Huif a

if i
ih{
Cin
 LIBRARY
STATE UNIVERSITY of NEW YORK
AGRICULTURAL and TECHNICAL
INSTITUTE
at CANTON, NEW YORK
Ot Rw 65=361h
 Digitized by the Internet Archive
in 2021 with funding from
Kahle/Austin Foundation

https://archive.org/details/languagethoughtO000unse_h4a4
FOUNDATIONS OF MODERN PSYCHOLOGY SERIES

Richard S. Lazarus, Editor

THE PSYCHOLOGICAL DEVELOPMENT OF THE CHILD, Paul H. Mussen

TESTS AND MEASUREMERTS, Leona E. Tyler

MOTIVATION AND EMOTION, Edward J. Murray

PERSONALITY ANO ADJUSTMENT, Richard S. Lazarus

GLINICAL PSYCHOLOGY, Julian B. Rotter

SENSORY BSYSHOLOGY, Conrad G. Mueller

PERCEPTION, Julian E. Hochberg

LEARNING, Sarnoff A. Mednick

LANGUAGE AND THOUGHY, John B. Carroll

SOCIAL PSYCHOLOGY, William W. Lambert and Wallace E. Lambert

PHYSIOLOGICAL PSYCHOLOGY, Philip Teitelbaum

EDUGATIQNAL PSYCHOLOGY, Donald Ross Green

THE NATURE OF PSYCHOLOGICAL INQUIRY, Ray Hyman

ORGANIZATIONAL PSYCHOLOGY, Edgar H. Schein

Roy E. Larsen Professor of Educational
Psychology, Graduate School of Education,
Harvard University; author of books and
numerous articles on the psychology and the
teaching of language, and of the Modern
Language Aptitude Test; consulting editor
of several scholarly journals in the fields
of language and educational psychology.
Language
and Thought
 be oe
Ca7

© Copyright 1964 by
PRENTICE-HALL, INC.,
Englewood Cliffs, N. J.

All rights reserved. No

part of this book may be
reproduced in any form,
by mimeograph or any
other means, without per-
mission in writing from
the publisher. Printed in
the United States of Amer-
ica. Library of Congress
Catalog Card No.: 64-
17072

LANGUAGE AND THOUGHT, Jokn B. Carroll

PRENTICE-HALL FOUNDATIONS

OF MODERN PSYCHOLOGY SERIES

Richard S. Lazarus, Editor

PRENTICE-HALL INTERNATIONAL, INC., London

PRENTICE-HALL OF AUSTRALIA, PTY., LTD., Sydney
PRENTICE-HALL OF CANADA, LTD., Toronto
PRENTICE-HALL OF INDIA PRIVATE LIMITED, New Delhi
PRENTICE-HALL OF JAPAN, INC., Tokyo
PRENTICE-HALL DE MEXICO, §$.A., Mexico City

Designed by Harry Rinehart

C-52270(p), C-52271(c)

hie aj. ‘ae

 Foundations
of Modern Psychology
Series
The tremendous growth and vitality of
psychology and its increasing fusion with
the social and biological sciences demand a new approach to teaching at the
introductory level. The basic course, geared as it usually is to a single text
that tries to skim everything—that sacrifices depth for superficial breadth
—is no longer adequate. Psychology has become too diverse for any one
man, or a few men, to write about with complete authority. The alterna-
tive, a book that ignores many essential areas in order to present more
comprehensively and effectively a particular aspect or view of psychology,
is also insufficient. For in this solution, many key areas are simply not
communicated to the student at all.
The Foundations of Modern Psychology is a new and different ap-
proach to the introductory course. The instructor is offered a series of
short volumes, each a self-contained book on the special issues, methods,
and content of a basic topic by a noted authority who is actively con-
tributing to that particular field. And taken together, the volumes cover
the full scope of psychological thought, research, and application.
The result is a series that offers the advantage of tremendous flexibility
and scope. The teacher can choose the subjects he wants to emphasize and
present them in the order he desires. And without necessarily sacrificing
breadth, he can provide the student with a much fuller treatment of
individual areas at the introductory level than is normally possible. If
he does not have time to include all the volumes in his course, he can
recommend the omitted ones as outside reading, thus covering the full
range of psychological topics.
Psychologists are becoming increasingly aware of the importance of
reaching the introductory student with high-quality, well-written, and
stimulating material, material that highlights the continuing and exciting
search for new knowledge. The Foundations of Modern Psychology Series
is our attempt to place in the hands of instructors the best textbook tools
for this purpose.
 {
ig

re lhe Ww replat
ia e s a
Sos te, PN
( Ga eh enum :
 Preface
In writing this book on the psychology
of language and thought I have tried to
make it come as close as possible to my ideal of what a presentation of this
topic should be in an introductory course in psychology. In my use of
scientific linguistics as a foundation and point of departure, I have pro-
duced a treatment that I believe you would find somewhat different from
those in typical texts. I have been quite deliberate about this, however,
because I believe that the psychological study of language and thought
requires first an accurate knowledge of what language is.
Thus, more is said about language than about thought. But this is more
a reflection of the advanced state of our knowledge about language and
the primitive state of our knowledge about thought than of my true
opinion about the relative importance of the topics. We need very much
to know more about thinking than we do. One of the major themes of this
book is that thought and cognition are presupposed by language—that
speech is a consequence of some kind of thought or cognition, even though
language structure may channel or influence thought.
The psychology of language and thought has only recently become a
well-respected and eagerly pursued field of research. I have attempted to
describe a sample of the more useful and interesting results obtained thus
far, along with an account of the theoretical developments that underlie
them. I feel certain that the psychology of language and thought as a body
of knowledge already has important implications for psychology in general,
for particular problems in education, and for everyday life. But at many
points I have had to indulge in speculation, going beyond the sure data
we have. Frequently I have had to characterize statements or ideas as
being only “probable” or “approximate” or even just “possible.” I hope
that in so doing I may arouse your interest, in the wish that some day a
more objective account of the psychology of language and thought, better
grounded in observation and experiment, can be written.

John B. Carroll

vit
 Contents

LANGUAGE AND COMMUNICATION

Page 1

Two Major Functions of Language.

Communication Svstems.

THE NATURE OF LANGUAGE

Page 8

NO The Structure of Language: Overview.

of Language: Sounds. Phonemes.
The Raw Materials
Forms. Constructions. M ethodological
Problems in Linguistic Analysis. Analysis of the Content
System of the Language. Linguistics and Psychology.

Vit
 THE LEARNING OF LANGUAGE
Page 30
a
'
The Course of Language Development in the Child. ad)
Meaning as a Problem for Psychology.
Second-Language Learning and Bilingualism.

ASPECTS OF LANGUAGE BEHAVIOR

Page 44

The Production of Speech. Some Aspects of Talking. The Construction

of Utterances. Language Behavior from a Statistical Point
of View. The Perception and Understanding of Speech. Reading.

INDIVIDUAL DIFFERENCES IN LANGUAGE BEHAVIOR

Page 66

Differences in Language Ability. Language in Mental Retardation.

Language in Aphasia. Language in Abnormal Mental States.
Stuttering. Language in Deaf Children.

Contents
1x
 COGNITION AND THINKING
Page 75

The Development of Thought in the Child. The Motivation

of Thinking. Concepts. Problem-solving as the Manipulation
of Concepts. Logic and the Psychology of Thinking.

LANGUAGE AND COGNITION

Page 89

Language and Concepts. The Role of Language in Cognitive Development.

Some Experimental Methods of Studying Verbal Mediation Processes.
The Linguistic-Relativity Hypothesis. Language
and Thought: A Final Look.

SELECTED READINGS
Page 112

INDEX
Page 114

Contents
 Language
and Communication “Speech,” wrote Benjamin

Lee Whorf,! “is the best show man puts on.” It is the task

of this book to elucidate the full meaning of this state-

ment, by describing exactly what this “show” consists of,

and by attempting to explain how man is able to “put on”

such a marvelous display. More than that, we shall try to

say how this capacity helps man in his thinking, and

suggest how it can sometimes work against his best in-

terests. We shall discuss how individuals acquire language

skills, how they differ in their facility in language, and how

speech functions are disturbed in aphasia, the psychoses,

1B. L. Whorf. Language, thought and reality. Cambridge and

New York: M.1.T.-Wiley, 1956, p. 249.
 and other disorders. All this knowledge, it is hoped, will aid you in understand-
ing the role of language and thought in your own and others’ behavior and
supply a necessary background for applications of the psychology of language
in other branches of psychology, in education, and elsewhere.
In sheer volume, speech behavior can yield impressive statistics. It may be
an interesting exercise for you to measure the average rate (words per
minute) at which you speak in spontaneous conversation, and then to estimate
the probable number of words you speak in a day, a month, or a year. A person
would not have to be a particularly talkative individual to speak a billion
words in a lifetime. A professional writer who averages 2000 written words
as his daily output would turn out 730,000 words a year; it has been reckoned
that the German psychologist Wilhelm Wundt published 53,735 printed
pages in the 68 years of his career, or about one word every two minutes,
night and day.
On the input side, the intake of speech or writing by the average individual
is tremendous. A student attending classes and holding frequent conversations
with fellow-students might hear 100,000 words in a day. If he has a modest
reading speed of, say, 300 words per minute and spends five hours a day
reading, he would cover 90,000 words a day. Such a student, then, could
easily be exposed to three-quarters of a billion words a year.
Not all these words would be different, of course; indeed, it is likely that
about one in ten is the word the. Notice the distinction between a word as
a type and a word as a token: The simply as a word is a type, whereas each in-
stance of it in use is a token. In telephone conversation, 50 common word types
make up about 60 per cent of all the word tokens. Nevertheless, the number
of types found can be very large, if large enough samples are taken. Even
though an individual may not use certain words more than once in a very
long time, if he is highly educated he may have a vocabulary of well over
100,000 different words, particularly if one includes all the proper names
of people and places that he knows. Because of certain methodological dif-
ficulties, research has not yet given clear answers on the magnitude of in-
dividual vocabularies, but we can confidently say that word types constitute
the largest single set of different learned responses of human beings.
But speech behavior cannot be measured solely in words. The sounds that
compose the words, the inflections of the voice, and the ways the words are
composed and arranged are all essential elements of speech. In addition, a
talker is most likely to accompany his speech with gestures and facial ex-
pressions that add emphasis or nuance. Of even more importance to a
psychologist, the behavior of the talker represents some kind of message,
and behind this message one is tempted to infer the operation of a host of
psychological processes commonly identified under such names as perceiving,
desiring, willing, thinking, believing, and feeling. Around 1900, a favorite
method of psychology was to ask subjects to “introspect,” that is, to make
careful verbal reports on their own mental processes. Nowadays, less con-
fidence is placed in such subjective reports, but it remains true that what a
person says or writes constitutes overt behavior that is potentially grist for
the psychologist’s mill if he will take the trouble to study it objectively and
with due regard for other kinds of information about behavior.
Language
and
Communication
 It will be well to define our terms. Speech behavior is that overt activity
in which the muscles controlling the diaphragm, the larynx, and the various
parts of the mouth are used to produce utterances exhibiting regularities that
depend on a system of vocal communication we call /anguage. In order for a
language to exist, there must be a speech community, that is, a group of
individuals who are able to communicate with one another because they have
learned to respond to one another’s utterances in consistent ways. In certain
psychological experiments, the experimenter and the subject in effect con-
stitute a miniature speech community; the experimenter arranges matters,
say, so that the subject gets rewarded only if he learns to say “zik” when he
sees “MUQ,” or the experimenter plays a game with the subject in which the
latter must discover what classes of stimuli are to be called “zugs.” Natural
languages like English, Chinese, or Navaho have speech communities composed
of thousands or even millions of speakers who have all learned a large number
of responses in common. Furthermore, the regularities found in the language
systems used by these speech communities are both numerous and complex.
In theory, a language can be a system underlying any set of responses of
which human beings are capable; thus, it is not entirely inappropriate, in some
contexts, to speak of “the language of mathematics” or “the language of
flowers.”’ However, for our present purposes, we shall use it only for the sort
of system that underlies the oral communication of a speech community.
Occasionally it may be useful to extend the concept of language to include the
system of gestures and facial expressions that ordinarily accompany speech
behavior, but this system is largely dependent on speech behavior and does
not exhibit the degree of complexity shown by the spoken language system.
It is beyond the scope of the present treatment to discuss the possible
“linguistic” status of still other systems of behavior that may play a part
in communication, like the “expressive movements” exhibited in various per-
formances such as handwriting that G. W. Allport and P. Vernon have
studied,2 the styles of culturally conditioned behavior called the “silent
language” by the anthropologist Edward T. Hall,? or the language of visual
symbols described by Ruesch and Kees.*
Writing, however,is a system of communication that has a special rela-
tionship to spoken language in that it depends largely on the prior existence
of spoken language. Phylogenetically, man learned to talk before he learned
write.
to write, and ontogenetically, the child learns to talk before he learns to
reason, written language must always be regarded as spoken lan-
For this
writing system and
guage “written down” in a particular conventionalized
the structure of a language
phrased, often, in a special written style. Studying
its limita-
solely in its written form, although useful for some purposes, has
this method totally ignores the sound system of the
tions; for example,
cal research
language and its possible effects on the structure. In psychologi
words as
and experimentation, it can be misleading to use written or printed
expressive movement. New York:
2G. W. Allport and P. E. Vernon. Studies in
Macmillan, 1933.
1959.
3K. T. Hall. The silent language. New York: Doubleday,
University of California
4 J. Ruesch and W. Kees. Nonverbal communication. Berkeley:
Press, 1956.
Language
and
Communication
 stimuli without taking proper consideration of the way in which subjects
may respond to these stimuli in terms of spoken language. To give a simple
illustration, a homograph like LEAD can be highly ambiguous.

There is a cartoon depicting two prehistoric men wondering, now that they
had learned to talk, what they would talk about; the humor of it lies mostly
in the fact that language would probably never have developed unless it had
served some function. We can think of language as serving two major func-
tions: (1) as a system of responses by which individuals communicate with
each other (inter-individual communication); and (2) as a system of re-
sponses that facilitates thinking and action for the individual (intra-individ-
ual communication).
It seems almost too obvious to say that language functions in interpersonal
communication—in conveying information, thoughts, and feelings from one
person to another and in providing a means by which people control each
other’s behavior. Such a statement, however, is couched in everyday parlance
and is actually difficult to translate into the terms of a purely objective,
scientific account of interpersonal behavior. For example, it leaves such terms
as “information,” “thought,” and “feeling” undefined, and fails to explain
how anything as apparently insubstantial as language can control behavior.
It is no wonder that philosophers and psychologists have had difficulty in
clarifying their concepts of the function of language. We shall take the point
of view that an objective account of the role of language in communication
can be achieved only through an analysis of people’s behavior as they use or
learn to use language; such an analysis will be attempted in Chapter 3.
Once an individual has learned even a small portion of the responses in-
volved in language, he can start to use them in intrapersonal communication,
that is, in “thinking” and in the facilitation of his own behavior. For one
thing, the individual can respond to his own speech behavior, either with
more speech behavior, or with action; he can, for example, respond to verbal
representations of previous experience, long after the original experience, and
he can give himself commands to act. Furthermore, many language responses
come to correspond to what we ordinarily call “concepts.” Indeed, they
function as names of concepts and therefore can be used as stimuli for
evoking and manipulating concepts. This function may be illustrated by
the behavior of a person performing an exercise in mental arithmetic. The
words he uses in performing this task (for example, “3 times 9 is 27; 7 times 9
is 63, carry 2, so 73 times 9 is 657”) stand for concepts (numbers and
operations with numbers) that he can manipulate as verbal forms far more
easily than as concrete things he might have to count or put together
manually. The decimal number system provides a well-organized set of verbal
concepts and operations by means of which an individual can communicate
with himself (and others). It is far better designed for Auman use than the
binary system (that uses only the digits 0 and 1), however appropriate the
latter may be for electronic computing machines.
Language Most systems of concepts are nonnumerical, of course, but the same prin-
and
Communication
ciple holds. The evidence suggests that, in general, people can “think” better
when they have a good stock of well-learnéd concepts and their names.
Because of the intimate connection between language and conceptualiza-
tion, we will devote considerable attention, in Chapters 6 and 7, to the nature
of concepts, how concepts are learned, and what role they play in behavior.

In order to study language, it is necessary to gain some perspective on the

relation between language and what is called communication. Communication
is a concept whose scope of application is rather flexible. In the most general
sense, communication occurs when some kind of energy is transferred from
one place to another, for example when a disturbance occurring on one side of
a pond is “communicated” to the other side by a series of waves, or when the
energy applied to a sense organ is “communicated” to the brain over a nerve.
One might also argue that the appearance of smoke somewhere in the dis-
tance “communicates” to an observer that a fire is burning. It could be said
that the smoke is a “sign” of fire. Our experience constantly furnishes ex-
amples of events or stimuli that are in some way “evidences” or “signs” of
various states of affairs, purely through the operation of various physical,
chemical, or biological effects, but we do not ordinarily think of such “signs”
as constituting elements in a “sign system” like language nor even as mes-
sages in a “communication system.” A cough may be the sign of a cold, but
it is not ordinarily perceived as a contrived, artificial symbol analogous to a
word.
We are, after all, interested in a particular kind of communication: com-
munication between persons (or, sometimes, within a person). The commu-
nication system that enables human beings to communicate has two major
aspects: (1) a physical and biological system in which the communication
takes place, and (2) a sign system in which messages are formulated.
If we examine the total system in which communication takes place, we
observe that the number of links over which messages may travel in passing
from one person’s nervous system to another’s may be very large. A separate
branch of engineering has developed around the problems of making sure
that messages flow over the purely physical links (air, telephone wires,
microphones, and so on) with minimum error and expense. Various biological
and psychological specialists are concerned with the characteristics and capa-
bilities of the biophysical links in the system—the nerves, the speech ap-
paratus, and the auditory receptors (if we consider only the transmission
of oral language).
Of much more relevance to psychology, however, is the study of sign
systems. Let us examine the sign system of one of the simplest communication
arrangements we can think of—that whereby a thermostat controls the
operation of a heating plant. Here we observe the three essential properties
of all sign systems, namely:

1. A finite set of discrete signs. In this case there are only two signs: the
flow of an electric current, or the absence of such a current. The thermostat, Language

and
Communication
 which originates messages, is able to produce both these signs in such a way
that they are readily discriminable by the sensing mechanism (probably an
electric relay), which turns the heating plant on or off.
2. Referential function of the signs. Each of the two signs in the system
represents or reflects a different state of objective reality—(a) a temperature
as high as, or higher than, a given setting, or (b) a temperature lower than
the given setting. These signs are generated by the mechanism of the
thermostat.
3. Arbitrariness of the sign system. Deciding which sign is to represent
which state is (at least in principle) arbitrary; it depends only on how one
arranges the mechanism.

Consider now how the sign system of a natural language also has the
above three properties. The finite set of discrete signs of a natural language
consists of the sounds, combinations of sounds, and arrangements of sounds
that recur in samples of messages in that language. These signs constitute what
may be called the expression system of the language. They have a referential
function in the sense that signs show consistent correspondences or relation-
ships to states of affairs other than themselves—states of affairs either in ob-
jective reality, in the psychological state of the speaker, or in the message in
which they appear. The set of correspondences between the signs of a language
and such states of affairs may be called the content system of the language, or
simply, the meaning system of the language.
Signs show various degrees of relationship with objective reality. There is
little trouble defining the class of objects or events to which a word like
telephone refers, but it may be difficult to identify the referents of words like
jealousy, teach, concept. Some signs, like Hi and Thanks bear referential
relationship only to certain kinds of social situations. Other signs refer to rela-
tions between referents; words like im, of, and and have this function. The
particular sequential arrangement of signs may have a referential function;
this is evident in the comparison of the strings man bites dog and dog bites
man, The word fo in an infinitive phrase like to be may be said to refer to
something in the grammatical structure of the sentence in which it appears.
Still other signs, namely, the basic sound elements of a language (pho-
nemes, to be discussed in Chapter 2), have no referential function in them-
selves, but constitute the component parts of other signs that do have
referential meaning; frequently, these basic signs carry critical differences
between referential signs. For example, the words bin and pin in English
might be the same were it not that we discriminate their initial sounds. The
two words permit (noun) and permit (verb) are differentiated by their
patterns of stress or accent.
Finally, the ways in which the expression system of a particular language
corresponds to its content system are essentially arbitrary. There is no reason,
other than a historical one, why pin stands for a small pointed object instead
of a storage receptacle. That two languages might happen to use two highly
similar signs for the same referent could be a pure coincidence; usually the
signs are quite different. There are cases, of course, where language signs
seem not to be wholly arbitrary, as where expressions imitate animal cries
Language
or where they appear to exhibit what is known as phonetic symbolism. In
and
Communication
_ cases of phonetic symbolism, the particular sounds of a word
are presumably
in some way correlated with the meaning of the word. For instance,
it has
been claimed that the meaning “smallness” is associated with
the vowel
sounds in such words as little, teeny, pin, nip, whereas “‘largeness”’ is
associated
with the vowel sounds in such words as large, huge, pool, ocean, and so on.
But even though people show some consistency in their affective response
s to
pure sounds or nonsense syllables, the role of anything like phonetic
sym-
_bolism in word formation or in carrying meaning is difficult to demonstra
te.
_ And even if phonetic symbolism is a significant factor, the essentially arbitrary
character of signs still remains. The variations in words for animal cries in
different languages are startling—compare, for example, gnaf-gnaf in French
with bow-bow.®
The sign systems of natural languages exhibit a high degree of complexity
in their structure, even though complexity is not an essential feature of a sign
system. In the next chapter we shall explore this fact and its implications.

5 For a representative recent study of phonetic symbolism, see M. S. Miron, J. abnorm.

Soc. Psychol., 1961, 62, 623-630.

Language
and
Communication
The Nature of Language A language is a socially

institutionalized sign system. It is the result of centuries

of gradual development and change at the hands of many

generations of speakers, but at any one point of history

it exists as a set of patterns of behavior learned and

exploited in varying degrees by each member of the speech

community in which it is used. In studying the structure

of a language we study the characteristics of these be-

havior patterns.

The behavioral science that concerns itself with the

description and study of languages as sign systems is de-

scriptive linguistics. Because of the rigor of its method-

8
|
|ology and the precision of its results, descriptive linguistics is one of the most
|advanced of the behavioral sciences; even so, the description of a language
_ system still presents many fundamentally unsolved problems.
If a psychologist were asked to describe the characteristics of a sign system
_ as a set of behavior patterns, his first impulse would probably be to attempt
to discover regular correspondences between stimuli and responses. He would
try to find out what words the speakers of a language have learned to speak
in given situations, and what responses these speakers would make to given
_words. That is, he would try to find correspondences between the content
system of a language and its expression system.
The immediate results of such an approach would probably be quite trivial.
_ They might consist of little more than a list of facts such as what one might
predict by studying a dictionary. They would not begin to describe the ulti-
_ mate structure of the total sign system. For one thing, the weight of evidence
suggests that, except perhaps in limited cases, the correspondences between
the content and the expression systems of a language are extremely complex.
_ But more importantly, such an approach would be fraught with the danger
that the description of the expression system would be made to depend on,
or be biased by, the description of the content system. This is precisely the
kind of mistake that has often been made in traditional grammar. For
example, the definition of the grammatical category called noun has some-
times been made to depend on a conceptual classification of the things pre-
sumably symbolized by nouns: “persons, places, things, and abstract ideas.”
You can doubtless find many reasons why such a definition is unsatisfac-
tory.
Descriptive linguistics has found it necessary to adopt a very different
approach, namely, the concentrated study and description of the expression
system prior to any attempt to coordinate it with the content system. A
completely rigorous description of the expression system of a language must
make no appeal to anything we ordinarily call “meaning.” It is not denied
| that this expression system is, or may be, connected with a system of meanings,
_ but the statement of the former must be independent of any statements about
the latter. In most of this chapter we shall be talking about expression sys-
tems of languages: The question of their relationships to content systems will
be postponed to the end of the chapter, or alluded to only briefly by sug-
gesting what kinds of linguistic items are “meaningful.”

The basic problem of the linguist has been to develop a description of the
sign system of a language, particularly its expression aspect (in the sense
defined previously) ,from whatever materials he finds relevant for the purpose.
In doing so he finds it necessary to adopt some theory or model of language
_ structure, if only to have some categories into which to classify the findings.
Historically, these models were at first fairly simple; they became more
| elaborate and refined as more and more languages were studied, and more and
_ more aspects of languages were investigated. It was recognized from the very
| start, however, that models of language developed on the basis of such Western
| The Nature

9
of Language

|
 languages as English and Latin would not necessarily serve for the descrip-
tion of non-Western languages. (We avoid calling any language “primitive” ;
even the languages of “primitive” tribes have often proved to be fully as
complex, at least in structure, as languages familiar to Western civilization.)
It is obvious that the raw material of which language signs are com-
posed is sound. The first thing that the linguistic field worker does is to write
down the speech uttered by his informant, initially in great detail in order
not to miss anything that might be significant. In order to do so, the linguist
has to be an expert in phonetics, the science of speech sounds, because it is
frequently the case that the languages he studies present types of speech
sounds unknown in the more familiar languages. The detailed record of the
speech sounds composing a text is known as a phonetic transcription. Some
linguists use for this purpose the list of symbols provided by the International
Phonetic Alphabet (IPA); others modify this drastically, or make up their
own symbols.
Having a phonetic transcription of his text before him, and possibly a
rough translation, the linguist next tackles the job of teasing out the struc-
ture of the language. He would remember that a fundamental characteristic
of a sign system is that it has a finite set of discrete, distinctive signs. The
three general kinds of signs that the linguist would look for are these (each
kind will be explained in more detail later):

1. Distinctive basic sounds—the types of sounds (vowels, consonants, tones

or pitches, relative intensities, and durations or pauses) that are not mean-
ingful in themselves but go to make up and differentiate the signs of a lan-
guage, which may in turn have some sort of meaning or grammatical function.
They are called phonemes.
2. Forms—sequences (not always continuous, however) of phonemes that
constitute the basic grammatically functioning elements of a language. Some
forms are very short, perhaps composed of a single phoneme, like the form
that converts dog to dogs or cat to cats. Medium-sized forms are those that
we ordinarily think of as words, like dog, exercise, abracadabra. Other forms
can be longer than words, like the form go through with, because this is a set
pattern whose meaning cannot be predicted from the separate words. Many
forms appear in different ‘phonemic shapes” depending on the context in
which they appear; in fact this is true of the plural-forming element just
mentioned, because in dogs the form appears as the phoneme /z/ and in cats
it appears as /s/. (For meaning of slant lines, see footnote 2, page 14.)
3. Constructions—patterns or arrangements of forms, the particular forms
composing them being replaceable with other forms. An example is the noun
phrase construction composed of a modifier and a head noun, illustrations of
which are the noun phrases green cheese, green hat, this hat, this car, this
green car, etc. The grammatical function of a construction will generally
depend on the classes of forms that can compose it. In order to study con-
structions, one must have made an appropriate classification of forms.

This model of language implies that language structure is hierarchical.

Constructions are composed of forms, and forms in turn are composed of dis-
tinctive basic sounds, or phonemes. In fact there can be a multiplicity of
The Nature
of Language
Io
 levels in the hierarchy: Forms can be combined to make other forms, and
constructions can be combined to make still other constructions.
The general procedure of linguistic analysis has been to work “from the
_ bottom up.” In theory, the linguist’s first task is to identify the phonemes;
having done so, he identifies forms, and in turn, constructions. In practice, it
is difficult to identify phonemes until at least some tentative identifications
of forms have been made; each part of the analysis is continually revised in
the light of findings in other parts until a satisfactory total description of the
language is built up.

THE BAW NATERIALS OF LANGUAGE: SQUNDS

The varieties of sounds people can produce are very large in number—
| far larger than occur in any one language. Phonetics (to be carefully dis-
tinguished from phonemics) is the study and description of the varieties of
sound that occur or can occur in different languages and their dialects.
_ Phoneticians study the physical attributes of speech sounds, the manner in
which they are produced by the human speech mechanism, and (sometimes)
_ the manner in which they are perceived and discriminated by hearers.
Speech sounds can be classified by the ways they are produced. Vowels
are produced by shaping the oral and nasal cavities in such a way that the
_ sound issuing from the vocal cords or the windpipe has certain resonance
_ qualities. Vowels can be roughly classified by noting the positions of the jaws,
_ tongue, lips, and pharynx (which controls access of air to the nasal cavity).
_ For example, the vowel [i] 1 in the English word pin is classified as an un-
rounded medial high front oral vowel, because in producing it the lips are
_ unrounded, the highest part of the tongue is toward the front of the mouth
_ and in a medium high position, and it is not nasalized. Consonants are pro-
_ duced when some part of the speech apparatus presents some degree of
interference with: the passage of air from the windpipe; the major variables
that distinguish consonants from one another are (1) the point of articula-
tion (the place in the mouth where the greatest interference occurs, and the
surfaces involved), (2) voicing (that is, whether a voice tone accompanies
the articulation), and (3) the type of articulation (the kind and degree of
| interference that occurs). Take, for instance, the initial [p] in the English
_ word pat. The point of articulation is between the two lips, thus it is bilabial;
the articulation is not simultaneously accompanied by a voice tone, hence
it is voiceless; it is made by suddenly releasing the lips from a completely
closed, or stopped, position to an open position, with a slightly noisy ex-
pulsion of air, hence it is an aspirated stop. In contrast, [v] in vat is a
- labiodental voiced fricative because the point of articulation is between the
lower lip and the upper teeth, it is accompanied by a voice tone, and there
is air friction. Study of articulatory phonetics will enable the student to
produce a variety of sounds not found in English; for example, the sound of
the 6 in Spanish caballo, often symbolized [f], is a bilabial voiced fricative,
because the point of articulation is between the upper and lower lips.
1 Tt is a convention in linguistics that a symbol enclosed in brackets is to be considered
simply as a sound, that is, solely from the standpoint of phonetics.

The Nature
of Language
II
 Modern technology has made great contributions to the study of the
physical attributes of sounds. Sound waves themselves can be portrayed with
an oscilloscope (Figure 1). In addition, the sound spectrograph, which elec-
tronically plots the frequencies of the overtones in a sound wave against
time (Figure 2), has made it possible to establish beyond question that vowel
quality is directly correlated with the relative positions, on the frequency
scale, of two or three formants, or bands of strong resonance overtones, pres-
ent in the frequency spectrum of a sound. For example, regardless of the
fundamental pitch at which it is spoken, the vowel [i] in pin is produced by
a male adult speaker of English when he shapes his oral cavity so that the
principal formants, or resonance bands, are at about 250 and 2100 cycles
per second. A so-called “pattern playback device” has made it possible to
find out what kinds of sounds are perceived when given combinations and
sequences of formants are generated electromechanically. This device has
shown that many of the stop consonants (like [p], [t], and [k]) have quite
different acoustic characteristics depending on what vowel sounds they precede
or follow. Through careful cutting and splicing of magnetic tape, one can
demonstrate that a [k] sound before the diphthong [iy] (as in key) will
sound like [p] when grafted on in front of the diphthong [uw] (as in pool).
These findings raise the interesting psychological question of how two [k]
sounds can be perceived as being the same or similar when their acoustic
characteristics may be quite different. One possible answer is that we learn
to perceive sounds in terms of the’ way we articulate them; it is as if we
automatically refer any heard sound to the nerves and muscles that produce
them.

When we turn to the study of how sounds function in making up the dis-
tinctive signs of a language system, we realize, first, that a given language
system uses only selected parts of the total range of possible speech sounds,

00

ron pwn [VY

a ; : Figure 1.- A representa-

tion of the sound wave of
the word poor, with the
WI JW Yr NI ; wave of a 500-cycle tone
for comparison. (From
500 Cycles H. Fletcher. Speech and

tion. Princeton: Van

Nostrand, 1953, p. 33.)
The Nature
of Language
 a

Rreep Uwrboy
z | 2k +t iy és r.z
Figure 2. A sound spectrograph representation of the sentence The poor
_ boys lack teachers. (Courtesy H. L. Cramer, Laboratory for Research in
Instruction, Harvard University.)

and second, that sounds function as signs only as the users of a language learn
to recognize and produce differences in sounds that will produce differences
in the communicative values of the linguistic forms these sounds go to make
up.
A concrete illustration may help you understand this very important state-
ment more fully. Start with the observation that one can generate a large
variety of hissing sounds—the kind of sound we ordinarily represent by the
letter “‘s.” The tongue-tip can be placed in various positions—just back of
| the teeth, or next to the gums, or even quite far back near the roof of the
mouth—and the rest of the tongue may assume various positions. Speakers
of English would probably perceive most of these sounds as “‘s” sounds;
any one of them could be used in pronouncing a word like sim, though some
i
of them might be perceived as producing the words shin or thin. Nevertheless,
the range of tongue positions that are ordinarily used for the ‘“s”’ sound in
| English is relatively narrow. But the total range of possible English “s”
_ sounds includes two ranges of ‘‘s” sounds for speakers of Arabic—one pro-
_ duced with the tongue just back of the upper teeth, and one produced with
the tongue-tip somewhat farther back and with the back of the tongue raised.
Although these two ranges of ‘“‘s” sound are clearly discriminable by Arabic
speakers, speakers of English have great difficulty in distinguishing them
without training.
In linguistics, phoneme is a technical term for a range of sounds that the
-speakers of a given language perceive as functionally the same and discrimi-
nate from other ranges of sound. Thus, within the range of sounds that are
‘perceived as “‘s”’ by English speakers, there is one phoneme in English, but
there are two phonemes in Arabic.
Phonemes are the building blocks out of which meaningful or grammatically
functional forms are composed; furthermore, they provide the critical basis
The Nature
of Language

13
 TASLE 3

English Phonemes and Some Words

and Sentences Exemplifying Them
(As in the Author’s Dialect)
33 SEGMENTAL PHONEMES
20 Consonants 4 Semivowels

As In
Occurring as initial, medial, or final: Consonant Diphthong

/b/ as in buy /s/ as in so

/d/ as in do /t/ as in toe /h/ as in hoe bah /bah/
/{/ as in foe /v/ as in vow
/g/ as in go /2z/ as in 200 /t/ as in roe err /ar/
/k/ as in key /0/as in thigh
/V/ as in lie /8/ as in thy /w/ as in woe now /naw/
/m/ as in my /8/ as in show
/n/ as in no /é/ as in chow /y/ as in you boy /boy/
/p/ as in pay /j/ as in Joe
Occurring only as medial or final:

/z/ as in pleasure, rouge

/1/ as in singer, thing

9 Vowels,

Simple Vowel Followed by Semivowels

+ Consonant /-y/ /-w/ /-h/ /-r/ /-yr/

/i/ as in pit pea * * * pier
/e/ as in pet pay * * * pare
/x/ as in pat * * /eh/ + * *
/i/ as in roses * * * * *
/a/ asin putt, but * * /ah/ ** purr tt *
/a/ asin pot pie now pa par pyre
/u/ as in put, look buoy coo boo! poor *
/o/ asin * boy low oh! pore *
Joy Bevin 53 * * law war *

* Does not occur as a monosyllable in the author’s Lower Connecticut Valley dialect,
but may occur in other dialects of English.
+ An interjection of frustration or disgust.
** The hesitation form.
77 In the author’s dialect, this is a single r-like vowel, but it fits best into the pattern
if considered a diphthong with /a/.

for differentiating among these forms. One way in which we can identify the
phonemes of a language is to try to find pairs of forms (words, for example)
that speakers will identify as different in meaning or in use, but phonetically
the same except for one sound. Such are called minimal pairs. For example,
in English the words Sam and sham are a minimal pair that establishes the
The Nature
of Language

14
 12 SUPRASEGMENTAL PHONEMES

4 Pitches: /1/ (lowest), /2/, /3/, /4/ (highest).

4 Stresses: /’/ (primary), /*/ (secondary), /*/ (tertiary), /~/ (weak).
4 Junctures: /+/ (internal), /|/ (level), /||/ (rising), /#/ (falling, terminal).

A SAMPLE TRANSCRIPTION

He: /?dija+siyda + %hwaythaws>||/

She: /?n6w3||2kazay + w5hntid + ta + vizitda + smi#s6wniyan + myiw3ziyam2|
26a + laybroriyov + 3kahngris*|2a@nda + jéfarsan + ma*mohriyal? ||
4$hl2|2in + thw + 4awrz2# /
He: /?im4pahsibol! #/
He: Didja see the White House?
She: No, ’cause I wanted to visit the Smithsonian Museum, the
Library of Congress, and the Jefferson Memorial—(ex-
citedly) all in two hours!
He: (incredulously) Impossible!

sounds “‘s”’ and “sh” as different phonemes.” In Syrian Arabic the two words
saam (a proper name) and saam (“fasted’’) can be distinguished by an Arab
speaker both in sound and in meaning (even though they might be indistin-
guishable to a speaker of English); this minimal pair establishes /s/ and /s/
as different phonemes in Arabic, the two ranges of hissing sounds mentioned
on p. 13. It is not always possible to find minimal pairs for all pairs of
phonemes, but other kinds of evidence can be adduced to establish the list of
phonemes for any particular language or dialect.
The student who plans to do practical work in any branch of psychology in
which precise specification of speech stimuli and responses is required would
do well to become aware of the phonemes of his own language, or better still,
his particular dialect. He should also learn to make a phonemic transcription
of his own speech. Table 1 gives a list of the phonemes found in the author’s
dialect of American English, as symbolized in one widely used transcription
system; it also gives a sample phonemic transcription of a hypothetical con-
versation. A phonemic transcription represents only the sound units that are
distinctive in the language being transcribed, and requires only the number of
symbols that corresponds to the number of phonemes: It does not necessarily
“have to use the symbols of the International Phonetic Alphabet.
2 Caution: the letters “sh” here stand for one sound, that is, one phoneme. It would
perhaps be wiser to introduce phonemic symbols for the sounds of “s” and “sh”: /s/
and /%/, respectively. The slant lines are used to indicate that what is enclosed in them
is to be regarded as phonemic, as opposed to square brackets [ ] which enclose material
regarded as merely phonetic.
The Nature
of Language

T5
 Segmental phonemes include consonants, vowels, and semivowels (phonemes
that function either as consonants or as parts of diphthongs, like /w/ and
/y/); they constitute the sequential segments (though with some overlapping
and gliding) of syllables, words, and sentences. For example, the word strokes
consists of the seven segmental phonemes /strowks/.
Suprasegmental phonemes include phonemes of stress, pitch, and juncture
which occur simultaneously: with the segmental phonemes or which separate
them. Here, “stress” refers to the relative degree of intensity with which a syl-
lable is pronounced; in English, four such degrees can be distinguished. Com-
pare:

I am contént.
It is devoid of contént.
The table of conténts. (In some dialects.)
Satisfaction and conténtmént.?

Pitch refers to the relative height of the tone with which a syllable is spoken,
or to some contour of such tones. Again, four pitch phonemes can be distin-
guished in English; see the sample transcription in Table 1. In Mandarin
Chinese, the four tone contours with which syllables can be spoken generally
indicate different meanings: Four possible meanings of the syllable ch’, de-
pending on tone, are “seven,” “period of time,” “rise,” and “breath.” Juncture
refers to different types of transition between syllables or between clauses.
The difference between the words nitrate and night-rate is a matter of whether
a slight break, called an open transition and symbolized /+/, occurs be-
tween the phonemes /t/ and /r/. The differences between the following three
pronunciations of “yes” are carried by intonation and juncture phonemes (it
is hoped that the context and punctuation will suggest the pronunciation
adequately) :

“Definitely; yes!” /yésl

#/
aoe te. ESE) PY OSES /?yés||/ (said on the telephone by a
speaker at intervals while he listens to the person at the other end of the line)
“Well, what do you say? Yes?” /®yés||/

Even a simple utterance like the “‘m-hm” which is sometimes used in psycho-
logical experiments to indicate approval (constituting “reinforcement,” al-
legedly) can be said in a rather large variety of ways, but it is possible to
standardize this utterance if the speaker will stick to a given contour of stresses,
intonations, and junctures. A good contemporary textbook of linguistics, or
better still, an expert on the linguistics of English, should be consulted for
details.*
Phonemes are essentially classes or categories of sound; variation and inter-
gradation within a phoneme class can occur. Some of this variation is free,
3 An interesting function of stress that the author has observed is its “honorific” use in
indicating that something is “well-known,” “famous.” An ordinary person on the street
might be named Clark Gable, just like John Doe. But make him a movie star and you
have Clark Gable. Similarly, compare Abraham Lincoln, Abraham Lincoln; white house,
White House; Third Avenue, Fifth Avenue.
4 See Selected Readings, p. 113.
The Nature
of Language
16
|
some of it conditioned. For example, a person could freely use any of several
|varieties, or allophones, of /s/ in the word span, but the difference between the
_/p/ of span and that of pan is conditioned by whether a pause or a sibilant
| ee it. Phonetically, these two allophones of the /p/ phoneme are dis-
_ tinguished by the fact that in span, /p/ is unaspirated, whereas in pan it is
aspirated. You can feel the difference if you hold your palm about two
|inches in front of the mouth as you say these words. This conditioned varia-
tion is by no means inevitable: It is possible to pronounce pan with the un-
aspirated allophone (it will sound almost like ban, as if spoken with a French
accent) or to pronounce span with the aspirated allophone (it will sound
|“funny,” perhaps as if spoken with a German accent). The way English speak-
ers pronounce pan and span is just a matter of learning; the complexity of
this learning process should not, however, be underestimated. That this learn-
ing is automatic and unconscious is suggested by the occurrence of the ex-
pected variation in allophones even when English speakers are asked to pro-
nounce two completely new words like pown and spown There are many other
_ cases of conditioned variation in phonemes in English; indeed, such variation
_ occurs in every language.
Usually, the phonemes of a language can be classified phonetically so as to
exhibit a neat structure (see Table 2). For example, all voiced consonant pho-

aoa &

Chart of Some English Consonant Phonemes*

Type of Articulation +

Place Stop Nasal Fricative A ffricate

of Articulation + V— V+ (V+) V— V+ V— V+

Bilabial p b m
Labiodental f Vv
Dental 0 6
Alveolar t d n S Zz
Alveopalatal S Z ¢ j
Velar k g 1

V—: Voiceless
V+: Voiced
* To avoid complications, the semivowels r, y, w, and h and the lateral 1 have been
omitted.
+ The meanings of terms such as alveolar, fricative, etc., can be found in most standard
dictionaries, or you can attempt to infer these meanings from the characteristics of the
sounds themselves. For the meaning of the phonemic symbols, see Table 1.

nemes in English except the nasals have voiceless counterparts. This and other
facts about phonemes may be explained by the theory that each phoneme
represents a group of distinctive features present in a single sound. Voicing,
for example, is such a feature, which in English is present in all vowels and a
certain group of consonants.
The Nature
of Language

17
 FORMS

After a linguist has identified the phonemes of the language he is studying,

he goes on to establish the forms that constitute grammatically functional
signs in a language. Form is a general term for any linguistic unit with definite
(though possibly varying) phonemic content. Here are some forms in English,
presented in their conventional spellings:

good get un-

-ness forget re-
goodness forget-me-not construct
for goodness’ sake -ed
hot unreconstructed
berry house
mulberry hothouse go through with

Some of these are smaller than what we call “words”; © they are called bound
forms because they never occur alone. Others are words, or combinations of
words; these are termed free forms because they can occur alone. Every form
given, however, has a definite and distinct grammatical function. For ex-
ample, -vess has an essentially grammatical function; when added as a suffix,
it changes the grammatical use of words like good, thorough, and connected
in a consistent way. Forms like good, forget, construct, and hothouse, however,
have not only grammatical functions, but also correspondences to certain
events and attributes of reality as perceived by human beings.
The identification of these forms by techniques of linguistic analysis is not
easy. The mere fact that a certain string of phonemes occurs with high regu-
larity is not enough to establish it as a linguistic form, and we have pointed
out that appeal to “meaning” is not reliable. Some of the questions that might
arise, with the answers that would probably be rendered by linguistic analysis,
are the following: Is mul in mulberry a form? (Probably not.) Is the form
for in for goodness’ sake the same form as the first syllable of forget? (No.)
Can the first syllables of forget and forsake be regarded as a distinct form?
(Probably not in present-day English.) Is the syllable /li8/ occurring in Eng-
lish, establish, girlish, and delicious the same form? (No, except that /-i8/ in
English is the same form as that in girlish.) Is un- in unreconstructed the
same form as un in uncovered? (Yes, but only if uncovered is the form that
means “not covered” rather than “revealed.”) Ascertaining these answers
without appeal to the specific meanings of the forms is no mean trick; lin-
guists believe they can do it, however, solely by examining and comparing the
sequences in which the phonemes appear, or the ways in which the forms are
built.
It is difficult, also, to know whether to regard any given sequence of forms
as being itself a form, that is, a unit. We could argue that forget-me-not is a
unit because it can replace flower in most constructions, and because it is not
5Tt is very difficult to define word as a linguistic concept; the linguist cannot accept
the conventions of spelling, whereby “words” are separated by spaces, as evidence for
what is to be regarded as a word.

The Nature
of Language
18
Hlikely that any comparable form (for example, forget-thee-not) would be
freely constructed. For goodness’ sake is a more debatable case; it has some
‘of the earmarks of a construction (see below) since various other forms can
ibe freely substituted for goodness. On the other hand, the form is always an
‘exclamation or a parenthetical expression; it always has theishaperjor mes:
\sake; and its meaning has little to do with its specific content. I am inclined
to call it a form. You may want to consider whether go through with ought
'to be regarded as a form.
_ Any form that cannot be divided into two or more other forms is a mor-
_pheme. Good, -ness, get, forget, un-, re-, construct, -ed, hot, house, berry,
mulberry in the above list are all morphemes. (Forget and construct had two
‘morphemes in the languages from which they came, Anglo-Saxon and Latin,
respectively, but they do not have two morphemes in present-day English.)
Like phonemes, morphemes can exhibit both free and conditioned variation.
The morpheme -ing in working, being, and so forth shows more or less free
variation with -in’. (At least, a speaker could say working in some situations
,and workin’ in others; any variables that control this would be extralin-
guistic.) But the pluralizing suffixes spelled -s, -es, -en, -i, in rocks, dogs, roses,
oxen, and alumni and pronounced /-s, -z, -iz, -in, -ay or -iy/ can be thought
of as conditioned variants of a single Poronente which may be symbolized
_{Z}. The study of forms constitutes the branch of grammatical analysis called
morphology.

CONSTRUCTIONS

Morphemes do not just get strung out one after the other in any order
-whatsoever.® Sake for ness good or Is well John very might cause some puzzle-
|ment, even though they might serve as items on intelligence tests. The branch
of grammar that deals with the arrangement of forms into acceptable se-
quences is syntax.
In the most general terms, acceptable sequences occur in what may be called
constructions. Comme are patterns in which forms can be placed; they
_themselves are empty of specific phonemic or morphemic content. If we start
|with the words He went we can replace He with such words and phrases as J,
Al, or Alice’s sister, and went with such words and phrases as coughed and
said that it was raining, still preserving the basic construction. In fact, a con-
struction is a series of “slots” into which particular kinds of material can be
fitted. But each slot must contain a particular kind of material; if it does not,
the result either belongs to another construction (J Caesar is not the same
construction as J coughed) or is not accepted as meaningful by the speakers
of the language (J very is not an acceptable construction in English). The

6 For that matter, neither do phonemes. The patterns in which phonemes occur are
peculiar to each language, and quite rigid. The phoneme /y/ never occurs at the beginning
of a word or syllable in English, although a comparable phoneme does occur as a
syllable-initial in many other languages. Nor does the sequence /vtsr/ appear in English,
although there is no phonetic reason why it couldn’t. Naive subjects asked to make up
new words or spoken nonsense syllables almost always unwittingly conform to the
phonemic patterns of their language.
The Nature
of Language

19
 slots may contain single forms, or they may contain still other constructions.
Sometimes a slot can be left unfilled, in which case its content is called ‘“‘zero.”
A construction is like a recipe, or a computer program, as if it said, for ex-
ample, “Take anything in Class M, follow it with anything in Class 47, then
finish off with something from Class N.”’ Thus, as was said previously, in order
to describe constructions we must be able to identify the classes of things
that can go into their slots. Some of these classes are called form-classes,
corresponding roughly to what have been traditionally been called parts of
speech. Constructions, too, fall into classes.
Here, for example, are six groups of phrases or sentences, each group illus-
trating a particular kind of construction in English. In each group, there are
four examples, and in every case, the first two have single forms as their com-
ponents, while the last two may introduce further constructions substituted for
one or more of these forms.

(1) (2) (3)

He sings pay attention women and children
Demolition occurred have fun either you or I
The five boys chose Jim have had measles both the Republicans and
Women and children
, the D t
mention the fact Merce:
should be allowed to that he came big ones and little ones
leave first

(4) (5) (6)

was sick sick men year after year

seemed green consistent evidence beer after beer

appeared to be neither the red, white, and layer on layer
consistent nor plausibl blue fl ,
Ace OR aes ues sleepless night after
proved to be innocent of old newspapers and books sleepless night
this crime

A good deal of the grammar of a language can be formulated in terms of

phrase structure grammar. The principal tool of this kind of grammar is the
rewrite rule. This has the general form X —> Y, which is read, ‘““X is to be re-
written as Y,” where X is any construction or a component of one, and Y is
an expansion of X or its replacement by a particular form. For example, we
can start with the prescription for one of the primary types of sentence:

Sentence — Subject + Predicate

and successively derive particular sentences by using such rewrite rules as

Subject — Noun, Pronoun, or Noun Phrase

Predicate — Verb, or Verb Phrase
Noun Phrase > Determiner + Noun
Verb Phrase > Verb + Noun Phrase
Determiner — a, the, this, that, some

The Nature
of Language
20
 Noun — man, boy, book, train
Verb — read, stopped, drank

‘to yield such sentences as

1. A man read the book.

2. The boy stopped the train.
3. Some train read a man.
4. A train drank a book.

\(The rewrite rules given above are not complete; they are only illustrative.
The actual rewrite rules for the phrase structure grammar of English would
‘be much more complicated.) Not all these sentences are meaningful, but they
are all “grammatical” in the sense that they conform to the rules. It is possible
‘to formulate rules that will exclude “nonsense” sentences like (3) and (4)
above; the only problem is to know how far it is wise to carry such a proce-
(dure.
Consider, now, the following sentences:

. You solved those problems.

. You didn’t solve those problems.
Those problems you solved, these you didn’t.
Didn’t you solve those problems?
. Those problems weren’t solved by you.
. Solve those problems!
How you solved those problems!

Sentence 1 is clearly a “‘rewrite” from the basic construction Sentence >

“Subject + Predicate. You is a subject, solved those problems is a predicate
| derived from Predicate > Verb + Noun Phrase; Noun Phrase > Determiner
+ Noun.
| We may now ask: Is sentence 2 a “rewrite” from 1 by the expansion of
| solved to did solve to didn’t solve? Conceivably, yes, but there are several
'reasons why the intermediate step did solve is gratuitous and incorrect. Rather,
| the derivation of sentence 2 can best be regarded as a special kind of process
in grammar called a transformation. Specifically, sentence 2 is a negative
_ transformation of sentence 1. The sentence You did solve those problems
would be regarded as a quite different transformation of 1, an emphatic
_transformation.
_ Sentence 3 would be even harder to derive from 1 by the usual rewrite rule,
_for elements are transposed. This would be called, in fact, a transposed object
‘transformation. Sentence 4 introduces the imterrogative transformation; in
deriving this sentence, we must apply the negative transformation before the
_interrogative. The passive transformation is introduced in sentence 5, with a
negative transformation applied; note that it would be awkward to derive
“sentence 5 directly from 2, since the did element would have to be deleted;
it is neater to start with sentence 1, apply the passive transformation, and then
| the negative. Nevertheless, deletion of You to produce the imperative trans-
formation in sentence 6 seems reasonable enough. Sentence 7 illustrates one
variety of exclamatory transformation.

The Nature
of Language

2u
 Thus, some constructions are transformations of other constructions; it is
parsimonious to make a systematic study of these transformations in devel-
oping the grammar of a language.
Many morphological phenomena can be interpreted as transformations,
also. Here are some examples:

Transformation Of Result
Put in past tense : take took
Nominalize good goodness
Adjectivalize wood wooden
Verbalize intense intensify
Make past participle cover covered
Reverse action cover uncover
Make past participle uncover uncovered
Make negative covered uncovered

That uncovered can be produced by two different transformations, with two

different meanings, illustrates a most interesting application of this approach
to grammar. It can be applied to the interpretation of ambiguous sentences
like the following:

They are frying chickens.

This sentence is a transformation either of They fry chickens or of They are

chickens for frying. In actual use, the context would usually furnish the key
to which of these transformations was intended by the author of the sentence.
The point is that many “constructionally ambiguous” expressions cannot be
analyzed satisfactorily without some reference to their “transformational
history.”

METHODOLOGICAL PROBLEMS IN LINGUISTIC ANALYSIS

The results of linguistic analysis, so briefly sketched here, have not been
easily achieved, and there is still room for argument about the validity of these
results. Linguists do not all work within the same theoretical and methodo-
logical frameworks. Some theories postulate that an exhaustive description of
a language can be made solely by stating the items (phonemes, forms) of a
language and their arrangements; other theories view language as a set of
items to which various “processes” (such as transformations) can be applied
to produce meaningful utterances. In either view, language is a formal set of
facts that exist independently of people who might potentially use the lan-
guage.
Many linguists have insisted, with some justice, on studying only language
behavior that has actually been observed, instead of experimenting with pat-
terns of behavior that might occur. They would not dream of asking an in-
formant, “Would you say it this way?” (offering a postulated novel utter-
ance), because this might bias the informant’s judgment of the acceptability
of the utterance.

The Nature
of Language

22
 Until recently, linguists have generally restricted themselves to studying
the relatively long utterances of single speakers telling stories or giving remi-
niscences, and have paid insufficient attention to the speech of normal social
interaction, with its fragmentary sentences, pauses, and rephrasings.
It is often said that the complete statement of the grammar of a language
(that is, a description of all the form-classes, constructions, and grammatical
processes found in the language, with all their variants) would fill a very large
volume. This seems paradoxical when the claim has been made, as it has
indeed, that a child of normal intelligence “knows” all the essential gram-
matical structure of his language by the time he is six years old, if not before.
Either the feat of the child is actually greater than we think it is, or there is
something wrong with the assertion that a grammar of a language needs to be
voluminous.
Of course, the feat of a child in learning his native language is impressive,
but it would seem that these learnings could be listed and described in some-
thing less bulky than a Bible-sized grammatical treatise! How can we state
the child’s “knowing” of grammatical structure? Knowledge of grammatical
structure has two aspects: First, aside from certain problems of vocabulary
and sheer grammatical complexity, a six-year-old is capable, we are told, of
understanding any utterance by a speaker of the language, even though he
has never heard it before; and second, he is capable of uttering sentences that
will be accepted as meaningful and “grammatical” by other speakers of the
language, even though those sentences may never have been said by anybody
before. This capability of generating and understanding novel utterances is
the essence of language; actually, the most valid objective of linguistic anal-
ysis is to describe this capability as a set of learned responses common to the
users of a language. The attempt to describe linguistic structures solely on
the basis of samples of text, divorced from the situations in which the texts
were uttered or created, can never have more than a partial success.
In recent years, linguists have in fact been directing their attention more
closely to the many varieties of normal speech pattern. C. C. Fries,’ for ex-
ample, was one of the first to point out the structural differences in what he
called ‘“‘situation sentences” (those that could begin a conversation), “sequence
sentences” (those that continue a conversation without change of speaker),
and “response sentences” (utterances that continue a conversation, but by a
new speaker). These are illustrated, respectively, as follows:

1. Speaker A: I’m going to take the car for a grease-job.

2. Speaker A: Needs it pretty badly.
3. Speaker B: Yeah, prob’ly.

Noam Chomsky, a linguist, is chiefly responsible for the development of a

theory of grammar that explicitly tries to formulate the phrase-structures,
rewrite rules, and transformations that apply when a speaker of a language
generates a novel utterance or accepts a presented novel utterance as “gram-
matical.” He has emphasized the notion that linguistic analysis must provide

7C. C. Fries. The structure of English. New York: Harcourt, Brace & World, 1952.

The Nature
of Language

23
 a model for describing how speakers construct utterances, and how hearers
can understand or “construe” them when they hear them.®
The linguistic analysis of English (or of any other language) is still far
from complete. The objective is to be able to describe exhaustively all the con-
structions and transformations that function at the various levels of analysis—
from the complete utterance, through the sentence, clause, and phrase, down to
the smallest and most elementary linguistic form. If we are going to be able
to account for the linguistic behavior of the user of a language we need a com-
plete description of the linguistic patterns that are available to him as a
speaker and that must be interpreted by him as a hearer. As psychologists
we will be in a better position to utilize this information if it is stated in
purely formal terms, that is, without appeal to the “meanings” of form-
classes, constructions, and transformations, because the problem of meaning
is better left for the psychologist to solve. The labels we put on form-classes
and other categories may indeed suggest “meanings,” but these labels could
just as well be arbitrary. In fact, Fries has established four principal form-
classes that he refuses to label as anything else than classes I, II, III, and IV,
even though it is evident that they correspond fairly well with the classes we
ordinarily call nominals, verbals, adjectivals, and adverbials.
If we analyze complete utterances, we will find the following types of ex-
pression units or expression-types:

I. Nonsentential expressions.
A. Greetings, etc. Hi, How-do-you-do, Goodbye, So long, “Over.”
B. Calls and other attention getters (some of which can be inserted in
sentences): Hey! John! Well... /ah/.
C. Nonsentential exclamations: Oh! Ouch! Golly! Damnation!
D. Nonsentential responses to another speaker: Yes, no, O.K., m-hm,
Thanks, “Roger.”
II. Sentence-types.
A. Existence-assertions: The basic pattern is [Tere + a verb phrase],
the verb phrase including some form of the verb to be or occasion-
ally one of a small number of intransitive verbs (come, occur, live),
plus a nominal. Example: There is a problem here.
B. Predications: The basic pattern is: [Subject + Predicate], where
Subject — Nominal; Predicate — one of the following constructions
(all verb phrases) :
1. Linking verb + Nominal: “‘is his sister,” “was Tuesday.”
2. Linking verb + Adjectival: “is sick,” “was dedicated to truth.”
3. Linking verb + Adverbial: “is home,” “‘is in Paris.”
4. Intransitive verb: “rains,” “is swimming,’ “occurred,” “ex-
istseg
5. Transitive Verb + Object(s): “killed a rabbit,” “received a
letter,” “gave him money,” “elected him president.”

The major transformations that apply to sentence-types and leave them still
in the form of sentences or minimal utterances are the following:
8 See the excerpts from Chomsky that appear in Psycholinguistics, edited by S. Saporta.
New York: Holt, Rinehart and Winston, 1961.

The Nature
of Language

24
 a. Null—the declarative sentence. (According to Chomsky, any
sentence
that has not been subjected to any transformation is a “kernel
sen-
tence.””)
b. Negative.
c. Passive (applicable only to certain sentence-types containing a transitive
verb and at least one object).
d. Formal interrogative: Interrogative sentences that may be “pivoted”
on different components of a construction, sometimes on two or more, as
in “Who is going when in what vehicle?”
i. Yes-no questions: Pivoted on verb phrase. “Are you going?”
ii. Wh-question: Pivoted on subject, object, indirect object, adverbial,
or adjectival. “When are you going?”
e. Intonation question: Intonation contour applied to declarative sentence.
“You went?”
f. Echo-question: Formal interrogative question of another speaker re-
stated or rephrased with special intonation contour. Speaker B: “What
are you doing?” Speaker A: “What am I doing?”
g. Transposition: Part of predicate transposed. “Those problems I solved.”
h. Subject postponement (applies only to predications). “It is true that he
is here” < “That he is here is true.”
i. Imperative (applies only to predications). “Solve those problems.”
j. Exclamation: Pivoted on different components of constructions, parallel
to interrogative transformation.
k. Deletion: One or more components of a sentence-type may be deleted
in “sequence” or “response” utterances when the content of the deletion
is clear from preceding context.

By means of these lists, we can assign expression-types to the components

of the following possible utterance:

Hi there, Bud. Listen! What I’d like to know is, what the dickens is your name?

Or we should be able to take a complex sentence like the following and

analyze it as a complex series of constructions and expressions from basic
sentence-types:

The mathematical concept that has attracted the most widespread attention
from psychologists and linguists interested in communication theory is that of
entropy.

Here, the basic expression-type to which this sentence reduces is a predication

with a subject (“The mathematical concept . . . theory”), a linking verb
(“is”), and a nominal (“that of entropy”). The subject is in turn a nominal
phrase with a clause transformed from a predication with a transitive verb,
of the form “Concept. . . attracted. . . attention,” and this in turn contains
a construction transformed from a predication with a linking verb and adjec-
tival, of the form “Psychologists and linguists are interested in... .” We
haven’t space, of course, to give all the details of these processes.
The deletion transformation is particularly useful as a possible analysis
of the many “fragmentary” utterances heard in normal conversation: ‘The
The Nature
of Language

25
 bread, dear.” “Needs more power.” “By the door.”’ In many cases, of course,
it would be difficult to decide on the exact “transformational history” of such
an utterance; for example, “By the door” could be derived fom “It’s by the
door,” or “You'll find it by the door,” and so on. The point is that such ab-
breviated utterances can be referred back to basic types of utterances whose
use and function can in theory be clearly described; it is not necessary to
create a special grammar to take care of abbreviated or fragmentary utter-
ances.

ANALYSIS OF THE CONTENT SYSTEM OF THE LANGUAGE

Throughout the preceding discussion of the expression system of language,

we have tried to treat the problem of meaning as cautiously and as gingerly as
the linguist does. We have emphasized the reasons why linguists feel it is
necessary to study the expression system of a language independently of the
content system. At the same time, we have not hesitated to give labels to vari-
ous linguistic phenomena which will give some indication of their possible
“meanings.”
Even though linguists have often written about the desirability of making
an analysis of the content system of a language, their scant progress in doing
so is probably an indication that such an analysis would be impossible without
drawing on information about the use of a linguistic system by the speakers
of a language.
We can be sure that the meaning of a particular linguistic form or con-
struction cannot be studied in isolation—that is, independently of a particular
instance of its use—for it can have different meanings depending on the con-
text. This idea from the context theory of meaning can be accepted, but we
should recognize that the context does not necessarily yield the meaning of an
item: It only provides a basis—sometimes shaky at that—for deciding which
of a number of possible meanings it may have. A statement of this meaning is
necessary, regardless of what information may be supplied by context.
Semantic analysis of a language would involve the listing of all its forms,
constructions, and transformations and the giving for each one of these a state-
ment of all the possible relationships it may have to the content system. In
this way, the denotations (see p. 40) of these linguistic elements would be
stated.
For example, suppose we are interested in stating the denotations of the
morpheme /mezn/ and its variant /men/. To do this, we would need to collect
a large number of instances in which this morpheme is used: We might find
such instances as: Man is mortal, Man the oars! Manning table. The child is
father to the man. Five men came. Man-crazy. Through a long process of
asking informants to interpret these sentences, we might eventually arrive at
an analysis something like this for man in Man is mortal: Grammatically, it
is subject; the denotation of the construction is that something is “predicated”
about this subject, or in view of the use of a linking verb and an adjectival,
some “attribute” of “man” is offered. Also, grammatically it belongs to the
form-class we call nouns: it is a “count-noun”’ because it may be preceded by
the article a. The fact that no article precedes it at all, however, signals or
The Nature
of Language

26
 denotes that it is to be taken “generically,” that is, man in general rather
than any particular man; in fact, it includes “woman.” (Man is one of a very
few nouns for which omission of the article signifies generic meaning. We can-
not form the analogous expression “Automobile is expensive.”) We finally
come to the pure lexical denotation of /mzn/, namely, member of the human
species. Similar analyses would have to be made for the other instances of
/mzn/ in our sample until we could be sure that no new denotations could be
found.
All forms, constructions, and processes have grammatical meaning, that is,
meaning concerned with the constructions in which they are found or to which
they may be applied. Most forms also have lexical content, that is, some
kind of reference to states of affairs outside the utterance or text in which
they are found. Some grammarians have claimed that lexical meaning is a
property of only certain major parts of speech (nouns, verbs, adjectives, and
adverbs), implying that the remaining parts of speech are “function words”
with only grammatical meaning. But it is clear that many “function words”
(for instance, prepositions like in, on, with, despite) have lexical components
that are relatively easy to define.
Lexicography, which has been practiced for centuries, is the art of listing,
in some rational order, all the items of a language with their meanings, or
“definitions.” Exactly what units are to be listed and how their meanings are
to be defined have always been problems. Ideally, a dictionary should list
all the morphemes, forms, and constructions that occur in a language. In
practice, the lexicographer has usually taken the word as his unit. He will
sometimes list compound words and forms like forget-me-not when they occur
with high frequency in samples of the language or when their meanings cannot
be predicted from their constituent forms. Ideally, also, each possible gram-
matical and lexical meaning for each item should be given, but in practice,
the lexicographer tries to specify the various parts of speech (form-classes)
in which a form may be found, and to indicate the “meanings” that are
possible in each case. There are no absolute rules to follow in deciding how
many entries should be made for a form, how many meanings a form has, or
how those “meanings” should be stated. It is interesting, however, to study
the practice of different dictionaries in these respects; they vary widely. It is
for this reason that it is ill-advised to use any ordinary dictionary as a basis
for developing methods for measuring vocabulary size, as some psychologists
have tried to do.
The principal function of a dictionary “definition” is to provide information
that will enable the user to fit the meaning and use of an unfamiliar word into
the system of meanings he has already acquired. ‘‘Definitions’” may contain
synonyms, synonymous expressions, or reference to particular attributes of
the item being defined. Typical contexts in which an item may be found are
given (for example, “strike: in baseball, . . .”). That these procedures gen-
erally work reasonably well does as much credit to the user of a dictionary as
to its maker.
Linguists do not attempt to legislate the grammar and structure of a lan-
guage; they merely describe what they observe. Likewise, a dictionary does
not attempt to set standards of usage or fix the meanings of words; it is only
a repository of information concerning word usages of wide enough currency
The Nature
of Language

27
 time of publica-
to be regarded as significant in the speech community at the
tion.
of a lan-
We can go beyond the dictionary in studying the content system
examine the total stock of lexical items for the number of
guage. We can
nate and superor dinate
distinctions that are made, and the number of subordi
the items we
terms they contain in a given domain. For example, what are
different ages and sexes? They may be repre-
have in English for persons of
sented as follows:

(Male) (generalized) (Female)

ie — (oldster?) — ?
man —_- adult — woman
(boy) — adolescent, — (girl)
teenager
(boy) — _ child, “kid” — (girl)
youngster
? — infant, baby — ?

e
There are quite a few gaps in this grid. We have no separate and distinctiv
terms for babies, children, and adolescents of the two sexes, and no special
terms at all, really, for old people. Compare Italian bimbo (boy-baby), bimba
girl-baby), Latin semis, old man.
On the other hand, English has a relatively rich vocabulary for colors and
shades (red, pink, orange, magenta, and so on) in comparison to many other
languages. It is of interest to study some of the distinctions offered by the
lexicon of English: For example, what is denoted by the following words with
the general meaning “give’—grant, donate, contribute, bestow, present, be-
queath, dispense, award? The difference between bestow and present has
partly to do with the relative social status of giver and receiver; such a
difference is not mentioned in most dictionaries.
Making statements concerning the “denotation” of certain form-classes
and constructions is difficult. For example, it is sometimes claimed that the
meaning of the “subject” of a sentence like the man swam is “actor” and that
of the verb is “action.” This seems to be contradicted, however, in Bob received
a letter, since Bob didn’t have to “do” anything to receive a letter, nor is
“receiving” an action. On the other hand, it may be noted that as the child
learns language, nearly all the verbs he learns earliest (like eat, bite, drop,
pull, hit) refer to definite actions. The “action” component of the verb form-
class is so dominant that it is possible that it transfers to words like have,
receive, and owe, which also belong to this form-class, in the sense that native
speakers “feel” that having, receiving, and owing are in some way “actions.”

There are at least three points at which linguistics and psychology have
clearly common interests. (We must remind ourselves that linguistics is essen-
tially a behavioral science.)
One is the possibility of ‘“universals” in grammar and in language struc-
The Nature
of Language
28
ture. Anything that is universal in natural languages is likely to have psy-
chological significance as a basic property of human communication, For
example, it is probable that all languages have transformations in their
grammar and that all languages have devices for asking questions. Thus far,
unfortunately, linguists have made very little advance towards cataloging
universal properties of languages.®
A second is the possibility of significant differences between languages in
the kinds of relationship they exhibit between their expression and content
systems, and the possible implications such differences may have for the
cognitive behavior of the speakers of those languages. A small start has been
made in investigating such differences; the results will be discussed in Chap-
Bole
A third is the possibility of making a psychological interpretation of gram-
matical structure. Although linguists have justifiably avoided any appeal to
psychological considerations, it is possible that the grammatical phenomena for-
mally described by the linguist can be even more parsimoniously described in
terms of what may be called their “psychological motivation” or “dynamic
logic.” That is to say, given a certain linguistic construction, we may be able
to find a psychological motivation for its existence and to show its relation-
ship to other constructions in a way that cannot be done by formal analysis.
One justification for this assertion is that linguists have frequently found it
necessary to re-do their analyses in the light of new discoveries. For example,
the discovery of phonemes of stress, intonation, and juncture in the 1940’s
made it possible to make a thorough revision of treatments of morphology
and syntax. Comparable discoveries in the psychological sphere could lead
the way to still further revisions of statements of linguistic structure.
It seems clear that psychological analysis of linguistic behavior needs to
proceed “from the top down” rather than “from the bottom up” as in usual
linguistic procedures. In formulating an utterance, it is probable that a speaker
selects its over-all construction (that is, its basic expression-types) rather than
the particular words that will compose it, much less the phonemes of those
words. The future task of psychological linguistic analysis, then, is to de-
scribe the situational and psychological concomitants and antecedents of
the major expression-types listed on page 24. The processes by which these
basic patterns are modified by “rewrite rules” and by transformations should
then be studied from the point of view of psychological motivation. There is a
real challenge in the task of determining whether these processes indeed have
any psychological reality beyond the formal analysis produced so elegantly by
linguists, and whether, therefore, these processes play a fundamental role
in thought and action.

9 The volume edited by Greenberg, listed in the Readings, may stimulate the interested
student.

The Nature

of Language

29
The Learning of Language Many species of animals

—ants, bees, birds, and wolves, among others—have

systems of communication that have in rudimentary form

some of the characteristics of human language. In one or

more of these systems one can find such features as: use

of the vocal apparatus, seemingly arbitrary signs, and

means for communicating information, feelings, and emo-

tions. However, these systems are for the most part

passed on by heredity, unchanged from generation to

generation, and they are all extremely simple.

Human language, in contrast, is always learned. Each

child must learn his language from scratch, and the sur-

30
prising thing is how rapidly, relatively speaking, he picks up whatever lan-
guage he is reared in (and how fast he can learn another language during
his very early years). The only hereditary element in human language is that a
normal child is born with the complex neural equipment he needs to learn
and produce it. Despite patient efforts, it has thus far been impossible to
teach primates (like chimpanzees) to speak more than a very few words,
and then only with difficulty. This poor showing seems to be due partly to
the primates’ poorly developed neural equipment for controlling the speech
musculature, and partly to the limitation of their capacity for handling a
complex symbolic system. It may be noted, incidentally, that some animals
(dogs, for example) have shown considerable capability in learning to re-
spond differentially to a small vocabulary of words and phrases, but this is
only what one might expect in view of the capacity of animals for discrimina-
tion learning.

THE COURSE OF LA JAGE DEVELOPMENT IN THE CHILD

Relevant to the learning of language are at least three interrelated se-

quences of development: (1) “cognitive” development, that is, a child’s
capacity to recognize, identify, discriminate, and manipulate the features and
processes of the world around him; (2) development of the capacity to dis-
criminate and comprehend the speech he hears from others in his environment;
and (3) development of ability to produce speech sounds and sequences of
speech sounds that conform more and more closely to the patterns of adult
speech. The last of these depends on the second, and both of them depend at
least partly on the first. It is sometimes claimed that listening-comprehension
ability also depends on the development of speech ability, and although this
may be to some extent true, it is not a necessary dependency, as witnessed
by the occasional case of a child who develops good listening comprehension
without a corresponding ability to talk. The great dependency of speaking on
hearing is demonstrated by the enormous difficulties encountered in teaching
deaf children to speak in anything approaching a normal manner.
Actually, we have most information about the development of a child’s
speech responses; we know much less about the development of his ability to
perceive and interpret speech, and still less about his cognitive development
in the early years. In what follows, we pay primary attention to speech de-
velopment; cognitive development is considered in Chapter 6.
The organically determined cries of an infant in the first two or three
months gradually give way to the apparently random vocalization known as
the “babbling stage.” There is some ground for thinking that this stage is
biologically determined, since it occurs in many babies who are subsequently
found to be congenitally deaf. At the same time, the amount of babbling has
a significant dependence on the presence of adults who reinforce this babbling.
Although certain trends can be observed in the kinds of sounds emitted by
an infant in the babbling stage, these sounds have little bearing on the
phonemes of the language the child is to learn. In fact, some children virtually
stop babbling when they begin to learn their first “words.” It is at this point
that true language development starts.
The Learning
of Language
31
 to discriminate
Even before the end of the babbling period, infants start
first aspects of speech they learn to
speech forms. Probably some of the
stress and intonat ion pattern s that communicate
discriminate are certain basic
desires of adults. But at about the age of 11 months they
the feelings and
may infer that they
start learning to obey simple verbal commands, and we
ration of words and phrases.
can begin to respond to the total phonemic configu
stand them in good stead when they start
Doubtless these discriminations
speech themselv es, typicall y around 12 months of
to learn to produce real
by which a child learns the phonolo gy, vocabul ary, and
age. The process
even though
grammar of his language is actually fairly long and arduous;
very great deal to learn. All aspects
progress may appear to be rapid, there is a
of development progress simulta neously and are interrel ated.
gradual
A child learns the phonemes of his language through a process of
iation. Initially, the words he speaks may seem to have only an
different
pre-
approximate similarity to the words in the adult language that they
that even at this stage the child
sumably imitate, but the evidence suggests
has a phonemic system of his own, even though rudimentary. For example,
one
a child who pronounces car, cat, and cup as /tah, tet, top/ has only
phoneme, /t/, where the adult language has two, /k, t/. It may be several
years before the child’s phonemic system on the production side approximates
that of the adult language. According to one theory of child language de-
velopment (that of the linguist Roman Jakobson), the gradual differentiation
of a child’s phonemic system is correlated with the manner in which he
acquires the various distinctive features contained in phonemes (see p. 17).
The vocabulary development of a child is at first rather slow, six months
after he has said his first “meaningful” word, he may still know only a handful
of words. There comes a stage, however, when acquisition of vocabulary is
amazingly rapid; this seems to occur when in his cognitive development
the child has reached the point of perceiving that things, events, and prop-
erties have “names.” During this “naming stage” he learns to ask questions
like “What’s that?” “What’s that called?” “What does that mean?” By the
time the child reaches school, say by age six, his vocabulary is often quite
impressive, particularly if he has had rich verbal experiences in his environ-
ment, One estimate of average Grade I vocabulary, 23,700 words, is demon-
strably faulty, but the number of different morphemes known by a first
grade child could easily reach 7,500.
The manner in which a child learns the grammar of his language is still
very poorly understood. In effect, the child has to perform the feat of making
a kind of unconscious linguistic analysis of the language he hears, trying
various patterns until he finds the patterns that are accepted and understood
by his wards and that get him what he wants. That even at a fairly early
age he produces incorrect analogical formations like taked instead of took
demonstrates his capacity to respond to patterning in language. After the
child learns to say single words (which function grammatically as if they
were complete sentences), at about 20 months he begins to use simple two-
word constructions in a kind of “grammar” of his own.
Here, for example, are some two-word sentences recorded from the
utterances of a boy during the period 19-22 months: ?

1M.D.S. Braine. Language, 1963, 39, 1-13.

The Learning
of Language

32
 see boy my mommy nightnight office allgone shoe
see sock my daddy nightnight boat allgone vitamins
see hot my milk ra et allgone egg
re oa
do it byebye plane see fan Be ie
push it byebye man aa
close it byebye hot more taxi
buzz it more melon

Thus far, no study has been reported concerning when a child starts to
learn the active use of transformations. He could, of course, learn the various
transformations as independent constructions rather than learn them as
variants of the same constructions, but it is difficult to imagine that the child
would not take advantage of the transformational property of grammar.
The psychologist Roger Brown ? has demonstrated that children learn the
form-classes of words rather early. At least, by the age of about four they
have learned that the construction @ miss (where niss is a made-up word)
denotes a “‘count noun,” that is, something that comes in separate entities like
stones, but that in the construction Have you ever seen any niss? a “mass
noun” is denoted, that is, a substance like sand or water that doesn’t ordi-
narily come in identifiable specimens. They also recognize that a gerund,
nissing, denotes an action. Brown points out that in early stages of language
development, most nouns are concrete, tangible objects while most verbs are
observable physical actions. It would seem that very early in the course of
language development, children form concepts of the form-classes we call
nouns and verbs. (Doubtless they also form a concept of the adjectival form-
class, although Brown did not include adjectives in his study.) Nevertheless,
when children are asked to give word associations, their responses are not like
those of adults; that is, they are unlikely to give words in the same form
class as the stimulus. Perhaps this is simply because children have not learned
the “idea” of the word-association task.

In philosophy, where the study of meaning has had a secure and honored
place for centuries, the problem of meaning is generally approached by con-
sidering the relationships that are said to hold between signs and the things
to which they refer, or better, between signs, their referents, and the users of
signs. In the previous chapter we pointed out that linguists are extremely
cautious in their use of the notion of meaning; they are in any case more
preoccupied with the expression system of a language than with the content
or meaning system to which it presumably corresponds. Since this book is
about the psychology of language and thought, we cannot pretend to give
any satisfactory treatment of meaning from the standpoint of philosophy or
linguistics. But since “meaning” is somehow contained or involved in lan-
guage behavior, psychology can justifiably be expected to render an account
of the concept of meaning that is valid within its own frame of reference.
We shall try to suggest the outlines of such an account.

2R.W. Brown, J. abnorm. soc. Psychol., 1957, 55, 1-5. The Learning
of Language

ei)
 with the
Some psychologists feel that psychology should try to dispense
as possible. They feel it is dangero us to assume
concept of meaning as much
any “ideas,” “thought s,” or “meanin gs” which are ‘‘ex-
that there can be
by language in communi cation, because such things cannot be de-
pressed”
able
scribed or observed objectively. With this view we can have consider
the explanat ion of a great many
sympathy, but on the other hand we feel that
experimental and observational results would be extremely awkward unless
we permit ourselves to assume that certain kinds of unobservable, “covert”
mental events or responses take place. Introspection and subjects’ verbal
reports give valuable leads to what these covert events may be like, and under
certain conditions we can even detect the occurrence of such events. In
1935, for example, Neal Miller conditioned a psychogalvanic response (PGR)
to the letter T by pairing it with a mild electric shock delivered to the subject.
Later, when the subject was instructed to think “4, T, 4, Tica eter tore
series of successive stimuli, 4 to the first one, T to the second, and so on, it
was possible to detect the PGR only in response to the even-nu mbered stimuli.
If the PGR recorder had not been connecte d to the subject, there would have
occurred a series of “mental” events that would have been complete ly un-
observed. It is useful to assume that similar events occur in a great many
situations—both in daily life and in psychological experiments, even though
we usually have no ready means for detecting them. We must, however, be
sure to justify the inferences we make about such covert events, usually by
treating them as reduced or latent forms of responses that we can observe
and control.
The proper solution to problems of “meaning” comes, we think, from the
description of the ways in which human beings learn and use language signs
in a speech community. We assume that anything we may want to say about
meaning in the speech community as a whole can be accounted for by con-
sidering what is true for the individual members of the group taken in
aggregate.
The word meaning is itself a linguistic form; we must explicate the use
(“meaning”) of this word at the very same time that we are explicating the
concepts to which it corresponds, or the situations in which it arises. Solv-
ing the problem of meaning is essentially a bootstrap operation.
To construct a psychological theory of meaning, we can make use of several
kinds of information available to psychologists: (1) observations of children
learning linguistic behavior in naturalistic settings; (2) the paradigms of
learning yielded by psychological theory and experimentation; (3) the
results of experimentation in the teaching of linguistic behavior to human
beings; and (4) experimental studies of the linguistic behavior of mature
speakers of a language, that is, persons who have already acquired a system
of linguistic habits based on “meaning.”
Start with some simple cases. Even before he learns to speak, a baby learns
to recognize a particular speech form as a sign of some stimulus or class of
stimuli. The speech form may be a particular intonation contour, or it may
be a sequence of segmental phonemes. In either case, the learning paradigm
that seems to fit this case most directly is that of classical conditioning, where
a conditioned stimulus (such as the sound of the word dog) presented simul-

The Learning
of Language

34
Figure 3. Establishment
of a@ meaning response
through classical condi-
tioning. The joint pres-
entation of UncS and
CS must occur often
enough to allow a reliable
CR to occur.

taneously with, or just before, an unconditioned stimulus (a real dog, or a

picture of one), comes independently to evoke a conditioned response similar
to the unconditioned response evoked by the unconditioned stimulus (see
Figure 3) .3
But what, exactly, is the unconditioned response to the sight of a dog
or a picture of a dog? Some writers on this subject have tried to identify
such a response with overt responses, such as patting, withdrawal, signs of
emotion, but this line of reasoning is unnecessary and probably incorrect. It
seems sufficient to say that before a child starts to learn the meanings of
linguistic signs, he learns to make pure perceptual responses to objects and
events in the world around him. He recognizes certain stimuli or stimulus
configurations as being similar to configurations he has experienced before.
Certain faces, toys, items of wearing apparel, foods, and so on, come to have
perceptual identities in the child’s experience; the same can be said of qual-
ities of experience such as colors, sizes, intensities of sound, and experiences of
touch, as well as experiences of motor action (such as pulling, hitting, eating).
Perceptually, these experiences are of constancies; for example, a favorite
doll is recognized as a constant, identifiable experience no matter from what
angle it is viewed, and no matter how it is felt or touched. How such stimulus
configurations can be recognized as constancies is not directly our problem
here. For convenience, let us make the reasonable assumption that there are
in the repertoire of young children a large number of perceptual identifying
responses to common experiences. These responses are prior to, and inde-
pendent of, language; deaf children have them. They are covert responses
that are ordinarily unobservable, but their presence can be inferred from the
overt signs of recognition that the child often makes.
This long digression was necessary to establish what kind of responses
function as the unconditioned responses when an infant is conditioned to
3 For information on classical conditioning and other basic learning concepts applied
1964,
in this chapter, see S. A. Mednick. Learning. Englewood Cliffs, N.J.: Prentice-Hall,
a volume in the Foundations of Modern Psychology Series.

The Learning
of Language

35
 respond to a linguistic sign by classical conditioning.* But it is well known that
the conditioned response is seldom precisely the same as the unconditioned re-
sponse. So it is in the learning of a response to a language sign. When a child
becomes conditioned to respond to a linguistic sign such as “dog,” he does not
have the full unconditioned response; he does not hallucinate the sight of a
dog! Rather, the conditioned response is some fractional representation of the
identifying response to dogs, or to particular dogs. Or more generally, it is a
fractional representation of the identifying response to whatever perceptual in-
variant is involved in the linguistic sign. Many psychologists call it a mediat-
ing response, because it can become a stimulus for further behavior.
Thus, in a simple case, “understanding” of a linguistic sign occurs when
it evokes the conditioned response related to the unconditioned response that
would be evoked by the stimulus or stimulus configuration which this sign
“represents.” This conditioned response may be called a “meaning response.”
Both unconditioned and conditioned responses here may be covert and in-
accessible to external observation by any ordinary means. From the stand-
point of the hearer, the “‘meaning”’ of a linguistic form is the conditioned
response it evokes. Even in the early stages of language learning, this mean-
ing may be quite complex and may contain both denotative and connotative
components (these terms are discussed below). It may be partially “incorrect”
from the standpoint of adult language. But learning the “correct” meanings
of linguistic signs may be as much a matter of sharpening and revising percep-
tions as of learning anything about the signs. Learning that “dog” does not
apply to horses is partly a matter of refining the perceptual responses in-
volved in each case.
Parenthetically, it should be noted that linguistic signs themselves are
stimuli which the child has to learn to recognize as perceptual constancies,
just as he has to learn to recognize other kinds of stimuli. A word has to be
recognized no matter who says it, or how. Also, linguistic signs themselves
have certain perceptual qualities (for example, the smoothness of an m sound
or the shrillness of a sibilant) that become associated, even if only very
marginally and peripherally, with meaning responses; this fact may account
for some of the findings of “phonetic symbolism” (pp. 6-7).
So far we have talked only about meaning responses in the hearer of lan-
guage. We must also account for the behavior of the speaker who utters the
linguistic signs to which the child responds, and also for his learning of signs
when he was first learning language.
Since speech is a motor response, the learning model that seems most
appropriate for explaining it is the operant paradigm; an operant response
is one whose strength is a function of the degree to which it has been followed
by rewards (positive reinforcements). In the case of speech, the reinforce-
ment is always social, for it is provided by other persons in the individual’s
environment. B. F. Skinner ® has drawn attention to several ways in which a
speech response may arise. It may be learned as an echoic response, that is,
4It will be noticed that in this account we assume that the perceptual response is an
involuntary type of response similar to those of the autonomic system, and therefore
subject to classical conditioning. See Mednick, ibid., pp. 52-53 for a comparison of operant
and classical conditioning.
5B. F. Skinner. Verbal behavior. New York: Appleton-Century-Crofts, 1957.

The Learning
of Language
36
 If on one or more occasions On later occasions the
events occur as below likelihood will be increased
that events will occur as below

BASIC Motive; Motive;

PARADIGM
FOR AN In presence of In presence of
OPERANT
SP.
Sy (a discriminative
stimulus)
There occurs
Rj (an operant) Rj
Followed by

grein
ECHOIC Motivee
VERBAL (generalized) (approval) (generalized)
OPERANT In presence of In presence of

S: = adult to Se
provide San
and and

D
S: = heard verbal S$. = any heard
stimulus verbal stimulus

There occurs

Ri = echo of Si Ri i = an “imitation” of
D
Followed by Si

rein
Motivex Si Motiver
(e.g. thirst) (e.g. milk) (e.g. thirst)
In presence of In presence of

S> = adult S$? = adult

There occurs

Ri = utterance Ri utterance

(e.g.’’milk”) “manding’’ a stimulus

Followed by that will satisfy
motive

rein
Motivessnccncciieier-nere
creer enorite S¢ Motives
(approval)
In presence of In presence of

S> = adult S$: = adult

Si=a perceived .$; — the object

object

There occurs
Ri = utterance Ri = name of the
which names Si object
Followed by

Figure 4. Operant paradigms for the learning and maintenance of verbal

responses.
 as an imitation of a heard stimulus which the parent, say, may reward if it is
sufficiently similar to the stimulus. Or it may be learned as a mand—as a
response which starts out as a random speech utterance but which is adequate
to cause the parent to provide a stimulus that happens to satisfy some current
need of the child. For example, a parent might take a random utterance on
the part of the child as sufficiently close to ball to make him think the child
is asking for (manding) a ball, whereupon it is given to the child, thus satis-
fying the desire that (one may suppose) he happens to have for the ball.
Such a sequence, repeated several times, may enhance the probability that
ball or something like it will be uttered by the child whenever he wants a ball.
Still another way in which a verbal response can be acquired, according to
Skinner’s notions, is as a tact. A child who for any reason makes a particular
verbal response in the presence of (in conéact with) a given objective stim-
ulus, and is rewarded for doing so, may learn to make this response, or some
variant of it, whenever he experiences the relevant stimulus.
It will be noted that we have said nothing about “meaning” in this account.
Skinner’s formulations concern only the objective relationships between
certain stimuli and certain responses; in his view, the “meanings” of the lin-
guistic forms that happen to be involved in verbal responses can be completely
accounted for by stating the contingencies under which the verbal responses
occur.
Skinner’s paradigms can be demonstrated experimentally; in fact, they
are matters of common observation. With the mand paradigm one can train
a child to make a certain verbal response whenever he has a particular need;
and with the tact paradigm one could readily train a child to name something
with any arbitrary verbal response one might like. The only difficulty that
might be encountered in either of these cases is that one might have to wait
a long time before the desired response occurs, to give one the opportunity to
reward it in an appropriate stimulus context. This difficulty can be avoided
by first teaching the child to make echoic responses, and then chaining mands
and tacts to these. At a later stage of the child’s development, he could be
taught to make texting responses (that is, verbal responses to printed or
written text stimuli) as discriminative operants, and these in turn could be
chained with mands, tacts, and other kinds of verbal responses. One could
thus in theory build up a quite elaborate system of verbal responses in the
child. In fact, this formulation underlies the “programed instruction” or
“teaching machine” movement.
Compelling as Skinner’s formulations may be, they present certain
theoretical difficulties, and not all psychologists are satisfied with the solu-
tions that Skinner has proposed. The theory as a whole banks heavily on the
concept of reinforcement, but not everyone is willing to accept the proposition
that reinforcement is the crucial factor in learning.® Also, the theory cannot
easily account for the fact that a language response learned in one way
generally is immediately available for use in other ways. A child who has
learned to understand a word (through classical conditioning) may later use
it as a mand (that is, a means of satisfying some need) or as a tact (that

6 This matter is discussed at length in the book in this series about learning, by Mednick,
already cited.

The Learning
of Language
38
is, aS a name of some object or event) without going through the particular
behavioral processes theoretically required in learning mands or tacts. It
seems desirable to postulate behavioral links among these different processes;
after all, they all occur in the same organism—an organism complex enough
)
surely, to allow for such links.
If we look again, more closely, at the paradigms postulated by Skinner,
we notice that in all cases there must be covert perceptual responses to the
rewards (in the case of mands) or to the discriminative stimuli (in the case
of tacts). In the process of operant conditioning, then, classical condition-
ing, or something very much like it, must be going on in parallel. That is to
say, a response that we may regard as a “meaning response” is conditioned
to the reinforcement (in the case of a mand) or to the discriminative stimulus
(in the case of a tact). Since the meaning response is a conditioned, covert
perceptual response to a linguistic sign, whether it arises in learning to speak
or to comprehend language, once learned it can function in any of these con-
texts, and this fact would account for the transfer that takes place from one
behavioral context to another.
So far we have arrived at the conclusion that in the early stages of lan-
guage learning, “meaning” arises from the fact that many linguistic forms
evoke conditioned, covert perceptual responses. Eventually, a child becomes

Figure 5. Classical con-

ditioning accompanying
mands and tacts.
 s it, meaning
aware of or generalizes this meaning relationship; as He perceive
the one hand, and objects,
is a direct correspondence between words, on
affairs in the world of his experience.‘
events, qualities, and other states of
we have called the “naming stage” when the
This perception occurs in what
that objects have names and that the meaning s of
child becomes aware
at a concept
unfamiliar words can be explained to him. Thus, he arrives
like mean-
of the word meaning, that is, he learns how to use expressions
ing and to mean something.
in
For the psychologist, however, the meaning relationship lies wholly
is a relations hip between experienc es that we call “languag e
behavior. It
s” or
signs” and other experiences that may be called “‘meaning response
“con-
“mediating responses” that are gathered together or organized into
cepts.”
We are now ready to sketch how this account of meaning deals with several
issues concerning meaning that are of particular importance in psychology.

Denotative Meaning
In all the paradigms of
verbal learning we have presented, the child gradually learns what range of
situations yields the highest probability of social reinforcement. That is,
through processes of discrimination learning and stimulus generalization, a
child learns what properties or patterns of stimulation are critical for social
reinforcement when he utters a given linguistic form. For example, he learns
what characteristics an animal has in order for it to be called a “dog.” To
the extent that this learning on the part of the child corresponds to com-
parable processes of learning on the part of other members of the speech
community, we may say that the child has learned the denotative meaning of
the form in the speech community. We can describe the denotative meaning of
a form by specifying the properties or patterns of stimulation which are
essential—that is, criterial—for its socially approved use in the speech com-
munity. Dictionary definitions are successful to the extent that they can do
this.
In theory, this analysis can apply to every item in a linguistic expression
system—that is, not only to the words that are names of objects, events,
and attributes in the physical and biological environment, but also to the
words that name abstractions and relationships, and to words, forms, and
constructions that have a purely grammatical function. Sometimes the pat-
terns of stimulation that are criterial for the use of grammatical elements
are solely verbal, for example, the grammatical context that evokes the use
of the infinitive marker to.
As the uses of words and other elements of a linguistic expression system
are being learned, a corresponding development of implicit mediating re-
sponses takes place. This probably occurs by virtue of the paradigms of
classical conditioning of perceptual responses that we have given. The per-
ceptual constancies, or invariants, that the child acquires preverbally are to
some extent modified and re-sorted along the lines dictated by the referential

7In her autobiography, Helen Keller describes vividly how as a deaf-blind child she
first became aware of this relationship.

The Learning
of Language
40
patterns of the language symbols he learns. The denotative meaning of a
linguistic form is reflected in a “concept”—a bundle of implicit mediating
responses which are linked with the properties and patterns of stimulation
that are criterial for that form in the speech community.

Connotative Meaning
As an individual accumu-
lates experience with the patterns of stimulation corresponding to a given
linguistic form, he responds not only to the criterial attributes of these pat-
terns but also to the noncriterial attributes—attributes that occur with these
patterns in either external or verbal contexts with considerable regularity but
do not govern reinforcement by the speech community. For example, “like-
ableness” may be a frequent attribute of “dogs,” but it is irrelevant to the
denotation of the word dog. An individual’s responses to noncriterial
attributes become attached, through conditioning processes, to the meaning
responses or concepts evoked by a linguistic form. That part of the mean-
ing response which does not correspond to criterial attributes may be called
the connotative meaning of a form. Fundamentally, connotative meaning is
an individual matter because it depends on the experiences that an individual
has happened to have. Since the experiences of individuals in a speech
community are in general rather similar, there are many similarities among
the connotative meanings they have. But to the extent that people’s experi-
ences and attitudes differ, connotative meanings can also differ. Even on the
assumption that people agree on the denotation of a word like Democrat,
we still cannot say that they will agree widely on its connotation. In Chapter
7 we describe some methods for studying the connotations of linguistic sym-
bols and the concepts to which they correspond.

Meaning fulness
The concept, or meaning
response, associated with a word experienced in a wide variety of contexts
will expand in the extent of its connotative meaning, and this richness of
connotation may be called ‘“‘meaningfulness.” One way of measuring meaning-
fulness, due to C. E. Noble, is based on the rate at which subjects give verbal
associations to a word. Words of very rare or limited use or nonsense syllables
are found to have low degrees of meaningfulness, although it is difficult to find
nonsense syllables that are completely devoid of meaningfulness by this
measure.

Situational Meaning
The problem posed here is
this: What is the “meaning” of a sentence uttered in a particular situation,
and how is it related to the meanings of the linguistic forms and construc-
tions that compose it? A sentence can be likened to a computer program;
in fact, that is precisely what it is: a set of directions for the human thinking
machine. The hearer or the reader of a sentence constructs its meaning by
following the “directions” it provides in terms of the concepts and conceptual
relationships it evokes, also utilizing whatever further information he may
have concerning the situation in which he hears it. This process may be called
The Learning
of Language
4I
 ted, like
interpretation. A string of linguistic signs that cannot be interpre
r,” is devoid of situatio nal
“words straighten poverty without every encounte
meaning.

Intention vs. Meaning

If a sentence is like a pro-
gram for a thinking machine, it is also an artifac t created by the speaker.
The speaker’s intention in creatin g a senten ce is not necessarily correlated
with the situati onal meanin g the senten ce has for the hearer. The speaker
as
may not be fully successful in creating a sentence that will be interpreted
ful in creati ng a senten ce
he intends it to be, or he may be only too success
that he knows will be interpreted in a way that will deceive the hearer.

SECOND-LANGUAGE LEARNING AND BILINGUALISM

There are cases when the environment of a child contains more than one
language—as when each parent speaks a different language, or when the child
is reared in a bilingual community. In the early years of childhood, such
children can often learn more than one language with ease. The circumstances
of learning are like those of a mother tongue in each case. Sometimes there
are interferences, of course: Occasionally responses from one language
system will intrude into speech in the other language. It appears that learning
is most successful when the situations in which the two languages are learned
are kept as distinct as possible: For example, the child learns one language
from one parent, the other language from the other parent. This conclusion
is in accord with the theoretical proposition that the cues for competing re-
sponses should be kept as distinct as possible. Resulting from such a learning
situation is what may be called coordinate bilingualism, because the two lan-
guage systems tend to be parallel and independent of each other, with inde-
pendent sets of meaning responses. There is no good evidence that such
bilingualism retards mental development; most instances where retardation
has been reported can be explained as the result of attempting to teach the
child in a language he has not learned adequately.
Past the age of early childhood, it appears to be much more difficult for an
individual to learn a second language system coordinate to a well-learned first
language. Typically, a person learns a second language partly in terms of the
kinds of meanings already learned in the first language. In this type of bi-
lingualism, called compound bilingualism, second-language responses are
grafted on to the first-language responses, and both are made to a common
set of meaning responses. Other things being equal, the compound bilingual
is less fluent in the second language, and the kinds of expressions he uses in
the second language bear tell-tale traces of the structure of the first language.
Efficient methods of teaching second languages attempt to duplicate certain
features of the learning situation that produces coordinate bilingualism—
maximizing the degree to which new language responses are made to objective,
nonverbal situations or to contexts utilizing previously acquired responses in
the new language, and minimizing the use of the native tongue except where
greater efficiency in teaching is attained by using it to explain meaning and
The Learning
of Language
42
grammatical points. In view of the large number of new habits that must be
made as highly automatic as possible, successful second-language learning re-
quires a considerable investment of time, a major proportion of which must
be spent in repetitive drill. Audiovisual devices, such as tape recorders and
teaching machines, can be of considerable assistance in language learning.
Sometimes language teachers overlook the importance of conducting drill in
accordance with principles of learning. For example, drill is probably of little
use unless there is more or less constant feedback of information to the learner
concerning the degree to which he is approximating the desired responses.
According to widespread opinion, many of the difficulties the learner has
with the phonology, vocabulary, and grammar of the second language are due
to the interference of habits from the first language. To some extent this may
be true, but it has also been reported that with very careful shaping of new
responses by a proper schedule of reinforcement, the problem of interference
is considerably reduced.
In thinking about the learning and teaching of second languages, we may
find it useful to refer to some distinctions made by the anthropologist E. T.
Hall. When conducted in a school situation, second-language learning tends
to be largely what Hall calls formal learning—learning guided by conscious,
deliberate effort on the part of the learner; there is also considerable infusion
of what he calls technical learning—learning guided by the application of rules
and logic. Very little of it is similar to the kind of informal learning—which
takes place “out of the learner’s awareness”—that occurs in much early first-
language learning. Although formal and technical learning may have some
place in second-language learning, it is probable that a faster, more appro-
priate kind of learning can be attained by shifting the balance in favor of
“informal” learning.

8. T. Hall. The silent language. New York: Doubleday, 1959.

The Learning
of Language

43
 Aspects
of Language Behavior Now that we have studied

the nature of language as a sign system (Chapter 2) and

the basic principles involved in the learning of language

(Chapter 3), we are ready to go into some detail on how

language behavior actually takes place. The bulk of this

chapter will be concerned with the production and under-

standing of speech—in both physiological and psycho-

logical aspects. Here we will find it useful to consider

language behavior from a statistical point of view. The

latter part of the chapter, however, will consider the

application of the psychology of language to reading.

44
 The neurological processes involved in speech are extremely complex. Not
only must all the muscles controlling the speech mechanism be precisely
coordinated in order to produce acceptable sounds, but also the utterance must
be composed and arranged in such a way as to be meaningful. Here we shall
discuss only a few important generalizations from scientific studies of the
neurophysiology of the speech apparatus.
It is generally agreed that speech functions take place in only one hem-
isphere of the brain, usually the side of the brain opposite to hand preference;
thus, the left hemisphere is the “dominant hemisphere’’ of most right-handed
people. In doubtful cases medical specialists can determine laterality by
noting which side affects speech when sodium amytal is injected into one of
the arteries supplying blood to the brain. There are cases on record of complete
hemispherectomy; if the hemisphere excised controls speech, the patient will
not be able to learn or relearn language unless he is still quite young (no
more than 10 years of age, say).
Beyond this, brain physiologists either cannot agree on the precise func-
tion of the several cortical areas, or feel the evidence is insufficient to allow
any definite conclusions. In the nineteenth century, Paul Broca’s discovery
(1861) of a “speech area” in the left temporal lobe, along with various other
observations of the speech difficulties of individuals with identifiable brain
lesions, held out the hope that it might be possible to assign specific lan-
guage functions to particular regions in the brain, and a number of such areas
were marked out. Careful sifting of the evidence accumulated since then,
however, does not support any such simple account. For example, when a
lesion occurs in a certain brain region, accompanied by a particular behavior
symptom, it is impossible to ascertain what the function of that brain region
is, because it may be either an area that originates some kind of neural im-
pulse, or one that integrates and transmits impulses received from another
area. Further, brain lesions are seldom well localized; often they are the
scattered areas served by a dysfunctioning blood vessel. On the other hand,
Wilder Penfield’s technique of observing patients’ speech reactions during
stimulation of the surgically exposed cortex promises to enable us to assign
functions of speech behavior to specific areas of the cortex (see Fig-
ure 6).
We can make certain inferences about the neural control of speech pro-
duction from the phenomena caused by delayed auditory feedback. These
rather astounding effects were discovered with the introduction of the tape
recorder. When equipment is specially arranged so that whatever a person
as in
says is fed back to his ears through headphones, not instantaneously,
a short time lag of up to one second, his speech shows
normal speech, but with
pat-
certain kinds of disturbances—changes in rate, intensity, and temporal
The effects are most
tern—that sometimes sound like a bad case of stuttering.
there
pronounced if the time-delay is about one-fifth of a second. Although
to resist them, every normal-he aring
are individual differences in the ability
normal speech in-
person shows some effects. The phenomenon suggests that Aspects

of Language
Behavior

45
 Vocalization :
Superior speech cortex
(supplementary motor)

Rolandic
Vocalization
Lip motor area
Voice
Jaw
control Tongue
Throat

Anterior speech Posterior speech

cortex (Broca) cortex (Wernicke)

Ideational speech

Figure 6. Speech areas in the dominant hemisphere (side view, below;

medial section, above), after W. Penfield and L. Roberts. According to these
authors, three areas, as shown, are devoted to the ideational elaboration of
speech, and two areas to its vocalization. (Adapted from Speech and brain
mechanisms by Penfield and Roberts, by permission of Princeton University
Press, copyright 1959.)

volves a perceptual self-monitoring process which is interfered with by the

delayed feedback.

Speech in ordinary discourse occurs at various rates depending on the

situation and the characteristics of speakers. The situation evokes speech
partly as a function of the extent to which the individual has been reinforced
or rewarded in similar situations. For example, William Verplanck has shown +
that people are more likely to continue a conversation if their interlocutors
agree with them than if they disagree with them; at least, for the people used
in Verplanck’s study, agreement was apparently more reinforcing than dis-
agreement. (In reinforcement theory, a stimulus that enhances the probabil-
ity of a response is defined as a reinforcer.) A previous history of reinforce-
ment or nonreinforcement is one possible explanation for the large individual
differences in the degree to which people participate in discussions, ask
questions in class, or tell jokes.
An even more basic trait is a person’s characteristic speech rate when he
is well motivated to talk, a rate that varies widely from individual to individ-
ual. It is not so much a function of the rate at which a person speaks the
1W.S. Verplanck. J. abnorm. soc. Psychol., 1955, 51, 668-676.
Aspects

of Language
Behavior
46
words themselves, as of the degree to which he interrupts the flow of his
speech with pauses and hesitation signs like [ah, or, om]. Frieda Goldman-
Eisler has demonstrated consistency within the individual: For the slower
speakers, the momentary “bursts” of rapid talking usually occur with highly
redundant, automatized utterances, giving way to slower speech when the con-
tent is more involved and less redundant.”
Most people, if not all, exhibit hesitation phenomena in their speech, with
either filled pauses containing sounds like those mentioned above, or simple
unfilled pauses, that is, silences. Here again, there are wide individual dif-
ferences; some people have trained themselves to avoid at least filled pauses.
There is some evidence that these two types of pauses have different functions
in speech. Unfilled pauses result from thinking to formulate ideas; filled
pauses are more likely to reflect anxiety. Study and observation of these
hesitation phenomena can have two uses; first, they are presumably reflections
of psychological processes in the construction of sentences, and second, they
can be shown to have significance for the clinical psychologist. In fact, any
deviation of spontaneous speech from what a simple reading of an edited
transcript would be can conceivably have some psychological significance,
and this includes all the various kinds of false starting and backtracking, voice
quavers and tremors, and abnormal stress and intonation patterns that a
trained linguist might detect in a highly detailed phonetic transcription of an
utterance. A number of clinical psychologists have found these phenomena
useful in identifying sources of anxiety and conflict in patients undergoing
therapeutic interviews.

THE CONSTRUCTION OF UTTERANCES

How does a speaker who has learned his native language well construct
utterances? Our discussion will revolve around the speaker, because we can
assume that a writer merely writes down, very deliberately and perhaps in
highly edited form, the sentences he imagines he would say as a speaker.
The differences between spoken and written language are for the most part
obvious and need not be commented upon here.
For the purposes of a scientific discussion, we have to make the assumption
that everything a speaker may say is completely determined, that is, de-
pendent on antecedent conditions. The causal relations may often be very
complex; indeed, we have to speak of multiple causation, because any act
of speech may be affected by a large number of antecedent conditions operat-
ing simultaneously. It is our task to establish generalizations concerning the
way antecedent conditions affect a given kind of event. It is often surprising
how satisfactorily a given speech act can be accounted for retrospectively.
For example, in a classic article on “the psychology of invention in a simple
case,” the educational psychologist Thorndike * asked students to guess or
make up the meanings of rare or nonsense words like amerce, besom, and
debrag. Even though most of the responses to any given stimulus were

2F. Goldman-Eisler. Brit. J. Psychol., 1951, 42, 355-362; 1954, 45, 94-107.
3 EB. L. Thorndike. Psychol. Rev., 1949, 56, 192-199. Aspects

of Language
Behavior

47
 different, Thorndike believed he could give plausible reasons for the guesses.
One student, for instance, defined amerce as “to duck under water,” ob-
viously influenced by the word immerse.
A retrospective account, of course, is not a satisfactory demonstration of
the power of scientific generalizations; we can validate our generalizations
better if we show that we can predict future events with a high degree of
success. To what extent can we do this in the case of language?

Word Association Studies

One of the most exten-
sively studied kinds of verbal behavior is what is called free association. (We
didn’t say “thoroughly studied,” for there are still plenty of problems in it
to challenge the researcher.) The standard technique is to present a subject
with a number of words (usually printed) and ask him to respond to each
word, orally or in writing, with “the first word that comes to mind.”’ Various
kinds of stimuli can be presented, but, chiefly through force of tradition, most
of the research has been based on the Kent-Rosanoff (K-R) list, named after
two psychologists who developed it in 1910 as a device for eliciting verbal
responses that might reveal personality, motivation, and affect. (The list is
seldom used for that purpose now.) It consists of 100 nouns and adjectives,
a few of which can also be taken as verbs. The free association task is ad-
mittedly not highly similar to anything that occurs in normal everyday
behavior, but it serves as a relatively simple, circumscribed setting for
laboratory studies of verbal responses.
Here we are chiefly interested in how well we can predict an individual’s re-
sponse to a word. Even without knowing much about him except that he is a
native speaker of English, our predictions will be reasonably good—very good
in the case of some words, for lists have been drawn up of the more common
responses to the K-R lists in large representative samples of adolescents and
adults. In one such listing,* over the whole K-R list, the most common, or
“primary,” responses accounted for 37.5 per cent of all the responses; the
second most common responses accounted for 13.7 per cent of the total;
and the third most common accounted for 8.0 per cent; in all, the three
most common responses accounted for 59.1 per cent of all the responses.
These figures represent a degree of prediction that would be highly satis-
factory in other contexts, say, predicting the weather. But they can be made
even better by taking account of certain characteristics of the respondents.
For example, the age of the respondent makes a considerable difference.
Whereas adolescents and adults tend to give what have been called para-
digmatic responses (that is, responses that are of the same part of speech as
the stimulus), children tend to give syntagmatic responses, that is, responses
that might naturally follow the word in a sentence: for example, BRIGHT—
sun. It has been claimed that this result means that children do not have
their language responses organized into form classes as well as adults do, but
it may also mean, as suggested on page 33, that children have a different
concept of the task of giving “the first word that comes to mind.’ That. is,
4W. A. Russell and J. J. Jenkins. The complete Minnesota norms for responses to
100 words from the Kent-Rosanof word association test. University of Minnesota
Department of Psychology, 1954.
Aspects

of Language
Behavior

48
the kind of set with which the task is approached may be an important de-
terminant of response.
The role of set is seen, for example, in the consistent individual differ-
ences ° among people of college age in the tendency to respond with words
that are opposites (antonyms) of the stimulus words, or with words whose
meaning contrasts in some way with that of the stimulus word (for instance,
moth as a response to BUTTERFLY). Some people give such responses on
almost every occasion; others rarely do. Therefore, if we can make a prior
determination of an individual’s tendency to make this class of response, we
can predict particular responses to the K-R test with even more accuracy than
we can from the general population statistics.
Of course, it is also possible to arrange cues that have a high probability
of triggering specific responses. One way of doing this is to give the first letter
of the response desired, and the number of letters in it. Try these stimuli:

Another way is to give the stimulus word only after several other words have
been presented. Davis Howes and Charles Osgood found that whereas the
normal percentage of hell as a response to DARK was very small, it rose
markedly when the word was given in the context of the words devil—fear-
ful—sinister.
All these experiments tell us something about the organization of verbal
response repertoires in those who have learned a language well—namely, that
when the context arouses a certain set, a given stimulus has a high probability
of evoking one of a small number of responses that can be specified in ad-
vance. The greater the number of background context factors, the more
highly determined (that is, predictable) is the response. We could expect
this generalization to apply not only to responses in a free association test but
also to verbal behavior in general.
Having persuaded ourselves that speech behavior can be treated scien-
tifically in a deterministic system, we can now inquire into the possible
stimulus-response patterns that operate in a normal speech situation. To
make the problem concrete, let’s consider the antecedents of the following
utterance:

“Tt’s getting stuffy in here; would you mind opening the window?”

You can easily imagine the situation in which this utterance might occur.
The first part might be called an observation concerning the situation; the
second part, a request. Look at the first part. In form, it is a predication about
it; but notice that in English we have a large class of utterances beginning
with it and making observations about general conditions: the weather (7’s
warm today), the time (it’s Tuesday, it’s late; it’s five o’clock). Lin-
guistically, perhaps, we might say that it is a substitute for the weather,
the atmosphere, the day, or the time. But there are certain expressions
where any substitute for it would be extremely awkward and rare. For
5 J. B. Carroll, P. M. Kjeldergaard, and A. S. Carton. J. verb. Learning verb. Behav.,
1962-63, 1, 22-30. Aspects

of Language
Behavior

49
 example, /t’s the first week of July would hardly be said as The week is the
first one in July. Furthermore, normally the pronoun it is used only as a
substitute for some previously named nominal, or as a “temporary sub-
ject” (as in the sentence to follow). It seems reasonable to assume that the
construction Jt + (predication-about time or weather) is a highly predictable
response to any situation in which the speaker has occasion to comment on
time, weather, or other general environmental circumstances; that is, it is
a response that is learned as a separate and distinct unit.
The formulation of an utterance It’s getting stuffy in here is normally very
rapid; perhaps it is gratuitous to think that the speaker’s performance can
be dissected into a series of decisions made in sequence. Nevertheless, hesita-
tions where the speaker seems to search for the right word (as in Jt’s getting—
er, well, stuffy in here), suggest that selection takes place on two levels; first,
on a grammatical level (selection of a construction or a transformation of one,
including the selection of form classes), and second, on a lexical level (selec-
tion of particular words, in appropriate form classes, to fit in the construc-
tions). Thus, having selected the construction i¢ + (predication about time
or weather) the speaker proceeds to fill in his construction. But this construc-
tion in turn demands another construction, namely, some form of predication;
the form selected (linking verb + adjectival) is presumably a response to
an experienced quality (the atmosphere). Other components of the situation
—current time, perceptible change in conditions or the immediate surrounding
environment, and place—dictate filling in this nest of constructions in such
a way as to produce the utterance It’s getting stuffy in here.
It is worth noting that a declarative construction was selected as the form
for the sentence as a whole. The speaker might have said Jsn’t it getting stuffy
in here? or How stuffy it’s getting in here! Linguistically, these are both
transformations of the basic declarative construction. An interesting question
is this: If one of these transformations had been used, is there some sense in
which it was psychologically a transformation? That is, was there some
process by which the finished utterance was derived from a “kernel” declara-
tive sentence? ® We cannot currently answer this question. George Miller has
developed evidence suggesting that the speed with which people can make or
find transformations of sentences presented to them is related to the number
and order of transformations that are involved in the process, but this says
nothing about how sentences are formulated in the first place.’ A person
could learn to formulate nonkernel sentences as easily as kernel sentences.
For example, the second part of the hypothetical utterance cited above is
linguistically a request in the form of a yes-no transformation of You would
mind opening the window. But there is little reason to suppose that the
speaker formulates the utterance as a kernel and then transforms it. The
actual unit of selection may, in fact, be a construction already transformed.
In the present case, Would vou mind . . . is such a frequent way of formulat-
ing a polite request that it may well be, idiom-like, learned as a distinct unit;
that it conforms to a certain linguistic pattern is psychologically of no more

6 See p. 25.
7G. A. Miller. Amer. Psychologist, 1962, 17, 748-762.

Aspects
of Language
Behavior
50
interest than the fact that an idiom like Mark you well deviates from the
normal pattern for the imperative transformation. Exactly what is a selection
unit in the behavior of an individual depends on his behavioral history, and
only secondarily on linguistic units. Many whole utterances are learned as
units, sometimes with little meaning, or incorrectly as in the case of a child
in the first grade who thought the flag salute began J pledge a legion to the
flag. Most of the available studies of verbal behavior that we can report—
studies of free association, controlled association, naming behavior, cate-
gorizing behavior, and so on—have more to say about the selection of lexical
units, or “content” words, than about the selection of grammatical construc-
tions, and thus what we have said here about selection processes is largely
speculative.
If we refer to the list of basic expression-types in Chapter 2 (p. 24), we
see that at the highest level of utterance formulation, the “choice” is between
four kinds of nonsentential expressions or two kinds of sentence-types. It
would not be difficult to specify the situations that evoke each of the four
nonsentential expressions (greetings, calls, exclamatories, and responses to
another speaker).
The basic situation evoking a sentence-type (as distinguished from a non-
sentential expression) is one in which “information” concerning some stim-
ulus—either objective or subjective—is handled. If the sheer existence of the
stimulus is in question, the situation calls for use of an existence-assertion
sentence type; if some perceived attribute of the stimulus is in question,
selection of some form of predication occurs, the exact form depending on
the kind of attribute involved. For example, an attribute perceived as a
“variable quality”’ becomes a predication with a linking verb and an adjective
construction; a large class of other perceptions is encoded into predications
with an intransitive or transitive verb.
Likewise, the situations evoking certain transformed constructions can be
specified and experimentally verified. The writer, for example, has studied
the elicitations of statements, questions, and imperatives in a miniature two-
person game situation. As might be expected, an utterance will normally be
in declarative form (the null transformation) if the speaker perceives his
information to be greater than that of his hearer; it is likely to be in the form
of a question if he perceives his information to be less than that of his hearer.
It can be in imperative form if he desires his hearer to perform an action or do
something to arrive at some particular state of affairs, but social amenities
permit the unadorned imperative only when the action is for the hearer’s own
benefit (Drink some water) ; otherwise the imperative (Give me some water)
is either supplemented with the expression Please or is replaced by questions
(Will you give me a glass of water?) or statements (I’d like some water).
Experimental results have also suggested that the choice of what element
in a situation is to be the subject of a sentence is an important determinant
of whether the active or passive voice will be used. In reporting a baseball
game we might say “Jones hit Smith with a fast ball” if we are primarily
interested in what Jones did, but we would be likely to say “Smith was hit
with a fast ball” if we are detailing what happened to Smith and why he was
sent to the hospital.
Aspects

of Language
Behavior
51

A Sbt
 Victor Yngve ® has developed a model of language structure which sug-
gests some of the processes whereby we produce sentences. He assumes that
all rewrite rules (p. 20) in a grammar are binary, that is, that no construc-
tion can be expanded into more than two other constructions. For example,
a noun phrase construction can be expanded into another noun phrase and
a prepositional modifier construction, as in pictures — pictures + of Rome.
The process of producing a sentence, according to Yngve, is one of succes-
sively expanding constructions until every construction is either expanded
into two other constructions, or filled in with words. Since uttering a sentence
is a temporal matter, the “left-hand” parts of any construction must be
expanded before the right-hand parts. Yngve observes that we tend to avoid
making too many successive expansions of left-hand parts of constructions,
because every time we do, the ut-
terance (and, perhaps, the expan-
sion) of the right-hand parts is (Noun phrase)
that much more delayed and less
likely to be remembered. The Lh RII
phrase very much more clearly Lie
projected pictures of Rome (dia- L[3
grammed in Figure 7) contains Ll4
five successive left-hand expan- L{s
sions and one right-hand expan-
sion. Yngve suggests that the very much more clearly projected pictures of Rome
number of nested left-handed ex-
pansions (which he calls _ the Figure 7. Left-hand (L) and
“depth” of a sentence’s grammar) right-hand (R) expansions of a
is generally limited by human phrase.
memory storage capacity. His hy-
pothesis is that depth seldom ex-
ceeds the “magic number 7 + 2” that George Miller has pointed to as the
measure of human memory storage capacity. The transformations noted by
other grammarians are in Yngve’s view mainly useful in enabling us to avoid
excessive grammatical depth. For example, instead of saying ‘““What what
what he wanted cost in New York would buy in Germany amazed us” (as
we would have to say using only active verbs) we can say ‘““We were amazed
by what could be bought in Germany for the cost in New York of what he
wanted.”

Tabulations of the frequency of occurrence of phonemes, letters, words,

and other language units are of practical usefulness in such enterprises as
developing teaching materials, designing systems of stenography, and break-
ing secret codes. These tabulations have also been of interest to psychol-
ogists and others interested in language behavior, and they are particularly
useful in studying speech production and reception.
8 V. Yngve. Scientific American, June 1962, 206 (6), 68-76.
Aspects

of Language
Behavior
52
 G. K. Zipf some years ago noticed that in English and several other
languages there was a rather close relation between word frequency and
word length and interpreted this as evidence of a natural law of efficiency in
language structure: Imagine how inefficient it would be to write or speak a
language in which all the common words were long and the rare ones short
(if, indeed, there could be sufficient variety in short, rare words). Zipf also

Figure 8. The relation between word length and word frequency (in terms
of word tokens occurring in typical samples of text). (Data from G. A, Miller,
E. B. Newman, and E. A. Friedman. Inform. and control, 1958, 1, 370-389.)

noticed an even more interesting relationship—what he called the rank-

frequency law. If you take a compilation of the frequencies of words in large
samples of English, such as the well-known Thorndike-Lorge count,’ arrange
the words in order of frequency, and then plot the logarithm of the word
frequency against the logarithm of the rank of the word in frequency, you
will usually get an almost perfect straight line with a negative slope of 45°.
Mathematically, this implies that frequency times rank will yield a constant
value for the whole frequency range in a given sample. Zipf believed that
deviations from this straight-line relationship, as might be found in samples
of the speech of children or of schizophrenics, are valuable indicators of

New York:
9 E. L. Thorndike and I. Lorge. The teacher’s word book of 30,000 words.
Teachers College, Columbia University, 1944. Aspects

of Language
Behavior

53
 Figure 9. Cumulative
word-frequency distribu-
tion for the Lorge Maga-
zine Count (4.5 million
words), plotted on loga-
rithmic-normal coordi-
nates. (Adapted from
data of D. Howes. In J.
Money (ed.). Reading
disability. Baltimore:
Johns Hopkins Press,
1962.)

developmental or other abnormalities in language behavior. Davis Howes 1°

has elegantly demonstrated that this is true for the speech of aphasics, al-
though he believes the word-frequency distribution is better described as a
logarithmic normal distribution (see Figure 9).
The word-frequency distribution, however it is analyzed, reflects a relation-
ship between the number of types (different words) in a sample and the
number of tokens (total words). (The terms type and token were introduced
in Chapter 1, page 2.) ‘Sometimes the type-token ratio (the number of
different words divided by the number of total words) is used as a measure
of the diversity or richness of vocabulary in a sample, but it should be noted
that this ratio will tend to decrease as sample size increases, other things
being equal, because fewer and fewer of the words will mot have occurred in
the samples already counted. A measure of vocabulary diversity that is ap-
proximately independent of sample size is the number of different words
divided by the square root of twice the number of words in the sample.
The development of a mathematical theory of information transmission
by Claude Shannon, a communications engineer, has enabled psychologists
to gain new perspectives on statistical studies of language behavior.!! The
10 Davis Howes. A quantitative approach to word blindness. In John Money. Reading
disability. Baltimore: Johns Hopkins Press, 1962, p. 135.
11C, Shannon and W. Weaver. The. mathematical theory of communication. Urbana:
University of Illinois Press, 1949.
Aspects

of Language
Behavior

54
key idea is a method of measuring the “information” contained in a message.
“Information” is used here in a special sense—it really means “‘the informa-
tiveness of the symbols in a message relative to one’s expectations of those
symbols.”’ Here are three ten-letter “messages,” in each of which the last
letter has been deleted; try to guess what letter has been deleted:
AAAAAAAAA-; PRRNWBITK-; GENERATIO.-. For the first one, prob-
ably you will guess that the deleted letter is A, and if indeed the tenth
letter is A it will not be very informative for you. In fact, if the tenth
letter is something other than A, say F, it will probably strike you as
very informative. Since the second message is a random sequence of let-
ters, you would have considerable uncertainty as to what the next letter
might be, and therefore information as to what symbol it actually was
would doubtless be very “informative.” For the last message, you can be
almost certain that the letter N was the one deleted, and again, if you
then learn that it was indeed N you will not be much surprised; that is,
the symbol N as it appears in the message is not very informative, whereas
if it turned out to be L, say, you would be much surprised. Note how un-
certainty prior to the receipt of the symbol is related to the informativeness
of the symbol after you have received it: The less certain you are, the more
informative the symbol is, but at the same time, the more certain you are,
the more informative an unexpected symbol is. Information theory provides
a mathematical way of measuring the informativeness of the symbols of a
message (and thus of the message as a whole) in terms of the probabilities
of those symbols, but it would take too much space to explain this and it is
not really necessary for an elementary understanding of information theory
ideas.”
One important, easily understood idea in information theory is that the
amount of information in a symbol is (other things being equal) directly
related to the total number of symbols from which one has to choose. If there
is only one symbol available, say the letter A, “messages” will be simply
strings of A’s and the successive symbols can carry no information at all, in
the sense defined above. If there are two symbols, say the digits 0 and 1,
one will have a 50 per cent chance of guessing each successive symbol in a
message composed of a random sequence of such symbols. Each successive
symbol will in this case carry one bit of information, according to a formula
that says the number of bits is the logarithm, to the base 2, of the number of
equally-likely alternative symbols. If there are 32 symbols from which one
can make random sequences, say the 26 letters of the alphabet plus a space
and 5 punctuation marks, this same formula says the number of bits of
information carried by each symbol will be 5.
The other important idea in information theory, for our present purposes,
is that the amount of information in a symbol is decreased if the symbols are
not random but in some degree dependent on (predictable from) each other.
We have already seen how the symbol N at the end of the ‘“message”’
GENERATION carried little if any information, because one could predict
it from the previous nine symbols. This idea is important in studying the

Garner. Uncertainty and structure as psychological concepts. New York:

12 See W.R.
Wiley, 1962. Aspects

of Language
Behavior

59
 statistical structure of language, because it allows us to measure the amount
of
of redundancy of samples of language text. Redundancy is the property
symbols from the context. High-
texts that allows us to predict missing
in-
redundancy texts tend to be repetitive and to contain relatively little
per symbol. A text with zero redundan cy would be one in which
formation
maxi-
no symbol is predictable from any other, and thus there would be a
A list of words selected at
mum amount of informativeness per symbol.
random from the dictionary for some purpose would constitut e a text with
low redundancy.
Psychologists interested in speech production, speech perception, and
various problems of verbal learning have found much use for information
measurement in their experiments because it provides a precise way of
quantifying the amount of material perceived or learned; further, when
information is quantified in this way, the relationships discovered are simple
and direct. For example, Miller, Heise, and Lichten showed that an individ-
ual’s ability to identify a word through a considerable amount of noise is
inversely related to the logarithm of the number of different words from which
he has to choose (see Figure 10). It is not a great jump from this result to the
further result that the intelligibility of a word is, other things equal, dependent
on the logarithm of its frequency in large samples of verbal material, and it
turns out that the amount of “information” (in the technical sense) that one
can gain by listening to a single word through noise is approximately constant
for words of different frequencies. That is, even though a rare word is iden-
tified through noise with difficulty, the average amount of information
transmitted by that word, in view of its low frequency of occurrence, is about
equal to that transmitted by a common word which can be identified much
more easily. These results suggest that the psychological mechanisms whereby
stimuli are perceived and learned have built-in limits for handling the in-
formational aspects of stimulation.
This interpretation is also suggested by experiments using samples of
language symbols in which the amount of redundancy is artificially controlled,
thus controlling the average amount of information in a sample. Samples
with various “orders of approximation” to English can be constructed (see
Table 3). A zero-order approximation is constructed by stringing out words
chosen at random from a dictionary without consideration of their frequency
in the language. A first-order approximation is like the zero-order except
that the words are chosen in proportion to their frequencies in the language.
Second, third, and higher orders of approximation are constructed in a some-
what artificial manner by depending on people’s language responses to small
strings of symbols. For example, to form a second-order approximation one
starts with a word chosen at random, A, and asks someone to use it in a
sentence; whatever word, B, follows the given word, A, is used in the string.
Then word B is given to another person to be used in a sentence, and what-
ever word, C, follows B is used as the next word in the approximation. This
process is repeated until one has a sufficiently long string of symbols, A, B,
C,..... A third-order approximation is constructed by using a set of two
words, AB, as stimuli for a sentence to yield a third word C, then using words
BC as stimuli to yield D, and so on.
Aspects
When these artificially contrived language samples are used in experiments,
of Language
Behavior
56
 oe 2 oe =

ee -
ee
é SZ See
rs
: OC Oe oo. ie

8
Stes

aS oS

oC. UE
Ȏ\
—oe Ss;
_
US

SON
27)

Sc
i en
SASS eS
SS
— CK Oe

Figure 10. Percentage of words correctly identified as a function of the

size of the vocabulary from which they have to be selected, for various signal-
to-noise ratios. At 0 db, the signal strength is equal to the noise strength. At
+12 db, the signal is 12 decibels (loudness units) above the noise level,
while at —12 db, it is 12 decibels below the noise level. (Drawn from data of
G. A. Miller, G. A. Heise, and W. Lichten. J. exp. Psychol., 1951, 41, 329-
335.) ;

it is found that the higher the order of approximation, the easier it is for
people to perform various tasks involving the passages—to learn and re-
member them, to reconstruct them by guessing, to type them, or to read them
aloud. One interpretation of these results is that the response to these complex
stimuli depends on the absolute amount of information they contain, and
that an individual can readily handle only so much information in a given
amount of time. A slightly different interpretation is this: The higher the
order of approximation to actual English structure, the more the sample will
approximate the particular phonemic, morphemic, and syntactical structures
the individual has learned; and it is these structures which correspond to the Aspects

of Language
Behavior

57
 3 Bete o
TABLE 2

Strings of Words at Various Orders

of Approximation to Normal English Text

Zero order (words chosen at random from a dictionary) :

alas pillory nautical.
Combat callous irritability migrates depraved temporal prolix
probability of occurrence
First order (words chosen independently, but with their
of normal English text):
proportional to their frequency of occurrence in large samples
of are its go studies the our
Day to is for they have proposed I the it materials
of the following not over situation if the greater.
ty with which each word
Second order (words chosen in such a way that the probabili
is intended to be proportio nal to the frequency with which it follows the one
appears
preceding word in large samples of normal English text):
so what is dead weight
Goes down here is not large feet are the happy days and
that many were constructed the channel was.
y with which each word
Fourth order (words chosen in such a way that the probabilit
which it follows the three
appears is intended to be proportional to the frequency with
preceding words in large samples of normal English text) :
and depart for home.
We are going to see him is not correct to chuckle loudly

ll, 1951, pp. 84-85.

As given by G. A. Miller. Language and communication. McGraw-Hi

response habits that function in comprehending and responding to language

materials.
We must carefully avoid supposing that frequency counts and information-
theory measures throw anything but a faintly reflected light on language
structure as it was described in Chapter 2. A language system as such has no
statistical structure, it being merely a set of regularities of symbolic behavior
in a speech community. Frequency counts and information theory are useful
in characterizing language samples produced by or experienced by human
beings, but they disclose relatively little about the nature of the language
systems in which these samples are coded.

Any thoughtful person who listens to a continuous flow of speech in a

language he does not understand will appreciate the astonishing feat per-
formed by the listener in perceiving and interpreting that speech. The listener
who knows the language does not hear it as a seemingly random concatenation
of sounds. Instead, from the sound waves, he is able to select and respond to
particular units of various sizes.
There has been a great deal of investigation, for obvious practical reasons,
concerning what can be done to a speech wave and still leave the speech
message it bears reasonably intelligible to a person who can understand the
Reais
of Language
Behavior
58
 original message. For example, how much noise and interference can safely
be added to the speech wave? In what way can the speech wave be distorted?
The detailed results of these studies cannot be presented here, but one im-
portant conclusion from them can: namely, that the speech wave is highly
redundant. Not every part of the speech wave is a completely new, independ-
ent bit of information; instead, some parts are repetitions of signals pre-
sented previously, or at least they are predictable from the signals by any-
body who has learned to interpret those messages. Thus, if noise or distortion
is introduced, the chances are that enough information will come through to
enable the hearer to infer the whole message. In fact, the intelligibility of
speech is amazingly resistant to distortions of a speech wave.
A speech wave would have a certain amount of redundancy even if the
message contained nothing more than nonsense syllables. But it will contain
much more redundancy if normal speech samples are used. It has been esti-
mated that the redundancy of normal speech samples is probably better than
50 per cent.
There are many sources of redundancy. Among them are these: phonemes
vary in frequency (therefore, a missing phoneme is more likely to be a common
one than a rare one) ; phonemes cannot be strung out randomly (for example,
a string of consonant phonemes can be only so long before it has to be followed
by a vowel); the inventory of morphemes in our language does not include
all the possible combinations of phonemes; morphemes themselves vary
in frequency; morphemes cannot be strung out randomly—they have to be
put in grammatical sentences; and finally, there are practical limitations on
the things we are likely to say—a sentence like ‘“‘The sieve larch was then
dissimilar in neither miracle” is highly unlikely though grammatical.
The more redundancy in a message, the greater the chance that it will be
understood despite noise, electronic distortion of the signal, or low speech
volume or poor intelligibility of the speaker. But redundancy is helpful only
to the extent that the hearer is able to take advantage of it. If the listener
doesn’t know the structure of the language, or has no feeling for what is
likely to be said, redundancy is not very helpful. The speech in an airport
control tower is statistically highly redundant, but in noise it is practically
unintelligible to the uninitiated.
Even without noise, speech signals have to be sufficiently discriminable
to allow the hearer to recognize the units of the message. For normal speakers
and hearers, phonemes are rarely confused, even though they may be very
close in sheer acoustic terms. It will not do to appeal solely to redundancy or
context to explain this lack of confusion; the phonemes themselves are
distinguishable even in minimal pairs. In fact, evidence suggests that for the
speakers of a given language, phonemes have acquired distinctiveness; that
is, the speakers have learned to discriminate classes of sounds, even though
they may be indiscriminable to speakers of other languages.!? Further, it
appears that for certain kinds of phonemic distinctions, hearers’ sensitivity
is heightened in those regions along acoustic dimensions that are most critical
in making those distinctions. For example, subjects clearly discriminate the

13 A. M. Liberman, K. S. Harris, H. S. Hoffman, and B. C. Griffith. J. exp. Psychol.,

1957, 54, 358-368.
Aspects

of Language
Behavior

59
 e the differences among
small changes along the dimensions that produc
change s of equal magnitude, elsewhere on
/b, d, g/, but fail to discriminate
mic difference.
those dimensions, that do not happen to produce a phone
the fact that the number
There is a possible psychological significance in
about 12 to 65; with 33
of segmental phonemes in languages varies from
median . A langua ge cannot have too
phonemes, English stands at about the
a very small number , the morph emes of the
few phonemes, because with
, to be distinc tive, and
language would have to be longer, on the average
a langua ge cannot
they might not be readily remembered; on the other hand,
lves might be less
have too many phonemes, because the phonemes themse
easily discriminable.
165 words per
The rate of normal uninterrupted English speech is about
average word length may vary, perhaps syllable s per minute
minute; because
syllable s per minute. If
would be a better measure; that figure is about 265
e adequa tely—t hat is, if he knows
the hearer of such speech knows the languag
used—h e would probabl y report that he can
the grammar and the lexicon
at this rate. By use of a special device for speedin g up
“anderstand” speech
would happen if
tape-recorded speech without changing its pitch (the latter
has been found that speech
one simply increased the speed of the tape) it
(At 450
is still intelligible when presented at up to 2.5 times normal rates.
rate is still below that at which printed words can be
words per minute, this
ood.) Actuall y, compreh ensibil ity is a very difficult thing to
read and underst
listener ’s subject ive
measure objectively; one has to depend largely on the
ated rate.
report as to whether he can “follow” speech presented at an acceler
speech would also depend both on the char-
The comprehensibility of speeded
itself— the complex ity of its gramma r, the difficulty
acteristics of the speech
vocabul ary used, and the abstrac tness or technica lity of the conten t—
of the
it seems clear that in
and on the competence of the listener. Nevertheless,
the use of speeded speech to transmi t
many cases it would be possible through
information much more efficien tly than usual.
How does a listener understand speech, whatever its rate may be? We
have likened a sentence to a “program” for a computing machine by calling
it a computer program for a thinking machine, that is, the human central
nervous system. Unlike a true computer program, where the smallest error
may lead to a complete stoppage, sentences have enough redundancy so that
even if they contain “errors” like “slips of the tongue,” they will normally
still be understood.
Whereas the speaker formulates his utterances first by selecting major sen-
tence-types and transformations and then by filling them in with appropriate
forms, the listener must apply these procedures in reverse order, as it were.
That is, all that is available to him is the sequence of forms; from this he must
“construe” the sentence in some particular grammatical pattern. Occasionally,
the heard sentence will be constructionally ambiguous and permit two or
more interpretations, and he may guess wrong even in the presence of ade-
quate context. For example, the exploitation of the workers might be con-
strued as a transformation of (they) exploit workers when it was intended
as a transformation of workers exploit. But the very fact that the listener
chooses an interpretation demonstrates the potency of grammatical structure.
Aspects Generally, the over-all pattern of a sentence in English can be detected rather
of Language
Behavior

60
early in the sentence. The form-classes of the first few words in a sentence
usually indicate, for example, whether it is to be an existence-assertion or a
predication, and whether it is in a null, a question, an imperative, or some
other kind of transformation.
We may regard each decoded grammatical construction as a discriminative
stimulus for some response in the listener. We must assume that these re-
sponses are covert, not immediately observable by any normal means. For
example, a null transformation is a discriminative stimulus for an orienting
response to note the information contained in the utterance; a question trans-
formation is a discriminative stimulus to search for the information requested
by the question-signal (an auxiliary verb, in the case of a yes-no question, or
a “wh-word” in the case of a multiple-response question), and so forth.
O. Hobart Mowrer has proposed that a sentence in the predicate form (like
Tom is a thief) is an arrangement for conditioning the meaning response
produced by the predicate is a thief to the meaning response to the subject
Tom. That is to say, the meaning reaction to is a thief now gets connected to
the meaning reaction produced by Tom, and the hearer’s subsequent behavior
in the presence of Tom may bear this out. Mowrer’s proposal is interesting
and probably on the right track; it is very limited, however, for a particular
utterance can hardly be a “conditioning arrangement” unless one ignores evi-
dence that conditioning rarely occurs in one trial; more importantly, it fails
to do justice to the complexities of sentence construction. How, for example,
shall we deal with a sentence like Tom is not a thief? Could the word not
cancel the conditioning arrangement so facilely?
According to the evidence as we see it, a sentence is a series of discrimina-
tive stimuli, learned by the speaker of a language, which in effect “program”
the mediating responses of the hearer in such a way that certain constructions
are put on the sentence and corresponding mediating responses are evoked in
the hearer. If A hears B say, ‘Tom is a thief,” this sentence is a series of
discriminative stimuli for mediating responses that represent:

1. a predication
2. a declaration (it being in the null transformation) on the part of B; that is,
A learns that B entertains a belief about Tom
3. Tom (the context supplies information as to which Tom is meant)
4, is: present tense, current predicative
5. a thief (one of a class of thieves)

or some such list.

In countries with formal systems of education, the age at which children

learn to read is somewhere between five and seven, fixed more by tradition
than by any rational considerations of the best age to start to read. A few
children manage to learn to read before they go to school—often with minimal
help from others; a fairly substantial number of other children are delayed in
their progress; and, of course, there are those who do not have the opportunity
to learn to read until they are adults, if then. But psychological considera- Aspects

of Language
Behavior

61
 experiments, suggest that
tions, and evidence from a number of educational
have gained mastery of essential
children can be taught to read as soon as they
phonol ogy, its most commo n grammatical
features of the spoken language, its
vocabu lary. (Wheth er such early reading is de-
constructions, and a-basic
debated here.)
sirable from a developmental point of view cannot be
really, do we mean by “readi ng”? A written text is a representation
What,
phonetic transcriptions
of a possible spoken utterance. Except for special
s give an imperf ect representation of the
devised by linguists, writing system
in a spoken utteran ce. To be sure, some writing
actual sounds produced
those of Spanis h or Finnish , convey fairly accurately the seg-
systems, like
egment al phonemes
mental phonemes, but no writing system represents supras
has a conside rable master y of the
at all adequately. Only an individual who
how a written text might reason ably be
spoken language is able to infer
limited cues supplie d by the text and its punctu ation.
spoken, from the
is another case of the use of context to supply missing data in a
(This
ructing
message.) We can define reading, ultimately, as the activity of reconst
spoken messag e from a printed text,
(overtly or covertly) a reasonable
es to the reconst ructed messag e that would
and making meaning respons
our dis-
parallel those that would be made to the spoken message. (Recall
cussion of how it is that we comprehend a spoken message.)
g that
The learning of reading, defined in this way, is obviously somethin
may take a considerable amount of time. We take it for granted that normally
a learner has already attained some control of the spoken language before
the
he tries to learn to read, although there are many situations throughout
particula r difficulty be-
world where the learning of reading is attended with
cause the child is asked to read a language not his own before he has suffi-
ciently mastered the spoken form of that language.
One major goal in learning to read is to learn to respond to written texts
in accordance with the writing system, that is, in accordance with any regular
or partially regular correspondences between spoken sounds and written
symbols that may exist in this system. The standard orthographies or con-
ventional writing systems associated with various languages (some languages
have two or three such systems) vary in the simplicity and regularity of these
correspondences; most of them use an alphabetic principle whereby phonemes
are directly represented in written symbols. In the case of some languages,
like Finnish, Spanish, and Turkish, these correspondences are simple and
highly regular; since they can be learned quite readily, this aspect of the task
of learning to read can be made relatively easy. At the opposite extreme is
the orthography of Japanese, where three parallel systems of orthography,
only two of which are even partially phonemic, must be learned.
The standard orthography of the English language presents a peculiar
problem. It is incorrect to say that English is “ynphonetic.” This adjective
cannot be applied to a language, in any case, because languages necessarily
have phonetic aspects; nor can it properly be applied to a writing system,
because writing systems are dependent more on the phonemic than the
phonetic properties of the languages they represent. The grapheme-phoneme
correspondences of English orthography are somewhat irregular, although not
as irregular as one might think. It is possible to formulate a set of rules for
Aspects
“translating” a printed text into segmental phonemes so that more than 95
of Language
Behavior
62
per cent of the phonemes would be correct. The rules would be fairly com-
plicated; one of them might be, for example, that the letter C followed by E
or I is to be translated by the phoneme /s/, unless E or I is followed by a
further vowel letter, in which case it will correspond to /S/.
A great deal of argument and a limited amount of study and experimenta-
tion have been devoted to the question of how the reading of English can
best be taught in view of the facts about its orthography. It is recognized that
a child can be taught to respond correctly to whole words, with no reference
to the sound correspondences of their letters; the argument is over how early
in reading instruction the child can be taught to take advantage of such cor-
respondences. One widespread opinion (actually a misinterpretation of certain
research studies) is that a child cannot be so taught until he has a mental age
of about seven, but this is demonstrably false, for many children do learn to
read in terms of grapheme-phoneme correspondences well before they attain
this mental age.
Apparently, the way in which these grapheme-phoneme correspondences
is introduced is important. The learner needs to be presented with systematic
sets of instances from which he can readily learn the discriminative function
of those letters or combinations of letters which are fairly sure guides to
pronunciation. For example, at some point the child needs to see that there is
a regularity in such pairs as at-ate, fat-fate, hat-hate, mat-mate, not by learn-
ing rules or verbalizing them, but by achieving consistent discriminative re-
sponses. At the same time he needs to learn that he must often adopt a some-
what experimental trial-and-error approach to the letter combinations in
words like come, dome, home, some. He must also be taught to take account
of context in such experimentation, or educated guessing.
Concern with grapheme-phoneme correspondences, or “phonics” as it is
often called, should not distract us from the necessity for the reader to attain
a rapid visual perception of printed words as wholes. It is probably wise to
give some training and practice in this in the earliest stages of reading, by
means of a “sight vocabulary.” (The sight vocabulary, however, can be
chosen so as to take optimum advantage of the regularities that exist in
English orthography while catering to the learner’s need to know common
“irregular” words like the, to, what.) Experiments in visual pattern percep-
tion suggest that recognition of words as visual patterns can be accelerated
by (1) drawing attention to the shapes of the parts of these patterns, that is,
the letters, (2) giving practice in writing or tracing these parts, and (3)
building up the frequency of exposure to these patterns. By means of the
tachistoscope (a rapid-exposure device), it can be shown that mature readers
recognize common words in not much more time than they need to recognize
single letters, around one-tenth of a second. Other things being equal, recog-
nition time appears to be inversely related to the frequency with which a
word has been seen by a reader in his past reading (see Figure 11). These
results suggest that each word in a basic reading vocabulary should be
presented many times over. At the same time, vocabulary control need not go
so far as to exclude words that the beginning reader can interpret on the basis
of the phonic habits he has built up.
Efficient reading entails well-coordinated eye movements of the saccadic
variety (the ones by which the eyes jump from one fixation to another). At Aspects

of Language
Behavior

63
 Figure 11. The relation-
ship between word fre-
quency and speed of
tachistoscopic perception.
Each set of points shows
the average amount of
exposure time (millisec-
onds ) required by 20 sub-
jects to recognize words
of the indicated frequen-
cies with a given per-
centage of accuracy.
(Adapted from data of
D. Howes. In J. Money
(ed.). Reading disability.
Baltimore: Johns Hobp-
kins Press, 1962.)

one time it was thought that reading could be improved by training these eye
movements themselves, but it now appears that poor eye movements are an
effect, not a cause. Mechanical devices for pacing reading performance are
valuable mainly for arousing effort and attention on the part of the learner;
they are of little use if the learner has not acquired the basic perceptual
responses which are prerequisite to rapid reading.
In the mature reader, reading speeds are a function of the reader’s training,
his purpose in reading (light skimming and concentrated study are at op-
posite poles of a continuum), and the difficulty of the material. There are
various ways to measure the difficulty, or “readability,” of prose. One is to
use one of the several formulas that have been developed, dependent usually
on crude measures of the grammatical complexity and vocabulary level of the
material. Another is the “cloze” procedure developed by Wilson Taylor,
whereby the readability of a passage is evaluated by deleting, say, every tenth
word in a text and seeing how successful, on the average, the members of a
panel of readers are in guessing the missing words from the context. As com-
pared with formulas, the latter procedure seems more successful, for example,
in detecting that the prose of Gertrude Stein is fairly difficult despite her use
Aspects
of common words and sentences of moderate length.
of Language
Behavior

64
 Because of the many factors in reading speed, definite standards have little
meaning unless one specifies the purpose of reading, the difficulty of the
material to be read, and the method of measuring comprehension. There is
probably a certain limit to the rate at which verbal information in any
form can be taken in by a human being. If we take as a standard the degree
of comprehension that can be attained when material is presented orally at
a normal rate (say 165 words, or 265 syllables, per minute), the rate at
which an individual can read material of the same difficulty with the same
degree of comprehension will probably not be over three or four times the
speech rate (that is, not over 500-700 words per minute). (Strangely, there
is no research evidence that has given a reliable and sensible answer to this
problem; the few research reports available confuse comprehension with
memory for details.) When reading rates much higher than this are reported,
they are probably attained with a lower standard of comprehension or with
greater than normal use of redundancy and context. The claim that reading
speed and comprehension are positively correlated is true only up to a point,
and only within groups of people whose average reading ability is well below
reasonable standards. In the case of a poor reader, reading speed can hardly
be improved without improving comprehension as well. Comprehension, in
turn, is improved by teaching the reader not only to recognize words faster,
but also to respond more swiftly to the grammatical signals in a piece of prose
and to attain a wider and richer vocabulary.

Aspects

of Language
Behavior

65
Individual Differences
in Language Behavior The members of a speech

community obviously differ greatly in their ability to use

and understand language. Our most precise information

about these differences comes from the study of per-

formances in the more or less controlled situations we

call tests. The so-called intelligence tests probably enjoy

the widest general use, and a major class of these are

verbal intelligence tests, which measure an individual’s

understanding of words and his ability to manipulate and

apply abstract verbal concepts. There are, in addition,

many other kinds of tests involving language habits: free

association tests, sentence completion tests, articulation

66
tests, reading readiness tests, foreign language aptitude tests, and so forth.
The old concept that “intelligence” is a general trait or characteristic of
the individual has given way to the notion that the classification of human
abilities into various “traits” is a somewhat arbitrary operation, justifiable
only when it can be demonstrated that the behaviors classified under a given
trait tend to occur together more often than would be expected by chance.
Even after we have demonstrated that they do, we still have the problem
of explaining this result. The behaviors may have been learned together,
for some fortuitous reason, or there may be some common explanation for
their co-occurrence, such as a single genetic trait that makes them both
possible.

In studying differences in language ability, therefore, we should start with

the widest possible sampling of language performances and see whether
there are any systematic ways in which the members of a speech community
vary with respect to these performances. We must then see what explanations
for these systematic differences we can find.
Information about the basic dimensions on which people vary has been
provided largely by the statistical technique known as factor analysis. This
is a procedure for studying the intercorrelations of measures that are ap-
plied to representative samples of people or things to identify the basic ways
in which these people or things differ. Each such way can be called a dimen-
sion or factor. Combing the results of a number of studies, we find the
following kinds of more or less independent dimensions of ability in the
domain of language behavior.

1. Verbal knowledge. One of the most pervasive ways in which people

differ is in their knowledge of the vocabulary and structure of the English
language. This dimension of ability is measured in greater or lesser degree
by the following types of tests: vocabulary tests (in fact any test in which
attainment depends to a considerable extent on knowledge of the meanings of
words, particularly the rarer and more abstract words) ; tests of knowledge
of “correct” (that is, socially approved) English grammar and usage; spelling
tests (to the extent that they involve more difficult vocabulary) ; and certain
tests of sentence completion that are scored for the extent to which people’s
responses agree with those of the majority. Verbal knowledge is highly cor-
related with the extent, variety, and richness of an individual’s concepts, at
least to the extent that these concepts are symbolized by words.
2. Abstract reasoning abilities. Although the factor analysis of reasoning
abilities has given results that are somewhat ambiguous and variously in-
terpreted, it seems clear that people differ markedly in their ability to perform
reasoning tasks, whether concrete or abstract. Some results suggest that
inductive reasoning ability is somewhat independent of deductive ability, and
that both in turn are independent of serial reasoning ability—the ability to
think through a chain of inferential reasoning steps. Individual
3. Ideational fluency. This factor represents an individual’s facility in Differences
in Language
Behavior

67
 calling up as many ideas as possible about a given topic or theme; the number
of different ideas, rather than their quality, is at issue. A typical test of this
ability is the one in which subjects are asked to write down as fast as possible
the words for as many “round things” as they can think of in a brief time.
4. Word fluency. This ability is something like ideational fluency, but
concerns the ability to think of words with certain formal characteristics—
such as those beginning with certain letters when spelled. It seems to depend
partly on the individual’s knowledge of phoneme-grapheme correspondences,
and also possibly on what we may call “phonemic coding ability,” that is, the
ability to encode what is heard into a form that can be put in a memory
storage and used at a later time. Certain tests of word fluency and of
phonemic coding ability have been very successful in predicting facility in
learning a foreign language.
5. Fluency of expression. This factor represents the individual’s facility
in formulating ideas—that is, given an idea, his facility in putting it in gram-
matically acceptable words and constructions. It can be tested in various
ways; one is to ask the individual to think up a large number of expressions
for praising the virtues of a political candidate.
6. Grammatical sensitivity. This is the ability to recognize the functions of
form-classes and constructions and to perform tasks requiring the ability to
perceive these functions. It is an important factor in aptitude for learning a
foreign language.
7. Naming facility. This factor represents the ability to respond rapidly
with the names of things, shapes, colors, and so on, when a series of such
items is presented in rapid succession.
8. Oral speech ability. This represents ability to speak effectively and
coherently in a more or less formal speaking situation.
9. Articulation ability. This represents individual differences in the speed
with which speech sounds or utterances are, or can be, articulated.

This list of language abilities is by no means exhaustive, and remains to

be clarified by further research.
Many of these factors have great practical significance in education and
personnel selection. Because verbal knowledge and reasoning abilities are so
advantageous in scholastic work, typical verbal intelligence tests, such as the
Scholastic Aptitude Test of the College Entrance Examination Board, tend
to stress the measurement of these abilities, along with quantitative reasoning
abilities. Several of the factors are of unique importance in the learning of
foreign languages, as indicated above, and are measured by the Modern Lan-
guage Aptitude Test. But some of the other kinds of ability we have listed
have not as yet shown any important relation to any kind of practical work.
Contrary to expectation, for example, the fluency of expression factor seems
to be unrelated to skill in English composition.
What about the causation or etiology of these individual differences in
learning ability? To the extent that some of these abilities are included in
what we usually call “intelligence,” this question about causation might
require us to discuss what it is that spreads human beings out on the scale
Individual
from genius to idiocy, but there are more appropriate places than this for such
Differences a discussion. Certainly both genetic and environmental influences are at work.
in Language
Behavior

68
 There has been relatively little research on the possible hereditary etiology
of those aspects of language ability that are largely independent of intelli-
gence. We have some evidence of a genetic basis for the special difficulties
that a certain small percentage of children, particularly males, have with
beginning reading and spelling, and later, with learning of foreign languages,
even though the children may be otherwise quite intelligent. This syndrome
is called specific language disability and is characterized by difficulty in encod-
ing and storing auditory (phonetic) information and tying it with visual
symbols. It may have a close relation to the word fluency factor mentioned
earlier.
For the most part, we can probably attribute individual differences in
verbal abilities to environmental influences—differential opportunities to learn
language behavior, and differential conditions of motivational arousal and
reinforcement. For example, individual differences in the verbal knowledge
factor are clearly related to socio-economic status, amount of schooling,
parents’ occupations, and other variables that indirectly measure opportunity
and motivation to learn language. This seems plausible since the full de-
velopment of native language skills, even though the groundwork is laid
in early childhood, requires the whole period of childhood and adolescence.
In fact, the acquisition of vocabulary is a process that goes on to some degree
even in old age. Because vocabulary knowledge is little disturbed even in
senility, the relative degree to which certain other abilities are lost is claimed
to be an indication of intellectual deterioration.

In the early years of life, some children are slow in learning to talk. Such
retardation can mean any one of a number of things. Sometimes the retarda-
tion is associated with a pattern of growth in which motor development runs
ahead of mental development. Sometimes it is psychogenic—that is, the
result of emotional problems or unfortunate learning sequences. But it may
also mean that the child is mentally retarded due to some organic or con-
stitutional defect in the nervous system. Certainly delayed or arrested lan-
guage development is one of the most universal characteristics of mental
retardation when properly diagnosed. Idiots, with IQ’s below 20, rarely attain
any language use beyond a small number of isolated words and the compre-
hension of simple commands. In other grades of retardation, language develop-
ment is delayed in proportion to the extent of mental deficiency. Even the
babbling stage is delayed. One study reported that in children with an IQ
of 51 to 70, babbling occurred, on the average, at 20.8 months, word use at
34.5 months, and sentence use at 89.4 months. The corresponding figures
for a normal child would be, approximately, 4 months, 12 months, and 20
months. Along with these late starts one finds, of course, very low ceilings
of development; a child with an IQ of 50 will on the average rise only to a
mental age of 7 or 8; language usage of such a child is restricted to relatively
simple sentences and a vocabulary of a few hundred words. There is, however,
read
a class of “educable mentally retarded” children who can be taught to Individual
between grades three and seven, IQ’s for such
and write up to somewhere Differences
in Language
Behavior

69
 children range from 50 to 75. But all these children have considerable trouble
with language, with both its motor and its conceptual aspects. Speech defects
are especially common, and difficult to remedy.
The study of the slow-motion language development of mental defectives
might throw much light on how grammar develops and how various gram-
matical phenomena rank themselves in conceptual difficulty.

Aphasia (etymologically, “lack of speech”) is a term that is ordinarily

applied to a condition in which a person who has already acquired language
competence suddenly and dramatically loses some or all of this competence
because of brain damage. The appropriateness of this term in the case of what
is sometimes called “childhood aphasia” is somewhat questionable, although
there is indeed a condition in children, distinguishable from mental retarda-
tion, where delay in language development is associated with some specific
kind of brain injury or maldevelopment resulting from prenatal condition
or birth trauma. Here, we shall focus attention on aphasia in adults.
The language disturbances found in aphasia are exceedingly diverse, and
of differing degrees of severity. Some patients lose all capability of speaking
and understanding, but most have residual capacity in certain performances.
Early classifications of aphasia were not based on a sufficient number of cases
or on sufficiently comprehensive clinical examinations. At present, it is widely
accepted that there are no distinct “types” of aphasia, although it is recog-
nized that cases differ not only in severity but also in kinds of loss. In some
cases the loss seems to lie predominantly in speech “reception” (recognition
and understanding), whereas in others, defect is manifested chiefly in a re-
duced ability to express thoughts. Among the latter, some are chiefly handi-
capped by an inability to find particular words for concepts (anomia), and
others by incapacity to form coherent sentences (syntactical aphasia).
Aphasic disorders may also be accompanied by deficits in visual and
spatial perception—for example, inability to recognize common objects and
symbols, or to recognize the equivalence of stimuli. Poor control of the
speech musculature, even inability to swallow, may occur in some cases. But
a central characteristic of all aphasias is loss of ability to conceptualize and
to manipulate concepts by language symbols, or as Kurt Goldstein puts it,
a loss of ability to use abstractions and to generalize. The nature of this loss
can be better understood when it is recognized that the aphasic loses not only
the ability to express himself by language—he also loses the ability to express
himself by gestures or other symbolic movements, or to understand such
gestures when others try to use them.in communicating with him.
Some psychologists have been tempted to use these diverse patterns of
aphasia to throw light on the possible neural mechanisms and organization
underlying language behavior. Charles Osgood has postulated a model of lan-
guage behavior as a multistage process that involves three levels: a projection
level at which stimuli are “projected” in appropriate areas of the brain; an
integrative level in which these stimuli are recognized and integrated with
Individual
Differences
other stimuli in appropriate grammatical constructions; and a representational
in Language
Behavior
70
level at which the significances of language signs are recognized and manipu-
lated. The model also assumes three processes in the handling of information:
decoding, association, and encoding, each of which can operate at each of the
three levels of organization just mentioned. Aphasic symptoms, according to
Osgood, represent losses of function at one or more of these levels and in
one or more of these processes, and he has had fair success in making the symp-
toms of individual cases of aphasia fit this model. Some of the results from the
work of neurologists and brain surgeons support Osgood’s general point of
view, but other investigators still incline to the belief that the phenomena of
aphasia are too complex to be dealt with in this way. For example, one school
of thought, represented by Davis Howes, holds that most of the phenomena
of aphasia can be explained as due to a general loss of the strength of verbal
habits. A closely related opinion is that of Hildred Schuell and James Jenkins,
who hold that aphasia is a unitary phenomenon characterized by general loss
of language competence on both the receptive and the productive side; accom-
paniments of aphasia such as disturbances in visual perception and poor
control of speech musculature are not essentially aphasia, for they can occur
without aphasia.
Despite very active research on the subject, our notions of the nature of
aphasia are still quite indefinite. Fortunately, lack of agreement on the theory
of aphasia has not impeded the development of therapeutic procedures.
Therapy usually requires considerable time, but can be successful in many
cases. Apparently relearning often involves the development of new neural
pathways for language functions.

SANSUARE th
LANGUAGE IN

In practically every form of neurosis or psychosis, there is some alteration

or disturbance of speech and language behavior. From the mildest type of
neurosis, in which the patient may express his anxieties by a slightly abnormal
repetitiousness of speech, or by noticeable hesitations or speech blocks, to severe
schizophrenias in which speech is fantastically garbled in what is often called
“word salad,’ we have evidence that these abnormal mental states can in-
fluence both the form and the content of speech. On the other hand, all these
behaviors represent extreme forms of phenomena that occur in normal in-
dividuals—slips of the tongue, overrepetition of certain words and phrases,
and even garbled speech. They apparently represent lapses in attention and
in the central control of speech processes, as well as the operation of mo-
tivational dynamics. Freud presented a classic account of how slips of the
tongue may often be interpreted as evidences for repressed motives. The “word
salad” speech of the schizophrenic can be regarded as the result of three
tendencies: (1) the disorganization of syntactical behavior, that is, an
extreme form of the kind of disorganization that occurs when we make false
starts and change the grammatical construction of a sentence; (2) paranomia,
the tendency to replace the ordinary names for things with substitute names—
neologisms and other highly personal symbols; and (3) a heightened lability
of verbal association. The first tendency is understandable enough when Individual
thought itself is disorganized; the second may be the result of fearsome or Differences
in Language
Behavior
71
 things (analogous to
aversive associations evoked by the ordinary names of
ct a personal idio-
taboo words) or an unconscious motivation to constru
apart from others. The third
syncratic language system which sets the patient
eeling associa tion that can be pro-
tendency accentuates the kind of free-wh
drugs, or even in certain states of religious
duced by alcohol and certain
normal inhibit ions against absurd or trivial intraverbal
ecstasy, when the
e form and
connections are released. Despite such deviations in languag
lose the underly ing phonol-
content, there is little evidence that patients ever
Speech disturb ances can be looked upon
ogy and grammar of their language.
as a special kind of transfo rmation of languag e habits.

its
Stuttering (also called stammering) is a disorder of speech that makes
ce in some small proporti on of children (estimate s range from 1 to 5
appearan
as
per cent), usually about the age of two to four years. It has been defined
“a disturbance of rhythm and fluency of speech by an intermitt ent blocking,
a convulsive repetition, or prolongation of sounds, syllables, words, phrases,
in
or posture of speech organs.” Incidence is somewhat higher in boys than
girls.
In view of all the research that has been devoted to the matter, it is
surprising that there is no generally accepted, single explanation for stuttering.
It is evidently a disorder of the motor control of speech. Theories of stuttering
focus either on heredity or environment. Environmental theories stress either
organic (neurological or chemical) factors or functional (psychogenic) etiol-
ogy. Those who prefer functionalist theories assert that stuttering arises
when circumstances make a young child anxious about his speech—that is,
fearful that he will not be able to speak properly. It is claimed that the very
existence of the concept of stuttering channels the behavior of parents and
teachers into unwittingly fostering the development of stuttering out of the
normal speech errors that nearly every child makes. This theory has been
neither confirmed nor disproved by the research results available thus far; it
seems prudent to recommend, however, that in raising children parents should
avoid creating anxiety concerning minor speech errors. But it is also possible
that stuttering is an accompaniment of anxieties and conflicts that are deeper
than a mere concern with speech behavior. In many cases, successful cures of
stuttering in adolescents and adults have been effected by protracted psy-
chotherapy in which the patient is enabled to rid himself of intense conflicts,
often concerned with matters of self-identity, sex-typing, and psychological
maturity.
Stuttering can also be analyzed and treated simply as a motor response,
associated with the effort of uttering a syllable. There has been some success,
for example, in treating it as an operant response and putting it under the
control of rewarding or aversive stimulation.! The stimulation is provided by
an unpleasant sound presented by earphones; a stutter can be rewarded by
turning off this sound momentarily as soon as the stutter appears, or it can be
Individual
1B. Flanagan, I. Goldiamond, and N. Azrin. J. exper. anal. Behav., 1958, 1, 173-177.
Differences
in Language
Behavior
72
punished by turning on this sound momentarily. As might be expected, reward
increases the incidence of stuttering beyond the individual’s normal rate, and
punishment decreases it. However, the effects are restricted to the experi-
mental situation, and are not startling. A much more dramatic effect has
been obtained by C. Cherry and B. M. Sayers in England.2 They claim that
stuttering can often be totally inhibited, in an experimental situation, by what
they call a “shadowing technique” in which the stutterer is asked to say aloud,
as close to simultaneously as possible, what he hears another person say or
read aloud. A variant of the procedure is to have the stutterer read a passage
aloud with another person; the stammering is reduced or absent even if the
other person starts reading something else, or saying gibberish. Cherry and
Sayers interpret these results as indicating that stammering can be reduced
by distracting the stutterer’s attention from his own voice. They also have
shown that stammering can usually be totally inhibited by using a white noise
(a noise containing all frequencies of the audible range) to mask out the
normal feedback from the voice to the ear provided by bone-conduction.
These findings give promise of helping us work out a better theory of the
motor aspects of speech production, and this may eventually point the way
towards more reliable therapy for stuttering than is now available.

The development of language in children who are congenitally deaf or suffer

a marked hearing loss before they acquire speech raises two particularly
interesting theoretical questions: (1) Since these children may have normal
or even superior intellectual functioning, what kind of intellectual functioning
and symbolic behavior do they have prior to the acquisition of true language
responses? And (2) since language responses have to be acquired without the
normal dependence on sound, can the language fluency of normal speech ever
be attained without this dependence, and if so, by what means? Of secondary
but by no means insignificant interest are the sign or gesture languages that
are often learned and used among the deaf as an apparently normal means of
communication (in fact almost as “mother tongues”) and the teaching of lip-
reading.
Deaf children who have not acquired language, not even a sign language,
can perform nearly all of the perceptual and cognitive tasks that hearing
children of roughly comparable ages can perform, as long as these tasks do not
involve language in any way. Any retardation that deaf children may show
(typically, development is delayed by about one year) is likely to be due to
the fact that their experiences of the world are inevitably limited in certain
ways; it is not necessarily due to the absence of language. Research by Hans
Furth in America and by Pierre Oléron in France has rather well established
that deaf children without language can acquire concepts, compare mag-
nitudes, remember sequences and associations, and solve simple problems
involving forms, colors, and the like. These performances are generally well
above the level of cognitive functioning that can be secured from primates.
Individual
2 C. Cherry and B. M. Sayers. J. psychosom. Res., 1956, 1, 233-246. Differences
in Language
Behavior

73
 ” without
These findings suggest strongly that there can be a kind of “thought
language, and this possibility will have to be kept in mind during the ensuing
discussion of the relationship of language and thought.
The task of teaching a deaf child to speak has been called one of the most
the
difficult teaching tasks there is; in fact, from ancient times up to about
eighteenth century it was believed impossible, and deaf children were treated
of
essentially like idiots. The completely deaf person has no proper way
monitoring his own performance, kinesthet ic-tactua l feedback is a far cry
from the auditory feedback upon which the hearing person depends. Fur-
thermore, this kinesthetic-tactual feedback bears virtually no relation to the
visual stimulation the deaf person has to use to interpret what is spoken to
him by others. Lipreading, or “speech reading” as it has come to be called
(because it involves more than just the lips), is at best a difficult and some-
what unreliable form of communication, since many phonemic contrasts of
auditory language disappear and a great many words are in effect homonyms.
For example, the phonemes /m, b, p/ are visually identical, and therefore
the speech reader must rely heavily on context and probabilities to distin-
guish such words as mill, bill, pill. It is nevertheless possible to teach a deaf
person to speak with fair intelligibility, and to understand the spoken language
of others. The learning process is exceedingly slow and arduous. The evidence
strongly suggests that it is not merely a truism that the critical difficulty for
the deaf in learning language is their inability to hear large amounts of lan-
guage spoken. No amount of reading of written language seems to be able to
fill this gap; the grammatical patterns of printed words cannot impress
themselves on the mind as in spoken language. This conclusion adds force to
the conviction that writing cannot serve as an independent system of com-
munication. It is interesting to note, however, that the manual sign system
of the deaf does function as a normal system of communication; it has a
grammatical and semantic structure of its own. That it is often learned by
deaf children considerably later than the normal period of first language
learning seems to indicate, contrary to the opinion which has sometimes been
expressed, that the normal age of learning language (say, from age one to age
four) is not a maturational “critical period” in the sense that language not
learned at that time will never be learned.

Individual
Differences
in Language
Behavior

74
Cognition and Thinking In this chapter we are go-

ing to talk about the psychological analysis of a series of

concepts that we refer to every day with such words as

knowing, thinking, and believing. Ordinary language does

not yield any precise conceptions of the denotations of

these words. Take a word like think: Consider the variety

of meanings denoted in the following contexts.

I’m just thinking. ... What do you think about him?

... 1 wish I had thought of that possibility... . I

think he’ll come soon. . . . I couldn’t think of his name.

. . . He was thinking of his childhood. . . . I think that

person is insane. . . . Think this through carefully... .

75
 The root meaning here seems to be “some kind of covert (unobservable)
behavior in response to stimuli that may be absent, not immediately present
to the senses.” But in some of the above sentences the word has overtones
too,
of belief, or remembrance, or expectation, or mental discovery. Often,
it implies some kind of self-generated, more or less prolonged covert activity
oriented towards a desired outcome such as a plan of action, a piece of
writing, or the solution of a problem. The layman is usually concerned with
all these kinds of thinking: He wants to think “correctly” or more “effi-
ciently,” to solve problems more swiftly, or to make ingenious mental dis-
coveries when he searches for them. To ask the psychologist for prescriptions
to improve thinking, however, is like asking an engineer for suggestions for
improving the process of locomotion. Just as the engineer could only give help
if he is asked about specific kinds of locomotion and the specific conditions
that might obtain, so the psychologist cannot say much about a general con-
cept like thinking unless specific kinds of thinking are at issue.
This is also the problem when the question of the relation between lan-
guage and thought is raised. Obviously, many kinds of “thinking” are non-
linguistic. Some musicians report being able to “listen” to the music they
are composing, before writing it down on paper or even playing it on an
instrument; this sort of activity would qualify as a nonlinguistic kind of
thinking, and there are parallels in other spheres of activity—for instance,
thinking through a planned aerial maneuver, a swimming stroke, or a dance
step. In fact, we may lay it down as a proposition that any kind of behavior
that can be observed overtly may also be represented in covert form. Overt
speech can therefore be represented in what is sometimes called “inner
speech.”
Even the process of recognition is a form of covert, unobservable behavior,
as was pointed out in Chapter 3 (p. 35). We can, of course, often obtain
overt, verbal reports of what is recognized, but that recognition is inherently
a covert process seems to be a reasonable conclusion from the fact that under
certain conditions we can obtain a verbal report concerning the recognition
of a stimulus which is not objectively present. For example, hungry people
will sometimes report seeing faint pictures of foods when they are actually
shown nothing but a blank screen, and subjects can also be induced to report
recognition of words in a tachistoscopic experiment even when the exposure
contains no words at all.1
If, therefore, recognition of external stimuli is a covert process, which may
or may not be overtly reported (and an overt report may or may not be
verbal—it may consist simply of pressing a key), it is an easy step to the
notion that there may also be recognition and processing of zmternal stimuli.
One of the most challenging tasks for the experimental psychologist is to
obtain objective evidence for such internal responses—at least, to build a
sufficiently water-tight case for the inferred existence of such responses, if
they exist.
Early behaviorists like Watson actually assumed the existence of such
1]. Goldiamond and W. F. Hawkins. J. exp. Psychol., 1958, 56, 457-463. These in-
vestigators showed, incidentally, that even when the exposures were blank the subjects
tended to report words (nonsense syllables) in proportion to the frequency with which
they had experienced them in prior learning trials.

Cognition
and Thinking
76
responses, but assumed them to be the reduced action of the speech muscula-
ture. This theory provoked a series of experiments to find out whether thought
activity could go on without any kind of response detectable by electronic
monitoring of the musculature. The net result of this series of investigations,
which were plagued by almost insuperable methodological difficulties, was
to support the propositions that mental activity occurs in the central nervous
system and that it need not be accompanied by activity in the motor system.
It was demonstrated that perceptual and thinking activities could occur and
be remembered even when the motor system of an individual is completely
immobilized by a drug of the curare family. When activity is detected in the
motor system, there is reason to think that it is controlled by the central
nervous system.
Asserting that thinking is a central rather than a peripheral process
does not exclude the possibility that under certain circumstances peripheral
processes may interact with central processes, either facilitating them or in-
hibiting them. We have already seen (pp. 45-46) a clear example of this in the
case of the motor processes of speech; even though speech is controlled
centrally, that control appears to depend upon a continuous feedback of in-
formation from the periphery—both auditory and kinesthetic—and a dis-
turbance of this feedback causes alterations in the kind of control exercised
centrally.
Another interesting case is that of the muscular movements that some-
times accompany silent reading. It has been shown” that mature, expert
readers can read even fairly difficult material without any detectable innerva-
tion of the speech musculature, but when the material becomes difficult
enough, they will start making subvocal movements of the tongue and lips,
as if subliminal speaking of the material will help them comprehend it. This
finding is also, possibly, evidence for the proposition that some forms of
mental activity develop through the gradual suppression of their overt
counterparts. According to some theoreticians,* “inner speech” (that is,
thinking which follows verbal patterns) develops in the child only in this way
—when it is no longer rewarding for the child to say everything aloud. On the
other hand, it is at least equally reasonable to assume that speech normally
originates as a total process involving both central and peripheral activity,
and that the child only gradually learns to suppress the peripheral component
when he so desires. This latter idea is the one to which I subscribe; it allows
us to assume that there can be some kind of mental content that is not
necessarily accompanied by speech activity. It makes more plausible, too, the
finding that prior to language acquisition children manifest cognitive activity
of considerable complexity—as in congenitally deaf children.

2A. W. Edfeldt. Silent speech and silent reading. Chicago: University of Chicago
;
Press, 1960.
3L. S. Vygotsky. Thought and language. Cambridge and New York: M.1.T.-Wiley,
1962.

Cognition
and Thinking

(GE
 Lp
WER
THE
Vy WYS MS CH BR Ser Ba yK
DEVELOPMENT
PY wr
OF
wre
ysgr ee Y
THOUG ey 8 IN THE CHILO

The Swiss psychologist Piaget and his associates have been responsible for
an intense and protracted program of research on the development of thought
in the child; their interest in the language development of the child has
actually been secondary. Piaget distinguishes four main periods in the de-
velopment of the child’s thought, and since in their chief features his findings
have been confirmed by other investigators, they deserve major emphasis here.
The average child in Western culture passes through the following stages
of mental development:

acquisition of perceptual invariants: to two years of age

preoperational intuitive thinking: two to seven years of age
concrete operational thinking: seven to eleven years of age
formal, propositional thinking: eleven upwards

The ages given may be thought of as mental ages, in order to apply to children
exhibiting different rates of development; of course, some hapless children
never arrive at some of the later stages. The stages are cumulative; even in the
stage of formal propositional thinking one is still acquiring perceptual
invariants.
The first stage lays only the foundation for thought development. It is the
stage during which, as we have already mentioned on page 35, the child
learns to identify the main features of the world around him and some of
their essential properties. He has to learn to perceive certain aspects of his
environment as invariant despite the various forms in which they may appear;
these perceptual invariants may be thought of as the basis of thought and
language. The child learns the “meanings” of these percepts not only in
terms of their direct sensory qualities but also in terms of the way objects
and surfaces react to the various kinds of manipulative responses (touching,
hitting, biting, and so on) that he finds he can make to them. Toward the end
of this period the child has built covert, internalized, representational re-
sponses around these perceptual invariants, for he can delay his responses
to a stimulus to a time when it is absent.
In the next stage of mental development distinguished by Piaget, the child
wrestles with further problems in the interpretation of his environment,
namely, the understanding of relationships among the perceptual invariants
he has come to recognize. He must arrive at elementary concepts of space,
time, and causality, but in so doing he remains for a considerable time in a
“preoperational stage” in which he makes what Piaget calls intuitive judg-
ments about relationships. For example, if the child is shown two rows
of beads, each containing four beads but with one row spaced further apart
than the other, he will in this stage consistently act and behave as if the more
widely spaced row actually has more beads. Likewise, the child in this stage
will maintain that a tall beaker, into which water from a low, wide beaker
has been poured contains more water (or occasionally, Jess water) than was
present when the self-same water was in the low wide beaker. The child
Cognition
and Thinking
78
has not arrived at a notion of the conservation of number or quantity. He
attends to only one property of experience at a time, and cannot see how
two or more properties (such as height and width) can interact or trade off
with each other.
As a result of further learning through experience, the child eventually
passes into the stage of “concrete operational thinking.’”’ He has acquired
concepts involving complex relationships, such as that of the conservation
of amount, weight, volume, size, and number, and has attained what Piaget
calls reversible thinking—that is, thinking that can trace a physical opera-
tion back to its starting point and account for the transformations in its
appearance. He can classify objects into groups of different sizes on the basis
of different qualities; he can arrange objects in order of magnitude with
respect to a given attribute, and he can perform such operations as sub-
stitution and the recognition of equivalences. But all his thought is bound to
actual, tangible, visible materials and objects. He cannot at this stage imagine
possible, potential relations among these objects, or manipulate possible rela-
tions among absent objects.
These latter capabilities develop, according to Piaget, only during the
stage of formal, propositional thinking, that is, at around the start of ado-
lescence for most children. It is during this stage that the child starts to think
in terms of purely logical propositions which can be stated and tested against
facts drawn from other experiences. This is the stage at which the child begins
to be able to deal effectively with formally stated syllogisms.
Piaget and his associates have been principally concerned with describing
the stages through which the child passes in development toward adult think-
ing. Their research program may be looked upon as a very elaborate testing
enterprise, with an effort to understand the results noted in terms of a
logical analysis and description of the mental operations or processes in-
volved. They have made no attempt to investigate the possible effects of
specific, deliberate tuition of mental processes, even though Piaget has in-
sisted that mental development occurs only through processes of learning.
Some American and British research has suggested, however, that although
t
Piaget’s stages are correct in their sequence, children’s mental developmen
can be hastened somewhat through specific teaching. For example, young
of
children in the preoperational stage are, according to Piaget, incapable
l letters, but American research shows
identifying the shapes of alphabetica
that this behavior can be produced in these children through appropriate
learning
schedules of discrimination training. Even though such accelerated
even the psychologist ! )
can sometimes be effective, what the adult (and
the large number of steps and the great
sometimes fails to appreciate is
to progress
variety of experiences that the child needs to go through in order
from one stage to another.
g of the
The unifying theme in the work of Piaget is the gradual unfoldin
“model” of the universe around
individual’s ability to construct an internal
on that model so as to draw conclusi ons
him and to perform manipulations
history of his environ ment or the probabl e results of
about the probable past
ment. The ability to
possible actions that could be taken upon that environ
al meaning s of the
do this is the essence of all “thinking” in the nontrivi
by Piaget correspond
term. The four stages of mental development listed
Cognition
and Thinking

79
 to four stages in the working through of any process of thinking. The pre-
thinking stage in which “perceptual invariants” are acquired by the infant
corresponds to a stage of concept formation or concept attainment in which
the basic entities which function in any particular context must be identified
and recognized. The preoperational, intuitive stage may correspond to a type
of “incubative” thought reported to occur even in adults when concepts in-
volved in a problem are allowed to interact somewhat freely. The concrete
operational stage corresponds to a stage when one experiments either overtly
or covertly with the tangible referents of these concepts. The formal, proposi-
tional stage corresponds to the process of constructing alternative hypotheses
regarding a problem, or linking together a series of inferences concerning a
situation.

No process of thinking occurs without a cause. One class of causes is sub-

sumed under the heading of motivation. In infancy, primary drives such as
hunger, thirst, and the need for warmth provide a basis on which certain
objects (such as foods, blankets) are discriminated, recognized, and built
into concepts, but it is difficult to account for all the behavior of this period
without also making reference to secondary, learned drives. In childhood,
thinking is motivated not only by the need to solve problems concerned with
the child’s interactions with other people and with his environment, but also
by a “need to understand” or to know, reinforced by experiences in which
knowledge about the environment has been put to good use in solving prob-
lems of adjustment to it. Possibly this kind of motivation has its roots in the
“orienting reflex,” described by Russian psychologists as the primitive tend-
ency noted in both animals and young children to pay attention to any novel
stimulus.
Adults’ thinking is also motivated, whether very diffusely, as in day-
dreaming and reverie, or very specifically, as when a particular problem
urgently needs to be solved. An especially strong motivation for thinking
arises from what Leon Festinger calls cognitive dissonance—a state of affairs
that occurs whenever two ideas are in marked conflict, as when one is pre-
sented with an objective fact that appears to undercut one’s cherished beliefs.*
Festinger shows that people are strongly motivated to reduce such cognitive
conflict—either by changing their attitudes, seeking more information, or
restructuring or reinterpreting the information available to them.
In studying the motivation of thinking in problem-solving, the effect of
instructional sets is noteworthy. The subject can be set to direct his thinking
in a given way depending upon the task (German: Aufgabe) that he is given;
this set has sometimes been called the ‘determining tendency.” If a subject is
given an Aufgabe to multiply the pairs of numbers given him, he will per-
form this task continuously whenever pairs of numbers are presented, but he
can equally well be switched to a very different kind of response simply by

4L. Festinger. A theory of cognitive dissonance. New York: Harper & Row, 1957.

Cognition
and Thinking
80
being told to add. Similar phenomena have been noted with controlled asso-
ciation tasks, as where the subject is given a series of directions like: “give
the opposite of: SMALL ... ; a superordinate of: CAT .. . ; a subordinate
to: FOOD. ...” It is as if the effect of the instruction is to program the
subject to take a certain course of mental activity. The reader may notice
the similarity of this kind of instruction to the programing effect which was
claimed for a grammatical sentence in Chapter 4. A heard sentence is for a
listener a kind of problem in thinking, and the effect of its grammatical
structure is to induce sets for the responses to the stimuli referred to by the
sentence.

Any analysis of thinking must accord an important role to what we call

concepts. This term has cropped up in this book rather frequently, but we
need to elaborate it still further.
The first concepts formed by the young child are the perceptual invariants
of objects, sensations, sounds, and feelings that we have already mentioned.
They are internal representations of classes or categories of experience. As the
child learns language, he learns socially reinforced names for these categories
of experience. He can even shape his behavior around internal representations
of concepts; for example, a child at a certain age can take a pencil and draw a
square on demand. Not all concepts can be overtly manifested in this way, of
course, but a child who can correctly recognize instances of a particular con-
cept and distinguish them from noninstances thereby demonstrates his acquisi-
tion of the concept.
Not all concepts are built out of raw sensations, either. Apparently some
,
concepts are built out of partial similarities in the responses to sensations
are internal, it is tempting to say that some
and since some of these responses
“opposite-
concepts may be built out of other concepts. Take the concept of
that one
ness,” which must be built out of instances in which it is noticed
d with the other extreme.
extreme of any dimension of sensation is contraste
concepts like “number, ” “relation, ” or
Similar analyses may be made for
has been studied by Piaget. We can now broaden
“randomness,” whose genesis
repre-
our definition of concept by asserting that any concept is the internal
es being either the
sentation of a certain class of experiences, these experienc
nt, or responses to other
direct response to aspects of the external environme
experiences.
experiences may
In theory, an infinite number of concepts are possible, since
ily constructed by
be classified in an infinity of ways. A concept can be arbitrar
three and five
combining other concepts: “All Colorado spruce trees between
acres or more.” But most con-
feet in height situated on U. S. farms of 100
e, science, and the arts, are based on
cepts used in daily life, or in commerc
which have been found useful in some way. It has
classifications of experience
discovery and formula tion
taken intelligence of a high order to make the first
“entrop y,” or, in the
of certain concepts like “gravitation,” “relativity,”
These are indeed classifi cations
psychological sphere, “operant conditioning.”

Cognition
and Thinking
8I
 of experience in the sense that there can be instances and noninstances of each
of them; that they play a role in interpreting experience follows from our
consideration of the use of concepts in thinking.
Concepts may vary in their degree of novelty and complexity for the
individual. For an individual without considerable training in mathematics and
physics, attaining a concept like that of entropy may be quite difficult because
he may first have to attain an extensive series of prerequisite concepts. For a
young child, attainment of the concept oppositeness may be equally difficult;
he is unlikely to attain it until he has experienced oppositeness in a number of
dimensions and notices a common pattern in these dimensions. On the other
hand, many concepts may be very simple to acquire; often an individual can
learn them by simply reading or hearing a verbal formulation of them; for
example, the concept “card with two triangles and a red border” could prob-
ably be rather easily learned in this way by intelligent adult subjects. Most
concepts an individual has to learn in school are of intermediate difficulty;
usually the individual has to acquire them both by studying verbal formula-
tions and practicing the recognition of instances and noninstances. The stu-
dent of law or history, for example, will probably learn the concept tort in this
way.
You must be careful to note what definition of concept learning is being
used in a given instance. One definition has the virtue of complete objectivity;
according to it, a person has learned a concept when he can with a high degree
of reliability discriminate between instances and noninstances. This definition
is usually satisfactory, but many individuals who know a concept by this
definition are not able to formulate the concept verbally (or in whatever mode
of communication is appropriate, for example, in visual or acoustic terms)
or to communicate it to others. In fact, several experiments have shown, ap-
parently, that it is possible to learn a concept without being aware of the
basis for it; the individual simply learns a response to the significant features
(that is, the “‘criterial attributes”) of positive instances of a concept without
being aware of this response. In one experiment,” Lorraine Bouthilet had sub-
jects memorize a series of pairs, like elephant-path and recognize-zero. She
then presented them with multiple-choice items like the following:

hexameter: (1) bib (2) tax (©); fat «<@) get

Many subjects were able to choose the correct answer—tax—just on a

“hunch,” not realizing that the correct answer was always formed out of
letters included in the stimulus word. Because such “unconscious” concept
formation is possible, in some contexts, it is useful to define concept learning
in terms of the ability to recognize instances and the ability to formulate
descriptions, or to construct instances of the concept.
The role of verbal formulation in the attainment of concepts has been very
little studied even though it is one of the prime methods of instruction used in
schools. There is the danger of teaching merely parrot-like verbal formula-
tions; William James, for example, recounts how some students were asked
5 Lorraine Bouthilet. The measurement of intuitive thinking. Unpublished Ph.D. thesis,
University of Chicago, 1948. Cited by R. Leeper, in S. S. Stevens (ed.). Handbook of
experimental psychology. New York: Wiley, 1951, p. 745.

Cognition
and Thinking
82
whether it would get warmer or colder as one dug a hole deeper and deeper
into the earth; they could not reply, but prompted by their teacher they
parroted the sentence “The interior of the globe is in a condition of igneous
fusion.”
Perhaps because of anecdotes like these (which do not really prove any-
thing), the power of verbal formulation in teaching concepts has probably
been underestimated. Verbal formulation should be valuable particularly when
it is followed by copious presentation of positive and negative instances. A
number of relevant experiments in the context of programed instruction in-
dicate that the teaching of concepts can be accomplished by the presentation
of “rules” and “examples,” in that order, more effectively than by the pres-
entation of examples followed by rules.
Extensive studies of concept attainment without the aid of externally
supplied verbal formulations have been conducted by Heidbreder, Bruner,
and others. Most of these experiments involve relatively simple combinatorial
concepts utilizing dimensions of color, shape, number, and spatial position.
Their basic design is to present subjects with a series of stimuli. The subject
is told which stimuli exemplify, and which do not exemplify, the “concept,”
which the experimenter has decided on in advance. For example, in one of
Edna Heidbreder’s experiments the subject has to learn which pictures of an
array are to be labelled with a particular nonsense syllable. Such experi-
ments are more properly experiments in problem-solving, because the criterial
attributes (things, forms, numbers, colors, and so on) are already well known
to the subjects (usually college students). Nevertheless, it is of interest to
study what processes occur in the course of a student’s attainment of the
solution.
Heidbreder’s original notion that there is a natural order of concept
attainment, in which concrete concepts (such as birds, faces) are more easily
learned than abstract concepts (such as number and spatial position) seems
to have been effectively disproved by evidence that the critical variable is the
number of stimulus properties perceived by the subject.®
In conducting variations of this type of experiment, J. Bruner, J. Good-
now, and G. Austin” found that the ease of learning a concept was partly a
function of its sheer complexity (the number of dimensions employed in it,
and the like) and also of its logical structure; other things being equal,
conjunctive concepts are easier than relational concepts; the most difficult
of all are disjunctive concepts. A conjunctive concept is defined as one for
which a specified combination of attributes is criterial (for example, “red fig-
ures with borders”) ; a disjunctive concept is one for which any of two or more
alternative combinations of attributes is criterial (“either a red figure or
one with 2 borders”); and a relational concept is one in which a specified
relation between attributes is criterial (“fewer figures than borders”).
or
Bruner and his associates were also interested in the “strategies”
(college- age) subjects in solving the
“cognitive styles” adopted by their
problems of concept attainment presented to them. These strategies, it should
be borne in mind, apply to a situation in which the subject is shown a positive

63-64, in this series.

6 For further details, consult S. A. Mednick. Learning, pp.
and G. Austin. A study of thinking. New York: Wiley, 1956.
7J. Bruner, J. Goodnow,
Cognition
and Thinking

83
 instance of a concept and then asked to guess which other stimuli in an array
are also positive instances; he is informed of the correctness of each suc-
cessive guess. Four strategies were distinguished, as follows:

1. Simultaneous scanning: systematic trial of alternative hypotheses, with

careful account taken of the information obtained from each success or
failure.
2. Successive scanning: trial of one hypothesis at a time but without
taking full advantage of the information supplied by successes and failures,
so that some of the guesses are actually redundant or inconsistent.
3. Conservative focusing: trial of conservative variations of the selected
focus or positive instance.
4. Focus gambling: drastic changes of focus, made in the hope of hitting
on the criterial attributes by a process of elimination.

This kind of concept-formation experiment illustrates situations in which the

individual develops and tests hypotheses (that is, “tentative” internal repre-
sentations of experiences) concerning the concepts he is to acquire. Such
behavior parallels, at a simple level, the behavior of the scientist seeking
regularities in the phenomena he is studying.
There are, however, kinds of concept-attainment tasks where the concepts
are so difficult or the attributes so lacking in salience that learning is gradual
and hypotheses seem of no avail. In such cases, subjects find they must resort
to “spectator behavior,” simply waiting for the presentations to suggest suit-
able hypotheses.
Whatever the case may be, the most interesting object of study in these
concept-formation experiments is how the individual arrives at hypotheses
to test, for testing a hypothesis is itself relatively easy. Fast learners in these
experiments are those who are facile in constructing hypotheses, either on
account of some general trait (intelligence?) or on account of their previous
acquisition of a rich variety of patterns of response likely to be useful in such
experiments. Thus, transfer of prior learning (‘learning to learn”) can
be effective in concept-attainment tasks.

During a lifetime, an individual acquires a goodly stock of concepts. He

may also have acquired names (words or phrases) for many of these concepts,
but it is not necessary for all concepts to have names. Some remain at a
kinesthetic or perceptual level: For example, the concept of the lever is
utilized by a farmer when he pries up a stone, even though he may not
verbalize it with either of the words lever or pry. In such a case it might be
thought that we could dispense with the notion of concept and assert that
the response of the farmer is a direct learned response to a particular kind of
problem, namely a stone which is hard to move. Nevertheless, the fact that
the farmer may exhibit considerable planful behavior—going to get a crow-
bar, digging a socket for it, and finally moving it in a certain direction—
Cognition
and Thinking
84
suggests that there is more than a direct overt response to the problem situa-
tion. On the other hand, the farmer might be hard pressed if someone asked
him to explain how even a not-very-strong child can move, with a crowbar,
a stone much heavier than himself.
All problem-solving—that is, thinking oriented toward the solution of prob-
lems—can be regarded as the manipulation of concepts that are evoked by the
total situation and that may or may not be relevant to the task at hand.
Depending on the nature of the situation, this manipulation may be at one
extreme wholly internal, that is, not accompanied by detectable overt be-
havior, or at the other extreme, it may be almost wholly overt, directly in-
volving relevant aspects of the environment. The former extreme would be
exemplified in the solution of a numerical problem by a lightning calculator
using mental arithmetic, the other extreme would be represented by the
solution of a mechanical puzzle by manipulating it with guided trial and
error. We must also recognize the utility of vicarious forms of interaction
with the environment, such as making pencil sketches or physical models,
solving mathematical equations on paper, or verbally formulating tentative
conclusions. Among the factors that may determine whether an individual
will solve a problem are the following:

1. The individual’s repertoire of relevant concepts.

2. The concepts evoked in the individual by the structure of the problem.
3. The individual’s skill in manipulating the concepts evoked, his strategy
of solution, his flexibility in changing his mode of attack, and his ability to
perceive the relevance of a concept.

These points can be illustrated by reference to what is probably the most

famous series of experiments on problem-solving—Norman Maier’s experi-
ments with the “two-string problem.” ® A subject is introduced into a room
with two pieces of string hanging from the ceiling and told that he is to tie
their ends together; they are too far apart for him to reach both ends at
once. The room is bare except for a chair, a piece of wire, and a pair of pliers.
Several solutions of the problem are possible; one of them, however, seems
to be particularly difficult for subjects to attain. This is the solution in which
the end of one of the pieces of string is to be weighted with the pliers, set
swinging, and caught after the subject has moved to grasp the end of the
other string. The critical concept is that of a pendulum, and success usually
follows as soon as the subject sees that he must make one of the strings into
a pendulum. For many subjects, however, the situation does not evoke this
concept readily; hanging strings are not perceived as potentially swinging,
nor are the pliers perceived as a weight rather than as a tool. (The failure
been
to see that an object can have a function other than its usual one has
Judson, C. N. Cofer, and
called “functional fixedness.”) However, A. J.
who have somehow been reminded of
S Gelfand? found that subjects
introduced to this problem (for example, by doing
pendulums before being
to solve the
some memory work that involves the word pendulum) will tend

SN.R. F. Maier. J. comp. Psychol., 1931, 12, 181-194.

1956, 2, 501-507.
9 A. J. Judson, C. N. Cofer, and S. Gelfand. Psychol. Repts.,
Cognition
and Thinking

85
 problem more quickly than otherwise. Yet another technique for evoking
the relevant concept, noted by Maier himself, is for the experimenter “acci-
dentally” to brush against one of the strings, setting it swinging; in this case,
however, the objective situation is changed so that it is more likely to evoke
the relevant concept. Interestingly enough, in this latter case subjects are
usually not even aware of the cue provided by the experimenter—a fact
which casts suspicion on the dependability of subjects’ verbal reports in
studies of the thinking process.
We have discussed the cues to concepts provided by the problem or task
itself, and the role of the individual’s repertoire of concepts. But equally
important is the way in which these concepts are manipulated in the process
of arriving at solutions. Many problems require a sequential processing of
concepts wherein each process yields some partial answer or tentative result
to be subjected to still further processes, as, let us say, in long division. Fac-
torial studies of individual differences in reasoning ability suggest that success
in this kind of problem-solving partly depends on the individual’s ability to
retain these partial answers in short-term memory so that they are available
for the next steps.
The steps themselves, of course, must be correct if the final solution is to
be attained. Very often these steps take the form of inferences: since A 1s
the case, it follows that B; since B is the case, C follows; and so on, when 4,
B, and C are propositions or statements of fact. The study of what kinds of
inferences can properly be made from given propositions or premises is part of
the domain of logic, a subject whose relations with psychology now must be
considered.

LOGIC AND HOLOGY OF THINKING

The following problem was among a series presented to a number of college

students who had no special training in logic:

Mrs. Cooke had studied home economics in college. “Youth is a time of rapid
growth and great demands on energy,” she said. “Many youngsters don’t get enough
vitamins in their daily diet. And since some vitamin deficiencies are dangerous
to health, it follows that the health of many of our youngsters is being endangered
by inadequate diet.” Does it follow that the health of many youngsters is being
endangered by inadequate diet? Give your reasoning.

Mary Henle !° reported that most of her subjects agreed with this reasoning.
And most of these did so with no amplification beyond a restatement of the
reasoning given by Mrs. Cooke. The well-known fact that inadequate diet is
endangering the health of many youngsters seemed to influence the readiness
of the subjects to endorse this reasoning. And yet some subjects must have
had an intimation that the reasoning as stated was faulty. One said, “Correct,
if we assume that the youngsters are lacking those vitamins in their diet which
endanger health.” Another said, “It seems to follow, assuming that the

10 Mary Henle. Psychol. Rev., 1962, 69, 366-378.

Cognition
and Thinking
86
 deficient vitamins are also the vital ones.’’ We have no way of knowing how
many subjects sensed the faulty reasoning but “corrected for it” by adding
or assuming premises.
A great deal of the reasoning we do, and even the reasoning we see in news-
papers, speeches, and the like, would not stand up under careful analysis from
the standpoint of formal logic. Formal logic can be thought of as a branch of
mathematics which enables one to test whether a stated conclusion validly
follows from stated premises. The study of the syllogism is a kind of semantic
analysis of the meanings of propositions using the words all, some, not, and,
or, and several others (with carefully defined meanings) and the study of
the degree to which sets of these propositions could be a consistent map
or counterpart of selected aspects of the real world (or even an imaginary
world). Textbooks of logic present various easily learned procedures for
ascertaining the validity of inferences, and we shall not try to sketch them
here.
Though logic and psychology are independent in a formal sense, the way
we attempt to think logically—apart from training in formal logic—may be
influenced by psychological processes which are worth studying.
In reasoning, we can use language to construct a “map” of a possible ob-
jective reality; the validity of the reasoning is perceived when the map is
recognized as self-consistent. A simple reasoning process occurs when we have
to figure out a satisfactory route between two isolated points in a city with
whose general layout we are only vaguely familiar. We know how to get
from A to B, and from C to D, say, but the question arises, does going to B
take one too far out of the way to get easily to C and thence to D? Think-
ing out the route from B to C will help us perceive whether the whole plan is
efficient. A somewhat analogous procedure is followed by a person trying to
evaluate the validity of a reasoning process such as the one about vitamin
deficiency; the verbal statement is a “program” for a construction of reality
whose validity is to be tested. “Here are youngsters who don’t get enough
vitamins in their daily diet,” a person could say. “And some vitamin defi-
ciencies are dangerous to health. But are these ‘dangerous’ vitamin defi-
ciencies included among the deficiencies asserted to occur in those young-
sters? No, we are not told, so the conclusion doesn’t necessarily follow.”
The “map” is not completely consistent. (The “Euler circles” and “Venn
diagrams” presented in texts of formal logic are mechanical ways of translat-
ing syllogisms into “maps” and evaluating their consistency.)
In practice, as we have seen, people are likely to change or add premises
to make the conclusions accord with their knowledge or belief. And people are
distinctly troubled when they have to deal with premises stated abstractly
or “nonsensically,” as in the following:

All X are Y. All skyscrapers are three-legged chairs.

Some Y are Z. Some three-legged chairs fly.
Therefore, some X are Z. Therefore, some skyscrapers fly.

In dealing with syllogisms presented formally, like these, people tend also
to be influenced by what some investigators have called an ‘‘atmosphere

Cognition
and Thinking

87
 tends to evoke ac-
effect”: Affirmative premises create an atmosphere that
conclus ions, regardl ess of the absolut e logic of the
ceptance of affirmative
matter.
life is so seldom
Because the reasoning we do or experience in everyday
prone not to test its logical validity
stated in explicit syllogistic form, we are
in formal logic do not seem to produce
even if we can do so. Even courses
startling changes in students ’ ability to spot faulty reasoni ng.
ic in-
Faulty reasoning does not stem solely from failure to test syllogist
of premises , or the setting
ferences. Frequently there is uncritical acceptance
‘‘fluorida-
up of improper equivalences between terms. If it is asserted that
a medicati on,” it is necessar y to know how people
tion of the water supply is
like “fluorid ation” and “medica tion” before one can ap-
understand terms
praise the validity of the statement or use it in further reasoning.
words
Many of the effects of persuasive language arise through the use of
meanings . ‘““Ragern ess” and “aggressi ve-
with emotionally toned connotative
used to refer to the same kind of behavior, say, in
ness” might properly be
n or an executive , but the former term has a higher position on
a politicia
to
the semantic differential Evaluative Scale (p. 103) and is thus more likely
arouse favorable attitudes toward the person to whom it is applied.
to
A wryly delightful example of the skillful use of connotative meanings
suppress the evocation of undesirab le emotional attitudes is the language of
undertake rs, as reported by Jessica Mitford in The American Wav of Death.
Instead of the body or the corpse, they use the name of the deceased: Mr.
Jones, or Mrs. Smith, or whatever, thus arousing the responses that had been
made to the living person. They do not speak of digging and filling a grave,
but of opening and closing it. The person (not the body) is interred rather
than buried, not in a cemetery or graveyard but in a memorial park. The word
death is avoided at all costs: A death certificate is a vital statistics form, and
the deceased didn’t die, but expired.

Cognition
and Thinking
88
 Language and Cognition Even though, as we have

seen in Chapter 6, many kinds of thinking are possible

without language, language can obviously play a large role

in thinking—a role that will be explored in this chapter.

Among the ideas that we shall want to examine are these:

that language may facilitate thinking, allowing it to be

more complex, efficient, and accurate; that language may

in some cases inhibit or misguide thinking; and that the

structure of a particular Janguage may channel thinking

and thus cause the users of that language to think either

ee
et*eee,
ee
Oe more or less efficiently and accurately than they would if

they were to use another language, or even to arrive at

89

ee
ae
 from what speakers of
different conclusions or different solutions to problems
the other language would do.

both animals and

In the last chapter we studied the processes by which
representation
children acquire concepts; we defined a concept as an internal
concept s are acquire d without
of a class of experiences. We saw that many
language.
s? It is pos-
What is the relation of language responses to these concept
an underlying
sible, of course, for a child to learn a language response without
underst anding it, or
concept—he may learn simply to echo a word, without
But learnin g to use a word in a
he may use it in an inappropriate context.
it in such a way that it will be consistently
meaningful way—that is, using
lies that the child has acquire d the concept which
socially reinforced—imp
s the linguist ic response . The child who can use the word ball for the
underlie
do has, we
same class of experiences the members of his speech community
and he will use this word when he
may say, acquired a concept called ball,
If he shows any tendenc y to overgen eralize the
encounters new instances.
concept—say, to call a strawbe rry a ball—he will be correcte d. If he under-
the
generalizes the concept—that is, fails to apply it to certain experiences—
use it for that class of
tendency will probably be corrected when others
experience.
a
One function of linguistic forms is to provide a cue for the formation of
The adult who tells a child that there is such a thing as a
new concept.
platypus, for example, alerts the child to the existence of a possible class of
experiences; pictures and descriptions of platypuses then help the child to
fix the boundaries of this class of experiences. Perhaps the child will never
actually see a live platypus. Even the word unicorn is a name for a possible
class of experiences and the child who learns this concept would at least be
able to identify a unicorn if one ever presented itself! Some concepts are
explicitly imaginary, like i = \/ —1 in mathematics; they refer to a con-
ceivable and useful class of experiences that will, it is known, never occur in
reality.
One characteristic of a language that can be used in general communica-
tion (whether it is “natural,” like English or Chinese, or “artificial,” like
Esperanto or Basic English) is that it provides words or linguistic forms
sufficient to catalog or describe all or nearly all the experiences or classes of
experience that occur to the user of the language. Of course, languages vary in
the sizes of their vocabularies; the vocabulary size of a language is chiefly a
function of the state of advancement of the civilization which underlies it.
But the “core vocabularies”—that is, the vocabularies of everyday parlance
and general use in writing—of all languages are of roughly the same order of
magnitude, say, somewhere around 10,000 words, and furthermore, they
show a surprising degree of correspondence. That is to say, the speakers of
the world’s languages agree to a considerable extent on the concepts they
have found convenient to symbolize with words. To a large degree, this is
because of certain uniformities in the physical and biological environments of
Language
and Cognition
go
mankind: sun, moon, water, fire, stone, flower, bug are names of physical or
biological entities found nearly everywhere, and in nearly every language
there is one principal linguistic form for symbolizing them. In Helen Eaton’s
Semantic Frequency List for English, French, German, and Spanish, 662
concepts are listed that in each of these four languages have words occurring
in the first thousand words in order of frequency; this list starts off (in
English) with the “concepts” a, able, about (concerning), about (approx-
imately), be about to, above. In Basic English, an essentially artificial lan-
guage based on English, developed by C. K. Ogden, it is possible to express
any nontechnical idea within a vocabulary of only 850 words.
In Chapter 2 we stressed that a language is made up not only of words,
but also of other linguistic forms both smaller and larger than words, and of
constructions. It seems reasonable to suppose that each linguistic form and
construction symbolizes a concept as we have defined it: an internal repre-
sentation of a (possible) class of experiences. This leads us to examine the
kinds of concepts symbolized in language. The form-class “allegiances” of
linguistic forms have a very interesting relevance here: On the one hand,
linguistic forms are learned as members of form-classes because of their
positions in constructions; since they are experienced in this way a form-class
is itself a concept (a class of experiences). On the other hand, the concepts
contained in a form-class themselves represent experiences which may tend to
have common elements among themselves and hence may tend to form a class.
These two tendencies result in the investing of a form-class with a certain
“meaning.” For example, the form-class of transitive verbs which includes
hit, kill, throw, drop, and other physical actions also includes the verbs have
and owe, which therefore may tend to absorb a meaning of ‘“‘physical action”
because they are classified with physical actions.
We can return, then, to the analysis of linguistic form-classes in Chapter 2.
What we there called meanings are now to be called concepts—internal
representations of classes of experiences. For speakers of English, we can
postulate the broad organization of major form-class concepts shown in
Table 4. These classes very nearly exhaust the conceptual structure on which
English form-classes are based: Linguistic forms that do not fit clearly into
any of these form classes are chiefly items that perform miscellaneous func-
tions in the mechanism of the grammar—mostly, they are signs or markers of
form-classes or of constructions, or directions for interpreting a certain con-
struction.
The constructions listed on page 20 can also be postulated to corre-
spond to concepts or classes of experience, and the meanings given there
can be taken to be approximate descriptions of those classes of experience.
These particular constructions are the patterns for complete utterances; the
underlying class of experiences they symbolize therefore include the commu-
nication situations in which these constructions are used. For example, the
subject-predicate construction which underlies the sentence Jim saw Bob
represents not only that class of experiences in which something is asserted
about a nominal (Jim, in this case), but also that class of experiences in which
information is furnished to a hearer (rather than being requested of him).
Many of the concepts of language are learned without the learner’s being
aware of them. This is true of most grammatical concepts. Even though these
Language
and Cognition
gi
 Major Form-Class Concepts

Approximate Conceptual Meaning—

Linguistic The Class of Experiences
Class Manifestation * That Includes:
s
Pr Se e
a

Nominals Nouns, pronouns, Objects, persons, ideas, and relations whose

noun phrases location or distribution in space, actually
or metaphorically, can be specified

Adjectivals Adjectives, adjective Qualities or attributes perceived as applying to

phrases nominals, either on an all-or-none basis
(presence-absence) or in terms of degree

Verbals Verbs, verb phrases Events, relationships, or states whose location

or distribution in a time dimension can be
specified

Adverbials Adverbs, adverb Qualities or attributes perceived as applying to

phrases adjectivals and verbals, either on an all-or-
none basis or in terms of degree

Prepositions, preposi- Relations of spatial, temporal, or logical posi-

Prepositionals
tional phrases tion relative to nominals

Conjunctives Conjunctions Logical relations occurring whenever any two

or more members of any class (or construc-
tion) are considered together

to be
* In each case it is to be understood that derivations from other form classes are
included.

concepts refer to certain classes of experience, many people never become

aware of these classes. If presented with two sentences with partially similar
structure, many people have difficulty in identifying the analogous parts.
For example, what word in the second sentence has the same grammatical
function as Aim in the first sentence?

We showed him the way to get there.

He wanted to buy his mother a present.

Some constructions, in fact, carry concepts that often seem to be hard to

attain on a conscious level. For example, the construction the X-er . . .,
the Y-er . . . (as exemplified in the faster I ran the better I felt) carries the
concept of correlation, which is sometimes hard for beginning students of
statistics to grasp. These same students can often be helped by referring to
this linguistic construction and its meaning.
It is probably at this point—that is, the grammatical construction—that
language structure begins to be of real help in aiding thinking beyond what
Language
and Cognition
Q2
could happen without language. By means of grammatical constructions one
can learn, remember, and manipulate more complex concepts, such as:

the boy’s hat

herbivorous mammals
the top of the Rock of Gibraltar
preoccupation with litigation
excess of income over outgo
psychologists aroused over ethical problems
two right turns after each left turn

Thinking aided by language is called reasoning, and the ability to reason

depends largely on the ability to formulate steps in an inferential process
in terms of language. An interesting question is, how complete and “gram-
matical” does such a formulation have to be? It is hard to get evidence on this
question, for reasoning usually occurs internally. There have been attempts
to observe these internal processes by the method of réflexion parlée (French
for “spoken thinking’’), in which the individual is asked to “talk aloud”
his reasoning processes. There can be no guarantee, of course, that the
subject can give a full verbalization of his reasoning processes, even with
every intention of doing so. Furthermore, it is possible that the very process
of producing an overt verbalization affects the course of the reasoning process ;
one study found that committing thought to writing too quickly delays its
full development, and the same thing may easily be true of “spoken thinking.”
In fact, most verbal reports obtained by this method seem to be just that:
reports of something that has already happened. The speaking a subject
does in the course of solving a reasoning problem does not seem to be iden-
tical with reasoning processes but instead a rather inaccurate and hazy report
of them. It is thus difficult to obtain evidence on the proposition enunciated
earlier in this paragraph—that the ability to reason depends on verbally
formulated inferential steps. Nevertheless, the content of reasoning processes
frequently can be stated only in verbal terms, even though the end result of
a reasoning process may be an action, such as a decision to buy an automobile.

Animals can perform many tasks that appear to require “thinking” or at

least some internal process which is not immediately open to observation.
For example, an adult monkey can be trained to perform the “double alterna-
tion problem”—that is, to learn that, in a sequence of trials in which he can
look for food under either a box at the right (R) or a box at the left (L),
the sequence which will always get him food is RRLLERRLE = Some
kind of symbolic activity appears to be involved because the monkey has
to remember, or keep track of, whether he has looked under a given box
before. We cannot say whether the monkey “counts” in any sense; all we
can really know is whether he can learn to perform the task.
One of the favorite strategies of psychologists in studying the mental
development of young children is to give them some of the same problems
that they give animals. That is, a child is introduced into a situation analogous
Language
and Cognition

93
 nter may
to what might be presented to a rat or a monkey; while the experime
(saying somethi ng like
talk to the child, he will talk only in general terms
will studious ly avoid telling the child
“We're going to play a new game”’) and
or what rules are to be followed . The advanta ge
what the experiment is about
e is that it allows the experime nter to study the speed and
of this procedur
or
accuracy with which the child can “figure out” for himself the rules
and merit in this ap-
principles of the experiment; there is much interest
proach. Unfortunately, experimenters have only rarely taken the additional
prin-
step of seeing whether their young subjects could also be taught the
ciples of the experime nt solely by verbal instructi on. For example, the double
alternation problem described above has been tried with children. It has been
found that they cannot learn to perform it until they are about three-and-a-
half years old at the youngest, and from that age until about the age of five,
they cannot verbalize the rule by which they perform it. By the age of five,
most normal children can both learn the task and verbalize it. There is no
systematically collected information about how early children can be taught
the double alternation problem by verbal instruction alone.
It seems obvious, however, that if a child has already learned a concept
verbally in the course of his everyday development, he would be more likely
to perform successfully in any problem situation where this concept is critical,
even though it does not figure explicitly. This conclusion is supported by a
number of experiments. C. C. Spiker, I. R. Gerjuoy, and W. O. Shepard
tested a group of children aged three to five and divided them into those who
could say something like “middle-sized” as the way to describe the middle-
sized member of a series of three stimuli, and those who could not.’ This
capability was then found to be highly correlated with the child’s performance
in a concept-attainment experiment where it was necessary to choose the
middle-sized stimulus from sets of three stimuli in which the absolute sizes
of the stimuli varied. (When the absolute size of the middle-sized stimulus
was constant, prior learning of the concept “middle-sized” turned out to be
irrelevant because the children could learn to respond to the absolute size
of the critical stimulus.) It should be noted, incidentally, that the training
period of the experiment itself was too short to allow learning of the concept
of middle-sized-ness in the case of children who had not acquired it before.
Apparently, learning a concept of this complexity is something that takes a
considerable amount of time; probably it depends on a variety of other
learnings that would have to be explicitly provided for if it were to be taught
in an experimental setting.
There has been much interest in the question of whether language responses
help or hinder nonverbal behavior in ways that go beyond the ones indi-
cated above. Does having names for stimuli help one respond differentially to
them? Does it help one remember them, or use them in further problems?
This question has now been investigated fairly extensively, both with chil-
dren and adults, although the answers we have are not always clear or
convincing, and we have no settled theory to explain the results.
One thing seems clear: Having names for things does not alter our ab-

1C. C. Spiker, I. R. Gerjuoy, and W. O. Shepard. J. comp. physiol. Psychol., 1956,

49, 416-419.

Language
and Cognition

94
 solute capacity to discriminate among these things when they are extremely
similar. Perhaps you have heard that workers in dye factories learn many
more names for colors and hence become better able to discriminate colors.
The latter part of this statement is not quite true, if by “discrimination” we
mean the ability to detect a small difference between two stimuli when they
are juxtaposed, as in a psychophysical experiment. Highly skilled workers in
occupations dealing with color or taste or any other sensory dimension
are on the average no better able to make psychophysical discriminations
than the average person (unless they have been selected for sensory ability
in the first place). But the special names they learn for colors do help in one
way: They facilitate communication, and, what is more interesting for the
present discussion, they enhance the ability of people to recognize and iden-
tify particular hues from memory. This has been demonstrated by R. W.
Brown and E. H. Lenneberg in an experiment with American college women.2
In one part of their experiment, they established that a series of colors dif-
fered in what they called codability. Highly codable colors were those which
the women named easily and promptly, and for which there was high agree-
ment on names. In the second part of their experiment, they showed that
the codability of a color was significantly related to how well it could be
recognized in a task such as the following: A subject was shown four colors
simultaneously for three seconds; then after a half minute she had to find
these colors in a large chart containing 120 colors systematically arranged.
When the subjects were asked how they performed this task, they reported
that they named the colors to themselves while the colors were exposed, and
then used the names they remembered in finding the colors on the large
chart.
The advantage of words in various sorts of tasks in which perceived im-
pressions have to be “stored” and remembered in some way has been demon-
strated in numerous experiments. Sometimes the words are ones that the
individual has already learned in his language in the normal course of
experience (as in the Brown and Lenneberg experiment); sometimes they
are words or nonsense syllables whose meanings are learned in the initial
phases of the experiment. The superior potency of a word as a carrier of a
sense impression is revealed even when an experiment is designed so that an
equal amount of attention is paid to the stimuli during initial learning. K. H.
Kurtz and C. I. Hovland ? had one group of children circle on a sheet of paper
the words that went with a series of objects being shown to them, while an-
other group circled pictures of these objects. One week later, the first group
of children were better able than the second group to recall or recognize the
objects that had been shown. Also, the possibility of modifying the learning
of subjects by varying the ‘“‘meaning” of words or other verbal responses has
been clearly shown in various experiments. If two different objects (or
stimuli) are assigned the same name by the experimenter, the two objects
are more likely to be responded to in the same way than if the objects are
given different names. Winifred Shepard * found that teaching a child to
call a series of red, orange, and yellow lights by the same nonsense-syllable
2R.W. Brown and E. H. Lenneberg. J. abnorm. soc. Psychol., 1954, 49, 454-462.
3K. H. Kurtz and C. I. Hovland. J. exp. Psychol., 1953, 45, 157-164.
4W.O. Shepard. Child Devt., 1956, 27, 173-178.
Language
and Cognition

95
 button-pushing response to all these
names will cause him to generalize a
only to the red light.
lights even though the original training was
, words assig ned to stimu li can modify an individual’s
If, as appears above
the individual can be deceived by
responses to those stimuli, it is possible that
chang ed in ways that are not com-
these labels, or at least, have his responses
were aware of them. This fact is the basis of
pletely what he might desire if he
prone. S. I. Hayakawa pointed out
certain semantic fallacies to which we are
payment to the unemployed was
many years ago that if a certain kind of
likely to be perceived favorably,
called “social insurance benefits” it was
likely to be perceived unfavorably.
whereas if it was labelled “relief” it was
iment al evide nce on how verbal labels can
Let’s look, however, at exper
sometimes deceive a person.
by L. Carmichael, H. P.
A classic experiment on this problem was done
subjec ts were briefly exposed
Hogan, and A. A. Walter. They found that when
and later asked to reproduce them, the
to the figures shown in Figure 12,
by the labels assign ed to the figures at the
reproductions were influenced
tended to be reproduced
time of original exposure. For example, the ()-C)
“eyeglasses” whereas it might
as something like (7) if it had been labelled
ed “dumbb ells.” Further experi-
be reproduced =_) if it had been labell

Stimulus Word list Il

Word list |
figures

Bottle cS Stirrup

ea (( Letter "C"

a Hat
Beehive

=: Figure 12. Figures used

by Carmichael, Hogan,

Eyeglasses CC) Dumbbells 9”d Walter in their ex-

periment on the influence
of language on percep-
tion. To one group of
Sun subjects, the stimulus fig-
Ship's wheel ures were presented la-
belled with the words of
Word List I, and to an-

Gun == Broom other group of subjects

with the words of Word
List II. (Adapted from
Two Eight L. Carmichael et al. J.
exp. Psychol., 1932, 15,
73-86.)

Language
and Cognition

96
mental analyses of this phenomenon? support the conclusion that the label
presented by the experimenter tends to “channel” the stimulus function of the
figure in the direction of the concept represented by the label; unless the
subject has prolonged opportunity to study the figure, or the delay period is
relatively short, it is principally this “concept” that is remembered, rather
than some direct representation. Indeed, even subjects who are not shown
any verbal label will invent their own labels and their later reproductions
of the figures will often reveal the nature of these labels. It should be noted,
incidentally, that the use of a label, whether by the subject alone or also by
the experimenter implies that the label refers to a concept; thus, the figure
is perceived as being one of a class of similar experiences named by the
concept.
A label is not particularly useful when it does not readily refer to a well-
learned class of experiences. For example, efforts by several experimenters
to teach people to recognize novel visual patterns better by assigning non-
sense syllables to them have not been successful. There is, in fact, a series of
experiments which purport to demonstrate that stimuli can ‘acquire distinc-
tiveness” solely by having verbal responses attached to them. The theory is
that when a verbal tag has been attached to a stimulus, the implicit responses
to that verbal tag enhance the total discriminability of the stimulus in relation
to other stimuli. Actually, there is no good evidence for any such effect. The
experiments show only that words themselves are discriminable to varying
degrees, or that subjects make varying use of words in mediating discrimina-
tions; these conclusions have been reached also through other types of
experiments.
An experiment conducted by Kathryn Norcross has sometimes been cited
as evidence for the “acquired distinctiveness of stimuli.” ° She taught chil-
dren the names zim and zam for the faces in one pair, and the names wug
and kos for the faces in another pair. Later she had the children learn a
motor response (pushing a particular button) for each face, and found that
the responses were less easily learned to similarly named faces (zim, zam)
than to dissimilarly named faces (wug, kos). But since her procedure called
for the child to say the correct name for each face before making his motor
response, her experiment may be regarded as showing merely that it is harder
to learn associations to relatively similar verbal stimuli than to dissimilar
verbal stimuli—a conclusion which has been repeatedly demonstrated in verbal
paired-associates experiments. It is difficult to conceive of an experimental
design for demonstrating “acquired distinctiveness of stimuli” that will not
be subject to the criticism that the discrimination is made in response to
words, or more generally, verbal mediation, rather than to characteristics of
the stimuli that are somehow invested in them by the words assigned to them.
The hypothesis that discrimination responses can be made to verbal mediators
is interesting enough in itself, and seems a more reasonable interpretation of
the experimental facts.
People vary in the degree to which they notice and concern themselves
with the various kinds of attributes that characterize the things and events
5D. T. Herman, R. H. Lawless, and R. W. Marshall. Percept. Motor Skills, 1957,
7, Monog. Suppl. 2, 171-186.
6K. J. Norcross. J. exp. Psychol., 1958, 56, 305-309.
Language
and Cognition

97
 ready than the average person
of the environment. A forester will be more
of trees. A machinist would more
to notice differences among various kinds
en right -hand ed and left-handed bolts
readily notice the difference betwe
would. Obviously, these differ-
than the ordinary household do-it-yourselfer
gh learning, but there is evidence
ences in response-tendency come about throu
learning. The very existence of
that language can play a special role in this
or for different values of a dimension
contrasting words for different categories
ries or values , and if a person has to learn to
draws attention to these catego
in his speech community, he
use these words in a way that is acceptable
the corresponding stimuli. The
must of necessity notice and discriminate
among stimuli more notice-
effect of language is thus to make the differences
wise be. Evide nce for this effect comes
able, or salient, than they would other
iment descr ibed on pp. 108-1 09.
from a cross-linguistic exper
role of verbal mediat ors in behavi or is so well attested that
The important
ation, too, for many other-
it can hardly be denied. It supplies a ready explan
as a child mature s. In his early
wise incomprehensible changes in behavior
enviro nment tend to be direct —the outcome
years, the child’s responses to his
n stimuli (or classes
of immediate connections that have been learned betwee
early) and respon ses,
of stimuli, for stimulus generalization occurs very
the child attains concep ts
either by classical or by operant conditioning. As
interna lly, he is able to respon d to the
which he can retain and respond to
For example, at some
environment in an indirect, less immediate manner.
fy the larger of two
point in a child’s development he can be taught to identi
wherea s at an earlier stage
stimuli no matter what their absolute sizes may be,
te sizes. He is presu mably responding
he can respond only in terms of absolu
than in terms of direct percep tions of stimuli.
in terms of a concept rather
s have though t that the change over from direct to mediating-
Some psychologist
s about the time
response behavior takes place at one particular stage, perhap
ge. Exami natio n of the total array
the child begins to solidify his use of langua
ure sugges ts, howeve r, that this change over
of pertinent experimental literat
partic ular concep t is, to be sure, learne d more or
is a gradual process; each
lty and in the time
less once and for all, but concepts vary widely in difficu
they are
required to learn them, and hence also in the age of the child when
ts are not learne d until relativ ely
likely to be learned. In fact, some concep
. If the learni ng of a concep t is accom panie d by the
late in life by some people
ular verbal respon se, the potenc y of the concep t in behav-
learning of a partic
panied
ior is likely to be enhanced; concept learning is more likely to be accom
by overt verbal learning, the older the indivi dual is.
lies in
One other striking effect of the development of verbal mediators
better able to state and test hypothes es. In fact, this
making the individual
to be highly correlate d with mental developm ent as a whole as
trend seems
obtained, ironicall y
measured by mental tests. Interesting evidence for this was
shown that high-IQ children were
enough, in an experiment in which it was
prob-
less successful than children of average IQ in solving certain kinds of
lems in which many irrelevan t stimuli were present.’ Apparent ly, the high-IQ
children were hampered because they spent time developing and testing
hypotheses concerning the irrelevant stimuli, whereas the children of average

7S. F. Osler and G. E. Trautman. J. exp. Psychol., 1961, 62, 9-13.

Language
and Cognition
98
IQ learned which stimuli were relevant by simple associational learning
processes.

eh co ayws<p

We have plenty of evidence for inferring the existence of concepts and other
mediating processes by noting that some such processes must function in
experiments of the sort we have been describing. The question now is, can we
study these processes for themselves? Can we describe them more completely
and get at their fundamental nature?
We shall describe and discuss a number of different methods that have
been devised. Each method involves the eliciting of certain kinds of overt,
objective responses from which reasonable inferences about the nature of
the underlying mediating processes may be drawn. Each method yields inter-
esting and suggestive evidence, but no one method can give the whole answer
we want.

Free Association
The free association ex-
periment has already been introduced in Chapter 4 as a way of showing that
verbal behavior is fairly predictable. The responses given by a person to a
verbal stimulus in the free association experiment are presumably mediated
by the concept this stimulus evokes and hence give evidence of the nature of
this concept. A person who is presented with an ambiguous stimulus like
LIGHT and asked to give a series of associations to it will generally give
either a series like heavy, feather, weight, and so on, or a series like dark,
color, white; he will rarely switch from one concept to the other.
The free association experiment is quite sensitive to the influence of set.
It is perceived by people as a particular kind of task; that is to say, “think-
ing of the first word that comes to mind” means different things to different
people. To young children it often tends to mean “think of the next word you
would use in a sentence,” so that to the stimulus LIGHT a child might give
the response bulb (light-bulb). To adults, it is more likely to mean, “think
of another word in the same part of speech as the stimulus and with a partially
similar meaning.” To some adults, it seems to mean, “think of a word that
contrasts in meaning with the stimulus as much as possible.” The different
sets with which people approach the free association task and the effects of
these sets have not yet been adequately studied. We do, however, have ex-
tensive compilations of the frequency of responses to standard stimulus lists
in representative American populations; they give evidence of the nature of
certain concepts in these populations. Here, for example, is a list of all the
words given in response to the stimulus LIGHT, and their frequencies, in a
sample of 1008 college students:*

norms for responses

8 From W. A. Russell and J. J. Jenkins. The complete Minnesota
noff Word Associatio n Test. University of Minnesota,
to 100 words from the Kent-Rosa
Department of Psychology, 1954.
Language
and Cognition

99
 647 dark shade(s), red, brightness, yellow, heaven,
78 lamp color, weight

30 bright shadow, street, health, green, hair, electric,

25 sun blue, morning, match, clear
23 bulb look, shine, high, out, glass, electricity,
16 heavy, day awaken, star, sky, cheery, read, warm, sunshine,
12 house spot, club, feather, object, brown, electric
see bulb, hand, Edison, sound, hard, study, dark or
10 window heavy, waves, apple, path, air, luminous, beam,
5 switch, soft, time, post, fair, flame, lamb, eyes, dirt,
candle, black, easy, creamy, hurts eyes, love, world, lift,
heat, darkness, desert, snow, lightning, year, truth, ceiling,
white boat, daytime, earth, beach, glow, head
4 night

The frequency distribution somewhat resembles the frequency distribu-

tion of all the words in large general samples of English text; in fact, Davis
Howes ® has pointed out that the summed frequencies of all the words ob-
tained in free association responses are highly correlated with the frequencies
of those words in English in general; he therefore regards the word association
test as little more than a special way of “tapping into” a person’s total
repertoire of responses. This does not explain, however, how the particular
associates of a word get to be selected.
Probably the best way of explaining word association responses is to
assume that a stimulus evokes some part of the concept named by the stim-
ulus, that is, some part of an assemblage of mediating processes. Exactly what
parts of this assemblage will be evoked depends on the set with which the
individual approaches the task and on other factors. The overt response is
then a name elicited by the particular fraction of the mediating response
evoked. In order to give a series of associations to the same stimulus, an
individual may have to change his set slightly, or deliberately “think of”
the stimulus in different ways. He usually exhausts the most immediate and
“obvious” associations in the first few seconds of the response period; the
more remote associations take much longer. Generally, the word association
test is given in such a way as to require only one or a small number of
responses to each stimulus; analysis of data is based on the responses of
large numbers of people. The results are thus not only a function of what
aspects of a given concept are most frequently found in a group, but also of
what “‘sets” are most likely to be adopted by the respondents. If we analyze
the results given above for LIGHT, it would seem that only about 2.2
per cent of the sample interpreted the word as the opposite of Heavy, while
most of the rest interpreted it either as the opposite of dark, or as a noun.
Further, a majority of each group approached the word with an “‘opposite-
ness” set to give words like dark, heavy, and other “contrast” words.
There have been some elaborate attempts to infer from free association
data the concepts underlying the stimuli and responses. All that we can say
here is that the data support the notion that most concepts contain, as it
were, a cluster of attributes which have been experienced by the individual

9D. Howes, J. abnorm. soc. Psychol., 1957, 54, 75-85.

Language
and Cognition
100
in association with the concept. For example, the concept underlying the word
(railroad) train may contain representations of other concepts like fast,
powerful, hard, dangerous, and boring, depending on the experiences of the
individual with trains. Not all of these attributes are criterial, of course; in
fact, few of them are. The task of stating exactly what a ¢érain is, is a far cry
from the task of giving “free associations” to the word ¢érain.
Be that as it may, the results of word association compilations are of great
use in predicting the outcomes of various kinds of experiments involving
words. For example, it is much harder to learn pairs of words selected at
random than it is to learn pairs of words that are frequently given to each
other as associations, or that give the same sets of associations (for example,
MAN and GIRL are seldom given as associates of each other, but in common
yield the association WOMAN). Also, word association results can predict
the outcomes of verbal experiments in transfer of training: If one learns a
pair A-B, it is then easier to learn the pair A-C if B and C are associates
according to free association data.
We have already mentioned (page 41) the use of free association pro-
cedures to define what has been called meaningfulness. The number of
different associations that a stimulus elicits in a person in a certain period
of time, say one minute, gives a measure that, when averaged over a repre-
sentative sample of respondents, will very well predict the ease of learning to
pair that stimulus with another stimulus. This measure of meaningfulness can
be applied either to real words or to nonsense syllables. Meaningfulness,
measured in this way, can be interpreted as an index of the variety of experi-
ences represented in a concept, and learning is facilitated when there are a
variety of associated experiences with which to form connections.
is it
When a subject responds to BUTTERFLY with the word MOTH,
a “direct,” immediat e connecti on between the words, or is
because there is
even evokes an
it because the individual “thinks of” a butterfly, possibly
have no
image of it, and then names an object similar to a butterfly? We
at present, which amounts to saying
good way of answering this question
d on either basis. The most parsimon ious in-
that the data can be explaine
is to assume that all associati ons are direct and immediat e, pure
terpretation
a sort of
responses to stimuli. It would even seem that associations form
of experime nts in verbal learning
network in the mind, and that the results
noting how far away on the network any
are most easily accounted for by
of free associati on data. Operation ally, such an in-
two words are in terms
A more reasonabl e
terpretation may be in order, but it seems a little too pat.
reflect the extent to
interpretation is that the associations between words
of attributes or related
which the underlying concepts share representations
le experime nt of W. A. Russell and L. H.
experiences. Consider the remarkab
that their results could be predicted very well solely
Storms,!° who found
they found was that if
from the norms for free association responses. What
he could later learn a new
4 student first learned a pair like DAX-WAR,
than if he had not earlier learned DAX-
response for DAX, JUSTICE, better re-
associati on norms is this: A highly popular
WAR. The relevance of free
JUSTICE), and a highly
sponse to WAR is PEACE (but almost never
Psychol., 1955, 49, 287-293.
10 W. A. Russell and L. H. Storms. J. exp.
Language
and Cognition
Io!
 popular response to PEACE is JUSTICE. The results tempt one to conclude
that the transfer of learning operates by relying on a chain of association
bonds, WAR — (PEACE) > JUSTICE, without any reference to the mean-
ings of these words. But an alternative hypothesis is possible and has not been
checked: that subjects construct a reason for associating WAR and J US-
TICE on the basis of their meanings (for instance, “‘a just war is one fought
for justice”), and that such constructions are more likely to occur to the
subject in the case of words that the experimenter has paired because the
association nouns allow him to regard their associations as chained.

Osgood’s “Semantic Differential” Technique

Charles Osgood, a_ psy-
chologist who has been one of those chiefly responsible for developing the
theory of the verbal mediator, took the bold step of deciding to get subjects
to report directly on the nature of their concepts. But instead of asking them
an open-ended question like “What is your concept of X?”, he devised an
ingenious method adapted from rating-scale procedures. He asked his sub-
jects to rate their concepts on a series of scales.*?
Let us digress a moment to consider the psychological basis of the process
of rating. Suppose we take any two words at random, say tree and stone, and
ask a group of people to indicate in what respects these concepts differ, as
indeed they must. Among some of the answers we are likely to get are these:
A tree is alive, while a stone is inert; a tree is relatively flexible, a stone is
rigid. That is to say, the mention of any two concepts evokes a series of
perceptual or conceptual dimensions in which they differ. Furthermore,
many of these dimensions are recognized to exist in varying degrees. There-
fore, it is possible to ask a subject to conceive the dimensions as represented
by a straight line and to assign concepts positions on this line. For example,
on the dimension flexible-rigid, I might assign tree and stone positions as
follows:

tree stone

because in my experience stones have been about as rigid as anything can be,
whereas trees tend to be somewhat flexible, but not very much so—not so
much, say, as a rubber band. Note that these are solely probabilistic state-
ments; certainly some kinds of trees are much more flexible than others, and
some kinds of “stone” (sheets of mica, for example) are somewhat flexible.
They merely express my “average” concepts of tree and stone.
It is evident that to get a fully rounded idea of my concepts of tree and
stone, an investigator would have to get ratings of them on a large variety
of dimensions. How many dimensions, indeed, would be needed? In working
out an answer to this question, Osgood collected 50 dimensions named by
different pairs of adjectives and then resorted to the statistical technique of
factor analysis (see page 67) to see whether this list of dimensions, or

11¢. E. Osgood, G. J. Suci, and P. H. Tannenbaum. The measurement of meaning.

Urbana: University of Illinois Press, 1957.

Language

and Cognition

102
scales, could be pared down to a relatively small number of “basic” dimen-
sions. This required the collection of data from a large group of respondents
(as usual, college students) who rated 20 different concepts on the 50 scales.
The somewhat unexpected result was that the 50 dimensions reduced them-
selves to three basic dimensions, or “factors”:

1. Dimensions like good-bad, pleasant-unpleasant, sacred-profane, reduce

themselves to what Osgood called an Evaluation factor, because the ratings
of concepts on all these dimensions tended to be intercorrelated.
2. Similarly, dimensions like strong-weak, large-small, and heavy-light re-
duced themselves to what was called a Potency factor.
3. Finally, dimensions like active-passive, fast-slow, and sharp-dull re-
duced themselves to a basic Activity factor.

Now, it is obvious that these three factors are not the only independent
dimensions by which concepts can be described. For example, stories can be
rated along the dimension serious-humorous, a dimension which is largely in-
dependent of any of the three dimensions listed above. Nevertheless, repeated
studies in a variety of cultures and with different sets of scales and concepts
have rather consistently pointed to these three dimensions as the ones that
apply to concepts most generally and most saliently.’”
This is an important result, one that we must pause to consider rather
thoroughly. It suggests that three important aspects of any “concept” cor-
respond to its positions on the three basic dimensions indicated—how ‘“‘good,”
how “strong,” and how “active” it is perceived to be. Few concepts are
neutral on all three of these scales. The evaluative scale is intimately con-
nected with a basic psychological process: the satisfying, rewarding, or re-
inforcing property of a stimulus (or conversely, the displeasing, punishing, or
nonreinforcing property of a stimulus). Thus, one’s concept of any class of
stimuli includes an assessment of its average reward value—either to oneself
or to the society with which one identifies. The potency scale is probably
connected with our perceptions of the effort that could be exerted on us by a
stimulus or the effort that would be required to resist it; the activity scale
has to do with the rapidity of movement expected of a stimulus-object—a
temporal matter, whereas the potency scale is more connected with space.
Indeed, the three dimensions identified in Osgood’s factor analyses are
global, or as Heinz Werner would say, syncretic—merged together from
several other distinct dimensions. Perhaps this is an artifact of the procedure
of investigation and analysis, but perhaps also it reflects the syncretic char-
acter of some of our thinking. For example, size and weight are both measured
by the potency dimension, perhaps reflecting the fact that it takes children
a long time to realize that size is not necessarily correlated with weight—
that big things, like balloons, can sometimes be quite the opposite of heavy.
A person’s concepts can be regarded as being located in a “semantic space”
of whatever number of dimensions are accepted as fundamental. (Although I
of
have mentioned only three dimensions, there can be others.) The position
be averaged over individual s, to give
concepts in this semantic space can

12 C. EK. Osgood. Amer. Psychol., 1962, 17, 10-28.

Language

and Cognition

103
 in Figure 13. For some
results such as we have attempted to depict graphically
it is useful to measur e (1) the polarity of concepts,
experimental purposes,
c space, and (2)
that is, their distance from the center or origin in the semanti
t their ‘“meanings” are.
the distances between concepts, that is, how differen
“meani ng” of a concept can be in-
It is unlikely, however, that the whole
tial techniq ue, for the class of experie nces
dexed by the semantic differen
cannot be complet ely describ ed in the dimensi onal
represented by a concept
artificiality
terms required by the technique. Note also that there is a certain
s are essentia lly idio-
in averaging results over many individuals. Concept
particul ar experie nces, both verbal
syncratic, dependent on the individual’s
ty between the concept s of differen t individ uals
and nonverbal; any similari

al space
Figure 13. Locations of selected concepts in the semantic differenti
active-pa ssive, and good-bad . Con-
defined by the three scales strong-weak,
are in lower-cas e letters, concepts rated bad are in capitals.
cepts rated good
for 360 words.
(Data of J. J. Jenkins et al. An atlas of semantic profiles
Amer. J. Psychol., 1958, 71, 688-699. )

ACTIVE

engine TORNADO
RAGE ced
FLEA ANGER ke
SCALDING leadership
i birth
irt happy : neogress
si MOSQUITO
sweeping brother pas heave
merica
MAD sex eating car
STEAL CRIMINAL light COP
mother DEVIL
DANGER bright courage God
gs
NASTY TROUBLE kitchen boat money justice
sister eat income ech
eal chure
SIN FRIGHTFUL PAIN see family
brilliant

CROOKED FEVERISH 44 song™€ memory farm Bible

DIVORCE woman Prony bath dowo “piano complete
en BAD lady fe gleaming clean health faith
PUTR i
aby
FRAUD charming graceful color "Ie
; holoy spatah
DISCOMFORT ood
NEUROTIC beautiful arimcccmcod!
GRIEF RANCID+ constant
fragrant
bodice sweet STENCH
rose barn root
STARVING HEARTLESS
WEAK STRON
refined lake house
blue stars
LEPE!
R flower comfort Sunday trees
rose garment peace
LATE flowers stem
SICKNESS SLIME
INFERIOR peter bed
DEFORMED silk
GLOOMY smooth moon
INDIFFERENT
cushion sleep
STAGNANT chair
mild calm
DREARY relaxed table statue
LAGGING

LAZY

PASSIVE
is a coincidence resulting from parallel experiences! One is tempted to call
the semantic differential, instead, an “experiential” differential, since it in-
dexes individuals’ experiences or attitudes as classed into concepts. Groups of
individuals are sometimes widely separated in their concepts; consider, for
instance, how far apart pacifists and West Point graduates would be in their
concept of ARMY.

Other Procedures
in the Study of Verbal Mediating Responses
The free association and
semantic differential techniques are the only means devised thus far that are
simple and convenient enough for cataloging the properties of concepts. But
there are many techniques for studying the role of concepts and their asso-
ciated mediation responses—various kinds of discrimination learning, concept
formation, and verbal transfer experiments, some of which have already
been described. These experiments clearly demonstrate that some kind of
mediating process must intervene between the stimuli presented at the start of
the experiment (the “initiating stimuli”) and the final overt responses made
by the subject (the “terminating” responses).
In Chapter 3 (pp. 35-36) we showed how meaning responses are established
most probably by classical conditioning. This being the case, we can use the
techniques of classical conditioning to investigate the relations between these
meaning, or mediating, responses.
A reasonably convenient procedure is to condition a psychogalvanic reflex
(PGR) to a word, X, by presenting the word several times together with a
fairly loud buzzer sound. It is then possible to measure the degree to which
the presentation of another word, Y, will elicit the same degree of PGR. This
would be a measure, presumably, of the degree to which the mediated responses
underlying the words have aspects in common. Bernard Riess!® used this
technique with children and found that at ages 8 and 11, homonyms pro-
duced a greater degree of generalization than synonyms, whereas this rela-
tion was reversed at ages 14 and 18%.
Arthur and Carolyn Staats have done much recent experimentation around
meas-
the idea that even the noncriterial aspects of a concept, such as those
differenti al, can be condition ed. In one of their experi-
ured by the semantic
as many
ments,!* subjects were told that they would later be asked to recall
ul words. The series
as possible of a series of nonsense syllables and meaningf
and 108 different words, but each of
contained six different nonsense syllables
presented , at random points in the series, with
the six nonsense syllables was
the words paired
a different one of 18 words. There was nothing special about
pen, four, this, sand,
with four of the nonsense syllables—words like with, car,
were words like thief, bitter,
but the words paired with XH, for example,
with YOF were words like beauty,
ugly, sad, worthless, sour, and those paired
smart. After the series was presented , subjects
win, gift, sweet, honest, and
the nonsense syllables on semantic differenti al scales. It
were asked to rate
rated significan tly dif-
was found that the syllables XEH and YOF were

13 B. F. Riess. J. exp. Psychol., 1946, 36, 143-152.

1957, 54, 74-80.
14 C. K, Staats and A. W. Staats. J. exp. Psychol.,
Language
and Cognition
105
 ferently, and it could be inferred that a positive evaluative meaning had been
conditioned to YOF and a negative one to XEH. Most of the subjects were
unaware that there was any special relation between the nonsense syllables
and the words, and results from those few who did become aware of the
relation were not used. This experiment suggests that the meaning responses
measured by the semantic differential can be conditioned to new stimuli. The
experiment also shows, incidentally, how the conditioning of meaning re-
sponses plays a role in the development of a concept—for in this experiment
subjects were forming concepts around the nonsense syllables YOF and XEH.

ISTIC-RELATIVITY HYPOTHESIS

Now that we know something about the possible effects of language re-
sponses on thinking, we are in a better position to consider a formidable and
persistent question that has been raised many times by philosophers and
others interested in fundamental issues of human life. This question is the
following: Does the structure and lexicon of the language we happen to speak
affect our perceptions of the world and our dealings with it in any way
that would be different if we happened to speak another language? This
question has apparently arisen because languages obviously differ among
themselves in many ways beyond the mere fact that different sounds are used
to express the same meaning. The concepts symbolized by the forms, form-
classes, and constructions of any one language do not always have exact
counterparts in other languages; some would affirm that they never have
exact counterparts in other languages. Anybody who has tried to make a really
faithful translation from one language to another becomes painfully aware of
these differences. If this is so, it would appear to follow that the thinking
processes of the speakers of one language are not the same as the thinking
processes of the speakers of any other language. Indeed, it would appear that
a bilingual using one language thinks differently from the way he thinks in
his other language.
The idea that the structure of one’s language affects one’s thought proc-
esses may be called the linguistic-relativity hypothesis, because it asserts that
thought is relative to the language in which it is conducted. Sometimes, also,
it is called the linguistic Weltanschauung hypothesis, because it is asserted
that a particular language implies a unique “world-view” or perception of
reality. A variant of this idea was suggested by German philosophers in the
nineteenth century who claimed that Aristotle’s logic would have been very
different if he had been a Mexican. The most articulate modern spokesman
for the theory of linguistic relativity was the linguist Benjamin Lee Whorf,
and in one of the best statements of his point of view he wrote:

The background linguistic system (in other words, the grammar) of each language
is not merely a reproducing instrument for voicing ideas but rather is itself the
shaper of ideas, the program and guide for the individual’s mental activity, for his
analysis of impressions, for his synthesis of his mental stock in trade. Formulation
of ideas is not an independent process, strictly rational in the old sense, but is part
of a particular grammar and differs, from slightly to greatly, as between different
grammars. We dissect nature along lines laid down by our native languages. The
Language
and Cognition

106
categories and types that we isolate from the world of phenomena we do not find
there because they stare every observer in the face; on the contrary, the world is
presented in a kaleidoscopic flux of impressions which has to be organized by our
minds—and this means largely by the linguistic systems in our minds. We cut nature
up, organize it into concepts, and ascribe significances as we do, largely because we
are parties to an agreement to organize it in this way—an agreement that holds
through our speech community and is codified in the patterns of our language. The
agreement is, of course, an implicit and unstated one, BUT ITS TERMS ARE ABSO-
LUTELY OBLIGATORY; we cannot talk at all except by subscribing to the organization
and classification of data which the agreement decrees.1°

Before we can appraise the worth of the linguistic-relativity hypothesis,

we should take a good look to see whether languages differ in their meaning
structures as much as is claimed. Unfortunately, few systematic attempts
have been made to compare languages in this respect, and it is a very difficult
task in any case. It does little good to use a bilingual dictionary to find
semantic and structural differences between languages, because it is the dic-
tionary-maker’s task, in this case, to identify correspondences, not differences,
between the languages it deals with. Most of the differences between lan-
guages which have been cited in support of the linguistic-relativity hypothesis
are striking, but possibly quite isolated cases, and it is always difficult to tell
exactly what meanings are involved in each case. According to Whorf,
in Shawnee, an Indian language, the expressions for “I push his head back”
and “I dropped it in the water and it floated (bobbed back)” use the same
basic verb form meaning “occurrence of a condition of force and reaction,
per-
pressure back, recoil” (see Figure 14). Whorf infers that the Shawnee
ceives these two situations as highly similar, whereas the English speaker
regards them as quite different. But how can we be sure? Are the English
Finally,
phrases given here adequate translations of the Shawnee expression?
use the same verb form to refer to what are rather
just because the Shawnees
actions, are we to infer that a similar mental process occurs
different physical
a single
in each case? We could doubtless find instances in English where
For example, is breaking a fast
verb form is used in highly different situations.
a stick? And are we perhaps not dealing
in any real sense similar to breaking
would be rarely
with “dead metaphors”—as in the term breakfast, which
English speakers as a real breaking of a fast? When we cite
regarded by
differences in the mental
differences between languages as evidence for
that this is really no evidence
processes of their speakers, we must realize
possibility of such differences in cognition that
at all; it merely points to the
appropriate investigation . That is, in order to estab-
might be confirmed by
in cognition, we must
lish that language differences correspond to differences
cognition or other forms of
gather independent evidence about differences in
nonlinguistic behavior.
Brown and Lenneberg
In an earlier section we described the finding of
and remember colors was
that the ability of speakers of English to recognize
. Some colors were found
related to the “codability” of those colors in English
name them promptly and with
to be highly codable—that is, subjects could

(J. B. Carroll, ed.) Cambridge and

reality.
15 B. L. Whorf. Language, thought, and
M.1.T. -Wiley , 1956, pp. 212-214 .
New York:
Language
and Cognition
107
 Original
force ia tee:
center + -tepe- + -n- == + eq = ni-kwaskwi-tepe-n-a

locusathead byhandaction causeto | push his head back.

another

kwaskwi (or kwask)

Condition of force xb a
and reaction, pressure
back, recoil
+ -ho- + -to- — ni-kwask-ho-to
eas locus at cause to I drop it in water and |
eae water surface the inanimate it floats (bobs back).

Figure 14. The English sentences I push his head back and I drop it in water |
and it floats are unlike. But in Shawnee the corresponding statements are
closely similar, emphasizing the fact that analysis of nature and classification
of events as like or in the same category (logic) are governed by grammar. —
(From B. L. Whorf. Language, thought, and reality (J. B. Carroll, ed.).
Cambridge-New York: M.1.T.-Wiley, 1956, p. 235.) |

high agreement; others were difficult to name. Lenneberg and an anthro-

pologist, J. M. Roberts, investigated whether the same phenomenon would —
hold for speakers of the Zuni language, an American Indian language spoken ~
in New Mexico. Striking differences between English and Zuni appeared;
colors that were highly codable in English were not always highly codable —
in Zuni, and vice versa. And furthermore, Zuni speakers had more trouble in
recognizing and remembering colors that were poorly coded in Zuni but well
coded in English, and conversely for color ranges better coded in Zuni than in —
English. This experiment, published in 1956, was one of the first to give strong
support to the linguistic-relativity hypothesis, although in the limited area
of color perception.'®
Another experiment done with American Indian languages suggests a fur-
ther way in which language structure may influence behavior. Joseph Casa- _
grande worked with groups of Navaho children living on the reservation.
Some spoke only Navaho; others spoke only English, having been brought up
in English-speaking Navaho families. Now, the Navaho language has the in-
teresting characteristic that certain verbs of handling—the- Navaho equiv-
alents of to pick up, to drop, to hold in the hand and so on—require special
forms depending on what kind of thing is being handled. There are eleven
different forms, one for round spherical objects, one for round thin things,
one for long flexible objects, and so forth, and the Navaho child has to learn

16 —. H. Lenneberg and J. M. Roberts. Indiana U. Pubs. in Anthrop. and Linguistics,

Memoir 13, 1956.

Language
and Cognition
108
these in order to speak his language grammatically. Casagrande determined,
first, that very young Navaho-speaking children did in fact know and use
these forms correctly. He then compared the Navaho- and English-speaking
children, matched for age, with respect to how often they used shape, form,
or material as a basis for sorting objects, rather than color; he used sorting
tasks that are usually performed by very young children on the basis of color.
He found that Navaho-speaking children had a tendency to perform the
sorting task on the basis of form at distinctly younger ages than the English-
speaking children. Apparently, the fact that Navaho grammar requires the
child to pay attention to the shapes, forms, and materials of things makes
him more likely to guide his behavior on the basis of this aspect of his environ-
ment. Language, however, is not the only influence that can produce this
result: Middle-class English-speaking children in metropolitan Boston per-
formed the sorting task in about the same manner as the Navaho-speaking
and
children, probably because of their abundant experience with shapes
forms in the toys they had played with."
ical
These experiments hardly touch upon ““world-views” or philosoph
speakers of Zuni, Navaho, and English;
orientations that might differentiate
s impose
there have been as yet no convincing demonstrations that language
ical orientati ons. It is, in fact, difficult to find aspects of
different philosoph
kind. What we do
language structure that would suggest differences of this
s they re-
find is differences among languages with respect to the categorie
to, and there is some promise that
quire their speakers to pay attention
confirm the relevance of these categorie s in directing
further research may
contrast to the
behavior. For example, is it significant that in English, in
to indicate the sex of a
situation in many other languages, we are forced
she)? Does this mean that we are
person we refer to with a pronoun (he,
author of a scientific article is a man or a
overanxious about whether the
woman?
can help or hinder
The question of whether the structure of a language
d. If we could find cases
problem-solving has not yet been definitively answere
a concept where another does not, we
in which one language has a code for
s of the first languag e would be more successful
might expect that speaker
ental paradi gm exempli fied
than speakers in the other, according to the experim
of Spiker, Gerjuoy , and
by Cofer’s finding (p. 85) or the experiment the
have been frustra ted by
Shepard (page 94). Efforts to find such cases con-
have develop ed ways of coding the
fact that apparently all languages
the environment that do not
cepts required for the ordinary transactions in
y all languages, even those
involve advanced science. For example, probabl
“middle-sized,” and in a con-
of primitive societies, have ways of saying
adult speakers of various languages
cept-attainment task we would not expect
in their ability to attain the concept
to differ, other things being equal,
ed” would not necessa rily have to correspond
“middle-sized.” (“Middle-siz
with a single word.)
seem to outweigh the differences.
In fact, the similarities among languages
differ ential techniques are used to
Even when free association and semantic
gs in
In E. E. Maccoby et al. (Eds.). Readin
17 J. B. Carroll and J. B. Casagrande. n, 1958, pp. 18-31.
Holt, Rineha rt and Winsto
social psychology (3rd ed.). New York:
Language
and Cognition
109
 index differences between the concepts of speakers of different languages, the
results are similar; differences can usually be interpreted as due to known
cultural factors. Semantic differential research in a variety of cultures has
shown that the same most salient factors show up: evaluation, potency, and
activity. The positions of concepts in this semantic space are generally the
same, and when they are not, explanations of the differences are usually
readily at hand. For example, rain is a relatively unpleasant, depressing con-
cept for speakers of English in an American college community, but it is quite
“good” and “pleasant” for the Hopi Indians, for whom rainfall is scarce but
essential.
To sum things up, the linguistic-relativity hypothesis has thus far received
very little convincing support. Our best guess at present is that the effects of
language structure will be found to be limited and localized.
This is not to deny that in learning a second language, a person will often
be forced to channel his ways of expressing ideas differently. In Russian, “I
had a book” is expressed as “to me was book.” The English verb to be
corresponds to either of two Spanish verbs, ser and estar, depending upon
whether a state of affairs is conceived as relatively permanent, or temporary:
“My father is a doctor’ uses ser, while ‘The door is open” uses estar.
Spanish thus makes a distinction which English does not ordinarily make. But
the contrast of temporary-permanent is equally available to speakers of
English and Spanish when it is truly critical.

LANGUAGE AND DUGHT; A FINAL LOOK

From an early age, human beings develop internal processes that represent
sensations and perceptions in such a way that they can be stored in memory
and later brought into consciousness and manipulated in the absence of the
stimuli that originally evoked them. Human beings can be aware of and
respond to these internal processes, and when they learn language they are
likely to call them by such terms as “thinking,” “imagination,” “imagery,”
“ideas,” “concepts,” “beliefs,” and so on,
The child normally grows up in a social environment in which a particular
language is in use among those who people that environment. This language
exhibits a relationship to the internal processes of these language users in
their own transactions with their environment.
In the early stage of language learning, the child’s own preverbal internal
processes are conditioned to the symbols used by others in his environment,
but as the child assimilates the structure of his language, his internal proc-
esses become more and more like those of the speech community as a whole,
at least insofar as these internal processes are represented in a language.
Thinking is the conscious or unconscious manipulation of internal proc-
esses for oneself, usually in some particular direction such as the solution of
a problem. Communication, whether through language or through other
means (such as music or painting), is behavior in which the initiator of the
communication seeks (whether successfully or not) to arouse certain internal
processes in the recipient of the communication and possibly to secure certain
overt responses on his part.
Language
and Cognition
IIo
 Language symbols—or, rather, the internal processes that underlie given
language symbols for the individual—may figure prominently in thinking and
often determine its direction. The concepts named by language symbols—
that is, verbal mediating processes—are “tools” of thought in these two
senses: (1) They provide at least some of the internal stimuli and stimulus-
producing responses that carry forward the sequences of events from the
external stimuli initiating the process to the overt responses terminating it.
And (2) they represent organizations of internal processes (acquired through
learning or past experiences) that are potentially critical in determining
whether a given sequence of thought will eventuate in successful or rewarded
overt response. The possession of particular concepts acquired through past
experience is a major factor in the solution of problems or the performance
of tasks; indeed, the teaching of such concepts is one of the major functions
of education. These concepts are usually, though not always, coded lin-
guistically; some of the most important concepts for the solution of problems
—concepts of identity, similarity, comparison of magnitudes, spatial position,
temporal sequence, causation, and the like—are coded in the lexical and
grammatical structure of a language. Nevertheless, many intellectual tasks
can be performed without the use of linguistic codes.
Do specific language codes have an influence on the thinking processes?
In principle, they can, if the above line of argumentation is accepted, and
in certain instances they do. But it is unlikely that speakers of different lan-
guages have, by virtue of the languages they speak, different “world views,” Bs

or different degrees of capacity to solve certain problems. There are more

similarities than differences in the way language codes symbolize concepts,
because these concepts are the result of the transactions of human societies with
a physical and social environment that has many uniformities over the world.
Even if there are differences, the basic intelligence of man is usually suffi-
cient to overcome them.

Language
and Cognition
IIt
 Selected Readings
General, and Chapter 1
Brown, R. Words and things. Glencoe, Ill.: Free Press (Macmillan), 1958.
Carroll, J. B. The study of language. Cambridge: Harvard University |
Press, 1953.
Miller, G. A. Language and communication. New York: McGraw-Hill,
1951.
Saporta, S. (ed.). Psycholinguistics: a book of readings. New York: Holt, —
Rinehart, and Winston, 1961.

Chapter 2
Bloomfield, M. W., and L. Newmark. A linguistic introduction to the his-
tory of English. New York: Knopf, 1963.
Francis, W. N. The structure of American English. New York: Ronald, -
1958.
Gleason, H. A., Jr. An introduction to descriptive linguistics. New York:
Holt, Rinehart, and Winston, 1961. (Revised edition.)
Greenberg, J. H. (ed.). Universals of language. Cambridge: M.I.T. Press, —
1963.
Ullmann, S. Semantics: an introduction to the science of meaning. New
York: Barnes and Noble, 1962.

Chapter 3
Church, J. Language and the discovery of reality. New York: Random
House, 1961.
Cofer, C. N., and B. S. Musgrave (eds.). Verbal behavior and learning:
problems and processes. New York: McGraw-Hill, 1963.
Skinner, B. F. Verbal behavior. New York: Appleton-Century-Crofts,
1957.

II2
 Chapter 4
Garner, W. R. Uncertainty and structure as psychological concepts. New
York: Wiley, 1962.
Miller, G. A., E. Galanter, and K. H. Pribram. Plans and the structure of
behavior. New York: Holt, Rinehart, and Winston, 1960.
Mowrer, O. H. Learning theory and the symbolic process. New York:
Wiley, 1960.
Penfield, W., and L. Roberts. Speech and brain-mechanisms. Princeton:
Princeton University Press, 1959.

Chapter 5
Myklebust, H. R. The psychology of deafness. New York: Grune and
Stratton, 1960.
Osgood, C. E., and M. S. Miron (eds.). Approaches to the study of
aphasia. Urbana: University of Illinois Press, 1963.

Chapter 6
Bruner, J. S., J. J. Goodnow, and G. A. Austin. A study of thinking. New
York: Wiley, 1956.
Humphrey, G. Thinking: an introduction to its experimental psychology.
New York: Wiley, 1951.
Hunt, E. B. Concept learning: an information processing problem. New
York: Wiley, 1962.
Hunt, J. McV. Intelligence and experience. New York: Ronald, 1961.

Chapter 7
ent of
Osgood, C. E., G. J. Suci, and P. H. Tannenbaum. The measurem
meaning. Urbana: University of Illinois Press, 1957.
New York:
Staats, A. W., and C. K. Staats. Complex human behavior.
Holt, Rinehart, and Winston, 1963.
Technology
Whorf, B. L. Language, thought, and reality, Cambridge:
Press, and New York: Wiley, 1956. (Edited by J. B. Carroll.)

Selected
Readings
113
 Index
A Behavior, language, 22
Bilabial consonants, 11, 17
Abilities, 67
Bilingualism, 42, 106
Abnormal mental states, 71-72
Abstractions, 70
Bit, of information, 55
Acquired distinctiveness, 59, 97
Bound forms, 18
Bouthilet, Lorraine, 82
Adjectivals, 24, 92
Brain, speech areas of, 45-46
Adverbials, 24, 92
Braine, Martin D. S., 32n
Affricate consonants, 17
Broca, Paul, 45
Age, and vocabulary, 69
Allophones, 17
Brown, Roger W., 33, 95, 107
Bruner, Jerome, 83
Allport, Gordon W., 3
Alphabet, International Phonetic, 10, 15
= C
Alveolar consonants, 17
Alveopalatal consonants, 17 Calls, as attention-getters, 24
Anglo-Saxon, 19 Carmichael, Leonard, 96
Animal language, 30 Carroll, John B., 49n, 109n
Anomia, 70 Carton, Aaron S., 49n
Anxiety, and speech, 47 Casagrande, Joseph B., 108-109
Aphasia, 54, 70-71 Categories, imposed by languages, 109
Arabic, 13, 15 Cherry, Colin, 73
Aristotle, 106 Child:
Arithmetic, 4 free association in, 48
Articulation: deaf, 31
ability, 68 language learning, 23, 30-33
point of, 11, 17 learning to read, 61-65
tests, 66-67 Chinese, 16, 90
type of, 11, 17 Chomsky, Noam, 23-25
Artificial languages, 90 Classification, of forms, 10
Aspiration (phonetics), 11, 17 “Cloze” procedure, 64
Association: Codability, 95, 107
controlled, 81 Cofer, Charles N., 85, 109
free, 33, 48, 66, 99-102, 105, 109 Cognition:
verbal, 71 and thinking, 75-88
Atmosphere effect, 87—88 and language, 89-111
Attributes, criterial, 40-41, 82, 101 Cognitive behavior, 29
Aufgabe, 80 Cognitive dissonance, 80
Austin, George A., 83 Cognitive styles, 83
Azrin, Nathan H., 72n College Entrance Examination Board, 68
Color vocabulary, 28
B Communication, 1—7, 29, 110:
Babbling, 31, 32 systems of, 5, 30
Basic English, 90 Composition, skill in, 68

Tid.
Comprehensibility, 60 Expression system, 6, 9, 29, 33, 40, 51
Concept formation, unconscious, 82, 91 Expression-types, 24-25, 29, 51
Concepts, 4, 5, 40, 41, 70, 73, 78, 80, 104, Expressive movements, 3
106, 109, 111: F
attainment of, 81-84
and form-classes, 92 Factor analysis, 67, 102
and language, 90-93 Factors of language ability, 67-68
and problem-solving, 84-86 Feedback, 45, 77
Concrete operational thinking, 78-79 Festinger, Leon, 80
Conditioning: Finnish, 62
classical, 34-36, 39, 98, 105 Flanagan, B., 72n
operant, 36, 39, 81, 98 Fletcher, Harvey, 12
Conjunctives, 92 Fluency of expression, 68
Connotative meanings, 41, 88 Foreign language aptitude tests, 66, 68
Conservation of number, 79 Foreign languages, learning of, 42
Consonants, 11-12, 14, 17 Form, 10, 13, 18-19, 22, 27, 34, 91
Constructions, 10, 19-22, 27, 28, 50, 91 Formal propositional thinking, 78, 79
Content system, 6, 9, 26-29, 33 Formants, 12
Context: Form classes, 20, 28, 33, 48, 50, 61:
and meaning, 26 learning of, 33
and redundancy, 56 Formulation,verbal, 82
Correlation, concept of, 92 Frequency of language units, 52-54, 63,
Covert behavior, 76-78 100
Cramer, H. Leslie, 13 Freud, Sigmund, 71
Criterial attributes, 40-41, 82, 101 Fricative consonants, 17
Friedman, Elizabeth A., 53
Fries, Charles C., 23-24
Deaf, language learning of, 31, 73-74, 77 Functional fixedness, 85
Declarative form, 51 Functions of language, 4
Decoding, 71 Function words, 27
Definitions, dictionary, 27 Furth, Hans, 73
Delayed auditory feedback, 45
Deletion transformation, 25 G
Denotations, 26, 28 Garner, Wendell R., 55n
Dental consonants, 17 Gelfand, Sidney, 85
Determiner, 20 Gerjuoy, Irma R., 94, 109
Determining tendency, 80 Gestures, 3
Dialect, 15 Goldiamond, Israel, 72m, 76n
Dictionary, 27, 28, 107 Goldman-Eisler, Frieda, 47
Diphthong, 14 Goldstein, Kurt, 70
Discrimination: Goodnow, Jacqueline, 83
sounds, 59 Grammar, 9, 19-21, 23, 27, 29, 52:
stimuli, 37-39 learning of, 32
Distances, semantic, 104 Grammatical “depth,” 52
Distinctive features, of phonemes, 17, 32 Grammatical sensitivity, 68
Double alternation problem, 93, 94 Grapheme, 62-63
Drives, 80 Greenberg, Joseph, 29n
Greetings, 24
Griffith, Belver C., 59n
Eaton, Helen, 91
Echoic responses, 36, 38
H
Edfeldt, A. W., 77n Hall, Edward T., 3, 43
Education, and concept teaching, 111 Harris, Katherine S., 59n
Emphatic transformation, 21 Hawkins, William F., 76n
Encoding, 71 Hayakawa, S. I., 96
English, 13, 19, 20, 21, 24, 62, 90, 108, 109 Heidbreder, Edna, 83
Environment, 72, 111: Heise, George A., 56, 57
and individual differences, 69 Hemispherectomy, 45
Esperanto, 90 Henle, Mary, 86
Euler circles, 87 Heredity, 30-31, 69, 72
Exclamations, 24 Herman, David T., 97n
Hesitation phenomena, 47
Exclamatory transformation, 21, 25
Existence-assertions, 24, 51, 60 Hoffman, Howard S., 59n
Hogan, H. P., 96
Experience, classes of, 81, 91, 105
Hopi Indians, 110
Expressions:
22 Hovland, Carl I., 95
constructionally ambiguous,
Howes, Davis, 49, 54, 64, 71, 100
nonsentential, 24

Index
TES
 Hypotheses, in concept-formation experi- Lenneberg, Eric H., 95, 107, 108
ments, 84 Lexicography, 27
Liberman, Alvin M., 59n
Lichten, W., 56, 57
Ideas, 34 Linguistic relativity, hypothesis of, 106-110
Ideational fluency, 67 Linguistics, descriptive, 8-9:
Imperative form, 51 methodology, 22-26, 29
Imperative transformation, 21, 25, 61 Linguistic units, 18, 24, 27, 59
Individual differences, 47, 65 Linking verb, 24, 51
Informant, 22 Lipreading, 74
Information, 4: Logic and thinking, 86-88
bit of, 55 Lorge, Irving, 53
theory, 54-58
M
Inner speech, 76, 77
Instruction, verbal, 94
Maier, Norman, 85, 86
Mand (response), 37-38
Intelligence, 111:
Marshall, R. W., 97n
tests, 66
Meaning, 6, 9, 16, 24, 78, 88, 91, 102:
Intelligibility, of speech, 58-59
analysis of content system, 26-28
Intention, 42
connotative, 41
International Phonetic Alphabet, 10, 15
Interpretation, 42
denotative, 40-41
as psychological problem, 33-42
Interrogative transformation, 21, 25, 61
situational, 41 -42
Intonation, 16, 29, 34
Meaningfulness, 41, 100
Intransitive verb, 24
Introspection, 2, 34
Mediating processes, 36, 40, 97-106, 111
Mednick, Sarnoff A., 35n, 36n, 38n, 83n
Invariants, 40
Memory:
Italian, 28
aided by verbal learning, 95
J storage capacity of, 52
Mental retardation, 69-70
Jakobson, Roman, 32
Methodology, of linguistics, 22-26, 29
James, William, 82
Miller, George A., 50, 52, 53, 56-58
Japanese, 62
Miller, Neal, 34
Jenkins, James J., 48n, 71, 99n, 104
Minimal pairs, 14-15, 59
Judson, Abe J., 85
Miron, Murray S., 7n
Juncture, 15, 16, 29
Mitford, Jessica, 88
Morpheme, 19, 26, 27, 32, 60
Morphology, 19, 22, 29
Kees, Weldon, 3
Motivation, of speech, 46, 71
Keller, Helen, 40n
Mowrer, O. Hobart, 61
Kent-Rosanoff list, 48, 99n
Kernel sentence, 25, 50
Kjeldergaard, Paul M., 49n
Names, 94
Kurtz, Kenneth H., 95
Naming facility, 68
Naming stage, 32
L
Nasal consonants, 17
Labels, verbal, 96, 97 Natural languages, 3, 90
Labiodental consonants, 17 Navaho language, 108, 109
Language: Negative transformation, 21, 25
in animals, 30 Neurosis, effect on language, 71
and cognition, 89-111 Newman, Edwin B., 53
definition, 3 Noble, Clyde E., 41
functions of, 4 Noise, 59
learning of, 23, 30-43 Nominals, 24, 92
primitive, 10 Nonsentential expressions, 51
as sign system, 8-28 Norcross, Kathryn, 97
silent, 3 Noun, 9, 20:
statistical structure of, 52-58 count, 33
written, 47 mass, 33
Language units, frequency of, 52-54, 63, Null-transformation, 25, 61
100 Number system, 4
Latin, 19, 28
Lawless, R. H., 97n (o}
Learning: Object, of transitive verb, 24
discrimination, 31 Ogden, C. K., 91
formal, informal, and technical, 43 Oléron, Pierre, 73
of language, 23, 30-42, 110 Operant response, 36, 76 (see also Con-
of second language, 42-43 ditioning) :
Leeper, Robert, 82n paradigms, 37

Index
116
Oral speech ability, 68 Recognition, 76
Orders of approximation, 56-58 Redundancy, 56, 59, 60
Orienting reflex, 80 Referents, 33:
Orthographies, 62 referential meaning, 6
Oscilloscope, 12 Réflexion parlée, 93
Osgood, Charles E., 49, 70-71, 102-103 Reinforcement, 16, 36, 38, 43, 46
Osler, Sonia F., 98n Response:
Overt response, 35 echoic, 36, 38
mediating, 36, 40
P texting, 38
Paradigmatic responses, 48 Responses:
Paradigms, for learning of language, 34, 37 overt, 35
Paranomia, 71 paradigmatic and syntagmatic, in free
Passive transformation, 21, 25 association, 48
Pattern playback device, 12 perceptual, 35, 39
Penfield, Wilder, 45—46 Response sentences, 23, 25
Perception, 70: Reversible thinking, 79
of printed words, 63, 76 Rewrite rule, 20, 21, 23, 29, 52
of reality, 106 Riess, Bernard, 105
of sounds, 12 Roberts, J. M., 108
Perceptual invariants, 40, 78, 80 Roberts, Lamar, 46
Perceptual response, 35, 39 Ruesch, Jurgen, 3
Persuasive language, 88 Russell, Wallace A., 48n, 99n, 101
Philosophy, and study of meaning, 33 Russian, 110
Phoneme-grapheme correspondences, 68
Phonemes, 6, 10, 12-17, 19, 22, 29, 59, 74:
Ss
learning of, 32 Saccadic eye movements, 63
segmental, 14, 16, 60, 62 Sayers, B. M., 73
suprasegmental, 15, 16, 62 Schizophrenia, 71
Phonemics, 11 Scholastic Aptitude Test, 68
Phonemic symbols, 15 Schuell, Hilda, 71
Phonetics, 10, 11 Second language learning, 42, 110
Phonics, 63 Semantic differential, 88, 102-105, 109-
Phrase: 110
noun, 20 Semantic fallacies, 96
verb, 20 Semantics (see Meaning)
Phrase-structures, 23 Semantic space, 103, 104, 110
Piaget, Jean, 78, 79 Semivowels, 14, 17
Pitch phonemes, 14-16 Sentence, 20, 21, 41:
Polarity, of concepts, 104 “depth” of grammar, 52
Predicate, of sentence, 20, 24 as discriminative stimuli, 61
Predication, 24, 50-51, 61 situation, sequence, and response sen-
Preoperational stage of thinking, 78 tences, 23
Prepositionals, 92 Sentence completion tests, 66
Prepositions, 27 Sentence types, 24, 51, 60
Printed words, perception of, 63 Sequence sentences, 23, 25
Problem-solving, 80, 84-86, 109 Set, 49, 80, 99
Processes, 22 Shadowing technique, 73
Production, of speech, 73 Shannon, Claude, 54
Programed instruction, 83 Shawnee, 107, 108
Pronoun, 20 Shepard, Winifred O., 94, 95, 109
Psychogalvanic response (PGR), 34, 105 Signs, 33
Psychology, and linguistics, 9, 28-29 Sign system, 5-7, 9
Psychosis, effect on language, 71 Silent language, 3
Situation sentences, 23
Size of vocabulary, 27
Rank-frequency relationship, 53-54
Skinner, B. F., 36, 38, 39
Slips of the tongue, 71
Rate:
Sound spectrograph, 12, 13
of reading, 64-65
of talking, 2, 46, 60 Sounds:
discrimination of, 59
of writing, 2
of language, 10-12
Readability, 64
perception, 12
Reading, 61-65:
readiness tests, 67 Space, semantic, 103, 104, 110
silent, 77 Spanish, 62, 110
Reasoning, 93:
Specific language disability, 69
Spectator behavior, 84
ability, 67
Spectrograph, sound, 12, 13
faulty, 87-88
Index
Te
 Speech, 3: Transfer of training, 101
as indicator of anxiety, 47 Transformations, 21-25, 29, 50, 60-61:
intelligibility, 58-59 learning of, 32-33
multiple causation, 47 Transition, 16
neural control, 45 Transitive verb, 24
parts of, 20, 27 Translation, 106
perception and understanding, 58-61 Transposed object, 21, 25
production and formulation, 45-52 Transposition, 25
rate of, 60 Trautman, Grace E., 98n
Speech community, 3, 8, 34 Turkish, 62
Speed (see Rate) Type (word), 2, 54
Spelling, 18 Type-token ratio, 54
Spiker, Charles C., 94, 109
Staats, Arthur W., 105
Staats, Carolyn K., 105 Understanding, of linguistic sign, 36
Stammering, 72-73 Units, linguistic, 18, 24, 27, 59
Statistical structure of language, 52-58 Universals, in grammar, 28-29
Stein, Gertrude, 64 Usage, standards of, 27
Stop consonants, 12, 17 Utterances, 3, 22-24, 29:
Storage capacity of memory, 52 formulation of, 29, 47, 50, 60
Storms, Lowell H., 101 fragmentary, 25
Strategies, in concept attainment, 83
Stress, 15, 16, 29 Vv
Structure: Velar consonants, 17
of language, 9-11 Venn diagrams, 87
of phoneme systems, 17 Verb, 20
Stuttering, 72-73 Verbal intelligence tests, 66
Subject, of sentence, 20 Verbal knowledge factor, 67
Suci, George J., 102n Verbal mediators, 97, 98
Suffixes, 19 Verbals, 24, 92
Syllogism, 87 Verbal transfer, 105
Symbolism, phonetic, 6, 36 Vernon, Philip E., 3
Symbols: Verplanck, William S., 46
phonemic, 15 Vocabularies, 2, 27, 28, 69, 90:
phonetic, 11 child, 32
Synonyms, 27 sight, 63
Syntactical aphasia, 70 size, 57
Syntagmatic responses, 48 Voiced consonants, 17
Syntax, 19, 29 Voiceless consonants, 11, 17
System: Voicing, 11
communication, 5, 30 Vowel, 11, 12, 14
content, 26-28, 33 Vygotsky, Lev S., 77”
expression, 6, 9, 29, 33, 40, 51
writing, 62 Ww
Walter, A. A., 96
T
Watson, John B., 76
Tachistoscope, 63, 76 Weaver, Warren, 54
Tact (response), 37-38 Werner, Heinz, 103
Tannenbaum, Percy H., 102n Whorf, Benjamin Lee, 1, 106-108
Taylor, Wilson, 64 Word, 2, 18, 27:
Tests, 66-67 intelligibility, 56
Texting response, 38 length, 53
Thinking, 110: Word association, 33, 41, 48
in child, 78-80, 110 Word fluency, 68
and cognition, 75-88 Word frequency, 53, 64
as construction of internal models, 79 Word length, 53
and language, 74, 89-111 ‘“‘World-view,” 106, 111
motivation, 80-81 Writing, 3-4, 47, 62
as motor activity, 77 Wundt, Wilhelm, 2
Thorndike, Edward L., 47, 48, 53
Thought, 4, 29 (see also Cognition, Think- Y
ing) Yngve, Victor, 52
Token, 2, 53-54
Traits, 67 Z
Transcription:
phonemic, 15 Zipf, George K., 53
phonetic, 10 Zuni, 108, 109

Index

118
 Date Due

Printe dinUSA
Cat. No. 23 233
BRODAR:
 65-3614
S
S S

and Thought
J.

Language
Carroll,
C27
0 966€€000 Orc9 €
Nou} pue adenduey ‘||assig UYor }jOUeD
220°S0Id
uojue4 ye ABojouysay jo a6a}|09 ANNS
7
ret
if
ih

ithe
iit Rn
THe

Htet
Tura ras
sist

ati
iE
ti

i
irbite
nit