cognitive abilities. This supports the relevance
of simultaneously considering attachment in relation to later IQ in the current study.
INTERRELATIONS BETWEEN PARENTING AND IQ

Parenting has mainly been appraised according to childrearing behavior, conceptualized
along two key dimensions: support and negative control (Aunola & Nurmi, 2005). Support
encompasses the affective nature of the parentchild relationship and refers to a variety of
related behavior including warmth, acceptance, involvement, autonomy, monitoring and the
establishment of guidelines. Negative control encompasses parents efforts to control their
childrens behavior by means of coercion, inconsistent, conflictual and/or harsh discipline, or
punishment. Negative control (high coercion, harsh punishment and lack of responsiveness)
has been repeatedly identified as a predictor for later IQ (e.g. Pougnet, Serbin, Stack, &
Schwartzman, 2011; Ryan, Martin, & Brooks-Gunn, 2006; Tamis-le Monda, Shannon,
Cabrera, & Lamb, 2004).
As suggested by Warren and Brady (2007), a useful theoretical framework for
understanding how highly supportive parenting may impact childrens cognitive development
(and consequently, IQ), and how controlling parenting may prevent this development is the
transactional model of development (Sameroff, 1975). According to this model, supportive
188
parents adapt their behavior in response to changes in the child in a way that directly supports
further development. This adjusting mechanism supports the childs exploration, as
previously described, and consequently his or her IQ.
Numerous research studies have focused on maternal parenting as having an impact on a
childs outcomes (e.g. Busch & Lieberman, 2010; Landry, Smith, Swank, Assel, & Vellet,
2001). However, a growing body of literature includes both mothers and fathers in work on
child development, since paternal parenting, as well as maternal, has been correlated to child
outcomes (e.g. Bronte-Tinkew, Carrano, Horowitz, & Kinukawa, 2008; Cabrera, Shannon, &
Tamis-LeMonda, 2007; Pougnet, et al., 2011). In line with these results, a recent study (Ryan,
et al., 2006) suggested that predictions of childrens IQ are more accurate when both the
mother and fathers parenting are considered simultaneously, instead of separately from each
other. From the childs point of view, a supportive parent may increase the benefits of the
other supportive parent, due to the coherence of their parenting, demonstrated as a significant
predictor of child adjustment (Teubert & Pinquart, 2010). Also, from the parents point of
view, a couple of supportive parents may develop superior interpersonal skills and emotional
regulation, positively impacting on their relationships with their child. This positive attitude
towards the child could be associated with positive child outcomes such as higher IQ
(Cummings & Davies, 1994).
Based on all this previous research, parenting in the current study was assessed using the
two key dimensions (support and negative control). These two dimensions were provided
simultaneously by both mothers and fathers. The main goal of the current study is to consider
these two dimensions as predictors of later IQ.
THE CURRENT STUDY

The current study extends the current literature in two ways. First, the combination of
attachment representations and parenting (support and negative control), in the prediction of
preschoolers IQ is new. We decided to focus our study on these two predictors since, to the
best of our knowledge; attachment, parenting and IQ have rarely been considered together in
the same study. Also, in accordance with Stievenart et al. (2011), IQ has been considered
separately for reasoning and verbal IQ in order to explore their different and specific effects.
Second, the current study is based on a person-oriented approach (Bergman & Trost, 2006;
Laursen & Hoff, 2006), whereas previous studies have usually been conducted in a classical
variable-oriented approach in which the concepts are considered as variables, and the
measures and their interrelations are studied over time. In person-oriented research, IQ is
treated in its intrapersonal context. The population is classified a priori into specified or
clustered groups (e.g. children with insecure attachment patterns or negative parenting).
Attempts are made to establish the external validity of the groups by asking whether their
members also differ according to variables not used to create the groups, such as IQ (von Eye
& Bogat, 2006). When these variables differentiate the groups, they are identified as
predictors since their exposure to them increases the chance of a predicted outcome
(Greenberg, Speltz, DeKlyen, & Jones, 2001). Because person-oriented analysis focuses on
the number of individuals displaying different predictors, either alone or in combination, it
provides important information about the indicators, for example, the configurations of
189
predictors which matter most. In this way, it can be considered as a critical method for
identifying at-risk children (Haapasalo, Tremblay, Boulerice, & Vitaro, 2000). This is the
reason why we chose to use a person-oriented approach for this study.
We designed our study as a person-oriented approach in the context of a two-wave
longitudinal research project, to predict later IQ using both childrens attachment
representations and parenting. Because of the associations between IQ and both attachment
representations and parenting, cluster analysis was used to distinguish groups of children
according to their personal (attachment representations) or environmental characteristics
(parenting) or both. In line with the multiple predictor perspective, children displaying both
disorganized or insecure attachment representations and negative parenting (low support and
high negative control) were expected to display lower IQ than children with only one
predictor.
METHOD
Sample and Procedure
This study was part of the longitudinal H2M (hard-to-manage) children research
program, which was conducted by the Psychological Sciences Research Institute Institute of
Research in Psychological Sciences at the Universit catholique de Louvain (UCL), Belgium,
in collaboration with Saint Luc University Clinic in Brussels. Data was collected from a
group of 346 French-speaking Belgian children (55.84% boys), some of whom had been
referred to specialists for their behavior problems and others of whom were developing
normally. The non-referred cohort (n = 236) was recruited when the children were 30 to 91
months old and enrolled in the first to third kindergarten sections in various primary schools
in the French-speaking part of Belgium. Note that kindergarten schooling in Belgium covers
children from three to six years of age. The referred group (n = 110) was recruited from
pediatric units in Belgian hospitals where the children had been referred for externalizing
behavior (arousal, opposition, agitation, aggressiveness and non-compliance). The referral
had to have been made by a doctor after a diagnosis of externalizing behavior which was the
immediate and principal reason for the referral. Children displaying substantial language
delays or developmental disorders were excluded from the sample. At the time they were
recruited, all the children were attending normal schools.
Externalizing behavior is considered here as a continuum ranging from normal to
pathological levels. There was a certain overlap between the two subsamples, since several
non-referred children displayed moderate to high levels of externalizing behavior, and several
referred children displayed only moderate levels. By considering both referred and nonreferred children, we maximized the distribution on the continuum from normal to
pathological behavior. Note that the way referrals for externalizing behavior in Belgium are
treated means that very young children can be referred for behavioral problems to a
pediatrician or to mental health services, but in the absence of a neurological or biological
origin of the problem, neither diagnosis nor systematic treatment is provided under the age of
seven. In this context, the referred children in our sample had not undergone any systematic
treatments including medication between the three two waves of assessment, although a few
190
of them (9%) had taken part in psychomotor activities conducted by physiotherapists. The
frequency of such activities, their relation to externalizing behavior and its correlates were
subject to controls in this study and found not to have any significant effect on the variables
under consideration.
The parents were informed about the study and asked if they would be willing to
participate in a longitudinal research program. They were assured that the data would remain
confidential. Informed consent was obtained from all the adult participants.
The data presented here come from two waves of assessment, one at the outset of the
research program (T1) (n = 346, 31.79% referred children), and the other 24 months later
(T2) (n = 260, 34.61% referred children). Attrition (24.85%) is common in longitudinal
research. Comparison of families with complete and incomplete data revealed no systematic
differences in either the socio-demographic variables or those under investigation.
The mean age of the children was 54.7 months (SD = 11.35) at T1 and 77.3 months (SD =
12.30) at T2. Most of the children lived in two-parent families (87%), but 13% of the
parents were separated or divorced at the onset of the study. Of the families who took part in
the study, 20% had only one child, 46% had two children, and 34% had three or more
children when the study began.
The educational level of the parents was measured by the number of years of education
they had completed. Some 27% of mothers had completed 12 or fewer years of schooling
(corresponding to the end of secondary school in Belgium); 56% had completed three more
years (corresponding to undergra duate study); while 17% had achieved a four-year degree or
more. The comparable figures for fathers were 37%, 42% and 21%.
At T1, both parents were instructed to complete a questionnaire, independently and
without consulting one another, assessing their parenting behavior towards the target child.
The questionnaire was given by the preschool teachers to the parents of non-referred children
and by the research assistants to the parents of referred children. The parents had to complete
it at home within the following two weeks. At T1, children were also visited at school and
examined individually in a quiet room. The children completed an assessment tool designed
to assess childrens attachment representations (Bretherton, et al., 1990). At T2, children were
exposed to the Block Design and Information subtests of the Weschler Preschool and
Primary Scale of Intelligence (WPPSI-III).
Measures
Outcome Measure
A brief evaluation of intellectual abilities (IQ) was carried out using two subtests of the
WPPSI-III (Wechsler, 2004): Block Design and Information. These subtests are considered to
be the most representative subtests of reasoning IQ and verbal IQ. They have been found to
correlate closely with the full scale IQ (Anastasi & Urbina, 1997). This was confirmed in the
subsample of referred children (n = 125) for whom a complete IQ assessment was available
(reasoning IQ: r = .74, p < .01; verbal IQ: r = .77, p < .01). Standardized scores were used in
the current study (Wechsler, 2004). The correlations between verbal and reasoning IQ were
moderate (r = .46, p < .01). In the non-referred group, the mean score of the verbal IQ subtest
was 9.48 (ranging from 1 to 18, SD = 3.01), while the mean score of the reasoning IQ subtest
was 9.86 (ranging from 1 to 19, SD = 3.17). In the referred group, the mean score of the
191
verbal IQ subtest was 9.42 (ranging from 1 to 16, SD = 2.76), while the mean score of the
reasoning IQ subtest was 9.37 (ranging from 1 to 15, SD = 2.14). No significant difference
was highlighted between these groups (verbal IQ: F(1,394) = .36, p = .85; reasoning IQ:
F(1,397) = 2.23, p = .14).
Measure of Predictors
Parenting, especially the childrearing practices, was assessed at T1 by the mothers and
the fathers separately, with the preschoolers form of the Evaluation des pratiques ducatives
parentales-Preschool and primary school form designed for parents of two-to-seven-year-old
children (EPEP, Meunier & Roskam, 2007). The EPEP-PPSF is a 40-item instrument yielding
nine factors: Positive Parenting, Monitoring, Rules, Discipline, Inconsistent Discipline, Harsh
Punishment, Ignoring, Material Rewarding, and Autonomy. A five-point Likert-type scale is
provided for each item, ranging from never to always. This instrument has been validated
on 565 French-speaking mothers and fathers of children developing normally and shows good
psychometric properties (Meunier & Roskam, 2009). For the nine-factor solution extracted in
the validation study, Cronbachs alpha ranged from .59 to .90; the total percentage of variance
explained was 61.36%. Confirmatory factor analyses from the validation study showed that
two second-order factors covering the supportive and controlling dimensions of parenting
emerged from the initial factor solution. The supportive factor was composed of Positive
Parenting, Autonomy, and Rules and included items such as When my child seems to have a
problem, I discuss with him/her what is wrong.
The controlling factor included Discipline, Harsh Punishment, and Ignoring and included
items such as When my child does something that is not allowed, I only talk to him/her
again when he/she behaves better. The fit measures in the validation study demonstrated an
acceptable fit to the data, with CFI = 0.94, RMR = 0.03, and RMSEA = 0.05 (Hu & Bentler,
1999; Meunier & Roskam, 2009). Similarly to the study of Meunier et al. (2011) and in order
to limit the number of predictors in the analyses, a support-to-negative-control ratio was
calculated in the current research both for the mothers and the fathers. A score of 1 thus
indicates equal amounts of support and negative control, whereas a score of less than 1
indicates more negative control than support, and a score of greater than 1 indicates more
support than control. While parenting was mostly appraised in previous studies from the
perspective of the mothers, in the current study it was assessed considering both the mothers
and the fathers ratios. More than 80% of the families in the sample were hence two-parent
families; in families where parents were separated, both the mother and the father were
involved in childrearing. A single factor, parenting, was computed by averaging the
mothers and the fathers support-to-negative-control ratio (r = .61).
The Attachment Story Completion Task (ASCT) was used to assess the childrens
attachment representations (Bretherton, 1990; Bretherton, et al., 1990). The French
translation, the Histoires completer (Fr-ASCT), was used in this sample, with both nonreferred and referred children. The ASCT has been cross-validated in several studies using
different coding systems, and by comparison with childrens responses to current or previous
separation/reunion episodes as presented in the SSP (e.g. Gloger-Tippelt, et al., 2002;
Solomon, et al., 1995). These comparisons support the view that story completions reflect the
childs attachment representations. The administration of the task was video-recorded. The
procedure includes five story stems: (a) the child figure causes an accidental mishap (spills
juice at the dinner table), (b) is hurt (falls off a rock in a park), (c) is afraid (of a monster in
192
the bedroom), and experiences (d) a separation from and (e) reunion with parents (the parents
leave for a trip while the grandmother looks after the children). In addition, a preliminary
story stem (a birthday party) serves as a warm-up procedure to introduce the child to what he
or she is expected to do.
The narratives the children produced were coded by clinical research assistants using the
Q-set procedure, Cartes pour le Codage des Histoires completer (CCH), which was
developed by Pierrehumbert and his colleagues (CCH: Miljkovitch, Pierrehumbert,
Karmaniola, & Halfon, 2003). After view- ing the video recording of the whole set of six
stories, the coder scores 65 items, each describing potential characteristics of the narrative
(Pierrehumbert, et al., 2009). Items focus either on the content (e.g., the child portrays the
parents as available) or on the formal characteristics of the narrative (e.g., the child enacts
emotions within the story). The first step of the CCH coding procedure was to sort the items
into seven piles (free distribution) from the most to the least characteristic of the childs
narrative. The second step consisted of a forced distribution, keeping only a specified number
of items in each pile. Each item received a score between 1 (least characteristic) and 7 (most
characteristic). Q-correlations are computed with the scores of the forced distribution, by
comparing the childrens individual Q-set description with the criterion sort provided by
experts for a prototypical child (Miljkovitch, et al., 2003). In the current study, only the
security and disorganization Q-scores were used for analysis, scores varying from +1.00 to
1.00, with a higher score indicating greater security or disorganization.
To maximize the coding validity, 20% of the first wave of video-recorded ASCT, for
both the referred and the non-referred groups, were coded separately by two independent
coders. The agreement between the two coders for the two Q-scores of attachment patterns
was computed using intra-class correlations (ICCs). These coefficients illustrate the
advantage of taking into account differences in scoring means for each coder (Howell, 1998,
pp. 550-553). The ICCs for the security and disorganization Q-scores were .80 and .86
respectively. These values were considered as good, although higher ICCs between coders
have previously been reported using the same procedure (.94 and .90 respectively)
(Miljkovitch, Pierrehumbert, Bretherton, & Halfon, 2004; Miljkovitch, Pierrehumbert, &
Halfon, 2007). However, our values were similar to those that were recently reported for a
Spanish sample of 30 randomly selected cases, with a total of 10 judges (Pierrehumbert, et al.,
2009). Their ICCs for the two Q-scores were .81 and .81, very similar to our results.
Statistical Analysis
Our data analysis process was the following. First of all, cluster analysis is recommended
for exploring data and identifying subgroups (Pierrehumbert & Miljkovitch, 2000). Wards
minimum variance method of analysis was used, as this method is particularly effective in
obtaining contrasting groups. However, recent work (Husson, L, & Pags, 2009) has
suggested that cluster analysis is more efficient when performed after a principal component
analysis (PCA). First, we performed a PCA on parenting and attachment Q-set scores,
measured at T1. This analysis was carried out without rotation. The stopping rule was data
driven in the sense that the first two principal components matched largely the two themes of
interest. The first axis referred to parenting, with 30.06% of the variance; and the second axis
referred to attachment, with 21.91% of the variance. The following principal components
were attached to subgroups of initial variables without clear interpretability.
193
Second, we perform a hierarchical clustering on the data. Instead of performing it on the

original variables, we use the first two principal components of the previous analysis. Preprocessing the data through a factor analysis serves as a guarantee of obtaining a perfect
euclidean space. Such a space is protecting us against mischoosing the distance used in the
clustering algorithm. Furthermore, "PCA can be viewed as a denoising method which
separates signal and noise: the first dimensions extract the essential of the information while
the last ones are restricted to noise" (Husson, Joose, & Pages, 2010, p. 2). We select the first
two axes as "signal". It is shown that "without the noise in the data, the clustering is more
stable than the one obtained from the original" set of variables (Husson, Joose, & Pages,
2010, p. 2)s. The algorithm chosen for the hierarchical clustering is the Ward's method. The
number of clusters was chosen on the dendrogram, based on the biggest jump of withininertia.
In a second step, the subgroups revealed by the cluster analysis were compared with
respect to the IQ at T2, with the statistical significance level set at .05. Previous studies have
mainly used analyses of variance (ANOVA) to compare the performance of subgroups. In
post hoc analyses, Tukeys range test was used.
RESULTS
Preliminary Analysis
Pearson correlations were computed between the security Q-score, the disorganization Qscore and parenting (mothers support and negative control, fathers support and negative
control). Parenting displayed a low correlation with disorganization (r range from .06 to .13)
and security (r range from -.0 to .15) Q-scores. These low correlations highlighted the
independence of the two predictors, attachment Q-scores and parenting.
Cluster Analysis
Without any constraint being forced by ourselves, a three-cluster solution emerged. The
descriptive statistics are presented in Table 1. Preschool children in Cluster 1 were lower in
the security Q-score, revealing insecurity, and higher in the disorganization Q-score,
revealing disorganization, than the children in other clusters, while parenting was moderately
high in support and low in negative control. Cluster 1 comprised 19.3% of the sample (n =
80).
Preschool children in Cluster 2 (23.4% of the sample, n = 97) were characterized by
moderate security and disorganization Q-scores but negative parenting (higher in negative
control and lower in support), both for the mothers and the fathers, than in other clusters.
Cluster 3 covered 57.3% of the sample (n = 238) and represented normally developing
children. Preschool children in Cluster 3 were secure (higher Q-scores) and organized (lower
Q-scores), with parenting mainly focused on support (higher scores, with lower scores in
negative control), for the two parents.
194
Comparison of Clusters 24 Months Later

The childrens IQs were compared at the 24-month follow-up (T2). No main effect was
observed for the reasoning IQ (F(2,304) = .24, p= .78, = .02), while a main effect was
observed on the verbal IQ (F(2,304) = 3.44, p= .03, = .02). Post-hoc tests showed that the
children in Cluster 3 (secure and organized children with positive parenting) were less at risk
of displaying lower verbal IQ than children from Cluster 1 and Cluster 2. The differences
between the other remaining clusters were not significant.
Table 1. Descriptive statistics for the three clusters which emerged
from the cluster analysis
Attachment
Mothers
Fathers
Age
(in years)
Gender
Group
Cluster 1
Mean
95%
(SD)
CI
Security
-.11(.03) [-.17; -.05]
Disorganization -.08(.02) [-.13; -.03]
Support
4.05(.05) [3.97; 4.19]
Control
2.54(.05) [2.43; 2.64]
Support
3.84(.06) [3.71; 3.96]
Control
2.54(.06) [2.41; 2.67]
3.84(.13) [3.58; 4.10]
Cluster 2
Mean
95%
(SD)
CI
.38(.02) [.34; .42]
-.44(.01) [-.47; -.41]
3.86(.05) [3.76; 3.86]
3.11(.05) [3.02; 3.21]
3.85(.04) [3.76; 3.94]
2.85(.05) [2.73; 2.97]
4.09(.11) [3.87; 4.30]
Cluster 3
Mean
95%
(SD)
CI
.45(.01) [.42; .47]
-.50(.01) [-.52; -.48]
4.22(.02) [4.17; 4.27]
2.28(.03) [2.22; 2.35]
4.04(.03) [3.99; 4.10]
2.36(.03) [2.30; 2.43]
4.22(.07) [4.08;4.35]
Male
Female
Non-Referred
Referred
64
33
53
44
97
186
52
116
122
238
59
21
55
25
80
CI: confidence interval.
Table 2. Descriptive statistics for childrens IQ at the 24-month follow-up according

to their results in the cluster analysis
Verbal IQ
Reasoning IQ
Mean (SD)
95% CI
Mean (SD)
95% CI
Cluster 1
9.34(3.05) a
[8.54; 10.14]
9.79(2.94)
[9.01; 10.56]
Cluster 2
9.43(2.70)
[8.79; 10.17]
9.34(2.91)
[8.55; 10.05]
Cluster 3
10.35(3.28)a
[9.99; 10.90]
9.53(3.81)
[8.93; 10.03]
CI: confidence interval.
a
: Tukey post-hoc test revealed significant mean differences between Cluster 1 and Cluster 3 (mean
difference = -1.00, p =.08).
CONCLUSION
The main objective of this study was to identify preschool children who were at risk of
having low IQ. We employed a multiple predictor perspective, including childrens
attachment representations and parenting as predictors. A cluster analysis identified three
195
subgroups. Comparisons between these clusters revealed IQ-related differences 24 months

later. Our main hypothesis was that children displaying both disorganized or insecure
attachment representations and negative parenting (low support and high negative control)
were expected to display lower IQ than children with only one predictor. This hypothesis was
supported by the results.
First of all, the cluster analysis distinguished the preschoolers according to both their
attachment representations and parenting, without isolating one predictor. Spontaneously, this
procedure distinguished preschool children with insecure and disorganized attachment
representations and a moderate level of parenting (Cluster 1), from those with negative
parenting and a moderate level of attachment representations (Cluster 2), and those displaying
no particular predictor (high levels of parenting and attachment representations) (Cluster 3).
This highlighted the relevance of considering simultaneously both parenting and childrens
attachment representations.
The main result of the current study is that the preschool children displaying both security
and organization, with positive parenting (low in negative control and high in support), were
less likely to have a lower, especially verbal, IQ 24 months later than the other groups of
preschool children. These results were in line with those of Stievenart et al. (2011), although
these authors used a classical variable approach without taking parenting into account,
suggesting a negative impact of childrens insecure and disorganized attachment
representations on verbal IQ.
The combination of predictors including security and desorganization of childrens
attachment representations and positive parenting might be considered as embedded in a more
broadly positive caregiving context. Conversely, Main and Hesse (1990) suggest that,
especially in cases of disorganization, the parent was seen as both a source of comfort and a
source of fear by children due to his/her specific parenting. Disorganized children were
therefore faced with a dilemma: whether to approach or to withdraw from their parent (Green
& Goldwyn, 2002). This dilemma was considered an obstacle for the development of
organized strategies, accounting for variance in intellectual development (Lyons-Ruth,
Repacholi, McLeod, & Silva, 1991). Furthermore, Aber and Allen (1987) advanced a secure
readiness to learn hypothesis: when attachment security is not ensured, the child is forced to
allocate excessive attention to monitoring attachment strategies, rather than to environmental
information improving his or her knowledge. Secure children are more willing to explore
their environment and solve problems autono- mously, resulting in improvements in their IQ
(De Ruiter & van Ijzendoorn, 1993) than insecure and disorganized children. These children
are unable to cope with stress in difficult situations and lack a clear and coherent strategy for
using the caregiver as a source of comfort. Thus, they do not learn any strategies for dealing
with attachment needs. They are therefore always monitoring any stressful event, and this
leaves them completely unavailable for anything else, such as environmental information. As
a result, their IQ does not improve. This could account for the positive interrelations,
observed in Cluster 3, between security and desorganization of attachment and later IQ, in a
broader context of positive parenting. Parents adopting parenting methods characterized by
high support but low negative control tend to promote secure and organized attachment in
their children. Parents appear to facilitate their childrens attachment representations of
themselves as lovable, and of the parent as available, by using support, responsiveness and
the promotion of the childs individuality (De Wolff & van Ijzendoorn, 1997; Karavasilis,
Doyle, & Markiewicz, 2003).
196
Preschool children in Cluster 2 were mainly characterized by negative parenting (low in

support and high in negative control), with moderate security and disorganization, but this
does not seem to impact later IQ. This result might be regarded as surprising, since it is not in
line with previous studies (e.g. Pougnet, et al., 2011; Ryan, et al., 2006; Tamis-LeMonda, et
al., 2004). How- ever, the specificity of the present study is its use of a multiple predictors
perspective.
This was not the case with the previous studies reported above, which consider only one
predictor in isolation in a variable-oriented approach. In the context of our person-oriented
approach using a multiple predictors perspective, our results suggest that negative parenting is
insufficient in itself to predict later IQ. However, the findings concerning verbal and
reasoning IQ were only computed with one subtest of the WPPSI. They need to be replicated
in future studies with a more complete evaluation of verbal and reasoning IQ.
All these results lead to some important considerations for clinical purposes. The
preschool children most at risk of displaying low IQ later on (Clusters 1 and 2) may benefit
from intervention concerning attachment relationships in the context of early intervention
programs, focusing on the prevention of deleterious developmental cascades between
cognitive and affective development (Masten, et al., 2005). Attachment relationships could be
modified through interventions focusing on parenting, which can have considerable impact on
childrens attachment (Bakermans-Kranenburg, van Ijzendoorn, & Juffer, 2008).
A recent study (Stievenart, Roskam, Meunier, & Van de Moortele, submitted) illustrated
the importance of childrearing practices, as a component of parental sensitivity, with regard to
the development of attachment representations, and especially the organization of those
representations; this is similar to what our present results suggest. Interventions can promote
parental development through the implementation of positive support, rules, autonomy
demands and monitoring techniques, and a decrease in negative parenting, harsh punishment,
ignoring and material rewarding techniques. These negative strategies can be progressively
replaced by more sensitive and supportive techniques. Positive parental development can be
expected to improve childrens secure and organized attachment representations, allowing
them to explore the environment and consequently to improve their cognitive abilities.
While innovative and important from both clinical and research perspectives, the present
study is by no means definitive. Our results were obtained using data on preschool childrens
attachment representations.
It would be interesting to replicate them with attachment behavior, and to explore the
patterns in older children, including adolescents. Also, we only considered childrens
attachment representations and parenting as predictors of IQ; other predictors (such as socioeconomic status, parents attachment representations, and childrens externalizing behavior)
are obviously also important.
The consideration of additional early predictors would lead to numerous other
combinations of predictors. Further person-oriented analyses could be used to identify valid
and reliable indicators allowing for the early identification of low IQ. Another limitation of
our study is that we did not distinguish between avoidance and ambivalence in insecure
children. We were therefore unable to be more precise about the predictors relating to these
two kinds of insecurity.
197
REFERENCES
Aber, J. L., & Allen, J. P. (1987). Effects of maltreatment on young children's socioemotional
development: An attachment theory perspective. Developmental Psychology, 23(3), 406414.
Ainsworth, M. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A
psychological study of the strange situation. Oxford England: Lawrence Erlbaum.
Anastasi, A., & Urbina, S. (1997). Psychological testing (7th edition). Prentice-Hall: Englewood Cliffs, US.
Aunola, K., & Nurmi, J.-E. (2005). The role of parenting styles in children's problem
behavior. Child Development, 76(6), 1144-1159.
Bakermans-Kranenburg, M. J., van Ijzendoorn, M. H., & Juffer, F. (2008). Less is more:
Meta-analytic arguments for the use of sensitivity-focused interventions. In F. Juffer, M.
J. Bakermans-Kranenburg & M. H. van Ijzendoorn (Eds.), Promoting positive parenting:
An attachment-based intervention. (pp. 59-74). New York, NY: Taylor & Francis Group/
Law- rence Erlbaum Associates.
Bergman, L. R., & Trost, K. (2006). The Person-Oriented Versus the Variable-Oriented
Approach: Are they Complementary, Opposites, or Exploring Different Worlds? MerrillPalmer Quarterly: Journal of Developmental Psychology, 52(3), 601-632.
Bretherton, I. (1990). Communication patterns. Internal working models and the
intergenerational transmissions of attachment relationships. Infant Mental Health Journal, 11(3), 237-252.
Bretherton, I., Ridgeway, D., & Cassidy, J. (1990). Assessing internal working models of the
attachment relationship: An attachment story completion task for 3-year-olds. Attachment
in the preschool years: Theory, research, and intervention. (pp. 273-308). Chicago, IL:
University of Chicago Press.
Bronte-Tinkew, J., Carrano, J., Horowitz, A., & Kinukawa, A. (2008). Involvement among
resident fathers and links to infant cognitive outcomes. Journal of Family Issues, 29(9),
1211-1244.
Busch, A. L., & Lieberman, A. F. (2010). Mothers' Adult Attachment Inter- view ratings
predict preschool children's IQ following domestic violence exposure. Attachment &
Human Development, 12(6), 505-527.
Cabrera, N. J., Shannon, J. D., & Tamis-LeMonda, C. (2007). Fathers' Influence on Their
Children's Cognitive and Emotional Development: From Toddlers to Pre-K. [Article].
Applied Developmental Science, 11(4), 208-213.
Cummings, E. M., & Davies, P. T. (1994). Maternal depression and child development.
Journal of Child Psychology and Psychiatry, 35(1), 73-112.
de Ruiter, C., & van Ijzendoorn, M. (1993). Attachment and cognition: A review of the
litterature. International Journal of Educationnal Research, 19, 523-540.
de Wolff, M., & van Ijzendoorn, M. (1997). Sensitivity and attachment: A meta-analysis on
parental antecedents of infant attachment. Child Development, 68(4), 571-591.
de Klyen, M., & Greenberg, M. T. (2008). Attachment and psychopathology in childhood. In
J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and
clinical applications (2nd ed.). (pp. 637-665). New York, NY US: Guilford Press.
198
Gloger-Tippelt, G., Gomille, B., Koenig, L., & Vetter, J. (2002). Attachment representations
in 6-year-olds: Related longitudinally to the quality of attachment in infancy and mother's
representations. Attachment and Human Development, 4(3), 318-339.
Green, J., & Goldwyn, R. (2002). Annotation: Attachment disorganisation and psychopathology: New findings in attachment research and their potential implications for
developmental psychopathology in childhood. Journal of Child Psychology and Psychiatric, 43(7), 835-846.
Greenberg, M. T., Speltz, M. L., DeKlyen, M., & Jones, K. (2001). Correlates of clinic
referral for early conduct problems: Variable- and person-oriented approaches.
Development and Psychopathology, 13(2), 255-276.
Haapasalo, J., Tremblay, R. E., Boulerice, B., & Vitaro, F. (2000). Relative advantages of
person- and variable-based approaches for predicting problem behaviors from
kindergarten assessments. Journal of Quantitative Criminology, 16(2), 145-168.
Howell, D. (Ed.). (1998). Mthodes statistiques en sciences humaines. Paris: De Boeck
Universit.
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6,
1-55.
Husson, F., Joose, J., & Pages, J. (2010). Principal component methods - hierarchical
clustering - partitional clustering: why would we need to choose for visualizing data?
Technical Report Agrocampus, 1-17.
Husson, F., L, S., & Pags, J. (2009). Analyse de donnes avec R. Rennes: Presses
universitaires de Rennes.
Karavasilis, L., Doyle, A. B., & Markiewicz, D. (2003). Associations between parenting style
and attachment to mother in middle childhood and adolescence. International Journal of
Behavioral Development, 27(2), 153-164.
Landry, S. H., Smith, K. E., Swank, P. R., Assel, M. A., & Vellet, S. (2001). Does early
responsive parenting have a special importance for children's development or is
consistency across early childhood necessary? Developmental Psychology, 37(3), 387403.
Laursen, B., & Hoff, E. (2006). Person-Centered and Variable-Centered Approaches to
Longitudinal Data. Merrill-Palmer Quarterly: Journal of Developmental Psychology, 52
(3), 377-389.
Lewis, M., & Feiring, C. (1991). Attachment as personal characteristic or a measure of the
environment. In L. E. Associates (Ed.), Intersections with attachment. Hillsdale, New
Jersey.
Lyons-Ruth, K., Repacholi, B., McLeod, S., & Silva, E. (1991). Disorganized attachment
behavior in infancy: Short-term stability, maternal and infant correlates, and risk-related
subtypes. Development and Psychopathology, 3(4), 377-396.
Main, M., & Hesse, E. (1990). Parents' unresolved traumatic experiences are related to infant
disorganized attachment status: Is frightened and/or frightening parental behavior the
linking mechanism? In M. T. Greenberg, D. Cicchetti & E. M. Cummings (Eds.), Attachment in the preschool years: Theory, research, and intervention. (pp. 161-182). Chicago,
IL US: University of Chicago Press.
199
Main, M., & Solomon, J. (1986). Discovery of an insecure-disorganized/disoriented

attachment pattern. In T. B. Brazelton & M. W. Yogman (Eds.), Affective development in
infancy. (pp. 95-124). Westport, CT US: Ablex Publishing.
Masten, A. S., Roisman, G. I., Long, J. D., Burt, K. B., Obradovi, J., Riley, J. R., et al.
(2005). Developmental Cascades: Linking Academic Achievement and Externalizing and
Internalizing Symptoms Over 20 Years. Developmental Psychology, 41(5), 733-746.
Meunier, J. C., & Roskam, I. (2007). Psychometric properties of a parental childrearing
behavior scale for French-speaking parents, children, and adolescents. European Journal
of Psychological Assessment, 23(2), 113-124.
Meunier, J. C., & Roskam, I. (2009). Validation of the preschool and primary school form of
a questionnaire assessing parents' childrearing behavior. Journal of Clinical Child and
Adolescent Psychology, 38(1), 166-175.
Meunier, J. C., Roskam, I., Stievenart, M., Van de Moortele, G., Browne, D. T., & Kumar, A.
(2011). Externalizing behavior trajectories: The role of parenting, sibling relationships
and child personality. Journal of Applied Developmental Psychology, 32(1), 20-33.
Miljkovitch, R., Pierrehumbert, B., Bretherton, I., & Halfon, O. (2004). Associations between
parental and child attachment representations. Attachment and Human Development, 6
(3), 305-325.
Miljkovitch, R., Pierrehumbert, B., & Halfon, O. (2007). Three-year-olds' attachment play
narratives and their associations with internalizing problems. Clinical Child Psychology
and Psychotherapy, 14(4), 249-257.
Miljkovitch, R., Pierrehumbert, B., Karmaniola, A., & Halfon, O. (2003). Les reprsentations
d'attachement du jeune enfant. Dveloppement d'un systme de codage pour les histoires
complter. Devenir, 15(2), 143-177.
O'Connor, E., & McCartney, K. (2007). Attachment and cognitive skills: An investigation of
mediating mechanisms. Journal of Applied Developmental Psychology, 28(5-6), 458-476.
Pierrehumbert, B., & Miljkovitch, R. (2000). La "prsomption de continuit" des modles
d'attachement. Incitation ou entrave la cration scientifique? In R. Tessier & G. M.
Tarabulsy (Eds.), Attachement et Dveloppement. Le rle des premires relations dans le
dveloppement humain. (pp. 206-234). Qubec: Presses de l'Universit du Qubec.
Pierrehumbert, B., Santelices, M. P., Ibez, M., Alberdi, M., Ongari, B., Roskam, I., et al.
(2009). Gender and attachment representations in the preschool years: Comparisons
between five countries. Journal of Cross-Cultural Psychology, 40(4), 543-566.
Pougnet, E., Serbin, L. A., Stack, D. M., & Schwartzman, A. E. (2011). Fathers' influence on
children's cognitive and behavioural functioning: A longitudinal study of Canadian
families. Canadian Journal of Behavioural Science/Revue canadienne des sciences du
comportement, 43(3), 173-182.
Ryan, R. M., Martin, A., & Brooks-Gunn, J. (2006). Is One Good Parent Good Enough?
Patterns of Mother and Father Parenting and Child Cognitive Outcomes at 24 and 36
Months. [Article]. Parenting: Science & Practice, 6(2/3), 211-228.
Sameroff, A. (1975). Transactional models in early social relations. Human Development,
18(1-2), 65-79.
Solomon, J., George, C., & De Jong, A. (1995). Children classified as controlling at age six:
Evidence of disorganized representational strategies and aggression at home and at
school. Development and Psychopathology, 7(3), 447-463.
200
Stievenart, M., Roskam, I., Meunier, J. C., & van de Moortele, G. (2011). The reciprocal
relation between childrens attachment representations and their cognitive ability.
International Journal of Behavioral Development, 35(1), 58-66.
Stievenart, M., Roskam, I., Meunier, J. C., & van de Moortele, G. (submittedaccepted). The
stability of young childrens attachment representations: The moderating effects of
parenting, IQ and attachment-related life events. Journal of Applied Developmental
Psychology.
Tamis-LeMonda, C. S., Shannon, J. D., Cabrera, N. J., & Lamb, M. E. (2004). Fathers and
Mothers at Play With Their 2- and 3-Year-Olds: Contributions to Language and
Cognitive Development. Child Development, 75(6), 18 06-1820.
Teubert, D., & Pinquart, M. (2010). The association between coparenting and child
adjustment: A meta-analysis. Parenting: Science and Practice, 10 (4), 286-307.
van Ijzendoorn, M., Dijkstra, J., & Bus, A.-G. (1995). Attachment, intelligence, and language:
A meta-analysis. Social Development, 4(2), 115-128.
van Ijzendoorn, M., Schuengel, C., & Bakermans-Kranenburg, M. J. (1999). Disorganized
attachment in early childhood: Meta-analysis of precursors, concomitants, and sequelae.
Development and Psychopathology, 11(2), 225-249.
van Ijzendoorn, M., & van Vliet-Visser, S. (1988). The relationship between quality of
attachment in infancy and IQ in kindergarten. Journal of Gen tic Psychology, 149(1),
23-28.
von Eye, A., & Bogat, G. A. (2006). Person-Oriented and Variable-Oriented Research:
Concepts, Results, and Development. Merrill-Palmer Quarterly: Journal of Developmental Psychology, 52(3), 390-420.
Warren, S. F., & Brady, N. C. (2007). The role of maternal responsivity in the development of
children with intellectual disabilities. Mental Retardation and Developmental Disabilities
Research Reviews, 13(4), 330-338.
Wechsler, D. (Ed.). (2004). WPPSI-III: Manuel d'interprtation Paris: Les Editions du Centre
de Psychologie Applique.
INDEX
#
21st century, 77
A
abuse, vii, 27
academic performance, 7, 9
academic progress, 89
academic success, 90, 105, 110
access, 47, 135, 175
accounting, 10, 99, 100, 138, 195
achievement test, 2, 8, 57, 93
acid, 71
acquisitions, 177
actuality, 56, 73
acute stress, 34
adaptability, 176
adaptation, 36, 38, 39, 40, 42, 43, 70, 71, 74
adenosine, 71
ADHD, 15, 103, 174, 180
adjustment, 89, 188, 200
adolescents, 19, 20, 31, 32, 55, 180, 181, 196, 199
adulthood, 31
adults, 19, 78, 115, 119, 121, 176, 179, 181
advancement, 104
affective intensity, 175
AFQT, 80
Africa, 72, 73, 74, 75, 76, 81
African-American, 17, 19, 20, 22, 23, 24, 25, 26, 27,
29, 72, 74, 83
age, 18, 19, 20, 21, 22, 23, 24, 26, 27, 29, 34, 36, 41,
45, 58, 65, 66, 67, 87, 90, 92, 93, 94, 97, 99, 100,
101, 102, 103, 105, 112, 114, 115, 155, 163, 174,
176, 178, 179, 185, 189, 190, 199
aggression, 49, 199
aggressiveness, 189
agonist, 28
alcohol dependence, 20
alcoholics, 17, 20, 29
alcoholism, 30, 32
alertness, 118
algorithm, 137, 193
allele, 18, 21, 32, 33, 71, 73, 74, 76, 77
allergy, 183
ALS, 62, 64
alters, 31
ambivalence, 196
American Psychiatric Association, 91
American Psychological Association, 51, 68
amino acid, 71
amygdala, 46
anatomy, 12
ANCOVAs, 22, 24
anger, 42
ANOVA, 23, 24, 193
anxiety, 9, 13, 170, 175
anxiety disorder, 9, 13
apex, 49
apples, 77
appraisals, 37
aptitude, 31
Aristotle, 80
arithmetic, 9, 25, 59, 131
arousal, 189
articulation, 173
Asia, 72, 74
aspartate, 19, 33
assessment, 1, 2, 9, 10, 12, 13, 14, 57, 63, 66, 67, 88,
101, 104, 105, 107, 111, 112, 116, 117, 118, 119,
120, 121, 122, 123, 167, 171, 180, 183, 189, 190
assets, 104
asthma, 177
at-risk populations, 88
atrophy, 170
attachment, 185, 186, 187, 188, 189, 190, 191, 192,
193, 194, 195, 196, 197, 198, 199, 200
202
Index
attachment theory, 187, 197

attitudes, 187
autism, 50, 65
automaticity, 50
autonomy, 177, 187, 196
avoidance, 45, 49, 51, 196
awareness, 47, 120, 175
B
babbling, 173
base, 110, 133
batteries, 8
behavioral assessment, 90
behavioral problems, 104, 189
behaviors, 27, 41, 43, 44, 67, 91, 104, 187
Belgium, 169, 185, 189, 190
belief systems, 49
benefits, 42, 47, 188
BIA, 163, 164
bias, 11, 43, 66
biological processes, 7
birth weight, 31
births, 81, 93
blacks, 70
blood, 1, 21
blood flow, 1
body size, 36
boredom, 42, 174
brain, 1, 7, 10, 17, 18, 27, 29, 30, 31, 32, 34, 36, 37,
38, 45, 51, 52, 53, 70, 77, 82, 84, 179, 181
brain activity, 31
brain functioning, 1
brain size, 32, 70
brain structure, 27
brain-derived neurotrophic factor (BDNF), 18
browser, 130
businesses, 129
Butcher, 78, 79, 82
C
candidates, 18
capitalism, 49
cardiovascular risk, 31
caregivers, 89
caregiving, 195
cascades, 196
cash, 9
catechol-O-methyltransferase (COMT), 18
categorization, 174
cation, 79, 94, 179
caucasians, 22
causal inference, 167
causality, 28, 166
causation, 77
central executive, 45, 46, 47
cerebral cortex, 173
cerebral hemisphere, 177
certification, 91
challenges, 103
Chicago, 16, 52, 53, 85, 110, 168, 197, 198
Child Behavior Checklist, 91, 103, 106, 107
child development, 32, 88, 100, 188, 197
childhood, 18, 31, 87, 100, 102, 104, 105, 106, 107,
108, 109, 166, 197, 198, 200
childrearing, 186, 187, 191, 196, 199
children, 2, 8, 13, 14, 15, 17, 19, 20, 21, 22, 24, 25,
26, 29, 30, 32, 33, 34, 42, 43, 58, 66, 67, 68, 78,
87, 88, 89, 90, 91, 92, 93, 94, 97, 99, 100, 101,
102, 103, 104, 105, 106, 107, 108, 109, 110, 119,
121, 168, 169, 170, 171, 172, 173, 174, 175, 176,
177, 178, 179, 180, 181, 182, 183, 185, 186, 187,
188, 189, 190, 191, 192, 193, 194, 195, 196, 197,
198, 199, 200
chimpanzee, 52
China, 2
chromosome, 31, 32, 72, 73, 77
civil servants, 2
classes, 89
classical conditioning, 41, 44
classification, 126
classroom, 100, 104, 105, 106, 170
clients, 114, 117
clinical application, 197
clinical assessment, 13
cluster analysis, 90, 185, 186, 189, 192, 193, 194,
195
clustering, 76, 193, 198
clusters, 72, 76, 185, 193, 194, 195
coding, 71, 78, 83, 84, 191, 192
coercion, 187
cognition, 6, 16, 18, 19, 27, 33, 50, 51, 52, 87, 88,
90, 100, 104, 187, 197
cognitive abilities, 1, 11, 13, 14, 15, 33, 68, 88, 104,
105, 109, 144, 165, 167, 168, 173, 187, 196
cognitive ability, 2, 4, 6, 7, 10, 12, 18, 19, 20, 21, 29,
32, 33, 35, 63, 68, 82, 88, 90, 92, 103, 137, 138,
142, 162, 163, 164, 200
cognitive activity, 173
cognitive deficit, 76
cognitive development, 30, 45, 89, 110, 187
cognitive function, 18, 19, 27, 30, 72, 77, 88, 92,
105, 125, 170, 173, 174, 175, 177, 181
203
Index
cognitive performance, 15, 18, 19, 29, 74, 77, 78, 79,
80, 81, 88
cognitive process, 2, 6, 8, 9, 13, 88, 93, 108, 180
cognitive profile, 170, 175
cognitive skills, 2, 14, 18, 89, 101, 170, 199
cognitive tasks, 10, 15, 48
cognitive theory, 10
cognitive tool, 43
cognitive variables, 14
coherence, 107, 188
collaboration, 6, 189
college students, 15
color, 37, 70, 74
commercial, 1, 6, 9
common symptoms, 176
community, 84, 100
Comparative Fit Index (CFI), 60, 61, 63, 96, 97, 98,
99, 100, 143, 145, 151, 154, 161, 191
competition, 44
complexity, 51
compliance, 189
complications, 74, 93, 177, 178
composition, 60, 61, 80
comprehension, 25, 45, 48, 55, 56, 90, 93, 172
computation, 48, 131
computer, 14, 48, 131
computing, 167
conception, 70, 77, 88
conceptualization, 100, 104
conditioning, 43
conduct disorder, 107
configuration, 173
conflict, 52
conscious awareness, 46, 47
consciousness, 41, 50
consensus, 1, 3, 56
consent, 190
construct validity, 8, 91
construction, 39, 49
consulting, 190
consumption, 80
contingency, 41
controversial, 171
convergence, 60
cooperation, 91
correlation, 1, 2, 7, 8, 12, 32, 48, 63, 77, 79, 81, 88,
90, 91, 93, 94, 97, 98, 99, 100, 103, 113, 114,
116, 123, 135, 137, 138, 139, 147, 148, 156, 158,
160, 162, 165, 167, 172, 190, 192, 193
correlation coefficient, 81
cortex, 28, 30, 32, 46
cost, 164
covering, 191
CPT, 90
creative potential, 173
creativity, 33, 173
cues, 36, 41, 43, 93
culture, 8, 37, 46, 52
curriculum, 105
D
daily living, 123, 175
danger, 41, 80
data analysis, 107, 129, 133, 156, 166, 192
data collection, 90, 91
data set, 63, 131, 134
database, 74, 75, 134
death penalty, 117, 119, 120, 121, 125, 126
decoding, 93, 109
decoupling, 47
defendants, 124
deficiency, 30, 89
deficit, 83, 109, 174, 179
degradation, 177
delinquency, 104
Delta, 74, 75
democracy, 49
dendrogram, 193
denial, 175
denoising, 193
Department of Education, 55
Department of Labor, 93
depression, 170, 175, 197
deprivation, 30
detectable, 44
detection, 31, 109, 173
developmental change, 91
developmental disorder, 84, 189
developmental process, 33, 104
developmental psychopathology, 198
deviation, 181
Diagnostic and Statistical Manual of Mental
Disorders, 91
diagnostic criteria, 20
dieting, 42
direct measure, 6
directionality, 22
disability, 111, 124, 126, 127
discordance, 70
discrimination, 16, 173
diseases, 9, 85, 155, 175
disequilibrium, 72, 77, 78
disgust, 42
disorder, 83, 107, 125, 172, 174, 176, 179
dissatisfaction, 108
204
Index
dissociation, 31
distress, 177
distribution, 72, 74, 77, 84, 116, 127, 130, 189, 192
diversity, 71, 77
divorce rates, 81
dizygotic, 156
DNA, viii, 21, 71, 78, 82, 84, 85
DOI, 14, 85
domestic violence, 197
dominance, 43, 157, 158, 162, 177
dopamine, 17, 18, 28, 29, 30, 31, 32, 33, 34
dopamine agonist, 31
dopaminergic, 17, 18, 27, 28, 30
drawing, 123
drive theory, 42
drug addict, 28, 32
drug addiction, 32
drug dependence, 27, 28
drugs, 27, 28
dyslexia, 108
E
East Asia, 72, 73, 74, 75, 76, 81
ecology, 104, 106
economic status, 196
ECPA, 183
editors, 130
education, 19, 21, 22, 23, 31, 33, 36, 87, 90, 106,
108, 171, 190
educational attainment, 19, 31
educational settings, 2
educators, 170
EEA, 35, 36, 39, 40, 43, 44, 46
elaboration, 40
elementary school, 14, 80, 88, 90, 102, 104, 108
eligibility criteria, 93
emotion, 100, 104, 105
emotion regulation, 105
emotional disabilities, 58
emotional health, 109
emotional intelligence, 175
empathy, 73, 85, 89, 175
empirical studies, 6, 7
employees, 81
encoding, 41
encouragement, 90
endurance, 3
enemies, 49
engineering, 30
England, 14, 78, 197
enslavement, 80
environment(s), viii, 19, 20, 29, 30, 31, 35, 36, 39,
44, 45, 49, 50, 81, 82, 118, 121, 155, 158, 162,
163, 164, 167, 168, 170, 174, 175, 177, 178, 186,
187, 195, 196, 198
environmental characteristics, 189
environmental effects, 19
environmental factors, 19, 82, 170, 178
environmental influences, vii, 16, 20, 33, 155, 163,
167
environmental resources, 101, 187
environmental stimuli, 47
equality, 95, 98
equilibrium, 176
Estonia, 32
ethnic background, 20
ethnic groups, 20, 22, 23, 24, 26, 27, 29
ethnicity, 21, 29, 90
ethology, 52
euphoria, 27
Eurasia, 72
Europe, 72, 74
everyday life, 51
evidence, 1, 11, 13, 18, 32, 33, 47, 48, 50, 56, 57, 59,
60, 74, 79, 80, 81, 91, 95, 97, 102, 105, 109, 111,
114, 116, 117, 118, 119, 122, 123, 125, 170
evolution, 38, 39, 40, 41, 42, 44, 50, 52, 53, 70
evolutionary psychologists, 37, 38
examinations, 2, 172
execution, 88, 111, 125, 177
executive function, 19, 48, 88, 89, 90, 91, 92, 101,
102, 103, 110, 173, 174, 179, 180, 183
executive processes, 35, 39, 44, 45, 61
exons, 72
expertise, 16
exposure, 7, 81, 188, 197
external validity, 188
externalizing behavior, 89, 91, 92, 189, 196
F
face validity, 1, 91
factor analysis, 12, 14, 56, 57, 58, 60, 66, 67, 68,
165, 166, 167, 193
families, 20, 78, 79, 93, 167, 190, 191, 199
family environment, 176, 186
fear, 37, 42, 46, 195
feelings, 27, 40, 42, 175, 176
fertility, 36
fidelity, 172
financial, 30, 47
first dimension, 193
fitness, 50, 71, 74
five-factor model, 51
205
Index
fixation, 74
flexibility, 174, 180
fluid, 3, 4, 10, 11, 12, 48, 51, 63, 78, 106
fluid intelligence, 11, 12, 48, 51, 63
Flynn effect, 113, 116, 117, 118, 119, 120, 121, 122,
123
fMRI, 10, 12, 28, 32
food, 33, 39, 40, 41, 42, 43, 44
food intake, 43
football, 3
Ford, 57, 67, 100, 105
formation, 71, 135, 163
formula, 113
foundations, 52, 68, 109, 168, 181
fragments, 21
France, 169, 182
freedom, 25, 41, 52, 67, 81, 137, 140, 142, 145, 151,
154, 160, 161, 164
friendship, 108, 166
frontal cortex, 28
frontal lobe, 10
G
GABA, 34
gender differences, 10
gender effects, 21
gene promoter, 84
general intelligence, 2, 3, 11, 15, 35, 37, 39, 40, 41,
42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 63, 72, 73,
78, 79, 81, 106
general knowledge, 88
generativity, 18
genes, viii, 10, 11, 18, 28, 30, 33, 34, 72, 73, 74, 77,
79, 84
genetic code, 71
genetic components, 159
genetic drift, 70, 72
genetic factors, vii, 10, 15, 29
genetic marker, 74
genetic predisposition, 81
genetics, 20, 32, 33, 52, 69, 81, 155, 165, 167, 168
genome, 70, 72, 77, 78, 79, 84
genomic regions, 72
genomics, 69
genotype, 17, 18, 19, 21, 22, 24, 25, 26, 27, 29, 155
gifted, 58, 169, 170, 171, 172, 173, 174, 175, 176,
177, 178, 179, 180, 181, 182
giftedness, 172, 177, 179, 183
glucose, 7, 18, 32
God, 49
grades, 2, 58, 59, 88, 90, 93, 94
graduate students, 59, 65
group membership, 102

grouping, 22, 70, 91, 92
growth, 87, 88, 91, 92, 93, 94, 95, 96, 97, 98, 99,
100, 102, 105, 107, 109
growth factor, 97, 99
growth models, 98, 99, 107
growth rate, 97, 99, 100
guidelines, 63, 107, 187
H
hair, 74
handedness, 177, 183
harmony, 174, 176
health, 31, 81, 105
health care, 81
height, 70, 77, 82, 85, 124, 147
hemisphere, 173, 177
heritability, 14, 18, 19, 20, 29, 34, 77, 85, 162, 182
heterogeneity, 90, 175
heuristic processing, 47
high school, 19, 93
high school degree, 93
Hispanics, 74
history, 2, 12, 13, 14, 36, 40, 49, 52, 69, 70, 71, 87,
108
homogeneity, 178
homozygote, 78
hormones, 177
host, 71
human, vii, viii, 1, 2, 3, 4, 6, 8, 12, 14, 16, 18, 20, 27,
30, 31, 32, 33, 34, 35, 39, 41, 44, 46, 47, 49, 52,
57, 66, 69, 70, 72, 73, 76, 77, 79, 80, 81, 82, 83,
84, 85, 104, 157, 179
human behavior, vii, 32, 49
human brain, 77
human cognition, 18
human development, 104
human genome, 72, 73
human intelligence, viii, 1, 2, 3, 4, 6, 8, 12, 14, 27,
32, 33, 66, 77, 79
human subjects, 27
Hunter, 12
hyperactivity, 56, 83, 174, 179
hypothesis, 18, 32, 33, 60, 61, 159, 175, 195
I
IBD, 76
iconic memory, 7
ideal, 41, 49
idealism, 49
206
Index
identification, 2, 7, 8, 96, 109, 167, 173, 176, 178,

181, 196
ideology, 49, 50
image, 3, 4, 13, 37, 47
immune system, 41
impairments, 58, 169
improvements, 195
impulses, 7
impulsivity, 34, 90, 91
in utero, 173, 177
in vivo, 31
inattention, 91, 106
income, 19, 20, 21, 22, 23, 36, 93
independence, 193
independent variable, 22
individual differences, 6, 13, 35, 36, 48, 51, 52
individual students, 90
individuality, 195
individuals, viii, 2, 7, 18, 19, 28, 72, 76, 78, 79, 81,
94, 111, 114, 119, 137, 156, 180, 188
inequality, 29
inertia, 193
infancy, 198, 199, 200
infants, 41, 179
infection, 74
inferences, 81
information processing, 15, 16, 67, 88
information processing speed, 67
informed consent, 20
inhibition, 88, 89, 103, 174, 176, 180
injuries, 177
insects, 71
insecurity, 186, 193, 196
instinct, 52
integration, 28, 81
intellect, 175
intellectual disabilities, v, 111, 126, 200
intelligence, vii, viii, 1, 2, 3, 4, 6, 7, 8, 9, 10, 11, 12,
13, 14, 15, 16, 17, 27, 30, 31, 32, 33, 34, 35, 36,
41, 44, 45, 47, 48, 49, 50, 52, 57, 66, 67, 68, 69,
70, 72, 74, 76, 77, 78, 79, 80, 81, 108, 116, 126,
127, 137, 165, 168, 169, 170, 172, 173, 174, 175,
176, 178, 179, 180, 181, 183, 200
intelligence quotient, 35, 169
intelligence scores, 78, 81
intelligence tests, 2, 3, 4, 8, 14, 45, 67, 126, 127,
170, 172
interdependence, 100
interface, 130
internal consistency, 59, 90, 91, 112, 113, 115, 117,
118, 125
internal working models, 197
internalizing, 199
interpersonal skills, 188

interpretability, 9, 171, 192
interrelatedness, 88
interrelations, 188, 195
intervention, 13, 15, 80, 87, 196, 197, 198
intron(s), 72, 78, 79
IQ scores, 4, 17, 19, 21, 24, 26, 27, 28, 79, 80, 113,
116, 121, 122, 125, 171, 172
isolation, 76, 170, 196
issues, viii, 10, 11, 13, 14, 67, 69, 124, 126
J
Jordan, 3
justification, 122
K
kindergarten, 89, 101, 102, 106, 107, 108, 189, 198,
200
kinship, 156, 157
L
labeling, 136
language development, 88, 108
latent variables model, 164
LEA, 107
lead, viii, 4, 39, 80, 82, 124, 173, 174, 175, 177, 186,
196
learners, 108
learning, 1, 3, 6, 8, 13, 14, 17, 27, 28, 29, 31, 34, 37,
41, 42, 43, 44, 48, 50, 51, 58, 88, 89, 100, 101,
103, 106, 107, 109, 110, 112, 125, 127, 173, 174,
177, 178, 181, 182
learning disabilities, 14, 58, 107, 177, 182
learning environment, 6
learning task, 43
left-hemisphere, 18
lesions, 177
level of education, 19, 23, 178
lifetime, 70
ligand, 27, 29
light, 70, 79
linear function, 158
literacy, 2, 88, 101, 107, 108
local conditions, 71
localization, 177
loci, 10, 70, 71, 72, 74, 76, 78, 82, 157
locomotor, 177
locus, 31, 72, 73, 74, 76, 78, 82
loneliness, 42, 107, 108
207
Index
longevity, 36, 52
longitudinal study, 106, 109, 199
long-term memory, 173
love, 42, 49
Luo, 7, 14
M
machinery, 41, 44
Mackintosh, 2, 4, 7, 8, 14, 32
magnitude, 120
major depression, 9, 11
majority, 4, 7, 47, 58, 72, 77, 88, 114, 124
malaria, 74
malingering, 117, 119, 121
maltreatment, 197
man, 13, 16, 167, 168
management, 48
manipulation, 6
marketability, 9
Mars, 32
Maryland, 112
material resources, 101
materials, 81, 132
mathematics, 9, 87, 92, 93, 102
matrix, 52, 60, 135, 137, 138, 139, 148, 149, 156,
157, 163
matter, 13, 41, 120, 158, 189
MCPH1, 72, 73
measurement, 14, 16, 48, 66, 67, 91, 94, 95, 97, 112,
115, 117, 120, 124, 125, 126, 150, 158, 167, 168
media, viii, 80, 124
median, 2, 8, 60, 74, 75
medical, 93
medication, 189
membership, 70
memory, 7, 8, 9, 18, 29, 38, 39, 45, 47, 48, 50, 51,
55, 56, 57, 58, 59, 60, 61, 63, 65, 88, 89, 105,
131, 133, 172, 173, 176
memory capacity, 48, 50
mental ability, 13, 166
mental health, 189
mental model, 35, 45, 46, 47, 49
mental planning, 174
mental retardation, 10, 58, 111, 125, 126
meritocracy, 80
meta-analysis, 9, 14, 66, 88, 89, 90, 114, 118, 186,
197, 200
metabolism, 18, 32
metaphor, 2
meter, 95
methodology, 65, 66
methylation, 85
Middle East, 72
military, 44
Minneapolis, 165
minorities, 93
misuse, 28
model specification, 105, 142
models, 3, 4, 6, 12, 35, 43, 45, 46, 47, 49, 50, 51, 52,
60, 63, 67, 70, 81, 88, 91, 92, 94, 95, 97, 104,
109, 136, 144, 148, 150, 151, 153, 155, 156, 157,
159, 162, 163, 164, 165, 166, 167, 170, 197, 199
modules, 36, 37, 38, 40, 46, 49, 50
morbidity, 155
motivation, 2, 34, 42, 56, 112, 118, 122, 174, 177
motor behavior, 44
MRI, 33, 34
mRNA, 18, 30
multidimensional, 87, 88, 90, 101, 170
multiple factors, 142
multiple regression, 81
multivariate statistics, 68
music, 42
mutations, 30, 77
myopia, 177
N
naming, 133
narratives, 192, 199
National Academy of Sciences, 30, 31
National Center for Education Statistics, 106
natural selection, 38, 50, 77
natural tasks, 43
nausea, 42, 43, 176
NCP, 141
negative consequences, 28
negative effects, 121
negative outcomes, 32
nerve, 7, 181
nervous system, 41
Netherlands, 55, 83, 105
neuromotor, 176
neurons, 45, 108
neuropsychological tests, 174
neuropsychology, 180
neuroscience, 50, 51, 104
neurotransmission, 30
neurotransmitter, 18, 27
neutral, 72
NFI, 96, 97, 98, 99, 100
nightmares, 176
normal children, 174, 177
normal distribution, 63, 132
North Africa, 74
208
Index
NRP, 93
nucleotides, 71
nucleus, 18, 27, 28, 29, 33, 57
null, 74, 162
nutrition, 81
O
obesity, 33
observable behavior, 157
Oceania, 72
oculomotor, 173
operant conditioning, 40, 41, 43, 44
operating system, 129
opportunities, 89, 90, 100, 101
organ(s), 44, 77
organism, 39, 40, 70
organize, 45
overlap, 2, 189
overtime, 91
P
pain, 42, 45
parallel, 36, 37, 45, 88
parallel processing, 36
parameter estimates, 60, 137, 139, 144, 156, 162
parental consent, 59
parenting, 185, 186, 187, 188, 189, 190, 191, 192,
193, 194, 195, 196, 197, 198, 199, 200
parenting styles, 197
parents, 19, 20, 21, 23, 43, 87, 88, 89, 90, 158, 174,
176, 178, 179, 187, 188, 190, 191, 192, 193, 199
participants, 7, 20, 21, 29, 48, 65, 190
partition, 156
path model, 158
pathology, 198
pathways, 46, 82
PCA, 192, 193
PCR, 21, 31
Pearson correlations, 193
pediatrician, 189
peer group, 108
peer relationship, 107
peer review, 80
penetrance, 77
perfectionism, 173
perinatal, 177, 178
perseverance, 171
personality, 2, 6, 11, 29, 35, 51, 52, 70, 167, 171,
173, 199
personality research, 35, 167
personality traits, 11
person-oriented approach, 185, 186, 188, 189, 196,
198
phenotype, 10, 19, 32, 34, 74, 77, 78, 79, 82, 156,
157, 158, 159, 163
phenotypic variability, 19
physical well-being, 88
physics, 37, 52
physiological, 179
pilot study, 33
plants, 41, 71
plasticity, 51, 53, 84, 174, 177, 178
Plato, 2
playing, 48, 173
pleasure, 27, 30, 42
policy, viii, 87, 101, 106
policy makers, 87
polymorphism, 18, 21, 22, 23, 27, 28, 29, 31, 32, 33,
34, 72, 82, 84, 85
population, 8, 20, 29, 34, 60, 65, 70, 71, 74, 76, 77,
78, 80, 113, 116, 121, 170, 171, 172, 188
population size, 78
positive feedback, 28
positive reinforcement, 28, 29
positron, 29
positron emission tomography (PET), 27, 29, 30
poverty, 81, 93, 106
predictive validity, 1, 9, 90, 91, 108
predictor variables, 22
prefrontal cortex, 45, 47, 51, 52, 101
pregnancy, 20, 177, 178
prematurity, 177
preschool, 87, 88, 89, 90, 93, 103, 104, 105, 106,
107, 185, 190, 194, 195, 196, 197, 198, 199
preschool children, 105, 107, 185, 194, 195, 196,
197
preschoolers, 107, 185, 186, 188, 191, 195
preservation, 178
president, 83, 87
prestige, 43
preterm infants, 178
prevention, 108, 196
primary school, 189, 191, 199
primate, 45
priming, 37
principal component analysis, 192
principles, 13, 16, 45, 49
prisons, 118
probability, 71, 124
problem behavior, 87, 90, 106, 197, 198
problem solving, 2, 9, 38, 42, 47, 93, 101, 174
professionals, 65
profit, 187
209
Index
programming, 48, 129
project, 14, 89, 104, 129, 130, 133, 137, 167, 189
proportionality, 149, 150
proposition, 47
psychological adaptations, 40
psychological development, 176
psychological processes, 109
psychologist, 2
psychology, viii, 6, 12, 35, 36, 38, 44, 46, 50, 51, 52,
59, 104, 106, 107, 126, 165
psychometric properties, 91, 114, 191
psychopathology, 165, 197
psychosocial stress, 82
psychosomatic, 175
public interest, 165
punishment, 49, 187, 196
Q
quality of life, 165
question mark, 132
questioning, 120, 175
questionnaire, 190, 199
R
race, 21, 22, 24, 33, 69, 70, 71, 76, 77, 80, 81, 82,
83, 84, 179
random mating, 158
rate of change, 94
rating scale, 105
reaction time, 6, 11, 13, 14, 165
reactions, 47, 187
reading, 9, 13, 48, 87, 88, 89, 90, 91, 92, 93, 102,
103, 105, 106, 108, 109, 124, 131
reading difficulties, 109
reading skills, 9, 88
reality, 70, 119, 123, 125
reasoning, 1, 18, 19, 45, 48, 52, 55, 56, 57, 58, 59,
60, 61, 63, 80, 172, 173, 185, 188, 190, 194, 196
recall, 173
receptors, 17, 18, 28, 29, 32
recognition, 45, 57, 63, 93, 108
recommendations, vii, viii, 111, 120
recovery, 60
reflexes, 176
regression, 139, 156, 165
regression analysis, 165
regression model, 156
regression weights, 139
reinforcement learning, 27, 28, 31
rejection, 38, 107
relatives, 78, 158, 159

relevance, 40, 185, 187, 195
reliability, 48, 57, 58, 59, 68, 93, 107, 113, 114, 115,
117, 118, 120
religion, 53
REM, 173
replication, viii, 11, 13
reproduction, 35, 36, 40, 44, 50, 70
requirements, 38, 40, 45, 46
researchers, vii, 3, 4, 6, 9, 58, 65, 87, 172, 173, 186
resilience, 107
resources, 44, 47
response, 10, 19, 28, 30, 33, 40, 47, 88, 89, 132, 175,
188
response time, 10
responsiveness, 187, 195
retardation, 108, 111, 112
reversal learning, 28, 31
right hemisphere, 173, 180
risk, 2, 8, 28, 76, 106, 177, 185, 186, 189, 194, 196,
198
RMSEA, 60, 61, 63, 96, 97, 98, 99, 100, 143, 146,
152, 154, 161, 162, 191
root, 88, 97, 148, 162
Root Mean Square Residual, 143, 146, 152, 154, 161
Royal Society, 52
rules, 37, 44, 60, 68, 147, 148, 176, 196
S
safety, 42, 49
sampling error, 114
sanctions, 49
Scandinavia, 118
scatter, 171
school, 2, 8, 14, 16, 59, 65, 87, 88, 89, 90, 91, 92,
100, 101, 103, 104, 105, 106, 107, 108, 109, 110,
111, 158, 163, 170, 174, 176, 178, 179, 189, 190,
199
school achievement, 14
school adjustment, 91
school examination scores, 2
school failure, 8, 106
school performance, 8
school psychology, 16, 59
school success, 110
schooling, 11, 102, 103, 189, 190
science, viii, 13, 84, 108, 165, 166
scientific method, vii
security, 42, 187, 192, 193, 195, 196
self-concept, 89
self-control, 91
self-efficacy, 89
210
self-interest, 49
self-regulation, 34, 88, 101
sensitivity, 28, 175, 196, 197
sequencing, 18, 84
services, 109, 111, 189
SES, 19, 20, 21, 22, 23, 24, 26, 27, 29
sex, 20, 29, 42, 43, 155, 175
sexual feelings, 42
shame, 42, 174
shape, 63
short-term memory, 12, 48, 51, 56, 57, 58, 90, 173
showing, 42, 121, 158
sibling, 19, 156, 158, 163, 164, 165, 166, 199
signals, 28, 40, 41, 42
significance level, 193
signs, 124, 151, 175
single-nucleotide polymorphism, 78
skewness, 61, 67
skin, 70, 74
sleep disorders, 170, 176
SNP, 18, 72, 73, 74, 75, 76, 78, 79
social adjustment, 112
social behavior, 80, 90, 91, 103, 107
social class, 19, 30, 49
social cognition, 52, 101
social competence, 88, 91, 104
social construct, 83, 84
social control, 49
social development, 89, 176
social exchange, 37, 50
social integration, 170, 176
social justice, 82
social learning, 40, 41, 43, 44
social life, 80
social organization, 70
social phobia, 13
social psychology, 51
social relations, 199
social sciences, 68
social skills, 87, 89, 90, 92, 104, 109
social status, 35, 36, 42, 49, 80, 81
socialization, 8
society, 29, 30, 53, 80, 170, 177
socioeconomic status, 80
sociology, 81
software, 129, 130, 164, 165, 167
solution, 36, 41, 42, 44, 48, 55, 63, 149, 185, 191,
193
Spain, 74, 83, 85
spatial ability, 2
spatial learning, 30
special education, 2, 58, 116, 126, 171
specialists, 100, 189
Index
speciation, 70, 71
species, 38, 43, 70, 71, 76
specific thinking, 174
specifications, 137
speculation, 119
speech, 48, 173
spelling, 174, 177
spindle, 73
split-half reliability, 113
Spring, 6, 9, 14, 81, 82
stability, 18, 30, 66, 114, 115, 118, 120, 127, 155,
171, 179, 182, 198, 200
standard deviation, 22, 25, 58, 61, 72, 80, 97, 113,
138, 147, 156
standard error, 60, 137
standardization, 6, 31, 34, 58, 63, 114, 115, 117, 118,
123, 170
states, 28, 32, 35, 43, 47, 49, 52, 71, 79, 87, 89, 101,
108, 111, 112, 118, 120, 121, 125, 174, 176
statistics, 63, 97, 99, 141, 142, 145, 158, 162, 193,
194
statutes, 112
stereotypes, 37
stimulation, 37, 177, 178
stimulus, 7, 41, 179
stomach, 176
storage, 7, 48, 108, 173
strategy use, 8
stratification, 31, 77
stress, 173, 175, 177, 186, 195
striatum, 28, 30, 33
structural equation modeling, 94, 105, 135, 165, 167
structure, 1, 2, 3, 4, 6, 8, 10, 11, 12, 13, 14, 15, 16,
41, 50, 56, 63, 66, 80, 106, 165, 198
student achievement, 90, 94, 103
style, 36, 198
subgroups, 71, 94, 102, 179, 185, 192, 193, 195
sub-Saharan Africa, 72, 74, 76
substitution, 65, 72
succession, 71, 175
Supreme Court, 111, 125
survival, 35, 40, 44, 50
symptoms, 176
syndrome, 30, 182
synthesis, 30
T
tactics, 44
target, 7, 141, 190
teachers, 87, 89, 90, 106, 190
techniques, vii, 21, 81, 196
technology(ies), vii, 6
211
Index
test items, 112, 114, 118
test scores, 7, 30, 33, 90, 112, 123, 126
testing, vii, 1, 2, 11, 13, 20, 35, 59, 65, 67, 81, 115,
118, 132, 150, 180, 197
testosterone, 173, 177, 181
textbook, 134
thalamus, 30
thoughts, 47, 48
three-dimensional object, 39
three-stratum model, 4
tics, 85
tides, 71
tinkerers, 42
top-down, 47, 49
trade, 145
traditions, 50
training, 16, 105
traits, 20, 36, 52, 70, 71, 74, 77, 81, 155, 166
trajectory, 106
transformation, 148, 149, 150
translation, 18, 156, 191
transmission, 43, 155, 173
traumatic experiences, 198
treatment, 172, 177, 189
twins, 19, 155, 156, 157, 158, 159, 160, 167
U
underlying mechanisms, 106
United Kingdom (UK), 11, 13, 33, 51, 104, 111, 116,
117, 127
United States (USA), 2, 30, 31, 65, 81, 129
V
validation, 115, 191
variables, vii, 6, 21, 22, 31, 57, 58, 59, 60, 61, 63,
94, 101, 112, 131, 135, 136, 137, 140, 143, 146,
150, 151, 152, 154, 160, 161, 162, 164, 178, 188,
190, 192, 193
variations, 79, 83, 156
varieties, 71
vector, 131, 133, 137, 138
velocity, 181
Verbal IQ, 10, 61, 194
violence, 49
vision, 36, 53
visual system, 36, 39
visualization, 7, 31
vocabulary, 9, 25, 33, 106
vomiting, 176
vulnerability, 178
W
waking, 176
Wales, 78
walking, 177
Washington, 51, 68, 104, 106, 107, 108, 109, 125,
126, 129
waste, 80
water, 44
watershed, 40
Watson, James, 81
weakness, 56, 92, 173
web, 84, 129, 132, 133, 135, 137
Wechsler Intelligence Scale, 4, 16, 17, 20, 34, 55,
56, 57, 66, 67, 68, 112, 127, 137, 167, 169, 171,
179, 183
welfare, 80
Western countries, 2
WISC subtest, 63, 126
Wisconsin, 19, 33, 179
withdrawal, 45
wood, 197
word recognition, 93, 109
working memory, 6, 12, 17, 18, 19, 27, 29, 34, 35,
37, 38, 39, 44, 45, 47, 48, 51, 55, 57, 63, 65, 67,
89, 90, 101, 110, 134, 172
worldwide, 72
worry, 172
Y
Yale University, 72
yield, 148

Inteligence Quotiend

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

Inteligence Quotiend

Încărcat de

Drepturi de autor:

Formate disponibile

NEUROSCIENCE RESEARCH PROGRESS

NEUROSCIENCE RESEARCH PROGRESS

NEUROSCIENCE RESEARCH PROGRESS

Copyright 2013 by Nova Science Publishers, Inc.

Library of Congress Cataloging-in-Publication Data

Library of Congress Control Number: 2013937394

Published by Nova Science Publishers, Inc. New York

Human General Intelligence as a Domain General

Subtests, Factors, and Constructs: What Is Being Measured

Race, Genomics and Intelligence: Slight Return

Multi-group Latent Growth Curve Modeling: The Influence

Error in the Measurement of Low IQ: Implications

Using R for the Analysis of Cognitive Ability and Behavior

Contribution of Intellectual, Psychological, Developmental

In: Intelligence Quotient

THE UTILIZATION OF INSPECTION TIME

DEFINING HUMAN INTELLIGENCE

The Utilization of Inspection Time as a Supplement to Standardized Tests ...

Figure 1. A model of g or general intelligence.

INTELLIGENCE AND IQ TESTS

Figure 2. A model of g with the Wechsler scales.

Figure 3. A higher order model using the Wechsler scales.

AN ALTERNATIVE TO TRADITIONAL IQ TESTS;

The Utilization of Inspection Time as a Supplement to Standardized Tests ...

IT correlate with standardized measures of intelligence, typically near a value of

CRITERION VALIDITY: USING IQ TESTS TO MAKE PREDICTIONS

The Utilization of Inspection Time as a Supplement to Standardized Tests ...

The Utilization of Inspection Time as a Supplement to Standardized Tests ...

The Utilization of Inspection Time as a Supplement to Standardized Tests ...

The Utilization of Inspection Time as a Supplement to Standardized Tests ...

In: Intelligence Quotient

IQ AND THE D2 DOPAMINE RECEPTOR GENE

Keywords: Addiction, dopamine, DRD2, intelligence, learning

DOPAMINE AND COGNITION

IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics

association of DRD2 Taq1A and N-methyl-D-aspartate receptor 2B subunit (GRIN2B) gene

SOCIOECONOMIC STATUS (SES) AND IQ

IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics

on the subcategories of Verbal Comprehension (VC), Freedom from Distractibility (FD),

DRD2/ANKK1 TaqI A Genotypes

Socioeconomic Status (SES)

Age and Gender

All values are mean standard deviation (SD).

IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics

Linkage Disequilibrium (LD)

Age and IQ Scores in the Total Sample by Racial/Ethnic Group

IQ Scores in the Total Sample by DRD2 Genotypes

IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics

Table 4. Verbal Comprehension (VC), Freedom from Distractibility (FD), Perceptual

All values are mean standard deviation (SD).

Childrens Index Scores by DRD2 Genotypes in Three Racial/Ethnic Groups

IQ Scores by DRD2 Genotypes within the Three Racial/Ethnic Groups

IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics

IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics

dopamine receptor D2 and catechol-O-methyl transferase genes and cognitive function.

IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics

IQ and the D2 Dopamine Receptor Gene in Children of Alcoholics

Plomin, R. (1999) Genetics and general cognitive ability. Nature 402(Suppl.):C25-C29

In: Intelligence Quotient

HUMAN GENERAL INTELLIGENCE AS A DOMAIN

THE MODULARITY DEBATE IN EVOLUTIONARY PSYCHOLOGY

In the area of intelligence, it is difficult to conceptualize individual differences as noise,

Human General Intelligence as a Domain General Psychological Adaptation