REFURBISHING OUR FOUNDATIONS

AMSTERDAM STUDIES IN THE THEORY AND

HISTORY OF LINGUISTIC SCIENCE

General Editor
E.F. KONRAD KOERNER
(University of Ottawa)

Series IV - CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board

Henning Andersen (Buffalo, N.Y.); Raimo Anttila (Los Angeles)

Thomas V. Gamkrelidze (Tbilisi); Hans-Heinrich Lieb (Berlin)
J. Peter Maher (Chicago); Ernst Pulgram (Ann Arbor, Mich.)
E.Wyn Roberts (Vancouver, B.C.); Danny Steinberg (Tokyo)

Volume 56

C.F. Hockett

Refurbishing our Foundations

 REFURBISHING
OUR FOUNDATIONS
ELEMENTARY LINGUISTICS FROM AN
ADVANCED POINT OF VIEW

C. F. H O C K E T T
Cornell University

JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA

1987
Library of Congress Cataloging in Publication Data
Hockett, Charles Francis.
Refurbishing our foundations.
(Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguis-
tic theory, ISSN 0304-0763; v. 56)
Bibliography: p.
Includes index.
1. Linguistics. I. Title. II. Series.
P121.H635 1987 410 87-29961
ISBN 90 272 3550 3 (alk. paper)
® Copyright 1987 - John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other
means, without written permission from the publisher.
 PREFACE
This essay is a sort of inverted Festschrift (a Schriftfest? a Tfirhcstsef? a
in honor of my most exacting instructors: my students. It is
intended especially for my colleagues —many of them former students ( * ) —
Frederick B. Agard, Edward L. Blansitt Jr, *Ann Bodine, Dwight L Bolinger, * J .
Marvin Brown, William M. Christie Jr, * James E. Copeland, *Ronald Cosper,
*Philip W. Davis, *Robert J. DiPietro, *James W. Gair, Paul L Garvin, Toby D.
Griffen, Robert A. Hall Jr, Michael A. K. Halliday, Eric P. Hamp, Henry M.
Hoenigswald, Dell Hymes, *Ashok Keikar, Gerald B. Kelley, *James Kilbury, *D.
Robert Ladd Jr, Adam Makkai, Valerie Becker Makkai, *WilIiam R. Merrifield,
*Rocky V. Miranda, William B. Moulton, *David L Olmsted, *Dennis E. Peacock,
*Velma B. Pickett, Kenneth L. Pike, Ernst Pulgram, *Albert J. Schütz, Hugh M.
Stimson, Lenora A. Timm, the late W. Freeman Twaddell, *Ralph Vanderslice,
the late Carl F. Voegelin, Florence M. Voegelin, *Willard Walker, *Stephen Wal­
lace, Linda R. Waugh, Roger W. Wescott, and *John U. Wolff, as a token of my
appreciation for their recent joint enterprise in my behalf [Essays in Honor
of Char/as F. Hockett (Leiden: E. J. Brill, 1983)].
In October of 1984 an earlier version of this material was presented as a
lecture series entitled "Refurbishing the Foundations of Linguistics," under
the auspices of the Department of Linguistics and Semiotics of Rice
University, Houston TX, Those in attendance at some or all of these lectures
included Sydney M. Lamb (the chairman of the department), Lily Chen, James
E. Copeland, Philip W. Davis, Lane Kauffmann, Douglas Mitchell, Stephen Tyler;
Stephen Wallace; Katharina Barbe, Elizabeth Cummings-Culliton, Cynthia Ed-
miston, Don Hardy, Lillian Huang, Daniel Mailman, Alan Rister; and, on one
happy occasion when he was able to fly down from Austin, Winfred P.
Lehmann. Their questions and comments proved invaluable both in revealing
errors and gaps of which I had been unaware and in suggesting ways to
correct them. It goes almost, but not quite, without saying that remaining
deficiencies are all my own.
In writing this essay I have proceeded somewhat as though preparing a new
course for my old students, treating afresh, with the benefit of more
experience and further pondering, many of the topics we discussed during
their apprenticeships. In keeping with that, and reflecting the atmosphere WE
used to establish together in the classroom, throughout the essay the
pronoun "you" — conveniently ambiguous as to number — means the read­
er(s), and "we" is the INCLUSIVE first-person plural. In the same spirit, through
all stages of preparation the subtitle appeared as Elementary Linguistics
vi PREFACE

from an Advanced Point of View Out, not just an allusion to but also a takeoff
on Felix Klein's famous Elementarmathematik von höheren Standpunkte aus
(Leipzig: . G. Teubner, 1908-9). But at the last moment I could not quite
muster the audacity to put it into print that way, so out dropped out. In either
form the subtitle is appropriate. We deal here with elementary issues, but I am
not addressing novices. I speak to my fellow sophisticates, asking that for
the nonce they join me in setting aside a measure of that sophistication, that
together we may explore a new and different path.
So as not to interrupt the flow of the exposition, most critical apparatus,
together with a fair amount of subsidiary commentary, is relegated to the
notes. The phenomenal expansion of the ranks of professional linguists in
recent years makes it ever more likely that different investigators will
independently think some of the same thoughts, but also increasingly difficult
to know when that has happened — both because the volume of literature is
unmanageable and because the terminologies of different schools of lin­
guistics verge on mutual unintelligibility. I am sure that many of the ideas I
present as my own in this essay will sound to some readers like echoes. I
have no desire to slight anyone, and will be delighted to acknowledge, with
apologies for my ignorance or oversight, any omissions of credit called to my
attention.
C F. Hockett
145 North Sunset Dr
Ithaca NY 14850 USA
My 1987
 CONTENTS
0. Introduction 1
0.1. Attitude 1
0.2. Angle of Approach 1
0.3. Prospectus 2
1. The Shape of Speech 4
1.1. The Dimensions of Syntax 4
1.2. Order 4
1.3. Spacing 5
1.4. Direction: More Dimensions 6
1.5. Iconic and Arbitrary 9
1.6. Iconicity and Dimensionality 10
1.7. Iconicity in Language 10
1.8. The Proximity Principle 12
1.9. Tension 13
1.10. Multidimensional Language? 13
1.11. Summary 15
2. Hearing Utterances 16
2.1. Depth Perception 16
2.2. The "Syntactic Sense" 17
2.3. A Learning Demonstration 18
2.4. Examples 19
2.5. Old Wine in New Bottles 21
2.6. Summary 24
3. The Hearer's Evidence 26
3.1. Introduction 26
3.2. A Classification 26
3.3. Constituents of the Utterance 27
3.4. The Roles of the Ingredients 29
3.5. Lessons from Written English 29
3.6. Lessons from Headlines 32
3.7. The Hat Game 33
3.8. Inferences 35
3.9. Being Realistic about Language 35
viii CONTENTS

4. Hearing Words 37
4.1. Modes of Listening 37
4.2. The Task of Identifying Words 37
4.3. Strategies for Word-Identification 39
4.4. Implicit Motor-Matching 39
4.5. Gestalt Perception 40
4.6. Code-Switching 43
4.7. Norm Shapes 44
4.8. The Evidence of Borrowing 45
4.9. Summary 47
5. Sounds, Words, and Redundancy 48
5.1. The Phonemic Theory 48
5.2. Words 50
5.3. The Word in Phonology 52
5.4. Slurvian 57
5.5. The Uses of Redundancy 60
5.6. Summary 63
6. Why Morphemics Won't Work 65
6.1. A Fish Story 65
6.2. Allusions 66
6.3. Idiomatic Phrases and Tongue-Twisters 67
6.4. More Puns 68
6.5. Onomatopoeia and Secondary Associations 68
6.6. Blends and Lapses 71
6.7. Malapropisnns, Doublets, Contamination, and Taboo 72
6.8. Metanalysis and Abbreviation 73
6.9. Naming 75
7. From Particle to Resonance 77
7.1. The Road to Morphemics: Bopp 77
7.2. The Road to Morphemics: Morphophonemics 79
7.3. The Morpheme in America 81
7.4. The Great Agglutinative Fraud 82
7.5. The Transducer Fallacy 84
7.6. Repair or Replace? 86
7.7. How Does an Utterance Mean? 90
7.8. Syntax, Morphology, Syntagmatics, and Paradigmatics 95
7.9. Summary 96
CONTENTS ix

8. How Language Means 97

8.1. The Logic of Resonance Theory 97
8.2. Social and Individual 99
8.3. The Meanings of Meaning: Reference and Denotation 103
8.4. Mental Images 105
8.5. The Meanings of Meaning: Denotations and Connotations 108
8.6. Internal Storage 110
8.7. Summary 113
9. The Craft of Speaking 114
9.1. Hearing and Speaking; Choice and Chance 114
9.2. The Work Space 115
9.3. Choosing Words 117
9.4. State Parameters 119
9.5. An Example 122
9.6. Analysis 126
9.7. Generalization 129
Notes and Commentary 131
References 162
Index 172
 0. INTRODUCTION
0.1. Attitude. No one in any culture known to us denies the importance of
language. Partly because it is important, partly just because, like Mount
Everest, it is there, we should like to know how it works. To that end, people
from time immemorial have examined it or speculated about it, trying to
come up with cogent commentary.
What one sees of language, as of anything, depends on the angle of view,
and different explorers approach from different directions. Unfortunately,
sometimes they become so enamored of their particular approach that they
incline to scoff at any other, so that instead of everybody being the richer
for the variety, everybody loses. That attitude has been called the "eclipsing
stance." The early followers of Noam Chomsky adopted this stance, but they
were by no means the first: some of us post-Bloomfieldians came close to it
in the 1940s (though Leonard Bloomfield himself never did), and so, appar­
ently, did the Junggrammatiker in the late 1870s. But it is a wrong position to
take, even toward those who have themselves assumed it. It is obviously
impossible to see all of anything from a single vantage point. So it is never
inappropriate to seek new perspectives, and always unseemly to derogate
those favored by others. Or, to use a different figure: the blind man touching
the tail has reason to say an elephant is like a rope, but no right to claim an
elephant is not also like a wall or a tree-trunk or a snake.
I don't mean we shouldn't be critical. I do mean we should try to be most
wary just of those propositions that we ourselves hold, or have held, closest
to our hearts — above all, those we come to realize we have been taking for
granted. Scientific hypotheses are formulated not to be protected but to be
attacked. The good hypothesis defends itself, needing no help from enthusi­
astic partisans.
0.2. Angle of Approach. Our title speaks of the "foundations" of linguis­
tics. Where do we find them?
That question has received answers of two sharply different sorts. Some
investigators ally natural language with logic and mathematics, and place the
foundations of all three in an ideal world of pure logic. Others see language as
a feature of everyday human conduct, and believe that the foundations of our
discipline must be empirical in the same way as are those of biology and
physics. These two views have both been with us for a long time: witness, for
example, the disagreement a century ago between the rationalist Heymann
Steinthal and the empiricist William Dwight Whitney. Although I have a clear
preference, in keeping with the attitude described in §0.1 I do not here ven­
ture to declare that one of the views is right and the other wrong, nor even
that both are right (or both wrong). Instead, I simply announce that the one
adhered to throughout this essay is the second.
But that still leaves considerable choice as to how we proceed.
2 INTRODUCTION

Among investigators who take the empirical approach, it is a habit of long

standing to refer to the members of a speech community as SPEAKERS. In
conformance with that usage, and perhaps partly because of it, the discus­
sion of language design has usually focused on the producer of utterances
rather than on any other participant in the communication process. Chomsky
hinted at a more balanced view some years ago when he replaced the term
language-SPEAKERS by language-USERS, but he seems never to have followed
up on the possible consequences. Here we do: we consider language-in-action
primarily from the point of view not of the speaker but of the HEARER.
To be sure, this approach is not unprecedented, but it has generally
received short shrift. I am convinced it has much to teach us. For me it has
clarified several issues that would otherwise have remained puzzling. On the
other hand, some aspects of language cannot be discerned at all from this
perspective, since they are hidden behind more prominent features in the
foreground. For those we resort, as necessary, to other approaches.
Our fundamental question can be phrased as follows: WHEN WE HEAR SOMEONE
SAY SOMETHING IN A LANGUAGE WE KNOW, HOW DO WE KNOW WHAT IS SAID?
That may suggest we are headed into a study of the psychology of lan­
guage rather than of language itself. I prefer to say only that we are moving
IN THE DIRECTION of the psychology of language. I think one can go a long wav
in that direction before one finally reaches anything to which the psychol­
ogists would stake an exclusive claim, and I shall not intend to go that far.
It is of course possible, and perfectly legitimate, to define linguistics in
such a narrow way as to preclude our question (surely that was Bloomfield's
mature preference). Our heritage of two centuries of research and field
experience enables us, today, to observe, analyze, and describe any language
in the world in just as much detail and with just as much precision as we wish
and have time for. On the basis of such descriptions we can generalize and
typologize; by comparing certain of them in a different way we can recon­
struct ancestral stages and describe how specific languages have changed in
course of time; and from such diachronic data we have achieved a firm
understanding (except for one point still in dispute) of the major kinds and
mechanisms of language change. What is more, all of this can be done, and
has been done, without seemingly saying anything at all about the properties
of the organisms that manifest language behavior.
Between linguistics, thus narrowly conceived, and psychology in an equally
constrained sense, there is a region for which I have no name but which I
think is the appropriate place for the development of a THEORY OF HEARING
AND SPEAKING. My aim here is to work toward the formulation of such a theory.
The staging area for the venture is just linguistics: that is, no psychology —
at any rate, none of the special assumptions of any one school of psychology
— but just our understanding (however limited it still remains) of how
language works. The ability to use language implies the possession of certain
capacities. Our move away from linguistics proper in the direction of psy­
chology is limited to an attempt to infer what those capacities must be.

0.3. Prospectus. Terminology is always a problem. For researchers, writers,

and readers alike, an efficient terminological system is a well engineered road
INTRODUCTION 3

that speeds them on their way — provided they want to go where it takes
them. For those who prefer to explore cross-country it is not a help but a
barrier. I have therefore assiduously avoided the verbal and diagrammatic
frames of reference of any contemporary school of linguistic theory, choos­
ing instead the most time-worn and colorless terms I could find, using as few
of them as possible, and trying to introduce each in a serviceable defining
context.
But at the outset there are a few common terms that we must take for
granted (subject to subsequent revision and refinement). In particular,
everybody knows what a WORD is, and the lay notion, though vague, will serve
our needs until deep into the exposition. The traditional conceptions of
SYNTAX and MORPHOLOGY assign to the former the ways in which words are
used in and as utterances, to the latter the patterns of formation of
individual words; these definitions will also work for us, even though the
boundary between the two compartments can be no sharper than our definition
of "word."
We start with syntax (chapters 1-3); go to words (4-6), to morphology
(6-7), to meaning (8), and finally shift the focus briefly from listening to
speaking (9). In chapter 5, when the main line of argument has properly set
the stage for it, we also turn our attention briefly to phonology.
You will find one major innovation in the treatment of syntax, one in the
discussion of morphology, and considerable turmoil about phonology.
The new slant on syntax is a realistic alternative to the cumbersome and
(in my view) arbitrary machinery of transformations and their progeny in the
diversified transformational-generative traditions. I was put on its trail about
ten years ago when I found myself asking the seemingly very odd question,
"WHERE is deep structure?" — only to realize that I had had the essence of
the answer in the early 1950s, several years before I invented the term "deep
structure," but had then had my attention diverted.
The treatment of morphology here is a corrective for what I think has been
a fundamental error in our theorizing (on this side of the Atlantic) for several
decades; my suspicions about this also date from about 1950.
The phonological questions in chapter 5 are ones I long thought I had
answered definitively, even if not to everyone's satisfaction, years ago. But
the new angle of approach has forced me to abandon those answers, without
being sure how to replace them. In the chapter I set forth what is involved
and suggest tentative conclusions; fortunately, the residual uncertainty in
this part of the essay does not much affect the argument in the rest.
 1. THE SHAPE OF SPEECH
1.1. The Dimensions of Syntax. In classical Greek, the term
whence our modern syntax, meant a setting-out in orderly array of either
soldiers or words.
The order one can achieve depends in part on the site. On a parade ground,
troops can be maneuvered in neat rows and columns. When the Persian army
reached Thermopylae in 480  nothing like that was possible; the pass was
too narrow. On a plain or at sea, one can try to encircle the enemy. In star
wars each ship has three degrees of freedom of motion instead of two, so
encirclement is meaningless; only englobement will do.
These are all matters of SYNTACTIC DIMENSIONALITY: the geometrical
properties of the space in which soldiers, or words, are to be deployed.
Speech is one-dimensional. No matter how large our army of words, in
speaking we dispatch them through a Thermopylae so narrow that they must
go single file. Given two words, U and V, the only possibilities are that U
precede V and that V precede U. They cannot be simultaneous; they cannot
be lined up in space instead of in time.
1.2. Order. That it can matter which comes first is clear from pairs like
these:
match book: book match
red hot : hot rod
follows page 18b : page 18b follows
race horse : horse race
dog bites man : man bites dog
Of course, languages are not all like English. Latin is sometimes said to
have had "free" word order. It is true that whether one said canis virum
mordet 'the dog bites the man' or interchanged the words 'dog' and 'man' and
said virum canis mordtet, the identifications of biter and of bitten remained
the same. (To reverse the roles, one altered not the word order but the
individual words, changing canis to canem and virum to vir) But that does not
mean that word order in Latin was of no importance. There is more to
meaning than just the identification of the actor and the goal of an action.
The two Latin sentences differed in emphasis, and the speaker uttered the
one or the other according to what emphasis was wanted.
Actually, order is never totally irrelevant — a point we might as well make
right now and get out of the way. To be sure, by an act of will a difference
can be set aside for specific purposes. Such an act is demanded of us in
 THE SHAPE OF SPEECH 5

childhood when we learn arithmetic. The teacher insists that two and three
and three and two are the same. And so they are — but that is a very special
NUMERICAL sameness, achieved precisely by agreeing to ignore anything in
which the two expressions differ. In other contexts those differences come to
the fore. Suppose someone asks you how many children you have, that you
answer five, and that they then say How many boys and how many girls?
Unless in fact you have two sons and three daughters, it would be deceitful to
respond with Two and three.
1.3. Spacing. In speech, given syntactic linearity, so that UV and VU are the
only possibilities, and assuming that no third word intervenes, there is
nevertheless an additional factor: U and V can be closer together or farther
apart. How many degrees of separation are distinctively different may depend
on the language. I suppose that if they are excessively far apart —say, U
today and V tomorrow — then they are not likely to be perceived as having
anything to do with each other. Setting aside that extreme, it is not hard to
show that English has at least a two-way contrast:
If you're fat, don't eat; fast. : If you're thin, don't eat fast.
I don't think; I know! 1 don't think 1 knov either.
Thwillfindthem enormously They willfindthem enormously
destructive and maybe cruel to destructive and may be cruel to
their pets. their pets.
The largest and most luminous The largest and most luminous
objects known, galaxies and objects known, galaxies and
quasars were discovered only quasars, were discovered only
in the past fifty years in the past fifty years.
1 saw the two women, the Isawthe tvo vomeri, the
mother and the daughter. mother, and the daughter.
(Two people in all) (Four people in all)
One nice hot June day, the day One nice hot June day, the day
after I graduated from high after I graduated from high
school, I told her to go to school, I told her to go to
hell, and beat it. hell and beat it
Especially amusing is this one:
That bright red rose—I see its That bright red rose I see; its
thorn. I disregard the scent thorn I disregard.Thescent
it gives off—that's nothing! it gives off, that's nothing I
1 hate the scratches I got be­ hate. The scratches I got; be­
fore; I fuss about the pain, too. fore I fuss about the pain too
Much I think of the beauty! much, I think of the beauty.
6 1 THE SHAPE OF SPEECH

1.4. Direction; More Dimensions. We have compared words being uttered

with soldiers passing through a defile. But there is a crucial difference: the
soldiers can turn around and retreat; the words can't. In speech you can move
in only one direction: from earlier to later. The syntax of speech is not only
unidimensional but also UNIDIRECTIONAL. When, henceforth, we characterize
speech (or any other system) as LINEAR, we shall mean just this combination,
unidimensionality and unidirectionality.
There are communicative systems that are like speech in their syntactic
dimensionality; for example,W e s tAfrican drum-signaling. There are also sys­
tems that are very different.
Consider writing and drawing. A sheet of paper, or the wall of a cave, af­
fords a flat expanse on which marks can be set out: two dimensions instead
of the one of speech, but both of them spatial, with no inherent direction. The
ears can "move" only from earlier words to later ones; the eyes can turn
from any part of a graphic display to any other part—and back again. That is
because speech, happening in time, vanishes as it happens, whereas marks on
a flat surface are more or less durable.
Writing exploits the durability, but not the dimensionality. To make a writ­
ing system work, special conventions have to be established that nullify much
of the geometric potential. The surface is divided into narrow strips, and a
single line of marks is inscribed in each strip. The eyes are not to flit freely
from point to point, but are to examine one strip at a time, scanning the
marks in it in a specified direction and moving from strip to strip in an
equally arbitrary but conventionally fixed order. Of course, the eyes can skip
"back" to reread an "earlier" passage — and the words in quotation marks
give the game away: the conventions have linearized the display, thus forcing
its geometric properties FOR A READER to match those of speech. There is no
clearer evidence that writing is, at bottom, derived from speech than the in-
escapability of this imposed reduction of dimensionality.
Not all the marks we put on surfaces are writing in the strict technical
sense of a representation of speech, and some of them are very different
indeed. Yet even in some of our surface-marking practices that seem most
closely allied to writing, we exploit syntactic two-dimensionality.
In mathematical notation, two-dimensional arrays are commonplace, and if
any two entries in such an array are interchanged the meaning (whatever it is
in the particular context) is altered. Thus, no two of these are equivalent:
a b    d\
c d ac  b

Likewise, the following four are all different, though, except for the enclosing
parentheses in the last, they involve the same pair of letters:
1.THE SHAPE OF SPEECH 7

The first, "a-sub-n", would usually be the label for the last of a series of η
things. The second would in most contexts denote the product of a and n.
The third means a to the 17th power: n a's multiplied together. The last is one
of the standard notations for the number of combinations of n things taken a
at a time. Again: in a few cases vertical alignment means 'or': thus, " < " 'less
than' and " = " 'equal to' are combined in to yield the meaning 'less than
or equal to'.
But one need not turn to anything even as slightly esoteric as math­
ematics for examples of two-dimensional syntax. It is obvious, though of no
profound importance, in such forms of graphic play as crossword puzzles and
acrostics, and in verses where the first letters of the successive words spell
a significant word, or in which the words along the major diagonal (first word
of first line, second of second, and so on) form a key sentence, or the like.
Lewis Carroll's poem "The Mouse's Tail" (in Alice) was originally written in
lines arranged to look like the tail of a mouse, thus adding a visual pun to the
verbal one (figure 1).
Much nearer home, look at any addressed envelope (figure 2). An inter­
change of the location on the envelope of the two addresses would lead the
postal service to deliver the letter to the sender instead of to the intended
recipient. Until the stamp is canceled, the arrangement AND NOTHING ELSE
specifies who is transmitting the letter to whom. This arrangement, with the
meaning all agree on, is a CONSTRUCTION, just as is the arrangement of
adjective first and noun second, the adjective modifying the noun, in black
cot. At the same time, the envelope shows us how formal constructions
merge imperceptibly into the "mnemonics of location" of everyday life: the
meaning of a business letter depends on whether it is in the in-basket, the
out-basket, or the waste-basket; if we find a string on our finger we try to
recall what we put it there to remind us of, but a string in a kitchen drawer
gives rise to no such cogitation.
Beyond the two-dimensionality of surfaces we come to the purely spatial
three-dimensionality of sculpture and architecture, and to the different three
dimensions (two of space, plus time) of the silent screen. Beyond these is the
four-dimensionality of pantomime, of parlor charades, of the football officiai
signaling an infraction and penalty, and of such gestural-visual systems as
Ameslan (American Sign Language for the Deaf). And to any of these except
the last, one may append a "sound track," increasing the dimensionality even
more.
β
1. THE SHAPE OF SPEECH

Figure 1. Carroll's originel handwritten version of "The Mouse's Tail"

l.THE SHAPE OF SPEECH 9

Leonard Bloomfield
Sweet Grass Reserve
c/o Battleford Agency
Battleford Saskatchewan

Dr Edward Sapir
National Museum of Canada
Ottawa Ontario

Figure 2.

1.5, Iconic and Arbitrary. For a given individual or community C, if A

connotes, denotes, designates, identifies, is an index of, indicates, points to,
refers to, reminds  of, represents, signifies, stands for, suggests, or
symbolizes B, then A and B stand in a SEMANTIC relation: A MEANS;  is MEANT.
Each of A and  may stand also in other such relations, and the variety of
verbs just used shows that semantic relations are of divers sorts. We bypass
the subtleties and complexities for the present not because they are not
important (see chapter 8), but because they are irrelevant for our immediate
purpose. Note that we take all semantic relations to be EXTRINSIC, mediated by
C; that is, remove  and the relation vanishes, even if A and  are in
themselves unaffected. For example — as we all know but all occasionally
forget — the fact that the sound of the word soft means salt is not a property
of either the sound or the substance; it is a property of users of English.
If a particular A and  stand in a semantic relation for some C, then the
relation is ICONIC to the extent that C's perception of A resembles the
perception of B, and ARBITRARY to the extent that it does not.
Presumably no semantic relation could be totally iconic, for then A and 
would be indistinguishable. An exact replica of a tiger would roar and might
eat you. But some semantic relations are less arbitrary and some are more
so. A photographic portrait is more iconic than a cartoon caricature. In trying
to track down a fugitive, the sheriff gets more help from a picture than from
a name.
10 1. THE SHAPE OF SPEECH

1.6. Iconicity and Dimensionality. Iconicity is related to syntactic dimen-

sionality as follows. The real world, the universe of Bs, is multidimensional.
Only an equally multidimensional art form could portray reality with maximum
fidelity. And, in fact, live drama on the stage, for all its conventions and
deceptions, is the closest currently achievable imitation of life. Among other
things, it is the only medium in which people are represented by people. There
can also be a lot of iconicity in representations with only two spatial
dimensions. In the bulk of our visual experience we are accustomed to
"reading in" depth, so we don't pay much attention to the fact that a picture
of an elephant has usually not even the thickness of an elephant's skin, and if
time and sound are added, as in the cinema and television, the illusion can be
very vivid indeed.
Yet every reduction of dimensionality cuts something out. A statue of an
elephant may be unmistakeably recognizable, but it does not move. A
two-dimensional projection still retains features that distinguish it from a
comparable projection of a tiger or a woodshed, but it must not be viewed
from too flat an angle (figure 3). And in one-dimensional projection all
distinctive features are lost: nothing remains but abutting line segments.
Accordingly, if a system of constrained syntactic dimensionality nev­
ertheless has fidelity and power, as language so obviously does, it is
necessarily the case that its semantics is largely arbitrary.
1.7. Iconicity in Language. That is not to deny that there are traces of
iconicity in speech.
In the first place, since a spoken word or phrase is perceived as a sound, it
can mean iconically, in the strictest sense, only if that which is meant is also
a sound. Apparently every language has a few words that meet these spec­
ifications. That different languages have somewhat different forms for the
same meaning (English ding-dong = German bim-bam) signals a measure of
arbitrariness, but we already know we shall not find totally untainted iconism
anywhere. Perhaps the closest approach to it in speech is when, in the
discussion of language, we use the sound of a form as a name for the form,
speaking of "the suffix -ing** or "the word sing"
Secondly, if a word sounds like a sound made by, or otherwise associated
with, the thing it means (ho-ho 'train'), there is at least a mnemonically
useful tinge of iconicity.
Thirdly, if there really is such a phenomenon as synaesthesia, whereby
some sights, smells, tastes, or touches are more like some sounds, others
more like others, then a word can be indirectly iconic if it means a sight,
smell, taste, or touch that is like the kind of sound its sound resembles.
1. THE SHAPE OF SPEECH 11

A. Two-dimensional projection of an elephant.

. One-dimensional projection of an elephant.

Q One-dimensional projection of a woodshed.

Figure 3.
Fourthly: at a much larger size-level, a narration is iconic to the extent
that events are recounted in the sequence in which they occurred. This
iconicity is abandoned in the device of "flashback," though one might claim
that it is regained in another way, in that the account follows the temporal
sequence of some character's thoughts.
Even when a narrative sequence is faithful, another departure from
iconicity is often inevitable, in that the time scale for the recounting of
successive events cannot be related in any constant way to the time scale of
12 l.THE SHAPE OF SPEECH

the events. A character enters a room, takes the situation in at a glance, and
speaks. The entrance and the speaking can be described quickly; the state of
affairs in the room may require a paragraph or a page to set forth. Novelists
worry about this and seek ways of minimizing the loss of realism, but it is a
remarkable fact that as listeners or readers we adapt our time-scale quite
readily to accord with the narrator's needs.
1.8. The Proximity Principle. More pervasive than any of the foregoing is
an implication of iconicity that I shall call the PROXIMITY PRINCIPLE: if the
meanings of A and  belong together, then A and  belong together. To
portray a black cat, a painter does not normally sketch a colorless cat on one
part of his canvas and put a blob of black somewhere else (figure 4).

Figure 4 Portrait of a black cat

Likewise, in speech the most natural way to indicate that a cat is black is to
put the words next to each other and say black cat (or chat noir, in the other
possible order), rather than putting cot (chat) one place in an utterance and
black (noir) somewhere else.
That assumes that words can be no closer than adjacent. In fact, there are
two devices by which simultaneity can be faked. One is to try to say both
words at once and see what happens. Lewis Carroll tells of a character who,
when his life depended on choosing properly between Richard and William, in
desperation came out with Rilchiam. So instead of black cat one might say
blat or cock. Unfortunately, there is no guarantee that this device will elicit a
favorable response. The other method requires leisure for establishing con­
ventions in advance: you introduce a technical term— say, felix— defined to
mean 'black cat', and then use that single word instead of the two words in
either order.
We resort to both of these devices on occasion; yet most of the time the
closest two words can be is for one to follow on the heels of the other
1. THE SHAPE OF SPEECH 13

through the defile.

If our cat is not only black, but also small, then in French one can say petit
chat noir, following the proximity principle with both adjectives. English does
not allow this. In a phrase like former adjutant general, for example, generai
modifies adjutant but former modifies the whole phrase adjutant general, and
that is not the situation we are talking about. So we are forced to say liftte
black cat, with one of the adjectives separated from the noun, even if our
intention is that both adjectives should directly modify the noun.
1.9. Tension. It is interesting to note how the proximity principle some­
times comes into direct conflict with some habit of word placement.
Consider the English adverbs just and only. For many of us there is a
difference between You can't just sit there! and You just can't sit there! But
quite generally the second order is used for both meanings: the mechanical
habit of placing such a restricting adverb directly before the finite verb wins
out over the insertion of the adverb directly before what it modifies. On the
other hand, despite purist admonitions not to "split" our infinitives, an
expression like Now we'll try to quickly review these points is far commoner—
and has been for perhaps two hundred years, if my desultory observations of
nineteenth-century expository prose can be relied on— than the alternative
Now well try quickly to review .... Even the little marker to is not allowed to
separate the modifying adverb and the modified verb: in this case proximity
wins out over mechanical rule.
1.10. Multidimensional Language? In any case, a word has only two sides, a
starting side and an ending side. What can one do if there are three or more
adjectives for one noun?
If speech were two-dimensional and all words were square and of uniform
size, then we could squeeze four adjectives around a noun, as shown on the
left below; or, if we didn't care that some of the adjectives shared boundaries,
we could crowd in six, as on the right:
14 l.THE SHAPE OF SPEECH

Note that the two-dimensional noun phrase on the right defines half a Boy
Scout. There is no place even in two dimensions — by the conventions just
followed — for the other half. To provide for those, we could expand to three
dimensions with words as cubes, getting room not only for all twelve epithets
but even for a couple of extras. (If my calculation is correct, uniformly sized
tesseractic words in a language with four spatial dimensions would allow
thirty satellites to snuggle around a single nucleus.) Of course, if confined to
two dimensions we could cunningly make adjectives smaller than nouns
(though just as square), to obtain the following:

Given the poverty-stricken dimensionality of language as it actually is, I

have been unable to think of any scheme for getting all twelve adjectives
next to the same noun, short of hiring a chorus of a dozen and having them
declaim, in unison except at one point,
1.THE SHAPE OF SPEECH 15

That would seem to violate the claim made earlier that two words cannot pass
abreast through the Thermopylae of speech; here not two but a dozen would
be simultaneous. But they would also be unintelligible! And wouldn't they be
passing through twelve Thermopylae instead of just one? Our claim seems
secure.

1.11. Summary. Setting aside all these fantasies, including the choral
fantasy, let us turn back to language as it really is and summarize.
The aim in this first chapter has been to establish a frame of reference
within which we can recast our underlying question (§0.2) in a maximally
productive way. I think we can now do that.
The shape of speech is not the shape of the world we speak about.
Therefore the semantics of language must be largely arbitrary, only
minimally iconic. Even the proximity principle can come into play only in small
measure: given the linearity of speech, it constantly happens that words
whose meanings belong together cannot be spoken together.
Yet hearers understand.
How do hearers know when words uttered apart are to be interpreted to­
gether? How do they know when the adjacency of two words is to be ignored?
WHAT IS THE NATURE OF THE COLLUSION BETWEEN THE STRUCTURE OF UTTERANCES
AND THE STRATEGY OF LISTENERS BY VIRTUE OF WHICH CORRECT INTERPRETATION AND
UNDERSTANDING ARE POSSIBLE?
 2 HEARING UTTERANCES
2.1. Depth Perception. To zero in toward an answer to our basic question,
as it was reformulated at the end of the preceding chapter (§1.11), we begin
here with a comparison of the experience of listeners with that of viewers.
Whether we contemplate a real landscape or a painting, the image on the
retina is two-dimensional. But between there and the brain the incoming data
are processed in such a way that we perceive both in three dimensions.
Our visual habit of supplying depth when it is not objectively there is
perhaps brought into awareness most vividly by two-dimensional designs that
allow alternative interpretations:

Almost at will, we can see either of two faces of this figure as the closer. To
perceive the figure merely as an assemblage of line segments on a flat
surface, though possible, is much more difficult.
As part of the answer to our question', I propose that when we listen to
speech in a language we know our processing of the input is something like
our perception of pictures. A picture is flat, but WB see it in depth; speech is
linear, but we do not hear it that way. We register the actual linear sequence
of the words (which, as we insisted in §1.2, is always distinctive), but we also
"read in" diverse groupings and connections-at-a-distance without which the
words would make no sense. By a convention that is only partly metaphorical,
I shall refer to this process as "perception in depth," and to the structure
thus perceived as STRUCTURE-IN-DEPTH.
For several reasons, mainly the fact that (as we shall see) it does not
mean quite the same thing, I prefer this new term to the older and more
familiar "deep structure." But if we use the latter label loosely for a moment,
I can now state succinctly the answer to the question, mentioned in the
2. HEARING UTTERANCES 17

Introduction, that was bothering me a decade ago: "WHERE IS deep structure?"

Just as beauty lies in the eye of the beholder, so deep structure lies in the
ear of the listener.
As for pictures, so also our reading-in of depth in heard speech can be
brought most sharply to our attention by an ambiguous case. The did men
and women stayed home Did the young women go or did they stay home? The
sentence does not tell us. It can be read either way. You may even "feel" it
flip back and forth, like the hollow cube.
2.2. The "Syntactic Sense." Thirty-seven years ago, discussing kindred
matters, Martin Joos wrote:
Say in English The councillors all put their glasses on their noses and then get the
sentence translated into German, for English noses you vill get the singular noun-form
Nase, not the plural Nasen
Two columns later, after (by way of intentional misdirection) this and several
other examples have been discussed, he says
Without looking back at the sentence about the councillors, let [the reader]askhimself
vhether it referred to present time or past time Hewillfindthat he has made up his
mind onewor the other; and yet, on checking the wording, hewillfimithat hewasnot
told The English categorisation past-versus-present forced him to decide, even though the
decision had to be made at random
Maybe "at random" could be challenged, but the rest seems convincing. The
English verb put is one of a small set which, with a subject other than third
person singular, do not show the distinction between past and present. The
sentence is therefore ambiguous on this score. But this distinction, which a
handful of English verbs sometimes fail to show, is maintained by those same
verbs in other circumstances (He out them on: He puts them on) and by most
English verbs most of the time (They donned them : They don them). Thus,
argues Joos, users of English are so accustomed to hearing and heeding this
distinction that they "hear" the tense even in the minority of cases when it
is not overtly marked.
That the ambiguity in Joos's example is grammatical, not just semantic
(assuming that some such distinction can be maintained) is shown by two
additional facts.
One is that, although we speak traditionally of "past" and "present," these
are merely mnemonically convenient tags. The relation of the grammatical
category of TENSE to the actual location of events in TIME is highly
complicated. If one says The councillor puts his glasses on his nose there is
no ambiguity as to tense. But that sentence need not occur in a running
report of current action, as by an on-the-scene radio sports announcer. It
could be used in a vivid account of something that took place long ago, or in
18 2. HEARING UTTERANCES

stage directions referring to any time a certain play is performed. This

indefiniteness of temporal specification is not ambiguity as we usually mean
that word. It is part of the flexibility of language that allows us to deal with
an unlimited variety of circumstances with limited machinery.
The other bit of evidence for the grammatical status of the ambiguity in
Joos's example is that it is language-specific- As Joos says, our language
forces certain decisions on us because it has certain categories, but
Bach language does this within the range of its own categories, which will not be the same
range as in our language. In Chinese, for example, the translation [of the original
sentence about the councillors) lacks both the past-vereus-present dimension and the
singular-versus-plural dimension, these "meanings" can be indicated on occasion, but vhen
they are not, the Chinese listener leaves the question open—he says he is not even aware
that it is a question
Joos's discussion shows us that the hearer's ability to perceive speech in
depth is an acquired skill. The capacity to acquire this skill is doubtless
innate, and is probably what is sometimes referred to as the human "syn-
tactic sense"; yet the specific form it takes depends on the language of the
community in which one is raised.
But one can also acquire the skill for other languages.
When I studied Latin in high school, I would occasionally come to a sentence
in which virtually all the words were unfamiliar. In order to understand, I would
go through what foreign-language teachers call "dumbbell endurance deci­
pherment": first PARSE and then CONSTRUE. If we clean up the everyday usages
of these two terms just a bit, we can say that to parse is to find out what
each individual word is, while to construe is to discover, by trial and error,
which words go together and how. Of course, these two operations overlap. If
one actually did all of the first before any of the second, it would show only
the SURFACE STRUCTURE of the sentence; that is part, but only a part, of the
structure-in-depth that would be more completely revealed by the second.
The noninfant beginner on a foreign language performs these operations
deliberately, with props in the form of technical terms supplied by grammar­
ians. The noninfant beginner is laboriously undertaking to acquire, and the
grammarians's task is to try to characterize, the skill that the experienced
listener takes for granted.
2.3. A Learning Demonstration. I can guide you through a learner's ex­
perience of this sort by presenting a passage in a language you don't know.
I'll take a chance on Fijian; for your convenience I have marked primary and
secondary accents; pronounce the vowels as in Spanish, and pronounce  like
the English th of this.
E ā kanía na dálo na càuravóu.
2. HEARING UTTERANCES 19

The first word can mean in, on, at'; or it can mark the expression it
precedes as numerical; or, at the beginning of a clause, it can simultaneously
indicate that the clause is not subordinate and show that the subject of the
clause either includes both speaker and addressee or excludes both.
However, if the subject of the clause includes the speaker and the addressee,
or if the subject of the clause is not singular, one of a certain small number
of words immediately follows the e, and none of those key words is present in
the sample sentence.
The second word, ā, indicates past time.
Kania means 'eat', as also does a related form kana. But kania, in place of
kanat tells us that the eating is of some definite thing, whether or not that
thing is named by separate words in the sentence (kana, in contrast, would
refer only to the activity, as when, in English, we say to a child Don't talk
whileyou'reeating). Furthermore, the -a of kania says that the thing eaten is
not, and does not include, the speaker or the addressee.
The little word na can indicate future time, or can mark what follows it as
a common noun (like 'boy', 'island') rather than as either proper ('John', 'Fiji')
or numerical ('three', 'many').
Dalo means 'taro' (a staple root crop of Oceania) and cauravou means
'young man'.
That probably gives you enough information to guess at the meaning of
the sentence. Perhaps you will hesitate momentarily between 'The taro ate
the young man' and 'The young man ate the taro' before selecting the latter
as the more likely. Indeed, the former, however expressed, would be as
unusual in Fijian as in English.
Even a beginner at the language, however, would very early have more
information than that to use in the deciphering. For construing it is useful to
know these facts about Fijian: a particle marking the status of a clause
typically comes at the beginning of the clause; a time marker, if present,
immediately precedes the verb; if both the subject and the object of a
transitive verb are expressed other than by pronouns, both typically follow
the verb, and the object, if short, usually precedes the subject.
All that is automatic for one experienced in the language, just as compar­
able, though by no means identical, syntactic details are automatic for users
of other languages. Skilled listeners need no props from grammarians. Their
perception in depth— their HEARER CONSTRUING, which is another name for the
same thing—is normally instantaneous, effortless, and out of awareness.
2.4. Examples. Then how do we know that hearer construing happens?
Because, no matter how expert one becomes, one's conscious attention can
be drawn to the process by an ambiguity, by an unusual syntactic trope, or
by a passage that turns out to be unconstruable. Surely everyone does an
occasional double-take (perhaps especially often while reading newspapers?),
20 2. HEARING UTTERANCES

rescanning an opaque passage until suddenly things click into place.

Here are some passages that may make one aware of the construing
process.
The south court of Peruconsistsof a range of low hills that runs generally from north
to south.
There is no anomaly here, but when I encountered the sentence I had to check
to see whether it was the hills or the range that run(s) north and south,
before I could be sure of my visualization.
Rereading Kipling's "Mowgli's Brothers" for the first time in many years, I
was derailed momentarily by the following:
... vhen hewantedhoney, ... he climbed up for it, and that Bagheera shoved him hov to
do.
How to do WHAT? But then I realized that I had taken that Bagheera as subject
of showed instead of construing Bagheera as subject of showed and that as
object of do.
The small-scale ambiguity in the next passage will cause, I think, only a
momentary hesitation; it is eliminated by the end of the sentence:
In the autumn of 1886 Whitney's work was unexpectedly interrupted by a "severe
disorder of the heart," from vhich recoverywasslowand never complete — although a
glance at the record of his remaining eight years suggests that he accomplished more ailing
than most of us can manage in gloving health.
The next may be a bit more puzzling:
Historiography must involve abridgment. An accurate icon of vhat has happened in the
past would occupy as much space and time as the happenings themselves, and there is no
room for it An exact reduced-scale model is prohibited by graininess. There has to be
deletion. Kroeber faced up to this by concluding that the historian's task is not to tell
vhat has happened but not to tell vhat has not happened.
The syntactically most obvious interpretation of the next yields nonsense;
the intended interpretation requires an unusual and perhaps uncomfortable
cross connection:
Suppose that t hours after a piece of meat is put into a freezer its temperature is 72 - 3t
F. How fast is the temperature of the meat falling 3 hours after being placed in the
freezer?
Sometimes, speaking or writing in the heat of emotion or under the
pressure of deadlines, we get our syntax tangled and have no chance to
untangle it. A once prominent politician, successfully recovered from
alcoholism, is reported to have said this to a group of patients undergoing
therapy for the same trouble:
2.HEARINNG UTTERANCES 21

Alcoholics can have more pressure exerted by employers in the direction of getting him
motivated than any person or anything else.
And an American newspaper account renders a Persian original this way:
"I ask Mr Bani Sadr to remain humble and his morality not be changed by serving as
president." said Khomeini.
For these last two I think construing is impossible; there are too many loose
ends.
Quite different is the trouble in the following, a handwritten sign on an
inside door in a building on the Cornell University campus:
Forroom165 go back outside and enter through first door nearest Ezra Cornell statue.
Here there aren't ends enough: it is as though two atoms were in competition
for the single available valence of a third. Or, to use a different image, the
structure is at one point "overloaded" or "supersaturated."
To be distinguished from all the foregoing is the inept slogan You dont have
to te Italian to cook like one, in which environing words place different and
incompatible demands on Italian. To test this, try replacing the reference to
Italy by a reference to France or to Spain.
How we can manage to understand such utterances, despite their uncon-
struability, is a problem to which we will turn later (p.63, §5.5).
2.5. Old Wine in New Bottles. The preceding is probably enough to
demonstrate my point about hearer construing. But evoking ambiguous
phrases as a basis for theoretical discussion is a time-honored practice in
American linguistics, dating back at least to the 1940s, and some examples,
like Rulon S. Wells's dd men and women, have become classics. I am going to
review them here, to see what we can learn from them from our particular
perspective.
As we shall discover, the classic examples run a sort of gamut.
At one end we have Kenneth L. Pike's joke about the father who called his
cattle ranch Focus because it was where
The sun's rays meet
The sons raise meat
This is not very interesting from our current perspective because it rests on
etymological accident. To be sure, the two sentences sound the same and the
structures-in-depth are different. But sun and son (and likewise rays and
raise, meet and meat) don't seem to have anything to do with each other
EXCEPT that they sound alike. At any rate, we can be sure that if there is any
other association between the members of one of these pairs, it derives from
22 2. HEARING UTTERANCES

the homophony. (This sort of thing will concern us in greater detail in

chapters 6 and 7.)
Our first example of a pause contrast, back in §1.3, was really marginal in
this same way, since fast 'go without food' and fast 'rapidly' likewise have
only the connection of sounding the same.
Joos's example, The councillors at/ put their glasses on their noses, also
turns on homophony at the word level, but the ambiguity of put (present) and
put (past) is very different from that of words of remote meanings.
Murphy was dancing with the stout major's wife (Eugene Nida) does not tell
us who was stout. One structure-in-depth resembles that of the major's stout
wife, the other that of the wife of the stout mayor. The difference is a matter
of what goes with what more closely, and is easily diagrammed simply by
bracketing:
(stout)(major's wife) : (stout major's wife)
Or, as it is also said, in the first meaning the IMMEDIATE CONSTITUENTS (ICs) of
the phrase are stout and major's wife, whereas in the second they are stout
major's and wife. In speaking, it is possible to disambiguate by introducing a
brief pause at the point where parentheses stand back-to-back in the dia­
grams given above; that makes this pair like the pause contrasts presented
in §1.3, except that here the pause is unusual and a bit awkward.
But the case just dealt with is very much like our first example, for which,
also, simple marking of ICs suffices to display the difference:
(old)(men and women) : (old men) and (vomen)
In well punctuated written English there is no ambiguity between Chomsky's
a man eating fish 'human piscivore' and a man-eating fish 'ichthyous
anthropophage', but the two would usually sound alike. Once again, bracketing
(or awkward pausing) disambiguates:
(man) (eating fish) : (man-eating)(fish)
But the difference here is greater than for the preceding examples, since the
participating relations among words and phrases (the CONSTRUCTIONS) are
distributed differently. Mere bracketing does not show this. It shows which
constituents go together, but not how; it reveals part of the structure-
in-depth, but by no means all of it.
Thus, in old men and women the first word modifies, or is attributive to,
something that follows it, and the uncertainty is only as to how much of what
follows is modified; likewise, arid joins together coordinately something that
precedes it and something that follows it, and the only indeterminacy is as to
how much of what precedes. But in a man eating fish, fish is the object (and
the goal) of the verb eat and the phrase eating fìsh modifies man,
2. HEARING UTTERANCES 23

which is at the same time the subject (and actor) of eat, whereas in 
rnon-eoting fish, men is the object (and goal) of eat and the phrase
man-eating modifies fish, which is at the same time the subject (and actor)
of eat. To display all of the structure-in-depth, a diagram would have to show
these diverse connections, as well as the constituent organization and the
actual linear order of the words.
When we come to Chomsky's Flying planes con be dangerous, dis­
ambiguation by bracketing is not possible, since the ICs are the same for
either interpretation. Consider (1) Flying planes are dangerous and (2) Flying
planes is dangerous. Fly is an intransitive verb and in all three sentences
piones goes with it as actor, but in (1) it is subject while in (2), where the
verb is used causatively, it is object. In (1), flying is adjectivelike (a
participle) and attributive to planes, and planes is the subject of are In (2),
flying is nounlike (a gerund) and subject of is. All these differences remain
unresolved in the sentence with can be.
There are many situations in which bracketing does not serve to dis­
ambiguate. As already noted, words that belong together cannot always be
spoken together, and when they are not, bracketing is difficult or impossible.
In the 1950s this drove some grammarians to drink and others to trans­
formations, but both are only anodynes, not answers. Analysts must do what
listeners do. They must decide which words go together, whether spoken
together or not; and they must also decide HOW they go together—that is, the
constructions. As I remarked a few years ago,
Linearity means that single devices must serve multiple functions, whereupon structural
ambiguity becomes par for the course. We hear carbon fourteen, strontium ninety ; out of
context, "we do not know whether this is mention of t v o radioisotopes, or a marker giving
the distance to towns on the road ahead, or the final score in the game between Carbon
Free Academy and Strontium Senior High
— or even the names of the winning and placing horses. Bracketing would be
of no avail in this case.
Nor would it help in the tiresomely familiar, though easily circumvented,
case of ambiguity arising from uncertainty about the antecedent of a pro­
noun. If we know that only two people are involved, then John waved to Mary
as he left and John waved to Mary as she left are unambiguous, whereas
John waved to Bill as he left is not. But this same sort of mischief can be
worked on us without any pronouns. John left the car to get a pack of
cigarettes and John left the ar to get a lube job both strike me as
pragmatically unambiguous, because cars don't buy cigarettes and people
don't subject themselves to lube jobs. But consider John left the car to get
cleaned up. Or, in another vein, suppose a husband says to his wife /
24 2.ERINGUTTERANCES

can't toke you OLÌ to dinner looking like this. Is he being apologetic or
insulting?
Here, in the middle of this grove of ancient chestnuts, we may usefully
plant a Fijian example. Since in current Fijian-using and English-using cul­
tures no dolo 'taro' is not the sort of thing that can eat and no cauravou
'young man' is not the sort of thing that can be eaten, the abbreviated
sentences E ā kanío na dálo and E ā kanío no càuravόu, both perfectly good
Fijian, are both effectively unambiguous: a hearer would assume, unless there
were strong contextual evidence to the contrary, that the first sentence has
no expressed subject (actor) and the second one no expressed object (goal).
But na vua 'pig, pork' eats constantly until slaughtered and is then itself
eaten. Therefore, if heard out of context, E ā kanío na vuáka would be
ambiguous: either '(Someone or something unspecified or understood) ate the
pig' or 'The pig ate (something unspecified or understood)'. As with Flying
planes con be dangerous and Carbon fourteen, the difference is in the
constructions, not in the I structure.
Returning to English, consider How do you find the brakes in that car?,
which can appropriately be answered either The/re firm since / hod them
repaired or it's easy: /ust put your hand down between the front seats.
Similarly, How would you like to hondle it? can be construed to mean either
'Would you like to accept the responsibility for taking care of it?' or 'What is
your preference as to the manner in which it is handled?'
Examples like the last two give rise to a rather moronic informal word
game: you are on the lookout for ambiguity, and whenever you detect a case
you carefully respond to the interpretation the speaker obviously does NOT
intend:
A Shall we have Johnnie's teacher for dinner?
B. Pm afraid she'd be too tough.
A Aren't you taking the plane to Chicago?
B. Sure—if I can get it into my suitcase
Thus, in the late afternoon, if my wife says How obout  drink? I usually
respond with something like Thanks; /if /ove one. (Then I get her the drink.)
Which brings us not only to the cocktail hour but also to the other end of
our gamut. I don't see how one can construe the words How about  drink? in
one way as a request, in another as an offer. If that is correct, than the
indeterminacy here is not at all syntactic but purely situational, thus lying
beyond the range of our immediate interest.
2.6· Summary. Our basic question, as rephrased in §1.11, is about the
collusion between the structure of utterances and the strategy of listeners.
2. HEARING UTTERANCES 25

We said this collusion necessarily exists, because otherwise understanding

would be impossible.
In this chapter we have defined STRUCTURE-IN-DEPTH and LISTENER
CONSTRUING, and I have tried to persuade you, through examples, of their
reality. If I have succeeded, then we have taken a good first step toward the
answer to our question. The next step, in the next chapter, will be a closer
took at certain key features of the structure of utterances.
 3. THE HEARER'S EVIDENCE
3.1. Introduction. The examples in the preceding chapter were chosen to
demonstrate the reality of hearer construing. At the same time they may
have suggested that ambiguity in speech, far from being an occasional
blemish, is pandemic. Almost any utterance, wrested from context, can be
interpreted in more than one way if we really set ourselves to the task.
Normally, of course, our aim is just the opposite—we want to understand if
at all possible. (Even the silly game described in §2.5 shows this, since
effective pretending to misunderstand requires actual understanding.) Fur­
thermore, although we have been examining utterances in isolation, they don't
usually happen that way. In everyday consultative prose, not only is there
context but the participants switch roles constantly: the addressee of one
utterance is likely to be the speaker of the next, and if there is puzzlement
that next utterance can be a request for clarification.
When we have completed the main business of this chapter we will return
briefly to this topic (83.9). First, however, as promised earlier, we undertake a
more detailed examination of the evidence listeners can rely on as they
parse and construe. It is remarkable how skimpy and heterogeneous this evi­
dence can sometimes be, without seriously impairing intercommunication. Let
us try to categorize it.
3.2. A Classification. To begin with, we distinguish between (I) those bits of
evidence, of whatever sort, that are actually present or actually occur in the
communicative situation and (II) those that are present only as information
stored in the listener's brain ready to be evoked by what goes on. The first
category includes one or more messages (one of which is the utterance that
is to be understood); the second includes the listener's knowledge of the
codes involved. Within category I we must segregate (IA) the total commu­
nicative behavior of the speaker —the COMMUNICATIVE PACKAGE, as Henry Lee
Smith Jr called it — from (IB) everything else. If the speaker points at some­
thing, the pointing is part of IA, the thing pointed at is part of IB, and the
listener's familiarity with the speaker's convention of pointing (with the
finger in some cultures, with the lips in others) is part of II.
Since our concern is language, we break the communicative package into
(IA1) the utterance, which is the linguistic part, and (IA2) all the rest.
IA2 includes a portion which is produced by the organs of speech and is
audible: variations in tone quality, of speed of speech, of register and of
volume; inserted hems and haws; simultaneous or interrupting laughs and
sobs, sighs and gasps. This is the PARALINGUISTIC part. IA2 also includes
3.THEHEARER'S EVIDENCE 27

a KINESIC portion: all potentially observable body motions of the speaker,

from the grossest stylized gesture to the most subtle pulsing of a vein. On
occasion IA2 may include much more than this, but we here lump the rest
together. I single out paralanguage and kinesics for special mention not
because I shall have much to say about them, but because they are close to
language and one can wonder just where the boundaries are between them
and language proper. In my view, although each of the three categories
subsumes some matters not covered by the others, there aren't any
boundaries, just zones of gradual transition.
Here, in outline form, is a summary of the classification of hearer's
evidence we have just worked out:
L Actual things and events in the communicative situation:
IA h communicative package:
IA1. h utterance;
IA2 h speaker's remaining communicative behavior, including
paralinguistic and kinesic parte.
IB. The setting.
. h listener's internalised frame of reference.
Also, whatever portion of this may occupy the focus of our analytic attention
at a given moment, the term CONTEXT denotes all the rest.

3.3. Constituents of the Utterance. The utterance, IA1, is itself composite,

in that it typically incorporates several different sorts of clues to under­
standing. Every utterance includes one or more WORDS (unless we think of
"grunts" of assent and dissent and the like as exceptions). But those words
are in a specific WORD ORDER, and they are delivered with a particular combina­
tion of STRESSES and INTONATION (that is, a particular PROSODY). Furthermore,
the words are themselves of two more or less clearly different sorts,
CONTENTIVES and FUNCTORS,
Recall the Fijian sentence analyzed in §2.3: E a kania na dálo na càu-
ravóu 'the young man ate the taro'. The words dalo 'taro', cauravou 'oung
man', and kania (or the first part of that word, kani-) 'eat', are contentives;
the other words, and the -a at the end of kania, are functors. In English, faro,
young, man, and eat are contentìves; the, if, do, and the -s of eats are
functors. A contentive is what francophone linguists call a sémantème, a
functor what they call a morphème (but don't be misled here by English
morpheme, whose French equivalent is not morpheme but manème).
Contentìves are words (or word parts) with "semantic substance," as it were:
they refer to things, conditions, or events in the real world or in the world of
our imagination. Functors may also make such reference, but in broad, vague,
or situational terms. The pronoun /, for example, can refer to a real person
named Agnes Day or to a fictional character named Agnes Day, but only if the
28 8.THEHEARER'S EVIDENCE

real Agnes Day utters the word or we imagine the fictional Agnes Day uttering
it. On the other hand, anyone in any circumstances can refer to Agnes Day
by uttering the contentive Agnes Day.
However, the most important thing about functors is not what we have said
about them so far, but the fact that they serve to MARK SYNTACTIC ROLES AND
CONNECTIONS. In English, a phrase that begins with the is a noun, and its
syntactic ties must therefore be of the sort a noun can have. That leaves
many possibilities, but it also excludes many, for which the hearer therefore
does not have to allow. In Fijian, a phrase beginning with no may be a
common noun or may be a verb referring to future time, but there are
numerous things it can't be, and other factors in the utterance or in the
context resolve the residual indeterminacy.
There are boundary-line cases. Pertaining to is functorish in that it is
used much as are such prepositions as about, for, or of, but also conten-
tivelike in that it includes a thoroughly "substantial" verb pertain. French
articles are surely functors; yet it is the article, not the following noun (a
contentive) that distinguishes between te poete 'the stove' and ta poete 'the
skillet'. In many languages there are single words in which both a contentive
and a functor seem to lurk, though they can't be disentangled: m e n , ate,
worst, as compared with the transparent combinations boy-s, consume-d,
foul-est
At the extremes, though, the distinction seems clear. Its reality is en­
dorsed in various ways by the comportment of people with no special sophis­
tication in technical linguistics. Consider, for example, the first stanza of
Lewis Carroll's "Jabberwocky," in which the contentives, though all strange,
have syntactic roles clearly marked by appropriate word order and familiar
functors:
Twas brillig, and the slithy toves
Did gyreandgimble in the wab;
All mimsywerethe borogoves,
And the mane raths outgrabe
In the same vein is The gostok distims the doshes, coined by some logi­
cian to make a point much like the one being made here.
Again: people playing parlor charades usually have a signal they use for any
"itsy-bitsy" word instead of trying to act such words out. Not all functors are
itsy-bitsy words, but all itsy-bitsy words are functors. Still again: in writing
their language, the Japanese use Chinese characters (kanji) for contentives,
supplemented by a syllabary (kana) for many or most functors. An inter­
esting consequence is that literate Chinese who know no Japanese and no
kana can often get the general drift of a Japanese text; but included in what
they cannot get, since it is shown in the kana part, is the rather important
matter of whether a proposition is being asserted or denied!
3. THE HEARER'S EVIDENCE 29

3.4. The Roles of the Ingredients. Our problem, remember, is how it is

possible for listeners to understand what they hear, in the face of the neces-
sarily high measure of arbitrariness in unidimensional speech about a multidi-
mensional world. We proposed in §1.11 that there must be some sort of
collusion between the listener and the structure of utterances. In chapter 2
we characterized the listener's strategy as a penchant for perceiving speech
in depth. We are now asserting that, quite apart from any help afforded by
the context, there are clues in the utterance itself for the strategy to seize
on. The contentives of an utterance carry the bulk of its semantic import.
But its ingredients of the other three kinds — functors, word order, and
prosody — all help to show how the contentives should be construed. A way to
get at this is to consider circumstances under which one or another of the
ordinary constituents of utterances is partly or wholly suppressed.
3.5. Lessons from Written English. Ordinary written English affords a
splendid example. The words and word order are there, but intonation is
represented only very roughly by punctuation, and stress is not indicated at
all unless the writer resorts to italics and the like. A good reader infers a
suitable prosody from the information the written form does give, and adds it
almost without thinking, but the ability to do this is acquired slowly.
Elementary-school children who are still struggling with spelling-to-sound
correspondences can often be heard reading aloud word by word: each word
has its own stress and is sung to the same flat tune, except that after so
delivering the last word of a sentence the child may notice the period and
repeat the word with falling pitch.
Intonation serves primarily, i think, as a running attitudinal commentary on
what is being said or done. That perhaps covers even the use of intonation to
distinguish statement from question, as in He's going along(.) (falling —
definite — statement) versus He's going along? (rising — uncertain — ques­
tion). But intonation also marks syntax, sometimes overruling the ordinary
implications of word order: compare John screamed "Bill!" and "John!"
screamed Bit/.
In the above examples punctuation in the written forms indicates the
differences. But orthography often fails to do this. Thus, written Would you
like coffee or tea? uttered with a rise on coffee and a terminal fall presents
alternatives between which the respondent is to choose, but read with a
terminal rise constitutes a question that can be answered yes or no; the
written form obliterates the difference. And consider the familiar example
(here we go again, reroasting old chestnuts!) What are you reading, Whitney?
Said with a fall from high to low on reading and a rise from low to mid on
Whitney, this is a query addressed to someone named Whitney; the same
written form can be read off with the same fall but with a rise from mid to
high on the last word, in which case it is a query to an unidentified
30 3. THE HEARER'S EVIDENCE

addressee, with a guess that what is being read is by Whitney. In the first
member of the last pair there is a syntactic tie at a distance — "apposition"
or "coreference" or something like that —between Whitney and you; in the
second, the same sort of tie between Whitney and what. The same intona-
tional contrast, likewise neutralized in the written form, yields a comparable
syntactic difference but a more macabre semantic effect with H. L. Smith Jr's
word-sequence What are we having for dinner, Mother?
Faced with an inadequate writing system, good writers learn that part of
their responsibility is so to phrase what they have to say that the written
form will induce the proper prosody with minimum effort on the reader's part,
and without backtracking or double takes. But readers have to do their part
too, and even with the best of intentions it doesn't always work smoothly. I
misread the following (in an Agatha Christie story) when I first encountered it,
putting a stress on the word I have capitalized below:
[Someone is speaking:] "How can I make you understand? Everything Pve been brought
up to believe in, Howard abominates and wants to do sway with. And sometimes, you know,
I feel like he DOES "
— meaning that I feel as though he does do away with all those things. But the
larger setting actually calls for the reading / fee/ like HE does: that is, at
times, I, too, abominate all those things.
Similarly, returning to a Rex Stout murder mystery after an hour's inter­
ruption and finding my place, I read
TheSTYMIEwasthe motive.
But that was puzzling. 1 could not recall who had been stymied, or in what way,
that would have been the motive for the murder. Then I realized that the
sentence should be read as The stymie was the MOTIVE: what had been
stymied was the detective's efforts to solve the crime, and the cause of his
difficulty was the lack of any discernible motive. The difference in struc­
ture-in-depth, signaled by the different location of stress (and concealed by
the orthography), should be noted. With either prosody, the (grammatical)
subject of the sentence is the stymie, and the motive is a nominal predicate
attribute. But in each case the unstressed noun is the TOPIC—what used to
be called the "psychological subject," the specification of that about which
information is being given; while the stressed noun is part of the COMMENT —
the "psychological predicate," the part of the sentence that gives the infor­
mation.
Encounters with prosodically ambiguous written passages, like the fore­
going, teach one something of what prosody does in ordinary conversation. /
bought a CAKE and / BOUGHT a cake differ in a way that is made clear if we
put these words at the ends of appropriate longer sentences:
3. THE HEARER'S EVIDENCE 31

I was hungry, so I bought  CKE.

I was out of soap, so I BOUGHT a cake.
The same stress difference appears in the following pair of possible reports
by a guilt-ridden motorist walking into a police station:
Fm afraid I've killed a DOG.
Pm afraid Pve KILLED a man.
Here we come across a subtle bit of sexism, built into our language by our
male chauvinist predecessors and peculiarly difficult to eliminate: in the
pattern of the last two examples, the only objects that put the stress on the
verb are man and someone. Any other object noun, including woman, will
itself take the stress.
For most people one of the following lines will be read correctly only after
a double take:
Joe has a consuming passion—golf.
Joe has a consuming passion—he eats too much.
We have a catchphrase consuming PASSION, meaning an interest that
absorbs all one's time and energy. Because that phrase is familiar, the
two-word written sequence is apt to be read as though it represented that
phrase whenever we see it. But in the second example above, the same
written sequence actually represents a different phrase: a CONSUMING
passion, meaning an overwhelming desire to consume.
There are virtually endless numbers of possible paired phrases of that sort
in English. Sometimes they are "synonyms" — to the extent that any two
forms diffeTing in sound can be; for example, ICE cream and ice CREAM
refer to the same substance, and any semantic difference is one of over-
tones, deriving from dialect differences or from accidents of individual life-
history. But most are not. Thus, a white HOUSE is any house that is white, or
a splotch of whiteness that is a house rather than (as someone has perhaps
suspected) a woodshed or a tiger or something else, whereas the WHITE
house is the house that belongs to the people named White, or any house that
is white rather than (as someone has perhaps suspected) some other color,
or the President's residence in Washington. (Standard orthog- raphy makes a
few distinctions by capitalization, as White house for the Whites' residence
and White House for the President's, but that doesn't go very far.) An albino
BLACKbird, as is well known, is not a black BIRD. A WOMAN doctor is a
gynecologist just as an EYE doctor is an ophthal- mologist; a woman
DOCTOR is a doctor who is a woman, just as a woman GOLFER is a golfer
who is a woman. And here we encounter another bit of inherited sexism: with
the crescendo pattern just illustrated, the modifier can be woman or female
or even male, but not man. A doctor or golfer — or a professor, a grocer, a
carpenter —is assumed to be a man unless there is explicit indication to the
32 3. THE HEARER'S EVIDENCE

contrary.
Perhaps nothing demonstrates the importance of proper stress-distribution
better than the gaffes of radio and television announcers reading without re­
hearsal from hastily prepared copy. They are trained not to "throw away" the
ends of sentences, so they normally put the major stress on the last written
word whether it belongs there or not: Jones's Office Supply now presents art
CALENDAR — leading one to expect that well known commentator Arthur
Calendar, whereas what one actually gets is a quick review of upcoming
events in the arts, known in normal English as an ART calendar. Or — on a
nationwide network — Hurricane Daniel, which was hovering off the coast near
Biloxi, is now moving in LAND, which is absurd. On the other hand, when there
is an itsy-bitsy word near the end of a sentence, customarily mumbled in
standard English delivery (§3.8), they often carefully put an emphatic stress
on it: Tune in for this program again next Saturday, AT two o'clock, ON this
station — as though the audience must be warned not to tune in ON two
o'clock, AT this station. They also confuse the different lower levels of stress,
which we have not had occasion to discuss here: if the text says the tent was
destroyed and the arena will have to be rèCOVered = 'covered again', they will
read it as reCOVered - 'rescued'. Of course, our writing system is partly to
blame, since if it indicated prosodie features more directly this ridiculous
announcerese dialect might never have arisen.
Individual polysyllabic words with different stress patterns can be homo­
graphs, as the nouns CONtent, INvalid, aRITHmetic versus the adjectives
conTENT, inVAUd, arithMETic, the noun ENtrance versus the verb en-
TRANCE, or the adjective FREquent versus the verb freQUENT. They can
thus yield ambiguous written sentences. But so can other kinds of homo­
graphs, such as row rhyming with know or with plow. Therefore these ambi­
guities in writing reveal nothing about prosody, which has to do with the dis­
tribution of stresses and pitches over the constituent words of an utter­
ance, but relate rather to the identities of those constituent words.
3.6. Lessons from Headlines. So much for ordinary written English. The
conventions of headlines (and telegrams) carry matters a step farther: most
punctuation is omitted so that there is even skimpier indication of prosody,
and also many of the functors of the itsy-bitsy-word type are left out. Thus,
the headline
SHIP SAILS TODAY
is indeterminately a condensation of The ship sails today, of They ship the
sails today, or of The ship-sails are available today. The example is extreme
because two of the words can be taken as either nouns or verbs and the only
interpretation that makes no sense is taking them both as verbs. Of
3. THE HEARER'S EVIDENCE 33

the languages most familiar to us, English seems especially prone to this
kind of headline ambiguity, probably because the language has so many
noun-verb, noun-adjective, and adjective-verb homophones and homographs
and because word-final -s is ambiguous for the plural of a noun and the
present-tense third-person singular of a verb. Here are more examples:
BASEBALL TALKS IN NINTH INNING
SMOKING RISKIER THAN THOUGHT
HOT DOG FIRM IN NEW HANDS
QUEEN DUCKS RIDE TO U.S. ON CONCORDE
TEXTRON INC. MAKES OFFER TO SCREW COMPANY SHAREHOLDERS
Of a different sort are the next two, in which the ambiguities would be
removed if the stresses were indicated (on one or the other of the first two
words in the first, of the last two words in the second):
FARMER BILL DIES IN HOUSE
FORD GIVES SWAN.SONG
In the next set, one interpretation of immediate constituents yields one
meaning, another interpretation another:
TUNA BITING OFF WASHINGTON COAST
CROWDS RUSHING TO SEE POPE TRAMPLB SIX TO DEATH
CITY TO ADD TWELVE FOOT COPS
In the next there is uncertainty as to what is modified by terminal prepo­
sitional phrases:
BUILDINGS SWAY FROM SAN FRANCISCO TO L A
— and in this one the uncertainty has to do with the identity of an antecedent:
HE FOUND GOD AT END OF HIS ROPE
Finally, here is one in which an intransitive verb can be interpreted either as
having an actor-subject and a nominal predicate attribute or as causative
with an agent-subject and an actor-object:
NONSWIMMER DROWNS A HERO
3.7, The Hat Game. We can take one more step.
Headlines and telegrams, to be sure, constitute our most constrained
channel for the real, purposeful transmission of information in linguistic
form. But in our culture we also play all manner of word games. What we need
here is a special sort of word game, one that not only eliminates all functors
(headlines retain a few) but also somehow gets rid of word order, so that
34 3. THE HEARER'S EVIDENCE

there is really nothing left but contentives. We want an artificial system

whose syntactic dimensionality (§1.4) is zero: that is, the only syntactic
device is mere togetherness.
Our game will go like this. The poser draws the contentives from some
brief sentence, shakes them together in a hat, and presents the hatful to the
respondent. Respondents are to make any sense out of them they can, using
as few functors as possible. The natural way to do this is to assemble them
into some linear order, insert suitable functors, and add a prosody so that
the result is a sentence. But there may be several solutions, in which case
the respondent should weight the different possibilities in terms of relative
likelihood or reasonableness or the like.
Here is our first hatful:

Anyone can easily dream up at least a dozen very simple meaningful

assemblies of these ingredients. But for our purposes we can confine our
attention to the following two, each of which adds just three functors: (1) The
boy eats the pig; and (2) The pig eats the boy Number (1) is a bit awkward
because when we speak of eating the edible portions of a pig we are more apt
to use the word park. The awkwardness is thus LEXICAL, stemming from the
meanings of individual contentives. Number (2) is likewise slightly off key, but
for a cultural rather than a lexical reason: in our culture, pigs don't normally
eat people. However, both of these assemblies are surely more likely than
would be, for example, The boy eats at the pig, intended to be interpreted on
the pattern of The boy eats at the table
The second hat contains German contentives:

Here we know from the contentives themselves that the boy (Bube) does the
eating, because the verb essen refers to consumption by human beings; for
animals, including Schwein 'pig', there is a different verb, fressen.
But now consider the first hatful again. Assume that this zero-dimensional
message is transmitted to orthodox English-speaking Jews or Muslims in the
context of their own culture. Although neither of the suggested interpre­
tations really appeals, the second will be preferred to the first, because
3. THE HEARER'S EVIDENCE 35

people simply do not eat pigs.

3.8. Inferences. These hat-game examples underscore two points.
First: The cultural frame of reference brought to the communicative sit­
uation by the listener, and the listener's impression (accurate or not) of the
speaker's cultural orientation, both of which are parts of our category II
(§3.2), can never be neglected. They bear constantly on whether the listener
will understand, not uncterstand, or misunderstand This will be of special
importance in §3.9.
Second: Even contentives do some of the structural marking of utterances.
This can also be said in another way. Since there is no sharp boundary
between contentives and functors, we can say that even the purest of
contentives have a bit of the functor about them. Therefore, although a
message composed solely of such contentives can be, within itself, extremely
vague, adequate disambiguation can sometimes be supplied by factors
completely outside the utterance.
Naively, one might expect it to be otherwise. Since functors do such
important work, one might expect them to be highlighted in speech. Yet
typically they are not. It is worthy of note that in languages like English,
Spanish, Fijian, and Mandarin, which have stress systems (and setting aside
such pathological cases as American announcerese), itsy-bitsy words are
mostly ATONIC: that is, they are normally pronounced without stress, or even
mumbled, emphasis and clarity being reserved for contentives. This habit is
overruled only by communicative necessity, as when one clears up a
misunderstanding by saying something like / said put it ON the box, not IN
the box or / said that I DID go, not that I did NOT. In a language with no
stress system there cannot be atonic words in just this sense, but there are
often other devices for foregrounding and backgrounding, and, whatever the
device, the foregrounding of a functor is rare and special. [Obviously these
remarks do not apply to bound functors (inflectional affixes}. In Spanish, for
example, whether a stress falls on an inflectional ending, as in cantaré Ί shall
sing', or on the preceding stem (a contentive), as in cario 'i am singing', is
purely mechanical and has nothing to do with emphasis.]
3.9. Being Realistic about Language. Now that we have classified the ingre­
dients not just of utterances but of the whole communicative situation, we
must consider more carefully the point mentioned briefly in §3.1.
We there expressed surprise at the prevalence of ambiguous utterances as
revealed by the discussion in chapter 2. But in that chapter we had been
examining specimens in vitro, as it were — in artificial isolation, dissected to
find out how much information they carry and how much is left to their con­
texts. In this chapter we have continued to dwell on ambiguity, for the same
reason.
But if we now consider language once again in vivo, our surprise should be
36 3. THE HEARER'S EVIDENCE

stilled. Most typically, the comnunicative package as a whole, in its setting,

given the largely shared cultural expectations of the participants, is so
redundant that almost any small portion can be lost without seriously
impairing intelligibility.
We linguists must remember our inevitable professional bias and make
proper allowance for it. In considering any human transaction in which lan­
guage plays a part we naturally incline to see the language part as central
Accordingly, we are struck by ambiguity in that part, and tend to think that
the major role of context is to disambiguate.
That must be turned around
After all, although language and its derivatives (such as writing) nowadays
perform a remarkable range of functions, throughout human history most talk
has taken place among people who have known one another all their lives, so
that only minimal clues are needed, and the major biological function of talk
has always been to pool information and make joint plans. There was a time
when our ancestors, doubtless to be thought of as not yet fully human, had
to manage their affairs without the help of language because language had
not yet evolved, and still today, many transactions go on quite successfully
with no speech at all or with only minimal verbal comments by the partici­
pants.
When we look at the matter that way, we realize that the most important
feature of consultative prose is not that it can be disambiguated, but that it
can disambiguate.
The primary role of the noisy exhalation we call speech is to close gaps. If
vague words often suffice, that is because SPEECH HAS TO SAY ONLY WHAT ITS
CONTEXT LEAVES UNSAID,
 4. HEARING WORDS
4.1. Modes of Listening. Our next step is to examine the listener's per­
formance in finer detail.
We have said that in order to understand a heard utterance the listener
must construe the words of which it is composed. But to construe those
words one must first parse them: that is, recognize each word for what it is.
Consequently, what the hearer normally does is to LISTEN FOR WORD IDENTITY.
There are several MODES of listening, but this one is surely the commonest.
Let us immediately acknowledge that it is not always necessary to identify
every word. If the utterance is a familiar fixed phrase, like Good morning or
See you later, one need identify only the whole phrase. Or if it includes such
idiomatic expressions along with other material, the constituent words of the
idioms don't have to be separately identified, only the idioms as wholes. Fixed
expressions obviously play a vital role in rendering speech intelligible, but it
will be more convenient to ignore this until later (§§5.4-5).
Were we viewing language through the wide-angle lens of sociolinguistics,
we should surely be led to distinguish many more modes. Here we give pass­
ing mention to only one additional phenomenon. Sometimes a speaker drones
on and on and we get bored. If we really stop paying attention the situation is
of no interest to us, since our concern is with listening, not with not
listening. But the setting can require us to seem to be paying attention and
to be prepared with a suitable response should one be called for. What we can
do in this case is post a peripheral sentry, as it were, to monitor what is said
and to sound an alarm if a response is required. Then, as long as the alarm
does not sound, we can turn our attention to our own thoughts. This might be
called SENTRY LISTENING. One can develop great skill at it, but it is always a bit
risky — if the sentry goes to sleep the results can be devastating. Though I
should like to know how this works, it lies far beyond the elementary matters
dealt with in this essay.
4.2. The Task of Identifying Words. In quiet surroundings, the identification
of what is said to us is often seemingly effortless. With noise or impaired
hearing one has to be more attentive. A language we don't know very well
calls for closer attention and less ambient noise than one we have been using
all our lives. In any language, if a key phrase is blotted out we may have to
ask for a repetition. If the speaker uses an unfamiliar word, we may under­
stand from context or may be forced to seek clarification.
When a speaker's dialect is strange to us, or has an unusual foreign accent,
we can be baffled. Yet this doesn't always happen: often we understand de-
38 4. HEARING WORDS

spite an alien cast to the sound of the words. Minor differences of pronun-
ciation may not even register. For example, I conversed freely for years with
a number of colleagues from the northwestern part of North America before I
realized that their speech lacks the distinction I have between the vowels of
cot and caught, and I think they were equally unaware that my speech
maintains this contrast. Likewise, communication with east-coast Americans
was never hampered by my lack of their three-way differentiation in merry :
Mary : marry, though this difference, too, escaped my notice for a long time.
Also, one may understand without difficulty even as one notes a strange
pronunciation. I recall my lone encounter with someone from western Penn­
sylvania who pronounced marry with the stressed vowel I have in starry. I
don't remember all the circumstances, but the sentence in which the word
occurred was something like The kîds run off down there to get manrried.
Another isolated experience of this kind was overhearing a young man from
southern Delaware ask for a glass of wooter, using the vowel I have in book.
Doubtless in both cases there were also other differences from my own brand
of English, but no others came to my attention.
Anyway, noting unfamiliar pronunciations of specific words is not the same
as being aware of a general cast of difference. For most of us, it is just this
overall coloring that we note in a dialect of English other than our own —
northern versus southern, New York versus Chicago, tidewater versus pied­
mont, British versus American, and so on. The laity describe such differ­
ences in vague impressionistic terms, as when middle-westerners speak of a
southern "drawl" or a Texas "twang." Even phonetically trained specialists
can have a hard time pinning down the specific features that add up to yield
the overall effect of a dialect other than their own.
It is within an intimate group such as a family that the stock of shared
understandings is the greatest; correspondingly, talking is there apt to be the
most careless and marked by the wildest distortion. Probably this is more
noticeable in some families than in others, and my own may be unusual, but I
know we are not unique. Just recently, as my wife and I were driving around
town on errands, we looked up and saw a small airplane and I said "Oh, see the
wittoo aiwopwane!" — pseudo baby-talk, with w or oo replacing both / and 
That exemplifies only one of numberless types of distortion, some more
patterned, some more sporadic. The former are the easier to illustrate, but
the latter are probably more important in the long run, and will be discussed
more extensively later (§5.4).
On top of all these differences from idiolect to idiolect and from occasion
to occasion, speech generally remains intelligible whether the voice be high or
low in register, smooth or gruff, overvoiced or breathy, loud or soft, clear or
noisy; and these features of voice quality vary not only from speaker to
speaker but also, for a single speaker, with mood and setting.
And last but not least, in almost all the circumstances which elicit speech,
4. HEARING WORDS 39

we articulate carelessly.
Let us note in passing that the task of a reader is in general like that of a
listener. To be sure, writing eliminates some sorts of noise, but it introduces
others and has its own complexities and irregularities that must be dealt with.
If the writing system is phonetic (that is, alphabetical, nagari, or a syllabary),
then the sequence of graphic shapes tells the reader what sounds to produce
or imagine, and from that point on the processing is much the same as
though someone else had uttered the sounds. With a character system the
point of entry from graphics to the language proper is different, but it still
ends up with ordinary linguistic processing: parsing and construing, as de-
scribed in earlier chapters.
4.3. Strategies for Word-Identification. Given the enormous range of vari­
ation of the sound of words, how do we manage to recognize a word when we
hear it?
Actually, sometimes we don't. In what I have said so far I have probably
been a bit oversanguine about the efficiency of our hearing. My impression is
that in the course of an ordinary day people miss one another's remarks
dozens of times, usually but not invariably clearing matters up through
repetition or paraphrase. Nevertheless, the level of success is so high as to
render the question meaningful, and the variation is so great that the solution
is by no means trivial. Three answers have been proposed, and I think that in
a sense they are all correct (§§4.4-6).
4.4. Implicit Motor-Matching. When you utter a word yourself, you nor­
mally have several clues to its identity. One is that you presumably know in
advance what word you are going to say. Another is that you feel the motions
of your organs of speech as you say the word: KINESTHETIC FEEDBACK. A third
is that you hear the word: AUDITORY feedback. The second and third enable
you to monitor the first. If what you feel and hear is not in agreement with
your intention, you may cut the word off and start it over again. That is
OVERT EDITING, a phenomenon we will discuss in chapter 9.
When someone else utters a word, although you do not have the first of the
above three kinds of evidence, you do have the context of what has already
been said, and you do hear the word more or less clearly.
The first hypothesis about the process of understanding heard speech
turns on kinesthetic feedback. It proposes that, although listeners obviously
cannot have kinesthetic feedback from someone else's articulation, they in­
terpret what they hear by IMPLICIT MOTOR-MATCHING.
Consider an extreme case. We bend over a dying man to catch his faint
final words. Then we try various phrases ourselves, seeking one whose sound,
properly filtered, will match what we have heard. If we are lucky, we find the
one that the dying man must surely have been trying to say. This is EXPLICIT
or OVERT motor-matching. Under more routine circumstances, we may occa-
40 4. HEARING WORDS

sionally resort to overt motor-matching to help identify an isolated muffled

word in an otherwise clear context. But with enough experience at this — the
kind of experience every user of a language gets very early in life — actual
movements of the organs of speech become unnecessary; the appropriate
pattern of impulses within the central nervous system is enough.
I think there is evidence that we sometimes rely on this method, though
certainly there is no reason to assume that it is the only way in which we can
identify what is said to us. Matching of sensory input against stored acoustic
patterns may often suffice. For sounds other than those of speech this is
manifestly so. I recognize the sound of a flute, which I can play; but I rec-
ognize flute performances that are far beyond my technical capacity; I also
recognize the sound of a violin, which I don't play, and the sound of thunder,
which no one plays. Thus, for some kinds of sounds, since motor-matching is
impossible, it must be unnecessary. We must also remember that there are
individuals who for one or another reason cannot themselves speak, yet who
nevertheless understand the speech of others.
4.5. Gestalt Perception. In any case, it is difficult to see how implicit
motor-matching would be of much avail when we are trying to understand
speech in a dialect divergent from our own. For this we must call at least on
the second hypothesis, perhaps also on the third.
To explain the second hypothesis it will be helpful to turn for a moment
from hearing to vision and talk about reading instead of about listening.
In our local hospital, which I have visited frequently in recent years, there
are many doors that bear the following inscription in terge red letters:

FIRE CONTROL DOOR

KEEP THIS DOOR IN
THE CLOSED POSITON
DO NOT PROP IN AN
OPEN POSITION
I must have read this sign a dozen times before I spotted the misprint. Such
experiences are a commonplace. If we recognize written words without noting
every feature of their appearance, we have gotten the message and do not
bother further. That is why typos are occasionally overlooked by even the
most meticulous proof-reader.
There are psychologists who think that children in their earliest learning
stages recognize a written word just from its general visual contour. For
example, if in a given context all one has to decide is whether a printed word
4. HEARING WORDS 41

is "cat" or "dog", one need notice only whether its outline is

If these experts are correct, the child proceeds from wholes — relatively
small ones, to be sure — to parts, paying attention at any stage only to as
much finer detail as is necessary for identification of the word. In the long
run, of course, the detail criterial in one or another context comes to be just
the total set of conventions of the writing system, but there may be many
successively closer approximations before that final stage is attained —and
even then no one context requires attention to every criterial feature.
An alternative theory, older and perhaps still more widely held, proposes
that the child first masters individual letters and then learns to identify
written words in terms of the letters of which they are composed. You learn
to recognize and identify "c" and " a " and " t " and then you learn that those
three letters in that order form the word "cat".
It may be that neither of these theories is right to the total exclusion of
the other. But I am convinced that there is a significant measure of truth in
the whole-to-part proposal.
Along the same line, note that for literate users of English all of the follow­
ing, and many more, are immediately recognizable as one and the same word.
Despite the variations of detail, all present the same holistic pattern or
GESTALT, and it is this overall pattern that identifies the word for us:

Turning back now from writing to speech, I propose that the same prin­
ciples apply. A spoken word is a perceptual gestalt. In listening we register
the overall pattern and ignore —or set aside for separate handling—the finer
details. We understand heard speech despite great variations in dialect, voice
quality, and superimposed paralinguistic effects, because all these changes
are rung on an almost invariant skeletal pattern.
What is more, we usually do not have to hear every bit of what someone
says in order to register it and respond properly. We have all repeatedly had
the experience of recognizing a familiar and expected word or phrase against
a noise level high enough to render unrecognizable an unpredictable novel
utterance. An excellent example of this is being introduced to someone: the
introducer's words are clear until the stranger's name is uttered, but with
that name, which in the United States can be anything from Prokoko-
42 4. HEARING WORDS

fieskovitch to Ng, we suddenly discover that the ambient noise-level is higher

than we had realized.
Of course, the identifications we make when we think we have heard clearly
are not always what the speaker intended. But I take that as further evidence
for the hypothesis: we listen for the acoustic stigmata of familiar and ex­
pected wholes (perhaps often also feeling, via implicit motor-matching, for
the corresponding kinesthetic sensations), and tend to pay attention to finer
detail only as the circumstances show us we must. The sound we hear (of
speech or otherwise) regularly lingers for a few moments in short-term
memory; for those few moments it can be "played back" for reexamination if
necessary.
Our task of identification is sometimes easier and sometimes harder. If we
know that the answer to a question is in all probability going to be meat or
fish, we could tell which is uttered just by noticing if it begins with a nasal
consonant. Alternatively, we could listen for the presence or absence of a
voiceless spirant. There might even be adequate clues that have nothing to
do with the sound of the words: perhaps we know that our interlocutor loves
fish, and perhaps as he gives his answer he smacks his lips in anticipation of
a favorite food. (This last possibility takes us back to §8 3.2 and 3.9, where
we emphasized that an utterance is usually only one part, often only a minor
part, of the whole communicative package; but it also takes us away from our
current focus of interest.)
As simple as the example of meat and fìsh is, it underscores two essential
points about the role of gestalt perception in listening for word identity.
The first is that, because speech is typically redundant, you don't always
have to heed the same evidence. If channel noise or dialect difference (which
is another kind of noise) obscures some clues, there are usually others you
can rely on.
The second is that, because of context, you don't have to be constantly on
the alert for the whole lexicon of the language, only for the words that have
some reasonable chance of being uttered in the particular setting. Or, as it
has been put aphoristically, WE LISTEN FOR WHAT WE EXPECT TO HEAR.
This maxim is confirmed by all manner of everyday experience. An espe­
cially telling informal test was reported years ago by the late Yuenren Chao:
Once over the telephone I asked Robert W. King ...: "Pam you ungelfpangg thob I fay?"
tad he answered promptly: "Yes, I understand you perfectly, but you talk as if you had
something in your mouth."
There are also many jokes of the "garden-path" sort, that exploit the lis­
tener's reliance on expectations by telegraphing one ending and then deliv­
ering a different one; each recounting of such a joke constitutes a tiny test
of part of our hypothesis:
She's beautiful, and a good dancer, and her coffee tastes like sham POO.
4. HEARING WORDS 43

4.6. Code-Switching. Gestalt perception, even as bolstered by implicit

motor-matching, is probably not enough to account for some of our more
remarkable feats of understanding. The third proposal is that as, say,
Chicagoans get more and more practice at listening to British English, or
Danes at dealing with Norwegian, they build up a SWITCHING-CODE: certain
sounds or arrangements of sound in the alien dialect come to be coded
automatically into the proper sounds or combinations of sounds in the
listener's own dialect, and the intended word is recognized by assembling the
latter.
It seems to me that this is somewhat like the older theory of reading
mentioned earlier. Suppose children first learn only the printed shapes of the
letters, but that they are then taught that
cursive  corresponds to printed C;
cursive a corresponds to printed a;,
cursive t corresponds to printed t .
Having learned these correspondences, they can now infer that
cursive cat corresponds to printed cat.

So this proposal does not conflict with either of the other theories. Cer­
tainly it is entirely compatible with the notion of the gestalt perception of
whole words; in fact, it seems likely that the listener does sometimes one,
sometimes the other, and sometimes both at once, each as a supplement to
the other.
Thus, I know from long experience that almost always where a southerner
pronounces a stressed long laminalized low front unrounded vowel (as in /,
my, mine, eye, ride), the corresponding word in my own dialect has a stressed
diphthong going from low central unrounded toward high front unrounded
(ah-ee) (That I can DESCRIBE this, by the way, may be due to my training in
phonetics, but that i KNOW it has nothing to do with such special education,
since millions of linguistically unsophisticated Middle-Westerners know it as
well as I do.)
What I am not sure of is how much I call on my knowledge of this corre­
spondence when dealing with southern speech. My impression is that usually I
don't need to — I simply recognize the word as a whole (gestalt perception).
But how would I handle an unfamiliar word, presented in a context that made
its meaning clear? If subsequently, in using the word myself, I replaced the
southern vowel by my regular equivalent, I think that would be evidence for
the code-switching hypothesis.
Obviously one cannot rely on code-switching unless one has identified the
speaker's dialect. Certain southern speakers have a diphthong that begins
44 4. HEARING WORDS

approximately like my vowel in law (low back rounded) and then moves
toward high back rounded: aw-oo. Many northwestern speakers use a
diphthong which to my ears is identical with the southern one just described.
But if I hear an unidentified stranger say law-oost I do not know what word
has been uttered unless I know whether the speaker is southern or
northwestern. The southern law-oost is equivalent to my lost, the
northwestern one to my loused.
4.7. Norm Shapes. We have now spoken at length of distorted or deviant
pronunciation. Deviant from what? How can we justify the implication that
some pronunciations of a word are less "normal" than others?
In fact, it seems to be the case that for any one user of a language, any
ordinary familiar word has a stable NORM SHAPE. It is toward this norm shape
that the user's articulatory movements are aimed when uttering the word, and
(perhaps because of implicit motor-matching) it is this norm shape the user
is usually under the impression of having heard when someone else utters the
word—provided the difference of pronunciation is not too extreme.
Although each word HAS a norm articulatory-acoustic shape, we must not
say that a word is that shape. Once a heard word has been recognized, only
its identity matters; what it sounds like and what particular clues have been
heeded in identifying it are no longer of any consequence. The truth of this is
endorsed by the measures people take to be understood in the face of
difficulties. In literate English-using society a puzzled look from one's re­
spondent may be met not by repetition or paraphrase but by spelling a word
out. In China and Japan the equivalent procedure is to draw a character in the
air. These methods identify the target word without any direct reference to
its phonetic shape; they are thus usable not only to circumvent ambient noise
but also to distinguish among words that sound alike. In illiterate communities
these techniques are unavailable. But there is one other method, available and
used everywhere.
This method involves another major mode of listening. If we miss a word, or
hear one with which we are not altogether familiar, we may ask for a
repetition, and then listen not for word identity but for WORD SHAPE. AS
speakers, if we realize that our interlocutor is listening in this way, we often
help by enunciating more slowly and clearly as we repeat the word (CLARITY
NORM speech, as over against ordinary careless articulation).
In our culture one of the commonest contexts for this is the social
transaction mentioned earlier in another connection: introductions. The
introducer says I'd like you to meet Miss Mrmph. It doesn't matter what the
lady's name really is; invariably this is what it sounds like. To be sure, other
kinds of words can also do this to us, and often enough, even if we manage to
4. HEARING WORDS 45

catch the meaning of a whole utterance, we may remain uncomfortable until

we have identified and pinned down the shape of a missed word.
Yet it is also possible for a user of a language to recognize and understand
a word without being quite sure what its norm shape is. This is perhaps not
likely to make any difficulty except when such users are trying to say the
word themselves. Imagine an interchange like this:
A Hesaidhe'd bring his sk-, ska-, skelpwithhim.
E His what?
A. Well, it's skelp orsquelpor something like that. You know—that funny
guitarlike instrument he brought back from Kerguelen.
It is also possible for a word to have two or more different norm shapes.
Familiar examples are PRO-gress and PROG-ress, or pres-TIJ-ous and pres-
TEE-joust or MAG-a-zine and mag-a-ZINE, or REC-og-nize and REC-kon-
ize In English there are literally thousands of cases; in languages confined to
small communities the phenomenon may be rarer. Ordinary users of English
would certainly say that progress is the same word whether pronounced the
one way or the other, just as they would say that the city name Syracuse (in
upstate New York) is the same in any of its eight pronunciations (beginning
like serial or like serrated, final consonant as in obtuse or as in refuse,
primary stress on first or on third syllable). This judgment is sometimes
based on identity of spelling, vet in some instances of identical spelling and
different sound and meaning (as lead the verb and lead the metal), the lay
opinion of same or different goes with the sound; in the case of bear 'carry'
and tear 'ursus', same in spelling and sound, it goes with the meaning.
If in such cases as PRO-gress and PROG-ress some people used the one
pronunciation and some the other, we would simply treat the matter as a case
of dialect (or idiolect) difference. But it can easily happen— and often does—
that a single speaker uses now one pronunciation and now another. Con-
fronted with this, we can begin to wonder if we are dealing with two norm
shapes of a single word or with two distinct, even if almost interchangeable,
words. I think this is one of the situations where to insist on a simple
either-or answer would distort the facts. There are more intergradations
between "one and the same word" and "two different words" than we have
been inclined to allow for.
4.8. The Evidence of Borrowing. That view is supported by some of the
things that happen in borrowing. If we hear an unfamiliar word in an alien
dialect —or, for that matter, in a foreign language—and seek to add it to our
own active vocabulary, we will surely resort to listening for word shape, but
we may then treat the evidence in either of two ways. One way uses code-
46 4. HEARING WORDS

switching: we map the succession of sounds in the word's norm shape in the
alien dialect into the proper succession of sounds in our own dialect. The
other way rests on PHONETIC SIMILARITY: we match the successive sounds of
the alien form as closely as we can with the sounds of our own, or even, in
the extreme case, introduce a previously foreign sound into our speech.
Years ago, in New York City, I met a man whose surname I took to be
Coulter, with the diphthong in the first syllable that I have in Cae, colt, cold.
But I was corrected, and informed (in no uncertain terms) that the name was
Kalter, which in my dialect has the low back rounded vowel found also in call
The source of the trouble was that the New York pronunciation of the man's
name has a mid back rounded vowel that occurs also in their pronunciation of
caff, but that I do not have anywhere, though it is close to the first part of
my diphthong in colt. Apparently I had not yet built into myself a sufficiently
detailed switching code for the conversion of this strange dialect into my
own. So in "borrowing" the surname I relied on phonetic similarity—and came
up with a pronunciation that was recognizably wrong to my New York friends.
With ordinary words, as over against people's names, we are more tolerant.
Our late colleague Raven I. McDavid was born and raised in South Carolina, so
in words such as /, like, ride, fine, cider he had the long laminalized vowel I
mentioned earlier (§4.6). Normally, having heard from a northerner a new word
with the ah-ee diphthong (as in my pronunciation of these same words), in
adopting the word for his own use he would replace the northern diphthong by
his laminalized monophthong. Doubtless some of the words in his vocabulary
had just that history. But in the case of the word tiger, which he learned from
some northern user, he imitated the northern model, and thus pronounced it
essentially as I do.
The closest parallel to this I have found in my own speech is that, though I
regularly pronounce oo (as in spook) rather than yoo (as in yoo-hoo) in
words like tune, duke, news, tube, so that I say mailing toob, I call the London
Underground the tyoob.
One might consider these examples suspect on the grounds that both
McDavid and I received special training in recognizing and imitating alien
speech sounds. In the case of my tyoob the suspicion is fully justified. But
McDavid's tiger was apparently learned in his youth, before the acquisition of
any special expertise.
Something like these two cases must have been involved in the origin of
such doublets as hoist and heist, burst and bust, curse and cuss, clerk and
Clark, wrestfe and wrastle, hoop with the vowel of spook or that of book.
That is, first different dialects develop different pronunciations of what was
at some earlier time a single word with a single pronunciation, then users of
one dialect borrow the word from the other, imitating its sounds instead of
replacing them by the corresponding sounds of their own dialect.
Relying on phonetic similarity rather than on a switching-code is really a
4 HEARING WORDS 47

separate mode of listening. Let us call it PHONETIC LISTENING. Perhaps we

should think of it as a submode rather than as a completely distinct mode,
since we resort to it only when seeking to learn the shape of a word (and not
always even then), not when just trying to identify a word. But that subsumes
a number of situations. It is the mode we are forced to use as best we can if
the speaker's diatect is unfamiliar, or when we start out to learn a new
language; it is the submode used by actors learning parts in dialects other
than their own, by mimics seeking to imitate familiar public figures, and by
linguists in the field trying to figure out the sound şystem of an alien lan­
guage. Competence can be increased by systematic schooling in phonetics,
but some people with no such formal training are nevertheless amazingly
skillful at it.
The same two contrasting ways of dealing with alien material appear in our
human handling of nonspeech sounds. Some individuate are able to produce a
remarkably accurate imitation of the barking of a dog, the meowing of a cat,
the neighing of a horse, the chugging of a steam locomotive, or even the
flushing of a toilet, using just their organs of speech. Forms like woof, meow,
neigh, or choo-choo (§1.7) are very different from such imitations. An ono-
matopoeic word is devised by filtering a nonspeech sound through the con-
straints of the sound system of a language, resulting in some rough acoustic
approximation allowed by that system. That is somewhat like monolingual
English users doing the best their English phonemic system will allow in
imitating an expression in some other language, instead of heeding, and
trying to imitate, finer detail. A duck hunter would entice no quarry by
shouting quack-quack. But by setting aside the sound systems of their lan­
guages and resorting to direct acoustic imitation some hunters do very well.

4.9. Summary. In this chapter we have examined the listener's task of

parsing, ignoring for the nonce that in some cases not all the constituent
words of an utterance have to be identified for the utterance to be under­
stood. We have seen that there are several different MODES OF LISTENING, used
under various circumstances and for diverse purposes. The chief of these is
LISTENING FOR WORD IDENTITY; the next commonest is LISTENING FOR WORD
SHAPE. In the former, in addition to simple acoustic recognition, we make use
on occasion of at least three different devices: IMPLICIT MOTOR-MATCHING,
GESTALT PERCEPTION, and SWITCHING-CODES. In listening for word shape, some­
times all that is needed is closer attention or a lowering of ambient noise. But
if the speaker's dialect is alien one may have to resort also to code-switching
or, if the requisite code has not been acquired, to the submode we have called
PHONETIC LISTENING.
 5. SOUNDS, WORDS, AND REDUNDANCY
5.1. The Phonemic Theory. With the substance of the preceding chapter
clearly in view, we are now ready (as foretold in the Introduction, §0.3) for
our foray into phonology. For this we by no means abandon the fundamental
orientation of the essay: it is clearly the hearer's response, rather than a
speaker's activity, that is criterial for the sound shape of forms — even for
speakers, it is what they hear themselves say that counts. Within this frame,
as we shall see, the gestalt approach to perception casts new light on a
number of old questions.
The NORM SHAPES of words, for any one user of a language, are built out of
small recurrent units or features of sound (and of articulation), of which the
total stock, for a single user, is not merely finite but relatively constrained:
from a dozen or so to a hundred or so, depending in part on the language but
also on just how we choose to count them. If two norm shapes differ in sound
at all, they differ in at least one such feature. There are no sets of norm
shapes that grade into one another continuously — as do, for example, the
range of pitches attainable by plucking a stretched string, or the degrees of
loudness with which one can yell Fire!
This assumption, of the discreteness and the efficiently small size of the
sound-articulation repertory, is the first and most important proposition of
the phonemic theory. The empirical evidence for it is so overwhelming and so
well known that it needs no review here. The principle holds whether we think
of the fundamental units of sound as "phonemes," as "allophones," as "dis­
tinctive features," as "components," or merely as "speech sounds" in the
sense in which that term was being used a century ago. Preference for one or
another of these labels depends on aspects of phonemic theory other than its
prime principle, and will not concern us.
It is crucial to remember that what we have just said applies to norm
shapes, not to what Sapir called the "actual rumble of speech" as it would be
recorded by objective apparatus. We have already asserted that the latter
varies wildly, sometimes to the point of not being recognizable even to
experienced native users of the language (see also below, §5.4-5).
This vital difference, between the sound shape of norm shapes, which is
normally what listeners think they are hearing both in their own speech and
in that of others, and the actual sound of actual utterances, can be under­
scored anecdotally.
In 1938 I spent about two hours recording words in Peoria, a southern
Central Algonquian dialect, from an elderly woman who was surely one of its
last surviving users (even for her it had been dormant since childhood, when
5. SOUNDS, WORDS, AND REDUNDANCY 49

she had learned it from her grandparents). In such a brief session, my tran­
scription was necessarily impressionistic. That means that I wrote down, in
phonetic symbols, WHAT I THOUGHT I HEARD. The degree of phonetic precision of
such a transcription varies with hearing conditions, the keenness of the
recorder's ears, and the thoroughness of the recorder's training, but no
matter how favorable these factors may be the transcription necessarily
remains impressionistic. A linguistically accurate record, in contrast, would
have shown not what I thought I heard but WHAT THE INFORMANT THOUGHT SHE
SAID —which is to say, what she heard herself say when saying the words with
clarity-norm enunciation, not the actual sound but that sound INTERPRETED in
terms of the articulatory-acoustic targets at which she was aiming.
That is what Bloomfield was talking about when he claimed that in the field
he never got anything right for the first month. He didn't mean he was a poor
phonetician. He meant he couldn't hear the words right until he learned enough
of the language to know what they are supposed to sound like. To hear Me­
nominee correctly he had to hear it the way the Menominee do; to hear
French correctly one must hear it the way the French do. Neither of these is
the way we users of English hear English, nor the way we hear Menominee or
French, Nor is either the way an expert in phonetics hears Menominee or
French or English or any other language. Phoneticians, to the limits of their
expertise, hear words the way they DO sound, uninfluenced by any knowledge
of how they are SUPPOSED to sound, and for most linguistic purposes that is
fatal.
It is instructive to consider here, as a case in point, William Labov's report
from Martha's Vineyard, For the diphthong in house he records a dense range
of pronunciations. In all of them there is a glide from a central vowel toward
high back rounded. At one extreme of the range the beginning of the diph­
thong is low (thus something like ah-oο), and at the other it is mid (uh-oo).
Labov has a superb ear, so that his transcriptions are phonetically very
reliable. But, like me with my Peoria informant, he was recording what he
thought he heard, not what the speakers thought they were saying. Such data,
no matter how accurate, form no valid basis for historical or methodological
inferences except in the company of a proper interpretation by the funda­
mental principle of the phonemic theory. I cannot be sure what that inter­
pretation would be, but I suspect it is something like the following, Martha's
Vineyard English has two norm shapes for house (as doubtless for many
other words). One is indigenous, the other imported; the two are thus perfect­
ly ordinary doublets (§4,8), Speakers aim sometimes at an articulatory-
acoustic target near the higher extreme of Labov's range, the local norm for
the word, and sometimes, in imitation of the "summer people," at a target
near the lower extreme. Likewise, when hearing someone else say the word,
they tend to interpret it the one way or the other. The targets, thus, are
discretely different. That there should be a spread in the actual shots aimed
50 5. SOUNDS, WORDS, AND REDUNDANCY

at the two bull's-eyes is only to be expected, nor is it at all surprising for the
two ranges of distribution to overlap.

5.2. Words. Before carrying this exposition any farther I must interrupt to
try to justify my casual use, so far, of the term "word" (§0.3).
It is obvious that the concept of the word plays a central role in theories
about language. In grammar it has traditionally served as the boundary be­
tween the two great compartments of syntax and morphology. In phonology
its proper role has been the subject of extensive debate, still unresolved.
Some have been content to take words as givens and go on from there. In a
way, that may be legitimate, but surely not without explicit exegesis.
In 1926 Bloomfield defined words as MINIMUM FREE FORMS. That is, a word is
a form that can on occasion occur as a whole utterance, but at least one of
whose immediate constituents cannot so occur. So we have boy, boys, boyish,
boyhood, but not -s, -ish, -hood, which are not free; and not good boy, Boy
Scout, which have only free constituents.
Bloomfield's approach rests on a prior definition of FORM: to wit, any
meaningful (and hence potentially recurrent) part of an utterance. Wells
clarifies this by pointing out that the sound-sequence a man are, although it
occurs whenever someone says (for example) The sons and daughters of a
man are his children, is not a form because it is not meaningful. The ap­
proach also requires that we distinguish between a form and an OCCURRENCE
of a form. It does no harm, and can be helpful—perhaps even true —to think
of forms (not their occurrences) as registered somehow in the brains of
users of the language, as well as (sometimes) in reference works such as
dictionaries, ready to be replicated and used as needed: any such use is then
an occurrence of the form.
By and large, the forms that pass Bloomfield's test, in English and in other
familiar languages, are just what we were all calling "words" anyway. So what
Bloomfield's "definition" actually did was to make explicit something we (as
language-users) already knew but had not verbalized, revealing a criterion we
seem in fact to use in recognizing words in our own speech, and guiding us
(as linguists) in identifying them in the analysis of alien languages.
Viewed in this way, it turns out, not surprisingly, that the criterion doesn't
quite fit — doesn't quite give us what, for some reason, we seem to want.
Sometimes it yields a form that we would prefer to think of as less or more
than a word; sometimes it fails to force a choice between alternative deci­
sions. English the and a/an ar at most marginally free, but we certainly find
it convenient to treat them as words. If's is a bound form, then the immediate
constituents of the man who he/ped us open the front door's wife are the man
who he/ped us open the front door's and wife, and the first of these is
certainly not what we should like to call a single word; but the alternative, to
call 's a word, is equally uncomfortable. Back BIRD is two words; is
5. SOUNDS, WORDS, AND REDUNDANCY 51

BLACKbird two or one? The criterion allows either decision and forces
neither.
So in our traditional habits of word-occurrence definition there seems to be
some criterion other than, or in addition to, Bloomfield's.
One of these inheres in the typical shapes of forms. Evidence of this sort
is plainer in some languages than in others, but never so simple as some­
times made out to be. For example, in Finnish, Hungarian, and Czech, it is
claimed, a stress falls on the first syllable of each word (and nowhere else).
But words that are functors (§3.3) may be atonic; the second elements of
compounds can lose or reduce the stress; and the stress may begin only with
the first vowel of the syllable, or it may begin with the final consonant of the
preceding word before a word-initial vowel rather than with the first seg­
mental phoneme (vowel or consonant) of the word itself. So even this clue to
word boundary, and thus to word identity, may in some cases be obscure or
partial.
Nevertheless, clues of this kind, however complicated, may be significant
for the users of the language. Consider the following English example. In Walt
Disney's Mary Poppins an effort was made to foist on the public, as a single
word, the expression supercalifragilistic expialidacious. Of course that was
fraudulent; it is a two-word phrase. It is irrelevant that much of the material
in the expression is nonsense. The elements -ist and -ic typically occur close
to the ends of words, and only a limited number of other bits can follow them.
Similarly, ex- is prefixlike —not necessarily initial, but close to the beginning
of the word and with only certain pieces allowable before it. So if the expres­
sion is English, as it was certainly intended to be, then there is a word
boundary exactly where I have shown it above, between -ic and ex-.
Still a third criterion is implied by the discussion in the preceding chapters
of this essay: for a listener, a word is a RECOGNITION UNIT, a PERCEPTUAL
GESTALT. If you hear someone say I drove downtown yesterday you recognize
each of four successive stretches of speech-sound as a familiar hunk. In
your processing of what you hear the whole stretch of sound is segmented
as the orthography indicates. You do not take the sound of Idroved or
rovedow or townyes as a unit because there are no such units. This criterion,
like BloomfiekTs, is certainly not perfect: that is, it does not always yield just
what we think we want. Especially, it fails to distinguish between words and
idiomatic phrases, which are assuredly recognized, just as are words, from
their overall acoustic patterns; this is a fact of crucial importance on which
we will have more to say later (§5.4). For that matter, in the above example I
am not sure that we should class the adverb downtown as a single word
rather than as a phrase — despite the tradition of writing it with no hyphen or
space—though either way it is an idiom and a recognition unit.
In many cases these three criteria (and others as well?) all agree, and then
there are no disputes about word identity. Sometimes there is at least ma-
52 5. SOUNDS, WORDS, AND REDUNDANCY

jority rule. Sometimes there is not even that, and whenever the criteria
disagree then so do people. But maybe arguments in such cases are beside
the point. I think we have here one more situation in which our scholarly
preference for sharp lines of demarcation and unambiguous categories is
discordant with reality. In fact, word identity in the stream of speech ISN'T
always straightforward and unambiguous. Rather, it appears that there are a
number of "degrees of cohesion" between successive identifiable elements in
an utterance. If the cohesion is loose enough, everyone agrees that there is a
word boundary; if it is tight enough, everyone agrees that there is not; but
there is no single degree of cohesion that can serve securely as the dividing
point between loose enough and tight enough.
The whole issue of words in utterances, versus segments smaller or larger
than words, should be thought of as one aspect of spacing (§1.3). Among
literate users of English, spacing disagreements are revealed by punctuation.
There is great variation in the use of commas and semicolons, both of which
presumably denote a separation (be it syntactic, phonetic, or both) of an
order higher than that usually shown by mere word space. But there is also
great inconsistency in the three-way choice among word space, hypheń, and
no-space, as in field work, field-work, and fieldwork, all of which are written
though the pronunciation, as I hear it, is invariably FIELD work with dimin­
uendo stress.
In this essay, with our primary emphasis on the hearer, we shall henceforth
take the recognition-unit criterion for word identity as the basic one: A PER­
CEPTUAL GESTALT IN CONNECTED SPEECH IS A WORD UNLESS THERE IS CLEAR
EVIDENCE THAT IT Is MORE THAN A WORD. If it is indeed a word, then it is also, in
most cases, a minimum free form, but we take that property to be
secondary. Although we cannot expect this shift of emphasis to solve all our
problems, at least we will see them from a different angle.
5.3. The Word in Phonology. The one feature of phonological theory in
which we can have complete confidence is that a norm shape of a word has a
determinate phonemic structure for any one user of the language. That is, a
single word, uttered with CLARITY-NORM pronunciation and listened to for WORD
SHAPE by a native user of the language (thus, at least by the speaker), has
such a structure imposed on it by that user's habits of perception and
interpretation. That is about as much as had been claimed by classical pho­
nemic theory as late as the 1940s, in any of its semiindependent recensions
(Geneva, Prague, Copenhagen, British, American); all of those schools of
thought dealt almost exclusively with WORTPHONOLOGIE. It is also all that was
stated or implied in the preceding chapter in the first section of this one.
What, then, of SATZPHONOLOGIE? — what about the phonemic structure of
utterances produced in ordinary, rapid, conversational interchange, and lis­
tened to for word identity rather than for word shape? Setting apart obvious
5. SOUNDS, WORDS, AND REDUNDANCY 53

contractions (like don't for do not, for which see §5.4), do words uttered as
parts of longer utterances retain the phonemic structure and identity they
have in isolation?
The classical theory, to the small extent that it faced these questions,
proposed various alternative answers, none definitive. Then, around the turn
of the half-century, there was a lively debate about it in American linguistics.
Both sides cast their arguments in the operational frame of reference
prevalent in that tradition at that time: the orientation of thefieldworker,who
wants to know what tests to use and how to interpret the results to achieve
an accurate and useful description of a language. The issue: RESOLVED, that
words and word boundaries must be known and heeded before phonemic
analysis can be accurately completed. The pros were represented most
vigorously by Pike. Surely the most strident voice among the cons was mine.
The matter was never settled. The argument was simply broken off in the
middle as a new and seemingly revolutionary dispensation (TG) preempted the
podium.
The altered frame of reference under development here does not solve the
problem, but does shed some new light on it.
It turns out that a good deal depends on whether one thinks of words
primarily as minimum free forms or as recognition units.
If we accept the Bloomfield criterion as a definition, as most of us did in the
1940s and '50s, it follows that words are FORMS, and hence lexico-grammatical
units, defined by semantic criteria (§5,2), On the other hand, the analyst's
decisions about phonemics are based not on meaning but on sound. This view,
the reasons for it, and its consequences were set out perhaps most plainly
by Wells (and notice his significant reference to the hearer):
A purely phonemic transcription, by definition, records all and only the significant
distinctions that can be heard If two utterances or parts of utterances sound perfectly alike
to native speakers of the language to vhich they belong, their purely phonemic
transcription is identical, even if they differ in grammar or lexicon.
Th rationale of this stipulation is clear. Phonemics takes the point of view of the
hearer. Now the hearer, in order to interpret correctly an utterance that he hears, must
rely on t v o separate sources of information: (a) the heard sounds (supplemented, it may
be, by the sounds of previous or folioving utterances); (b) the extra-linguistic context
(including his knovledge of vhat the utterance may or must mean). For the purpose of
sharply distinguishing betveen vhat can be learned from one source and vhat can be
learned from the other, phonemics makes a point of recording nothing but vhat is
conveyed by (a). All else belongs to grammar (and lexicography). This is vhy one vishes
to avoid intersection ... in phonemics but not in grammar. ... We do not say that
phonemics and grammar must be separated, but only that the separation is feasible, and
serves the above purpose.
54 5. SOUNDS, WORDS, AND REDUNDANCY

Given the Bloomfield criterion and this orientation, it would seem to follow
that words by their very nature can be of no phonological relevance. Even if
there are phonetic phenomena in some language whose occurrences regularly
and unambiguously mark word boundaries, it is only those phonetic phenom­
ena that are phonological entities, not the words or the boundaries.
But there are two catches in this.
First, it involves accepting Bloomfield's minimum-free-form criterion as a
DEFINITION, whereas, as we saw in §5,2 we prefer to interpret it merely as a
by-and-large characterization of units whose identity is independently known
to us.
Second, Wells speaks of things "sounding perfectly alike to native speak­
ers," I think this criterion is indispensable. But what determines whether two
utterances or parts of utterances sound the same or sound different to
users of the language? Hearing speech is an enormously complex skill. Might
not a user hear a word boundary at a certain spot in an utterance even
though no phonetically sophisticated field-worker, no precision laboratory
equipment, could detect anything special at that spot?
These uncertainties both point toward our preferred alternative to the
Bloomfield approach: an acceptance as basic of the recognition-unit criterion.
From that we immediately get an affirmative answer to the question just
posed. To be sure, native users may hear a word boundary at a certain point
partly because of features audible at that point — the sorts of features N,
Trubetzkoy called DEMARCATORS, English word boundaries are sometimes,
though not invariably, so marked: witness the classic example ceased
aching : cease taking : see staking, and note that gray-day milk is not
necessarily grade-A milk, that up-lying strata need not be applying for
anything; that the beginning of the game need not be a big inning. But they
may also hear a boundary at a point where there is no such characteristic
audible feature, because a perceptual gestalt, a contour of sound up to the
point, is a recognizable word, as is a contour of sound starting at that point,
whereas no contour stretching across the point is. The recognizability of the
contours may be enhanced by the kinds of phonetic features Trubetzkoy
called CULMINATORS: e,g,, each word (or each major word, as over against
particles) may bear one and only one stress, so that somewhere between two
successive stresses there must be a word boundary. An atomistic bias might
make us assume that listeners can only infer segments from boundaries. It
seems more likely that they often infer boundaries from segments.
If we turn for a moment to writing instead of speech, we quickly find that
inferring boundaries from segments is no trick at all. The Greeks and Romans
managed just fine with no spaces between words, and so can we, as shown by
the following quote from John Steinbeck:
5. SOUNDS, WORDS, AND REDUNDANCY 55

youcanboastaboutanythingifitsallyouhave
For speech, a concrete example ( / drove downtown yesterday) was pre­
sented in §5.2; here we state the case in general terms. Let us symbolize the
continuous train of sound waves (the speech signal) of utterance (a) by the
sequence of letters ABCDEFGHIJ. Experienced users of the language, hearing
this, recognize the segments ABCD, EEC, and HIJ as words, but not ABC,
DE, DEF, or the like. By virtue of that, they hear a word boundary between
D and E and one between G and H Perhaps there could be another utterance,
say (b) KLMNDEOP with a word boundary between KLM and NDEOP, where
the piece DE, acoustically identical with the DE in (a), is all in the same word.
But that would not matter — it would not interfere with the users' correct
locating of word boundaries in utterance (a).
Of course, such interference can also happen, in either writing or speech,
though mostly in trivial bits. Our practice of not always writing a hyphen
between the members of a compound produces some real ambiguities and
some bogus ones, the latter being instances where an incorrect spacing by
the reader can yield a nonexistent though realistic form. Thus, I once care­
lessly read "tightrope" as TIGH-trope before hastily correcting myself. Early
in their reading careers, children may guess incorrectly at an unwritten
boundary; to this day, when I see "awry" or "bedraggled" my first impulse is
to convert them into sound as AW-ry and BED-raggled. Somewhat less trivial
is the tale I once heard from a Methodist minister, about the militant atheist
who carelessly scrawled "God is nowhere" on his child's slate:
GODISNOWHERE
—whereupon the child read it as "God is now here."
in speech, for the general case, imagine an utterance (c) ACEF/KMO that
can be recognized as consisting of words ACE and FIKMO or, equally well, of
words ACEFI and KMO (but not of ACE, Fl, and KMO, there being no word of
shape FI). A speaker might intend it one way and a hearer interpret it the
other; or different hearers might hear it different ways; or a single hearer
might hear it one way, realize that the interpretation doesn't fit the context,
and do a double take; or, finally, the speaker might intend the utterance as
ambiguous — as a pun. The only English examples I can think of at the
moment are inexact, because we have too many demarcators. By "inexact" I
mean that there is structural ambiguity only if the enunciation is slightly
careless, or the setting a bit noisy, so that the hearer has to fill in from
spotty acoustic evidence. Under such conditions, I once misheard the name
Juan Bigby Hynd as one big behind, and once misidentified a decanter as
Eddie Cantor— but under such conditions, if the noise is bad enough, anything
can be misheard as anything else. The examples are realistic only because
most of our vocal-auditory communication takes place in fairly noisy
settings. A more convincing case is the following famous French couplet,
56 5. SOUNDS, WORDS, AND REDUNDANCY

because even with theatrically limpid delivery, unless the speaker pauses at
the commas either line can be misinterpreted as the other:
Gal, amant de la reine, alla, tour magnanime,
Gallament de l'arène, a la Tour Magne, a Nîmes.
Under our approach, which defines words primarily as perceptual gestalts, I
propose that words ought to be thought of as neither phonological nor gram­
matical, but as superordinate to that dichotomy and relevant for both. If we
should decide that recognizing words and word boundaries is a prerequisite
for correct phonemic analysis, that nevertheless would entail no "mixing of
levels" (as the jargon used to put it), because it would be neither a GRAM­
MATICAL prerequisite for PHONOLOGICAL analysis nor a PHONOLOGICAL basis for
GRAMMATICAL analysis; instead, it would be a holistic background requirement
for either and both. That has been the avowed or implied opinion of many
linguists, especially those who have silently taken words as givens (§5.2). In
the light of the apparently unavoidable fuzziness of our terminology, in par­
ticular the multiple and changing usages of such words as "grammatical," I
cannot know for sure that this is the view Pike was trying to promote in our
mid-century debate— for this to be so his meaning of "grammar" at that time
would have had to be strikingly different from mine —but I want to give him
the benefit of the doubt and say it is, and add that, although our approaches
were from opposite directions, I now think he was as ctose to the truth as I
was.
I am not sure I fully understand the consequences. Consider our sample
utterance (c) ACEFIKMO. Remember that this is supposed to be a unique
historic event, produced by a particular speaker at a specific place and time
within earshot of certain specific hearers. Our proposal is that, even though
the hearing conditions are excellent and the articulation careful, the audience
cannot tell EXCEPT FROM CONTEXT whether its word structure is ACEFI KMO or
ACE FIKMO,but that some in fact take it the one way and some the other.
Are we to say in such a case that the single utterance has one phonemic
structure for those who interpret it one way, a different phonemic structure
for those who take it the other? Then how about a participant — e.g., the
speaker, intending it as a pun — who hears it both ways? Does the single
utterance have two different phonemic structures for the same person at the
same time?
Despite my insistence on the criterial role of hearers, I am not comfortable
with this outcome. I do not wish to abandon Wells's principle, only to modify it
to allow for reliance on the native user's hearing and for gestalt perception. I
do not like to abandon the notion that complete homophony for native users
means phonemic identity, but it may be that there is no alternative.
5. SOUNDS, WORDS, AND REDUNDANCY 57

5.4. Slurvian. Clarity-norm articulation and word-shape listening go together

in that the latter often elicits the former. And listening for word shape is, I
think, sharply different from ordinary listening for word identity (even though,
as we shall see in a moment, one can do both at almost the same time). But
clarity-norm speaking is not in the same way discretely different from
ordinary speech. Rather, it falls toward one end of a scale of precision of
enunciation, at the other end of which one finds speech so badly mangled that
the constituent words are entirely unidentifiable. Most talking strikes this
scale somewhere in the middle, with occasional rises toward greater clarity
and somewhat more common drops into mumbling.
In recent years, trying to gather evidence on all this, I have been observing
myself from time to time as a listener. It seems to me that much of the
speech I hear sounds just the way it "ought" to sound. That is, if I con­
sciously pay heed to the shapes of the words, rather than just letting the
word-identities register out of awareness, mostly either those shapes are the
ones I myself use or else they are other familiar norm shapes of "the same"
words, I don't imagine that the speech signal reaching my ears always con­
tains the full quota of stigmata for what I think I am hearing; surely the
seeming completeness is often the product of gestalt closure. And some­
times, although there is no question about word identity, I cannot be quite
sure which shape was uttered: for example, I have often recognized recognize
without being certain if it was pronounced recognize or reckonize
On the other hand, some utterances, particularly of high-pitched voices or
heard on television, escape me altogether: the clues in the signal are fewer
than the minimum my speech-decoding circuits must receive to fill in the
rest. Since I have some presbyotic deafness, the relative incidence of such
opaque passages is higher for me now than it used to be, and doubtless
higher than for younger people in general, but in every other respect my
experience is surely typical.
The speech I hear also contains many pieces of SLURVIAN: words or phrases
carelessly articulated and hence poor in acoustic clues but nevertheless
understood with no apparent trouble. These are quite different from the
passages I cannot hear well enough to decipher.
Investigators confronting such slurred material have often been baffled
trying to determine its phonemic structure, I vividly remember, for example,
trying to transcribe the ordinary slurred variant of Fijian 'thank you' (clar­
ity-norm shape Vináka vàka-lévu), and finally giving up in desperation. But I
have now found the reason for the difficulty. The reason is that slurred
expressions DON'T HAVE any phonemic structure.
No utterance, after all, has any phonemic structure except as imposed by a
hearer, and I propose that familiar slurred expressions ordinarily escape the
imposition, I think we normally identify such an expression directly from its
context and its gestalt contours _ having under-
58 5. SOUNDS, WORDS, AND REDUNDANCY

stood it, we don't bother — or our speech-decoding circuits, which normally do

the work for us, don't bother — to flesh out the perceptual skeleton to obtain
a full-blown norm shape. If this is so, then to some extent our understanding
of familiar slurvian is not dependent on the sound system of the language, or
even on the general fact that human languages have sound systems, but is
more like the way gibbons seem to understand one another's calls: holis-
tically, without dissecting them into constituent sound-features.
In producing a slurred expression it is as though speakers just nod their
articulatory organs casually in the general direction of the clarity-norm shape
of the phrase. The pitch and volume go up and down, but not so far as in
clear speech; the voiced-voiceless contrast is weak; articulatory closures are
imprecise or skipped; in moving from position to position the jaw, tongue, and
lips cut corners, partly muffling the vowel color. One and the same under­
lying clarity-norm phrase may show various degrees of slurring under varying
circumstances, but in the right context even drastically reduced versions
may be understood. However, whether a carelessly articulated expression is
slurvian depends not on its production but on its reception: for a hearer who
merely registers the acoustic outline and grasps the expression as a whole, it
is slurvian; for one who hears it in terms of clarity-norm pronunciation, or
who cannot make it out at all, it is not.
All this is hard to show in a written report, since to transcribe a slurred
phrase (even in ordinary spelling) is to impose a phonemic structure on it.
But that is part of my point. I suggest that our everyday speech makes use
of a great many slurred expressions (far more than we commonly realize),
and that although ordinarily we accept and digest them as wholes, there are
also circumstances in which we have to "phonemicize" them (that is, impose
a phonemic structure on them). Furthermore, in doing so we do not always
revert to the norm shape of the corresponding clarity-norm phrase, but come
out instead with something more compact. And I propose it is just in this way
that RESTRESSED FORMS and CONTRACTIONS arise.
Think of a phonemic norm shape as a crystal. In ordinary speech many
crystals retain their identity, but others go into solution. A slurred expres­
sion is in solution — fluid, amorphous. But it can be precipitated out, either in
its original shape or, often, in a different and simpler one.
The atonic shapes of functors mentioned in §3.8 are instances of slurvian,
and are sometimes precipitated out of solution in new shapes. For example,
we have a functor whose clarity-norm norm shape is represented by the
spelling have, but which in context is usually atonic. The slurred atonic shape
can hardly be transcribed, but it does get restressed on occasion as of
(pronounced uhv or ahv) rather than as have. The normal atonic (and slurred)
shapes of norms would and had are acoustically so similar that in getting
precipitated out of solution they are often confused.
We have a phrase whose norm shape is (or is represented by the spelling)
5. SOUNDS, WORDS, AND REDUNDANCY 59

do not In ordinary speech this phrase is almost always in solution, some­

times by itself, sometimes as part of a larger slurred expression. It would be
a mistake to think that the slurred shapes are what we write as don't. The
slurred shapes are much less coherent than what we indicate with that
spelling, which represents, rather, the clarity-norm shape of a standardized
recrystalized contraction.
Again: try to imagine a real, normally sloppy utterance hinted at in written
shape by Whyncha go with us? In order to represent this in writing I have had
to recrystalize the included slurvian phrase: the spelling whyncha may HINT at
slurvian, but it REPRESENTS the clarity-norm shape of a contraction just as
don't does, even though in this case the contraction has not been
standardized.
How do users rephonemicize a slurred form? If the unslurred and unshort-
ened equivalent is obvious, or if there is a familiar standard contraction,
there is no problem. Otherwise it is done by finding a clarity-norm phonemic
structure which, when aimed at in a sufficiently slovenly manner, will yield an
acoustic contour like that of the slurred form they seek to match. The proc­
ess is thus much like those by which users create onomatopoetic forms or
adopt forms from other dialects or languages into their own (§4.8).
Novelists have to do just this, as best they can within the constraints of
orthography, when they want to present realistic dialog. Short of recording
their story in oral form instead of in writing, there is no way they can
directly present the real slurred phrases of conversational prose. They have
to resort to recrystalizations, either standard or coined for the occasion. P.
G. Wodehouse had fun coining them: he invented orthographic "R?" (to be
read as the name of the letter, with British rless pronunciation) as a request
for clarification, and the contraction "Q!" (also to be read as the name of the
letter) for Thank you, the latter certainly resembles my own typical grunt of
acknowledgment.
Linguists in the field, trying to transcribe slurvian forms as I tried with
Fijian 'thank you', even if they resort to a rich repertory of phonetic distinc­
tions and symbols, are likely, despite their best intentions, to be doing just
what novelists do — except that since they are not native users of the
language their recrystalizations are bogus.
We can think of slurvian as a more primitive communicative system (more
primitive because it lacks duality of patterning) that stands in a sort of
symbiotic relation with ordinary language. Expressions constantly get im­
ported into slurvian from ordinary speech, by the simple mechanism of being
slurred and yet understood (if they are not understood the speaker resorts
to clearer articulation or to paraphrase, and no slurvian has been born).
Transfers in the opposite direction sometimes keep alive the tie between a
slurred expression and its clear-speech prototype, and sometimes create
contractions.
60 5. SOUNDS, WORDS, AND REDUNDANCY

Slurring followed by contracted recrystalization is one kind of RESHAPING

(different from that to be discussed in §6.7). English has quite a stock of
words known or believed to be the result of this sort of reshaping. All our
words with initial voiced th (as in this) ar functors, and all are restressed
reshapings of earlier words with voiceless th (as in thing). ( is a restressed
byform of off. Yeah, yep, and nope are recrystalizations of slurred yes and
no. Goodbye is supposedly from an earlier God be with your, ornery is a byform
of ordinary, maim is from mayhem, bed/am from Bethlehem, maudin from
Magdalene, and frantic from phrenetic. Sir William S. Gilbert (of Gilbert and
Sullivan) recrystalized the local slurred shape of Saint Mary Axe (a London
street) as Simmery Axe In Massachusetts, Worcester retains the spelling of
the British original, and a slurred pronunciation much like that of the original;
Wooster, in Ohio, is a recrystalization.
fn addition to items imported from ordinary language, the slurvian by-sys-
tem also contains material which, as far as we can tell, is native to it and
which collectively may even even trace back aff the way to the period before
duality of patterning appeared and prelanguage became language. These are
the various "vocal gestures," of assent and dissent, acceptance and rejec-
tion, surprise, annoyance, pleasure, and the like, which have long been rec-
ognized as not conforming to the ordinary phonemic habits of the language.
Like other slurvian items, they are sometimes precipitated out. For example,
in English slurvian there is a vocal gesture used to invoke silence; it is not
precisely transcribable, but somewhat resembles the sound of the two words,
hush and shush, that are its crystalizations. We don't call these words
contractions because there are no uncontracted equivalents (dictionaries
characterize many of them as "of imitative origin"), but otherwise they be-
long with don't,whyncha,goodby, ornery, and so on.

5.5. The Uses of Redundancy. Although we have mentioned redundancy by

name only twice so far (once in §3.4 and once in §4.5), its spirit has haunted
almost everything we have said. Before returning (in chapter 6) to our main
theme we must take a few moments to deal more directly with this crucial
phenomenon.
Readers of A. A. Milne's Winnie-the-Pooh will recall the account in chapter
4 of Pooh-Bear's first visit to Owl. Owl's home was particularly noteworthy
because its door, instead of having just a bell-pull or just a knocker, had
both. Furthermore, at Owl's request Christopher Robin had supplied some
signs for the guidance of visitors: those who wanted a response were to ring;
those who did not were to knock. Pooh read these signs carefully, forward and
5SOUNDS,WORDS,AND REDUNDANCY 61

backward, and then pulled and knocked the bell-pull and knocked and pulled
the knocker, all the while loudly calling out his identity and that he wanted an
answer. That worked, or some part of it did, for Owl opened the door.
In his actions Pooh was exploiting the principle of communicative redun­
dancy: to make sure the essence of the message gets through, despite
diverse sorts of unpredictable interference along the way, you send it (or
receive it) several times, or via several channels, or —most effective of a l l -
in a perceptual shape so different from that of any message with some other
meaning that, even if mangled in transmission, it will still be unmistakeable.
This principle has worked its way, by natural selection, into the essential
fabric of language, in that words with embarrassingly different meanings, if
likely to occur in identical or similar environments, tend to be acoustically
very different from one another. Bloomfield explains the more recent history
of Old English /ettan 'impede' and lǣtan 'permit' on just this basis. Both yield
modern /et. But, given their meanings, for either of the two to be
misinterpreted as the other could be annoying or even disastrous. Speakers
therefore learned to paraphrase, and in due time the 'impede' meaning was
lost except in a few fixed expressions (without let or hindrance, a let ball)
which are quite commonly misunderstood or reshaped (a net ball.)
Absolute identify of sound is not necessary for this outcome. A couple I
know named their son Loren, which is acoustically very close to the father's
name, Warren; by the time the son was six they were calling him Luke. Ora/
and aural are enough alike that the latter is often replaced or supplemented
(at least among professional workers with language) by auditory. Yet the
factors involved in such competition of forms are too diverse and too
specific for safe prediction in the individual case. One might expect the
homophony of raise 'build up' and raze 'tear down (completely)' to lead to the
abandonment of one of the two, but they have both been around for a long
time with no signs of any such outcome. While watching football on television
I have trouble distinguishing quarterback and cornerback, yet the terms
persist.
There is a tale, probably apocryphal, of an American copilot whose inability
to distinguish clearly in speech between au dessus and au dessous led a
French pilot to crash trying to fly under a bridge instead of over it.
Another way to express the principle is to say that similarity of sound
tends to match, and to be matched by, similarity of meaning. (This is more
general than our Proximity Principle of §1.8, which could be regarded as a
special instance of it.) In the case of bee dancing, the redundancy principle in
62 5. SOUNDS, WORDS, AND REDUNDANCY

this form has won out completely. The angle between the line of the dance
and the vertical, in the hive, indicates the direction from the hive to the
source of nectar. I don't know how accurate the indication is, but an un­
certainty of five degrees would leave a target region less than nine meters
across at the end of a worker's flight of 100 meters, whereupon she could
easily zero in using vision and smell. The paralinguistic effects that ac­
company speech tend to work in a similar iconic fashion: if you express slight
anger by slightly raising your voice, then for more anger you raise it more. In
language proper, the phonemic principle quantizes the system of all physio­
logically possible speech sound — chops it up into discretely contrasting
regions — so that two utterances cannot be indefinitely similar in distinctive
sound: they are either identical or they differ in at least one whole phono­
logical unit (§5.1). Yet we do tend to give similar verbal responses to similar
sets of stimuli (otherwise we would never be understood!), and do tend to
interpret appropriately similar utterances in similar ways: the principle does
not have free play, but manifests itself to the extent permitted by the
workings of other, conflicting, features of language.
No matter what degree of magnification we use in our examination of hu­
man communication, and regardless of the angle of approach, we find redun­
dancy. Considering speech at the smallest size-level, we find that the pho­
nemic principle, as described in §5.1, is a sort of built-in redundancy: if the
construction of our speech organs and the keenness of our hearing allowed a
language to quantize the system of all physiologically possible speech-sound
into thousands of contrasting regions, instead of into hundreds or dozens
(§5.1), speech could carry ten or more times as much information per
successive sound unit. There is also redundancy in the fact that by no means
all combinatorially possible arrangements of the distinctive sounds of a
language occur as the shapes of forms — actually, only a relatively small
proportion of them do. Not all sequences of phonemes (or speech sounds, or
letters) make words; not all sequences of words make utterances. Of those
that do, some occur much more frequently than others, which increases the
redundancy even more.
And that is just language viewed in a vacuum. When we put talk back into
its natural setting, as part of communicative packages transmitted and re­
ceived in the course of human transactions, we find again what we remarked
on in §3.9: the total packages are in general so redundant that any relatively
small part can be masked out by noise without serious injury to mutual
understanding.
That is what makes slurvian possible. It is very rarely the case, in the
ordinary run of talk, that a word must be distinguished carefully from those
whose norm shapes differ from it by only one or a few distinctive features of
sound (recall the discussion in §4.5). Usually all the other words that might
conceivably occur in the same context sound quite different, so that the
5. SOUNDS, WORDS, AND REDUNDANCY 63

range of variation that will not lead to misunderstanding is large. Hence,

anyone who happens to be listening for word shape instead of just for word
identity will frequently hear TOTAL phonemic "overlap"—a b or an f for a norm
shape p, an n for a d, or even greater deviance from what the sounds would
be in carefully enunciated speech. And usually we understand anyway. Only
when we don't does the speaker turn toward clarity-norm enunciation.
The redundancy of the total communicative package also explains why we
can often understand even an "imparsible" utterance: one whose parts just
can't be fitted together by our syntactic patterns to make a coherent whole
(§2.4). Handling such material is like trying to assemble a jigsaw puzzle with
a few pieces missing, or with a few extras, or both. Given sufficient context,
we can supply the missing pieces or discard the extras, and thus get the
picture, or something enough like it as usually to make no practical dif­
ference.
Thus it is only at a first careless glance that one might see redundancy as
wasteful, something to try to eliminate or at least minimize. True enough,
redundancy is sometimes redundant, which is to say that there is sometimes
more of it than we need. That, however, is the price we pay for only rarely not
having enough. For when we consider the matter more fully we realize that the
undesirable thing is not redundancy but noise. There are circumstances under
which noise can be reduced, but to get rid of it altogether is a PHYSICAL
impossibility, just exactly as is the construction of a perpetual-motion
machine (and, at bottom, for the same physical reason). And the only
possible weapon for combatting noise is redundancy. Without a good measure
of well distributed redundancy, human communication as we know it could not
be carried on.
More than that (and often overlooked): it is only by virtue of redundancy
that the conventions of human communication can be LEARNED.
If the context in which you hear, or a small child hears, an unfamiliar word
serves to imbue that word with some meaning, it must be that there is over­
lap among the meanings of different ingredients in the whole communicative
package, which means that the whole is redundant. Otherwise you, and the
child, would have no clue, and the new word would remain opaque. The earliest
sense learned for a word not uncommonly deviates considerably from the
community norm —as in the case of the child who described a newspaper by
saying that it is thrown on the front porch and mama wraps up the garbage in
it. That is hardly a satisfactory adult definition of the word newspaper, yet it
does incorporate a measure of experiential truth. And relearning, readjust­
ment, and intercalibration with the communicative conventions of other
people continue not merely through childhood but as long as we live.

5.6. Summary. In this chapter we have singled out the DISCRETENESS and the
EFFICIENT SMALL SIZE of the repertory of sounds that distinguish norm shapes
64 5. SOUNDS, WORDS, AND REDUNDANCY

as the foremost principle of the phonemic theory. We have characterized

words as primarily PERCEPTUAL RECOGNITION UNITS rather than as minimum
free forms. Following Pike, we have proposed that words are basically neither
phonological nor grammatical but stand outside that dichotomy, and have
suggested that hearers, listening in terms of gestalt, may infer boundaries
from the abutment of recognizable contours even at points where there is no
characteristic phonetic marking. We have outlined a theory of slurring and
"recrystalization" that may help to explain fast-speech forms, restressed
by-forms, and contractions. Finally, we have taken a more systematic taok at
redundancy and its significance for language.
 6. WHY MORPHEMICS WONT WORK
6.1. A Fish Story. Some years ago I took our rowboat out on the river for a
bit of fishing. When I had shipped the oars and was reaching for the tackle,
my wallet slipped out of my pocket and fell overboard, directly into the open
mouth of a Japanese goldfish. I grabbed for it, but the fish leapt out of the
water and tossed it to another fish. That's known as carp-to-carp walleting.
Either a grin or a groan in response to this joke acknowledges that it has
had some impact. No theory of language that does not make provision for
that impact, it seems to me, is deserving of serious attention. Yet some of
our most widely held notions militate against such accountability.
Those notions, though they have earlier roots, achieved their current reg­
nant position only in the 1940s and 1950s, when there was a great prolif­
eration of morphological theory, especially on this side of the Atlantic, At that
time, in attacking the grammatical-lexical aspect of languages we were
following a pattern that had proved eminently successful in dealing with their
sound aspect, and were thus proposing that MORPHEMES are, on the gram­
matical level, what phonemes are on the phonological. Jerzy Kuryłowicz even
gave this parallelism a name: borrowing a term (though not a concept) from
mathematics, he called it ISOMORPHISM. The analogy afforded us a series of
working assumptions, not always clearly stated and not necessarily
consistent: first, that morphemes are, by definition, the minimum meaningful
units of a language; second, that any utterance in the language consists of
an arrangement of occurrences of its morphemes; and third, that the
meaning of the utterance as a whole is a function of the meanings of its
constituent morphemes and of their arrangement.
In addition, there was the general feeling that the correct morphological
description of a dialect, like the correct phonemic analysis, though perhaps
difficult to discover, is necessarily unique.
Any version of lexico-grammatical theory that incorporates all three of
these propositions leaves our fish story unexplained. The special impact of
the phrase carp-to-carp walleting is dependent on its phonetic resemblance
to the familiar expression wall-to-wall carpeting, which shares with it only
the morphemes to and -ing. It is thus obvious that some meaning-conveying
machinery other than morphemes and their arrangements must be at work.
None of us theoreticians of the 1940s and 1950s was so foolish as to ad­
here to the three propositions in the bald form in which they are set forth
above. They were hypotheses, not axioms, and our concern was to adjust the
theory, with minimal injury to its appealing simplicity, so as to fit the
observed facts more snugly. Nor did we necessarily allow our theoretical
66 6. WHY MORPHEMICS WON'T WORK

predilections of the moment to interfere with the ongoing task of describing

individual languages accurately. Some descriptions of the time cast the data
slavishly into a mold that might or might not be appropriate, but others set
aside theoretical preferences whenever they got in the way of realistic and
efficient presentation. Even so, the essential notion of an elementary, not
further divisible, unit, an "atomic morpheme," suffused our thinking. And
despite seemingly radical theoretical innovations in several different direc­
tions between then and now, I think it still does.
All that will be dealt with in detail later (chapter 7). First, just so there
can be no residual doubt in the matter, we present a catalog of familiar
homely phenomena for which the atomic morpheme notion cannot provide
without embarrassingly extensive supplementation — a tail big enough to wag
the dog. At the same time, our examples will afford broad hints of the nature
of the replacement for morphemic theory to be proposed in chapter 7.

6.2. Allusions. In March of 1984 there was a sudden heavy snowfall in New
York City. A few hours later, describing the situation, a radio announcer
included in his weather report the comment If you an make it here you can
make ft anywhere. Part of the impact of this was its direct application to the
current state of affairs. For many, another part was via the allusion to the
song "New York, New York" (by Fred Ebb and Joseph Kander, made famous by
Frank Sinatra), which contains the words "if I can make it there, I'd make it
anywhere." The further implications would vary from listener to listener, but
by and large people associate the song with a happy, upbeat mood, so that
the announcer's allusion served to soften somewhat the severity of the
situation.
In the 20th paragraph of chapter 15 of Agatha Christie's detective story
Murder is Easy the reader encounters the following sentence:
Normal, "well-to-do, respected, last sort of man, and so on
The protagonist is going through a list of suspects, considering each in turn
as a possible murderer. To know this, the reader need only have perused the
few sentences that precede the quoted one. But with no more context than
that, the phrase last sort of man is likely to be puzzling.
This phrase, also, is an allusion. It differs from our first example in that if
you don't recognize the former as an allusion you can understand its "literal"
meaning anyway, whereas in this case I think understanding with only the
immediate context is unlikely. But if you have started from the beginning of
the novel, then as you reach this point you are prepared. In the very first
chapter, a woman who is later one of the victims is speaking of her sus­
picions of a man who may have already murdered several people, and she
says
... the person is question is just the last person anyone would suspect
6. WHY MORPHEMICS WON'T WORK 67

Then, just a couple of pages before the passage we are discussing, consid­
ering another name on his list of candidates, our protagonist has said
The last person you'd ever think would be a murderer!
With that background, and with the knowledge that all the words in this
chapter are the hero speaking to himself, the reader not only understands
the phrase but finds it apposite.
This phenomenon of the CONTEXTUAL allusion is widespread in extended
written discourses. When it is suspected of being unintentional, someone pulls
out all the repetitions and sends them to the New Yorker for publication as a
column-filler under the heading "Fascination-with-Sound-of-own-Words Depart­
ment." When it seems to have been planned, as in the work of Thomas Mann,
critics speak in glowing terms of the artistic use of leitmotifs. In either case,
I think we can summarize the effect by saying DISCOURSE DESIGNS ITS DIALECT.
6.3. Idiomatic Phrases and Tongue-Twisters. By virtue of its alluding, if in
no other way, an allusion is idiomatic. But I think it can be shown that most
phrasal idioms involve allusion, though not always in the same way.
In the case of tf you can make it here you ccm make if anywhere the
primary impact was the literal interpretation of the words as they applied to
the weather; the allusion to the popular song was secondary. In the case of
kick the bucket almost the opposite is true. The primary impact, the
immediate application to the state of affairs being spoken of, is via the
idiomatic sense 'die'; the literal interpretation of the words is of subordinate
importance, and constitutes an allusion, however vague.
The latter is also the case for a tongue-twister, such as Peter Piper picked
a peck of pickled peppers or Rubber baby-buggy bumpers or (preferably in an
rless dialect) She saw Esau on the seashore. Such an expression is uttered
not because of the meanings of its constituent morphemes and of their
arrangement, but because it is tricky to say. True, each utterance consists of
ordinary English words put together in a routine fashion, and therefore has a
"literal" meaning. But that meaning is purely incidental, and therefore only
connotative or allusive.
However, the original source of an allusion may be forgotten. How many
people know the original reference of the expression in the limelight? Is it to
a citrus fruit or to oxide of calcium? What are the pad of padlock, the bon of
bonfire, the criss of crisscross, the sham and the rock of shamrock? What is
the source of to have too much of a good thing? What is the figure in the die
is cast? Does it refer to the tossing of a gambling cube, or to the making of
a mold for mass production? In any of these cases, if you think you know,
then whether you are historically right or not the form is an allusion for you.
If you don't know, then the phrase is a "floating" idiom, which we might say
bears the same relation to ordinary or "anchored" idioms that, in the parlance
68 6. WHY MORPHEMICS WON'T WORK

of literary critics, a faded or dead metaphor bears to live ones.

We should also note that this is a respect in which idiolects and family
dialects vary enormously. If I speak of everything being bock in routine, or
say, in response to a bit of mildly good news, That makes it nice, virtually
any user of English will get the gist, but for only a few of my close kin is the
phrase an allusion, and even for them the original context may be forgotten
so that only its emotional flavor is still elicited. Here, then, is one more re­
spect in which, were it not for redundancy (§5.5), language could not perform
its social function.

6.4. More Puns. For our next example we turn to Madison Avenue.
Kellogg makes a cereal called Crispix. Each individual piece consists of two
roughly hexagonal lattices, each bowled like a spoon rather than flat, stuck
together along the edges to form a sort of hollow lozenge. On the box,
directly under the name "CRISPIX", appears the slogan A crisp cross of corn
and rice. The phrase crisp cross here is a pun on crisscross. As in several of
our earlier examples, though part of the force of the words is their direct
application, "literally" interpreted, another part is indirect. For the literal
part, the cereal is certainly crisp and is certainly a cross between corn and
rice. For the indirect connection, the shape of the individual pieces renders
the term crisscross apt. Also, the letter "X" is a cross, so that "CRISPIX" can
be read as crispi ( = crispy) cross. To underscore this even more, the
inscription "CRISPIX" on the box is in large letters with broad strokes, and the
final "X" shows an ear of corn in the stroke that rises from lower right to
upper left intersecting a sheaf of rice in the other one. I am sure the
marketing experts who designed the name, the slogan, and the box were fully
aware of all this —in fact, that they worked as diligently as any poet to make
them all fit together.
I propose that, although not every allusion is a pun, every pun is an allusion.
To be sure, there is a difference between our carp-and-wallet example and the
present one, but it is not the difference between pun and nonpun. In the
former, the allusion is to a phrase which is totally irrelevant to the state of
affairs described. The story is therefore absurd, and if it is funny (as it is
intended to be) the humor derives from the absurdity. In the Crispix case the
allusion is to a semantically relevant phrase, whose meaning thus reinforces
that of what is actually said rather than conflicting with it.
Do not infer from these examples that the allusion in a would-be funny pun
is always absurd. I need only remind you that the Department of Classics at
Cornell is housed in the top floor of one of the oldest buildings on the
campus, and that, appropriately, they teach Attic Greek.
6.5. Onomatopoeia and Secondary Associations. Let us go back to the
cereal box and take a closer listen to the key word crisp. Does the word itself
sound crisp to you? I hesitate to call a substance "crisp" unless, when broken
6. WHY MORPHEMICS WON'T WORK 69

or ground up, especially by chewing, it produces a certain sort of sound, and

that kind of sound to my ears does somewhat resemble the sound of the
word itself. At any rate, the sound of biting into something crisp resembles
the sound of the word crisp much more than it does that of such a word as,
say, limp, bite, chew, slurp, chrysanthemum, or woodshed.
Be that as it may, the aptness of the word crisp for its meaning is
reinforced in another and, I think, much more important way: it belongs to a
family of words that begin with the same consonant cluster and whose
meanings all have to do directly or indirectly with sound; and it belongs to a
(much smaller) family of words that end as it does and have at least vaguely
similar meanings. For the former, consider
crisp, crack, crackle, cracker, crash, crap, crepitate, cricket, crimp, cringe, crimple, crinkle,
creak, creaky, croak, crov, croup, crunch, crumble, crumple, crunkle, crush, crust.
Some of these are semantically closer to crisp and some are farther away —
and, of course, there are also a good many words that begin the same way
but whose meaning is totally different (e.g., crown, crazy, cream). As for
words that end with the same cluster, although there are only about a dozen
in our dictionaries at least three of them seem appropriate to cite here:
(crisp), lisp, gasp, rasp
The relevance of these resemblances in sound paralleling resemblances in
meaning, however vague the latter may be, is that the occurrence of any one
word of such a set may have part of its effect indirectly, by reminding the
hearer of some of the other words of the set. To the extent to which this is
true, the mechanism of the impact on the hearer is like the mechanism by
which an allusion does its work.
The vast majority of cases of so-called onomatopoeia, and also most of
the forms whose sound, whether or not thought of as onomatopoetic by the
users of the language, is felt by them as particularly appropriate to their
meanings, are probably just instances of such paralleling resemblances. This
is not to deny the reality of onomatopoeia, which we have already affirmed
(§1.7), but just to insist that true instances of it are rarer and fainter than is
often thought.
Thirty years ago I referred to such indirect effects as due to the
"secondary associations" of linguistic forms. An example I used then was the
word vanish, part of whose force can be because of its resemblance in shape
and meaning to such words as banish and vanquish There are endless
numbers of such resemblant sets, some with many members and some with
very few. Since similarities in sound and similarities in meaning both range by
imperceptible steps from virtual identity to extreme vagueness, there are no
sharp boundaries marking one set off from another. For example, it would not
be utterly ridiculous to enlarge the -sp set given above by adding brisk and
70 6. WHY MORPHEMICS WON'T WORK

perhaps some other words with final -sk— even though these also form their
own family.
In 1933 Bloomfield listed a number of resemblant sets, either rhyming or
with identical initial consonantisms; for example,
flash, flare, flame, flicker, flimmer,
said to share reference to 'moving light'. I here supply some sets with differ­
ent sorts of similarity in sound, but without venturing to define their common
semantic features:
grin, groan;
squirt, spurt; squeak, speak,
stolid, solid;
skip, skimp, skim;
flip, flit, flick;
damp, dank, dark;
brink, brim, rim, ring, rink;
crumble, crumple; bathos, pathos; bother, pother; stable, staple; snigger, snicker; hurdle,
hurtle; doddering, tottering; girdle, kirtle;
somber, sober;
smelt, melt, meld;
whorl, whirl, vhirr, twirl, hurl(?), furl, curl, coil, curve, circle, cycle.
— but crumble and crumple belong also to the cr- set mentioned earlier, and
twirl belongs also to the following:
twirl, twist, twine, turn, tourniquet, torque, torsion, tort, torture
If we were confronted by a small amount of such material, our inclination
might be to try to break the words down into morphemes: say, kr- (spelled
cr-) 'small sound' and the like. But there are great difficulties.
First, which of our English words beginning phonetically with the consonant
cluster kr- contain this morpheme and which ones don't? The decision has to
be based on meaning, and, as already pointed out, the meanings shade into
one another so gradually that any choice seems arbitrary.
Second, granted that some of the words include at least one more-or-less
definable piece that recurs elsewhere, many of the pieces left over when such
recurrent ones are pulled off are what we are going to call CRANS — unique
constituents, like the cran- of cranberry. Where else, for example, does one
encounter the -ist of twist or the -ine of twine?
Also, it seems not only difficult but rather unrealistic to extract a
recurrent partial from solid and stolid, or circle and cycle, or vanish and
vanquish, or oust and out. This may seem like a separate difficulty, distinct
from the second, but in fact it is not. Were there a whole string of pairs of
English words related in sound and (at least vaguely) in meaning as are solid
6. WHY MORPHEMICS WON'T WORK 71

and stolid, we should not hesitate to recognize a potentially discontinuous

element s-olid and an infix <t>. Phenomena of this sort are abundant in some
languages. But in English, except for such verb sets as sing : sang, ring :
rang, sit : sat, pairs of the sort cited above are largely isolated, so that
analyzing the words requires recognizing more and more crans.
I conclude, therefore, that the fact of importance here is not whether a
similarity in shape rests on an obvious shared partial or on something else,
but the resemblance itself. ANY sort of similarity in sound between forms,
provided it accompanies some resemblance in meaning, may serve to remind
a listener of one form when the other is heard, and it is that sort of pho­
netic-semantic reverberation or "resonance" that concerns us.
6.6. Blends and Lapses. Resemblances somewhat like those just illustrated
are created or modified by every innovating blend.
One of the teachers of lower brass instruments in the Ithaca College School
of Music held a special autumn institute announced as n Octubafest. There is
no point in this new word unless the audience hears in it both October and
tuba, as well as noting that the two are being uttered simultaneously, as it
were, instead of in sequence. Likewise, Reaganomics is pointless unless one
hears in it both Reagan and economics.
There is always something special (in English, at any rate) about a new
blend. By appearing to be a slip of the tongue without actually being one, it
forces the listener to do a quick double take, and it is often devised and used
for that very reason. The impact depends on the setting. Most frequently it is
at least mildly humorous. Sometimes it is also absurd, as when a theatrical
program for children thoughtlessly dubbed itself "YOUTHEATER". Occasionally
it is a sharp pain in the gut, as with the sudden occurrence, in Alison Lurie's
novel Foreign Affairs, of the horrible commercial "euphemism" cremains. With
recurrence the distinctive character is apt to fade. Thus I doubt if the resi­
dents of Toronto any longer pay special attention to the street-name
Hurontario Likewise, smog and motel have tast any marked prominence as
blends, to become, in most of their occurrences, just ordinary colorless
words; it takes some unusual circumstance to remind us that smog is smoke
and fog, or that a motel is a motor hotel.
When a blend reminds the listener of the forms that have been blended, the
phenomenon is at bottom the same as that of an allusion reminding the
listener of the word or phrase alluded to. So a blend, like a pun, is a kind of
allusion.
Sometimes a blend not only appears to be, but actually is, a slip of the
tongue. Conversely, many or most slips of the tongue are the result of
indecisiveness on the part of the speaker as between two alternative words or
phrases and the resultant production of something that has ingredients from
each; so most slips of the tongue are blends. Long ago I cited the example of
72 6. WHY MORPHBMICS WON'T WORK

my own Don't shell so loud! in which shell was the product of indecision
between shout and yell. For the speaker, an accidental blend is different from
one thought out in advance and uttered on purpose. But for a listener they
are in the first instance all alike, and thus all allusions, the only difference
being that after producing a lapse the speaker may give signs of not having
meant to say what was actually said,
6.7. Malapropisms, Doublets, Contamination, and Taboo. A speaker wish-
ing to utter a form not learned very well may confuse it with a more familiar
form and come out with the "wrong" one. If the motive is elegance, then the
actually uttered form is a MALAPROPISM —an unintentional and absurd pun, A
girl of thirteen proudly announced to a friend, I'm administering now. Other
examples are I'm simply ravishing (for ravenous); the windshield factor (for
wind-chill); planter's wart (for plantar wart).
Some of these are hardly more than idiosyncratic deviations that may es­
cape detection and "correction" for years. One person I know discovered only
in her thirties that what she had picked up in childhood as next-store neigh-
bor is for everyone else next-door neighbor. Another was surprised, at about
the same age, to learn the common current equivalent of his for all intensive
purposes.
At the other extreme, a replacing form may come into regular and wide­
spread use. Then the only difference between a pair of which one is "right"
and one a "mistake" and a pair both of which are acceptable is the socio-
linguistic attitude of those involved. For many of us older Americans, the
expression / couldn't care less, meaning 1 don't care at all', is "right" and the
now very common alternative I could care less, intended to mean the same
thing, is downright ridiculous. But there are users of English whose attitudes
toward these two are exactly the reverse, and probably many who will accept
either with perfect calm—or who, perhaps, do not ordinarily even register the
difference.
All of the foregoing pairs could be classed as doublets, though somewhat
different in both origin and status from those discussed in earler sections
(§§4,8, 5,4), There are a good many idiomatic phrasal doublets like could(n't)
care less. a new lease [or leash] on life, that's the way the cookie crumbles
[or crumples]) hés the spittin' image [or spit and image] of his father. Also
conceivably classifiable as doublets, even though at a much larger size-level,
are the endless cases of slightly differing versions of jokes, nursery rhymes,
or folk tales. In all of these, if. you know one of a pair and hear the other, it
reminds you of the one you know, and the consequences (amusement,
embarrassment, whatever) follow from that. So here, again, we see at work
the mechanism of allusion.
When a less commonly used form is influenced by a better known one, the
effect need not be, as in ravishing for ravenous, a total replacement; instead,
6. WHY MORPHEMICS WON'T WORK 73

the shape may simply be altered in the direction of that of the better known
one. The original pronunciation of granary was GRANnery, the very common
pronunciation GRAINery shows the influence of grain. Such a partial change is
what is usually called CONTAMINATION, but its kinship with the other phe­
nomena under discussion is obvious. Another case is eye-dee-OLogy, more like
idea than the etymologically more "correct" id-dee-OLngy. Milch, as in milch
cow, is for most people so unusual that, despite the spelling, they are likely
to pronounce it like milk. English-using sports fishermen in the north woods
sometimes say they are out to catch Mister Narneys (the spelling varies):
this is a borrowing of Cree misti-narnēs 'big fish, big trout', contaminated by
English mister. The English fish-name grouper is reshaped from Portuguese
garoupa, the Jerusalem of Jerusalem artichoke is from Italian girasole
'sunflower'. The well known classic examples, French loans female, mushroom,
and crawfish or crayfish, belong here too.
Contamination is sometimes called ATTRACTION. In contrast, taboo words
sometimes manifest an effect that could be called REPULSION.
The weakest manifestation of repulsion is the replacement of a taboo word
by a shape similar enough to suggest it but different enough to be acceptable
in situations where the taboo form is not. Thus, from a socially more delicate
period in the history of modern English we inherit Jeez for Jesus, Jeepers
{Creepers) for Jesus (Christ), golly for God, darn or dern for damn(ed), heck
for hell. Such seemingly mild expletives as shucks and fudge may belong here
too. All of these are perhaps recrystalizations (§5.4) of intentionally muffled
pronunciations of the taboo forms, intentionally interpreted as slurvian.
More interesting is the repulsive influence of a taboo form on the shapes
of other words: an ordinary word whose shape is too much like that of the
taboo form (so that it might be misheard as the latter, or so that hearing it
would remind people of the latter) may be avoided or replaced — unless the
speaker is seeking to evoke just those associations. Bloomfield cites rooster
as a euphemism replacing cock and donkey in place of ass. By the ordinary
patterns for forming abbreviations in English, condo for condominium is only
second choice, but it is easy to understand why the most likely candidate has
always been avoided. Bloomfield mentions the obsolescence of coney 'rabbit',
variously replaced by rabbit, cottontail, or bunny, in New York City the place
name Coney Island shows a survival of the otherwise obsolete term;
Honolulu's Rabbit Island was named more recently. Given continuing male
dominance in our social order, one wonders if there is not also some veiled
line of association between the now obsolete word and Playboy bunny.
Repulsion, as much as attraction, is clearly a product of associative reso­
nance.
6.8. Metanalysis and Abbreviation. Knowing that Limburg, Braunschweig,
Hamburg, Frankfurt, and Wien are names of cities, and that limburger,
braunschweiger, hamburger, frankfurter, and mener are kinds of foods, one
74 6. WHY MORPHEMICS WON'T WORK

can conclude that hamburger is hamburg- + -er. Or, knowing cheeseburger,

beefburger, fishburger, lambburger, Tullyburger, and so on (including the
independent form burger), one can decide, rather, that hamburger is ham- +
-burger. The first part of the latter is a bit bothersome, true enough, since it
can hardly be the morpheme that refers to cured leg of pork; in fact, I think
ham (in the latter sense) is the one ingredient no burger ever contains. But
the -burger part seems incontestable.
How is one to choose between these two analyses? Historically the matter
is simple; the first alternative is the etymology, the second the result of a
subsequent metanalysis. But those historical facts can be of relevance only
for the rare individuals who know them, and even then are of at most con­
notative value (we will discuss this more later on). Descriptively, under the
morphemic hypothesis, a choice must be made, since the single word cannot
be both hamburg- + -er and ham- + -burger at the same time. Yet both
proposals are based on sound evidence.
Abbreviations present the morphemic theory with two sorts of difficulties. In
the first place, although an abbreviated form, by definition, is shorter than the
unabbreviated original, it usually strikes us not as morphologically simpler
than that original, but as more complex. Are we to recognize a morpheme
that consists not of material but of the dropping of material? That was
actually tried in the 1940s, but seemed like rather a great strain to put on
the morpheme notion. It is akin to the old school grammarian's trick of
supplying an unspoken but "understood" subject you for an imperative verb,
and to the whole machinery of "deletion" exploited in transformational-
generative theory. Working at language from the speaker's point of view, it
may be legitimate to imagine things thought but not spoken aloud. But in this
essay our point of view is that of the hearer, who has to figure out what is
meant by what is actually heard; in that frame, the notion of deletion seems
out of place.
Also, abbreviations can give rise to a problem somewhat similar to that
presented by hamburger. Indeed, we could take burger to be an abbreviation of
hamburger and then interpret cheeseburger and the like as having this
abbreviation as second constituent — though that would render the morphemic
analysis no easier. The term condominium surely begins with a prefix con-, as
in concentration, condemnation, condescension, confirmation, consecration,
and so on, and the second part bears significant resemblance to dominion,
domain, dominate, dominance. But the abbreviation condo (§6.7) is widely
used. That immediately seems to imply, under the morphemic theory, that the
word is condo- + -minium. Similarly, because of the abbreviation lab,
laboratory seems to be both /abor-atory and lab-oratory, microphone is both
micro-phone and mike-rophone;and so on. In British English, because of the
slang term preggie 'pregnant woman', pregnant would seem to be preg-nant as
well as— for a scholarly minority— pre-gnant. On top of all other difficulties,
6. WHY MORPHEMICS WON'T WORK 75

this sort of analytic procedure generates a remarkable surfeit of crans:

-rninium, -rophone, an -oratory that has nothing to do with speechmaking, a
completely kosher ham-, and thousands of others.
In all these cases, instead of choosing one analysis and rejecting the other,
the course of wisdom is obviously to reject the approach that demands the
choice. It is the resemblances that count, not what yields them. Take the first
example. The similarities that turn on the recurrent partiate hamburg- and -er
are genuine; so are those that turn on the recurrent, partial -burger. All of
them are potentially functional for a user of the language. That they involve
incompatible recurrent partiate is true, but why is that of any relevance?
Does it bother us because we think it bothers the user, or because it offends
our professional sense of morphological neatness?
6.9. Naming. If the morphemic view of morphology were true, linguistically
unsophisticated users of a language, consciously devising new words, would
tend to do so by sticking together meaningful parts, many of which would be
obviously minimal.
Something close to this actually happened in the century-tang evolution of
the international nomenclature for chemistry. By early in the present century
such a typical designation as
7-(1,2-dimethylpentyl)-5-ethyltridecane
was the answer to a morphemicist's prayer. Not only is it almost entirely
AGGLUTINATIVE (its meaningful parts assembled in sequence with virtually no
change of shape where they touch) but also the hierarchical structure, of
successive layers of immediate constituents, is marked either overtly (the
parentheses) or by the inherent nature of the ultimate constituents. Those
ultimate constituents are
7,1, 2, di, meth, yl, pent, 5, eth, tri, dec, ane
A name in this system is as explicit as the structural formula of the sub-
stance. Given either, the chemist can write down the other.
But the requirements of chemistry are special. Precision of denotation is
crucial, and connotations hardly count. It does not even matter that the whole
system is set forth in terms of writing rather than speech and that many of
the terms would be ambiguous or unintelligible in oral form. In other naming
situations other factors, especially connotations, tend to be more important,
and the neat row-of-bricks style characteristic of chemical terminology loses
its appeal.
We pass over names in fiction, often routine yet sometimes, especially in
children's stories, quite imaginative. But Madison Avenue needs more atten­
tion than given it in our brief discussion of Crfspix (§6.4).
The designers of proprietary labels try to make them suggestive and
memorable, yet unique enough to be registerable as trade marks. Some
76 6. WHY MORPHEMICS WON'T WORK

products, for example most foods, are so well known by everyday terms that
it would be risky to replace them. Pork is pork; all you can do is supply a
distinctive modifier, as the Silver Plotter pork sold by the P&C chain of
supermarkets in upstate New York, This is true even of some breakfast
cereals; e,g., Kellogg's Honey Smacks, two ordinary words with nothing
unusual except their juxtaposition. The only mark of proprietorship can be a
respelling, as with Kellogg's Rice Krispies, Ralston's Wheat Chex, and General
Mills' Kix. Or there can also be distortion: Quaker's Quisp sounds crisp but
also evokes memories of Porky Pig, General Mills' Cheerios, as well as
incorporating cheerio, cheery, and cheer, can be read as cheery ds, a graphic
pun on the shape of the individual pieces of the cereal. The earliest names
for our patented cereal products, about a century ago, were apparently fairly
sober, but in recent decades they have become more and more wild and have
been designed more and more explicitly to appeal to children, the chief con­
sumers. General Mills for a time put out Count Chocula, Frankenberry, and
BooBerry, all containing some allusion to ingredients or flavors as well as to
fictional figures; now there are names that apparently involve not even any
indirect reference of the former sort, like General Mills' Kaboom and
Ralston's Gremlins, Rainbow Brite, and Cabbage Patch Kids.
In other products carefully chosen ordinary words seem often to turn the
trick the advertisers want to turn: e,g,, the skin lotion Intensive Care
(Chesebrough-Pond's), the hair dye Lnving Care (Clairol), the anti-dandruff
shampoo Head and Shoulders (Procter and Gamble), The puns or punlike
allusions in others hardly need exegesis: Kleenex tissues (Kimberley-Clark),
Band-Aid bandages (Johnson and Johnson), Fluorigard tooth-hardening rinse
(Colgate), Turns antacid tablets (Norcliff-Thayer), Porcelana spot-removing
skin lotion (Jeffrey Martin), Sominex to induce sleep (J, R. Williams), Anusol
suppositories (Parke-Davis), To me the allusion in the antihistamine Contac
(Menley and James Laboratory) is obscure, but probably I am missing
something.
For prescription drugs name-inventors face another problem: proprietary
names should sound in a general way like, without actually being, official
generic terms in the pharmacopia (themselves close kin to the international
nomenclature of chemistry). The designers are often successful: I can never
remember whether inderai is the proprietary and propanoic} the generic name
of a hypertension medication I take or the other way around (I have just
checked, and the first way is right: inderai is Ayerst's trade name). In the
case of zyloprim (Burroughs-Wellcome) and allopurinol I would be equally
baffled except that I know the second is the generic because it alludes to
purines and purines cause gout and that is what I take it to prevent. It does
not matter that that line of association is personal, since ALL such
associations are personal, and all that varies is their strength and how widely
they are shared.
 7. FROM PARTICLE TO RESONANCE
7.1. The Road to Morphemics: Bopp. The morphemic way of thinking about
language is going to be hard to outgrow, for it is firmly rooted in the history
of our discipline. We shall be better able to deal with both its merits and its
defects if we are aware of that history.
One of the minor but persistent themes of nineteenth-century scientific
thought was what I have elsewhere dubbed GRANULARITY: the tendency to con­
ceive of everything as built up from small constituent particles, along with
some inclination to ascribe the properties of a thing to its constituents
rather than (as in the gestaltist view) to the way they are arranged. Dalton's
material atoms, an experimentally based revival of a notion that had been
purely speculative with the ancients, led the way in 1803. Thereafter came
molecules (Avogadro, 1811); cells in btology (Virchow and others, 1840s); in
mathematics, the reduction of numbers of all sorts to the positive integers
(Dedekind and others, 1870s); ions (Arrhenius, 1884); and belatedly, in phys­
ics, quanta (Planck, 1900). In psychology the theme manifested itself as "ele-
mentalism": both Herbart, early in the century, and Wundt, at its close, sought
to dissect complex mental phenomena into independent basic factors, though
there was little agreement as to the identity of the latter. In evolution and
heredity, as soon as the adjective "genetic" had begun (with Darwin) to take
on the meaning it now has, "genetic trait" was used to denote a presumably
minimum unit of inheritance, and as it subsequently became clear that many
features of human structure and behavior are transmitted not genetically but
by learning and teaching, "culture trait" made its gradual and unobtrusive
appearance. To be sure, many of these sorts of constituents — even Dalton's
atoms — turned out, in due time, not to be ultimate, but at least they were all
relatively small and simple as compared with the wholes of which they were
the parts.
The contrasting emphasis on pattern (gestalt), which is playing such a
large role in the present essay, began in earnest only in the 1860s, though
doubtless with many forerunners. Like granularity, it appeared first in chem­
istry (Kekulé), only later in psychology (1890, von Ehrenfels).
Long before Dalton, students of language, like all literate Europeans, had
had a kind of ready-made two-level granularity thrust upon them in the form
of words and letters. These could be, and were, interpreted validly as ele­
ments of speech as well as of writing: Rask and Grimm were perfectly aware
of the difference between visual shape and sound, and although their termi­
nology was not maximally convenient their colleagues were adept at discern­
ing when they were referring to the one and when to the other.
But it is only the appearance or the sound of a word that is built from
78 7. FROM PARTICLE TO RESONANCE

letters (or phonemes), not its meaning. If we heed meaning, the situation is
quite different.
Many words, thus, seem to consist of single meaningful parts: cat, dog,
wood, shed. Some are built by gluing such elementary words together: wood-
shed,blackboard (this is the agglutinative model followed throughout the long,
slow evolution of chemical nomenclature: §6.9). Some have meaningful parts
that resemble separate words with which they were formerly identical: the -ly
of suddenly is from Old English -lic, which as an independent word comes
down to us as like, Old English dōm gave rise, with divergence of sound and of
meaning, to both doom and the -dom of kingdom; the -ful of careful is altered
from an earlier full, which still survives also as a separate word; a single
earlier form gives us nose, the nos- of nostril, and the nozz- and nuzz- of
nozzle and nuzzle, even the past-tense marker -d, as in loved or hated, looks
like it might at some earlier stage have been the independent word did. And,
finally, some words which seem to be semantically composite are related to
semantically simpler ones not by added material but by an internal change:
men is in meaning comparable to boys, but it is not related in shape to man
as boys is to boy, similarly sang and sing, or the noun shelf versus the verb
shelve
Faced with evidence of this sort, thinking in historical rather than syn­
chronic descriptive terms, and working in a scholarly climate characterized
by granularity, Franz Bopp suggested that although languages today cannot
be dealt with entirely in terms of granular ultimate constituents, there was a
time when they could. The available historical evidence shows that many
meaningful internal changes are the residual influence of earlier suffixes now
lost. For example, while an earlier (pre-English) *mann has become modern
English man, the ancestral plural *manniz, which shows a suffix but no
internal change of shape, became first *menniz, then *menn, and finally men.
The inference is that, although in most cases the details are not recoverable,
ALL internal changes have arisen in this way. The surviving evidence also
enables us, as shown in the preceding paragraph, to trace many current
affixes back to earlier independent words; it is concluded from this that ALL
bound forms have such origin. It follows that at the very beginning, in the
"original" language or languages of our species, there could have been no
internal changes and no bound forms: the earliest human speech was not only
agglutinative, with each meaningful element retaining the same shape in all
environments, but also ISOLATING, with each such element a separate word.
Quite apart from the questionable meaningfulness, not yet recognized in
Bopp's era, of speaking of the "original" language, his argument strikes us
today as a ridiculous over generalization. But many of his contemporaries and
immediate successors found it persuasive. Whitney, for example, a model of
scholarly sobriety, insisted that all sound historical method requires accep­
tance of the doctrine of uniformitarianism (originally formulated by Hutton.
and Lyell in geology), whereby one projects into the past, by way of expla­
nation, only those processes and mechanisms that can be observed at work
7. FROM PARTICLE TO RESONANCE 79

in the present; and he felt that such method, applied to the linguistic
evidence, leads unavoidably to Bopp's conclusion.
7.2. The Road to Morphemic: Morphophonemics. As early as the 1870s
some scholars, notably Baudouin de Courtenay and Kruszewski, were making
serious efforts to deal with the synchronic facts of languages without refer­
ence to their known or suspected antecedents.
Such synchronic facts include many cases in which what in some sense
seems to be the "same" element does not always appear in the same shape,
but bears different guises in different environmentś. This phenomenon is
ALTERNATION. Even the independent English word like and the suffix  might
be regarded as an instance — were it not that the former shape now also
occurs as a suffix, yielding such direct contrasts as womanly versus
womanlike. Since they are in direct contrast, they must now be distinct
elements, despite the historical affinity. On the other hand, it is customary to
think of noun pluralization as a single thing in English, and it is a fact that
although the plural suffix has more than one shape (e.g., -s in cats but -z— so
pronounced — in dogs), the different shapes are never in contrast; so this is a
genuine case.
Alternations are of various kinds, and Baudouin worked out a classification.
With the subsequent advancement of phonemic theory, some of his types
turned out to be matters of subphonemic variation, and others were rejected
when it was shown that the shapes are in direct contrast, like like and 
But the residue after these adjustments is still large.
Many an alternation is easily explained in terms of history. For example, the
divergence betwen the pronounced final -z of dogs and the final -s of cats
would be understandable if it could be shown, or guessed, that at an earlier
stage z was pronounced in both words, but that subsequently the z of cats
became voiceless because of the voiceless t directly before it (and some­
thing much like this seems actually to have happened).
But if we do not know the history and do not want to guess at it, or if we
want to ignore what we know of it in order to achieve a purely synchronic
treatment, what then?
What happened was that MORPHO-PHONEMICS was invented as the synchronic
surrogate of history. The development took a while; the term MORPHOPHO-
NEME was devised only in 1927 (by Ułaszyn), and it was a decade or more
after that before the hyphen was dropped.
The original nature and purpose of morphophonemics, and ìts similarity to
and difference from history, are manifest in two of the earliest American
papers on the topic, both from 1939.
Swadesh and Voegelin class alternations as PATENT and NONPATENT, meaning
essentially what many of us soon thereafter were calling AUTOMATIC and
NONAUTOMATIC. That of the English noun-plural suffix between -z and -s is
patent (automatic), because it depends entirely on the phonetic nature of
what precedes. That between f in the singular wife and ν in the plural wives is
80 7. FROM PARTICLE TO RESONANCE

nonpatent (nonautomatic), because it depends on no such environmental

feature (witness invariant f in cliff : cliffs, and invariant v in grave : graves),
but is simply the way certain nouns behave. One might wish to signal this
nonpatent alternation by writing, say, a small cap F in one's notation for
those nouns, leaving the letters / and v free to indicate the invariant ones.
For this particular alternation in English it is perhaps more efficient just to
list the nouns themselves, but for comparable alternations in Tübatulabal, the
language from which they draw their extended example, special symbols are a
great help. The authors then make the following observation:
... alternations are the result of phonetic history, affected also by foreign borrowings and
analogical changes. The most efficient formulation of the synchronic facts is ordinarily not
the same as a reconstruction of the actual historical developments, but the process of
constructing morpho-phonemic formulae has some resemblance to that of historical-
phonological reconstruction.
Bloomfield, writing of Menominee, spells this out in greater detail:
The process of description leads us to set up each morphological element in a theoretical
basic form, and then to state the deviations from this basic form vhich appear when the
element is combined with other elements. If one starts frith the basic forms and applies
our statements ... in the order in vhich we give them, one will arrive finally at the forms
of words as they are actually spoken. Our basic forms are not ancient forms, say of the
Proto-Algonquian parent language, and our statements of internal sandhi are not his­
torical but descriptive, and appear in a purely descriptive order. However, our basic forms
do bear some resemblance to those which would be set up for a description of Proto-
Algonquian, some of our statements of alternation ... resemble those which would appear
in a description of Proto-Algonquian, and the rest ..., as to content and order, ap­
proximate the historical development from Proto-Algonquian to present-day Menomini
Before this paper Bloomfield had not been in the habit of using the terms
"morphophonemics" and "morphophoneme," and even here he overtly defines
the former word as equivalent to "internal sandhi." But "theoretical basic
form" and "alternation" had been in his technical vocabulary for a long time.
In 1933, for example, noting that Samoan words end only in vowels, he
proposed to provide for the forms tanisia Svept', inurnia 'drank', and ulufìa
'entered', which show the suffix -ia, and the forms tani, in, and l, with no
suffix, by setting up theoretical underlying bases tanis-, in-, and uluf-, and
letting the final consonant drop when it would be unpronounceable.
And others had been using morphophonemic notation of one sort or another
in their descriptions of specific languages. I give an example from my own
analysis of Potawatomi, because it will serve especially well to illustrate a
point to be made later.
So far in this essay we have been able to make do with ordinary orthog­
raphy, with occasional respelling to indicate a pronunciation. But at this point
we must resort to transcription. In what follows I enclose phonemic symbols
between slant lines, morphophonemic notation between vertical strokes.
7. FROM PARTICLE TO RESONANCE 81

Typical of alternation in Potawatomi is the difference between the bare

noun 'paper', and the allocated form 'my
paper'. The latter not only adds a prefix / n - / but also switches the vowels
around to different positions relative to the consonants, which looks pretty
messv. But suppose we write, morphophonemically, and
and formulate a rule for the deletion of some of these
vowels in actual speech: the vowel before a final consonant stays; otherwise,
starting from the beginning of the word, the odd-numbered vowels drop and
the even ones stay. (The total pattern is somewhat more complicated than
that, but never mind.) Now we can chop the whole words, in morphophonemic
transcription, into recurrent meaningful pieces of INVARIANT shape: a prefix
| and a noun-forming suffix
You will not be surprised to learn that this synchronic treatment parallels
history. In Fox, which mostly retains the vowels of Proto Algonquian, our two
words are /mesenahikani/ and /nemesenahikani/. So what history has done,
description can undo: through the magic of morphophonemics, embarrassing
alternations have been not only provided for but, in a sense, suppressed!
7.3. The Morpheme in America. The 1930s and early 1940s in American
linguistics were the Decade of the Phoneme; the 1939 papers of Swadesh and
Voegelin and of Bloomfield foreshadowed the Decade of the Morpheme. Almost
everyone took the existence of morphemes for granted. The problem was not
whether they exist, but to find them, to identify them, and to discover how
they work.
In the light of difficulties of the sorts set forth in the preceding chapter, as
well known in the Decade of the Morpheme as they are now, one can wonder
how faith in the atomic morpheme, even though never universal, could have
been so widespread. But it was bolstered by a number of factors.
One was the supportive parallel between our technical morphemic notion and
the lay attitude toward language, to which we were all heir merely as literate
participants in Western culture. The average nonspecialist, if pressed to voice
an opinion in the matter, would probably say that an utterance — or, better, a
"sentence," that being the laity's prestige term — consists of words, and that
the meaning of the sentence is a function of the meanings of the words and
of their arrangement. We specialists were sophisticated enough to know that
words are not the minimum units, but the chief modification needed to render
the lay statement correct was just to replace "word" by "morpheme."
Alternatively, the lay view could be taken as correct but incomplete. Sen­
tences, true enough, consist of words, but that is only part of grammatical
design, the part known as syntax. Words, in turn, consists of morphemes, and
that is morphology, the other part. The implied parallel between syntax and
morphology also promoted faith in the morphemic approach to the latter.
Within our professional tradition, even though Bopp's proposal had been
diachronic rather than synchronic, and even though it had been proved
untenable, the notion of agglutination, if not of isolation, as the ideal (§7.1)
82 7. FROM PARTICLE TO RESONANCE

apparently still lurked just below awareness in our thinking.

And finally, on top of all that, there was the factor described in §6.1:
Kuryłowicz's "isomorphism." Phonemics was turning out to be eminently
successful in phonology. A parallel approach to grammar-lexicon, with the
morpheme in place of the phoneme, bade fair to be equally enlightening.
In the effort to make morphemics work despite the difficulties, all sorts of
questions were asked, and various possible answers were explored. For ex­
ample, must a morpheme consist of phonemic material (certainly that was the
original notion), or can it have a zero shape, as in the plural sheep, or even a
negative one (§6.8), as in the plural phenomena / / , as compared
with the singular phenomenon / / ? Or can a morpheme consist not
of a shape but of a change of shape, as in men as over against man? But if a
morpheme consists of phonemic material, then how can we claim that two
different shapes, say the -z of dogs and the -s of cats, are the same
morpheme? Or does a morpheme have a phonemic shape at all?— is it, rather,
an elementary unit on another "level" whose presence at a particular position
in an utterance is SIGNALED by a particular arrangement of phonemic
material? Or, at the other extreme, do morphemes really have meaning? Are
they perhaps simply recurrent units larger the average than phonemes,
convenient to evoke in trying to account rigorously for what utterance-sized
arrangements of phonemes occur and what ones don't?
7.4. The Great Agglutinative Fraud. In the speculation just outlined, mor-
phophonemic ideas and techniques played a central role.
Despite the parallels between morphophonemics and history, it is correct to
insist that morphophonemes have no historical reality. (The empirical justifi­
cation for this is the unarguable foundation of all synchronic descriptive lin­
guistics: very few users of any language know anything of its history, yet
they manifest the behavior we are trying to describe.) But if morphophonemes
have no historical reality, then what sort of reality DO they have?
Before answering that question let us see how it came to be asked.
In the historical development of our discipline the expression "morphopho-
nemic alternation" came first, "morphophoneme" only later. The older ex­
pression, even if shortened to "morphophonemics," can be interpreted as
referring to situations in which the morphemic structure is constant and the
phonemic structure is not. But when one gets used to the short form then a
different interpretation suggests itself. It may not make any more sense to
infer the existence of morphophonemes from the reality of morphophonemics
than it would to decide that since there are turkey drumsticks there must be
turkey drums, or to deduce from the existence of clever linguists that there
must be clever languages. Yet some such inference was implied by "morpho-
phoneme"—here was a noun, and it had to name something, real or fictitious.
But what?
In describing languages it is frequently useful to set up theoretical under­
lying forms. Since the notation for such a theoretical form was a morpho-
7. FROMPARTICLETO RESONANCE 83

phonemic formula, the individual symbols in it, whether or not also used in the
phonemic notation for the same language, came to be conceived of as one of
two things: either they were themselves morphophonemes, or else they were
symbols standing for morphophonemes.
That is how our question came to be asked, and there are the answers that
were proposed in the Decade of the Morpheme. We will refer to them respec­
tively as the CONVENIENCE theory and the REALITY theory.
The convenience answer, plainly implied both by Swadesh and Voegelin and
by Bloonnfield (§7.2), proposed that a morphophonemic symbol is merely a
descriptive device. If we decide to write a small cap F in our notation for the
basic shape of such English nouns as leaf, beef, knife, wife, to indicate the
alternation between phonemic / f / in the singular and / v / in the plural, then
the symbol is simply shorthand for a description of that alternation. So a
morphophoneme is nothing more than a symbol in a morphophonemic
notation.
But why "MERELY" a descriptive device? If we find one technique helpful and
others awkward in describing a language, that must reflect something about
the language, and surely it is the latter, rather than our convenience, which is
relevant. The discovery that one descriptive technique is easier than others
should thus be taken as a signal telling analysts to look more closely at their
data.
It was perhaps that reaction, in my opinion an entirely proper one, that led
some investigators to the second answer, the reality theory, which interprets
morphophonemes as units IN THE LANGUAGE. These units are something like
phonemes, but function at a deeper level. They are the smaller (phoneme-
sized) units of which morphemes are composed; they are represented in
actual speech by phonemes.
This can be made clearest by resorting again to phonemic and morpho­
phonemic notation, distinguished by the kinds of bracketing introduced in
§7.2. The reality theory of morphophonemics proposes that when users of
English utter / k l i f / , / k l i f s / , /wayf/, /wayvz/, /greyv/, /greyvz/ they are
also, internally and inaudibly, "uttering" |klif|, |klifz|, |wayF|, |wayFz|, |greyv|,
|greyvz|. When people speak, phonemes OCCUR—and so do morphophonemes.
Note that with these morphophonemic transcriptions we can now, just as in
the case of the Potawatomi example of §7.2, cite all the participating mor­
phemes in seemingly invariant shape: noun stems |klif-|, |wayF-|, |greyv-|, and
plural suffix |-z|.
In 1953 Floyd Lounsbury tried to tell us what we were doing with our clever
morphophonemic techniques. We were providing for alternations by devising
an "agglutinative analog" of the language and formulating rules that would
convert expressions in that analog into the shapes in which they are actually
uttered. Of course, even such an agglutinative analog, with its accompanying
conversion rules, could be interpreted merely as a descriptive device. But it
was not in general taken that way; instead, it was taken as a direct reflection
of reality. We seemed to be convinced that, whatever might superficially
84 7. FROM PARTICLE TO RESONANCE

appear to be the case, every language is "really" agglutinative.

The extreme of this tendency, for me its reductio ad absurdum, was a
proposed way of handling the English plural men. To recognize this as the
plural of man is, of course, correct: that's what users take it to be when they
hear it. But the proposal went far beyond that: in order to make it conform to
the agglutinative ideal we said that it is really a sequence of two morphs, the
first one, / m e n - / , being an allomorph of the morpheme {man}, while the
second, of phonemic shape zero, is an allomorph of the noun-plural
morpheme.
Not only was Lounsbury's diagnosis correct, but also what we were doing
verged on nonsense and occasionally, as in this analysis of men, crossed the
line.
Indeed, the problems —and pseudo problems —to which the reality theory of
morphophonemics has given rise are almost beyond belief. On the one hand,
the introduction of morphophonemes produced a terrible confusion between
phonemics and morphophonemics, and of both with phonetics, which has
lasted from Baudouin's times to our own and shows no sign of dwindling. On
the other, the promotion of "theoretical underlying forms" to the status of
hidden realities was carried over into syntax in the earliest transforma­
tional-generative theory, diverting attention from problems of real substance,
concealing but not eliminating the unstated assumption of underlying
agglutination, and dispatching researchers on a seemingly interminable
snipe-hunt. Even today, many of the issues to which linguists in the
Chomskyan tradition devote their attention are, it seems to me, no more than
artifacts of their frame of reference. (But let us remember that there is
nothing new about that. The Decade of the Phoneme was characterized by
irrelevant questions about phonology, and the Decade of the Morpheme by
equally impertinent morphological queries. Progress in linguistics seems to be
made mainly by painfully discovering which issues are real, which ones
spurious.)
7.5. The Transducer Fallacy. Moreover, the reality theory reinforced, and
was reinforced by, a phraseology of LEVELS which has now been molding our
discussions of language design for half a century or more.
Different levels seem to present us with different basic units: for example,
morphemes were proposed as being, on the morphemic or grammatical level,
what phonemes are on the phonemic. It follows that additional units suggest
additional levels. So if we are going to speak of morphophonemes it seems
there must be a morphophonemic level, and allophones or phones imply an
allophonic, phonic, or phonetic level.
Also, in this conception there must be conversion principles of some sort
to move us from one level to another (like Lounsbury's rules for converting
expressions in the agglutinative analog into their actually spoken forms). This
I formalized, years ago, by positing TRANSDUCERS in tandem, a level sand­
wiched between each adjacent pair, and the output of each constituting
7. FROM PARTICLE TO RESONANCE 85

the input to the next. A transducer is an apparatus, for example a

microphone or a loudspeaker, that converts input into output in accordance
with a fixed code. So the proposal is that somewhere inside a speaker's
central nervous system there is a device that transduces a stream of
morphemes into a stream of morphophonemes, a device that transduces that
in turn into a stream of phonemes, and a peripheral device, including also the
organs of speech, which maps that into sound waves (called by engineers the
SPEECH SIGNAL). For the hearer this is turned around. Just as in old-fashioned
railroad telegraphy a receiving operator applies Morse code in reverse to
transduce clicks or buzzes into a sequence of letters, the listener first
transduces the speech signal into a sequence of phonemes, then that into
morphophonemes, then that into morphemes, and so on.
At the receiving end this model suggests a sort of assembly line — or,
better, a "disassembly" line, like the digestive tract. The mouth can be
working on a fresh bite of food as the stomach is processing the mouth's
output from the preceding bite. In the same way, a listener's "phonemicizer"
can be processing the most recent bit of incoming speech signal as the
"morphophonemicizer" is working on the preceding one and the "morphe-
micizer" on the one before that. Thus several transducers work at the same
time. But since they are hooked up in series, they have no independence of
function.
There are many versions of the levels-and-transducers notion, most of
them only hinted at in a general way by this description. Of course the
terminology varies; for example, the transformational-generativists and their
successors speak not of levels but of "components." Allowing for this diver­
sity of detail, the model is virtually universal. I know of no current school of
linguistic theory that does not incorporate it in one guise or another.
Apart from the way it reinforces the other misconceptions to which the
reality theory of morphophonemes has led, the snare in this conception, the
focus of the fallacy, is the insistence on processing in SERIES. I propose,
instead, that the listener may operate on the incoming signal in several ways
at once, in PARALLEL.
What would happen if an alien space-ship exploded and a piece of it fell to
earth? If you have read science fiction you know. Reporters and the military
would converge on the site. Photographs and samples would be dispatched to
various laboratories. Each would analyze the data by its own techniques and
send the findings promptly to all the others, as well as to some central
headquarters where, eventually, all would be integrated and interpreted.
I suggest that a listener's performance may be something like that. The
linearity of speaking (§1.4) does not imply linearity of the internal processing
of heard speech. A human being is entirely capable of doing several things at
once. It may be difficult to devote conscious attention to more than one thing
at a time, but as we pay heed to one activity others can be moving along
smoothly and efficiently out of awareness.
86 7. FROM PARTICLETORESONANCE

So the fact, established in our first three chapters, that listeners parse
and construe dosn't mean that they can't be doing anything else at the same
time. And even the decipherment of morphological structure may well involve
operations in parallel rather than in series.

7.6. Repair or Replace? In §6.1 we presented the original unpolished version

of morphemic theory in the form of three tentative assumptions. The third of
these asserts that the meaning of an utterance is a function of the meanings
of its constituent morphemes and of their arrangement. During the Decade of
the Morpheme it was this proposition that was most obviously inadequate.
Minimal meaningful units there may be, and utterances may be built from
them, and they may contribute essentially to the meanings of utterances, but
there are utterances whose meanings cannot be completely accounted for in
that way. So, unless we were willing to discard the whole morphemic idea—too
radical a proposal to be seriously entertained at the time —we had either to
abandon the notion that morphemes are meaningful (some investigators
explored this alternative) or to recognize additional meaning-conveying
mechanisms.
The treatment in my 1958 textbook was perhaps reasonably representative
of the latter alternative. Morphemes were there supplemented in two ways.
First, it was acknowledged that a composite form may have a meaning not
entirely the product of its constituent morphemes and their arrangement.
Any form, elementary or composite, whose meaning cannot be inferred from
its structure is an IDIOM. Thus all morphemes are idioms, since, being min­
imal, they have no structure; but so are enormous numbers of larger forms
(see passim above in chapter 6).
Second, it was proposed that some of the semantic impact of an uttered
form may be not because of its morphemic (or idiomatic) make-up but be­
cause of secondary associations of its phonemic shape (discussed in §6.5).
On this I now quote a sizable passage, which both points up the futility of the
supplementing procedure and affords a clue to the real solution. The following
is from pages 296-7 of the text:
We can say that tvo vords or phrases are acoustically similar if, under conditions of
some extraneous noise, and lacking defining context, one might be misheard as the other.
Acoustic similarity is a matter of degree, since vith enough noise any vord could be
misheard as any other; but ve are concerned only vith the closer similarities. Any given
focal word (that is, a specific word in vhich ve are interested) is thus surrounded by a
vaguely defined family of vords vhich are more or less acoustically similar to it. The
members of the family will in general have the videst variety of meanings, and yet it may
often happen that some members of the family vill resemble the focal vord not only in
acoustic shape, but also in meaning....
Some of the paralleling resemblances in both sound and meaning are accounted for by
the grammatical structure of the forma Thus sighting amd lighting share the terminal
shape -ingt as well as a feature of meaning, and ve find this shape so widespread in
association vith this meaning that ve break the -ing off as a separate morph, occurring in
7. FROMPARTICLETORESONANCE 87

both sighting and lighting as well as in many other forms. But beyond the grammatically
induced resemblances, there is a vast and complex tracery of subsidiary resemblances
vhich ve do not regard as due to the grammatical structure of the forms: they are, as it
were, "accidents." Once the -ing is broken off from sighting and lightingt we are left
vith a pair sight and light, vhich still resemble each other partly in sound and in
meaning; but ve do not find it feasible to recognise morphs s-, I-, and -ight. And yet there
is really no sharp line of demarcation betveen parallel resemblances vhich ve choose to
account for grammatically and those vhich ve do not—this is borne out by the frequent
disagreements, among laymen and specialists alike, as to the "proper" grammatical
analysis of some forms.
The broad hints of chapter 6 and the historical survey of the preceding
sections of this chapter can now be brought to a head.
Recall the figure I used in passing in §6.1: we can make the morpheme no­
tion cover the real facts only by providing embarrassingly extensive supple­
mentation—a tail big enough to wag the dog.
But, with a tail like the one just described, who needs the dog?
More soberly: the machinery we have introduced to prop up morphemics is
so powerful, and empirically so realistic, that it can also do all the work
morphemics was devised to do. So instead of supplementing we supplant.
Here is how that works.
In the first place, with idioms defined, and with every morpheme by defi­
nition an idiom, we can discard the notion of morpheme and just acknowledge
that idioms come in various sizes. Some idioms, true enough, are tiny and
compact and don't seem to be divisible into smaller pieces that are also
idioms. But that is a matter of degree, and even for the smallest and clearest
instances there are often (as revealed in chapter 6) cross-cutting
associations.
In the second place, why should we think of the impact of an idiom, small or
large, as anything different from the effect of a "secondary association"?
And why "secondary"? Why not recognize that ALL the effect of a heard
utterance is a function of associations between things in it and things the
hearer has experienced before? Some of these associations are extremely
strong and are universally shared (or virtually so) by the users of the
language. The pieces that give rise to these strong associations find their
way into dictionaries and grammar-books, and, if they are small, may be
dubbed "morphemes" by a morphemicist. But, as hinted (I am glad to say) in
the quoted passage, there is no neat boundary separating these from those
features or patterns that give rise to vaguer associations. And certainly this
was demonstrated over and over again in chapter 6.
I had better express that also in another way. In the quoted passage it is
proposed that there is a crucial difference between those features of an
utterance that are caused by its formal, official grammatical (or lexico-gram-
matical) structure, on the one hand, and, on the other, those that rest on
"accidental" resemblances to things in other utterances. Morphemes are the
88 7. FROM PARTICLE TO RESONANCE

minimum pieces in the official grammatical structure; the accidental resem­

blances (or resonances) are secondary associations. But there are no objec­
tive grounds for the distinction between official grammatical structure and
"accidental" similarity. It is a difference of degree, not of kind, with all man­
ner of intergradation between the strongest and most pervasive resonances
and the weakest. Now, there is no way to reduce the features of utterances
responsible for secondary associations to the status of morphemes, but to
reduce morphemes and their effects to associations is easy. The "official
grammatical structure" of utterances is in the last analysis nothing more
than that which correlates with the clearest and most widely shared of the
resonances. But it is not the formal grammatical structure that yields the
resonances; it is the resonances that induce the grammatical structure. So
we abandon the epithet "secondary" and simply recognize a continuous range
of strengths of resonance.
What we have now said twice in slightly differing terms can also be stated
in the jargon of so-called "natural generative phonology." (Remember that in
the TG tradition "phonology" includes morphophonemics, and it is only that
part that concerns us here; nonphonemic variations are a separate category
and should not be allowed to complicate the issue.) Joan B. Hooper, in a book
published in 1976, presents a classification of alternations that is reminiscent
of Baudouin's (§7.2), but then provides for unusual, unproductive, or cross-
cutting associations via VIA rules (a notion credited to T. Vennemann): if two
forms are associated for a particular user of the language but are not related
in any of the "regular" ways, then for that user there is a via rule that con­
nects them. An example is the alternation between Spanish -ct- and -ch- in
lactor 't suckle, to suck' and leche 'milk', or nocturno 'nocturnal' and noche
'night'. There are pairs similar in meaning whose shapes show this difference,
but the alternation is not productive, and any one user's usage may or may
not be affected by any particular instance of it. Bloomfield classed similar
phenomena in Menominee (§7.2) as MORPHOLEXICAL alternations, and did not
provide for them in his ordered morphophonemic rules; but Hooper takes no
note of this precedent. Hooper makes a valiant effort to distinguish between
via rules and her various sorts of morphophonemic rules, but in the last
analysis the difference reduces to one of degree. In her terminology, the
essence of our resonance theory can be stated simply as ALL ALTERNATION
RULES ARE VIA RULES.
Think of a child learning its first language. Anything the child hears is, by
definition, an utterance. Some utterances may consist of a single word, but
never of a bound form. The use and meaning of a bound form are necessarily
inferred by the child from the whole utterances in which it plays a part and
the contexts in which those utterances occur.
What misleads us linguists about this is our sophisticated practice in
working with informants. We consciously look for recurrent pieces, smaller
than words as well as of word size, and tabulate and tentatively identify them
as they turn up. And we write them down in notebooks, perhaps marked as
7. FROM PARTICLE TO RESONANCE 89

bound forms by a preceding or following hyphen, but otherwise indistin­

guishable from our notations of other observed forms. What we are doing is
treating bound forms as though they words — handling morphology as though
it were like syntax. But the child learning a language does not do that. The
child's acquisition of significant features smaller than words takes place in
the indirect inferential way just described. For that matter, so does most of
the child's acquisition of whole words, or of idiomatic phrases, but there is the
difference that, words mostly being free forms (§5.2), one MAY be heard in
isolation, or in an explicitly defining context, whereas bound forms never are.
It is true that users (usually adults) occasionally do what we analysts do:
some pieces smaller than words may come to be recognized just as are whole
words. Recently in my reading I encountered the word ernuiatability, which I
had surely never seen or heard before. I admit that I had to do a double take
to identify the constituents emulat(e) and ability before I could understand,
and I think that was the performance merely of a literate user of English, not
of a specialist in linguistics. It is even true that a piece smaller than a word
may recur with such definite shape and meaning that it is pulled out and
made into a word in its own right. I doubt if we can count my introduction of
the term cran, in §6.5, as an instance, nor yet Pike's adjectives etic and
emie, since for this (as over against emutatability) he and I have too much
technical sophistication. But ism 'doctrine', which dates from around 1790,
and oiogy 'branch of learning', from some twenty years later, are surely
examples. This phenomenon is similar to the coining of an abbreviation (like
burger, §6.8), but I think it is not quite the same. Even if it were much com­
moner than it seems to be, it would not conflict with our proposed new orien­
tation in morphology, since I am in no way denying the reality of meaningful
pieces smaller than words, just classing them IN THE FIRST INSTANCE along with
all other potentially resonance-triggering features of utterances.
Onto the trash-heap along with morphemics goes a pair of hoary notions
that ought to be recognized as bugaboos, not as the useful working
assumptions they are often thought to be: uniqueness and minimality (§6.1).
Since it is resonances that count, not the particular arrangements of
material in an utterance that produce them, there is no motivation for
uniqueness in one's analysis into constituents, only (in some cases) for a
recognition by the analyst of the relative weights of each of any set of
cross-cutting similarities. Microphone can be micro- + -phone—that is, can
have (for a particular user) a significant resemblance to microscope,
microcosm, microorganism and to telephone, megaphone, dictaphone — and
can at the same time significantly resemble its abbreviation mike. There
would be incompatibility in these if we were breaking the word into recurrent
meaningful pieces, but not if we are merely noting resonances.
Likewise mike, occurring in a context which tells us it is the abbreviation of
microphone, although seemingly indivisible, can nevertheless give rise,
indirectly and perhaps more faintly, to all the resonances of the unabbre­
viated microphone. So a smaller piece can have part of its impact via
90 7. FROM PARTICLE TO RESONANCE

resemblance to, and association with, larger pieces — whereupon minimality

ceases to be of any special significance.
7.7. How Does an Utterance Mean? It is necessary to distinguish between
what an utterance means and what the speaker means by uttering it. The
classic illustration of this is Bloomfield's child saying I'm hungry not to obtain
food but to postpone being put to bed (1933, p. 142). The truth of an uttered
assertion depends only on its meaning (and on the facts); whether it is a lie
depends also on what the speaker means by saying it (cf. §9.4, p. 120).
At the moment we are concerned only with the meanings of utterances, not
with speakers' intentions.
Although we speak of utterances carrying meaning just as of a pipeline
carrying oil, we know perfectly well that the parallel is faulty. The meaning of
an utterance is not in the utterance as the oil is in the pipeline; rather,
awareness of the meaning is triggered or induced in the hearer BY the utter­
ance. That is, to be sure, equally true of the grammatico-lexical structure of
an utterance, and of its phonological structure: NONE of these are in the
sound waves that pass from speaker to hearer; ALL are triggered in the
hearer by the incoming signal. Although this is sometimes forgotten, there is
nothing esoteric about it; it is just the difference between any transmission
of matter or energy on the one hand and, on the other, any transmission of
information.
A heard utterance, even if novel, conveys meaning to a particular hearer
insofar as it includes meaningful pieces (that is, FORMS, §5.2, which we have
seen in §7.6 could equally well be called IDIOMS) familiar to that hearer,
together with clues, of the sort discussed in chapter 3, as to how the
meanings of some of those pieces are to be related to one another.
Of course, the meaning of a meaningful piece, also, is not in the piece; a
piece has no meaning except as it is interpreted. A piece is meaningful, in the
first instance, if it reminds the hearer of something heard before in settings
that served to define it (§8.1). Traces of such earlier experiences remain in
the hearer's central nervous system, and are evoked by hearing the piece
again: we say that the traces RESONATE with the incoming stimulus. A piece is
also meaningful (and hence is a form) if its meaning can be inferred, by the
operation which we called "hearer-construing" in chapter 2, from the forms in
the utterance that are already familiar to the hearer. The total meaning of
the utterance for a particular hearer, then, depends on which traces resonate
and with what relative strengths, and on these inferences.
The forms in an utterance are of various sizes. Words are primary, but
there are also forms smaller and larger than words; for example, by definition
any SENTENCE is a form. Larger forms typically include smaller ones, and
overlap is common.
As asserted a moment ago, any participating piece, large or small, may be
unfamiliar to the hearer and yet be understood, in its setting, because of the
familiarity of other participating forms. Most often such a newly interpreted
7. FROMPARTICLETORESONANCE 91

piece (especially a relatively large one), having performed its communicative

duty, merely vanishes. But sometimes, in the particular context, the utter­
ance serves to define the novel piece, which is thereby added more or less
firmly to the hearer's lexicon. Whether this happens or not, every heard
utterance must to at least some slight extent restructure the balance among
the hearer's points of potential resonance. Such restructuring is not part of
the meaning of the utterance, but it does affect how subsequent utterances
will be interpreted. This is how forms take on special meanings within the
limits of a single situation (DISCOURSE DESIGNS ITS DIALECT, §6.2), always with
more subtle longer-term results. The biographically earliest stages of this
constant readjustment are just the infant's acquisition of language.
Often we achieve practical understanding even though the sound of an
utterance is not fully registered; the typical redundancy of the whole com­
municative package (§3.9) makes this possible. Even when the hearer hears
everything, which internal traces resonate more strongly, which less strongly
or not at all, depends on the setting, on the particular hearer's past expe­
rience, and on the hearer's orientation of the moment. It is thus an unsur­
prising commonplace that telling a joke to an audience of six people may
elicit a half dozen varied responses.
A hearer normally does not react equally to all the forms in a heard ut­
terance; some of them, indeed, often evoke no response at all, and may even
pass unregistered. The potential effect of a form may be MASKED if, in the
particular setting, other forms give rise to stronger resonances, or the
meanings of accompanying forms are incompatible with the meaning of the
form in question, or the functors in the utterance relate some of the
contentives but leave others stranded. For example, if you hear They issued 
new ports catalog last week, the attendant circumstances would have to be
most unusual for you to be reminded of either felines or pieces of wood,
although you have indeed heard both cat and tog. (So has the speaker, who
might nevertheless vigorously deny having uttered either form!) If a friend
tells you Old Benny kicked thebucketlast night, you may register that Benny
is dead and note that your friend is being a bit flippant in reporting it, yet
wholly suppress any image of foot-action on pails. And, provided you have
never been exposed to Pike's pun (§2.5), if in the course of a popular talk on
optics you hear the lecturer say The surfs rays meet, the setting may leave
you totally unaware that you have also heard The sons raise meat.
Our traditional atomistic theories of grammatico-lexical design have taken
masking, in an extreme form, for granted. That is what has rendered it so
difficult for us grammarians to cope—-except by departing from our preferred
frames of reference — with even such simple phenomena as puns, to say
nothing of the deeper subtleties of poetry and literature.
The hearer's reaction is also conditioned by what kind of utterance is
expected. Discourse is of many sorts, their nature and number varying from
culture to culture, and each sort elicits its own proper mode of listening, in
much greater variety than suggested in §4.1. We do not attend to a poem as
92 7. FROM PARTICLE TO RESONANCE

we do to a joke or a stock-market report or the proof of a mathematical

theorem. Listeners who know they are about to hear a joke may register a
pun or double entendre that would generally be missed in a more solemn
context; a listener expecting verse will register features of rhythm and
sound-recurrence that are easily (though not invariably) overlooked when
they occur, as though by accident, in other sorts of talk. In the run-of-the-
mill practical exchange of consultative prose, where attention is focused not
on words but almost wholly on the activities the talk is serving to coordinate,
one is likely just to skim the surface of heard utterances, ignoring much that
in another setting might be noticed. At the opposite extreme, as Joos once
pointed out, for fully grasping a finely crafted written discourse fifty years of
rescanning and pondering may not be enough.
If you pick two words at random in a language you know well, you may be
inclined to say that they are totally unrelated: that is, that there is no sig­
nificant resemblance between them in either sound or meaning, so that nei­
ther would reverberate when the other occurs. You may even be able to find
some single word that strikes you as remarkably "isolated" from all others.
My own current candidate for this lonely status is eckankar, which as the
designation of a religious movement just doesn't hook onto anything else.
Kwashiorkor and sigatoka as names of diseases are isolated in the same
way. If I remember correctly, shamrock, handicap, and ordeal, when I first en­
countered them, struck me as queer in that they include familiar meaning­
ful parts but have meanings totally unrelated to the meanings of the parts. In
time such associations, positive or (as for shamrock) negative, tend to fade.
One of the rules about hearers' reactions is that a new or relatively rare form
is more likely to evoke consciously noticed subsidiary reverberations than is
an old, familiar one. Hearing noisome is more apt to activate one's trace of
noisy than the other way around. And though there is no greater similarity in
shape between these two than between handsome and handy, noisome is much
more likely to remind one of noisy than handsome of handy.
As a check on the assertions just made, you might find it interesting to
test your reactions to the word pairs listed in the Notes.
The variation from one person to another in such associations is consid­
erable, since it arises from the total life-history of each user of a language,
and individual life-histories differ markedly even in the most severely reg­
imented culture (§8.2).
Although every use of speech, even silently to oneself, must alter the
balance of potential resonances at least a little, most of this goes on out of
awareness. But sometimes one can observe an association-changing episode. I
learned the word pigtail in childhood, but made no connection between the
hairdo and anything else until one day, at about thirty-five, while driving
through the countryside, I passed a hog barn and noticed the tails on the pigs.
I was over fifty when I first consciously related horizontal to horizon (and
then immediately also vertical to vertex), and at least sixty-five when it
occurred to me that the hap- of happy could be the same as that of perhaps.
7. FROM PARTICLE TO RESONANCE 93

I have known the city-name Bordeaux for decades; only last week it struck me
that it might be bord d'eaux. Surely the professional linguist spends more
time wondering about such matters than most people, but if this sort of thing
did not occasionally happen to the laity, we would have no folk-etymologies.
The intricate tracery of weaker associations among forms, even if it does
its work largely out of our awareness, has a more important role than we are
apt to realize. Consider what happens when we undertake to learn a truly alien
foreign language —one as different from our own as Chinese is from English,
or Swahili from Russian. The first words encountered are terribly hard to
grasp and retain, because they just don't make any sense: the arbitrary
nature of the relation between sound and meaning which holds for all Ian-
guages, but which is difficult to perceive in a language you know well, stands
out like a sore thumb. But after we have by brute force mastered fifty words
or so, each subsequent new word is easier. As varied as the initially acquired
vocabulary may be, in both sound and meaning, yet there are helpful
resemblances: the associative network that we must build for the new lan­
guage has begun to come together. The degree of similarity may actually be
no greater than is implied by the conformity of all the words to the same
sound system. But as you master that system, learning to make and to heed
the distinctions of sound that the users of the language heed and to ignore
the ones they ignore, even that becomes a real help.
There can also be cross-language resemblances which assist the beginner,
not just by virtue of cognation or borrowing (as in the case, say, of English
and French) but also through sheer accident. When I started to learn Chinese,
the particle de was easy, perhaps because of its virtual identity in shape with,
and its mirror-image resemblance in function to, French de. roughly, Chinese
A de  = French  de A And gèi 'give, for, for the sake o f was learned
instantly, possibly because of its resemblance to German geben.
For those who have already acquired some knowledge of several languages,
such cross-language associations can be even more significant. My practical
control of any language but English is extremely limited. Yet, when I am
racking my brain in search of le mot juste, occasionally the first word that
comes to mind is not English but French or German, or from one of the other
languages of which I have a smattering; if it is therefore unusable, it still may
lie on the search path to the word I finally choose. An American who knows
Spanish is probably less likely to refer to the Rio Grande River than one who
does not. Many common expressions are redundant for those with certain
sorts of multilingual experience and not for others: clam chowder soup,
minestrone soup, pita bread, pizza pie, roast beef with au jus, a new inno­
vation, Lets continue on — even the Mississippi River and the Guadalquivir
River, but who can be expected to know that? The echt monolingual who
learns bicycle and its colloquial abbreviation bike does not hesitate to use the
latter in referring to a child's tricycle. The same character would doubtless
also speak unhesitatingly of the dog population of a neighborhood. We snobs
who know a bit of Latin and Greek may find such usages amusing or offensive
94 7. FROM PARTICLE TO RESONANCE

or both (I was uncomfortable in §1.1 with the expression a Thermopylae,

though I let it stand), but we really shouldn't, since we are surely guilty of the
same thing in countless cases in which we know nothing of one of the
languages involved.
The examples just given are blatant enough to make the point, but they are
trivial. Of much deeper significance is the impossibility of matching today the
elegant English prose of good nineteenth-century writers (such as, in our own
field, William Dwight Whitney), which derived its power from the writers'
thorough classical education. Writers of English can no longer turn out such
prose, and readers in general can no longer react to it whole-heartedly,
though actually it requires no remarkable depth of mastery of the classical
languages to perceive the difference between this prose and what is written
today and, without necessarily disparaging current usage, to admire the
former.
It is clear from the above that one of the long-accepted principles of de­
scriptive linguistics requires modification: that which proposes the irrel­
evance of history for synchronic description. Statistically the principle is
ALMOST right, since surely only a very tiny minority of the world's language-
users have ever known or cared anything about the antecedents of their
respective languages. But for those who do, that information — or mis­
information —, although in a sense it is about the past, nevertheless forms
part of their present total language competence, affording potential res­
onating points just as does any familiarity with contemporary foreign
languages. In the perspective of this essay, it does not make sense to think
of an IDIOLECT as anything less than an individual's TOTAL LANGUAGE COM­
PETENCE, whatever bits and snatches of different languages, or different
historical stages of languages, it may include.
It may be that hearers also develop relevant resonance points that do not
correlate with what we would ordinarly think of as forms. I can understand the
overtones of stolid by proposing that when I hear that word my internal trace
for solid resonates along with that for stolid. But how do we account for the
impact of 'Round the rough and rugged rock the ragged rascal ran? Do I have
a resonance point for a series of spaced-out r-sounds? Or for consonantal
assonance, regardless of what consonant it is? How about our reaction to the
regular rhythm of verse? To rhyme? I do not have the answers to these
questions, but it seems to me that with the resonance conception of mor­
phology we can approach such matters realistically, whereas with the
atomistic morphemic notion we are crippled from the outset.
In another direction, I do suggest that language users develop resonance
points, of a sort, for the individual speech sounds of their language, maybe
even for the constituent distinctive features of those sounds.
I know that this proposal runs counter to our usual interpretation of
morphemes as minimum meaningful pieces and of phonemes as minimum
meaningless but differentiating elements. The standard approach tells us the
English word pin is a single morpheme consisting of three successive pho-
7. FROM PARTICLE TO RESONANCE 95

nemes; Bloomfield's series pin : tin : tan : tack (1933, p. 79) reveals them. But
we are unable to break the shape into smaller parts for which we can find
meanings of which the meaning of the whole is a function; so the word is a
single morpheme.
In the resonance frame of reference we do not ask so much. The meaning
of shamrock cannot be derived from the meanings of sham and rock, but
those forms are present in shamrock just the same. Just so, if we find mean­
ings for isolated English P, i, and n, that is enough; they do not have to
combine to yield the meaning of pin.
And, indeed, an isolated English ρ has a vital kind of significance in that it
is not just a brief collision of the lips followed by a puff of breath. An
isolated English wh is not just a physiological movement, not (as Sapir said a
half century ago) just the act of blowing out a candle; it is a SPEECH SOUND.
As such it retains, even in isolation, its tiny residue of the meaningfulness of
all the English speech in which it has been heard. This was Sapir's point in
1933; it has taken me fifty years to understand.
More than that: a heard speech sound in an alien language will evoke the
same sort of resonance. If you hear, in such a language, an utterance be­
ginning with a voiceless dorsovelar stop, you know that you have heard a [k],
a SPEECH SOUND, not just a cough.
We human beings, with or without special training, simply do not react to
speech the way we do to any other sort of sound. Perhaps this explains why it
has always been so difficult to do "pure" phonetic research, aimed at
straightforward, undiluted description of the articulatory motions of speak­
ing: we cannot suppress our instinctive desire for meaningfulness, and it gets
in the way. Perhaps this also helps account for our continuing difficulty, noted
in §7.4, in untangling phonemics and morphophonemics.
7.8. Syntax, Morphology, Syntagmatics, and Paradigmatics. The reader
will have realized by now that the changed view of morphology proposed in
this essay constitutes, at bottom, a return to Saussure.
Saussure did not propose cutting words up into smaller constituents; the
syntaxlike view of morphology of the Decade of the Morpheme was mainly an
American notion. Saussure treated morphology in terms not of dissection but
of association. The discussion in part 2, chapter 5, section 3 of the Cours
makes this perfectly clear, and is echoed by what we have said in §7.7 and by
what we shall say in chapter 8. His "rapports associatifs" (which later came
to be called PARADIGMATIC) are just our "resonances." They hold between a
form actually spoken and an indefinite number of more-or-less similar forms
in the user's internal storage; and Saussure contrasted these with the RAP-
PORTS SYNTACMATIOUES which hold between different forms in what is actually
said.
I think we have sometimes gotten into trouble by trying to equate syntax
and syntagmatics, morphology and paradigmatics. We should not do that,
because it would entail some radical and confusing redefinitions of terms. But
96 7. FROM PARTICLE TO RESONANCE

it is worth while to see to what extent they correlate.

Thus most, but not all, syntagmatic connections are between words and
phrases, and are therefore syntactic. Yet we have to admit instances in
which, for a particular user on a particular occasion, such ties hold between
the parts of a single word: we discussed the case of emulatability in §7.6, and
Bloomfield (1933, p. 235) cites bittersweet and blackbird. At least, an unusual
and unfamiliar word may sometimes be understood via a recognition of its
parts and an inference as to how the parts are to be related, and that is
syntagmatic. So some morphology is syntagmatic, and some syntagmatic ties
are morphological.
Similarly, most paradigmatic effects are resonances evoked by individual
words, and are thus morphological. But some are not, since in every language
there are also idiomatic phrases which users recognize and respond to as
wholes.
We see that the correlation of syntax-morphology with syntagmatics-
paradigmatics is only partial. And it varies from language to language. While it
is nonsense to think of all languages as "really" agglutinative, it is probably
true that every language has some measure of morphological agglutination,
and certainly the case that some languages have more of it than others. And
the more agglutinative a language is, the more obvious and important —by
definition —are its syntagmatic connections within individual words.
However, there is another correspondence that turns out to be not just
close but exact. Recall the distinction made in chapter 2 between parsing and
construing as two different stages or aspects of the hearer's processing of
an incoming utterance. A moment's thought shows that the former relates to
paradigmatics, the latter to syntagmatics! So here, also, our "new" ori­
entation turns out to be more Saussurean than I would have expected.
7.9. Summary. In this chapter and the preceding one we have shown how the
atomic morpheme notion arose in the history of our discipline, pointed out
the weaknesses of the morpheme theory of morphology, and outlined a more
realistic and efficient resonance theory in its stead — a theory that in
surprising measure harks back to Saussure.
 8. HOW LANGUAGE MEANS
8.1. The Logic of Resonance Theory. In earlier chapters of this essay we
have encountered over and over again difficulty in drawing a sharp boundary:
between linguistics and psychology (§0.2); between syntax and morphology
(§0.3); between iconic and arbitrary (§1.5); between language and paralinguage
and each of these and kinesics (§3.2); between functor and contentive (§3.);
between two different words and two norm shapes of a single word (§4.7);
between single word and idiomatic phrase (§5.2); between one resemblant
family of words and another (§6.5). Even in phonology, we find clearcut cate­
gorization only in hearers' interpretations of the incoming signal when lis­
tening for word shape, not when merely identifying heard words and not in the
actual sound of speech (§5.1). Because of our customary (though usually
unstated) working assumption of absolute sameness, along with our liking for
airtight categories, such boundary difficulties tend to worry us.
As a matter of fact, in the more informal phases of our scholarly work,
even as in our everyday affairs, when we call two things "same" or "different"
we may be using any of a large number of criteria, often without overt
specification and yet usually without getting into trouble. At one extreme, we
may be impelled to deny the equivalence of what might be thought of as two
occurrences of a single form: you and I both said Good night! to Joe last
evening, but you said it cordially and I said it in disgust, or you said it with
your Down East accent and my dialect is Midwestern. At the other extreme, in
the discussion of language history one might quite legitimately declare that
English tooth and Latin dens are the same — meaning, in this case, that each
is the endpoint of an unbroken tradition from one and the same word in the
common ancestor of English and Latin. By that criterion, Latin dentatus and
English dentate are NOT the same, though one can easily find a criterion by
which they are.
It would be ridiculous to propose that as analysts of human language (or of
anything else) we try to get along without making identifications. As language
users we obviously make them. Not everything in a message we receive has
to be familiar, but if it includes nothing that we have experienced before
there is no way we can understand. Since we do understand some of what we
hear, it follows that we have heard some of it before, which means
recurrence, which means sameness. As analysts we must follow the lead of
users in this as in other respects; otherwise we are playing a pointless game.
This is what Bloomfield was summarizing and encapsulating in 1933 in his
"fundamental assumption of descriptive linguistics" (p. 78; original in italics):
" I N EVERY SPEECH-COMMUNITY SOME UTTERANCES ARE ALIKE IN FORM AND MEAN­
ING." All our investigation of language implies some such belief; he chose to
98 8. HOW LANGUAGE MEANS

make it overt.
Yet, so expressed, the assumption may go farther than necessary or justi­
fiable.
We usually think of the relation of sameness as reflexive, symmetric, and
transitive. "Reflexive" means that anything is the same as itself. "Symmet­
ric" means that if one thing is the same as a second, the second is the same
as the first. "Transitive" means that if one thing is the same as a second and
the second the same as a third, then the first is the same as the third. So
understood, the notion yields razor-sharp classifications: given any one thing,
anything else in the universe is either the same as or different from it, with
no middle ground.
That is all very neat and works just dandy in mathematics and for artificial
systems which, perhaps without the devisers being aware of it, have been
devised so as to render it neat (see the end of §1.2). But for empirical
application, as in linguistics, a modified relation may be more useful: an
"almost same," or perhaps a "very similar," with a little leakage in it. "Almost
same" is weaker than "(exactly) same," so that if two things are exactly the
same they are also almost the same, but not the other way around. Our
modified relation retains reflexivity and symmetry, but not transitivity. That
is, though it sounds funny to say so it is nevertheless true that anything is
almost the same as itself. Also, if one thing is almost the same as a second
then the second is almost the same as the first. But if one is almost the
same as a second and the second is almost the same as a third, it does not
automatically follow that the first is almost the same as the third; instead,
that requires independent determination.
Actually, "almost same," so described, is what we usually mean in our
everyday affairs when we just say "same." The precise logical usage that I
presented first is a sort of formal idealization of the vernacular sense, useful
in some realms of discourse because it can define comfortably sharp
boundaries and yet yield a close approximation to empirical truth. But with
other subject-matter these very properties may be injurious instead of
helpful.
Here is an example of almost sameness that may help with the linguistic
interpretation to be given in a moment. As is well known, the music of
Wagner's Ring des Nibelungen is built around a set of leading themes, each of
which represents some character, place, object, or other recurrent feature of
the story. Since the total performance time is about twenty hours, some of
these themes recur many times. Except that they don't — at any rate, not
exactly. Some of the themes are subtly altered from one appearance to the
next, in keeping with the developing plot and mood of the story, so that two
occurrences are at most "almost same." The difference between the first
presentation of a theme, near the beginning of the cycle, and its final
repetition toward the end can be so great that if the latter were heard
immediately after the former it would be hardly recognizable. But if the
audience has followed through the whole sequence, so that all the intervening
8. HOW LANGUAGE MEANS 99

links are there, the connection is clear.

On a vaster scale, yet in a subtler way, the whole Western musical tradition
is like that. To a person from some alien culture suddenly exposed to a
sample of our music, be it a Beethoven string quartet or a chip of punk rock,
the experience is baffling. Music means what it does to participants in our
culture because it reminds them of other music in their experience, all the
way back to childhood; thus, the richer an individual's personal musical
history, the more there is to resonate with newly heard music, and therefore
the more understanding and the more pleasure.
In the case of language, If I understand something you say to me today, it
can be (recall §7.7) because some of the forms you utter are almost the
same, in sound and meaning, as some I heard and understood yesterday (or
last year); and when I understood them yesterday (or last year) it may have
been because they were almost the same, in sound and meaning, as some I
had heard and understood the day before (or the year before); and so on. Yet
there might not be sufficient resemblance between the forms heard today and
those heard and understood many years ago to serve as a basis for me to
understand your new utterance. A series of pairwise almost-sames can add
up to a not-at-all-the-same of the endpoints.
But this example needs three amendments.
For ease of statement I made a simple identification between degree of
similarity of form-occurrences and proximity in time, but that is surely cor­
rect only in a very broad statistical sense. If A, B, C, ... N are form-occur­
rences each adjacent pair of which are almost same, but whose terminal
elements are not almost same, then I may have heard N yesterday and 
thirty years ago, and yet it may be my memory of  that enables me to
understand you when you say A to me today.
The second correction has to do with numbers. Where I have referred in the
example to only a handful of different dated form-occurrences, in real life, in
the total experience of any single individual, occurrences of almost-same
forms must number in the hundreds, the thousands, or even the millions.
Finally, although the preceding remarks were cast in terms of FORMS, they
apply equally to CONSTRUCTIONS. Someone says / have one white cat and one
bl cat Is the construction of black cat the same as that of wfvte cat?
How about One white cat and one shaggy dog? How about white cat and one
dog? These are all similar; they are also all different. No matter how elab­
orate a typology we recognize, there will be gray areas, not because our
methods are wrong but because that is the inherent nature of our material.
The user of a language does not require exact identity of constructions any
more than exact identity of forms. The minimum clue normally needed for
understanding a newly heard collocation is that its construction be ALMOST
the same as those of various collocations heard and understood earlier.
8.2. Social and Individual. No institution plays a greater role than language
in holding together the fabric of human society. At the same time, nothing is
100 8. HOW LANGUAGE MEANS

more intensely personal that the language habits of individuals, a tight-fitting

yet flexible garment around the ego of each. Moreover, speaking and hearing
speech are both, at bottom, performances of individuals. Group speaking, as
in choral recitation or song, is a very special sort of activity requiring careful
rehearsal; joint listening, as when several people work together in trying to
decipher a poorly recorded message, is individual listening accompanied by
consultation.
Properly impressed by the first of these two aspects of language,
Saussure placed it at the focal center of his theory, and Bloomfield, following
Saussure's lead as well as his own bent, was guided by it in formulating his
fundamental assumption of descriptive linguistics as cited in the preceding
section. We lose none of the power of that assumption, and add to its realism,
if we interpret his "same" as our "almost same": this allows speech com­
munities to blur into one another in space and in time, which is what they
typically do, whereas absolute sameness would imply the sort of sharp
boundaries that are rarely found in space and never (except upon the ex­
tinction of a language) in time.
For a multitude of legitimate reasons, our practical handbooks of languages
(and most purely theoretical treatises too) emphasize the social rather than
the individual, and in so doing tend to suggest sharp boundaries, uniformity
within a speech community, atomicity, and other aspects of absolute same­
ness. Yet many of these works are better and more realistic than the theory
on which they are ostensibly based. A dictionary's entries are (mainly) words
and phrases, implying not only atomicity at a certain size-level but also the
ability to distinguish between one atom and another. This basic organization
is dictated by the use for which a dictionary is designed: one has to be able
to look things up. But dictionary definitions are typically compound, presented
in sections and subsections arranged to keep similar usages together, more
diverse ones farther apart, and it is left to the user's discretion to decide — if
so impelled —whether a particular word-shape in, say, listed senses A3 and C1
is one word or two. Given the traditional format, which would clearly work best
if atomic morphemics were true, this organization is a sensible compromise
between the theoretical implications of exact sameness and the empirical
reality of almost sameness.
Moreover, experienced dictionary-makers and writers of grammar-books do
not try to do what cannot be done. They cannot include all the tiny twists and
turns of form and meaning developed in the constantly changing idiolects of
individual users or even of small subcommunities (think how long a dictionary
entry would have to be to present all the dialectal denotations of Main
Street!). Likewise, a sensible grammarian does not pretend to itemize "all"
the constructions of a language, but groups them together on the basis of
varying degrees of similarity and leaves room for overlaps, indeterminacies,
and analogical innovations. Some features of the language of any community
are in more general use than others and some change more slowly than
8. HOW LANGUAGE MEANS 101

others; a useful description is necessarily confined to the more widespread

and sluggish parts.
However, all this well motivated emphasis on the social aspect of language
has led us to give short shrift to the personal phase. We forget that (as
pointed out in §7.7) even in the smallest and most tight-knit speech com­
munity, however great and important the overlap of the idiolects, they never­
theless differ widely and change rapidly. A practical reference work neces­
sarily aims at breadth of coverage, which can be achieved only by omitting
ephemeral and local specificities. But those tiny twists and turns are part of
the reality of language-in-action, necessarily taken into consideration and
provided for by any realistic theory of language or, as in this essay, of
hearing and speaking. They cannot be cast aside as mere "parole" instead of
"langue," or as mere "performance" rather than "competence."
The personal aspect, and the crucial diversity of idiolects, were described
so beautifully over a century ago by Whitney that I cannot refrain from an
extended quotation (1867, lecture 1, par. 22):
The limits of variation of meaning; are, of course, very different in different classes of
vords. So far as these are designations of definite objects, cognizable by the senses, there
is little danger of our seriously misapprehending one another vhen v e utter them. Yet,
even here, there is room for no trifling discordance, as the superior knovledge or more
vivid imagination of one person gives to the idea called up by a name a far richer content
than another can put into it. Two men speak of the sun, vith mutual intelligence; but to
the one he is a mere ball of light and heat, vhich rises in the sky every morning, and goes
dovn again at night; to the other, all that science has taught us respecting the nature of
the great luminary, and its influence upon our little planet, is more or less distinctly
present every time he utters its name. The vord Pekin is spoken before a number of
persons and is understood by them all: but some among them knov only that it is the
name of an immense city in Asia, the capital of the Chinese empire; others have studied
Chinese manners and customs, have seen pictures of Chinese scenery, architecture, dress,
occupation, and are able to tinge the conception vhich the vord evokes vith some fair
share of a local colouring; another, perhaps, has visited the place, and its name touches a
store of memories, and brings up before his mind's eye a picture vivid vith the hues of
truth. ... just so is every part of language liable to be affected by the personality of the
[user] ...; and most of all, vhere matters of more subjective apprehension are concerned.
The voluptuary, the passionate and brutal, the philosophic, and the sentimental, for
instance, when they speak of love or hate, mean by no means the same feelings. ... It is
needless, however, to multiply examples. Not half the vords in our familiar speech would
be identically defined by any considerable number of those vho employ them every day.
Nay, vho knovs not that verbal disputes, discussions turning on the meaning of vords,
are the most frequent, bitter, and interminable of controversies?
Whitney here pays no attention to a difference implied by logicians when
they distinguish between two sorts of DEFINITION: respectively REAL and NOM­
INAL. A real definition tells how a particular thing or kind of thing fits into the
universe of things. The definition of a human being as a featherless biped,
102 8. HOW LANGUAGE MEANS

which follows the classical Aristotelian pattern of assigning the definiendum

to a class ("biped") and telling how to distinguish it from other members of
the same class ("featherless") is an example. If this definition is correct and
if you understand it, then when confronted by anything whatsoever you can
tell whether or not it is a human being. A nominal definition, on the other
hand, tells only how a particular term is (or should be) used. The preceding
sentences of this paragraph are examples of nominal definitions, since they
tell you how you should use the terms real definition and nominai definition if
you want to sound like a logician.
We heed this distinction in our reference works, in the difference between a
dictionary and an encyclopedia. The latter gives information about things (and
places and people); the former purports merely to give information about
words.
Yet it is significant that to consult either a dictionary or an encyclopedia
we look up a WORD. It seems to me Whitney was right to ignore the distinction,
not because the logicians are wrong, but because their point was irrelevant
for his purposes, as for ours. The logician's distinction suggests that one's
knowledge of the meanings of words can somehow be set off from one's total
sum of knowledge of the world. I see no nonarbitrary way of doing that. If
there is a boundary between word-knowledge and world-knowledge it is as
broad and hazy as all the other boundaries listed at the beginning of §8.1. I
see no objective basis for claiming that the meaning of the WORD Pekin was
the same for Whitney's globe-trotter and his stay-at-home, and that the dif­
ference was purely in the former's more intimate knowledge of the PLACE. I
think we have to say what Whitney said: that the WORD had a fuller meaning
for the one than for the other.
We can squeeze all the foregoing into a slogan. To do so we shall use a
notion developed by foreign-language teachers. Because French and English
share so much vocabulary, words in the two languages that resemble each
other in shape but not in meaning are "false friends": for example, the
unalerted user of English is likely to interpret French équipage as 'equipment'
rather than as 'crew'. But even in a single language, no one else ever means
by a form quite what you would mean by it, and what you yourself mean by it
today is doubtless a bit different from what you would have meant yesterday.
So we have our slogan: EVERY FORM YOU HEAR IS A FALSE FRIEND. This is almost
the opposite of Bloomfield's fundamental assumption. That doesn't mean he
was wrong and this is right; it means only that he was looking in one direction
along the personal-social scale and we are looking in the other; and both
directions are important.
It is interesting to note that our professional attention to the individual
aspect of language, and to the dialectic relation between the individual and
social phases, has tended to be keener in historical study than in descrip­
tive. That is why the historical perspective, though marginal to our present
interest, plays the part it does in the following succinct summary:
8. HOW LANGUAGE MEANS 103

All individuals acquire their speech habits, to start with, from those about them.
Even in the most homogeneous community, no t v o people have exactly identical speech
habits.
All individuals alter their speech habits throughout life, as new challenges are en­
countered and must be dealt vith by talking.
Since different individuate have different personalities and encounter different challenges,
their speech habits tend to be modified in different ways.
Therefore, as time goes by the speech habits of different individuals, even in a single
community, tend to diverge.
Within a functioning community this tendency is held in check by the continued
requirement of mutual understanding. The speech habits of community members are thus
constantly intercalibrated. Each adopts some innovations from others; each abandons
those personal innovations that impair crucial understanding.
Therefore, although the language of a community inevitably changes as time goes by, as
long as the community holds together the tendency tovard divergence is held in check and
all its members continue to be able to understand one another reasonably well (Sowesay
all of them use "the same language")
But if a community grovs and spreads out too wide, or splits into t v o or more daughter
communities out of touch with one another, the built-in checks on divergence cease to
operate among the daughter communities, and in course of time it can become so great
that members of different daughter communities can no longer understand one another's
speech. (Then v e are likely to say that a single language has become several.)

8.3. The Meanings of Meaning*. Reference and Denotation. We accept

here a working principle that Bloomfield insisted on (1933, p. 145): if two
forms (in a single language or dialect) differ in shape, we must assume that
they differ also in meaning, even if the difference is too subtle for an
observer to grasp. Bloomfield took this as a consequence of his fundamental
assumption of descriptive linguistics, cited above in §8.1; it follows equally
from our own superficially opposite guiding slogan.
Here, now, comes some calibration of the terminology we use in talking
about meaning. One cannot solve terminological problems either by word
magic (examining words to decide what they "really" mean or really "should"
mean — of all people, we linguists should know that!) or by objective lexi­
cographic study of the ways in which they have been employed. Yet a careful
examination of our habitual usages, both everyday and technical, is helpful,
since it can reveal ambiguities and obscurities that we can then try to avoid.
Referring back to §1.5, let us consider an A that means  for a particular
individual or group  Using an arrow for the connection (relative to C) be­
tween A and B, we have
A-> 
Just a few years ago I was made aware of a difference of interpretation of
the noun meaning that I had never before noticed. For me, in terms of the
104 8. HOW LANGUAGE MEANS

diagram just presented, the meaning of A had always been B. For a colleague
it had always been →. For him the word meaning was like any other gerund-
participle in -ing used nominally to denote an ongoing process or relation, so
that JoMs meaning is dear is parallel to, say, Their singing is beautiful. In
contrast, I had always taken the word as like the noun building, thus denoting
something meant just as building denotes something (already) built: JoMs
meaning is clear is like The buiiding is tail.
That ambiguity we will dispose of for our present discussion by setting
aside my colleague's interpretation and sticking to mine.
But what is the nature of B?
If, at table, I say to you Please pass the sait and you pass the salt, it
follows that for you and for me there is a tie between the word sait and salt.
So if salt is A relative to the two of us as C, it follows that  is salt — the
granular white stuff in the shaker.
The ties between words and things or situations that lead to this conclusion
are assuredly real, but there are nevertheless some other interpretations to
be considered.
We have to distinguish between CODE and MESSAGE (this is one of the two
distinctions melded by Saussure into his overloaded single opposition langue-
parole; the other, which is actually orthogonal to this one, is the social-
individual contrast discussed in §8.2). In the episode just recounted, we can
say that the word-occurrence salt REFERRED to the stuff you passed me. In
order to keep the cover term .meaning usefully flexible, we will henceforth use
REFERENCE for this kind of connection between things and events in an actual
episode.
But the causal sequence of events in our episode was possible only be­
cause of a connection or correspondence in the code, partly shared by you
and me, that I used in framing my message to you. Speaking now of the code
rather than of any one message, we might choose to say that the word
"denotes" the substance, just as we could say that in Morse code two dots
denote the letter T . That would take us back to our first interpretation, and
is how I handled the matter in 1958.I now consider it generally more useful to
follow the suggestion of E. A. Nida and say that the DENOTATION of A is those
properties of B, as perceived or imagined by C, by virtue of which it is
appropriate to use A to refer to B. Denotation is thus a second specific sense
of our general cover term meaning. Denotation is in the code; reference is in
messages.
The advantage of Nida's approach as over against my former procedure is
easy to show. I once read a wholly serious argument — by some logician
ignorant or scornful of Bloomfield's working assumption — f or treating morn­
ing star, evening star, and Venus as exact synonyms because they all stand
for the same object. (The author forgot, as did Wagner in Tannhäuser, that
the morning and evening stars can both be Mercury rather than Venus, so
we'll ignore it too.) Now, among the properties an object must have if it is
legitimately to be referred to by the term morning star is a contextual one:
8. HOW LANGUAGE MEANS 105

barring a cloud cover, it must be visible in the east at dawn. This property is
not required for the use of Venus, and definitely excludes using evening star.
We are saved all manner of unnecessary pseudological trouble by taking the
denotation of a term to be not specific things but only the conditions that
must be met by a thing, situation, or state of affairs in order for it to be
appropriately referred to by the term. Who would want to say, for example,
that unicorn is meaningless (that is, has no denotation) merely because there
are no unicorns? Our definition of the term rescues us from any such dis­
tressing consequence.
Some investigators distinguish between SEMANTICS and PRAGMATICS, meaning
by the former the connections between things in messages and things and
situations spoken of, by the latter the connections between messages and
their users (senders and receivers). Although that distinction is perfectly
valid, it cuts across the ones we are making here and will not particularly
concern us. In our perspective the pragmatic aspects of an actual episode of
speech are part of the reference of forms in the utterance, and are made
possible by the denotations of those forms. In the example given earlier,
please refers to the speaker's establishing of a context of situation marked
by politeness, and it can make that reference (and perhaps have that effect
on the addressee) because that is what the word denotes — that is the way
the word is habitually used.

8.4. Mental Images. Still another interpretation of the nature of  in our

diagram is to take it, not as an object or class of objects and not as a
property or condition, but as C's mental image of the object or property. That
is supposed to explain why we can speak of something in its absence: the
thing itself may not be in evidence, but our mental image of it is. This is an
ancient and honorable notion, but perhaps properly requires two arrows
instead of one, as well as a change of notation (C's mental image of B we
represent by /, leaving  to represent what it did earlier):
A→/→
Sometimes this figure is drawn as follows, in order to maintain graphically
the distinction between the public (lower line) and the private (upper line):

In either case, the second arrow is to be interpreted as generically like the

first: just as the word means (stands for, denotes) the mental image, the
mental image represents (stands for, means) the object. The connection
between the word and the object is not direct but mediate, the intermediary
being the mental image.
Bloomfield would have none of this, and I think he was right, but not for his
reasons.
106 8. HOW LANGUAGE MEANS

Bloomfield's reasons derived from his behaviorism, which he identified

(mistakenly, I think) with physical ism. We have no evidence, goes the
behaviorist argument, for what goes on inside other people's heads except
just what they say and do, so that it is circular to explain the latter in terms
of the former. This has a merit which we should appreciate and retain,
whatever our reaction to the rest of his position: it gets meaning out in the
open where we can deal with it publicly. And, true enough, when I ask you to
pass the salt you do not reach into your head for it; you reach for the salt
shaker. If you ask me to close the door, I cannot satisfy you by knitting my
brow and concentrating, but must bestir myself and manipulate the door.
But it is also a fact that I can say Apennyforyour thoughts, and that in
response to this you can, if you so choose, tell me what you are thinking.
This everyday occurrence is clear evidence for the reality of inner mental
activity. At the same time, properly examined it brings us forcibly back to
our earlier issue of the individual and the social.
For if you can tell me what you are thinking and I can understand you, it
must be that we share, to some extent, the semantic conventions of the code
in which you do the telling. But in order to share those conventions, we must
have both ACQUIRED them, directly or indirectly, in terms of shared or parallel
PUBLIC experience. If I know what you mean when you say you are thinking of
orange juice, it is because both of us have seen and tasted it and both have
learned to call it orange juice. Of course, my basis for understanding you
does not have to match your own experience in all detail (that would be
impossible). Using mostly words known to us both through shared public
definition, you can tell me of something I have never experienced, calling up in
my imagination something more or less parallel to your own mental image.
Recall Whitney's example (§8.2) of the place-name Pekin. The man who has
visited the capital of China can paint verbal pictures of it for the one who
knows it only by name, certainly not rendering the latter's response to the
word identical with his own, but nonetheless greatly altering it.
If we try to go beyond the preceding and ask, for example, whether orange
juice "really" tastes the same to you and to me, we have crossed the
boundary between the potentially public and the truly private and our question
becomes unanswerable. This is the boundary Bloomfield and other behaviorists
felt was crossed by any allusion to individual mental activity. For speechless
animals they were surely right, but, JUST BECAUSE OF LANGUAGE, the barrier is
pushed far inward for human beings.
To be sure, even apart from the many times our interpersonal transactions
wander perilously close to that barrier, there is constant danger of mis­
understanding. As we have said, every familiar form you hear is a false friend.
And the danger is the greater, the further removed the subject of discourse
from the immediate setting, since the latter can quickly give the lie to major
misinterpretations. But if idiolectal divergence never ceases, neither does
intercalibration. So we come once again to the intimate dialectic interplay
8. HOW LANGUAGE MEANS 107

between the individual and the social, and see that much of that interplay is
made possible exactly by the nature of language.
If I nevertheless agree with Bloomfield's rejection of the interpretation of
the meaning of a word as a mental image, my reasons are that this tra­
ditional proposal is in one way too complicated; in another, too simple.
It is more complicated than necessary because in the first instance —and,
for many purposes, also in the last—it suffices just to say that  KNOWS that
A denotes  Even this is redundant (though perhaps not objectionably so),
since in our original description of A and B we emphasized that the con­
nection between them exists only relative to some C. As we said in §1.5: that
saft means salt is not a property of either the word or the substance; it is a
property of users of English. If I can get you to pass the salt by using the
word salt, it is because you and I share the property of knowing —or of be­
lieving, or of laboring under the delusion — that salt denotes salt. Call it a
delusion, if you like that better: but note that if you are not a user of English
and want to communicative with people who are, then this is one of a large
number of, interrelated delusions that you will have to acquire. If a speech-
community has n users, its language is a folie à n.
The resort to mental images is at the same time too simple because it
considerably undercuts the complexity of the internal apparatus necessary
for  to acquire, to have, and to exploit such delusions (or such knowledge J.
We shall see this in detail later (§8.6). Here let us simply point out that if a
diagram is to include, along with A and , 's mental image of Bt it must also
include Cs mental image of A Representing these latter two by IB (instead
of just /) and by /Ą we have, in an appropriate two-dimensional arrangement,
IA → IB
↓↓
A → 
Furthermore, even if, as suggested earlier, all the arrows are generically alike,
they are surely of different species. The horizontal arrow connects a symbol
with a symbolized; the vertical arrow runs from a mental image of something
in the shared external world to that thing. The diagram is to be interpreted
dynamically, as indicating parts of actual sequences of activity. The vertical
dimension is the private-public axis, or covert-overt, or THINKING-ACTING; the
horizontal dimension is the symbol-thing axis, or denotandum-denotatum, or
THEORIZING-PRACTICING. A person can think the word salt or can hear it or say
it; the relation between these two (which is along the vertical axis) is the
same as the relation between thinking of salt and experiencing or handling
salt. The relation between the thought of saft and the thought of salt (which
is along the horizontal axis) is the same as that between soft and salt, and
both are the same as the relation between the discussion and planning of a
town board about a bridge wrecked in a storm and the actual activities of the
108 8. HOW LANGUAGE MEANS

crew that repair the bridge. Of course it is more complicated than that, but
this at least shows why three-point diagrams, such as those given earlier, are
inadequate.

8.5.The Meanings of Meaning: Denotations and Connotations. The occur-

rence of a particular contentive in a heard utterance, unless it is masked in
the way described in §7.7, calls to the attention of the hearer one or more of
its denotations. Typically the context is such as to suppress some of these
denotations and enhance others, but that can be thought of as part of
masking.
In addition to denotations, a heard form is likely to evoke diverse CONNO-
TATIONS.
(1) Any denotation may have experiential associations; e.g., blood, which is a
denotation of blood, may be associated by a hearer with injury, bleeding,
feeling faint, and so on. If the sight of blood sometimes evokes such re­
actions, then the mention of blood — the occurrence of the word blood— may
also give rise to them. There is nothing linguistic about the associations
themselves (speechless animals presumably have them too), but if a word can
evoke them then they are connotative values of that word. We shall refer to
these as EXPERIENTIAL connotations.
I am not sure I would know how to distinguish clearly between the deno­
tation of a word and its experiential connotations. Much of the reaction of
Whitney's world-traveler to the word Pekin (§8.2) should perhaps be thought
of as the latter rather than as part of the former.
The remaining sorts of connotations are all resonance effects.
(2) A heard form (most usually a contentive) may suggest forms with
which it often cooccurs in utterances: hermetically is apt to remind one of
sealed, brush of comb (and vice versa); body of soul (and vice versa); and so
on. If the occurrence of form Ą reminds one of form A2, then the DENO-
TATIONS of A2 are the SYNTAGMATIC CONNOTATIONS of A1.
These first two sets of connotations, to be sure, are likely to overlap
considerably.
(3) A heard form may remind one of other forms to which it is similar in
meaning, sound, or both. The DENOTATIONS of these other forms are the
PARADIGMATIC CONNOTATIONS of the heard form, respectively (3a) SEMANTIC and
(3b) PHONETIC The most important of these, as we have seen in many of our
examples, are those that are both semantic and phonetic.
The foregoing applies to functors, to forms that include functors, and to
forms on the contentive-functor boundary just as to "pure" contentives,
except insofar as some functors have little or no denotation. That resonance
effects can involve functors is clear from the joke about the man who went
into a drugstore to buy diapers; when the clerk said "That will be two
thirty-eight, plus seventeen cents for tax," he responded with "Oh, never
mind the tacks; we use safety pins."
Any and all of these various kinds of connotation can play a part not only
8. HOW LANGUAGE MEANS 109

in routine consultative prose but also in all other sorts of verbal transaction
or activity.
Their role in WORD-PLAY (and plays on words) has already been abundantly
illustrated (chapter 6) but we add one example here. Years ago one of my
fellow college students was a young woman named Goldie Blank. A writer who
knew her well published a story in one of the student magazines featuring a
character named Silverie Dash. It seems to me the only sort of connotative
connection that plays no part in this is that labeled 3b above. Some
headline-writers are fond of this sort of thing.
Many WORD GAMES, including many of those on television game shows, when
stripped of their elaborate trappings turn out to be in essence like crossword
puzzles. In crosswords one has two sorts of clues, formal and semantic. (The
former are graphic rather than phonetic, to be sure, but since our writ­
ing-system is alphabetic that is a minor matter.) One knows the number of
letters in an entry, and if some of the cross-cutting words have already been
entered then one has reasonable guesses about some of the individual
letters. The semantic clues range from literal to some involving the wildest
sort of far-fetched associations, experiential, semantic, and phonetic; without
the resonance network which underlies connotations, many of them would be
totally unsolvable.
In a very different sector of our culture, clinical psychologists sometimes
use a type of WORD-ASSOCIATION TEST devised originally by Karl Jung. The
stimulus is an isolated, and therefore presumably context-free, contentive
uttered by the psychologist; the response is supposed to be the first word
that comes into the patient's head. The response may reflect any of the
sorts of similarity outlined above: go may elicit away (syntagmatic) or come
(semantic paradigmatic); candy may elicit sweet (experiential) or dandy
(phonetic paradigmatic, which the psychologists call KLANG-ASSOCIATION); and
so on. The supposed clinical value of the test lies in subtle semantic
associations (e.g., that of tree : fìsh), of which the patient is unaware and
which might be suppressed if they were more obvious. Many years of clinical
use have shown that such hidden associations indeed exist, underscoring for
us the enormous complexity of the resonance network.
POETRY is regularly discussed and characterized in terms of SOUND and
SENSE, which at first thought would seem to be just our "phonetic" and
"semantic." In fact, however, the sound is not just sound and the sense is not
all of the sense. The sense referred to is mainly the denotations. The sound
is not effective because of pure onomatopoeia or because it is pleasant to
the ear as instrumental music is; rather, it has its impact because the words
are chosen and arranged to maximize the effect of all the connotative values,
especially the phonetic paradigmatic ones, and to make them reinforce the
denotations. In addition, the most powerful poetry adds to the machinery of
everyday speech the special devices of regularly spaced recurrence which we
class as rhyme, assonance, amd rhythm.
110 8. HOW LANGUAGE MEANS

The reasons for the effectiveness of good poetry can be underscored by

considering the two commonest sorts of not-so-good poetry. In PHONETIC
DOGGEREL (usually just called DOGGEREL), rhythm and rhyme are precise, but
the content is trivial so that the phonetic devices are wasted. In what I like to
call SEMANTIC DOGGEREL there may be a striking and perhaps profound mes­
sage, but there is no real control of rhyme, rhythm, or any other element on
the "sound" side to reinforce it — the author seems to think that writing
things out in oddly broken lines is all that is required. (It almost need not be
added that some would-be poetry is both kinds of doggerel at once.)
8.6. Internal Storage. Back in §5.2 we proposed that we can "think of forms
as registered somehow in the brains of users of the language, ready to be
replicated and used as needed; any such use is then an occurrence of the
form." We can now carry that farther. Of course, we are not talking about
brain architecture — not about the actual physical locations of things inside
our heads. We don't know or need to know how the brain works, other than
that it functions in some way to store and to process information. What we
have to say has to do only with what communications engineers call the
"logic" of that storage and processing.
What is the process by which a hearer recognizes a word or phrase?
If you see someone's face, to identify it as that of a particular person you
must have some record with which the appearance is compared and which it
matches. To identify a heard word one must have some internal record that
the perceived sound matches. Such internal records are what we referred in
§7.7 as "traces" of relevant earlier experiences. Each trace consists of a
paired form and meaning —or, better, it consists of KNOWLEDGE of a form and
of the denotations that go with it. Each trace is of what Saussure called a
LINGUISTIC SIGN and represented by his famous elliptical figure divided into an
upper and a lower half, the upper for the "concept" [our "(knowledge of) the
denotations"], the lower for the "acoustic image" [part of our "(knowledge of)
the form"].
But the form part of this pairing is not just an acoustic or phonetic shape;
it includes also an array of SYNTACTIC PROPERTIES: the shape is surrounded, as
it were, by a grammatical "aura" or "field." For a literate user in a literate
community, moreover, the phonetic shape is matched by a GRAPHIC SHAPE, to
constitute, in all, the following sort of package:
denotation(s);
form: shape: phonetic;
graphic;
grammatical field
Let me call such an assemblage of related information a PACKET. The
grammatical field in a packet is not the form's syntactic ties in a particular
occurrence, but a range of POTENTIALS extracted (by experience) from the
environments in previous occurrences, by virtue of which the form is felt to
8. HOW LANGUAGE MEANS 111

be likely to occur with certain syntactic ties and less likely, or unlikely, to
occur with others. Or we can think of the form as having certain VALENCES —
not, as in chemistry, just a certain number of possible bonds, all alike, but a
set of different KINDS of potential ties to accompanying forms. For example,
transitive verbs like abandon and sing have a valence for an actor, in addition
to which the former has a strong (compulsory) valence for a goal, the latter a
much weaker (optional) one; an intransitive verb like rejoice has only a
valence for an actor. Look and listen have no goal valence, but their gram­
matical fields include the fact that they can be "transitivized," the former
with at, the latter with to When the hearer hears a word its valences are
activated, and it is by virtue of them that the hearer is able to supply the
structure-in-depth.
The internal records for different linguistic signs are not isolated; they are
connected to one another in a complex interlocking network of "resonance
channels" along which impulses can travel to evoke all the sorts of conno­
tations described in §8.5.
There are doubtless also other sorts of packets. For example, if two words
share a denotation, perhaps knowledge of the denotation is in a separate
packet with connecting threads to it from the packets for the two different
words. Homophones, such as pearl and pur/, may be stored in a single packet
for one user of the language and in separate packets (with connecting res­
onance channels) for another — even if their overt use of and reaction to the
words appear identical. If your language is English but you know a little
French, do you have separate packets for English crew and French équipage
and a third packet for their shared denotation? Does the grammatical field
for the first include a flag specifying ENGLISH, while that for the second bears
a tricolore? (I imagine that in the brain of a real bilingual the information for
the two different languages tends to be somewhat more firmly compart­
mentalized, though it would be pathological — almost split personality — for
there to be no connecting pathways at all.) For some things the internal-
storage arrangement must be quite different. Thus, since the record of a
memorized poem often fades differentially, some parts being easily recalled
after others have vanished, perhaps to start with it occupies a whole series of
packets with special connecting linkages.
Be that as it may, as one listens and speaks and carries on private internal
speech the total apparatus is constantly changing, with the development of
new packets, the strengthening or weakening of part of an old one or of a
resonance channel, and so on.
Suppose, for example, that you are introduced to a Mr Joseph Catwood. If
you are fast about such things, you register the name and the man's
appearance in such a way that if you encounter either again, in actual
experience or just in imagination, you will recall the other. But if you are like
me, the name is apt to get lost, and even the record of the face may fade.
Or, in some cases, I retain both but lose the association, so that both name
and face are familiar but neither activates my memory of the other.
112 8. HOW LANGUAGE MEANS

How about the nanne of someone you have never seen? In my reading a few
years ago I came across mention of a brilliant nineteenth-century British
woman named Mary Fairfax Somerville, and was so intrigued that I obtained
and read her biography. I suppose my internal packet for her has a blank spot
ready to be occupied by a record of her appearance if I ever see her portrait,
but so far I haven't. And despite all that I learned about her, my memory of
her name has proved peculiarly unstable. At times I have been able to recall
none of it; at times I have dredged up Mary Fairfax but not Somerville. Once,
in fact, frustrated by my inability to get past Fairfax, I spent a half hour
tracking down the passage in which I had originally learned about her.
More recently, again while reading, I encountered the unfamiliar word con-
cinnity. The input itself supplied knowledge of the form's graphic shape, and
because of another sort of knowledge (that of English spelling-to-sound cor­
respondences, stored in my brain heaven knows where or how) also a phonetic
shape. Moreover, from certain features of the shape (concinnity like activity,
unity, parity etc. — this is one sort of morphological resonance at work),
together with what was required by the surrounding words, I knew it was a
noun, which is a succinct way of saying something about its grammatical field.
So there I had a new packet, lacking only one ingredient, for which the
context offered no clue: I didn't know what it meant.
When that happens in our society there is something we can do about it: we
can use a dictionary. The dictionary entry on a word supplies, in general public
terms (§8.2), just the kinds of information we have posited above for a user's
internal record of a linguistic form, in that it assembles a graphic shape, the
associated phonetic shape, grammatical information, and indication of mean­
ing via gloss and example.
That fact raises an interesting question. Did Saussure portray his "lin-
guistic sign" as he did, and have I posited the kind of internal storage-units I
have, because of the way dictionaries are constructed? Or are dictionaries
constructed as they are because through long trial and error they have
evolved to reflect more and more accurately the facts of language?
Although in linguistics, as in all branches of science, the relation between
observation and theory is a chimney lift, observations requiring revision of
hypotheses and hypotheses suggesting fresh observations, in the long run
the observation of facts leads the way. I therefore take the organization of
dictionary entries as in a general way support for our proposals about internal
storage.
At the same time, our description of internal machinery is patently in­
complete. Hearing and speaking involve more than just searching through
vocabulary. As hearers we also construe; as speakers, as we shall see in our
final chapter, we also select and assemble. There must be some sort of "work
space" where all this is done.
It might be suggested that our posited machinery does not really constitute
a THEORY, capable of guiding observation and yielding new information, but
8. HOW LANGUAGE MEANS 113

only a FAÇON DE PARLER, a terminology in which we can say just what is said
in other terminologies, no less and no more. In the spirit of self-criticism
announced in §0.1 this is the kind of question we should often put to our­
selves. It is particularly à propos here because of our insistence, a few
paragraphs back, that we are talking not about brain architecture but only
about design-logic. Yet I would like to think that as formulated above the
issue is put too harshly. Even terminologies can differ in the extent to which
they stimulate inquiry, and the terminology developed here has at least
stimulated mine. Beyond that I leave the matter open.

8.7. Summary. In this chapter we have tried to achieve a more realistic con­
ception of "same" and "different," have reconsidered the dialectic relation
between the individual and social aspects of language, have pinned down and
labeled several important meanings of meaning, and have suggested some­
thing about the storage and processing of language-related information in the
brain.
 9. THE CRAFT OF SPEAKING
9.1. Hearing and Speaking; Chance and Choice. At the beginning of this
investigation (§0.2) our question was: (A) WHEN WE HEAR SOMEONE SAY SOME-
­­­­­ IN A LANGUAGE WE KNOW, HOW DO WE KNOW WHAT IS SAID?
In chapters 1 through 8 we argued that hearers know what is said (that is,
they understand — when they do) by parsing and construing. Parsing means
identifying the words and phrases in an utterance, a task which turns on
gestalt perception and may require any of several different modes of listen­
ing (chs. 4-5). In addition, parsing serves to discover denotation and to evoke
all manner of connotations (chs. 6-8). For both parsing and construing, hear­
ers make use of the evidence in the heard utterance together with clues in
the context (§3.2). The latter may sometimes disambiguate obscure wording,
but mainly it is the other way around: the primary function of speech is to
disambiguate its context (§3.9). Although, in trial-and-error fashion, tentative
construing may induce reparsing, in the long run parsing is basic. Construing
is aided especially by functors (§3.3; but note §3.8). A hearer perhaps rarely
grasps all the meaning an utterance is capable of triggering (§7.7, p. 91), but
normally that does not matter, since usually only part of that meaning is
intended by the speaker or crucial to the matters at hand.
As we turn now, for this final chapter, from hearing to speaking, our ques­
tions are: WHEN WE SAY SOMETHING, (B1) HOW DO WE KNOW WHAT WE HAVE SAID? and
(B2) HOW DO WE KNOW IF IT IS WHAT WE MEANT TO SAY?
The answer to BI is the same as the answer to A
That may seem strange at first, because our tendency is to conceive of
speaking and hearing as opposites, like exhaling and inhaling or taking versus
giving. But it is easy to see that that is wrong. We hear our own speech as
well as that of others, so that what we transmit is a PROPER SUBSET of what
we receive. Furthermore, we understand our own speech by just the mech­
anisms by which we understand the speech of others; we can know what we
are saying only because we hear it. Feedback is not just a convenience but an
absolute requirement for production. That holds, as we shall propose in a
moment, even when we are silently talking to ourselves, "thinking in words."
The answer to question B2 is that you can know if you are saying the right
thing only by comparing what you hear yourself say with your knowledge of
what you want to say.
These answers imply that speaking is GOAL-DIRECTED. Granted that there is
an element of randomness in all organic behavior, and that some human
actions, including some acts of speaking, are automatic responses (on both
9. THE CRAFT OF SPEAKING 115

of these we shall say a bit more later on), it is nevertheless proper to think of
speech first and foremost as purposive.
To start with, WE SPEAK TO BE HEARD. Why and by whom depend on the cir­
cumstances, but it is hard to imagine a situation in which that basic motiva­
tion is missing. Sometimes, to be sure, your intended audience is only your­
self. When that is so, the overt motions and resulting sound of talking are
often (though by no means always) suppressed, since that saves energy and
you can "hear" yourself when speaking silently almost as well as when you
speak aloud. But even in this private case the proof of the pudding is in the
eating—the proof of the production is in the reception.
Also, most of the time WE SPEAK TO BE UNDERSTOOD, which means that we
make a more or less serious effort to use words and expressions with which
we think our addressees will be familiar. The direct opposite is not unknown-
technical or mystical jargon used to impress or confuse rather than to share
information or seek help — but for our purposes we need discuss that sort of
conduct no further.
Our answers to questions Bl and B2 propose that receptive ability is a
necessary condition for productive ability. But it is only necessary, not suf­
ficient. In at least one essential way speaking is different from listening to
the speech of others. Although it is true that (as suggested in §4.5) you
achieve efficiency in listening by basing your moment-to-moment expec­
tations on previous experience, nevertheless it remains a probabilistic affair.
But what is CHANCE for the hearer is CHOICE for the speaker. In listening to
the speech of others, the only decisions you have to make are about the
identity of words or phrases not heard clearly or in some other way con­
fusing. In speaking, you constantly have to decide what to say.
Generating an utterance is thus a creative act, properly to be compared
(allowing for differences of medium and scale) with writing a sonnet, com­
posing a sonata, trying to devise the proof of a mathematical theorem,
painting a portrait, or carving a statue.
9.2. The Work Space. WHERE is the work of creating done?
Forget about the sonata, the portrait, and the statue; think here only of
building something with words — say, some verses. In constructing the lim­
erick that I shall use as an example later in this chapter (§9.5), I sat with
pencil and paper before me, with a dictionary and a rhyming dictionary close
at hand, and with various sorts of relevant information stored inside my head
ready to be drawn on. Some of the work of creating was done overtly, with the
pencil and paper or by muttering to myself; some of it was done quietly,
inside my head. By the time the task was finished there were a good many
pieces of paper in the waste basket (or would have been, except that I was
keeping a record of the failures as well as of the successes so that I could
report in detail on the creative process).
116 9. THE CRAFT OF SPEAKING

People facing such tasks differ greatly in how they apportion the work be­
tween outside and inside. By his own account, if Martin Joos had a half hour
for a writing task, he would sit silent at his typewriter for most of that time,
working things over in his head, and then type out a final draft—perhaps only
a half page, but ready to go. At the other extreme, under the same circum­
stances, back in my days B.C. (= "Before Computer"), I would keep my
typewriter going for the whole time, crossing out abandoned efforts, fre­
quently inserting a new sheet of paper and copying over as much as I had of
the most recent version. I used to find it impossible to judge the
acceptability of what I had written unless it was cleanly typed, so even the
slightest slip of a finger could lead me to redo a whole page (in the course of
which I might not just correct the typo but also change something else).
What counts most in this comparison is not the differences but the
common features.
We assume that internal efforts at designing something are basically like
observable overt creative behavior: that what Joos was doing in his head was
in essence the same as I was doing at the typewriter. In fact, we propose that
the patterns and mechanisms of ALL silent internal behavior are at bottom
the same as those of overt conduct and perception. This notion is supported
by all manner of introspective reports, but derives also from the proposition
that the contours of private mentation must in general arise from an
internalization of overt activity and experience simply because there is
nothing else for them to arise from.
That is why we speak (§8.6 ρ 112) of an internal WORK SPACE: an interior
region which is at the focus of one's "mind's eye" (and ear, and maybe touch,
smell, and taste) during the creative process (just as the paper before you is
at the focus of your attention as you try to write something), and which is
also where you do the parsing and construing required to understand received
speech.
What if you are not working? Then what you observe in the work space
seems to be a sort of free-association running commentary, an "inner voice,"
not isolated but tangled up with all sorts of other impressions, both linguistic
and nonlinguistic (e.g., traffic sounds, music), and both auditory and nonaudi-
tory (e.g., faces, colors, aromas) — and all this superimposed on whatever
sensory input is arriving at the moment from your environment. Many inves­
tigators have described this inner stream of speech. Their accounts agree in
essentials; that they differ about details is only to be expected, since each is
reporting his or her own introspection.
But you are not just a passive observer of this stream of real, recollected,
and imagined experience. You can grab hold and try to direct it. With the
minimum of such conscious supervision one has dreaming. With the maximum
of direction one has thinking: planning, problem-solving, building. Instead of
passively accepting what your unguided inner voice says to you, you search
9. THE CRAFT OF SPEAKING 117

for words and constructions that will mean what you want to mean and you
force your inner voice to use them — and then maybe you also open your
mouth and turn the covert stream of speech into an overt one, or put pen to
paper and convert it into graphic form.

9.3. Choosing Words. Since hearing and speaking are such closely related
activities, we assume that whatever internal apparatus is used for the former
(§8.6) is available as a resource also for the latter.
Now, I read French just about as easily as English, looking up unusual words
no more often and using a monolingual dictionary when I do. Yet when I try to
speak French I am constantly at a loss for words. Why?
Let us look again at our scheme for internal packets of stored information
about linguistic signs (§8.6):
denotation(s);
form: shape: phonetic;
graphic;
grammatical field
In listening to the speech of others, or in reading, the point of entry into a
packet is via the form. In speaking, it is via the denotation. It would seem that
the connection within the packet between the form and the meaning is not
just a road traversed with equal ease in either direction. Rather it is a
dual-lane highway, half of which can be under construction, or closed for
repairs, while the other is open; or it is two one-way paths, each of which can
be widened and smoothed by repeated use while the other remains cluttered.
So I can read French until I turn bleu, and still not be able to speak it fluently;
I will learn to speak only by speaking.
Even in a language one has used since childhood finding the right words can
be tricky. The search is normally instituted from the meaning end, though
occasionally with some feeling as to what the shape should be like; it is then
conducted, no doubt, by chasing denotations and connotations around through
the resonance network, where there are bound to be some smoother paths
and some blocked ones. Here are examples of the kinds of troubles everyone
encounters:
First: we have all had the experience of knowing exactly what we want to
say and being certain that there is an exact word for the meaning, yet being
unable to find the word. You can misplace a word you have known perfectly
well, and which you would recognize instantly if you heard it; and you may
then feel obliged to try to recover it. For some of us this happens most
frequently with personal names (§8.6), but ordinary words are by no means
exempt. One morning neither my wife nor I could recall the slang expression
for a business situation in which a possible gain would be offset by a con­
comitant loss. We could think only of stand-off (in contests) and push (in
118 9. THE CRAFT OF SPEAKING

gambling), neither of which was right. (Later in the day, while occupied with
other matters — Aha! Trade-off!) I once lost the shape of the term for the
dingus on a map that gives the scale and explains the symbols. As an
experiment I stubbornly refused to look it up or ask for help, and it was three
years before the unmarked detours in my brain, circumventing the blocked
lane, finally led me to it: legend, (Of course I knew this was right as soon as I
found it, since the lane from the shape to the meaning remained unen­
cumbered.)
Second: we have also all on occasion known what we want a word for but
been forced to realize that we don't know the word. In our society this is
perhaps most likely in dealing with unfamiliar technical aspects of material
culture — plumbing if you are not a plumber, crocheting impedimenta if you
aren't crotchety, and so on — and is why we have cover terms like dingus,
doohickey, and thingamajig. Of course, we can always learn. Once in a hard­
ware store I managed to describe what I needed well enough to find out that
the specialists labor under the delusion (§8.4) that it is an awning cleat—a.
fantasy that now resides comfortably and accessibly in its own packet in my
own head.
We also pick up words for temporary use with no intention of retaining
them indefinitely. While traveling you stop for the night in a motel on a street
whose name you remember until leaving the next morning, but may never
again be able to recall even if you want to (nor can you later remember your
motel room number, though you may have retained it as long as you did the
room). Maybe, as some specialists suspect, the stored memory never totally
evaporates; but we should be careful about what we stow in closets with
doors that stick.
Third: one step beyond that is to know exactly the meaning for which you
want a word, including denotation and connotations, but to be forced to the
conclusion that our language supplies no such word. Once, in a belletristic
context, I needed an adjective something like feil (as in with fell intent), but
eviler and spookier. I couldn't find one, even with the help of a thesaurus, and
still think none exists. For this situation, as for the preceding, I imagine a
packet in which the pigeonhole for the shape is empty — except that in this
case it can't be filled.
To be sure, there are also cases of empty shape-slots that are tillable.
For example, in the course of doing research in a technical field we can be
led to a concept for which our terminology supplies no ready-made label.
Socially speaking — that is, in terms of the horizontal axis of the figure on
page 107 (§8.4) — a "concept" is simply THE POSITION OF A TERM IN A TERMI­
NOLOGY, where a "terminology" is understood to be a system of labels with
instructions for their use. For the brain of a participating individual — the
vertical scale in the diagram—a concept is just a packet. A concept for which
there is no label is an unoccupied terminological position; for the individual it
9. THE CRAFT OF SPEAKING 119

is a packet lacking a shape. In such a case we are often led to devise a

shape.
Similarly, upon the acquisition of a new baby, a new pet, a new car or boat,
or any other sort of new thing on which it is our habit to bestow a proper
name, our language and culture supply us with ways to do so. The shape is
often an old one, as when a child is named after an ancestor, but since the
denotation is new so is the packet.
Fourth: akin to not finding a word at all is finding the wrong one, or thinking
we have. Sometimes our articulation forges ahead of our selection of words.
Once, referring to the beer left in my bottle, I said "I can't just (pause) slug
this down all at once," and was then momentarily bothered because this use
of slug struck me as queer. (Later I looked it up, to find that even if the
usage was unusual for me it is quite current.) Whilst watching football on
television a friend said to me "let me know if they show pictures of the
coaches," and then, a moment later, corrected the last word to "announcers."
All manner of malapropisms (§6.7) and paralallies (§6.6) belong here.
Finally: quite different from any of the foregoing is the distressful feeling,
usually in emotionally charged circumstances, of not knowing what to say.
This is sometimes described as being "at a loss for words," but it isn't just
the shapes that are lost, it's the whole packets: we don't know what we WANT
to say. Sometimes we resort to nonspeech sounds; sometimes we stammer
out the beginnings of a half dozen words, all inappropriate, before quieting
down and trying to regain our cool. I would like to think of this as a momen­
tary disengagement of the whole language apparatus, except that (as sug­
gested earlier, §8.2) I really don't know where our language capacity ends and
our other behavioral repertories begin, and I doubt if this particular sort of
experience offers much of a clue.
Let us return now briefly to my problem of finding words in French. I don't
think there is any difference in kind between that and my occasional trouble
with my native English. The stoppages are all of the sorts we have itemized.
There are just more of them.

9.4. State Parameters, Although the circumstances which elicit talk are
bewilderingly diverse, there are some widely recurrent features. Every
speech-situation is a challenge. In each there is a goal to be achieved (or to
be abandoned); each presents a puzzle to be solved; each can be thought of
as a WORD GAME to be played by rules with which you are more or less
familiar.
Furthermore, many of the significant differences from one speech situation
to another can be characterized in terms of a relatively small number of
STATE PARAMETERS, somewhat reducing the seeming chaos. To get started at
this, here is a passage I wrote some years ago (1977, pp. 287-8, in an article
analyzing jokes):
120 9. THE CRAFT OF SPEAKING

A man rises in church to recite the Apostle's Creed in chorus with the rest of the
congregation. What he is to say is totally prescribed in advance. Even the phrasing and
intonation are ritualized. He can do it if he has memorized vhat he is to do, in its exact
linguistic form, otherwise he cannot. And, if he has, then he needs no help from
Stanislavsky in order to achieve an acceptable performance. Now consider a witness to an
accident who has been asked to dictate an account of what he saw to a police
stenographer. What the witness dictates has not been stored inside his head in a fixed
verbal form, ready to be "read off" to the stenographer. If he has memorized anything—
and, if he has not, he is valueless as a witness—it is the sequence of sights, sounds, and
smells at the accident. He calls these into awareness, scans them, and describes them in
words. It does not in the slightest matter what words he uses, as long as they convey the
right information.
It is a fact that a human being has both nonverbal and verbal memory, both nonverbal
and verbal imagination, and that when one speaks one may be either reading ... [a stored
text] or putting words together to describe something stored or imagined in nonverbal
form If it is true that one's memory of an event or fact can sometimes be stored more
securely by working out a verbal description of it and memorizing the description, it is
also true that one may rely on remembered or imagined events to help remember words
or phrases....
Remembering and posing a joke requires both of these typical human skills.... The poser
memorizes the exact wording of certain key parts of the joke [because they must be
recounted in just that form or the point will be lost]. ... The poser memorizes, in
nonverbal form, some skeleton sequence of the events covered in the joke. He makes no
effort to memorize the rest of the material he has heard in any one telling of the joke.
Instead, when he is about to tell it himself, he takes off from the part memorized verbally
and the skeleton memorized nonverbally, and fleshes out the rest to fit.
That tells us something of the possible FORM of the information to be con­
veyed by one's words in a speech situation: it may already be in the exact
wording one is to use (the Apostle's Creed), or may be in nonverbal form
requiring to be worded (the testimony), or partly each (the joke). In the case
of the witness, the information was supposed to be factual; if you invent a
joke or write a story, you are drawing on imagination as well as memory and
the discourse is fictional; if you know it is fictional but pass it off as factual,
you may be lying.
I say "may be" because it is not that simple. Once, when I was sitting on
the waterfront a few blocks north of the center of town in Myrtle Beach, South
Carolina, someone asked me how to get to the business district, and I said
"Just go north from here about a mile." My instruction was wrong, but not a
lie, because I was not intentionally giving false information. I am not sure why
I said "north" instead of "south." There are many conceivable reasons. One
could be momentarily confused, perhaps because of extensive experience on
the west coast, where, when one is facing the water, the direction to one's
right is indeed north. Or one might not know the local geography very well, and
9. THE CRAFT OF SPEAKING 121

think one really is south of the center of town. Or one could have habitual
trouble with directions, or just with north and south, or even just with the
WORDS north and south. In my case I think it was something like the last of
these, but surely this brief itemization of the possibilities is not exhaustive.
The path from experienced facts to a verbal report about them, as every
historian knows, is often tortuous and treacherous. Indeed, it is possible for
some combination of the kinds of factors I have listed to lead one to make a
true statement by accident, or even to tell the truth despite an intent to lie.
The IMMEDIATE SOURCE of the information can vary. In all three cases in the
quoted passage, the information (or misinformation) was in the speaker's
head, but often that is not so. If you are reading aloud to a child, the infor-
mation is not in you but in the book, already in an essentially linguistic form;
your task is just to is convert it from writing to speech. An announcer cover­
ing a baseball game for a radio audience gets the information in nonlinguistic
form by observation, or perhaps in part (in linguistic form) from spotters, but
performs the same sort of processing as was required of the witness working
from memory. A simultaneous interpreter at the U.N. differs from both,
receiving the information coded in one language and transmitting it in
another.
There are also situations in which the information put into the words is
beside the point: that is, is not determined by the purpose of the talking.
Some logorrhea may belong here — ceaseless chatter kept up only in order to
hear one's own voice. But there are other types. Suppose you are testing a
P.A. system. To save the trouble of each time dreaming up something to say
in this circumstance, someone long ago invented the formula Testing: one,
two, three. If someone needs to find you in an almost empty auditorium during
a blackout, you can say anything —or sing or whistle —as long as you keep it
up long and loud enough.
Another parameter in speech situations is the degree of URGENCY. The si­
multaneous interpreter must keep pace with the speaker whose message is
being processed, never falling more than a few words behind (though it may
often be necessary to delay that much because the source and target lan­
guages require the information to be presented in different linear se­
quences). In contrast, someone translating a novel or a textbook can try
various alternative target-language versions of a word, sentence, or para­
graph, take time to look things up in dictionaries, and so on. If you meet
someone on the street and want to utter a polite greeting, you must decide
rather quickly what to say or it is too late. If you enter a limerick contest
with a deadline two weeks away, you can dawdle.
Still a fourth parameter — the last we need mention —is the range of FREE­
DOM OF CHOICE allowed by the situation. This is narrowest in a ritualized event,
such as in reciting the Apostle's Creed, delivering one's lines in a play, or
even in greeting an acquaintance on the street. Of the examples we have
122 9.THECRAFT OF SPEAKING

given, it is greatest in the cases where the content of the words is totally
irrelevant, but also considerable in, say, a limerick contest with no prescribed
topic.
The OPPORTUNITY  BE CREATIVE is not exactly a parameter, but it varies
from one speech situation to another, and has intimate ties with choice and
purpose, as follows.
On the one hand, if there are no alternatives from which to choose there
can be no creation. Consider a thermostat. When the ambient temperature
falls below a preset threshold, its only available reaction is to turn the heat
on. There is no creativity in that; it is the sort of phenomenon we referred to
in §9.1 as an automatic response.
On the other hand, if there are no constraints at all then creativity is
equally impossible, because there is no standard against which to judge
performance. How can one recognize success if there is no possibility of
failure? That is why artists speak of the discipline of form: it is more of a
challenge to write a sonnet than to grind out "free verse" and, in the same
measure, success is more fulfilling.
So we can summarize: NO CREATIVITY WITHOUT FREEDOM; NO CREATIVITY
WITHOUT CONSTRAINT.

9.5. An Example. The actual process of working out what one is going to
say in a speech situation — the procedure involved in solving the word puzzle
— is easiest to follow in an instance with low urgency and relatively high
freedom of choice, simply because in such cases everything moves more
slowly. So let us observe the process of creating a limerick. The demon­
stration presented below is "honest," in that when I set out to prepare it i had
no more idea of what the outcome would be than you have as you start to
read it. As we go through the process of composition I will describe each step
as clearly as I can. Remember that it is this, the procedure, that concerns us,
not the quality of the outcome.
Since the product is to be a limerick, the words must be so chosen and
arranged as to fit the scheme prescribed for all limericks. Using dit for a less
prominent syllable, DA for a more prominent one, R ("rest") for a moment of
silence, and parentheses around optional items, we can display the RHYTHMIC
part of the required pattern as follows:
(dit) dit 01DA dit dit 02 DA dit dit 03 DA (dit) 0 4 R
(dit) dit 05DA dit dit 06 DA dit dit 07 DA (dit) 08R
(dit) dit 09 DA dit dit 10DA (dit)
dit 11DA dit dit 12DA (dit)
dit 13DA dit dit 14DA dit dit 15DA (dit) 16R.
The superscript numerals mark the sixteen "beats," which should be about
evenly spaced in time, though not necessarily as precisely as in music. This
9. THE CRAFT OF SPEAKING 123

spacing holds even though three of the beats (numbers 04, 08, and 16) coin­
cide with silence rather than with a prominent syllable. To maintain this reg­
ular spacing, if an optional dit is omitted an equivalent amount of silence
must replace it. In musical terms, the scheme can be described as consisting
of four measures of twelve-eight time, the third and fourth lines making up
the third measure.
The rest of the pattern (also, at bottom, a rhythmic matter) is a RHYME
SCHEME: the first, second, and fifth lines must rhyme, as must the third and
fourth.
The foregoing (rhythm and rhyme) constitute what the limerick form pre­
scribes as to SOUND (recall the end of §8.5). It also specifies something as to
SENSE. First: a limerick must have SOME sense, no matter how trivial. Sec­
ond: by tradition the content is light rather than serious. This tradition is so
strong that a limerick on a serious topic is immediately interpreted as ironic,
sardonic, or "sick." Third, and most important: the content should be se­
quenced in such a way that the last line, especially its end, delivers some­
thing unanticipated —a punch, like that of a joke, though not necessarily one
that would be effective in the joke genre. At the same time, the last line has
to rhyme with the first and second, so that the audience is given a consid­
erable clue in advance as to what the end will be. Herein lies the challenge for
limerickers, who must play the game fairly and still try to keep the audience
from outguessing them.
Within these constraints, the tight ones of sound and the looser ones of
sense, we are now free to follow our fancy. We can start with any clue we
choose. The proposal to compose a limerick is not a clue, only a program. A
clue would be either some possible topic (such as linguistics, or Picasso's
paintings, or elephants, or woodsheds) or some word or phrase (say, linguis-
tics, Picasso, or woodshed] whose sound is a challenge because finding useful
rhymes would be tricky. I shall here choose a clue of the second sort:
specifically, since its meaning is à propos in this essay, the word iinguistics.
We stress this word on its second syllable: linGUIStics. That limits how we
can use the word in our limerick. For example, we cannot let either the first
or the third syllable coincide with a beat. If the word appears at the end of
the first line it will fit into the prescribed pattern only as follows (similarly at
the end of the second or fifth):
(dit) dit 01 DA dit dit 0 2 DA dit lin03GUIStics 0 4 R
If we use it at the end of the third line (and similarly for the fourth), it can fit
this way:
(dit) dit 0 9 DA dit lm10 GUIStics
Either way, we must find appropriate rhyming words or phrases —two in the
first case, one in the second. Let me now search through the vocabulary
124 9. THE CRAFT OF SPEAKING

stored in my head and see what is available. I find staTIStics, DIStichs,

loGIStics, MYStics, balLIStics. But in this game we can use any aids we have
at hand. In Walker's Rhyming Dictionary I find sphraGIStics 'the science or
study of engraved seals' (Hey! An English word with the initial consonant
cluster sfr-!), paTRIStics, chròmaTIStics (with a secondary stress on the
first syllable) 'political economy with relation to wealth-production' (whatever
that means), possibly matèriaLIStics HèlleNIStics, chàracteRIStics, heu-
RIStics, and TRIStichs. You may be able to think of others: for example, a
place-name or a person's name or nickname. In sound, any of the listed words
would do, regardless of the location of secondary stresses, since in the
English poetic tradition such a stress can freely be either promoted to a DA
or demoted to a dit But in sense some of them are so odd that I should
hesitate to use them. I am not quite sure what they mean, and although a
limericker ought not to underestimate the sophistication of his audience his
limerick will fall flat if it is unintentionally unintelligible. From that point of
view, I think the safe ones are statistics, logistics, mystics, characteristics,
and perhaps ballistics and heuristics.
Consider meaning further. If our clue word is linguistics, then our verses
will presumably have something to do with language. Statistical methods are
not prominent in linguistics, but our discipline does involve the marshalling
and organizing of data (logistics) and discovery procedures (heuristics); and
language, linguistics, and linguists all have properties (characteristics). One
of the striking traits of certain would-be scientific linguists is that they
object, often violently, to scholars who issue weird and wonderful pronounce­
ments about language on the basis of pure intuition. The objective-minded
may derogate such characters as mystics. Unless all this is too solemn for
limerick treatment, then, our materials seem promising: maybe we can poke
fun at both scientism and mysticism.
Perhaps we could start something like this:
A beginning student of linguistics
— but the rhythm is no good. If we want to express a meaning of this sort in
the first line, we must find a rhythmically acceptable phrase that is more or
less synonymous:
A student who studied linguistics
A freshman who studied linguistics
A freshman who started linguistics
A student m love with linguistics
A student beginning linguistics
or the like. Let it go for now, and try possible content for a second line:
Manifested undesirable pseudophilosophical tendencies
9. THE CRAFT OF SPEAKING 125

which could be whittled down to the proper rhythm and rhyme this way:
Shoved mystical characteristics
But we may want to use mystics as the rhyming word in the last line, in which
case mystical should not appear here. Is there a rhythmically identical
near-synonym? Perhaps; but in searching for one I find a word of quite dif­
ferent meaning that suggests a possible way of going on to the third line:
Shoved typical characteristics
Of...
Now we need to specify, in the third line or in the third and fourth, what
undesirable traits the student was showing, and then tell what his instructor
did about it (... and sent him back to the mystics or the like). In the process,
we have to find a third-and-fourth-line rhyme. One well-known variety of
antiscientific thought is dualism, often associated (rightly or wrongly) with
Plato. So perhaps
Of DA dit Platonic.
His teacher, laconic,
... sent him back to the mystics.
Here is what we now have:
A freshman beginning linguistics
Shoved typical characteristics
Of DA dit Platonic
His teacher, laconic,
... sent him back to the mystics.
We could strengthen the connections in this by inserting soon at the begin­
ning of the second line, without hurting the rhythm because it would become
an optional dit. However, now that we have this much, it strikes me that our
limerick might be more effective as a warning than as a narration. I mean by
this that we could change the first two lines as follows, and the rest to
match:
If a student beginning linguistics
Shovs the typical characteristics
Of...
Now let us solve for the DA dit of line three. Rhythmically, two syllables
(THINKing) would be better, though three with stress on the second (con-
FUsion), could be squeezed in.
At this point I have turned to a dictionary for suggestions. Because of the /
of Platonic and teacher, I have scanned the entries beginning with that sound
126 9. THE CRAFT OF SPEAKING

and have found a word sufficiently insulting to fit. (The resulting assonance is
not compulsory in the limerick form, but is always welcome.) With a couple of
other changes, we get:
If a 01 student, in 02starting lin03 guistics,04
Shows05 firmly fixed06 character07 istics08
Of 09twaddle Pla10tonic,
His11teacher la12 conic-
'ly 13tosses him 14 back to the15mystics. 16

9.6. Analysis. What we learn from the foregoing confirms the generalities of
§§9.1-4 but goes well beyond them.
PURPOSE, YOU will observe in the example a HIERARCHY of purposes, some
more inclusive and some less so.
The overriding object was to present an instance of creative behavior that
could be dissected to show what the process involves (which is what we are
now doing). Within that frame the regnant aim was to compose a limerick.
That aim generated various subsidiary goals, such as choosing a topic or
clue, finding suitable rhymes, and so on, each as an immediate means to the
ultimate end. Indeed, at any hierarchical level, the pursuit of a goal can give
rise to subordinate goals the attaining of which may help one toward the
superordinate one.
In our example, some subsidiary goals were abandoned unachieved because
the work done to attain them led, by. chance, to a modification of a super­
ordinate goal that rendered the subordinate one superfluous. For example, at
one point the aims, each subsidiary to those listed before it, were (P1) to
build the limerick on the word linguistics, (P2) to make fun of both mysticism
and scientism; (P3) to find a synonym for rnysticai to use in place of that
word in the tentative third line Showed m ystical characteristics, so that
mystics would be freed for use at the end of the last line. But aim P3 was
abandoned when the word typicai, serendipitously encountered during a more
or less random search, suggested a different way of achieving aim P2.
(Another possibility in such a case, though it did not happen in our example,
is for a superordinate goal to be forgotten as a subordinate one is pursued.)
TRIAL AND ERROR. The seeking of any goal, overriding or subsidiary, is a
matter of trial and error. Such behavior is always in part random, though also
always guided by previous experience.
For example, how does one go about finding a usable rhyme for a given
word? — say, for the word rhyme itself? In English, two words form an exact
rhyme if and only if their norm shapes are identical in sound from the
stressed vowel to the end (in the present instance -ime), but different before
that. Some of the most effective rhymes, such as Ira Gershwin's gospel : :
poss'ble or Tuesday : good nets day, are not exact, or are rendered so only
by distortion, or involve phrases rather than individual words; but taking these
9. THE CRAFT OF SPEAKING 127

into consideration here would make our exegesis unnecessarily complicated,

so we set them aside. Given a word to be rhymed, there are various ways of
conducting the search. The least systematic is just to see if a word comes
to mind; that way leaves the greatest amount to chance. Or one may go
through the alphabet (starting with zero), trying each consonant before the
nucleus: I', bime, cime, etc. Those of us with formal knowledge of the English
sound system are likely, instead, to go through the attested initial conso-
nantisms: *pime, prime, *piime, *spime, *sprime, *splime, time, ..., *sfíme,
*sfrime, ... . Or one can consult a rhyming dictionary, a procedure that differs
from those already mentioned only in that it uses a public external record
instead of private internal ones.
ANALOGY AND CHANCE. Any of the foregoing operations in seeking a rhyme
is a way of being guided by previous experience (individual or, in the case of
the rhyming dictionary, collective). Recognizing that a word-shape so dis­
covered is an actual word also depends on previous experience, as does
discerning whether such an actual word is usable. In a very general way, this
kind of guidance by previous experience is what linguists have traditionally
called ANALOGY or ANALOGICAL CREATION. SO we might describe the process as
ANALOGICALLY CONDITIONED TRIAL AND ERROR. However, note that even here
chance plays a part, in that it helped shape both one's previous experience
and what one's memory retains of it.
EDITING. Trial-and-error situations vary in all the respects mentioned in
§9.4: the urgency with which a choice must be made, the number of alter­
natives from which one can choose, their relative likelihoods of being chosen,
and so on. When conditions allow a degree of deliberateness, much of the
process constitutes what I like to call EDITING: moving back and forth, and up
and down the hierarchy, pruning, replacing, and recasting (see §§ 4.4 and
9.7).
At the opposite extreme, AUTOMATIC RESPONSE (as we saw in §9.4) is simply
what happens in a special limiting trial-and-error situation: to wit, one where
there is only a single possibility to try and it is chosen with no delay.
Going through the alphabet (or through a memorized table of initial conso-
nantisms) involves responses that are automatic or nearly so. We memorize —
in fact, we overlearn — the order of letters in the alphabet. Since we have
done that, A reminds us of B, B of C, and so on. Whatever else A and B may
be, as the names of letters they are WORDS, and words have connotations.
That B follows A in the traditional alphabetical order is a special and very
strong syntagmatic connotation of each (surely A suggests B as strongly as
hermetically suggests seated). Depending on the situation, various of the
connotations of any word may more or less quickly and automatically leap to
mind when one hears or thinks of the word, and all such resonances are grist
for the trial-and-error mill.
128 9. THE CRAFT OF SPEAKING

REARRANGEMENT. At one point in the construction of our limerick we

considered the phrase a beginning student of linguistics. Later this was
amended to a student beginning linguistics. Neither of these survived in our
final version, but this step in the editing is the one relevant here. Although
the functors are different (an of being dropped), the change from the first to
the second is essentially a PERMUTATION: not a change of contentives but just
a juggling of the order of those already chosen. And although some of the
meaning is altered, a good deal of it is the same for both phrases.
Planning an utterance involves not only choosing words but also settling on
the linear sequence in which they will be uttered. Either sort of choice can be
tentative: a possibility is tried, found wanting, and replaced by another, as
the speaker seeks both the best words and their best arrangement. The
actual order of delivery, however effective or ineffective, reflects the
speaker's final decisions, a fact which correlates with our insistence in §1.2
that order is never nondistinctive.
It is here, in the speaker's procedure of programming utterances, that the
notion of transformations arises, and it is in this setting, and ONLY here, that
transformations have any objective reality.
In the original formal descriptive sense, a transformation is an ordered pair
of syntactical patterns which share certain connections: that is, if the
constituents in an expression that conforms to one of the patterns are
permuted to conform to the other, some things change, but certain basic
syntactic connections remain unaltered. Since our limerick-constructing
lacks any good illustration of this we draw examples from elsewhere. Consider
the following (adapted from Halliday 1985, p. 38):
The duke gave my aunt that teapot
The duke gave that teapot to my aunt.
That teapot the duke gave my aunt.
My aunt vas given that teapot by the duke.
That teapot vas given my aunt by the duke.
That teapot vas given to my aunt by the duke.
That teapot vas given by the duke to my aunt.
Any ordered pair of these seven sentences illustrates a transformation in
the formal sense just described. Although many features of emphasis and
highlighting differ from one to another, and although further subtle modifi­
cations of meaning can be achieved by varying intonation and stress-
placement, in all seven sentences give is the action, the duke is the actor, my
aunt is the recipient, and that teapot is the goal.
In the course of constructing an utterance about dukes, aunts, and tea-
pots, a speaker might first hit on any one of these, and then replace ¡t by one
9. THE CRAFT OF SPEAKING 129

of the others in order to modify the emphasis while maintaining the basic
noun-verb connections and thus also the meanings of those connections.
When that happens in the editing process, we have transformations in action.
Note that there is no reason to assume that the speaker always starts with
some simplest or most "natural" pattern and can get the others only as
transforms. Selecting one as basic or "underlying" is a convenience for
description but implies nothing about speakers' strategies.
Actually, we are not even entitled to assume that the planning process is
linear —one decision at a time —in the sense of §1.4. In §7.5 we insisted that
the linearity of speech implies no similar limited dimensionality in the
interpretation and understanding of a hearer, who may (and probably does)
process input in several ways at once. And now we must propose the same
nonlinearity for planning, (In the extended example of §9.5, true enough, the
creative process was presented as a series of steps. But we had no way of
describing it except in words, and the apparent linearity was an artifact of
the machinery of description, not a property of what was being described.) It
is in no way unrealistic to propose that in planning what is to be said a
speaker may simultaneously encounter several possibilities, consider several
aims, or make several compatible choices. So, for all we know, it is even
possible that in designing an utterance about teapots, aunts, and dukes a
speaker might first contemplate the action and the participants in some
multidimensional arrangement — after all, the actual event was multidimen-
sional! — and then choose among the alternatives given above as part of the
necessary linearization for production.
Be that as it may, it must be emphasized that for the hearer this is all
irrelevant. Only occasionally does an utterance as performed offer any clue
as to the order in which it was planned. It follows that THE HEARER DOES NOT
HAVE TO KNOW THE PLANNING SEQUENCE IN ORDER TO UNDERSTAND AN UTTERANCE, A
hearer is not obliged to work backwards through the speaker's tangled
process of creation in order to understand; it suffices for the actually
produced utterance to be parsed and construed along the lines set forth in
chapters 1-8.
9.7. Generalization. It must be acknowledged that the extended example of
§9.5 is only one sort of verbal behavior. If we consider other sorts, both
simpler and more complicated, do we find additional features and mech-
anisms? Or do we find just the same ones, in different balances and ratios?
Our discussion in §9.6 implies the second answer, and I think that is right.
Consider first a task as relatively minuscule as creating a sentence. To
make up a limerick I must have grist for the mill, and I must know what a
limerick is so that I will recognize success if I achieve it. Just so, to generate
a sentence I must have access to possible materials — words and phrases —
and to ways of arranging them, and I must know what a sentence is. To
130 9 THE CRAFT OF SPEAKING

produce a sentence that conveys a certain meaning I must, in addition, be

able to recognize when the words I choose and arrange express that meaning
rather than some other. Unless an automatic response suffices, I then work by
analogically conditioned trial and error toward the goal.
As evidence, note that in our everyday verbal interchanges we constantly
hear signs of OVERT editing: hems and haws, hesitations, pause-fillers like uh
and wull, abandoned false starts, "self-critical" comments such as or I shouid
say ..., and so on. To be sure, performance is sometimes smooth and free-
flowing. But we take that to mean either that the editing was done earlier so
that the speaker is delivering a memorized discourse, or else that the
current editing is covert, carried on silently within the speaker.
Turn now in the opposite direction, to consider more ambitious projects. I
can detect no essential difference between the process by which our limerick
was created and that by which I write an article, a monograph, or a textbook.
Of course the scale is different: the overarching aim is larger and there are
therefore more hierarchical layers of purposes. But in the pursuit of any one
purpose the pattern of development conforms just to the paradigm we have
given.
I do not even think our schema is confined to verbal behavior (recall the
end of §9.1). I can vouch from personal experience that it applies in toto to
musical composition. Clearly it holds exactly for the efforts of a mathema­
tician, or of successive generations of mathematicians, to prove a conjec­
ture and thus turn it into a theorem, or to disprove it so that it can be
discarded: one must know what constitutes a proof, just as in our illustration
we had to know what constitutes a limerick, and one is then guided by
relevant mathematical experience in the trial-and-error process of devising
candidates for proofhood and in testing to see if they hold. Both of these
examples might be considered relatively close kin of verbal behavior, but I
suspect that our paradigm is in fact completely general, found in every sort
of human goal-directed activity.
 NOTES AND COMMENTARY
References to the text are by chapter, section, and page. Cross-references
within the notes end with " n " . External references are by author and year;
numbers separated from a preceding year by a period (e.g., 1958.43-4) refer
to pages.

0. Introduction.
§0.1 ρ 1. Some of the wording used here (and originally drafted for use
here) was borrowed and paraphrased in Hockett 1984.
Many of the views presented in this essay supersede opinions expressed
by me in earlier publications. The more drastic cases are pointed out in
these notes.
"Eclipsing stance": Voegelin and Voegelin 1963.
§0.2 p 1. The "logic" view: Bar-Hillel 1953; Chomsky 1975 (written twenty
years earlier) points in this direction, even though Chomsky 1955 is quite
critical of Bar-Hillel; Strawson 1974; Katz 1981; very clearly in Postal 1986.1.
The empirical view: outstandingly Bloomfield, but also Boas, Sapir, and many
others — indeed, the whole nineteenth-century tradition of historical-compar­
ative linguistics.
Whitney 1872 is a critique of Steinthal 1871.
Diller (1971) tries to show that the difference between these two views of
foundations is just the traditional philosophical distinction between ration­
alism and empiricism. Surely we can use those labels as a matter of
convenience, but the twentieth-century difference, or even the nineteenth-
century one as illustrated by Whitney and Steinthal, is not to be identified
indiscriminately with that of earlier times. I think Diller's treatment is greatly
oversimplified. Yet I agree with his conclusion that the issue is essentially
unarguable — one can "disprove" either orientation if and only if one assumes
the other.
Given the "logic" orientation, Chomsky's basic assumption of the indepen­
dence of "grammar" (in his broad sense, for which see below, §0.3 ρ 3 η)
from probability and semantics may be valid. Given the orientation of this
essay, although I have always found his assumption tempting I think it is
unwarranted. This is part of the issue I argued in 1968, concluding that,
contrary to Chomsky's proposal, a language is not "well defined" in the
logico-mathematical sense. Particularly, syntax and semantics seem to blur
into each other inextricably (see §8.1 and §8.1 ρ 97 η).
132 NOTES AND COMMENTARY

§0.2 ρ 2. Chomsky's suggestion of "user" in place of "speaker": around

1960, since it occurs in the title of Miller and Chomsky 1963.
Considerations of hearer's point of view: incidental in many earlier works
(for example, in the passage from Jespersen quoted below in §2.2 ρ 18 η);
very explicit in chapter 11 of the second edition of Gleason (1961; not in the
1955 first edition), which also anticipates the approach taken in our chapter
3; insisted on in principle in Hall 1968.82, but without special follow-up; im­
plicit in a key paragraph (p. 72) of Hockett 1966; highlighted, for phonology, in
Wells 1947a, Hockett 1976 and 1980; highlighted, for grammar, in Hockett
1960a. Especially apposite is Francis 1983.16-9 (including a passage quoted
below in §3.2 ρ 26 n). Chapters 11 and 12 of Miller 1981 are extremely helpful
on the different roles of speaker and hearer and on conversational inter­
change. In a number of recent papers (1976, 1980a, 1980b, 1986), Straight
insists on the theoretical importance of the distinction between reception and
production. See also the collection of articles in McGregor, ed, 1986.
From a different angle, Yngve has proposed (1983, 1984, 1985) that we
should be trying to describe language-users rather than languages. I do not
fully grasp his argument, which involves an unfamiliar use of the word "gram­
mar," but suspect that he will find the change of emphasis developed in this
essay compatible, at least, with the approach he is recommending.
Our fundamental question: likewise, Chomsky (1962.57) asks "When you
hear a sentence, how do you interpret it?"
With the point of view of the hearer, adopted here, one should compare
certain other attested orientations. Nineteenth-century descriptive linguis­
tics, for example, was directed mostly toward language-learners, be they
schoolchildren contending with Latin, scholars delving into Sanskrit, or mis­
sionaries struggling with Chinese or Choctaw. American theory in the first
half of the twentieth century, on the other hand, was operational, oriented
toward the observer-analyst, who wants to know what observations to make
in the field and how to interpret the data. Although the leaders in this
orientation were Boas, Sapir, and Bloomfield, none of those three had much to
say about either field procedure or analytic technique (Boas 1911 is a partial
exception, Bloomfield 1942b a very minor one). The overt discussion of those
matters was left mostly to the post-Bloomfieldians (see below) of 1935-1960.
Twaddell (1983, especially pp 38-39), with some justification, lumps together
the audiences for these first two orientations. Early transformational-
generative theorists rejected the operational approach, and claimed also to
be neutral between speaker and hearer, but, as we shall see later, speaker-
orientation did become dominant, especially in so-called "generative pho­
nology." Some more recent followers or ex-followers of Chomsky (for
example, Postal 1986.1) treat a language merely as a collection of sentences
viewed as an abstract object, and disavow any implications at all for the way
NOTES AND COMMENTARY 133

the language is actually used. (For this comparison of theoretical orientations

see also §7.3 ρ 81 η.)
I use "post-Bloomfieldian," in the face of Hymes's objection (Hymes and
Fought 1975.7 η 3) that it is inept, for the generation of American linguists
who followed on Sapir and Bloomfield and were avowedly influenced more by
those two than by any other predecessors. Those who agree with Hymes are
free to replace the term, say by "neo-Bloomfieldian," wherever it occurs. The
work of these investigators, whatever we call them, is sometimes called
"American structuralism." I prefer to avoid that label, because I have never
been sure just what structuralism is supposed to be (Hymes and Fought, in
the work just referred to, assiduously avoid telling us) —unless it is just a
fancy way of referring to the twentieth-century emphasis (promoted
especially, though not exclusively, by Saussure) on system and pattern in
contrast to the somewhat atomistic nineteenth-century approach.
Psychology: the linguist who became most insistent on the independence
of linguistics from any sort of psychology was, of course, Bloomfield, who
sought (1926) to replace any reliance on psychology by a set of overt as­
sumptions (see also the characterization in Wells 1963). We would have to say
today that such investigators as Katz and Postal (see references in §0.2 ρ 1
η) go much farther than Bloomfield ever did in depsychologizing their dis­
cussion of language, but in a way very different from Bloomfield's. On Bloom-
field's procedure, the obituary comment of Sturtevant (1949) is interesting:
[His 1914) book was frankly based upon the mentalistic psychology of Wilhelm Wundt,
vhile the [1933] book adopts the behaviorist point of view ... . I t is important to note
that Bloomfield has far less to say about psychology in the second book than in the first;
he does not any longer base his treatment of language upon psychology of any particular
school. One may even suspect that if he had treated the material a third time he might
have declared the logical dependence of psychology upon linguistica
Although we are not doing in the text what this last sentence says, we are
surely insisting that, quite apart from anything else to which psychologists
choose to devote their attention, the psychology of LANGUAGE must be
founded on sound linguistics.
On the practical independence of analytic and descriptive technique from
psychology, consider such superb recent treatises as Bloomfield on Menom­
inee (1962), Grevisse on French (1964), Quirk and others on English (1985),
Chao on Chinese (1968a). In my view the last of these is one of the finest
descriptions ever written. It is also the most explicitly Bloomfieldian, with few
methodological modifications by Chao and very few accommodations to the
"refinements" of the post-Bloomfieldians.
§0.3 ρ 3. "WHERE is deep structure?" — compare Hall's somewhat broader
question (1985) about the locus existendi of language.
134 NOTES AND COMMENTARY

"Morphology": I think we can count this term as traditional in the sense

given in the text, although it was introduced into linguistics only in the 1860s,
by Schleicher, who got it from botany; before that, and by many grammarians
even after that, the structure of words was subsumed under two separate
headings, "accidence" (or "inflection") and "word-formation."
On terminology, it is especially important not to be confused by the radical
changes of meaning of basic terms introduced by Chomsky. The following
table of approximate equivalences should help. The traditional senses given
on the left are those intended throughout the present essay. Wherever, in
either column, a single term is regularly used in more than one sense, I have
appended subscripts:
TRADITIONAL CHOMSKYAN

a language (langne) a gammar3

set of all possible sentences in a language a language
a grammar2 = a description a grammar2
morphophonemics and phonemics phonology
morphophonemic symbols, morphophonemes systematic phonemes
phonemics = phonology systematic phonetics
phonemes surface phonemes
gramma1 = syntax + morphology syntax(?)
syntax syntax
morphology morphology(?) (partly in phonology)
clause sentence
lexicon lexicon
An example of how misleading the resulting conflict of definitions can be
will be found in Anderson 1985.6. By interpreting the terminology of earlier
investigators as though they meant, by such words as "language," what
Chomsky chose to mean by them, Anderson gives a completely false picture
of their orientation and aims. They were, in fact, uniformly interested in
LANGUE (SYSTEM, COMPETENCE), not in PAROLE (MANIFESTATION, PERFORMANCE;
see §8.3 p 104 n). Anderson's faulty remarks here are especially surprising
because in much of his ensuing examination of predecessors like Saussure
and Baudouin his scholarship is impeccable.
1. The Shape of Speech.
§1.1 p 4. "Dimension" is used here neither metaphorically (as in a mys­
terious "hidden dimension") nor as a high-toned synonym for "aspect" (as in
Jankowsky's book title Scientifìc and humanistic dimensions of language,
1985), but in its strict mathematical sense of linear independence. It takes
three numbers to specify the location of an airplane, one for longitude, one
for latitude, and one for altitude; none of these three is determined by the
others; air space is therefore three-dimensional.
NOTES AND COMMENTARY 135

§1.3 ρ 5. Spacing: the fifth example (a paraphrase of a French original in

one of Jules Verne's novels) is an excellent demonstration of the commu­
nicative importance (in English) of the serial comma, regularly omitted in
journalistic and legal written English and in most printed material in the
languages of continental western Europe. The final spacing example is
reworked from a verse in Espy 1975.18-9 ascribed to Mary Youngquist.
§1.4 ρ 6. My attention was called to the dimensionality of speech especially
by Stokoe's discussion of American Sign Language for the Deaf (Stokoe 1960,
1966; and see Hockett 1979.274-5).
Although the linearity of speech is obvious, few have considered either how
it is achieved or its consequences. Aarsleff (1982.146-224, esp. 157-8) re­
ports that several eighteenth-century French philosophes, beginning with
Condillac, emphasized the linearity of speech as an important factor forcing
people to "organize their thoughts" (my wording). Linearity is again high­
lighted by Saussure at the beginning of the present century (1916, part 1
chapter 1 §3, part 3 chapter 5 §2, and elsewhere). In 1951 Lashley, in a
discussion of the possible neurological basis of serial order, reasserts
Condillac's claim. So does Vygotsky, when he says [1934(1962).44] that the
child's thinking becomes rational only as it comes to be tied to language.
I think the Condillac-Lashley-Vygotsky proposal is questionable. As we shall
see in the sequel, the linearity of speech itself does not obviously imply
comparable linearity either in the processing of heard speech by a receiver
(§7.5) or in the internal preparation by a transmitter of what is to be said
(§9.5). Reasoning in words, like anything else in words, has to be linear, but
that need imply no linear constraint on reasonable THINKING. What is more,
are we to declare that data-processing by the deaf, using Sign Language,
cannot be reasonable just because the dimensionality of Sign is greater than
that of speech?
§1.4 ρ 7. Last paragraph: the reference is to American football. The offi­
cials now use an intercom and announce the penalties over the P.A. system,
but formerly the hand signals were all the fans got.
There is a remarkable exploitation of two-dimensional syntax in the shape
and placement of constituent figures in Michelangelo's fresco of "The Last
Judgment" in the Sistine Chapel, analyzed in detail by Steinberg (1980). The
direct representational (iconic) features need no comment. But Steinberg
shows that various additional meanings are conveyed by the arrangement of
figures along straight lines intersecting at various angles, each line brought
to the viewer's attention by an outstretched arm or leg, a pointing finger, or
the like. Each alignment joins characters or events related in Christian
tradition or in the social scene of Michelangelo's times. These cross-cutting
connections, registered largely out of awareness, surely have much to do with
136 NOTES AND COMMENTARY

the phenomenal impact of the fresco, and similar subtle devices doubtless
account for the power of many other great works of visual art.
§1.4 p 8. In the published version of Atice Carroll replaced this poem by a
much better one (though the typeset tail is less effective).
§1.4 ρ 9. Of course the envelope is a fake.
§1.7 ρ 10. See also below, §6.5. The iconicity of the shape of a form used
as a name for the form —what Quine (1940.23-6) calls the MENTION of a word
as over against its USE—was called to my attention by Ashok Keikar.
§1.8 ρ 12. Rilchiam etc.: Carroll's explication is in the Preface to The
Huntingof the Snark.
In what sense could we propose that uttering either a blend like błat or a
coinage like felix is "saying both words at once"? Of course the production of
a word takes time: the constituent sounds are consecutive, not simultaneous,
and so are the constituent morphemes of a "polymorphemic" word. Mulder and
Hervey (1980.123-4) nevertheless propose that the constituent meaningful
parts of a word can be thought of as functionally simultaneous in that there
is no choice as to order (or, if there is, as in matchbook versus bookmatch,
one can wonder if one is really dealing with single words).
§1.9 ρ 13. Recently, in the work of a very competent contemporary writer, I
even found "This synthesis enables the animal to, so to speak, build together
perceptions which are ..." — an infinitive split by an infinitive!
§1.10 ρ 13. With the initial remark compare Tesnière 1953.1-2; and see also
§8.6 p 111 n.
The rest of this section may seem like horseplay, but it helps make a point
already touched on in §1.4 ρ 6 n, and due to be underscored in §7.5 and §9.5.
2. Hearing Utterances.
Hearing and listening are, of course, not the same thing: one can hear
something, say an unexpected loud noise, without listening, and one can listen
without hearing, as when one's interlocutor responds to a question with
silence (on this, McGregor 1986.xix n 1 is only partly right). Whenever the
distinction between hearer and listener is relevant I have been careful to
choose the proper term, but in many passages, where either would do, I have
used whichever seemed to sound better. Note also that neither hearer nor
listener should be confused with ADDRESSEE, whose identity depends on the
speaker and is sometimes imaginary.
§2.1 ρ 16. The Necker cube was first used for the clarification of linguistic
patterning, as far as I know, in Hockett 1954 (imitated without credit by Miller
1981.131). The contrasting pair "surface and deep grammar" was introduced in
Hockett 1958 chapter 29, but the meaning given there to the second term
was somewhat different both from that later assigned to "deep structure" by
NOTES AND COMMENTARY 137

Chomsky (1964.10) and from that for which I am now introducing the more
explicit label "structure-in-depth." The word "depth" turns up in various other
contexts, not necessarily very closely related to our use here: e.g., in Yngve's
"depth hypothesis" (1961), or Mulder and Hervey 1980.156, "... not to present
the structure in all depth."
§2.1 p 17. Old men and women: Wells 1947b.
§2.2 p 17. Joos 1950. Excerpts from p. 702 col. 2 and p. 703 col. 2.
§2.2 p 18. Parsing and construing: Compare the following passage from
Jespersen (1922 §10.2.3):
What means does the English language possess to enable the man in the street, who is
no grammarian and has no need of learned terms like subject and object, to understand the
meaning of sentences? If he hears a sentence like "John saw Henry," it is, of course,
necessary for him at once to know who was the seer and who was seen. How is this
effected?
§2.3 ρ 18. On Fijian see Schütz 1985 (my view of the language, and hence
the terminology used here, differ slightly from his).
§2.5 ρ 21. Ambiguity and analysis: Woolley 1977b includes a brief review. He
cites many of the examples I have chosen to use here, including How atcut a
drink? (end of our §2.5), which apparently occurred independently to Adam
Makkai (to whom Woolley gives credit) and to me. Woolley's concern is with
the extent to which expressions that seem structurally ambiguous in citation
form can be disambiguated in actual delivery, a matter of considerable
theoretical importance since it suggests that, if we really take the hearer's
point of view, our phonemic analyses are incomplete. None of that conflicts
with what we are saying here, since our point is only that many utterances of
the sort are often NOT disambiguated except by context.
§2.5 pp 21-23. Examples: Pike's appears in 1982.119 (but he used it years
ago); Nida's is from memory. Chomsky's are from 1957a.
§2.5 ρ 22. In the second interpretation of stout major's wife there is a
problem as to what to do with 's. One possibility, not usually provided for by
the post-Bloomfieldians but suggested in Hockett 1960a §6, is to treat it not
as a constituent at all but as a MARKER of the relation between the con­
stituents stout major and wife. Chao (1968a.261 follows this pattern for a
comparable phenomenon in Chinese (he errs, unfortunately, in identifying it as
"item and process" rather than the more customary "item and arrangement"
approach). Blansett's 1983 title "Non-constituent connectives" obviously
refers to the same idea. Blansett cites not me but Muir 1972.73, indicating
that the notion arose quite independently in the post-Firthian British tradition
of systemic-functional grammar.
138 NOTES AND COMMENTARY

Immediate constituents: Bloomfield 1933.161 (and later passages); Wells

1947b; Pittman 48; Nida 1949; post-Bloomfieldians passim; and see below.
"Construction" is the traditional term for a syntactic relation; a dozen
different contemporary theoretical schools use a dozen different neologisms
for what upon careful inspection turns out to be essentially the same thing.
Especially, it should be noted that a TG "rewrite" rule, such as "S —> NP +
VP," is the statement of a construction; the fact that the type of the
constitute is listed before the types of the constituents, instead of after as
in many other formulations, is in this connection quite irrelevant.
(In passing we must note that some investigators in the post-Bloomfieldian
period — and also more recently — have fallen into the unfortunate habit of
using "construction" not for a syntactic relation but for a form built by such
a relation, properly called a COMPOSITE FORM, a CONSTITUTE, a CONSTRUCT, or,
following Saussure, a SYNTAGM.)
The terms "subject" and "object" are also (of course] traditional.
"Actor" and "goal" are Bloomfield's (e.g., 1933.173, describing the relation,
in Latin, of the subject respectively to an active or to a passive verb). Many
recent treatments of noun-verb syntax (following Fillmore 1968) use "agent"
where I use "actor"; I prefer to reserve that term for the subject in a clause
like Daddy's burping the baby,in which an intransitive verb is used causatively
and the actor (which would be the subject of the verb in noncausative use) is
the object.
Immediate constituents again: in the 1940s, 1950s, and early 1960s it was
sometimes assumed, by some grammarians, that immediate constituents
must be ADJACENT —so that, for example, in Are you going? the two words Are
and going could not be taken as belonging together. It was also sometimes
supposed (not necessarily by the same investigators) that a word or phrase
can participate in only one construction at a time, so that if, in John /oves
Mary, /oves and Mary stand in a verb-object construction, then John cannot
stand in a subject-verb construction with /oves, but, instead, must stand in a
subject-predicate construction in which the second constituent is the whole
phrase loves Mary Finally, there was a common working assumption that as
far as possible one should recognize only BINARY constructions.
Some of the most carefully wrought criticism of the period (e.g., Postal
1964) proves, on equally careful exegesis, to have been directed largely at
such superficialities as the three notions just mentioned. That doesn't mean
the criticism was useless; indeed, it helped us to recognize just how
ineffective some of our tentative working assumptions were. Yet American
"structural" (or "taxonomic") grammatical theory was by no means the
monolith it was made out to be in some of this criticism. (As evidence for the
variety and the experimentation, see my own 1945 treatment of Chinese
syntax.) Each of the three proposals described above was also rejected by
NOTES AND COMMENTARY 139

many, or was assumed tentatively only to explore what the consequences

would be. None of the three, of course, is accepted in this essay.
All three are rejected also in much contemporary grammatical analysis
outside the TG tradition. Consider, for example, Halliday's (1985) realistic
"polystructuralism" (my term for it), which recognizes several simultane­
ously functioning sets of structural relationships among the elements of a
sentence — and hearers can detect and react to all of them at the same time,
processing them in parallel (see §7.5 pp 85-6).
There was also a strong (though not explicit) tendency in the period men­
tioned above, among both the post-Bloomfieldians and their critics, to assume
a very restricted number of different constructions. In my 1958 Course I
fudged on this issue by speaking only of a relatively few "construction-
TYPES," evading the responsibility of presenting a detailed list of distinct
constructions (and Postal 1964 called me on it). Likewise, a TG rewrite rule of
the sort illustrated a few paragraphs back tended in practice, if not in
theoretical necessity, to lump together highly diverse composite forms;
structural features at the "underlying" level were few, much being left to be
taken care of by transformations. We shall see in the sequel (§8.1) that the
number of constructions in a language is in fact neither small nor large but
INDEFINITE, since we deal not with identities, only with similarities.
Despite the diversity of views and experiments of the post-Bloomfieldians,
the asides about TG in the last few paragraphs could nevertheless be
summarized reasonably well by saying that (from our empirical point of view)
the Chomskyan diagnosis was right but the proposed cure wrong.
§2.5 ρ 23. Quote from Hockett 1979.275.
It was "connections at a distance" (e.g., between antecedent and sub­
stitute), of the sort illustrated at the bottom of page 23, that suggested the
original notion of "deep grammar," back in the 1950s (see above, §2.1 ρ 16 η).
§2.5 ρ 24. Some further examples worth thinking about:
That shirt was made to wear out
I want to sit outdoors behind her house and vatch her garden.
Yesterday I saw a chipmunk sitting in the patio.
We tried to find more edible apples.
You can't eat too much spinach.
He shot the girl in the park.

3. The Hearer's Evidence.

§3.2 ρ 26. Consider here, again, the remarks quoted from Jespersen above,
§2.2 ρ 18 n. Also note the following passage from Francis 1983.18, of which
the bulk of the present chapter could be regarded as an expansion:
140 NOTES AND COMMENTARY

... the receiver of a linguistic message is not a passive machine reacting only to overt and
explicit items in a code. He is instead a partner in the enterprise, constantly making use of
his ovn experience with variation to reconstruct the message as it comes in, by means of
all kinds of associations, contextual information, and awareness of the situation and of the
idiosyncrasies of the speaker. This adaptability helps make communication possible in spite
of the fact that the surface [sic] codes of speaker and listener may differ in many ways
Smith used the term "communicative package" in lectures and discussions;
I do not find it in his published articles.
Paralanguage: Trager 1958.
§3.2 ρ 27. Kinesics: Birdwhistell 1952. On both paralanguage and kinesics,
Key 1975 is an excellent survey with a very full bibliography. Smith and
Trager assumed sharp boundaries between language and these other commu­
nicative modalities, as did I in the 1950s; but I now think, as Key seems to,
that the lines of demarcation we drew at that time were the product of
wishful thinking (see also below).
§3.3 ρ 27. Stress and intonation: I should perhaps distinguish at this point
between "word stress," which is a matter of word identity, and "sentence
stress," which is part of intonation, or prosody; the difference emerges in the
course of the discussion. ("Prosody" is not to be taken here in its special
Firthian sense.)
"Morpheme" and "moneme": the latter, which André Martinet tells me was
introduced by Henri Frei around 1940, is the better term because it avoids the
potential ambiguity. But so much of the work due to be criticized later in this
essay uses "morpheme" in its American sense that I have felt it better to do
the same.
The Trager and Smith treatment (1951) of American English stress and
intonation, which I followed in my 1958 textbook and in various other pub­
lications of the 1950s and 1960s, is still, in my opinion, the most satisfac­
tory analysis we have. Yet I think it is wrong, mainly by way of overelabora-
tion. The basic issue is whether prosodic features are patterned as are the
sounds and meanings of segmental forms, or by some other logic, and I
suspect the latter (see Ladd 1980): intonation falls in the treacherous
no-man's-land between the clearly linguistic and the clearly paralinguistic.
§3.3 ρ 27-8. Contentive vs. functor: the distinction is old (though not with
these labels). For English it was emphasized especially by . . Fries in his
recognition of "content words" and "function words" (e.g., 1952; see also P. H.
Fries 1983). My terms, used here, were introduced in Hockett 1958 chapter 31:
the replacement of Fries's "function word" by "functor" allowed for the sub-
sumption of bound markers (e.g., inflectional endings) along with particles and
separate words. .  Fries used "Jabberwocky" in 1952 (at the suggestion of
Aileen Traver Kitchin) to illustrate the difference; I used it in 1958.262-4; and
NOTES AND COMMENTARY 141

Chao used it in 1968a.501-3 in Chinese translation to illustrate the same

point for that language. Among Chinese grammarians the distinction (between
"full" and "empty" words) is traditional.
The general notion of the marking of syntactic connections grew, for me,
partly from . . Fries's work, partly from the analysis of Fijian, but mainly
from an investigation of CASE undertaken in the early 1950s (unpublished
paper, condensed to one section, §27.5, in 1958; but see McKaughan 1958 and
l'čuk 1986). I took cases not as noun roles relative to a verb (as Fillmore
1968, following Jakobson and Hjelmslev, defines them, in violation of
Bloomfield's fundamental principle of working from form to meaning), but
rather as formal MARKERS of such roles within nouns or noun phrases. The
first enlargement of my original notion was the recognition (inspired by
McKaughan's field findings) that such marking could be achieved by particles
as well as by inflectional elements. The second expansion was the realization
that the marking of noun-verb relations in the verb or verb phrase, comple­
mentary to that in the noun, is VOICE (defined broadly to cover also types of
transitivity). All of that — the syntax of case and voice — can be viewed as a
proper part of the syntactical marking under discussion in this chapter.
§3.3 ρ 28. "Jabberwocky": see, again, Carroll's explanation in the Preface to
The Hunting of the Snark.
§3.5 ρ 30. The Rex Stout example: Halliday's analysis of the distinction
(1985.15-17), which I saw after drafting this passage, is more thoroughgoing.
"Topic" and "comment": the first appearance of these in print is in Chao
1955 §9. I find "topic," without "comment," in my 1953-4 unpublished notes
on Georgian grammar. One of the two of us devised the terms in the early
1950s, and the other borrowed them (probably at the Linguistic Institute at
Indiana University in the summer of 1952), but thirty years later neither of us
could remember which did which. Halliday uses other terms.
§3.5 ρ 31. In the English stress-contrast examples I have ignored the
Trager-Smith distinction (1951) between reduced primary and secondary
stress (e.g., the White house 'house belonging to the White family' has
reduced primary on house, whereas the WNte House 'president's residence'
has secondary). I hear the difference quite clearly when presented with
minimal pairs, but am often uncertain otherwise, and many others report the
same trouble. In any case, this further distinction is not needed to support
the points being made.
§3.5 p 32. Announcerese: Bolinger 1982.
§3.6 pp 32 f f . All the headline examples except the first are from "The
Lower Case," a regular feature of the Columbia Journalism Review. The first
was a favorite of .  Fries.
§3.8 p 35. Further evidence for the relative communicative unimportance of
functors is that in many languages they seem to cluster on a remarkably
142 NOTES AND COMMENTARY

restricted portion of the sound system. In English, for example, a word-final

s-sound has five different functions in Mark's hat, Mark's here, Mark's seen it,
tad marks, and X marks the spot, a word-final d-sound has four different
values in Hed been there, He'd go on a moment's notice, He judged the contest,
and He's fudged the contest In Fijian, the preposed particle e has the
three-way ambiguity described in §2.3 (p. 19); na ~ a is the common article,
while na marks future time and a marks past;  ~  is the proper article
and the second person singular pronoun. In Chinese, atonic de is typically
described as having from three to five different functions, and atonic te two
or three. With this sort of homophony, it is not surprising that different
investigators often disagree as to how many distinct functors should be
recognized.
4. Hearing Words.
§4.1 ρ 37. That listening in general (not just to speech) is a positive
activity rather than just a passive experience has long been clear; see, for
example, the discussion in Hockett 1975.73-75. Yet I have searched through
the literature in vain for any hints of the distinctions of modes of listening
recognized here (Cowan 1965 comes the closest). But I must have missed
things; the matter is so obvious and so important that I cannot believe I am
here presenting it for the first time.
'Idiomatic expression" (or "idiomatic phrase") is awkward; we might do well
to adopt SYNTHEMEfrom French (Martinet 1980).
§4.2 ρ 38. The overall coloring of a dialect: what is responsible for it? I
suspect two factors in particular: the base of articulation (Sievers 1876;
Hockett 1976.85-6), which can give a tinge to both vowels and consonants,
and socially preferred voice quality. Both of these vary from community to
community just as does the articulation of individual speech sounds.
§4.4 p 39. Implicit motor-matching: Hockett 1976 (which anticipates a
number of the points made in this chapter and the next).
§4.5 p 40. On gestalt or pattern ("structuralism") versus atomism, see
Hockett 1983 §§10-12.
It is unfortunate that so many investigators, in both linguistics and
adjacent fields, have fallen into the habit of saying "structure" where the
proper term is "pattern": a sentence, or any other event (or thing) HAS a
structure; that structure may CONFORM to a pattern. Structure is in behavior
and belongs with parole, pattern is in culture and belongs with langue (see
§8.3 ρ 104 η).
I have never found it easy to follow Pike's discussion of "particle, wave, and
field" (in 1959 and various subsequent papers, including 1982), although
Halliday's reference to it (1985.169) helps. Pike may be making the same
distinction with these three terms that is usually made just by contrasting
NOTES AND COMMENTARY 143

element and arrangement; Pike's "wave" is then a gestalt in time, while his
"field" is one in space. Even if that is wrong, it seems to me Pike and I are
both trying to balance an overemphasis on things by paying more attention to
arrangements and interrelations.
My first published allusion to gestalt is in Hockett 1975.
§4.5 ρ 42. For redundancy in greater detail see §5.5 pp 60-4.
The aphorism: Cowan 1965.
Chao reported this story from memory twice, with slightly differing detail; I
cite the version given in 1968b.170.
Garden-path jokes: Hockett 1973.160.
§4.7 ρ 44. We insist here that the shape of a word is not to be identified
with the word itself. Then what is the word? As we shall see, partly in §5.2,
partly in §8.6, it is a sign in the Saussurean sense, and the shape is only one
part of the bundle.
On modes of listening, see again §4.1 ρ 37 η.
Most discussion of phonemic theory of the second quarter of this century
reads as though everyone believed the only kind of listening is for word
shape. That was particularly true of American theory, oriented as it was
toward the practical problem of "solving" the phonemic system of a language
encountered in the field. My realization that we don't usually listen that way —
that usually we are listening only for word-IDENTiTY — came very slowly. I
remember, forty years ago, being struck by a well turned remark of Chao in
defense of his use of his National Romanization of Chinese for teaching the
language. In that system, instead of indicating tones with diacritics (or —
horrors! — leaving them unmarked) one spells a syllable in different ways for
different tones: for example, instead of fēi, féi, fĕi, fèi one writes fei, fair,
feei, fey. In support of this practice Chao said (1948.11, emphasis added by
me): "This makes the spelling more complicated, but gives AN INDIVIDUALITY
TO THE PHYSIOGNOMY OF WORDS, with which it is possible to associate meaning
in a way not possible in the case of forms with tone-signs added as an
afterthought." I did not at the time approve of Chao's choice. It seemed to me
his argument could be used equally to defend Chinese characters, or
traditional English orthography with all its irregularities, as over against
phonemic transcription, and our reasoning at the time said that if the sound
of speech is enough for a listener then a phonemic transcription is enough
for a reader. Pedagogically that may be true. But Chao's phrase stuck in my
memory just the same, and its emphasis on the perceptual unity and integrity
of the word obviously ate slowly away at my earlier theoretical predilections.
I have no idea how extensive may be, in the long run, our necessary revision
of phonological theory in terms of gestalt perception. Most of my earlier
theorizing about phonemics was done in atomistic terms (as suggested at the
beginning of the preceding paragraph); every bit of it now requires rethinking.
There are only minor hints of this in the present essay.
144 NOTES AND COMMENTARY

§4.8 p 46. McDavid told me about his tiger forty years ago.
§4.8 p 47, first full paragraph: Credit for these remarks on onomatopoeia,
and also for much of the treatment of the recrystalization of slurvian given
below in §5.4, goes to Nancy 0. Lurie. She recently called to my attention the
general absence in North American Indian languages of tailored onomatopoetic
words for animal sounds, like our woof, meow, neigh. Instead of such words,
the Indians simply imitate the actual animal sound. However, many American
Indian NAMES of animals (especially of birds) are onomatopoetic in form: in
Algonquian, Fox kākākiwa 'raven', Cree kōhkōhkahōw 'great horned owl', Fox
titīwa 'bluejay' (with t retained before ī instead of being palatalized, as always
in ordinary words).

5. Sounds, Words, and Redundancy.

§5.1 p 48. On the phonemic principle and its history (with more or less at­
tention also to morphophonemics): Trubetzkoy 1938; Hockett 1955; V. Makkai
1972; Fischer-Jørgensen 1975; Kilbury 1976; Jakobson and Waugh 1979;
Hockett 1983.31-4; Anderson 1985. The justly famous footnote in Sapir
1921.56 ends with this summarizing sentence: "In watching my Nootka
interpreter write his language, I often had the curious feeling that he was
transcribing an ideal flow of phonetic elements which he heard, inadequately
from a purely objective standpoint, as the intention of the actual rumble of
speech." There you have it: the hearer's perspective, gestalt perception —
everything from which the present essay has grown.
§5.1 ρ 49. Bloomfield's "confession" about his phonetic ability in the field
was made in my presence, I think in Chicago in the fall of 1939. He meant the
remark only half-jokingly.
Martha's Vineyard:The marvelous report by Labov 1963.
§5.2 ρ 50. Bloomfield's definition of word: the 1926 formulation (Definition 11,
with examples) does not use the expression "immediate constituents," but
that is clearly the intent. A remarkable train of investigators have been
confused about this. Wells's clarification: 1947b.
Note the difference between Bloomfield's definition of a FORM and the
Saussurean definition of a SIGN: A form is a shape that HAS a meaning; a sign
is a two-faced unit consisting of the shape AND the meaning — or, more
accurately, of the mental images of the shape and of the meaning. (We shall
say this rather differently in §8.6). A good deal of misunderstanding between
scholars in the European Saussurean tradition and those in the Bloomfieldian
tradition has resulted from inattention to this difference.
The sons ond daughters ....: Wells 1947b.
The lovely example of the group genitive is direct from the lips of the late
Gracie Allen, in a show of which I saw a replay in the summer of 1986.
NOTES AND COMMENTARY 145

Bazeli (1952) points out that different equally significant criteria often give
slightly different results, (If they give sharply different—virtually orthogonal—
results, then the situation is not the sort that concerns us here.) We cannot
expect, for example, that the stock of elements that qualify as "words" by
one test will necessarily all have the property demanded by another. On the
other hand, if we use several criteria, then, even if they mainly agree, we get
a fuzzy boundary. So our choice is often between the neatness of precise
boundaries and the realism of fuzzy ones, and if we are after the truth we
had better settle for the latter.
§5.2 p 52. "Degrees of cohesion" (spacing): see the lists in Palmer and
Biandford 1924(1939).27-8,92.
§5.3 p 53. Of course, systematic external sandhi, as in Sanskrit, answers
this question in the negative. Our concern is with cases in which no such
obvious alteration of phonetic shape takes place (e.g., English or Spanish,
setting contractions aside).
The debate: Pike 1947, citing earlier literature; Hockett 1949; Pike 1952. A
different approach (still not the one presented here), with allusions to the
debate, in Hockett 1980. The quote: Wells 1947a.270-1.
§5.3 p 54. Demarcative (abgrenzende) and culminative (kulminative): Tru-
betzkoy 1938.29.
The notion of inferring boundaries from segments (going from wholes to
parts) is clearly gestaltist. It is inherent in the notion of phonological
immediate constituents (Hockett 1955.150-4, but tracing back to Pike and in
part to Bloomfield). And it is implied in any linguistic discussion that
concentrates on "groups" or "contours," be they intonational, rhythmic, or of
some other sort, rather than on their constituents. So this notion was behind
Scott's proposal (1948) of the MEASURE in Fijian phonology: a stable rhythmic
unit at a size-level larger than that of the syllable or word but smaller than
that of the phrase. I resisted this proposal for a long time because I didn't
know how it would work, but now believe that Scott was on the right track.
§5.3 p 55. The argument here may remind some readers of my 1949 paper
(referred to above). The solution proposed in that paper stemmed from the
atomistic assumption that segments must be defined by boundaries. Various
people took up the issue and handled it in one way or another, but neither my
original solution nor any of the alternatives proposed by others is satis­
factory. The real solution resides, I now believe, in the language user's ability
to infer boundaries from segments.
The "tightrope" case was actually a little different: I encountered the word
in phonetic transcription, with no mark of the boundary between the members
of the compound, in Palmer and Biandford 1924, and did a double take. Clearly
the phonetic transcription left out something distinctive.
146 NOTES AND COMMENTARY

§5.4 ρ 57. Compare Bloomfield 1933 §9.7 and ρ 388. The content of this
chapter was not included in the Rice University lectures mentioned in the
Preface; this section, and to a lesser extent the rest of the chapter, grew
from penetrating criticisms and brilliant constructive suggestions offered by
James E. Copeland and Philip W. Davis in response to those lectures. Credit is
also due Nancy 0. Lurie, as remarked in §4.9 ρ 47 η.
For the wonderful term "slurvian" I thank the unidentified author of a
Sunday-supplement article of many years ago.
In the 1940s I worried about the proper phonemic analysis of utterances
such as a rapidly delivered Ed hod edited it. At the end of my Manual of
phonology (1955.220-1) I distinguished between a FREOUENCY NORM and a
CLARITY NORM for phonemic analysis and indicated that there was at the time
no clear answer as to which norm is the proper one to use—maybe both, with
different but complementary results. But the hearer's orientation and the
gestalt approach adopted in the present essay lead us to the clarity norm as
the right choice. The recognition of slurvian eliminates any reason to seek a
phoneme-like analysis on a frequency basis. Likewise, it suggests a revision
of Hall's three-way distinction for French (1948.7) of archaic, slow colloquial,
and fast colloquial, though I am not sure just how the revision would run
except that it would doubtless involve a sliding scale, rather than a precise
number, of degrees of clarity.
Note, though, that the work the hearer's internal circuitry must do in order
for words to be "heard" in terms of their expected phonemic structure,
instead of as they actually sound, must play a part, in the long run, in the
gradual alterations of habits of articulation and hearing that constitute sound
change (in the neogrammarian sense of the term).
§5.4 p 58. On have and of: I can remember, as a child in grade school
learning to read and write, how insulted I was on being told that the slurred
third syllable in / would of gone was the word have. I wish my teacher could of
explained why —to this day I think she must of made a mistake.
§5.4 ρ 59. Duality of patterning is Martinet's DOUBLE ARTICULATION (1957).
For a guess about its inception in the evolution of human communication, see
Hockett 1979.275-6, 296-9.
§5.4 ρ 60. Prelanguage: see Hockett 1979.296.
§5.5 ρ 60. Redundancy (communication-theoretical sense): Shannon 1949.
§5.5 ρ 61. Let Bloomfield 1933.398-9.
§5.5 ρ 62. Bee dancing: Frisch 1950.
§5.5 ρ 63. Noise and perpetual motion: the connection is via thermody­
namics. Noise interfering with a signal is the effect of increasing entropy.
NOTES AND COMMENTARY 147

The second law of thermodynamics says that in a closed physical system

entropy always increases. And it is just that law of nature that prohibits the
construction of a perpetual-motion machine.
6. Why Morphemics Won't Work.
The revision of morphological theory developed in this and the following
chapters was anticipated most strongly, even if negatively, in a sort of
appendix (pp 42-3) of my 1961 "stratificational" paper, which I quote verbatim
here (only changing some typographical conventions):
Most of our descriptive thinking in the past few decades has been based on an unstated
assumption: that any utterance in a language, occurring in a specific context involving
specific speaker and hearers, has IN THAT CONTEXT a determinate grammatical struc­
ture, involving an integral number of grammatical elements in specifiable structural rela­
tions vith one another. We have quarreled extensively over the exact nature of the
elements and the relations, about heuristic criteria, and about their status as objective
parts of the universe or as convenientfictions.But the underlying assumption has scarcely
been challenged.
There are, in fact, certain types of utterances that should raise serious doubts about
the assumption. One of them is Don't shell to loud!, something I once said, in angry
irritation, to my noisy children. It is clear that an attempt to deal with this sentence
vithout information about the context would yield erroneous conclusions: the item shell
vas not, or vas not exclusively, the morpheme that is customarily programmed into that
string of phonemes. But if it vas not that morpheme, vhat vas it grammatically? We can
call this shell a "blend" of shout and yell; but no existing system of grammatical analysis
or theory makes provision for the building of a grammatical form by "blending."
Such utterances are not rare, but extremely common They occur not only as "slips of
the tongue" (whatever that means), but as planned puns, double entendres, plays on
vords, and variously in poetry and advertising. We can do three things about them: (1)
Ignore tham (perhaps as "ungrammatical"). (2) Regard them as deviations from "normal"
sentences, to be explained vith special machinery glued onto our basic theory for "nor­
mal" sentences. (3) Use them as evidence for some new and very different theory of the
generation of speech, that will provide at once for such "deviant" utterances and for all
"regular" utterances. If ve are really concerned, in linguistics, vith the discovery and
description of the place of language in the universe, I believe ve must seriously consider
the third alternative, no matter hov radical may be the revisions that are required in our
vays of thinking.
The examples in this chapter are almost all from English, but it is not only
experience with that language that has led me to the abandonment of the
atomic morpheme theory. At the end of the notes to this chapter I append a
very brief survey of additional embarrassing data, mainly from other lan­
guages.
§6.1 ρ 65. Isomorphism: Kuryłowicz 1949.
148 NOTES AND COMMENTARY

§6.3 p 67. The most thorough study of idioms is surely A. Makkai 1982.
§6.3 p 67-8. Faded metaphor: Recall Bloomfield's comment 1933.443): "The
picturesque saying that language is a book of faded metaphors' is the re­
verse of the truth, for poetry is rather a blazoned book of language." This is
a fine antidote for the common somewhat derogatory view, but actually both
assertions hold a measure of truth. We need only remember that the fading
can be very slow and that not only poets, but also we ordinary language-
users, are continually reblazoning our speech.
§6.4 p 68. The Crispix box described here was the original. A few months
after introducing the cereal Kellogg changed the packaging; the replacement
strikes me as not nearly so clever.
On puns see Hockett 1973.
§6.5 pp 68-9. Certainly the most indefatigable investigator in recent years
of the sort of phenomenon discussed here has been Wescott, whose articles
on the subject are conveniently available in a 1980 reprint volume—in which,
for English, see especially pp 303-405. Also Woolley 1976.
Secondary associations: Hockett 1958 ch 35.
§6.5 ρ 70. Bloomfield's treatment: 1933.245-6.
On crans: Christie (1984.56-8) feels that such pieces do not even satisfy
the criteria for recognition as morphemes. He is led by this, and by some
knotty problems in phonemics which he perceives as methodologically parallel,
to challenge one of our unstated traditional assumptions in synchronic
analysis (in effect, the notion that working down from utterances to minimal
elements and working up from minimal elements to utterances are strict
inverses). Although he does not explicitly say so, his conclusion clearly
involves rejecting the atomic morpheme theory.
§6.6 ρ 71. Blends: See the careful classificatory treatment by Algeo (1977).
Current English coinages are cited frequently in the same journal (American
Speech). For slips of the tongue see the articles reprinted and referred to in
Fromkin, ed, 1973, and Dell 1977.
Reaganomics. Russell and Porter (1982.206) treat this rather as Reagan
plus a suffix -(on)ornics which they find in a number of new formations of the
same period. Of course, we have to allow for individual differences, and some
people hearing the word for the first time may have felt it that way. To me,
the difference between the voiced g of Reagan and the voiceless k-sound of
economics was trivial, so I heard ALL of economics in the new word, as well as
ALL of Reagan.
§6.7 ρ 72-3. Malapropisms: Bloomfield 1933.154; Hockett 1958.287.
Contamination, reshaping: Bloomfield 1933.422-6; Hockett 1958.287-90,
390. Many more examples of contamination could be added, but we settle for
NOTES AND COMMENTARY 149

two: (1) volcanology (influenced by English volcano, from Italian) has almost
entirely replaced the older vulcanology (Latin Vulcānus, supporters of the new
shape can point out that Latin records also show a spelling with O). (2) Words
like controversial have competing pronunciations in -sh'l and in -see-'l. The
latter is a new formation: tack -al on at the end of controversy just as you
add it to form or mechanic or anything else. The former reflects an earlier
pronunciation with -see-'l modified by an intervening regular sound shift.
Taboo: Bloomfield 1933.400-2; Hockett 1958.399-400.
§6.8 ρ 74. Abbreviation: Hockett 1958.313-6. And see the Chinese example
at the end of the notes to this chapter.
§6.9 ρ 75. Chemical nomenclature: Crosland 1962.
"Agglutinative" and other typological labels: Steinthal (ed. by Misteli) 1893;
Jespersen 1922.79-80; Sapir 1921 ch 6.
Another topic worth mentioning here is our reactions to personal names.
These certainly derive in large part from the personalities of people known to
us who bear the names, but secondary associations (§8.5) are also relevant.
For many a user of American English, no amount of admiration for someone
bearing the name can keep Percival, Egbert, Casper, Homer, or Herkimer
from sounding queer. Of course it is relative to the language background.
Pincus and Hyman are odd in English, fine in Yiddish and Hebrew.
6.1-9 ρ 76. Further embarrassing data:
SPANISH VERB INFLECTION: am-α is indicative but beb-a subjunctive; am-e is
subjunctive but beb-e indicative. There is nothing about the shape of either
the stems or the endings to tell which is which; this can be determined only
with reference to other inflected forms of the same verb. Thus, only by
virtue of such paradigmatic associations can a hearer properly identify the
form he hears. Of course this sort of thing can be handled with atomic
morphemes, but the analyst has to set up the morphemes and devise the
morphophonemic notation on the basis of the total behavior of the verbs,
which means that the morphophonemic notation is simply shorthand for the
paradigm. The traditional statement, directly in terms of "principal parts and
paradigms," if properly streamlined, is more realistic and more economical.
For a more detailed example of this sort, using not Spanish but Yokuts, see
Hockett 1967a.219-22.
ENGLISH VERB INFLECTION: By standard morphemic procedure, one has to
recognize a past-tense suffix that appears (in various shapes) in judged,
sang, hid, and a past-participle suffix that appears (likewise in various
shapes) in judged, sung, hidden. Each of these morphemes has a basic shape
(namely, / - d / in both cases), some automatic replacements ( / - t / , / - + d / ) ,
and some irregular alternants. It is possible, but complicated, to deal with the
two forms independently. But the description of English verbs is rendered
150 NOTES AND COMMENTARY

much simpler, and probably more realistic in terms of how the language is
actually learned and used, if we forget about morphemes, define the past
tense as irregular (as we always have) when it is not made with -d or its
automatic replacements, but define the past participle as irregular only when
it is not identical in shape with the past tense.
FIJIAN (Arms 1974; Schütz 1985; my own notes). Verbs appear with either
of two suffixes, -a and -/, or with none:
TRANSITIVE PASSIVE INTRANSITIVE
laveta laveti lave 'raise up'
lovaa lovoi lova 'pour in a small stream'
gutuva gutuvi gutu 'sever'
raica raici rai look, see'
musuka musuki musu 'break or cut crosswise'
lomana lomani loma 'love, have mercy on'
lernoya lernoi lemo 'conceal'
savora savori savo 'pull tight'
ceruma cerumi ceru 'sip'
terega teregi tere 'touch lightly'
Fijian allows no word-final consonants. We are therefore tempted to set up
verb stems with final "thematic" consonants (raic- with thematic c, and so
on), and allow the loss of that consonant when no suffix follows to be
automatic (cf. Bloomfield's suggestion for Samoan, referred to in §7.2 ρ 80).
But that doesn't quite work. Historically, the word-final consonants of Proto
Malayo-Polynesian were lost in the Fijian lineage some three thousand years
ago, and there has been a great deal of analogical reshuffling since. In some
cases stems alike in shape except for thematic consonant show sharp se­
mantic differentiation, as labaya 'examine a set trap' and labato 'pierce,
murder'; in others there is little or none, as kuruta or kuruyo, both 'drive off
(birds or the like) with a stick'. There are some dissimilative tendencies: e.g.,
thematic  does not occur after either  or q earlier in the stem. There is
also some rather vague semantic correlation: many stems referring to
breaking, splitting, and other such forceful actions have thematic k, many
denoting motions or positions have v, and so on; but there are numerous
exceptions. Also, actual usage shows more fluctuation than the writers of our
reference works report. It seems unrealistic either to regard thematic con­
sonants as separate morphemes or to take them as parts of stems—or even
to do sometimes the one, sometimes the other. But those are the only
alternatives made available by the atomic morpheme approach.
ALGONOUIAN MORPHOLOGY. In his treatment of this (e.g., 1962), Bloomfield
distinguishes between ROOTS, which can begin a word, and MEDIALS, which
cannot. But some roots and medials resemble each other in sound and
meaning. For example, Menomineewāparmew'he looks at him' contains root
wāp- 'look', whereas enāpamsw'he looks thus at him' (with relative root en-
NOTES AND COMMENTARY 151

'thus') contains medial -āp- 'look'. Now are wāp- and -āp- the same mor­
pheme? If so, then the clean contrast between root and medial is lost, since
we are recognizing elements that can occur in either morphological position.
Or are they related forms — say, one a morpheme and the other a complex
form derived from it? If so, which is the single morpheme, what derivational
morpheme appears in the other, and what does that derivational morpheme
mean? Or are they simply distinct, though roughly synonymous, morphemes?
But that is tantamount to a rejection of the ordinary criteria for morphemic
analysis, which instruct us to look for MINIMUM meaningful shapes.
An isolated case like this would hardly be disturbing, but the languages are
shot through with them, and Bloomfield's treatment gives them full recog­
nition. I think it is clear that we cannot retain both the atomic morpheme and
Bloornfield's remarkably insightful portrayal of Algonquian morphology, and I
vote for the latter. (It is perhaps relevant that in his whole description of
Menominee the word "morpheme" occurs only once, and then only as a
parenthetical clarification.)
There are kindred problems in English, recently examined by Mulder and
Hervey (1980.122-44). They show that, following a strict formal procedure
which they derive ultimately from Saussure, and heeding only what they
define as "denotation" rather than all of meaning, it is not even possible to
identify the black of blackbird with the independent word black. That is
clearly a reductio ad absurdum. It follows that morphemic theory, in any
version which requires exact identifications, is unable to account for the
facts of language.
CHINESE comes about as close to the atomic morpheme ideal as any
language I know, especially among literate users, for whom morpheme iden­
tifications and distinctions tend to be determined by assignment of character
(that is, with a few obvious exceptions same character means same mor­
pheme). Yet there are great difficulties, brought about especially by abbre­
viation. In Beijing in the autumn of 1986 my wife and I memorized the full
name and address of our home base, and carried it around written on a card
in case our pronunciation was unintelligible. But we found that a taxi driver
would take us home if we asked simply for yíwài literally "one outside" (I
shall use double quotes here for "literal" renderings). This is a replacement
for běiwài "north outside," used because there is another institution also
entitled to be called běiwài, that other institution, being less well known, is
distinguished as èrwài "two outside." Běiwài, for our organization, is in turn
an abbreviation of the full official name which
taken morpheme by morpheme means "north-capital outside-country-language
learn-establishment"; but "north-capital" means 'Beijing' (or Bĕijing means
'north capital'); "outside-country-language" means 'foreign language'; and
"learn-establishment" means 'institute' or 'university'. So in an indirect sense
152 NOTES AND COMMENTARY

the two-syllable expression yíwài means 'Beijing Foreign Studies University'

(the official English name). But how can one divide up that total meaning so as
to allot its proper share to each of the morphemes in the two-syllable
expression?
A single case of this sort would be unimportant, but the language is riddled
with it (Chao 1968a.366-7,379,381ff, and elsewhere). Chao tries to allow for it
with his "principle of immediate analysis" (1968a.367), but upon inspection
this turns out to be not so much a principle as a label concealing the fact
(which Chao would have been quick to admit) that no clean-cut boundary can
be drawn between synchronic and diachronic: people draw from their
knowledge of their language only what they need to meet the exigencies of
the moment, and in using that part of their language they change it, while the
rest remains essentially unaltered. Furthermore, the evidence shows that this
kind of thing goes on constantly, not just in Chinese but in every language,
and has been going on ever since language began.
7. From Particle to Resonance.
Both the description of the problem in our chapters 6 and 7 and the
solution proposed in chapters 7 and 8 are paralleled, not exactly but closely
enough to be suggestive and useful, by Christie in 1980 chapter 5. I have also
found encouraging similarities in some recent work of Anttila (1976, 1977,
1980), especially his overt mention of gestalt.
§7.1 ρ 77. Granularity and gestalt: Hockett 1983 §§10-12.
§7.1 ρ 78. Whitney 1867 ch 2 second half.
"Isolating" and other typological terms: see §6.8 ρ 74 η.
§7.2 p 79. On the history of morphophonemics see Kilbury 1976. For the
views of Baudouin and Saussure on these matters, the relevant chapters of
Anderson 1985 give us the benefit of some first-rate historical scholarship.
History of the English regular plural: Peters (1985) says that what actually
happened was not really very much like what one would infer from the current
alternation. "Internal reconstruction" doesn't always work! I don't think that
seriously impairs our argument.
According to Kilbury (1976.31), H. Ułaszyn, an active member of the Prague
circle, introduced morphonema in Polish in 1927, in German in 1931; the expan­
sion to morphophonema came shortly thereafter.
§7.2 p 80. Bloomfield on Samoan: 1933.219. The same difficulty with this
way of handling the matter turns up for Samoan as for Fijian (see above,
§6.1-9 p 76 n).
§7.2 p 81. Potawatomi: Hockett 1948 §2.
§7.3 p 81. We can here usefully add a point to the comparison of linguistic
traditions made in §0.2 p 2 n. In the nineteenth century, and still for Saus­
sure, the element at the focus of attention in both descriptive and historical
NOTES AND COMMENTARY 153

work was the WORD. In his review (1923) of Saussure, Bloomfield suggested
that the SENTENCE, rather than the word, should be highlighted, but the post-
Bloomfieldians in fact concentrated on the MORPHEME; only with Chomsky did
the American emphasis change to the sentence. The British functional-
systemic grammarians, meanwhile, had long been taking the sentence —or,
more accurately, the CLAUSE —as their major point of departure. To be sure,
these emphases were far from absolute. For example, my own first learnèd
publication (1939) was on syntax.
"Sentence" vs "utterance": of course the two terms are not synonymous,
although they are occasionally confused even by specialists. An utterance is
any act of speech: utterances are the linguistic analyst's raw material.
Meillet's 1903 definition of sentence (p 355 in 8th edition, 1937), accepted by
Bloomfield in his 1926 postulates, still seems operationally the best, even
though syntactic connections-at-a-distance (§2.5 ρ 23 η) render the applica­
tion of the Meillet criterion somewhat fuzzy.
§7.3 ρ 82. Although Bloomfield spoke of morphemes in his 1926 "Postu­
lates" and his 1933 treatise, he had too much respect for the cantankerous
facts of languages to be as consistent and systematic about morphemes as
his followers soon became. Atomic morphemitis was largely a disease of the
American post-Bloomfieldians. Various heckling voices could be heard from
across the Atlantic if one took the trouble to listen (e.g., Bazeli 1949, 1951,
1952, 1954). Bolinger was an early American apostate (first 1948, then on
related matters in 1954, 1961, 1981); later Chafe joined him (1963).
The American thinking of the Decade of the Morpheme is well covered by
the articles reprinted in (and those referred to in) Joos, ed, 1957; Hamp et al,
1966, supplements this with contemporary criticism.
§7.4 p 83. Lounsbury 1953 §1.1.
§7.4 p 84. It was the reality theory of morphophonemes, along with the
transducer view (§7.5), that led to generative phonology, with its orientation
in terms of production and its relative neglect of the hearer.
Generative phonology rested, when it first began, on a tripod of assump­
tions: (1) the basic stability of morphemes (or of some such unit) despite
their tendency to appear superficially in more than one shape; (2) the
correctness of the theory of a relatively small "universal phonetic alphabet"
of binary oppositions (Jakobson and others, 1952, with subsequent minor
revisions); and (3) the conception of phonemics as a stratum sandwiched
between morphophonemics and phonetics, differing from the latter only in
that there are no redundancies. With this orientation, given rules for mapping
arrays of morphophonemic elements into strings of phonemes and rules for
mapping the latter into strings of sounds, the eliminability of the so-called
phonemic stratum is obvious.
154 NOTES AND COMMENTARY

If any leg of a tripod fails, the structure topples. But from the empirical
point of view all three legs of this particular three-legged stool are lame or
broken. The notion of the stability of morphemes has always been questioned
by some scholars, and is undergoing thorough examination and rejection in
the present essay. The Jakobson-Halle "universal phonetic alphabet" uses the
wrong mathematics (discrete instead of continuous with quantization) and is
both articulatorily and acoustically unsound (see Joos 1957); its continuing
popularity is a scandal. The interpretation of phonemics as a "level" is the
transducer fallacy: phonemics is meaningless or useless except in terms of
the hearer's position.
A great deal has happened in generative phonological theory since its
inception, but as far as I can see these three fundamental defects have not
been eliminated. That the atomic morpheme still reigns virtually unchallenged
can be seen in such an article as Halle 1977. (But see §7.7 ρ 92 η).
§7.5 ρ 85. In my own work the transducer fallacy was apparent though
subsidiary in 1955, dominant in Hockett 1961. Lamb's stratificational theory
(see A. Makkai and Lockwood, eds, 1973, for reprints of some early papers
and references to others) also incorporated the fallacy at the outset, but
with the development of relational networks it started to break away.
§7.6 pp 86-7. What I wrote in 1958 (cited here) scarcely deviated from
what Nida had said a decade earlier (1948) about the same sorts of
morphemically marginal phenomena, or what Householder (1952.266-7) had
said about "descriptive etymology" and "phonesthematics." But the emphasis
had changed. Nida had felt that some method must exist for distinguishing
between "official" morphemes and "accidental" resemblances, and Householder
had implied no major dissent from that opinion, whereas by 1958, though
dutifully reporting the prevalent view, I also expressed serious doubts.
§7.6 ρ 88. Bloomfield on Menominee morpholexical alternations: 1939 §3,
1962.58-61.
§7.6 ρ 89. Emulatability. When we replay (from short-term memory) or
rescan (in reading) an unusual word like this, in order to register its make-up
in terms of resemblances to more familiar words, we might almost speak of
another mode of listening (cf. §4.1, §4.7) — except that the process seems
also akin to syntactical processing (parsing and construing).
Pike introduced etic and emic in 1954.8.
§7.7 pp 90-1. On the close-grained and continual dialectic interplay of
performance and system in the individual, see below, §8.3 ρ 104 η.
§7.7 ρ 92. Fifty years: you will find this in Joos 1962 ch 5, but only by
rereading.
If kwasNorkor and sigatoka (not to mention Prokokofìeskovitch and Ng)
can be English words, then there is no reason why Halle's specimens
NOTES AND COMMENTARY 155

(1973.120), including Donaudarnpfschiffahrtsgesellschaft and katav, cannot

be. All it takes is for someone to use them. Just the same, the point Halle is
getting at is an important one.
The sha in shamrock: Read 1949 is an excellent survey of this sort of
thing, with examples such as the pun in pundit the bust in robust, the sex in
sextet. And how about the green in gangrene, the bottom (and the flea) in
phlebotomy?
We give here a list of word pairs for the reader's delectation. The order is
as random as I could make it. In each case, the only consideration is your
personal reaction: are the two words related (for you) or not, and, if they are,
in what way? The evidence you collect in this way from your own idiolect
should be much more convincing about the points made in the text than any
examples I might choose because they are persuasive to me:

orchard : vineyard acid : acrid brim : rim

enormity : enormous infant : infantry crayfish : fish
hog : hedgehog noisy : noisome house : husband
sorry : sorrow wife : hussy triple : treble
adult : adultery bylaw : bypass flounder : founder
grammar : glamour par : peer prudent : prurient
vindov : vind (that blovs) catnap : catnip male : female
pencil : pen gumbo: jumbo thimble : thumb
pen : feather bald : ribald tie (score) : tie (fasten)
recover : cover vixen : fox tire : retire
mouse : titmouse inch : ounce vreak : vrought
prostate : prostrate drink : drench sonnet : sonata
brothel : bordello turpentine : tar porridge : pour
devil : evil suppose : suppository shriek : screech
wrought : vork frail : fragile nine : noon
catsup : catnip pipsqueak : pipe recognise : reckon
revolve : rotate gleaming : gloaming moment : momentum
soluble : solvable debit : debt grain: corn
curry (flavor) : curry (favor) human: man soup tureen : nectarine
crov : escrov quadrant : quarter comb : brush
dale : dell glimmer : shimmer host : guest
vine: rine Monday : moon griddle : grill
sup : sip timber : lumber coven : Coventry
s t o v : store happen : happy ostensive : ostensible
156 NOTES AND COMMENTARY

sloth: slow vink: blink disciple : discipline

unkempt : comb killdeer : deer orgasm : organism
holistic : whole sweat : sveater girl bland : blond
temblor : tremble lord : lady shade: shadov
volcano : vulcanology guard: vard mead : meadov
nacho: natural chovder : povder vink : vench
ass : asinine mongoose : goose milk: milch
evade: avoid circadian: cicada Africa : affricate
burden (of song) : burden (load) coven : covenant pray : prayer
isle : island ear (on head) : ear (grain) chev : esche v
voodchuck : vood karat : carat regimen : regime
piane: plain valley : vale vork : vreak
pipe : fife corn (foot) : corn (grain) rage : outrage
actor : agent vertical : vertex furbish : furnish
rehearse: hearse month : moon invent : inventory
crevice: crevasse anode : cathode junction : joint
grin: groan plus : minus statue : statute
chalumeau : calumet bustard : bussard flower : flour
tight : taut ade: ale draver : drav
socket : sprocket bodice: body drag: drav
shiver : shimmer see · seer eleemosynary : alms
super : superb kidnap : catnap beseech: seek
vertex : vertigo vind (a vatch) : vinch ruddy : red
hussy : house vreck : vreak formica : mica
skirt: shirt deal : ordeal chord : cord
backward : back vard delay : dally flavor : savor
wife: hussy cat:scat lentil: lens
groom : bridegroom bizarre:  livid : hver
wastrel : waste grov : green sup : supper
ear : hear chipmunk : monkey decadent : decade
pita : pisa guerrilla: var horizontal : horizon
loathe : loath ambush : bush surgeon : sturgeon

§7.7 ρ 95. Sapir 1925.37, 1933 passim.

NOTES AND COMMENTARY 157

8. How Language Means.

§8.1 ρ 97. On our preference for sharp boundaries, the brief discussion in
Joos, ed, 1957.VI (about "identicals" and "graduals") is excellent.
For sound suggestions from Bazeli, see §7.3 p 82 n.
To the roster of fuzzy oppositions given in the first paragraph should be
added that between synchronic and diachronic. See §7.7 p 94 (middle of the
page), and —even more to the point—Christie's penetrating critique (1984.56)
of the Saussurian notion of a stable état de /angue.
§8.1 p 98. "Almost same": Hockett 1953.168; Bolinger 1954; Chomsky
1957b.232-3; Hockett 1960b.184, Bolinger 1961; Hockett 1977b.62.
§8.1 p 99. We see here, in the last paragraph of §8.1, why it is not possible
to count the number of constructions in a language (see §2.5 p 22 n par. 10).
§8.2 p 100. On Saussure see below, §8.3 p 104 n.
The relevance of idiolects and local dialects: Read 1962.
§8.2 p 103. The summary statement of variation and calibration is from
Hockett 1986.64-5.
§8.3 p 104. Code and message; social and individual: Saussure was right in
his insistence on the difference between code (system, institution, pattern,
culture) and message (act, event, structure, behavior). Investigators (e.g.,
Trager, Pike) whose terminologies do not in one way or another maintain this
distinction have always gotten into trouble. Saussure was also right in
insisting on the contrast between the social (shared, collective) and the
individual (personal, idiosyncratic). But he made a serious error, misleading to
a great many of his successors, when he identified these two oppositions
under the single pair of terms langue and parole, implying that the system is
only social, not individual, and that individual conduct is only idiosyncratic, not
patterned. (There was a strong but subtle nineteenth-century cultural theme
pressing toward this mistake of Saussure's, a theme which I hope to expound
presently in a revision of Hockett 1983.)
In fact, an individual's speech habits, at any moment, constitute a system
which underlies and conditions what the individual says and how the individual
interprets the speech of others; and every such episode of the use of lan­
guage modifies the individual's system at least a little. Furthermore, a lan­
guage as a system in this sense of the word exists ONLY in individuals. When
we — quite properly — call a language a social system we are implying a
somewhat different use of the word: there are a great many agreements or
parallels among the systems of the participating individuals (whose usages,
as we have seen, are constantly being intercalibrated); by virtue of these
parallels the participants can ordinarily manage to understand one another;
158 NOTES AND COMMENTARY

and although the whole set of parallels is only roughly defined, it can validly
be called a "system" in this slightly different but related sense of the word.
But, quite clearly (see §8.1 ρ 97 η), there is no such thing as a stable état
de longue of the sort Saussure proposed, in either an individual or a
community—or, if there is, it is stable for only a fraction of a second.
§8.3 ρ 105. In 1946 Morris proposed a trichotomy of semantics, pragmatics,
and SYNTACTICS, the last being the purely internal economy of a sign system,
the ways meaningful pieces can relate to one another in messages. Before
Morris, apparently unknown to him, Trubetzkoy (1938.17 ff), following Bühler,
had made a further distinction, separating KUNDGABEFUNKTION, which bears on
the sender or the sender's attitude, from APPELFUNKTION, which relates to
what is wanted from the receiver. But all these (except syntactics) are
subsumed under "semantics" (or "meaning") in our broad sense.
Almost needless to say, the precise but very narrow sense assigned to the
term "meaning" in logic, by which a proposition is meaningful only if it is true
or false, is hardly even a sixth cousin thrice removed of "meaning" in the
inclusive sense in which it must be taken in linguistics.
§8.4 pp 106-7. For Bloomfield's views on these matters see especially
1927, 1930, 1933, 1936, 1942a.
We can use the diagram on page 107, and the accompanying nonce def­
initions of "thinking" and "theorizing," to clarify Bloomfield's position and to
point out what in my opinion was his mistake. Bloomfield saw language (the
horizontal axis) as bridging the gap between the nervous systems of different
individuals, so that a stimulus on one person could elicit a very precise
response from another; and certainly, whatever else is the case, language
does do just that. As for problem-solving, he saw it as a function of language
used in consultation (perhaps in the limiting case just with oneself): that is,
of peimarily collective theorizing rather than of individual thinking. He did not
deny the existence of the diagram's vertical axis, but thought of it has having
to do only with obscure internal events that could be of no public concern.
The passage most often quoted from Bloomfield about events along the
vertical axis (private, so-called "mental," phenomena) is this (1933.34):
... It is a mistake ... to suppose that language enables a person to observe things for which
he has no sense-organs, such as the workings of his own nervous system.
Bloomfield's assertion here about what language does not do is obviously
correct. But apparently we do in effect have "sense-organs" with which we
can observe what goes on in some parts of our nervous system, since every
one of us has a complex internal private life, and on a physicalist assumption
this can be nothing other than such observation (see ch 9 passim). What
NOTES AND COMMENTARY 159

language does is to make it possible for us to report on this to others —

something speechless animals cannot do.
For further critique, see Hockett 1983 §14.
§8.5 ρ 108. The classification of connotations given here is not identical
with Saussure's, but certainly very much like it. I don't think I was influenced
by Saussure in working out my own. Perhaps that renders the similarity even
more significant.
§8.5 ρ 109. Poetry, and literature in general: In the heyday of the post-
Bloomfieldians, members of college literature departments did not like us very
well. Partly this was economic: at a number of institutions, linguists were
taking over elementary foreign-language instruction, long a prerogative of
scholars whose primary interest was literature. In part it was because they
did not really know what we were trying to do. But in part, perhaps, it was
because they suspected our orientation at the time was not designed to
come to grips with those aspects of language that do the most in making
literature possible. And if they really did sense this, I think they were right.
Later somè of them welcomed TG as an approach more compatible with their
own interests, but in that they were wrong. Chomsky talks about human
creativity, but I think his basic philosophical stance absolutely precludes it
(see Hockett 1968 and—deo volente —the forthcoming revision of 1983).
§8.6 ρ 110-3. Be sure to recognize that we are here discussing (a theory
of) internal storage of language material in the brain of A SINGLE USER. We are
not talking about some superindividual uniform ideal system in a "group
mind," floating like a cloud over the heads of the individuals who compose the
speech community (see §8.3 ρ 104 η). Furthermore, we have no reason to
assume that the details of internal storage are identical for different
individuals, even when observable usage seems to be indistinguishable. It is a
mistake to assume that similarity of external manifestation implies, beyond a
bare minimum, similarity of internal organization. So, for example, the infor­
mation stored in one user's head in a single packet with a multifaceted
denotation may be stored in another user in separate packets for the
different facets.
§8.6 p 111. On valence: when I introduced this term from chemistry into
syntax in 1958 (pp 248-9, 253-4) I thought it was my own idea. Almost three
decades later Moulton called my attention to its use by Tesnière (1953.9;
1959.238). I gladly relinquish any claim to priority, especially because in
checking into Tesnière's work I have found other ways in which his thinking
anticipated my own. Compare, thus, his discussion of dimensionality in
1953.3-4 with mine in the first chapter of this essay.

9. The Craft of Speaking.

§9.1 ρ 114. Despite our traditional bias in favor of the speaker, linguists
have actually said very little about the problem of speech-production (other
160 NOTES AND COMMENTARY

than, in phonetics, the mechanics of phonation). But there are some stim­
ulating exceptions, mostly by way of brief asides: Bloomfield 1962.44-0-441 on
anticipation, parenthesis, and afterthought in Menominee; Chao 1968a.129-135
on [pre-]planned and un[pre-]planned sentences; Householder 1971 chapter 2
on reasonable and unreasonable assumptions about interior (mental) pre­
cursors to speech; Yngve 1970 on forgetting and its implications; Joos 1962
ch 5, especially on rewriting and rereading.
Especially stimulating is the daring Pions ond the structure of behavior of
Miller et al, 1960.
My views as expressed in this chapter have grown over the years from
rough beginnings in the Introduction to my unpublished 1936 MA Thesis at
Ohio State University. Bits and pieces have been set forth in various papers,
such as 1967b (especially §11), 1973, 1976.79 (also on reception); related
papers, that bear on reception as well as transmission and whose subject
matter is not confined to language, are 1960b and 1964 (both reprinted in
1977a), and H10 and H11 in 1977a.
§9.2 ρ 116. The contours of private mentation: the statement in the text is
not equivalent to the notion of the mind of the newborn as a tabula rasa.
Against the claim that private patterns can derive only from an inter­
nalization of overt activity and experience, one might propose that some may
be innate. But that only renders the line of causality more circuitous.
"Innate" is an old-fashioned word for "genetically inherited." What is
genetically inherited is a matter of natural selection, and selection is at
bottom an exogenous process: that is, it is the external conditions of life that
determine whether an organism shall survive and reproduce and pass on its
genes. So here, again, we have the internal deriving from the external.
The inner "stream of speech": Lashley 1951 (p 184 in 1961 reprint) cites,
among others, Titchener's personal account; Vygotsky 1934(1962).44-52 and
passim; Jakobson and Waugh 1979.77-9. Behaviorists (e.g., Watson) thought of
inner talking as subliminal movements of the organs of speech, but there is
no reason why it should not be confined wholly to the central nervous system
(compare our comment about implicit motor-matching, §4.4 ρ 40).
§9.5 ρ 122. Compare the musical analysis of poetic rhythm here with that
given by Halliday 1985.8-18. He and I agree that the traditional machinery of
feet, iambs, dactyls, etc. is not well adapted to English.
§9.5 ρ 126. Is the completed limerick Draconian or Tsakonian? Either way,
don't miss the subtle allusion to the Peloponnesian War.
§9.6 ρ 127. On editing: see especially, in Hockett 1960b, the section so
headed. Also Householder 1976.483 (in an article that argues the incom­
patibility of the Chomskyan assumptions about innateness with the empirical
fact of human creativity in speech).
NOTES AND COMMENTARY 161

§9.6 p 128. The usefulness of transformations in description and pedagogy

is undeniable and not at issue here. In my own work I have found them
especially convenient in tying together the various uses of the Chinese
particle de. They are also clearly of diagnostic help in the process of
analysis: trying to passivize The butcher weighed two hundred pounds quickly
shows that it is not like The butcher weighed the meat. Since 1957 a thousand
papers exploring points of syntax have used this technique.
On our various transforms: give, of course, has a valence of three,
respectively for actor, recipient, and goal; compare Moulton's illustration
(1986) with the verb owe, which has the same valence. Following Moulton's
terminology, we could say that our seven transforms, though identical in
fundamental noun-verb ties (based on the valence of the verb), differ in
(SENTENCE) PERSPECTIVE. The valence approach also tells us what to do with
pieces like the prepositions to and by in some of our examples, or (see §8.6 ρ
111) with the at that "transitivizes" took, the to that transitivizes listem they
are overt VALENCE BONDS (maybe like the hydrogen bond in chemistry?), doing
in some syntactic circumstances what is done just by word order and verb
form in others.
Given the prominence and strength of Greek grammar in our linguistic
tradition, it is easy to fall into the trap of thinking that only such relations as
subject-verb or verb-object, which are formally backed by case inflection in
the classical languages, are "syntactic," and to interpret roles like actor,
goal, and recipient, or relations like topic-comment, as not syntactic at all
but something else, say "semantic." I know no way of justifying such a
segregation. The roles in question are of words or phrases, and are therefore
by definition syntactic; the relations in question are between words or
phrases, and are therefore by definition syntactic. All these roles and
relations have meanings and contribute to the meanings of sentences, which
makes them all semantic. It is incorrect to CONTRAST "syntax" and
"semantics" as though they were two separate "strata" or "components" of
language design. Here, again, I recommend Halliday's (1985) "polystruc-
turalism."
§9.7 ρ 130. On music: I kept a close record in the early 1960s of the
sequence of aims and decisions in the composing of a fairly elaborate piece.
The process fits in every way the paradigm presented in this chapter. What
limited testimony we have from the great composers also seems to agree,
despite many differences in detail. See also Longacre and Chenoweth 1986,
Chenoweth 1986.
 REFERENCES
Aarsleff, Hans. 1982. From Locke to Saussure: Essays on the study of
language and intellectual history. Minneapolis MN: University of Minnesota
Press.
Agard, Frederick ., Gerald Kelley, Adam Makkai, and Valerie Becker Makkai,
eds. 1983. Essays ¡η Honor of Char/es F.Hockett. Leiden: E. J. Brill.
Algeo, John. 1977. "Blends: A structural and systemic view." American Speech
52.47-64
Anderson, Stephen R. 1985. Phonology in the twentieth century. Chicago:
University of Chicago Press.
Anttila, Raimo. 1976. 'The metamorphosis of allomorphs." In Reich, ed,
1976.238-48.
— . 1977. "Dynamic fields and linguistic structure: A proposal for a Gestalt
linguistics." Die Sprache 23:1.1-10.
—. 1980. "Field theory and morphology." Lingua Posnaniensis 23.15-9.
Arms, David George. 1974. Transitivity in standard Fijian. University of
Michigan Dissertation; Ann Arbor Ml: Xerox University Microfilms.
Bar-Hillel, Yehoshua. 1953., "Logical syntax and semantics." Language
30.230-7.
Bazeli, Charles Ernest. 1949. "On the problem of the morpheme." Archivum
Linguist/cum 1.1-15. Repr Hamp et al, eds, 1966.216-26.
—. 1951. "Structural notes."Studies hy Members of the English Department,
Istanbul University 2.3-25.
—. 1952. "The correspondence fallacy in structural linguistics." Studies by
Members of the English Department, Istanbul University 3.1-41. Repr Hamp
et al, eds, 1966.271-98.
—. 1954. "The sememe." Litera (Istanbul) 1.17-31. Repr Hamp et al, eds,
1966.329-40.
Birdwhistell, Ray L. 1952. Introduction to kinesics: An annotated system for
analysis of body motion and gesture. Washington DC: Foreign Service
Institute.
Blansitt, Edward L. Jr. 1983. "Non-constituent connectives." In Agard et al,
eds, 1983, 180-8.
Bloomfield, Leonard. 1923. Review of Saussure 1922. The Modern Language
Journal 8.317-319. Repr Hockett, ed, 1970.106-8.
—. 1926. "A set of postulates for the science of language." Lg 2.153-64.
Repr International Journal of American Linguistics 16.195-202; Joos, ed,
1957.26-31; Hockett, ed, 1970.128-38.
REFERENCES 163

. 1927. "On recent work in general linguistics." Modern Philology

25.211-30. Repr Hockett, ed, 1970.173-90.
. 1930. "Linguistics as a science." Studies in PhiIology 27.553-7. Repr
Hockett, ed, 1970.227-30.
. 1933. Language. New York: Holt.
. 1936. "Language or ideas?" Language 12.89-95. Repr Hockett, ed,
1970.322-8.
. 1942a. "Philosophical aspects of language." In Studies in the history of
culture: The disciplines of the humanities (presented to Waldo Gifford
Leland). Menasha Wl. Pp 173-7. Repr Hockett, ed, 1970.396-9.
. 1942b. Offline guide for the practical study of foreign languages.
Baltimore MD: Special Publication of the Linguistic Society of America.
. 1962. The Menomini language. New Haven CT: Yale University Press.
Boas, Franz. 1911. "Introduction." in Handbook of American Indian Languages
(Washington DC: Bureau of American Ethnology, Bull. 40 Part 1) pp 1-83.
Bolinger, Dwight L. 1948. "On defining the morpheme." Word4-18-23
. 1954. "Identity, similarity, and difference." Litera 1.5-16.
. 1961. Generality, gradience, and the all-or-noneThe Hague: Mouton.
. 1981. "Consonance, dissonance, and grammatical it y : The case of wanna.
Language and Communication 1.189-206.
. 1982. "The network tone of voice." Journal of Broadcasting 26:3.725-8.
Chafe, Wallace L. 1963. "Some indeterminacies in language." Report on the
fourteenth annual round-table meding on linguistics and language studies
57-62. Washington DC: Georgetown University Monograph Series on
Language and Linguistics No. 16.
Chao, Yuenren. 1948. Mandarin primer: An intensive corse in spoken Chinese.
Cambridge MA: Harvard University Press.
. 1955. "Notes on Chinese grammar and logic." Philosophy fast and West
5:1.31-41. Repr Anwar S. Dil, ed, Aspects of Chinese sociolinguistics:
Essays by Yuen Pen Chao. (Stanford CA: Stanford University Press, 1976)
pp 237-49.
. 1968a. A grammar of spoken Chinese. Berkeley CA: University of
California Press.
. 1968b. Language and symbolic systems. Cambridge: At the University
Press.
Chenoweth, Vida. 1986. "Music as discourse." Word 37.135-9.
Chomsky, Noam. 1955. "Logical syntax and semantics: Their linguistic
relevance." language 31.36-45.
. 1957a. Syntactic structures. The Hague: Mouton.
. 1957b. Review of Hockett 1955. International Journal of American
Linguistics 23.223-34.
164 REFERENCES

. 1962. Paper and discussion remarks in Archibald A. Hill, ed, Third Texas
conference on problems of linguistic analysis in English: May 9-12, 1958.
Austin TX: The University of Texas.
——. 1964. Current issues in linguistic theory. The Hague: Mouton.
. 1975. The logical structure of linguistic theory New York: Plenum Press.
Christie, William M. 1980. Preface to a neo-Firthian linguistics. Lake Bluff IL:
Jupiter Press (Edward Sapir Monograph Series in Language, Culture, and
Cognition, no. 7).
. 1984. "Analysis vs. synthesis: An examination of the principles of
invariance and exhaustiveness." Forum Unguisticum 8.50-8.
Cowan, J Milton. 1965. "Attention." (Unpublished presidential address to the
Linguistic Society of America).
Crosland, Maurice P. 1962. Historical studies in the language of chemistry.
London: Heinemann Educational Books, Ltd.
Dell, Gary S. 1977. "Slips of the mind." In Michel Paradis, ed, The Fourth Locus
Farum: 1977. (Columbia SC: Hornbeam Press, Inc.) pp 69-74.
Diller, Karl Conrad. 1071. Generative grammar, structural linguistics, and
language teaching. Rowley MA: Newbury House Publishers.
Di Pietro, Robert, and Edward L. Blansitt Jr., eds. 1976. The Third Locus Forum:
1976. Columbia SC: Hornbeam Press Inc.
Espy, Willard R. 1975. An almanac of words at play. New Tork: Clarkson N.
Potter, Inc.
Fillmore, Charles J. 1968. "The case for case." In Emmon Bach and Robert T.
Harms, eds, Universals in linguistic theory. (New York: Holt, Rinehart &
Winston) pp 1-88.
Fischer-Jørvgensen, Eli. 1975. Trends in phonological theory: A historical
introduction. Copenhagen: Akademisk Forlag.
Francis, W. Nelson. 1983. Dialectology: An introduction. London and New York:
Longman.
Fries, Charles  1952. 77æ structure of English. New York: Harcourt, Brace &
Co.
Fries, Peter H. 1983. ".  Fries, signals grammar, and the goals of
linguistics." In Morreall, ed, 1983.146-158.
Frisch, Kurt von. 1950. Bees: Their vision, chemical senses, and language.
Ithaca NY: Cornell University Press.
Fromkin, Victoria A. ed. 1973. Speech errors as linguistic evidence. The Hague:
Mouton.
Gleason, H. Allan, Jr. 1961. An introduction to descriptive linguistics. Revised
ed., New York: Holt, Rinehart & Winston.
Grevisse, Maurice. 1964. Le bon usage: Grammaire Française, avec des
remarques sur la langue française daujourdhui. Gembloux (Belgique):
Editions J. Dugulot, S. A.
Hall, Robert Α., Jr. 1948. French Baltimore MD: Linguistic Society of America
(Language Monograph No. 24; Structural Sketches 1).
REFERENCES 165

-. 1968. An essay on language. Philadelphia: Chilton Books.

: 1985. "Meaning and the idiolect: t h e idioseme." In Jankowsky, ed,
1985.353-9.
Halle, Morris. 1977. "Morphology in a generative grammar." In A. Makkai et al,
eds, 1977.120-30.
Halliday, Michael A. K. 1985. An introduction to functionaf grammar. London:
Edward Arnold.
Hamp, Eric, Fred W. Householder, and Robert Austerlitz, eds, 1966. Readings in
linguistics II Chicago IL: University of Chicago Press.
Hockett, Charles F. 1939. "Potawatomi syntax." Language 15.235-48.
: 1945. "Formation of sentences." (Unsigned.) In Dictionary of spoken
chinese, pp 9-27. Washington DC: War Department Technical Manual TM
30-933.
: 1948. "Potawatomi." In four installments: UAL 14:1.1-10, 2.63-73,
3.139-49, 4.213-25.
: 1949. "Two fundamental problems in phonemics." Studies in Linguistics
7.29-51. Repr V. Makkai, ed, pp 200-10.
. 1953. "Short and long syllable nuclei (with examples from Algonquian,
Siouan, and Indo-European." international Journal of American Linguistics
19.165-71.
. 1954. "Two models of grammatical description." Word 10.210-31. Repr
Joos 1957b.386-99.
. 1955. A manual of phonology. Indiana University Publications in
Anthropology and Linguistics 11.
. 1958. A course in modern linguistics. New York: Macmillan.
: 1960. "Ethnolinguistic implications of recent studies in linguistics and
psychiatry." In William M. Austin, ed, Report of the ninth annual round table
meeting on linguistics and language study (Georgetown University Institute
of Languages and Linguistics, Monograph Series number 11) pp 175-93.
Repr Hockett 1977.107-23.
: 1961a. "Grammar for the hearer." In Jakobson, ed, 1961. Repr with a new
foreword in McGregor, ed, 1986.49-68.
. 1961b. "Linguistic elements and their relations." Language 57.29-53.
. 1964. "Scheduling." In Filmer S. C. Northrop and Helen H. Livingston,
eds, Cross-cultural understanding: Epistemology in anthropology. (New
York: Harper & Row) pp 187-203. Repr Hockett 1977.187-203.
. 1966. "What Algonquian is really like." UAL 32:1.59-73.
. 1967a. "The Yawelmani basic verb." Language 43.208-22.
. 1967b. "Where the tongue slips, there slip I." In To Honor Roman
Jakobson (The Hague: Mouton) pp 910-36 Repr Fromkin, ed, 1973.93-119;
Hockett 1977.226-56.
. 1968. The state of the art. The Hague: Mouton.
166 REFERENCES

: 1973. "Jokes" In M. Estellie, Smith, ed, Studies in linguistics in honor of

George L. Trager. (The Hague: Mouton) pp 257-89. Repr Hockett
1977.257-89.
. 1975. "If you slice it thin enough it's not baloney." American Speech
47.233-55 (issue for Fall-Winter 1972).
. 1976. "A new point d'appui for phonology." In Reich, ed, 1976.67-90.
. 1977a. The view from language: Selected essays 1948-1974. Athens GA:
The University of Georgia Press.
. 1977b. "A note on priorities." Forum Linguisticum 2.61-3.
. 1979. "In search of Jove's brow." American Speech 53:4.243-313 (issue
for Winter 1978).
. 1980. "Some historical notes on autonomous phonology." Forum Lin-
guisticum 3.95-106 (issue for December 1978).
: 1983. "The changing intellectual context of linguistic theory." In Mor-
reall, ed, 1983.9-42.
: 1984. "The uniqueness fallacy." In James E. Copeland, ed, New direc­
tions in linguistics and semiotics. (Houston TX: Rice University Studies) pp
35-46.
: 1986. "Language and 'scientific creationism.'" Forum Linguisticum
8.59-67 (issue for April 1984).
Hockett, Charles F., ed. 1970. A LeonardBloomfieldanthology. Bloomington IN:
Indiana University Press.
Hooper, Joan B. 1976. An introduction to natural generative phonology. New
York: Academic Press.
Householder, Fred W. 1952. Review of Harris Methods in structural linguistics.
International Journal of American Linguistics 18.260-8.
. 1971. Linguistic speculations. Cambridge: Cambridge University Press.
. 1976. "Innateness and improvisability," In Di Pietro and Blansitt, eds,
1976.479-85.
Hymes, Dell, and John Fought. 1981. American structuralism. The Hague:
Mouton. (Originally published in vol. 13, pt. 2, of Current trends in lin­
guistics, 1975).
Jakobson, Roman, ed. 1961. The structure of language in its mathematical
aspects. American Mathematical Society Proceedings of Symposia in
Applied Mathematics, vol. 12.
Jakobson, Roman, C. Gunnar M. Fant, and Morris Halle. 1952. Preliminaries to
speech analysis. Technical report no. 13 of the Acoustics Laboratory 2d
printing with additions and corrections). Cambridge MA: Massachusetts
Institute of Technology.
Jakobson, Roman, and Morris Halle. 1956. Fundamentals of language. The
Hague: Mouton & Co.
Jakobson, Roman, and Linda R. Waugh. 1979. The sound shape of language.
Bloomington IN: Indiana University Press.
REFERENCES 167

Jankowsky, Kurt R., ed. 1985. Scientific and humanistic dimensions of lan-
: Festschrift for Rotert Lado Amsterdam & Philadelphia: John
Benjamins.
Jespersen, Otto. 1922. language: its nature, development, and origin. London:
George, Allen & Unwin.
Joos, Martin. 1950. "Description of language design." Journal of the Acoustical
Society of America 22:6.701-8. Repr Joos, ed, 1957.349-58.
. 1957. Review of Jakobson and Halle 1956. Language 33.408-15.
. 1962. The five clocks. Bloomington IN: Indiana University Research
Center in Anthropology, Folklore, and Linguistics, Publication 22.
Joos, Martin, ed. 1957. Readings in linguistics. Washington DC: American
Council of Learned Societies.
Katz, Jerrold J. 1981. Language and other abstract objects. Totowa NJ: Row-
man and Littlefield.
Key, Mary Ritchie. 1975. Paralanguage and kinesics. Metuchen NJ: The
Scarecrow Press Inc.
Kilbury, James. 1976. The development of morphophonemic theory. Amsterdam:
John Benjamins B. V.
Kuryłowicz, Jerzy. 1949. "La notion de l'isomorphisme." Travaux du Cercle
Linguistique de Copenhague 5.48-60.
Labov, William 1963. "The social motivation of a sound change." Word
19.273-309.
Ladd, Robert S. 1980. The structure of intonational meaning: Evidence from
English. Bloomington IN: Indiana University Press.
Lashley, Karl S. 1951. "The problem of serial order in behavior." In Lloyd A.
Jeffress, ed, Cerebral mechanisms in behavior. (New York: Wiley) pp
113-36. Repr Sol Saporta, ed., Psycholinguistics: A book of readings. (New
York: Holt, Rinehart and Winston, 1961) pp 180-98.
Longacre, Robert E., and Vida Chenoweth. 1986. "Discourse as music." Word
37.125-34.
Lounsbury, Floyd. 1953. Oneida verb morphology. New Haven: Yale University
Publications in Anthropology no. 48. Introductory chapter repr Joos, ed,
1957b.379-85.
Makkai, Adam. 1972. Idiom structure in English. The Hague: Mouton.
Makkai, Adam, and David G. Lockwood, eds. 1973. Readings in stratifìcational
linguistics. University AL: University of Alabama Press.
Makkai, Adam, Valerie Bekker Makkai, and Luigi Heilmann, eds. 1977.
Linguistics at the crossroads Padova, Italy: Liviana Editrice, and Lake Bluff
IL: Jupiter Press.
Makkai, Valerie Bekker, ed. 1972. Phonological theory: Evolution and current
practice. New York: Holt.
168 REFERENCES

Mandelbaum, David, ed. 1949. Selected writings of Edward Sapir in language,

culture, and personality. Berkeley and Los Angeles CA: University of
California Press.
Martinet, André. 1957. "Arbitraire linguistique et double articulation." Cahiers
Ferdinand de Saussure 15.105-116. Repr Hamp et al, eds, 1966.371-8.
. 1980. 1980. "Synthematics." Word 31.11-4.
McGregor, Graham, ed 1986. Language f or the hearer. Oxford: Pergamon Press.
McKaughan, Howard. 1958. The inflection and syntax of Maranao verbs. Manila:
Publications of the Institute of National Language.
Meillet, Antoine. 1903. Introduction a létude comparative des langues Indo-
Européennes. Paris: Hachette. 8th ed, 1937.
l'čuk, Igor Α. 1986. "Toward a definition of case." Slavica 10.35-85.
Miller, George A. 1981. Language and speech. San Francisco CA: Freeman.
Miller, George A. and Noam Chomsky. 1963. "Finitary models of language
users." In Robert D. Luce, Robert R. Bush, and Eugene Galanter, eds,
Handbook of mathematical psychology. 3 vols. (New York: Wiley, 1963-5).
Miller, George Α., Eugene Galanter, and Karl H. Pribram. 1960. Plans and the
structure of behavior. New York: Holt.
Morreall, John, ed. 1983. The ninth Lacus forum: 1982. Columbia SC: Hornbeam
Press, Inc.
Morris, Charles W. 1946. Signs, language, and behavior. New York:
Prentice-Hall.
Moulton, William G. 1985. "On the use of Valence' in foreign language teach­
ing." In Jankowsky, ed, 1985.63-72.
: 1986. "Contrastive sentence perspective: English and German." In
Joshua A. Fishman et al, eds, The Fergusonian Impact. (Berlin: Mouton de
Gruyter), vol 1 pp 438-53.
Mulder, Jan W. F., and Sándor G. J. Hervey. 1980. The strategy of linguistics.
Edinburgh: Scottish Academic Press.
Nida, Eugene A. 1948. "The identification of morphemes." Language
24.414-41. Repr Joos, ed, 1957b.255-71.
: 1949. Morphology, the descriptive analysis of words. 2nd ed, Ann Arbor
Ml: University of Michigan Press.
. 1975. Componential analysis of meaning: An introduction to semantic
structures. The Hague: Mouton.
Palmer, Harold E., and F. G. Biandford. 1924. A grammar of spoken English on
a strictly phonetic basis. Cambridge: W. Heffer & Sons Ltd. 2nd ed, revised,
1939.
Peters, Robert A. 1985. "Historical development of noun plural -(e)s. Journal
of English Linguistics 18.25-32.
Pike, Kenneth L. 1947. "Grammatical prerequisites to phonemic analysis."
Word 3.155-72.
: 1952. "More on grammatical prerequisites." Word 8.106-21.
REFERENCES 169

-. 1954. Language in relation to a unified theory of the structure of human

behavior. Part I. Preliminary ed, Glendale CA: Summer Institute of
Linguistics.
-. 1959. "language as particle, wave, and field." The Texas Quarterly
2:2.37-54.
-. 1982. Linguistic concepts: An introduction to tagmemics. Lincoln NE:
University of Nebraska Press.
Pittman, Richard S. 1948. "Nuclear structures in linguistics." Language
24.287-92; repr Joos, ed, 1957.275-8.
Postal, Paul M. 1964. Constituent structure: A study of contemporary modets of
syntactic description. Bloomington IN: Indiana University Research Center
in Anthropology, Folklore, and Linguistics, Publication 30.
-. 1986. Studies of passive clauses. Albany NY: SUNY Press.
Quine, Willard V. 0. 1940. Mathematical logic. New York: Norton.
Quirk, Randolph, Sidney Greenbaum, GeofFrey Leach, and Jan Svartvik. 1985. A
Comprehensive grammar of the English language. London: Longman.
Read, Allen Walker. 1949. "English words with constituent elements having
independent semantic value." In Thomas A. Kirby and H. B. Woolf, eds,
PNIologica: The Malone anniversary studies. (Baltimore MD: The Johns
Hopkins Press) pp 306-12.
: 1962. "Family words in English." American Speech 37.5-12.
Reich, Peter A. ed. 1976. The second Locus forum: 1975. Columbia SC:
Hornbeam Press, Inc.
Russell, I. Willis, and Mary Gray Porter. 1982. "Among the new words."
American Speech 57:3.204-7.
Sapir, Edward. 1921. Language: An introduction to the study of speech. New
York: Harcourt, Brace & World.
: 1925. "Sound patterns in language." Language 1.37-51. Repr Mandel­
baum, ed, 1949.33-45; Joos, ed, 1957.19-25; V. Makkkai, ed, 1972.13-21.
; 1933. "La réalité psychologique des phonèmes." Journal de Psychologie
Normale et Pathotogigue 30.247-65. Repr (in English) Mandelbaum, ed,
1949.46-60; V. Makkai, ed, 1972.22-31.
Saussure, Ferdinand de. 1916. Cours de linguistique générale. Publiée par
Charles Bally et Albert Sechehaye. Second ed, 1922. References are to:
Edition critique préparée par Tullio de Mauro. Paris: Payot, 1980.
Schütz, Albert J. 1985. The Fijian language. Honolulu HI: University of Hawaii
Press.
Scott, Norman  1948. "A study of the phonetics of Fijian. Bulletin of the
School of Oriental Studies, University of London 12.737-52.
Shannon, Claude L. 1949. "The mathematical theory of communication." In
Claude L. Shannon and Warren Weaver, The mathematical theory of
communication. (Urbana IL: University of Illinois Press) pp 1-91.
170 REFERENCES

Sievers, Eduard. 1876. Grundzüge der Lautphysiologie. Leipzig: Breitkopf und

Härtel.
Steinberg, Leo. 1980. "A corner of the Last Judgment? Daedalus 109.207-73.
Steinthal, Hey mann. 1871. Abriss der Sprachwissenschaft. Erster teil. Die
sprache in allgemeinen. Also with separate title: Einleitung in die psy-
chologie und Sprachwissenschaft. Berlin: Dümmler. 2nd ed, 1881.
: 1893a. Charakteristik der hauptsäehlichsten typen des Sprachbaues.
Second ed, revised by F. Misteli. Berlin.
Stokoe, William C. Jr. 1960. Sign language structure. Studies in Linguistics,
Occasional Papers 8.
·. 1966. "Linguistic description of sign language." Georgetown University
Monograph Series in Languages and Linguistics 19.243-50.
Straight, H. Stephen. 1976. "Comprehension versus production in linguistic
theory." Foundations of language 14:525-40.
: 1980a. "Auditory versus articulatory phonological processes and their
development in children." In Helen G. Yenikomshian, J. F. Kavanagh, and
Charles A. Ferguson, eds., Child phonology, vol. 1, Production. (New York:
Academic Press, 1980) pp 43-71.
-. 1980b. "Structural commonalities between comprehension and produc­
tion." Revue de phonétique appliquée 55/56.313-6. Repr F. Lowenthal et al.,
eds, Language and /anguage acquisition. (New York: Plenum Press, 1980)
pp 177-80.
. 1986. "Theimportance and irreducibility of the comprehension/pro­
duction dialectic." in McGregor, ed, 1986.69-90.
Strawson, Peter F. 1974. Subject and predicate in Logic and grammar. London:
Methuen.
Sturtevant, Edgar H. 1949. Obituary of Leonard Bloomfield. The American
Philosophical Society Yearbook 1949.302-5. Philadelphia: 1950.
Swadesh, Morris, and Carl F. Voegelin. 1939. "A problem in phonological
alternation." Language 15.1-10. Repr Joos, ed, 1957.88-92.
Tesnière, Lucien. 1953. Esquisse dune syntaxe structurale. Paris: Klincksieck.
. 1959. Eléments de syntaxe structurale. Paris: Klincksîeck. 2nd ed, 1966.
Trager, George L. 1958. "Paralanguage: A first approximation." Studies in
Linguistics 13.1-12. Repr Dell H. Hymes, ed, Language in culture and
society: A reader in linguistics and anthropology. (New York: Harper &
Row, 1960) pp 274-9.
Trager, George L., and Henry Lee Smith Jr. 1951. An outline of English
structure. Studies in Linguistics, Occasional Papers 3.
Trubetzkoy, Nikolaj S. 1938. Grundzüge der phonologie. Prague: Travaux du
Cercle Linguistique de Prague, No. 7.
Twaddell, W. Freeman. 1983. "The study of syntax: past, present and future."
In Agard et al, eds, 1983.35-48.
Voegelin, Carl F., and Florence M. Voegelin. 1963. "On the history of struc­
turalizing in 20th century America." Anthropological Linguistics 5.12-37.
REFERENCES 171

Vygotsky, Lev Semenovich. 1934. Thought and language. References are to

the English translation by Eugenia Hanfmann and Gertrude Vakar,
Cambridge MA: The M.I.T. Press, 1962.
Wells, Rulon S. 1947a. Review of The intonation of American English by
Kenneth L. Pike. Language 23.255-73.
. 1947b. "Immediate constituents." Language 23.81-117. Repr Joos, ed,
1957b.186-207.
. 1963. "Some neglected opportunities in descriptive linguistics." Anthro­
pological Linguistics 5.38-49.
Wescott, Roger W. 1980. Sound and sense: Linguistic essays on phonosemic
subjects. Lake Bluff IL: Jupiter Press (Edward Sapir Monograph Series in
Language, Culture, and Cognition, no. 8).
Whitney, William Dwight. 1867. Language and the study of language. New York:
Scribner's. References are to the fifth ed, 1884.
. 1872. "Steinthal on the origin of language." North American Review
114.272-308. Repr as "Steinthal and the psychological theory of language"
in his Oiental and linguistic studies. 2 vols. (New York: Scribner,
Armstrong, 1873-4), vol 2 pp 332-75.
Woolley, Dale E. 1977a. "Iconic aspects of language." In Di Pietro and Blansitt,
eds, 1976.287-91.
. 1877b. "Homonymity, ambiguity, and grammatical theory." In A. Makkai
et al, eds, 1977.424-39.
Yngve, Victor. 1961. "The depth hypothesis." In Jakobson, ed, 1961.130-8.
. 1970. "I forgot what I was going to say." In M. A. Campbell et al, eds,
Papers from the sixth regional meeting, Chicago Linguistic Society.
(Chicago IL: Department of Linguistics, University of Chicago) pp 688-98.
. 1983. "Bloomfield's fundamental assumption of linguistics." In Morreall,
ed, 1983.137-45.
. 1984. "The law of componential partitioning." In Alan Manning Pierre
Martin, and Kim McCalla, eds., The tenth Locus forum: 1983 (Columbia SC:
Hornbeam Press, Inc.) pp 108-16.
. 1985. "Concepts of system in science and in linguistics." In Robert A.
Hall Jr, ed. The eleventh Locus form: 1984. (Columbia SC: Hornbeam Press,
Inc.) pp 63-70.

Aarsleff, Hans 135 162
abbreviation 73-5 89 149
151-2
accidence 134
acoustic image 110
acrostics 7
act vs system 157
activity, internal and external
116
actor (syntax) 23 24 111 138
161; and action 128
addressee 136
addresses on envelopes,
syntax of 7-8
adjacency requirement for
immediate constituents 138
advertising 68 148
Agard, Frederick Β ν 162
agent (syntax) 138
agglutination 75 78 81-2 96
149
agglutinative analog 83-4
Agglutinative Fraud, the
Great 82-4
Algeo, John 148 162
Algonquian languages 48-9 80
81 144 150-1
Allen, Grade 144
allophones 48
allusions 66-7 68
almost Ys exact sameness
98-9 157
alternation 79 80,
morpholexical 88
ambiguity 17-24 26 36, and
analysis 137, in writing
30-1
American Indian languages
144
American Sign Language for
the Deaf 7 135
American structuralism 133
American theory 65 132 143
INDEX
References are to pages.
Ameslan 7 135
analogical creation 127
analogically conditioned trial
and error 127
analogy 127 150
analytic technique 132
anchored idiom 67
Anderson, Stephen R 134 144
152 162
angle of approach in this
essay 1
announcerese 32 141
antecedent 23
Anttila, Raimo 152 162
Appelfunktion 158
apposition 30
arbitrary vs iconic 9
architecture, dimensionality
of 7
Aristotle 102
Arms, David George 150 162
Arrhenius, Svante 77
articulation, double 148
articulatory phonetics 95
associatifs, rapports 95
associations, secondary 86-7;
weak 93, vhy "secondary"?
87
assonance 94 109-10
atomic morpheme theory 66
149-52 153 see also
morphemic theory
atomism 91; YS gestalt 143
145
atonic forms 35 58-9
attitude in this essay 1
attraction 73
attnbutive 23
auditory feedback 39
Austerlitz, Robert 165
Austin, William M 165
automatic responses 114-5
127
automatic vs nonautomatic
alternations 79-80
Avogadro, Amedeo 77
Bach,Emmon164
Bally, Charles 169
Bar-Hillel, Jehoshua 131 162
Barbe, Katharina v
base of articulation 142
basic form theoretical 80
Baudouin de Courtenay, Jan
79 88 134 152
Baaell, Charles  145 153 157
162
bee dancing 61-2 146
behavior YS culture 157
behaviorism 106 160,
Bloomfield's 158
bilingual 111
binary oppositions,
phonological 153
binary requirement for
immediate constituents 138
Birdwhistell, Ray L 140 162
Biandford, F G 145 168
Blansitt, Edward L Jr v 137
162 164
blends 71 136 147
Bloomfield, Leonard 1 2 49 50
51 53 70 73 80 81 88 90
95 96 97-8 100 102 103
105-6 107 131 132 133 138
141 144 145 146 148 149
150-1 152 153 154 158-0 160
162-3
Bloomfield definition of word
50 54 144
Bloomfieldian language
description 133
Bloomfield's behaviorism 158
Bloomfield's fundamental
assumption 97 100 102
INDEX
Bloomfield's working principle Christie, William M Jr v 148
in analysis 103 141
Boas, Frana 131 132 163
Bodine, Ann ν
Bolinger, Dwight L v 141 153
157 163
Bopp, Franz 77-9
borrowing 45-7
botany 134
bound forms 89
bound markers 140
boundaries from segments,
inferring 51 54-6 145
boundaries, fuzzy or sharp 27
28 45 52 97 98 102 119 140 communicative package 26 91
157
breakfast-food names 76
British theory 137 153
Brown, J Marvin v
Buhler, Karl 158
Campbell M A 171
canonical or typical shapes of
forms, influence of 51
careless speech 38-9
Carroll, Lewis 7 8 12 28 126
141
case Í41
causative 138
Chafe, Wallace 153 163
chance and choice 114-5 127
Chao, Yuenren vi 42 133 137 connections at a distance,
140-1 141 142 152 160 163
characterization vs definition
54
charades 7
chemical nomenclature 75
149
Chen, Lily v
Chenoweth, Vida 161 163
167
child language 88
Chinese characters 28 44
Chinese language 18 93 137
138 141 151-2 161
choice and chance 114-5
Chomsky, Noam 1 2 22 23
131 132 134 137 153 157
159 168
Christie, Agatha 30 66
152 157
cinema, ¿lent 7
clarity norm 44 52 57 146; vs
frequency norm 146
clause 134 153
code vs message 104 157
code-switching 43-4 47
cohesion, degrees of 52 145
coining words 118
collective vs idiosyncratic 157
comma, serial 135
comment and topic (syntax)
30 141
140
communicative redundancy 61
see a/so redundancy
communicative situation 26
communicative systems,
dimensionality of 5-8
competence vs performance
101 134
components of language (=
subsystems) 85
components, phonological 48
composing music 130 161
composite form 138
concept 110; defined 118-9
Condillac, Etienne Bonnot
135
syntactic 139
connotations 159;
classification of 108
constituents 138; and
markers 137; of utterance
27
constitute 138
construct 138
construction types 139
constructions 7 22-4 138, and
forms 99; misuse of term
138; number of in a
language 139 157
construing 90, and parsing 18
96 137, and syntagmatics
96
consultative prose 92
contamination 72-73 148-9
173
content words 140
contentiyes 108; and functors
27-8 29 34 35 140
context 37 108; defined 27
contextual allusions 66-7
continuous vs discrete 154
contractions 58
convenience theory of
morphophonemes 8 3
conversion principles or rules
(between strata) 84
Copeland, James B v 146 166
coreference 30
cosmetics, names of 76
Cosper, Ronald v
cover terms 118
covert vs overt behavior 160
covert Ys overt editing 130
Cowan, J Milton 142 143 166
crans 70 74-5 148
creating sentences 129
creating words 118
creative acts 115
creative work, internal YS
external 116
creativity 122 159; and
innateness 160
Cree language 144
criteria, multiple 145
Crosland, Maurice P 149
cross-language resemblances
93
crossword puzzles 7 109
culminative and demarcative
145
culminators 54
cultural frame of reference
26-7 35
cultural themes 77 157
culture YS behavior 157
Cummings-Culliton, Elizabeth
v
Dalton, John 77
Darwin, Charles Robert 77
Davis, Philip W v 146
dead metaphor 68
deafness 37 57
Decade of the Morpheme 81
86 95 153
174
Decade of the Phoneme 81
Dedekind, Julius Wilhelm
Richard 77
deep and surface grammar 136
139
deep structure 3 16-7 133
136-7
definition vs characterization
54
definition, real and nominal
101-2
definitions, traditional YS TG
134
Dell, Gary S 148 164
demarcative and culminative
145
demarcators 54 55
demonstration of creating
122-6
denotation 104-105; vs
connotation 108-10; YS
reference 103-5
depth perception 16
describing languages 2 66 100
133
descriptive etymology 154
DiPietro Robert J v 164
diachronic and synchronic 94
152 157
dialect and idiolect 157
dialect coloring 142
dialect diferences 37-8 43-4
47
dialect divergence 103
dialects, family 68
dictionary 100 112; and
encyclopedia 102
different and same 97
Dil, A n w S 163
Diller, Karl Conrad 131
dimension 134
dimensionality 129 159; and
iconicity 10; of speech 135;
syntactic 4 34 135-6
direction 5
disambiguation 22-4 29-30
137
discrete and continuous 154
discreteness of sound
repertory 48 63-4
INDEX
distinctive features 48 94-5 false friends 102 106
distortion of pronunciation 38 falsehoods 90 120-1
divergence of dialects 103 familiarity, effect of 92
doggerel 110 family dialects 38 68
double articulation 146 Fant,  Gunnar 166
double takes, syntactic 19-21 feedback 39 114-5
doublets 46 72-3 Ferguson, Charles A 170
drawing 6 field procedure 132
dreaming 116 Fijian language 18-9 24 27 57
drug names 78 137 141 145 150
drum signaling, West African Fillmore, Charles 138 141
6 164
duality of patterning 59 60 Firth, John Rupert 137 140
146 Fischer-Jørgensen, Eli 144 164
Fishman, Joshua A 168
Ebb, Fred 66 floating idiom 67
eclipsing stance 1 131 focus of attention in different
editing 127-9 160; overt and linguistic traditions 132-3
covert 130 152-3
Edmiston, Cynthia'ν
folk etymology 93
Ehrenfels, Christian von 77
football American 135;
emic and etic 89 154
signals 7
empirical approach 1 131
forgetting 160; words 117-8
empiricism vS rationalism 131
form: and (Saussurean) sign
encyclopedia and dictionary
144; and meaning 117; and
102 shape 144
English, written 29-32 52 135 form of information to be
envelope addresses, syntax of worded 120
7-8 forms 50 90 110; and
errors in reading 30-1; in constructions 99
reading aloud 32 Fought, John 133 166
Espy, Willard 135 164 foundations of linguistics,
état de langue 157 158 found where? 1
etic and emic 89 154 Fox language 81 144
event vs institution 157 Francis, W Nelson 132
evidence for the hearer 26-36, 139-40 164
classified 26-7 Fraud, the Great
evocation of meaning 90 Agglutinative 82-4
evolution of language 78 146 free association 116
example of creating 122-6 free form 50
exogenous 160 freedom of choice in speech
expectation of hearer 91-2 situation 121-2
experiential connotations 108 Frei, Henn 140
109 French language 28 93 117
explicit motor-matching 146
39 40 frequency norm and clarity
external sandhi 145 norm 146
external vs internal behavior Fries, Charles Carpenter 140
116 160 141 164
Fries, Peter Η 140 164
faded metaphor 68
INDEX
Frisch, Kurt von 146 164
Fromkin, Victoria A 148 164
function vords 140
functional simultaneity 136
functional-systemic theory
137 153
functors 108; and contentives
27-8 29 34 35 140; not
highlighted in speech 35
141-2
fundamental assumption of
descriptive linguistics 97
100
fundamental question about
speaking 114
fundamental question of this
essay 2 29 114 132;
reworded 15 114
fuzzy boundaries 27 45 52 see
also boundaries
Gair, James W γ
Galanter, Eugene 168
garden-path jokes 42 143
Garvin, Paul L v
generative phonology 132
153-4, natural 88
generic drug names 76
genetics 160
Georgian language 141
German language 17 34 93
Gershwin, Ira 126
gerund 23
gestalt 40-2 47 51 54-6 57 77
142-3 144 145 146, vs
atomism or granularity 77
143 152
gestalt closure in hearing 57
gestalt learning 40-1
gibbons 58
Gilbert, Sir William S 60
Gleason, H Allen Jr 132 164
goal (syntax) 22-3 24 111 128
138 161
goal-directed behavior 114-5
130
grammar 81 132 134; and
phonology 56; and
semantics 17-8; theory of
independence of 131
grammatical aura or field 110
grammatical-lexical aspect of
language 90 91
granularity YS gestalt 77 143
152
graphic play 7
graphic shape of word 110
Great Agglutinative Fraud,
the 82-4
Greek grammatical tradition
161
Greek language 4
Greenbaum, Sydney 169
Grevisse, Maurice 133 164
Griffin, Toby D v
Grimm, Jakob 77
group genitive 144
group mind 159
Hall, Robert A Jr v 132 133
164-5 171
Halle, Morris 154 154-5 165
166
Halliday, Michael A K v 128
139 141 142 150 165
Hamp, Eric P v 153 165
Hanf mann, Eugenia 171
Hardy, Don v
Harms, Robert T 164
hat game 33-4
headlines, ambiguities in 32-3 idioms 147
141
hearer 111 129 132 139-40
hearer construing 19 90
hearer parsing and construing immediate constituents 22 50
of a vord 112
hearer's evidence 26-36;
classified 26-7
hearer's view 2 40-50 144 146
hearing and listening 136
hearing and speaking, theory
of 2 101 114-5
hearing, holistic 58
hearing, summary of 114
Heilmann, Luigi 167
Herbart, Johann Friedrich 77
Hervey, Sándor G J 136 151
168
hierarchy of purposes 126
Hill, Archibald A 164
175
historical and descriptive
102-3
historical-comparative lin-
guistics 131
history and morphophonemics
82-3
history, synchronic relevance
of 94
HjelmsleY, Louis 141
Hoenigswald, Henry M v
hoUov cube 16 136
homophony 61
Hooper, Joan  88 166
Householder, Fred W 154 160
165 166
Huang, Lillian v
Hutton, James 78
Hymes, Dell v 133 166 170
iconic and arbitrary 9
iconicity. and dimensionality
10; in language 10-2
identifications, making 97
identifying words 37-47
idiolect, defined 94
idiolectal divergence 103 106
157
idiomatic expressions or
phrases 37 51 67-68 86-7
90 96 142
idiosyncratic vs collective 157
imitating nonspeech sounds
47
138 144
implicit motor-matching
39-40 47 142 160
individual diferences 92 101
159
individual vs social 99-103 106
157
inferring boundaries from
segments 51 54-6
inferring segments from
boundaries 54
infinitive, split 13 136
infix 71
inflection 134 149-50
inflectional elements 141
176 INDEX

Katz, Jerrold J 131 133 167 Leland, Waldo Gifford 163

information, transmission of
90 Kaufifmann, Lane v letters and words 77
innate 160 Kavanagh, J F 170 levels 84, and units 84-5;
Kekule von Stradowitz, mixing of 56
innateness and creativity 160
Friedrich August 77 lexico-grammatical structure,
inner speech or voice 116 160
Keikar, Ashok v 136 formal or official 87-8
institution vs event 157
Kelley, Gerald Β ν 162 lexico-grammatical theory 65
intercalibration 103 106 157
Key, Mary Ritchie 140 167 see also grammatical
internal and external behavior
Kilbury, James v 144 152 167 lexicographic analysis 103
106 116 160 lexicon 91 134
kinesics 27 140
internal morphological lie 90 120-1
kinesthetic feedback 39
changes (ablaut), theory of
Kipling, Rudyard 20 limerick 122-3
origin of 78
Kirby, Thomas A 169 linearity 23 86 129; defined
internal reconstruction 152
Kitchin, Aileen Traver 140 6, of speech 135
internal sandhi 80
klang-association 109 linguistic sign 110 112
internal storage (in a person)
knowledge of forms and linguistics and literature 159
110-3 159
meanings 107 110 linguistics and psychology 2
internal work space (in a
Kruszewski, Mikołaj 79 linguistics defined narrowly
person) 112-3
Kund gabe funktion 158 or broadly 2
internalization 116 160 listener construing 25
intonation 27 29 128 140 Kuryłowics, Jerzy 65 82 147
167 listening: and hearing 136; by
intransitive 138 prediction 42; for word
introductions 41-2 44 LaboY, William 49 144 167 identity 37 47; for word
introspection 116 Ladd, D Robert Jr v 140 167 shape 44 47 52; modes of
invariant 81 Lamb, Sydney Μ ν 154 37 44 46-7 47 91-2 142 143
irrelevance of information to
language 134 literature 91-2; and lin­
be worded 121 language change, summarized guistics 159
irrelevant questions in history
103 live metaphor 68 148
of linguistics 84 language, dialect, idiolect Livingston, Helen 165
isolating 78 81-92 152 100-1 location, mnemonics of 7
isomorphism 65 82 147 language learning 63 80 91 Lockwood, David 154 167
item and arrangement 137 93 132 locus existend i 133; of
item and process 137 language, multidimensional language 157 159
itsy-bitsy words 28 32 13-5 logic 97-99; terminology of
language typology 149 162 158
Jabbenwocky28 140-1
logic approach 1 131 132-3
Jakobson, Roman 144 153 160 langue vS parole 101 134
lapses 71-2 119 147 148 Longacre, Robert E 161 167
166
Lashley, Karl S 135 160 167 losing words 118
Jankowsky, Kurt R 134 167
Last Judgment, The (fresco) loss for words, being at a 119
Japanese language 44
135-6 Lounsbury, Floyd 83-4 167
Japanese tinting 28
Latin language 4 18 Lowenthal, F 170
Jeffress, Lloyd A 167
Leach, Geoffrey 169 Lurie, Alison 71
Jespersen, Jens Otto 
leading theme (Leitmotif) in Lurie, Nancy  144 146
132 137 149 167
music 98-9 Lyell, Charles 78
jokes 42 65 92 143
learning from parts-to-whole Lyons, John vi
Joos, Martin 17 18 22 116
41 43; from whole-to-parts
137 153 154 157 160 167 Madison Avenue 68 75-6
(by gestalt) 40-1
Jung, Karl 109 learning language 63 88 91 Mailman, Daniel v
kana and kanji 28 93 132; second language 93 Makkai, Adam v 148 154 162
Kandor, Joseph 66 Lehmann, Winfred Ρ γ 167
INDEX
Makkai, Valerie Becker ν 144
162 167
malapropisms 72 148
Malayo-Polynesian languages
150
male chauvinism 31 31-2
Mandelbaum, David 168
manifestation vs system 134
Manning, Alan 171
marginal phenomena, mor­
phemic 154
markere 13 141, and con­
stituents 137, bound 140
Martha's Vineyard 49-50 144
Martin, Pierre 171
Martinet, Andre 140 142 146
168
masking 91 108
mathematical notation 6-7
mathematics 98 130 154
Mauro, Tullio de 169
McOalla, Kim 171
McDavid, Raven loor 46 144
McGregor, Graham 132 136
168
McKaughan, Hovard 141 168
meaning 3 9 65 90-5 130 158;
and form 117; and shape
103; dissected 10343,
evocation of 90; of speech
sounds 94-5
meaning-conveying machinery
65
Meillet, Antoine 153 168
Meillet definition of sentence
153
Mel'čuk, Igor 141 168
memorization 111
memory: short-term 42, verbal
and nonverbal 120
Menominee language 80 88
150-1 154
mental activity 106
mental images 105-8
mention and use 136
Merrifield, William R v
message vs code 104 157
metanalysis 73-4
metaphor 148, faded vs live
68
Michelangelo 135
Miller, George A 132 136
160 168
Milne, A A 60-1
mime see pantomime
mind's eye 116
minimality 89-90
minimum free form 50
Miranda, Rocky V v
Misteli, Franz 149
Mitchell, Douglas v
mnemonics of location 7
modes of listening 37 44
46-7 47 91-92 142 143 154
modification (syntax) 22-3
moneme 27 140
morpheme 27 65 70 82 140
153; in America 81-2,
questions about 82;
stability of 153-4
morphemic and morphological
theory 65-6 74 147
morphemic theory: discarding
87-8; supplementing 86
morphemics 89; history of
77-85
morpholexical alternations 88
154
morphology 3 50 134, and
syntax 81 89
morphology, syntax, syntag-
matics, and paradigmatics
95-6
morphophonemes 80 134;
theories of 8 3 153
morphophonemics 134; and
history 82-3; and morpho-
phonemes 82; and
phonemics 84 95; history of
79-86 144 152
Monreali, John 168
Morris, Charles W 158 168
Morse code 85
motor-matching, implicit
39-40 47 142 160
Moulton, William Gamaliel γ
159 160 168
Mulder, Jan W F 136 151
168
177
multidimensional language
13-5
multilingual purism 93-4
music 98-99 130 161;
composing 130 161
musical analysis of poetic
rhythm 122-3 160
names personal 111-112 117
149
naming 75-6 118-9
natural generative phonology
88
natural selection 160
Necker cube 16 136
neo-Bloomfieldian 133
neogrammarians 146
Nida, Bugene A 22 104 137
138 154 168
nineteenth-century linguistics
131 152-3
noise and redundancy 63
nomenclature of chemistry 75
149
nominal and real definitions
101-2
nonautomatic vs automatic
alternation 79-80
nonconstituent connectives
137
nonpatent vs patent alter­
nation 79-80
nonspeech sounds, imitating
47
nonverbal memory 120
norm shapes of words 44-5 48
Northrop, Filmer S  165
notation, mathematical 6-7
noun and verb 141 161
object (syntax) 22-3 24 138
occurrence 110
Olmsted, David L v
onomatopoeia 47 68-71 138
144
operationalism 132
order of elements 4-5 see
also -word order
organic behavior 114
178
orientations of schools of
linguistics compared 132-3
152-3
original language 78
overlap of phonemes 63
overloaded utterances 21
overt motor-matching 39-40
overt vs covert 160; editing
130
packets in internal storage
110 111 117
Palmer, Harold E 145 168
pantomime (= British mime)
7 phonesthematics 154
paradigmatic connotations
108 109
paradigmatics and parsing 96
paradigmatica, syntax, mor-
phology, and syntagmatics
9«
Paradis, Michel 164
paralanguage 26-7 62 140
parallel processing 139, YS
processing in senes 85
parameters of state in speech
situation 119-12
parole vs langue 101 134
parsing and construing 18 96
137; an unfamiliar word 89
parsing and paradigmatics 96
participle 23
particle, wave, and field 142-3
particles (vords) 141
patent vs nonpatent alter­
nation 79-80
pattern or gestalt 77 142
pattern vs structure 133
142-3 157
pattern-changing events 92-3
Peacock, Dennis Ε ν
Peoria language 48-9
perception in depth 16
perceptual gestalts or recog­
nition units 64
performance YS competence
101 134
permutation of words 128
perpetual motion 63
personal YS shared 157
perspective of this essay 1
perspective, sentence 161
Peters, Robert A 152 168
phonemes 48 65 82 134
phonemic overlapping 63
phonemic principle, history of
144
phonemic principle or theory
44 48-50 62 63-4 143
phonemic transcription 143
phonemiciaing 47 58 59
phonemics 134; and morpho­
phonemics 84 95; as a
stratum 153
phonetic doggerel 110
phonetic listening 46-7 47
phonetic paradigmatic conno­
tations 108 109
phonetic shape 110
phonetic similarity 46
phoneticians 49
phonetics 95
phonological structure 90
phonological theory 143
phonology 3 50 134; and
grammar 53 56; the word in
52-6
phrase, idiomatic 51
physicalism 106
Pickett, Velma Β ν
Pike, Kenneth L v 21 53 56
91 137 142-3 145 154 157
168-9
Pittmann, Richard 138 169
Planck, Max 77
planning 116
poetry 91-2 109-10
polystructuralism 139 161
Porter, Mary Gray 148 169
post-Bloomfieldians 133 153
159
Postal, Paul M 131 132 133
138 139 169
Potawatomi language 80-1 152
potentials 110-1
pragmatics 105 158
predicate (syntax) 138
prediction in listening 42
prelanguage 60 146
INDEX
prepositions 111; as valence
bonds 161
prescription-drug names 76
Pribram, Karl H 168
principal parts and paradigms
149
probability in listening 42
problem-solving 116
processing in parallel 139; vs
in series 85
pronoun 23
proprietary drug names 76
proprietary labels 75-6
prose style 94
prosodic features 140
prosody 27 29 140; vs word
stress 32
proximity pnnciple 12-3 61-2
psychiatry 109
psychological subject 30
psychology 2; clinical 109, of
langi+1duage 133
Pulgram, Ernst v
punctuation 29-30
puns 68 91
purism, multilingual 93-4
purposes 115 126; of speaking
115
quantization 154
question, fundamental, of this
essay 2 29 114 132;
reworded 15 114
Quine, Willard V 136 169
Quirk, Randolph 133 169
randomness 114-5
rapports associatifs et syn-
tag matiques 95
Rask, Rasmus 77
rationalism YS empiricism 131
rationality 135
Read, Alan Walker 155 169
reading 40-41; errors in 30-1
32
real and nominal definitions
101-2
reality theory of morpho-
phonemes 83 153
rearranging words 128-9
INDEX
recalling words 117-8
recipient (syntax) 128 161
recitation 120
recognition units 51 54 64
recognition-unit criterion for
word 51 54
recrystalization 58 59
redundancy 42 60-3 64 68 91
93 146 153; and noise 63
redundancy pnnciple 61-2
reference 104-5; vs denotation
103-5
reflexivity (logic) 98
Reich, Peter A 169
repatterning 91 111
rephonemicizing 59
repulsion between words 7 3
rereading: and rescanning 92
154; and rewriting 160
resemblances, accidental 87-8
reshaping 59-60 148-9
resonance 88 89 90 95;
between languages 93;
channels 111; from
nonforms 94-5; tests of
155-6
resonance effects as conno­
tations 108
resonance theory of mor­
phology 97-103
restressed forms 58
restructuring (= repattern-
ing) 91 111
reverberations between lan­
guages 93
rewrite rule (TG) 138
rewriting and rereading 160
rhyme 94 109-10
rhythm 94 109-10 168
rhythmic patterns 122-3
Ring des Nibelungen 98
Rister, Alan v
Romanization 143
Russell, I Willis 148 169
same and different 97 98
sameness, almost 98-9
Samoan language 80 152
sandhi 80 145
Sapir, Edward 48 131 132
133 144 149 156 169
Saporta, Sol 167
Satz phonologie 52-3
Saussure, Ferdinand de 94 100
104 112 133 134 135 148
144 151 152 153 157-8 159
169
scale of creative goals and
achievements 130
Schleicher, August 134
Schütz, Albert J v 137 150
169
Scott, Norman  145 169
sculpture, dimensionality of
7 170
searching for words 117-9 127
Sechehaye, Albert 169
second-language learning 93
secondary associations 68-71
86-7 148
segmental forms 140
segments from boundaries,
inferring 54 145
semantic doggerel 110
semantic paradigmatic conno-
tations 108
semantics 158; and grammar
17-8; and syntax 102 131
161
sentence 90 134 153; and
utterance 153; Meillet
definition of 153
sentence perspective 161
sentence stress 128; and word
stress 140
sentry listening 37
serial comma 135
series Ys parallel processing
85
sexism 31 31-2
Shannon, Claude 146 169
shape and form 144
shape and meaning 103
shared vs personal 157
Sievers, Eduard 142 170
sign language (for the deaf) 7
135
sign, linguistic (Saussurian)
110 143; and form 144
179
sign system (Morris) 158
silent screen 7
silent speech 115 116
simultaneity, functional 136
Sinatra, Frank 66
slips of the tongue 119 147
148
slurred speech 58
slurring and recrystalization
64
slurvian 57-60 146; basis of
62-3; native material 60
small size of sound repertory
48 63-4
Smith, Henry L Jr 30 140 141
Smith, M Estellie 165
social vs individual 99-103 106
157
Somerville, Mary Fairfax 112
sound alike and sound
different 58-6
sound aspect of language 65
sound change 146
sound symbolism 148
source of information to be
worded 121
spacing of elements 5 52 145
Spanish language 88; verb
inflection 149
speaker 2 132 159-60
speaker's view 2
speaking and hearing 114-5
speech communities 100-1
speech signal 85
speech sounds 48 94-5
spelling for word identifi­
cation 44
split infinitive 13 136
stances in linguistic theory 1
131
state parameters in speaking
119-12
Steinbeck, John 54-5
Steinberg, Leo 135 170
Steinthal, Heymann 1 131 149
170
Stimson, Hugh M v
Stokoe, William  135 170
Stout, Rex 30
180
Straight, H Stephen 132 170
strategies for vord identi-
fication 39
stratificational theory 147
154
stratum 153 see also level
Stravson, Peter F 131 170
stress 27 51; patterns of 31;
vord and sentence 140
structuralism 133
structure vs pattern 142-3
157
structure-in-depth 16-7 25 30
i l l 137
Sturtevant, Edgar H 133 170
subject (syntax) 23 24 30 138
subject, psychological 30
subject-predicate (syntax) 138
subject-verb (syntax) 138 161
supersaturated utterances 21
surface and deep grammar 136
surface structure 18
Svartvik, Jan 169
Swadesh, Morris 79 170
switching-codes 43-4 47
symmetric (logic) 98
synchronic and diachronic 94
152 157
synonyms 31
syntactic dimensionality 4 34
135-6
syntactic marking 141
syntactic properties 110-11
syntactic relations or ties
110-11 138
syntactic sense 17-8
syntactics (Morris) 158
syntagm 138
syntagmatic connotations 108
109
syntagmatics and construing
96
syntax 3 50 134; and
morphology 81 89; and
semantics 102 131 161,
dimensions of 4 34 135-6
syntax, morphology, syntag­
matics, and paradigmatics
95-6
syntherne 142
INDEX
system 133; different senses transitivity (logic) 98
of term 157-8; vs act or transitivity (syntax) 141
manifestation 134 157 transitivization (syntax) 111
system-changing events 92-3 161
systemic-functional theory transmission of information
137 153 90
trial and error 122-6 126-7
taboo 73 149
Trubetzkoy, Nikolaj S 54 144
tabula rasa 160 145 158 170
taxonomic 138
truth 90
tense and time 17-8 Tübatulabal language 80
tension betveen patterns 13
Twaddell, W Freeman v 132
terminology 2-3, defined 170
118-9, and theory 112-3
two-dimensional syntax 135-6
Tesnière, Lucien 136 159 170 Tyler, Stephen v
TG see transformational-
typical or canonical shapes of
generative theory forms, influence of 51
themes, cultural 77 157
typology of languages 149 152
theoretical basic form 80
theoretical underlying forms Ułaszyn, H 79 152
81 83 84 unconstruable utterances 21
theories of morphophonemes 63
83 underlying 80 129 139
theorizing vs thinking 107-8 underlying forms: in morphol­
135 158 ogy 81 83 84, in syntax 84
theory and terminology 112-3 uniformitarianism 78-9
thermodynamics 146-7 uniqueness of analysis,
things and vords 104 expectation of 65 89
thinking vs theorizing 107-8 units and levels 84-5
135 158 universal phonetic alphabet
time and tense 17-8 153-4
Timm, Lenora A v urgency of speech situation
121
Titchener, Edvard Bradford
use and mention 136
160
user 2 132
tones 143
utterance 26-7; and sentence
tongue-twisters 67-8 153, constituents of 27-8
topic and comment (syntax)
30 141 Vakar, Gertrude 170
Trager, George L 140 141 157 valence 111 159 161
170 Vanderslice, Ralph γ
transcription, phonemic 143 Vennemann, Theo 88
transducer fallacy 84-6 153-4 verb 111 138, and noun 141
161 verb inflection: English
transducers 84-5
149-50, Spanish 149
transformational-generative verb-object (syntax) 161
theory 84 85 88 132-3 138 verbal memory 120
153 159
Verne, Jules 135
transformations 3 23 128-9 verse 94 160
139 161
via rules 88
transforms 161 Virchov, Rudolph 77
INDEX
vocal gestures 60
Voegelin, Carl F v 79-80 81
131 179
Voegelin, Florence Μ ν 131
170
voice (syntax) 141
voice quality 142 38 57
vowel color 58
Vygotsky, Lev S 135 160 171
Wagner, Richard 98 104
Walker, Willard ν
Wallace, Stephen ν
Watson, John Broadus 160
Waugh, Linda R v 144 160
166
Weaver, Warren 169
Wells, Rulon S 21 53 56 132
133 137 138 144 171
Wescott, Roger W v 148 171
Whitney, William Dwight 1
94 101-2 131 152 171
Winnie the Pooh 60-1
Wodehouse, P G 59
Wolff, John U v
Woolf, H  169
Woolley, Dale E 137 148 171
word 3 27 50-2 89 143 152-3;
defined 144, identification
of in reading 39; in
phonology 52-6; isolated 92,
neither phonological nor
grammatical 64
word-association tests 109
word formation 134
word games 24 109 119
word identification strategies
39
word identity 45; listening for
47 57
word knowledge and world
knowledge 102
word magic 103
word order 4-5 27 29
word play 109
181
word shape 44-5; alternative
45; listening for 44 47 57
word stress and sentence
stress 140
words and letters 77
words and things 104
words remembered temporar­
ily 118
work space 112-3 115-6
world knowledge and word
knowledge 102
Wort phonologie 52
writing 6
written English 29-32 52 135
Wundt, Wilhelm 77 133
Yenikomshian, Helen G 170
Yngve, Victor 132 137 160
171
Yokuts language 149
Youngquist, Mary 135
zero-dimensional syntax 33-4