History of American

Linguistic Thought

Rice University

© 2005
 Foreword

History of American Linguistic Thought came into existence because I

occasionally taught a course entitled Modern Linguistic Theory. As best as I
can recall, I first taught it in 1969-1970. At that time, the course had no
historical perspective, and the material was presented as “what’s happening
now”, more or less. But as time passed, “what’s happening now” wasn’t
happening anymore, and it became “history”. The emphasis of the course
changed. In place of looking at “theories” as objectively as possible in order to
discriminate between the good ones and the less good ones, the goal was to
understand why linguists did (were doing) what they did. In retrospect, some
of the directions seem obviously misguided. And some still practiced, still are,
I think. Such determinations are often a matter of personal experience and
personal taste, and I will try to address some of these issues in the last chapter.
The title History of American Linguistic Thought (chosen in part so that I
can abbreviate it HALT) requires some immediate qualification. First, the
discussion here is limited to the 20th century, and to not all of the 20th
century, at that. The story starts in earnest in the mid-1920’s, but it looks
backward to some occurrences in the 19th century to help understand the
stances taken at this arbitrary beginning point. The story ceases about 1975
with a discussion of Government & Binding. Second, it is a Sketch (at best).
This is not intended to chronicle all varieties and all contributions to the
debate about language. It is not a true history. My intent is to identify what I
believe have been the principle currents in the discussion of language, and in
doing this I necessarily omit mention of the work of many (or most) linguists
in the 20th century. The omission is not to be taken as censure nor as an
implicit judgment of their irrelevance. Nor is it to be taken that all worthwhile
work in linguistics originates in the USA. I just had to choose. In another
place (Davis 1973), I have outlined some of the approaches to language
omitted here, and over the years, other authors have detailed and placed the
work of a variety of linguists in their appropriate contexts.
The purpose of HALT is to help understand why linguistics, the
professional inquiry into language, has done what it has and not something
else. Ultimately, I want to understand why linguistic thought — in the view
put forward here — progressed so little in the 20th century. One of the
arguments of the text will be that although there were protestations of
innovation (and even “revolution”), nothing much changed in the period be-
 -iv-

tween 1926 to 1975, and we have inherited many of those ways of thinking.
Since the 50+ years covered in HALT are the founding period of “modern”
linguistics, I believe that it is important to be familiar with the orientations
toward language that were established then. It is important to reflect on how
we do things now, to understand how much of it is just historical accident, and
to identify what is worth keeping and what, not.
If all these qualifications and amplifications were encoded in the title, it
would be something like Sketch of the History of American Linguistic Thought
from Some Time before 1926 until about 1975 with an Eye towards
Evaluation. Clearly, not as catchy as HALT.

August 25, 2005

Houston, TX
 Chapter 1

An Introduction to Thinking About Language.

Devil’s Advocate: Why a theory?

1. Introduction
There is something in humankind that abhors chaos, the absence of
pattern; and ‘theory’ and ‘science’ are one kind of response to that malease.
So is language a response of that sort to chaotic experience. The sensitivity to
chaos, the cognitive attraction to that which does not fit what we know and to
that which is different, is present in all living organisms as the biologically
universal orienting reflex. The equally universal capacity of intelligence to
habituate provides a way of imposing pattern and of removing chaos from
experience. The attraction which chaos has for all organisms is a concomitant
of intelligence, which exists to create pattern, the antithesis to chaos.
Intelligence ‘factored’ into ‘capacities’:

(i) Sensitivity to environment

(ii) Memory ... requires
(iii) Identity, metaphor, insight, creativity ... requires segmentation by
(iv) Focal attention, awareness ... but leaves the unsegmented for
(v) Automatic processing.

The application of these capacities ... which are not exclusively human ...
creates identities which are recurringly attributed to novel (chaotic because it
is novel) experience and which assimilate the experience to extant pattern.
The precipitate of the engagement of intelligence with experience yields a
‘memory’ of that activity. That residue is knowledge. Language is
knowledge. And science is knowledge. Being able to find my car this
afternoon and drive home is knowledge. But is language science; or is science
language?
Not all knowledge is the same.
Not all knowledge is susceptible to our in(tro)spection. We do not (cannot)
always know/be aware of what we know. Some of what we call language is in
this way overt; but most of language is not overt. It is covert. We can obtain
2 HALT

some information about language simply by asking certain questions of a

person who speaks that language. Other information is not so easily
accessible. “What do you call a four-legged animal than goes ‘oink’?” versus
“What sounds are in your language?”; or “What is the difference between ‘It
was yesterday that I forgot my keys’ and ‘Yesterday I forgot my keys’?”; or
“What is the difference between ‘That was a heavy book you hit him with’
and ‘That’s a heavy book you hit him with’?”; or “‘That’s a friend of my
mother’ versus ‘That’s a friend of my mother’s’?”. Such differences will
cause some difficulty ... in different ways ... for our understanding of
language, and for our theories of language.
The knowledge that constitutes science is overt and it is privileged. When
applied to language, science must make what is covert, overt.1 This overtness
... expressed as testability, and methodologically in the scientific method of
controlled observation ... is the source of the privilege of scientific knowledge.
We assign a higher (or different) value ... greater reliability ... to what we
know as science than we do to what we know as speakers of our own
languages and to what we know that allows us to find our way home in the
afternoon. Science deals with truth. It seems to make no sense to say that a
language can be true or false (or ‘better’ [‘civilized’] or ‘worse’ [‘primitive’])
... language may be ‘mistaken’, perhaps as when my son complained “You
been having that fork in your mouth!” But such manifestations of language are
not ‘false’. Similarly, when I get lost going home, that is inconvenient; I’ve
made a mistake, but again what I did is not false/untrue. But scientific
knowledge which has been shown to be mistaken is false/untrue.
Somehow, language has extruded some portion of itself as science, which
is capable of being constantly overt and its content, constantly falsifiable. The
science of language then turns back upon its origin in an attempt to draw the
remainder from the shadows and to make it, too, overt ... to create a science of
language. The scientific knowledge produced by this activity are the ‘theories’
of this discussion.
Science supports its special stature with its methodology. The public and
replicable techniques by which information is attained allows science to
transcend the immediacy of its experience and to become general or
universal. And to the extent that the information is not contradicted, it is
accepted as true. It is the methodology of science that maintains the
knowledge that is science as different from other knowledge, and as different

1 Some of the terms used in linguistics to label this desired quality have been ‘rigorous’,
‘explicit’, and ‘formal’. And the negative/absence of this has sometimes been called
‘mentalism’, ‘fuzzy linguistics’, or simply ‘not linguistics’.
 Introduction to Thinking about Language 3

from language. So that any knowledge that will be scientific must meet those
(some) methodological standards. To be a scientist you must act like one.
But if the method fails, so the knowledge fails. Whatever it may be, it
becomes not science ... If linguistics (‘the science of language’) is to be a
science, it requires a methodology, and it must work with what all can see.
But if much of language is covert, the information that the science of language
accumulates may not be testable/repeatable under the same conditions. But is
there an overt?2 To find an answer, for example, to the question about the
number of sounds in a language, we must know when two utterances are
repetitions. Consider the question of whether a speaker can ever repeat
exactly the same [the reduction of chaos to pattern again] word? How will we
decide what counts as a repetition?

[ ]
and this pair:
[ ] vs. [ ]

(i) I sprained my [
].

What we can grasp in a scientifically acceptable manner may not be the

whole of language. And if the scientific method guarantees that the patterns of
language will not be touched upon, which shall we discard [must we?]:

the data which are tainted or the status of science?

2. The Search for Pattern

Scientific inquiry is concerned with the search for pattern. To understand
is to see pattern, to reduce new observations/experiences to old. To see them
as distinct instances of knowledge. Kuhn (1970.62-63) cites this case of
anomaly and the reaction to it:
... Bruner and Postman asked experimental subjects to identify on short and
controlled exposure a series of playing cards. Many were normal, but some were

2 Linguists lack the mechanical equivalents of the cyclotron or the Hubble telescope. There
is no tool external to language itself which may stand as semi-neutral witness to the
otherwise ‘covert’. Positron emission tomography (PET) “can give quantitative information
about the function of the nervous system” (Sid Gilman & Sarah Winans Newman. 1987.
Manter and Gatz’s Essentials of Clinical Neuroanatomy and Neurophysiology7.
Philadelphia: F.A. Davis & Company. P. 242), and some now may suggest that PET
technology can fill the observational gap for language. Earlier, the technique/exercise of
phonetics promised to be the observational tool for language.
4 HALT

made anomalous, e.g. a red six of spades and a black four of hearts. Each
experimental run was constituted by the display of a single card to a single subject
in a series of gradually increased exposures. After each exposure the subject was
asked what he had seen, and the run was terminated by two successful correct
identifications.
Even on the shortest exposures many subjects identified most of the cards,
and after a small increase all the subjects identified them all. For the normal cards
these identifications were usually correct, but the anomalous cards were almost
always identified, without apparent hesitation, or puzzlement, as normal. The
black four of hearts might, for example, be identified as the four of either spades
or hearts. Without any awareness of trouble, it was immediately fitted to one of
the conceptual categories prepared by prior experience. One would not even
like to say that the subjects had seen something different from what they
identified. With a further increase of exposure to the anomalous cards, subjects
did begin to hesitate and to display awareness of anomaly. Exposed, for example,
to the red six of spades, some would say: That’s the six of spades, but there’s
something wrong with it — the black has a red border. Further increase of
exposure resulted in still more hesitation and confusion until finally, and
sometimes quite suddenly, most subjects would produce the correct
identification without hesitation. Moreover, after doing this with two or three
anomalous cards, they would have little difficulty with the others. A few subjects,
however, were never able to make the requisite adjustment of their categories.
Even at forty times the average exposure required to recognize normal cards for
what they were, more than 10 per cent of the anomalous cards were not correctly
identified. And the subjects who failed often experienced acute personal distress
[Emphases mine, PWD]. One of them exclaimed: ‘I can’t make that suit out,
whatever it is. It didn’t even look like a card that time. I don’t know what color it
is now or whether it’s a spade or a heart. I’m not even sure what a spade looks
like. My God!

The search for pattern in this example is personally or naturally motivated;

it is what we do to make sense of our experience. But there may be other areas
of experience in which different motivations for imposing pattern are present.
For practical reasons, the movement of celestial objects may be codified so
that their repetitive patterns can be used in travel by determining location. The
establishment of astronomy results. Or in the interest of reestablishing
property boundaries after they are destroyed in floods, the principles of
geometry and trigonometry are established. Or in order to maintain the correct
performance of sacred texts, a metatext ... a grammar, e.g. ’s and others’
grammars of Sanskrit, and the principles of a linguistics may emerge. But the
search for pattern may also be pursued for no (immediately) practical reason,
e.g. the Greeks’ elaboration of geometry and ‘pure research’.
3. Identity, pattern, and prediction
 Introduction to Thinking about Language 5

Pattern centers upon identity and recurrence. The identity may that of
the same thing on different occasions, e.g. identifying the North Star,
recreating a property line, or in repeating a word. In these cases, the context
for asserting identity is a matrix of time. The time is different while the
substance is the same. The difference between the two experiences is
removed from the perception and located elsewhere; and in order to do that
we have to create time (a ‘now’ and a ‘then’).
The problem of identity will arise in a different way, e.g. in seeing the
Morning Star and the Evening Star as the same entity, Venus. Or in seeing
Korean [s] and [ ] as the same. In addition to different times, there is now
difference in substance between the things equated. After all, a bright light in
the morning is not a bright light in the evening, and a [s] is not a [ ]. The
sameness now lies in the constancy of the context in which these appear. The
sameness is not directly ‘in’ the two experiences, but in their circumstance.
For example, [ ] is joined to a position before front vowels and [s] is not. In
place of internal sameness, the sameness is in the conditioning
environments.3 What matters is not the presence of difference but of the
recognition of systematic patterned difference. This ‘systematicity’ is what
allows us to perceive distinct experience as ‘same’.
The outcome is that pattern exists as a creation, an abstraction. An
actual difference is systematically ignored/removed leaving no difference. The
identity is the link between two experiences which lies in neither, but in our
relation to them. In that the link is in neither experience itself, it lies outside
them and in its own context. In the context of language, this yields naming,
e.g. {book}, that must be in discourse ... that must be in our experience as this
book, a book, some book, any book, etc. What we call ‘language’ — before we
get to it as linguists — operates in terms of ‘abstractions’ of this sort.
Dealing/reacting to experience in this way — whether covertly and
unaware or overtly and self-consciously — effectively allows us to transcend
the moment. We are no longer held to the immediacy of perception; we can
now reference other times and spaces, and other experiences. We gain in
generality. Generality, here, means fewer ‘units’.
A second consequence derives from the systematic and non-random
property of pattern (however we see it):

(1) 2 6 3 (9 4.5 13.5)

(2) 2 4 6 (8 10 12)

3 The distinction between the two contrasting examples is, of course, not discrete and is a
matter of degree. All perceptions of sameness are similar to those of the second sort.
6 HALT

(3) 13 39 5

Looking at the first three numbers in (1), we may be able to guess the next
three (and any others as long as we know what position in the series they
occupy). That is, X/2 yields the number following X if X > the number
preceding X; otherwise, 3X yields the next number. In (2), it is simpler. We
do not need to know where in the series a number falls; given any X the next
is X + 2. But (3) is different. We would be hard put, given just those three
numbers, to guess the fourth, or the fifth. The sameness/pattern of (1) is that
‘abstraction’:

X/2 yields the number after X if X > the number preceding X;

otherwise, 3X yields the next number

There is a superficial difference between examples such as (1) and earlier ones
such as the North Star. Given a visual glimpse of the night sky, we may react
rapidly to the question “Is what you see the North Star?” whereas if given the
number 30.375, it will take us some time to answer whether it is part of the
whole series of (1) or whether it belongs to a different series. 4 Not all series
will require a great amount of time; for example, in (2) we know immediately
that any even number will belong to it. The nature of such abstractions as (1)
and (2) — the sameness which holds them together and allows them to exist,
whether ‘slow’ or ‘fast’ — seems to allow us to predict or calculate possible
experience. And we seem to have grasped them as patterned only when we
can satisfy ourselves that such prediction is possible. If you cannot
predict/calculate whether 30.375 belongs in (1), you have not ‘understood’ it.
The same condition appears in (4):

(4) The dog yelped.

The cat meowed.
His teacher groaned.

Knowing what constitutes the first word, the second, and the third predicts
such utterances as

(5) His cat yelped.

The cat groaned.

4 The answer is ‘yes’; it does belong to the series of (1).

 Introduction to Thinking about Language 7

The sequence Prediction ––> Pattern ––> Understanding is commonly

accepted in scientific practice and in linguistics. If one can ‘predict’, then
there is a ‘pattern’; and if there is a ‘pattern’, we ‘understand’ the
phenomenon.5 But notice that there appears to be little prediction in the
example of the North Star.

4. Theory
To begin to express any experience in a self-conscious way, we require an
inventory with which to represent those perceptions, e.g. numbers off dials,
photographs of the events within a cyclotron, or a phonetic transcription. We
need to fix our experience so that we can make it overt and manipulable.6 For
convenience, we may call this the observational language. In the schema of
abstractions, there are two types: assumed primitives and created definitions.
For our example in (4), Det, Noun, and Verb may be taken as primitives; and
in (1), number will appear as a primitive. They are given. Definitions may
consist of expressions such as ‘Det + Noun + Verb’ or ‘3X’.7 A third type of
element may be necessary to constrain definitions. For example, Det and
Verb may be adjacent, i.e., although there is no *Noun + Verb + Det, there
may be Verb + Det + Noun (e.g. Catch the dog) in English. Further, elements
not be simultaneous: they have only linear relationships. Examples of such
limitations/ constraints are:

(6) The Complex NP Constraint (Radford 1981.218):

“No rule can move any element out of a Complex Noun Phrase Clause (i.e. no
rule can move any constituent X out of the bracketed clause in any structure of
the type ... [NP ... [N – [–S ... X ...] ...]”

(a) You gave up Linguistics –––>

Linguistics you gave up –––.
but
(b) It would be a pity for [NP [you[S to give up LinguisticsX]] –––>

5 Cf. the ubiquity of ‘rule’.

6 This first ‘fixing’ of course can have a determining effect upon what we say about the data,
for it, in fact, determines what the data are. In that way, there can be no such thing as ‘raw’
data; nothing is baked from scratch. It is all somehow been prepared for us.

7 Of course, other primitives are required as well to create these definitions.

8 HALT

For you to give up linguistics would be a pity.

*LinguisticsX for [NP[you [Sto give up ––– ]] would be a pity.
*LinguisticsX would be a pity for [NP[you [S to give up ––– ]]
.
(7) Phonetic similarity among allophones:

(a) E.g. English [h] and [ ]

(b) E.g. Japanese [h] and [f]

(8) ‘Parallel lines never intersect’ where ‘parallel’ is a defined

relationship between two lines perpendicular to some third
line.

This third type of conceptual tool is the axioms. Formally then, a theory is a
set of primitives and any axioms and the definitions that are based upon the
preceding.

5. Dimensions of theories
A theory is usually a theory of something, i.e., it is derived from (or
applied to) some data (experience), although it need not be in order to remain
a theory. Thus, empirical theories are distinguished from non-empirical
ones. An empirical theory is recognized as one that has a relation to data via
the observational language, e.g.

(9) (a) Det + Noun + Verb Theory

(b) Observational language
(c) a performance of (b) Data

As long as a theory is an empirical one, there is the probability that there

will be more than one of them for the same range, or overlapping ranges, of
data. One explanation for the existence of mosquitoes is that they are the
remains of the sun’s son, known as Kank . He was permitted just once to
guide across the sky the boat containing the fire which warms the earth; but in
place of adding wood to the fire gradually as the boat progressed across the
sky, he put it all on at once causing everything exposed to be burnt up. His
father was angry at him and spanked him ... spanked him so hard that he
turned into a cloud of dust and his bones scattered to become today’s
mosquitoes. That, of course, is not our explanation for the mosquito. What
happens when there is more than one understanding of a phenomenon? If we
 Introduction to Thinking about Language 9

decide that the phenomenon is the same, then the theories are in some sense in
competition. And if these are scientific theories, then the issue is what is
truth?8 They must in some way be evaluated so that we can determine, at
least, which is better (if not which is true). The evaluation may appeal to
various measures:

(10) Confirmability: circularity and the possibility of discon-

firmation. Possibility of checking ... overtness. The func-
tions of experimentation.
(11) Degree of fit: Preciseness. E.g. Keplerian celestial mecha-
nics versus Ptolemaic.
(12) Generality: Extension in that the theory makes sense of
a range of data in addition to those observed, of a different
sort. The incorporating, integrating ideal.

Confirmability may contribute to distinguishing between scientific and non-

scientific theories. The ‘myth’ is in principle not confirmable.
Linguistic theory, i.e., theories of language, make possible sets of
statements we call grammars, each appropriate to a specific language:

THEORY
Grammar1 Grammar2

The dog yelped

[run the dog the]
‘The dog ran’

‘The man spoke’

‘The cat walked’

Now in view of Grammar2, the specific linearity of Det + Noun + Verb no

longer holds, but linearity is still present. Each competing theory of language
may be judged on the merits of (i) the grammar(s) it makes possible for each
language and (ii) on the grammar(s) it allows for languages not yet observed.
The first is a matter of (11) ... the degree of fit or accuracy. The second is a

8 In the example of the mosquito, there is no competitive evaluation. The first explanation is
labeled as not scientific (It’s a ‘myth’.), and it does not enter into a comparison with the
biological explanation.
10 HALT

matter of (12), which for theories of language entails a characterization of

possible language. The Phonetic Similarity Constraint of (7) narrows the
conception of language for any theory which contains it. Again, a matter of
(11). It is now a matter of examining languages to determine whether the
theory is supported or contradicted.
It may be possible for two competing theories to be indistinguishable in
terms of (10) - (12). If such were the case, then they are identical with
respect to the data which they are attempting to order. Yet they may still
differ; and if so, then the difference is internal, i.e., with respect to the
primitives, axioms, and definitions themselves. The criterion which is
applicable at this point is:

(13) Simplicity: The theory which can do the same with less is
preferred.

This level of delicacy has not been reached for theories of language and
Simplicity is not yet an independent criterion; but Simplicity plays a role in
conjunction with Degree of Fit and with Generality. For example, pursuit of
the fewest number of phonemes (an example of Simplicity in a grammar) is a
desideratum limited by Degree of Fit.9 Simplicity will co-vary with Generality
so that the more general will be the simpler. Theories may be constructed so
that the more general grammar is also the simpler (manifested as the
‘shorter’).
The idea of evaluation is directed towards overtly and objectively
resolving the question ‘Is it right or is it wrong? Is it true?’ If the link between
Generality and Simplicity can be maintained, then the answer is made
objective. We simply calculate which theory (or grammar) is the simpler, and
that one theory (or grammar) must also, therefore, be the more general, the
better, and the more true. But can a theory still be wrong if it works? We may
decline to submit to evaluation, but admire the product of our theorizing on
other grounds. One way to avoid this evaluation is to claim that the theory is
simply a mechanical tool to aid in the manipulation of the data. This attitude
has been aptly called instrumentalism. Galileo in his de Revolutionibus
Orbium Coelestium espoused this approach:
... it is not necessary that these hypotheses should be true, or even probably; but it is
enough if they provide a calculus which fits the observations ...

9 The retreat from the abstract phonologies illustrated by Chomsky & Halle 1968 (which
were sanctioned by appeals to Simplicity) to more concrete ones (Kiparsky 1968) was
motivated by an appeal to Degree of Fit. The latter began to outweigh the former again.
 Introduction to Thinking about Language 11

The attitude of instrumentalism when it occurs within linguistics has been

called “hocus pocus” (Householder 1952). Such opinions may concern only a
portion of a theory (or be held only temporarily). Sir Isaac Newton, for
example, maintained such a reservation concerning the introduction of gravity
(action at a distance) into mechanics; and a belief in corpuscular mechanisms
and ether continued into the 19th century. While odd within a Euclidean
geometry, within a non-Euclidean one (such as the curved space of Riemann),
gravity becomes a consistent aspect of that space. It is no longer a ‘tool’ to
make the system work; it is part of the system itself. A similar reservation has
been made concerning the morphophoneme in a theory of language (Hockett
1961.42):

It has no status in language, but is evoked by our desire to make cross-stratum

correlations neat.

These partial or temporary opinions of instrumentalism have been

distinguished as descriptivism. The other extreme opinion — which we
assumed above — is generally termed realism and within linguistics (again
Householder 1952), it is recognized as “God’s truth”.
We have now several ways of carrying out the scientific endeavor of
creating a theory. Returning to the initial task of constructing a set of
statements for some interesting range of data, we now find that we make adopt
two contrasting attitudes, and this will introduce the possibility of further
diversity among theories. Our concern with finding the ‘true’
understanding/description of our data make prompt us to try to guarantee the
correct outcome before-hand. The idea is that the correct, most highly
evaluated description will emerge automatically.
One way to attempt this is to take the handling techniques or experimental
manipulations of the data (what one does in the laboratory) and match them
(convert them) into the definitions of the theory. This constrains what a theory
can be, for no notion may appear in the theory if there is no operation which
can be performed to produce it. Such care in the construction of theory has
been most notably characteristic of psychology and of linguistics as practiced
in the United States. The constraint is called operationalism, and in
linguistics it was directed toward the establishment of discovery procedures.
The conservative nature of operationalism also dictated that the process begin
with the most certain data, i.e. phonetics and not meaning. This is what gives
American structuralism of the 1930’s to the 1950’s its bottom ––> up
12 HALT

directionality.
Alternately, in place of trying to define pattern by a set of procedures, we
may assume it, using our best guesses and then weed out the competitors
using the evaluation criteria introduced above. Theories which result from this
practice are explanatory. They take on a top ––> down directionality. The
statements which they generate are called explanations while the operational
theories produce descriptions. In an explanatory theory, the data follow
deductively and categorically from the assumed best-guess theory. Other
possible explanations, in addition to the deductive-categorical, are statistical,
teleological, and historical.
The opposition between operational and explanatory theories magnifies
the variety of scientific practice. A final addition which needs to be mentioned
here is the opposition between taxonomic and nontaxonomic. While these
are characteristics of the component patterns of theories, they may also be
taken as prior attitudes to constrain the theories themselves. A taxonomic
theory is one which is limited to such patterns as those typically found in
biological classification. A nontaxonomic one is not so limited; it may contain
a taxonomy, it but may also exhibit patterns which are not stated in terms of
classes (or categories) and members.

6. Conclusion
All of the identified positions have been espoused at some time or other in
some theory of language. But this variety is not limited to linguists; it is not
solely our doing. Given these attitudes, it can at least be understood (in part)
why there are so many theories for us to deal with ... so many kinds of
linguistic thought.
 Chapter 2

Remarks
on
Hockett’s
“The Changing Intellectual Context of Linguistic Theory”
&
Kuhn’s
The Structure of Scientific Revolutions

1. Introduction
Hockett and Kuhn are both interested in the history of science.
Comparison of the two approaches provides some insight into our own
interest in linguistics and language.

2. Hockett
Hockett (1983) finds several themes which allow one to follow the flow of
intellectual activity in the nineteenth century. Two of these themes, which he
introduces with Pierre-Simon Laplace, are the notions of progress and
determinism. The latter forms the primary criterion for the distinction
between two allocations of phenomena. “There are two categories of science
because there are two fundamentally different kinds of things to be scientific
about” (Hockett 1983:14). The two categories of science are termed

Naturwissenschaft Geisteswissenschaft

Precise characterization of the opposition between the two changed during the
nineteenth century as more data became available (more discoveries were
made, aided in part by improving technology), and as conceptions of the data
were altered. Hockett (1983:20-21) identifies several oppositions as forming
the basis of the two kinds of science:

Physical Mental
Determinate Indeterminate
Synchronic Diachronic
2 HALT

Nomothetic Idiographic

It is only in Naturwissenschaft (or la philosophie naturelle or natural science)

that one can exercise the scientific ideal, and the history of science in the
nineteenth century is one in which the range of Geisteswissenschaft is reduced
and that of Naturwissenschaft is augmented. Geisteswissenschaft (or les
sciences morales) does not permit one to use the methods of Naturwissen-
schaft because the data do not exhibit pattern. They are not determinate, and
therefore they are not predictable (Hockett 1983:14 & 16):

In the realm of nature one can make timeless assertions: the valence of oxygen is
two, always has been, and always will be, whether the oxygen is on someone’s
bloodstream or in an interstellar cloud. But in Geisteswissenschaft there can be no
such generalizations. The way things are in the human world is constantly changed
by the willful actions of people. Therefore only a particularistic approach is possible
— whereupon the Geisteswissenchaften were also called the historical sciences, or
just history ... Perhaps it was not so foolish, after all, to propose that plants, animals,
and languages, all as then conceived, are sufficiently alike to merit assignment to a
single larger category.

But the boundary of this division shifts throughout the nineteenth century
and the character of the boundary is changed as well. In the initial state of
affairs, Geisteswissenschaft appears to have included all the life sciences, and
the contrast was nearly one of organic (Geisteswissenschaft) versus inorganic
(Naturwissenschaft). Geisteswissenschaft was guided the doctrine of by
vitalism (Hockett 1983:13):

The vitalist view had held that “organic” compounds, meaning those found
characteristically and exclusively in organisms, could not be built up out of raw-
materials except under the direction of the posited vital energy.

The first realignment followed from the discovery that organic results can
originate from inorganic sources. The synthesis of urea demonstrated that
such a vitalist view was incorrect, and then “it was demonstrated that the laws
of thermodynamics hold in organisms just as they do in nonliving organisms”
(Hockett 1983:13). The effect was not to change the way of working within
the Geisteswissenchaften; it simply resulted in the establishment of
physiology as a Naturwissenschaft and the removal of some phenomena from
one category to the other, leaving minds and the phenomena associated with
 Hocket & Kuhn 3

them (humans) subject to vitalism (Hockett 1983:14):

... the gulf between living and nonliving seemed narrower, [but] that [gulf] between
human and nonhuman yawned as unbridgeable as ever. It did not matter that human
physiology is much like other physiology. Obviously human beings have bodies,
which behave like matter because that is what they are ... But we also have minds,
and mind is a different sort of substance, obedient perhaps to different laws.

The place of language in the division depended upon whether one saw the
phenomenon as subject to individual will or whether the relevant phenomenon
was beyond the reach of that will. If we look at the data then emerging from
comparative study, it may appear that there are a succession of stages which
— because they seem to show a progression — are determinate in their
behavior. In particularistic observations on the behavior of individuals, the
conclusion must be that there is no determined shape to the data and it is
indeterminate; but in the aggregate and over time, there is determinate
behavior. Hockett (1983:16-17) cites William Dwight Whitney as
exemplifying the first view, and August Schleicher as maintaining the second
... with his proposed diachronic progression of languages from isolating to
agglutinative to inflecting. Ultimately, they represent two complementary and
noncompeting views of the same phenomenon; one does not have to choose
between them. However, it is the viewpoint represented by Schleicher which
first carries linguistics from the domain of Geisteswissenschaft to
Naturwissenschaft with the establishment of laws, which “showed a pervasive
regularity”, and which “seemed not to have any connection with the human
will”, and which were “in a sense, a mass phenomenon, affecting many people
at once” (Hockett 1983:22).
The viewpoint which Whitney represented remained subject to Geistes-
wissenschaft until two additional notions were made prominent and finally
united into one perspective: granularity (Hockett 1983:24-26) and pattern-
ing/arrangement (Hockett 1983:26-29). In the Naturwissenschaften, granu-
larity appeared in the form of molecules in chemistry, as cells in physiology,
and finally as the quanta of light in physics. The granular mode of thinking
was in the air and in linguistics as well; the grammatical tradition spanning
two millennia in which sentences were seen as composed of their parts, and
the longer experience with alphabetic writing systems made the extension of
particles to phonetics a natural one. Patterning in Naturwissenschaften is
identified with “the arrangement of parts” (Hockett 1983:26), and not with the
substance which implements that pattern. Such substance may in this view be
4 HALT

replaced completely without damage to the pattern; “in the course of time
every constituent atom of a person’s body is replaced, but the pattern persists
and continuity of identity is unbroken” (Hockett 1983:26-27). Within
linguistics, we can now see such pattern in syntax — if viewed as an
arrangement of words into hierarchical forms — and in the concept of the
phoneme. Patterns such as these may stand beyond the behavior and will of
individuals forming a constant identity, which the will and vagaries of the
individual may not touch. Pattern in this view is exclusively equated with
‘arrangement’ and ‘structure’, and ‘Gestalt’. Structuralism replaces
vitalism. The possibility of there being another mode in which pattern may be
present is no longer possible (Hockett 1983:29 and 32):

In syntax, Gestalt plays such a crucial role that if one takes it away there is nothing
left –– and this has been so from the very beginnings of the discipline in classical
antiquity ... the structure [pattern, PWD] of a thing, event, or system, if I understand
it aright, is nothing other than the pattern [structure, PWD] it manifests.

In this way, the second viewpoint maintained by Whitney is also transferred

from the realm of Geisteswissenschaft into Naturwissenschaft. Language can
now be interpreted as physical, determinate, synchronic, and nomothetic.

3. Kuhn
Kuhn (1970) proposes a different and more general scenario intended to
allow us to understand the ways in which scientific study is pursued. He
distinguishes several states of such activity:

(i) “Prehistory as a science” or “pre-paradigm period” (Kuhn

1970:21 & 163)
(ii) Paradigmatic science
(iii) Normal science
(iv) Crisis science
(v) Crisis resolution or the return to normal science

As an example of the prescientific condition Kuhn (1970:13) cites the

example of “physical optics”:

... anyone examining a survey of physical optics before Newton may well conclude
that, although the field’s practitioners were scientists, the net result of their activity
was something less than science. Being able to take no common body of belief for
 Hocket & Kuhn 5

granted, each writer on physical optics felt forced to build his field anew from its
foundations. In doing so, his choice of supporting observation and experiment was
relatively free, for there was no standard set of methods or of phenomena that every
optical writer felt forced to employ and explain. Under these circumstances, the
dialogue of the resulting books was often directed as much to the members of other
schools as it was to nature. That pattern is not unfamiliar in a number of creative [i.e.
not scientific, PWD] fields today, nor is it incompatible with significant discovery
and invention.

Recognizing these practitioners in the pre-paradigmatic condition to be

scientists, Kuhn also allows their work the status of ‘theory’. A movement
away from this condition occurs when (Kuhn 1970:10 & 17):

[there is an achievement] ... sufficiently unprecedented to attract an enduring

group of adherents away from competing modes of scientific activity.
Simultaneously, it was sufficiently open-ended [emph. mine, PWD] to leave all
sorts of problems for the redefined group of practitioners to resolve ... To be
accepted as a paradigm, a theory must seem better than its competitors ...

One of the conditions for normal science, then, is a successful evaluation of

one of the competing modes of thought, which becomes thereby the
paradigm for that science. “In its established usage, a paradigm is an
accepted model or pattern, and that aspect of its meaning has enabled me,
lacking a better word, to appropriate ‘paradigm’ here” (Kuhn 1970:23). The
others will lose adherents and fall into disuse. Kuhn (1970:17) sees this
weeding out process as irreversible in that such a field will not revert to the
condition in which many theories are again competing with no one of them in
the ascendancy (Kuhn 1970:17 & 19):

[initial divergences] ... do disappear to a very considerable extent and then

apparently once and for all ... There are always some men who cling to one or
another of the older views, and they are simply read out of the profession, which
thereafter ignores their work.

Sociologically, the practitioners are transformed from “a group ... interested

merely in the study of nature into a profession or, at least, a discipline” (Kuhn
1970:19), recognized by the presence of journals, societies, and a place in
the curriculum. Because there are fewer (or no) competing views, the
practitioner can now “take a paradigm for granted, [and] he need no longer, in
6 HALT

his major works, attempt to build his field anew, starting from first principles
and justifying the use of each concept introduced” (Kuhn 1970:19-20).
Textbooks come into existence.
At the paradigmatic stage, the subject matter becomes more textured; the
perspective provided by the paradigm brings certain questions to the fore and
places others in the background (Kuhn 1970:15):

In the absence of a paradigm or some candidate for paradigm, all of the facts that
could possibly pertain to the development of a given science are likely to seem
equally relevant. As a result, early fact-gathering is a far more nearly random
activity than the one that subsequent scientific development makes familiar.

The emergence of a paradigm does not require that the victorious view be
comprehensive (Kuhn 1970:23 & 24):

Paradigms gain their status because they are more successful than their competitors
in solving a few problems that the group of practitioners has come to recognize as
acute ... Normal science [emph. mine, PWD] consists in the actualization of that
promise, an actualization achieved by extending the knowledge of those facts that
the paradigm displays as particularly revealing, by increasing the extent of the match
between those facts and the paradigm’s predictions, and by further articulation of the
paradigm itself ... Mopping-up operations are what engage most scientists
throughout their careers.

The experimental activity of normal science (opposed to theoretical activity)

centers on three areas:

(i) “that class of facts that the paradigm has shown to be

particularly revealing of the nature of things” (Kuhn
1970:25)
(ii) “those facts that, though often without much intrinsic
interest, can be compared directly with predictions from the
paradigm theory” (Kuhn 1970:26)
(iii) “work undertaken to articulate the paradigm theory,
resolving some of its residual ambiguities and permitting the
solution of problems to which it had previously only drawn
attention. This class proves to be the most important of all ...
More than any other sort of normal research, the problems of
paradigm articulation are simultaneously theoretical and
 Hocket & Kuhn 7

experimental ...” (Kuhn 1970:27)

Normal science consists in the incorporation of additional information and

the ordered elaboration of the paradigm. Puzzles are solved; anomalies are
discovered and made integral (Kuhn 1970:79):

... the puzzles that constitute normal science exist only because no paradigm that
provides a basis for scientific research ever completely resolves all its problems ...
every problem that normal science sees as a puzzle can be seen, from another
viewpoint, as a counterinstance and thus as a source of crisis.

Some residual problems may resist incorporation and produce “a period of

pronounced professional insecurity. As one might expect, that insecurity is
generated by the persistent failure of the puzzles of normal science to come
out as they should. Failure of existing rules is the prelude to a search for new
ones” (Kuhn 1970:67-68). And this creates a crisis. The response has never
yet been the abandonment of the paradigm and a return to the pre-
paradigmatic condition (Kuhn 1970:77 & 79):1

... once it has achieved the status of paradigm, a scientific theory is declared invalid
only if an alternate candidate is available to take its place. No process yet disclosed
by the historical study of scientific development at all resembles the methodological
stereotype of falsification by direct comparison with nature ... The decision to reject
one paradigm is always simultaneously the decision to accept another, and the
judgment leading to that decision involves the comparison of both paradigms with
nature and with each other ... To reject one paradigm without simultaneously
substituting another is to reject science itself. That act reflects not on the paradigm
but on the man. Inevitably he will be seen by his colleagues as ‘the carpenter who
blames his tools’.

When a crisis condition comes to exist, it may be resolved in three ways

(Kuhn 1970:84): it may eventually be reduced within the old paradigm, it
may be set aside for future generations, or it may prompt the emergence of a
new candidate paradigm (Kuhn 1970:80):

1 “Though history is unlikely to record their names, some men have undoubtedly been driven
to desert science because of their inability to tolerate crisis. Like artists, creative scientists
must occasionally be able to live in a world out of joint — elsewhere I have described that
necessity as ‘the essential tension’” (Kuhn 1970:78-79).
8 HALT

... by proliferating versions of the paradigm, crisis loosens the rules of normal
puzzle-solving in ways that ultimately permit a new paradigm to emerge.

By a process similar to the emergence of the original paradigm, it may be

replaced as the dominant one. ‘Has your illness progressed?’ Does science
progress?

To a very great extent the term ‘science’ is reserved for fields that do progress in
obvious ways ... we tend to see as science any field in which progress is marked ...
But nothing that has been or will be said makes it a process of evolution towards
anything. (Kuhn 1970:160,162 & 170)

Pre-paradigmatic science cannot progress since there is no replacement, only

simultaneous competition. When one paradigm becomes dominant, inevitably
to be replaced by another, then there may be a sense of ‘progress’.2 Since one
never returns to a displaced paradigm, the impression is that progression-as-
replacement is also progression-as-improvement. Why, after all, would the
alteration occur if not as a (perceived) improvement? But since the new
paradigm will itself inevitably be replaced (they have always have been), the
‘progress’ is only a local one. The impression of progress is, in this way, a
redundant epiphenomenon.3
There is, however, a gradation in the “confidence in their paradigms”
which appears to differentiate (i) the arts from (ii) “history, philosophy, and

2 Kuhn (1970:161) sees the return (retrogression) of “art” to “primitive models” as the source
of the cleavage between “art” and “science”:

For many centuries, both in antiquity and again in early modern Europe, painting was
regarded as the cumulative discipline. During those years the artist’s goal was assumed
to be representation. Critics and historians, like Pliny and Vasari, then recorded with
veneration the series of inventions from foreshortening through chiaroscuro that had
made possible successively more perfect representations of nature ... even after that
steady exchange [between arts and science as illustrated by Leonardo da Vinci] had
ceased, the term ‘art’ continued to apply as much to technology and the crafts, which
were also seen as progressive, as to painting and sculpture. Only when the latter
unequivocally renounced representation as their goal and began to learn again from
primitive models did the cleavage we now take for granted assume anything like its
present depth.

3 This returns us to Laplace and his two characteristics of science: progression and
determinism. And now it is the former which appears the more important as a criterion of
science, but that ‘progress’ is now different. Hockett (1983:10) portrays it as a conscious
“collective march” having begun at least with the Enlightenment in the 18th century. This is
not the same as the revolutionary, eclipsing progress which Kuhn proposes.
 Hocket & Kuhn 9

the social sciences” and from (iii) the “natural sciences”. Education in these
three fields from (i) to (iii) relies increasingly upon the use of textbooks until
the last stages of instruction because the essential content is confidently
encapsulated in texts. In arts, “the practitioner gains his education by exposure
to the works of other artists”. In the mid-range areas of social science,
textbooks are employed, but “even in these fields the elementary college
course employs parallel readings in the original sources, some of them
‘classics’ of the field, others the contemporary research reports that
practitioners write for each other” (Kuhn 1970:165). In the natural sciences,
“the few [curricula] that do assign supplementary reading in research papers
and monographs restrict such assignments to the most advanced course and to
materials that take up more or less where the available texts leave off” (Kuhn
1970:165). The strength of this “confidence” in or “commitment” (Kuhn
1970:100) to one’s paradigm is evident when it has to be abandoned (Kuhn
1970:151-52):

The transfer of allegiance from paradigm to paradigm is a conversion experience that

cannot be forced. Lifelong resistance, particularly from those whose productive
careers have committed them to an older tradition of normal science, is not a
violation of scientific standards but an index to the nature of scientific research itself.
The source of resistance is the assurance that the older paradigm will ultimately
solve all its problems, that nature can be shoved into the box that the paradigm
provides. Inevitably, at times of revolution, that assurance seems stubborn and
pigheaded as indeed it sometimes becomes. But it is also something more. That same
assurance is what makes normal or puzzle-solving science possible. And it is only
through normal science that the professional community of scientists succeeds, first,
in exploiting the potential scope and precision of the older paradigm and, then in
isolating the difficulty through the study of which a new paradigm may emerge.

4. Conclusion
Using Kuhn’s paradigm for the history of science, linguistics has not just
recently become a science in the twentieth century. 4 It has had a long history
with its own paradigms (sometimes shared with other fields). This blending,
in which several distinct scientific interests can have a common notion, e.g.
the breadth/use of vitalism or structuralism, recurs within linguistics and

4 But consider the age of our own journals, societies, and curricula/departments. Hockett
(1948:566) asserts that “Linguistics is only in its beginnings”, whereas Whorf (1940:232)
describes linguistics as “a very old science” although in “its modern experimental phase ...
[it} could be called one of the newest”.
10 HALT

permits a wide range of activities ... all the while sharing a single paradigm.
Cf. Figure 1 from Southworth & Daswani (1974:8).
Returning to Hockett’s themes in this light, the transition from nineteenth
century to twentieth century science in the fields identified as
Geisteswissenschaft is the transition from the paradigm of vitalism to that of
structuralism (Hockett 1983:33):

...in the middle of the present century there was –– and perhaps still is –– a whole
complicated ‘structuralist’ movement, in fields as diverse as ethnology, literary
criticism, and mathematics some of whose participants proclaim their
methodological indebtedness to Prague.

Figure 1: Linguistics in the 20th century.

This paradigm, for the moment at least, behaves as normal science (Hockett
1983:34):

... I am convinced that the full impact of the Gestalt view has not yet been felt. If we
can learn to take that approach in a consistent way, I believe many of the problems
that beleaguered us in the 1930s, 1940s, and 1950s (and that have been largely
neglected between then and now) will turn out either to be spurious or to have simple
and satisfying solutions.
 Hocket & Kuhn 11

This shift of paradigms from vitalism to structuralism, then, prompts us to see

language in a different way and in different places than before. Hockett
(1983:40) summarizes his view of the reconceptualization of language:

Linguistics Linguistics
as as
Naturwissenschaft Geisteswissenschaft

Action Thought
Social Individual
Practice/Communicative Theory/Rehearsal

And he finally insists upon a physicalism (Hockett 1983:42):

I do mean that, in my view, there have been no developments either in linguistics or

in the scientific world as a whole demanding any major revision of the
Bloomfieldian physicalistic orientation ...

There is no crisis ...

 Chapter 3

Ferdinand de Saussure
Introduction

1. Introduction
If Saussure is responsible for the establishment of a linguistic paradigm, it
is his description of the linguistic sign that is the basis for such a claim. If
there is any single unifying concept among the varying schools and theories

Ferdinand de Saussure (1857-1913)

of the twentieth century, it is the tacit assumption (no one finds it necessary
anymore to cite Saussure as the original source) of the dual, Janus-like nature
of language, facing in one direction towards phonetics/sound and in the other
direction toward meaning/semantics/content, etc. It is his derivation of the
sign and the attendant attributes that has provided unity to linguistic theorizing
over the past eighty-plus years.1

1 I know of only one widely accepted view of language which escapes this generalization. J.
R. Firth and the London School of linguistics elaborated a concept of language from which
2 HALT

2. Orientation
Saussure develops his ideas from examination of a speech act — what
happens when (minimally) two people talk — and ironically perhaps arrives at
a characterization that is completely other.2 Saussure’s schematicization is
presented in Figure 1. It is symmetrical in that both A and B act as speaker

Figure 1: Saussurean representation of a speech act.

or listener indifferently, but asymmetrical in that at a given time the two roles
are complementary. Assuming that A is speaking, several components of this
event can be identified. First, there is a point in A where a ‘concept’ is
associated with a ‘sound image’; that point is delimited from the remainder as
psychological. Second, that portion within A wherein the sound image is
converted into muscular activity (articulation) is physiological; the remainder
is not. Third, that portion of the speech act that consists of the sound itself,
independent of both A and B, is physical. Fourth, that portion within B
wherein the sound is converted back into a sound image is physiological,
lying between the ear and the point described next. Fifth, that point where the
sound image is associated with a concept is finally psychological. Figure 2
adds this partitioning to Figure 1. That portion described as physical is op-
posed to the remainder in both A and B as outer (physical) to inner (physio-
logical and psychological). Those portions described for both A and B where
the ‘association’ is accomplished is contrasted to the remainder as psycholo-
gical to non-psychological. Those portions where A associates a concept with
a sound image, then converts the image into articulation, and finally the

the Saussurean sign is absent. The Prague School, Hjelmslevian glossematics, the work of
Bloomfield and the American Structuralists (or Post-Bloomfieldians), Pikean tagmemics, and
Transformational Generative Grammar in its various forms are all the direct intellectual
inheritors of Saussure.

2 A similar inspection by Bloomfield — in his anecdote of Jack and Jill — begins

analogously, but ends for Bloomfield with very different results (Bloomfield 1933:22-27).
And recall Harris’ (1992) lament for the “ordinary” or “lay language user”.
 Saussure: Introduction 3

resulting physical sound are collected to comprise the active portion of the
chain. The remainder — lying entirely within B — is passive. The psycho-
logically active is finally opposed to the psychologically passive as executive
to receptive.
As it turns out, Saussure finds language (langue) to be completely
other/different from these distinctions and elucidates his concept of language
by opposing its attributes to those he finds in this schematized speech act.
Language is above all a social phenomenon while the speech act of Figure 1

Figure 2: Saussurean speech act partitioned.

is not, and it is from the social property that several additional characteristics
arise.3 Although the speech act would appear to be ‘social’, it is not, given
Saussure’s particular use of the term social, by which he intends the
collectivity of individuals. And language is social to the degree that it is
common to that collectivity. It is the average. It is, as Saussure says, not
complete in an individual, but is identified as the ‘overlap’ that unites
individuals into that collective whole. The speech act, then, is not social
because it is unique and particular. It is individual and thereby opposed to the
collectivity wherein language is found. There is no collective speaking;
“execution is never carried out by the collectivity” (Saussure 1959:13).
The introduction of social allows Saussure then to distinguish language
(langue) from speech or speaking (parole), and it is just the latter which is
represented in Figure 1. It is from this conception of a social language that
additional differentiating properties arise. They are that:

3 Recall Hockett’s (1983) associating with Schleicher the idea that pattern may arise from the
social nature of a phenomenon.
4 HALT

Langue Parole

Nonwillful Willful
Passive Active
Homogeneous Heterogeneous

Both langue and parole may show their respective properties in several ways.

2.1 On the opposition of Active to Passive

Language having been separated from the speech act is passive, while the
latter is active. Parole is the use, the drawing upon the passive — the
collective sum or average. Language is passive ontogenetically, as well, in its
acquisition. It is a “storehouse” which is “filled” in the process of acquisition,
and seeing language as “a storehouse filled [emph. mine, PWD] by members
of a given community” (Saussure 1959:13) gives language yet another source
for its nonwillfulness, i.e., its ontogeny.4

2.2 On the opposition of Willful to Nonwillful

Because speaking involves choice — selection — it is willful, not just in
the sense of choosing what to say, but also how and when (or whether)
anything is said. Equivalently, speaking is a conscious activity, whereas
language is unconscious/nonwillful. Language never requires “premedita-
tion” (Saussure 1959:14) while speaking does.

... reflection enters [language] only for the purpose of classification (Saussure
1959:14).

The introspective examination of language is carried out by a linguist for the

purpose of description. Language is nonwillful, as well, because it cannot be
altered at will. This property is tied to the notion that language is social and
beyond the reach of the invididual, each speaker having only a portion of it
(Saussure 1959:71 & 72):

The masses have no voice in the matter ... speakers are largely unconscious of the

4 Recall Hockett’s (1948) desire to reduce language to terms of a stimulus-response

mechanism and a central nervous system. This will produce a ‘storehouse’ filled in a ‘non-
willful’ way just as in Saussure’s vision. Does Chomsky’s idea of a LAD have the same
qualities as Saussure’s and Hockett’s views of acquisition ... or does it differ?
 Saussure: Introduction 5

laws of language; and if they are unaware of them, how could they modify them?5

Recall from Hockett (1983:17):

Schleicher’s reasoning [concerning the absence of role of free will in determining

language change, PWD], never fully elaborated, was Hegelian in the same way as
was that of Karl Marx (1818-1883) at about the same time. Marx did not deny, for
example, that an individual entrepreneur can try to be a generous fellow if he so
chooses. He argued only that the capitalist system — the interlocking
relationships among people within which their decisions lead to one or another
result — involves internal stresses forcing it as a whole to develop in a certain
way, resulting ultimately in a relatively sudden system-changing outcome ...
Whitney, true to the spirit of the American frontier, would have been much more
concerned with the free-will nature of the individual decisions that lead (?) [This
is Hockett’s “?”, PWD] to the collective consequence; Schleicher would have
been more intrigued by what he thought was the inevitability of the consequences.

Hence, language is passive in a further way. Diachronically, while language

may not be willfully changed by individuals, their behavior is nevertheless the
unwilled source of language change. Language responds to changes in
speaking and may itself change.

2.3 On the opposition of Heterogeneity to Homogeneity

This opposition follows from several others. First, because language is a
social average, complete within a collectivity that excludes individual
variation, it projects ‘ideal’ speaker-listeners that are by definition homogene-
ous. But notice that in Saussurean terms such ‘ideal’ speakers cannot (?) in
fact exist since each speaker will be partial, incomplete, and not ‘ideal’ in
that she will not represent the entirety of language. Second, because language
is opposed to willful speaking, speaking is heterogeneous, as a function of
individuals, and individual occasions. Third, because language is not an
activity, a use, its localization must lie somewhere other than in the chain of

5 Cp. Whorf’s (1940:221) reference to “automatic, involuntary patterns of language”. It may

be noted here that Hockett himself (and others) attribute an absence of willfulness to sound
change because it occurs outside of awareness (Hockett 1965:202):

[sound change] is not REDUCIBLE to borrowing because the density distribution is

largely altered by innumerable tiny imprecisions of pronunciation and by constant
channel noise ... that take place totally out of awareness.

Hockett (1965:191) sees the regularity of change in sound change itself and it is the social
context which produces any contradiction to the regularity. This view is the inverse of the
one described in the quotation just below from Hockett (1983).
6 HALT

activity that is speaking (Saussure 1959:14):

It can be localized in the limited segment of the speaking-circuit where an

auditory image becomes associated with a concept.

I.e., it is psychological, existing in what Saussure terms the associative

center. Its nature is homogeneously psychological while speaking remains
composed of various kinds of phenomena — psychological, physiological,
physical.

2.4 The implications of these properties

Ultimately, language is patterned and regular whereas speaking is irregular
and unpatterned. We may see this projection of language upon the collective
society as a continuation of the perception of sound change in the nineteenth
century as outlined by Hockett (1983). The regularity there escaped from the
indeterminateness of Geisteswissenschaft by conceiving change as a mass
(social) phenomenon, which is beyond the reach of any individual to initiate
or to alter. The notion of social is primary for Saussure and the other
properties of language appear to follow from it. Yet from our point of view it
leads to some odd conclusions.
Assuming that language derives its homogeneity from its social nature, we
are directed to search for a homogeneous speech community where that
homogeneity will reside. Yet that search (e.g. linguistic geography) has not
found such a community.6 In the same search for homogeneity, American
structuralism begins with language and then fractures that concept into
dialects, and then further into idiolects, and finally into styles.7 The curious
result of that progression is that when homogeneity is ultimately found (?), the
social property is lost. For Transformational Generative Grammar, any person
who knows the language can be an ideal speaker-listener; the language is
complete in that individual. Homogeneity and social are not concomitant
properties in the American style of linguistics. The ‘language’ studied now is
the behavior of one speaker behaving in one style. Language is no longer a
collective, and it is complete in the individual. Language is contextually
bound to usage, i.e., a specific style in a specific speech act, and is no longer
the usage-free thing it was for Saussure. It is circumscribed by the terms of
that usage.

6 Cf. Saussure’s (1959:90) ‘idiosynchronic’.

7 Cf. Bloch (1947) and the description of the variants of English have: hǽv, h æ̆v, and v.
 Saussure: Introduction 7

3. The sign
The implementation of Saussure’s socially homogeneous language
depends upon his use of the sign, which is a composite of a concept and a
sound-image held together by a bond of mutual implication.8 This relation
establishes the concept as a signified and the sound-image as a signifier (cf.
Figure 3 from Saussure 1959:114). Neither exists independently of the
other as there is no up without the opposed down ... no left without the
contrary right. Both derive their existence relationally, rather than by their
own content.

Figure 3: A depiction of the sign relation.

The sign exists as a psychological reality independent of its manifestation,

e.g. “phonemes ... which suggest[...] verbal activity ... is applicable to the
spoken word only” (Saussure 1959:66). And (Saussure 1959:94):

The word-unit is not constituted solely by the totality of its phonemes but by
characteristics other than its material quality [emp. mine, PWD].

Signs acquire their existence and their character not from their content, but
from their place in a system, by their opposition to one another. The
matter/material that realizes them or the opposition is irrelevant.
The separation of the signs of language/langue from material expression
implies a second property of signs, the famed arbitrariness of the bond
between the signifier and signified. Given that language/langue exists
independently from its manifestation and that a signifier and signified take
their status from that mutual relation, it matters not what signifier bonds with

8 But notice that the ‘bond’ may experience degrees of necessity (Saussure 1959:75):

Latin necāre ‘kill’ became noyer ‘drown’ in French. Both the sound-image and the
concept changed; but it is useless to separate the two parts of the phenomenon; it is
sufficient to state with respect to the whole that the bond between the idea [i.e.,
concept/signified] and the sign [i.e., sound-image/signifier] was loosened [emph. mine,
PWD].
8 HALT

what signified. This systemic arbitrariness is supported by the empirical

observation that one cannot reason from sound-to-meaning nor from meaning-
to-sound.9

4. Language/langue and the theory

Language/langue is a system of arbitrarily constituted signs, defined by
their opposition to other signs. Language is abstract, but it is not an ab-
straction.10 Saussure takes care to emphasize the reality, the existence of the
system of signs (Saussure 1959:107):

Language then has the strange, striking characteristic of not having entities that
are perceptible at the outset [i.e., abstract] and yet not permitting us to doubt that
they exist [i.e., not abstractions] and that their functioning constitutes it.

Saussure adopts a realist’s position towards his object of study; he advocates a

God’s-truth belief in the actuality of langue and not a hocus-pocus attitude.
In denying the possibility of arriving at the system by means of a series of
analyses, Saussure rejects an operational kind of theory and proposes a
theory which is explanatory. There are several motivations for his position.
First, because language/langue has an abstract (though real) existence and
only an arbitrary association with its realization, it is not possible to reason
from sound-image (or from concepts) to the sign. 11 It is the (system of) sign(s)
which order(s) and shape(s) sound-images and concepts; and without prior
knowledge of the sign, one cannot know what portions to operate upon. The
segmentation is not a given. Any structuring of sound or thought results from
projecting the form of language/langue upon an otherwise formless purport.
Cf. Figure 4 (Saussure 1959:112). Given A and B in Figure 4, it is not
possible to see their organizations unless they have been given before; but

9 This arbitrariness is found in American structuralism in the patternless association

between levels, e.g. in the connection between syntax/morphology and phonology. Cf.
footnote 12. Grammar cannot be reduced to phonology (except arbitrarily) in the same way
that language cannot be reduced to physics/chemistry (except arbitrarily).

10 Recall my earlier use of ‘covert’.

11 This recurs in Chomsky’s (1960) advocating an explanatory theory in the face of an

impenetrable blackbox, the LAD.
 Saussure: Introduction 9

Figure 4: The relation of Saussurean form to substance.

that organization is just what an operational approach is trying to determine.12

Second, an operational discovery of the elements of language/langue would
lead to an abstraction. Consideration of the French alternation between mwa
and mwaz ‘month’, and attempting to establish a signifier would yield an
abstraction which is the ‘link’ between the two, but is neither, and which has
no status in parole, nor in langue (Saussure 1959:105):

In mwa (mois, as in le mois de Septembre ‘the month of September’) and mwaz

(mois, in un mois après ‘a month later’) there are also two forms of the same word,
and there is no question of a concrete unit. The meaning is the same, but the slices
of sound are different. As soon as we try to liken the conrete units to words, we
face a dilemma: we must either ignore the relation — which is nonetheless evident
— that binds cheval and chevaux, the two sounds of mwa and mwaz, etc. and say
that they are different words, or instead of concrete units be satisfied with the
abstraction that links [emph. mine, PWD] the different forms of the same word.13

The signifier component of the sign (cf. Figure 3) is an unresolved whole.

12 This recalls Hammarström’s (1978) distinction between internal and external and the
direct seizure of language through introspection (Hammarström 1978:20 & 22):

A linguist can study language externally. He may have to do so, or he may choose to
do so, but in both cases his description will be at least somewhat wrong and incomplete
... I have previously suggested ... that this kind of scrutinizing [of an internal object]
involves intuition or introspection. Intuition would imply a more direct procedure: one
can immediately tell that in English the definite article always precedes ... the noun.

13 This dilemma looks forward to Hockett’s later trilemma (1961:30):

(1) Knife- and knive- are the same morpheme.

(2) Knife- and knive- are phonemically different.
(3) A morpheme is composed of phonemes.

One of the propositions must be false.

10 HALT

Third, the discovery techniques would yield a list and not a system, thus
missing the structural essence of language/langue. And fourth, signs match
no entities of a fixed size in speaking. They are not uniquely equivalent to
‘words’ (e.g. porte-plume), nor to ‘locutions’ (e.g. s’il vous plait), nor to
‘sentences’. Beginning with one or the other, then, will not produce consistent
identification of the signs of the language system.
The theory is therefore explanatory. Saussure underscores — implicitly
— that the explanation is a deductive one, not historical, and seeks to establish
a synchronic linguistics. The explanation is not one of cause-and-effect, that
Hammarström (1978:26) attributes to the natural sciences but the ‘weaker’
deductive one. The system which Saussure suggests exists independent of
time ... it is unchanging ... thus cause-and-effect explanations can have no
home in language/langue. Given a sign relation as in Figure 3, the replacement
of the content of the signifier, or the replacement of the signified will have no
effect upon the system as long as the system, i.e., the structural relations
remain unaltered (Saussure 1959:94):

... these transformations are basically alien to words and cannot touch their
essence.

One can see additional properties in language change which set the history of
language off from atemporal language and which void historical explanations.
First, change affects only one term of the sign. It is phonetic or semantic, as
when the pronunciation of Germanic gast/gasti ––> Gast/Gäste without
changing the semantics. The history of language does not deal with signs, and
“to try to unite such dissimilar facts in the same discipline would be certainly
a fanciful undertaking” (Saussure 1959:85). The patterns of language are not
those of change (Saussure 1959:93 & 104):

... if one speaks of law [i.e., pattern] in synchrony it is in the sense of arrangement
[i.e., static], a principle of regularity ... [whereas] ...Diachrony supposes a
dynamic force through which a thing is produced, a thing executed.

Second, change like speaking is active; both are events. There will be no
‘events’ in language/langue, only “the momentary arrangement of terms”
(Saussure 1959:80 & 81):

The first thing that strikes us when we study the facts of language is that their
succession in time does not exist insofar as the speaker is concerned. He is
confronted with a state. That is why the linguist who wishes to understand a state
must discard all knowledge of everything that produced it and ignore diachrony.
 Saussure: Introduction 11

Implicit in this is that whatever pattern Saussure adduces for language/langue,

it will be static and taxonomic.
 Chapter 4

Saussure
The Theory

1. Introduction
First, further on the data. In his conceptualizing language, Saussure has
already excluded a class of phenomena from consideration, namely, all that
associated with speaking. Saussure eliminates — or at least seems to —
additional realms from inclusion within language. This further exclusion rests
on two assumptions: first, the primacy of language over parole and second,
his insistance that language is a system of signs. Since a sign is a mutual
implication of two terms — a signifier and a signified, neither without the
other has status within language. Absent its existence in the sign, a signifier is
not linguistic; and the same is true of signifieds. To study a signifier
independently of a signified is to engage in a study of physiology; and a
similar attempt at independent study of a signified takes us into psychology
(Saussure 1959:103):

A succession of sounds is linguistic only if it supports an idea. Considered

independently, it is material for a physiological study, and nothing more than that.
The same is true of the signified as soon as it is separated from its signifier.
Considered independently, concepts like ‘house’ ... belong to psychology.

Saussure appears to have eliminated patterns of semantics and phonology

from language and given them to other fields, but notice as well that language
projects a segmentation upon ‘thought’ and another upon ‘phonic substance’.
(Figure 1 is a modified version from Saussure 1959:112.) Yet those segments,

Thought

Signs

Phonic
Substance

Figure 1: ‘Levels’ in Saussurean theory.

2 HALT

each within its own realm — materially as ‘thought’-segments or ‘phonic’-

segments, continue to lie outside language.

2. The theory
Saussure suggests a different kind of existence for thought and sounds, an
interrelationship based on difference, whereas the relation which maintains
the language system is the opposition of values (Saussure 1959:117 & 121):

Instead of preexisting ideas then, we find in all the foregoing examples values
emanating from the system [of signs]. When they are said to correspond to
concepts it is understood that the concepts are purely differential and defined not
by their positive content but negatively by their relations to other terms of the
system. Their most precise characteristic is in being what the others are not
[Emphases mine, PWD] ... When we compare signs — positive terms — with
each other, we can no longer speak of difference; the expression would not be
fitting for it applies only to the comparing of two sound-images, e.g. father and
mother, or two ideas, e.g. the idea ‘father’ and the idea ‘mother’; two signs, each
having a signified and signifier, are not different but only distinct. Between them
there is only opposition. The entire mechanism of language, with which we shall
be concerned later, is based on oppositions of this kind and on the phonic and
conceptual differences that they imply.

Signifiers of distinct signs are mutually ‘different’, and similarly for

signifieds. The notion of ‘value’ is the functional equivalent of ‘contrast’ in
Bloomfieldian linguistics and in the following structural interpretation of
Bloomfield. Neither of value nor contrast admits partial similarity. The two
relationships are absolutes. If signs X and Y have value they are entirely
unlike each other. There is no partial similarity. If X and Y contrast, they, too,
are each absolutely and completely unlike the other ... on that level. If mother
and father contrast as morphemes, they are absolutely unalike.1 Given that
signs are positive units (mutual implications) of signifier/signified (whereas
the latter themselves are not), Saussure chooses ‘distinct’ to qualify the
relation between signs and ‘different’ to characterize the mutual relations
between signifieds and signifiers.

1 The American interpretations of Saussurean value do, however, allow one to incorporate
the observation that there nevertheless appears to be some similarity between the two
signs/morphemes. But to do so requires an amplification of levels beyond the single one that
Saussure maintains. The similarities between mother and father are lodged first in the
phonology that allows one to state a common occurrence of , , and /r/, and second in a
semantics that permits one to recognize a common presence of <parent> in both mother and
father. Yet, on the level of morphology, mother and father remain absolutely unlike. They
are unitary and not internally composite in exactly the same way that Saussure’s signs are.
 SAUSSURE: The Theory 3

A different interpretation of this relation of physiology (and psychology)

to language is possible. The American interpretations of Saussurean value
allow one to incorporate the observation that, for the absolutness of the
contrast, it nevertheless appears possible for there to be some similarity
between the two signs/morphemes. But to recognize that pattern requires an
amplification of levels beyond the single one that Saussure maintains. The
similarities between mother and father are lodged first in the phonology that
allows one to state a partial similarity based on a shared occurrence of , ,
and /r/, and second in a semantics, that permits one to recognize a common
presence of <parent> in both mother and father. Yet, on the level of
morphology, mother and father remain absolutely distinct. They are unitary
and not internally composite in exactly the same way that Saussure’s signs
are. This ‘looser’, ‘mediated’ interpretation of the sign leads to the standard
textbook schema of language represented in Figure 2.

Semantics

Grammar

Phonology

Figure 2: A common presentation of ‘levels’ in language.

There may be a system of signifiers and of signifieds, but Saussure seems

to find no pattern within them to discuss. There are no sames, no recurrences
... properties which we recognize as diagnostic of pattern. There is nothing,
therefore, to be said of them. Again signifieds and signifiers are in language
and participate in whatever pattern that may characterize language only by
virtue of their participation within signs. There is no independent study of
semantics nor phonology. There is only something grammar-like, and by
comparison of Figure 1 with Figure 2, this theory of language has only one
kind of pattern, and one level.

2.1 Size levels

There exists a second limitation within the system of signs, i.e.
Saussure’s grammar. Frequently, grammar is conceived as a hierarchy with
minimal units arranged into larger ones. The concern here is with the limits
placed upon the larger domains of pattern. Traditionally, that largest unit is
4 HALT

designated as the sentence, but Saussure notes that the most characteristic
property of sentences is their chaoticness (Saussure 1959:106 & 124):

If we picture to ourselves, in their totality the sentences that could be uttered, their
most striking characteristic is that they in no way resemble each other ... diversity
is dominant ... it [the sentence] belongs to speaking, not to language.

But do all combinations of signs share in the chaos of grammar, or do some

escape it and belong to grammar? If we propose a distinction between those
combinations of signs that are not language and those that are, then we must
be able to recognize which are which. Where, then, does the upper boundary
of sign combinations lie? It is not determinate. It exists where a language has
(Saussure 1959:125)

registered a sufficient number of specimens

such that there exists a fixedness (Saussure 1959: 124 & 125):

(i) “pat phrases”, e.g. à quoi bon ‘What’s the use?’

(ii) “idiomatic twists [that] cannot be improvised”, e.g. forcer la main à
quelqu’un ‘To force someone’s hand’
(iii) “idiomatic twists ... furnished by tradition”, e.g. facilité : facile, but
difficulté : difficile, not *difficilité.2
(iv) “syntagmatic types that are built upon regular forms”, e.g.
indécorable ‘undecoratable’ on the model of impardonnable,
intolérable, infatigable, etc.3

The line is not drawn by ‘size’, but by the degree of fixed conventionality, and
not all utterances satisfy this criterion. “Speaking is characterized by freedom
of combinations” (Saussure 1959:124). Apparently, most of syntax exhibits
this freedom and is therefore excluded from Saussure’s theory of language.
There can be no ‘creativity’ in Saussure’s concept of language/langue. It is by
the removal of this property — by grammaticization — that a combination
crosses the boundary from speaking/parole and enters into language/langue.

2 “These idiomatic twists cannot be improvised; they are furnished by tradition” (Saussure
1959.125).

3 A form like indécorable “already has a potential existence in language; all its elements are
found in [other] syntagms” (Saussure 1959.166)
 SAUSSURE: The Theory 5

Signs may have a rough correspondence to terms within traditional

grammar as outlined in Figure 3. Saussure (1959:114) adopts the term ‘word’
as a technical one to cover this diversity (Saussure 1959:113-14):

Being unable to seize the concrete entities or units of language directly, we shall
work with words. While the word does not conform exactly to the definition of the
linguistic unit ... it at least bears a rough resemblance to the unit and has the
advantage of being concrete, consequently we shall use words as specimens
equivalent to real terms in a synchronic system, and the principles that we
evolve with respect to words will be valid for entities in general [Emphases
mine, PWD].

SIGN

Some sentences Phrases Morphologically Simple

complex words words

Figure 3: Complexity within Saussurean signs.

A word, then, is a sign that corresponds to an utterance of indeterminate

complexity, and his theory is now one of words.4

2.2 Value
There are several properties that may be attributed to words, but the
primary one is still value. Saussure has used value to characterize the
relationships between signs and Saussure reapplies value to words. As noted
above, value is analogous to the more familiar notion of contrast or non-
identity. Saussure continues with the notion and reaches an extreme result.
Words form a system by their being distinct from other words. It is value that
interrelates words and yields the system he so insists upon. Saussure has
already said that all content/matter is/functions as the manifestation of words/
signs. How then to get at these words/signs if they have no content? A word
exists simply by its opposition to other words — a word is defined by what it
is not. Consider Figure 4. In (a), we see three geometrical shapes, each not the
other. We can understand (‘grasp’) them in that way by their contents as

4 The Saussurean ‘Word’ is the analog of the Chomskyan ‘S’.

6 HALT

‘circle’, ‘square’, and ‘triangle’. In this version, they are unconnected unless
we somehow add that relationship, e.g. ‘closed geometrical figures’. We can
build a connection into (a) by converting it to (b), where each is designated by
being not the others. We now know that ‘(not , not )’ is ‘ ’, that ‘(not ,

(not , not )

⇒ ⇒
(not , not ) (not , not )

(a) (b) (c)

(d)
Figure 4: From substance to form.

not )’ is ‘ ’, and that ‘(not , not )’ is ‘ ’. The interrelationship

between the three is now part of the ‘display’ in (b); but we still require ‘ ’, ‘
’, and ‘ ’ for the construction of (b). We may try to get further from the
specific positive content of geometry and to create a more neutral
interrelationship as in (c). But we have still employed a three-way distinction
between ‘solid line’, long-dash line’, and ‘short-dash line’. If (c) is placed in a
still larger context, we can recover/know which intersection is ‘ ’, which is ‘
’, and which is ‘ ’ by referring to the other (more distant) relationships,
which are in turn known by their relationship to the others. Each depends
upon the others, and if any of them changes, then they will all change
identities as well because they are the reflections of the relationships from one
point within all those relationships (Saussure 1959:110 & 121):

... elements hold each other in equilibrium in accordance with fixed rules, [and]
the notion of identity [e.g. ‘ ’] blends with that of value [e.g. ‘not , not ’]
and vice versa ... whatever distinguishes one sign from another constitutes it.

Version (c) may now be replaced with (d), in which reference to different
types of lines is removed. The version of (d) is now maximally efficient
(‘simple’) in requiring no positive primitive at all, no ‘ ’, no ‘ ’, no ‘ —’, no
 SAUSSURE: The Theory 7

‘- -’, etc. Just ‘not’ ... value. Language/langue is “a system of pure values”
(Saussure 1959:111), pure relationships. According to Figure 1, language/
langue “serve[s] as a link between thought and sound” (Saussure 1959:112)
while partaking of neither (Saussure 1959:113):

Linguistics then works in the borderland where the elements of sound and thought
combine; their combination produces a form, not a substance.

This conclusion is the lauded one, which Hockett (1983) identifies with
granularity and pattern. Having divorced langue from substance and the
individual speaker, its science attains the stature of Naturwissenschaft.
Theories of language can be further characterized as to whether they
assume as primitives things that occur in their theory-free observation of
language, primitives which contain positive ‘real world’ content, e.g.
phonetics or any portion of the human sensorium. Some do and some do not.
Those that do have been called empiric (n.b. not empirical), and those which
do not are non-empiric (n.b. not non-empirical).5 This opposition is one that
allows us to distinguish between and group conceptualizations of language,
both American and other:

Non-Empiric Empiric

Saussure Prague School

Hjelmslev Bloomfield/Post-Bloomfieldians
Firth Tagmemics
Stratificational Grammar Transformational Generative Grammar
Neurocognitive

This is a potential danger to a non-empiric theory. Saussure, by this attitude,

permits his theory to be so abstract that it may be a theory of more than
language (i.e., English, French, etc.). It is a theory of any system of
communication. If it is intended to be a theory of language alone, and to
describe what makes language, language, distinct from all other phenomena,
then it is mistaken in admitting within its range things which obviously are not
language. The mesh of the theory is not sufficiently narrow to exclude them.

5 Cf. the relevant chapters in Davis 1973. Stratificational grammar has evolved into
‘cognitive-stratificational’ grammar (cf. Copeland & Davis 1980), ‘relational network’
linguistics (Lamb 1994), and finally into ‘neurocognitive’ (Lamb 1999.). It has retained the
non-empiric character identified here.
8 HALT

The exclusion — if desired — may have been effected by assuming empiric

primitives (e.g. sound-images) or by assuming more or different patterns for
language. The patterns of words which Saussure attributes to language are,
however, not specific to human language, and Saussure has a theory of
semiological systems (semiology). Saussure has a theory more general than
any of the others we shall examine, but at the same time it says less in that it
attributes less rich pattern to that range of applications.

3. Patterns within signs

We now turn to the pattern of words/signs that Saussure perceives in
language. He first observes that while language is arbitrary within the sign —
i.e., there is no pattern — examining words/signs with respect to each other,
there can be less than complete arbitrariness. There occur samenesses; and
recurrences are apparent. Language is partially motivated. “In language
everything boils down to differences [Read ‘distinctions’ PWD] but also to
groupings” (Saussure 1959:128). That recurrence reveals first a pattern that is
associative. Cf. Figure 5 (Saussure 1959:129). Within the system, it is
possible to recognize that not all is negatively constructed. Having recognized

Figure 5: A Saussurean depiction of an associative relationship.

that some words are now complex by their entering into an associative
pattern, others will be simple. Complex words reveal a second, internal
patterning between their parts, a syntagmatic, both-and relationship. They
constitute syntagmatic solidarities (Saussure 1959:127).
Figure 5, of course, will require a reworking (or elaboration) of Figure 4.
These samenesses will have to be constituted as additional signs within the
system or the nature of system will have to be altered. Consider the following
set of data:

(1) heal (10) sixths

(2) six (11) foul
 SAUSSURE: The Theory 9

(3) slow (12) sloth

(4) merry (13) cloth
(5) wide (14) sloths
(6) sixth (15) wealth
(7) health (16) cloths
(8) well (17) filth
(9) mirth (18) width

Figure 6 represents a possible partial description of the data, but not the only

slow th
s
cloth
Root
Stem
Noun

merry er
slow est

Adjective

Figure 6: Another representation of Saussurean pattern .

one because there are uncertainties both in the data and in the theory.
Assuming that the patterns are to be attributed to English, we can first focus
on how to incorporate the patterns composed by the curly braces. Cf. Figure 5
and the slanted lines that radiate out from dé-faire. Does there exist in the
system of signs the associative pattern that collects the stems that may occur
before {th}? Or is there one that is the summarized occurrence before {s},
whether complex six-th or simple cloth-? Or before {er est}? Notice that in
Figure 5 dé is present four times and faire is also present four times ... and that
is without counting the potential of the two “etc.”. The depiction of Figure 5 is
10 HALT

not an unambiguous description of the data. But the alternative requires

something like Figure 6, in which paradigmatic (disjunction-based) classes
appear, e.g., ‘root’, ‘stem’, ‘noun’, and ‘adjective’. But what is the systemic
equivalent of Root, Stem, Noun, Adjective, etc.? Is it itself a sign? Saussure
(1959:110) responds:

... to base the classifications on anything except concrete entities — to say, for
example, that the parts of speech are the constituents of language simply because
they correspond to categories of logic — is to forget that there are no [Emphasis
mine, PWD] linguistic facts apart from the phonic substance cut into significant
elements [e.g. Figure 2.].

The description depends upon a latent system (Saussure 1959:130), which

must be a local configuration within the system of signs (Saussure
1959:130):

Our memory holds in reserve all the more or less complex types of syntagms,
regardless of their class or length, and we bring in the associative groups to fix our
choice when the time for using them arrives. When a Frenchman says marchons!
‘(let’s) walk!’ he thinks unconsciously of diverse groups of associations that
converge on the syntagm marchons! The syntagm figures in the series between
marchons! and the other forms determines his choice; in addition, marchons! calls
up the series montons! ‘(let’s) go up!’ mangeons! ‘(let’s) eat!’ etc. and is selected
from this series by the same process ... In reality the idea evokes not a form [i.e.
no sign, PWD] but a whole latent system that makes possible the oppositions
necessary for the formation of the sign.

Signs are it.6

[Version: September 14, 2005]

6 This less-than-maximally efficient description is ‘remedied’ in other theories, which

construct descriptions analogous to Figure 6. Recently (Davis 1993), views of language have
been proposed in which a return to the ‘distributed’ descriptions of Saussure is advocated (cf.
Fox 1994). The motivation is that there is no justification for the abstractions of Figure 6 and
that concepts of language which decline to invoke them are more exact conceptualizations of
language. Recognition that syntactic pattern is bound to and embedded within lexical matter
is more common.
The Saussurean separation of form from substance and the implication of this separation
(e.g. that the nature of the ‘substance’ itself injects no patterned relations into language) has
been partially rejected. In phonology (cf. Figure 2), for example, the ‘naturalness’ condition
(Chomsky & Halle 1968) illustrates a systematic intrusion of phonic substance into langue.
In semantics, the intrusion of substance is not generally acknowledged (but cf. Davis &
Saunders 1989). Importing psychological constructs into discussion of language is not the
same as recognizing a role for the substance common to psychology and linguistics.
 Chapter 5

Saussure
Limitations and Uncertainties

1. Introduction
There are several areas in Saussure’s conception/theory of language in
which we may see limitations and uncertainties. We have already seen that
Saussure appears to restrict the patterning of language to that of signs ...
roughly equivalent to what we would now call morphology (including
lexemes). Second, where pattern, i.e., associative pattern, is recognized within
language, it is expressed, seemingly, by clustering the signs which enter into
that pattern into one location within the system so that the relevant signs are
adjacent to one another. This, however, leaves a problem in how this
expression works more precisely. If the system of signs is created by the value
of signs, i.e., their being distinct from other signs, is the one sign in which we
are interested distinct from all signs equally? That is, is it the intersection of
contrasts with every other sign in the language? Compare the entry of Darn!
into the system of English langue. There would appear to be no one sign to
which it has a closer relation than any other. Where the sign of our interest
clearly has recurrences in complex signs, e.g. contaminat(e), contamination,
contaminant, contaminator, recontaminate, etc. is the relation of contaminate
to excite as direct a one as is the relation of darn to excite? Or is the relation of
contaminate to excite mediated by contamination and excitation? If not, then
how is the fact of associative relation registered into the system of signs that is
language/langue? And if the mediated alternative is the case, then how will
darn be integrated into the system such that its relations will recognize the
absence of any associative relation?
The key to the distinction seems to be the recognition of the syntagmatic
solidarities. A sign may be the mutual implication of one signified with one
signifier or it may be the mutual implication of two (or more) signifiers with
two (or more) signifieds. Darn will then have no relation with such a
syntagmatic solidarity. But then what is it related to directly? The only answer
would seem to be every sign in the language/langue. Contamin would then be
2 HALT

related to syntagmatic solidarities in which it also appears.1 But we still do not

know whether contaminat(e) has an immediate relation with excite or a
mediated one. It would seem to be that it must have the former. Otherwise,
how would we in some principled way relate darn to excite while excluding
contaminat(e) from such a relation? Compare the three possibilities from
parole:

(1) Darn them!

(2) Excite them!
(3) Contaminate them!

Any pair seems as related (or unrelated) as any other. But if we accept this,
then how can we distinguish the local presences of clusters of associative
relations from their absence? We must answer this question keeping in mind
that our own intuitive recognition of signs by their positive manifestations as
concepts and sound-images, e.g. [ ], is not available within language/
langue. That naive recognition must have some systemic equivalent (formal
expression) within language/langue, but Saussure does not give us explicit
answers to such problems, and we are left to work them out for ourselves ...
perhaps by extending the theory or, ultimately, by abandoning it for some
other.
Notice that the introduction of paradigms, or disjunctive classes, would
provide the basis for such a response. But that response would violate
Saussure’s constraint of theoretical realism in that such formal devices go
beyond the pattern which such a constraint will allow (Saussure 1959:137-38):

We can say that the sum of the conscious and methodical classifications made by the
grammarian who studies a language-state without bringing in history must coincide
[Emphasis mine, PWD] with the associations, conscious or not, that are set up in
speaking. These associations fix word-families, inflectional paradigms, and
formative elements (radicals, suffixes, inflectional endings, etc.) in our minds.

2. Uncertainties
One substantive area of uncertainty in this theory centers upon the extent
to which associative relations are to be attributed to language/langue.
Speaking of the Latin forms , , and (Saussure
1959:138):

1 I leave it as an unresolved problem just how we recognize that contamin is the first
member of the syntagmatic solidarity contaminant, but the second member of recontaminate.
 SAUSSURE: Uncertainties 3

The sounds of the three endings offer no basis for association, yet the endings are
connected by the feeling [!, PWD] that they have a common value which prescribes
an identical function. This suffices to create the association in the absence of any
material support and the notion of the genitive takes its place in the language
[Emphasis mine, PWD].

This may establish a single, simple sign and not an associative series because
the signified is also simple, i.e., ‘the notion of genitive’. Cf. Figure 1. The
suppletive relations between signifiers are no problem within this theory.
Since the theory is non-empiric, the positive association of a signifier(s) of
identical, similar, or completely different sound images is irrelevant for the
constitution of a sign, although methodologically it may be important.
Although the example in Figure 1 is an ‘inflectional ending’ and may
therefore be cast as an associative relation rather than a sign relation, such
variations as go ~ went would almost certainly involve a common sign. The
uncertainty here is that between pattern as an associative relation or pattern
as a sign.

Figure 1: Suppletion.

The same problem exists in considering a many-to-one relation between A

and B (Saussure 1959:104):

Take the two French phrases laf rsdüvã (la force du vent ‘the force of the wind’),
and abudf rs (a bout de force ‘exhausted’; literally ‘at the end of one’s force’). In
each phrase the same concept coincides with the same phonic slice, f˛ors; thus it is
certainly a linguistic unit [i.e. sign, PWD]. But in (il me force a parler
4 HALT

‘he forces me to talk’) f˛ors has an entirely different meaning; it is therefore another
unit [i.e. another sign, PWD].

The semantic distinction of ‘noun’ : ‘verb’ is the basis for the “entirely
different meaning”, and it is sufficient here to maintain a distinction between
two signifieds and therefore between two signs. But (Saussure 1959:108):

In the same vein, a word can express quite different ideas without compromising its
identity (cf. French adopter une mode ‘adopt a fashion’ and adopter un enfant
‘adopt a child’, la fleur du pommier ‘the flower of the apple tree’ and la fleur de la
noblesse ‘the flower of the nobility,’ etc.

Figure 2: Polysemy.

In separating n from v, Saussure seems to recognize obliquely the

presence of a signified ‘noun’ and ‘verb’ by permitting them to consistently
separate homonyms of this sort into two signs; while within the boundary of
‘verb’, adopter1 and adopter2 are perceived as the same signifier although
they “express quite different ideas”.
A constant uncertainty here (as is polysemy in the discussion of any
language) will be how to best recognize the boundary between the relation in
Figure 2 and that in Figure 3. Saussure’s methodological response to the
issue is (1959:108):

... there is identity because the same slice of sound carries the same meaning in the
two sentences. But that explanation is unsatisfactory, for if the correspondence of
slices of sound and concepts is proof of identity ... [as in the noun senses of force in
 SAUSSURE: Uncertainties 5

Figure 3], the reverse is not true. There can be identity without this correspondence
[e.g. Figures 1 and 2].

The boundary between the two cases, however, must be established in each
case (Saussure 1959:138):

... we never know exactly whether or not the awareness of speakers goes as far as
the analyses of the grammarian. But the important thing is that abstract entities
are always based in the last analysis on concrete entities. No grammatical
abstraction is possible without a series of material elements as a basis, and in the
end we must always come back to these elements.2

Figure 3: Homonymy.

In practice, the semantic side of the sign leans more heavily on the
“awareness of the speaker”, and the phonic side relies upon the presence of
“material form”. Together, they suggest resolutions to these problems, but

2 This reliance upon ‘concrete’ appears to be literal and not just confined to ‘awareness of the
speakers’ (Saussure 1959:139):

In English, the man I have seen apparently uses a zero-sign to stand for a syntactical
fact which French expresses by que ‘that’ (l’homme que j’ai vu). But the comparing of
the English with the French syntactical fact is precisely what produces the illusion that
nothingness can express something. The material units alone [Emphases mine PWD]
actually create the value by being arranged in a certain way ... a meaning and function
exist only through the support of some material form.

Old men and women (Wells 1947) would be a problem here.

6 HALT

because the theory is not an operational one, the product of their application
must finally be judged as to the accuracy of the representation of
language/langue which the linguist perceives; the correct one will “coincide
[Emphasis mine, PWD] with the associations, conscious or not, that are set up
in speaking” (Saussure 1959:138).3 The correctness of the grammar cannot be
recognized internally to itself.
There are without doubt uncertainties in the Saussurean paradigm and
areas to be worked through. As Kuhn suggests, this may be counted a
desirable aspect of a paradigm; it is suggestive of issues which require further
thought. Three areas of ambiguity in Saussure’s theory are:

(i) Size-levels within the one level, e.g. morphology versus syntax.
(ii) The amount of pattern within the size-levels of grammar that are
recognized, e.g. form classes.
(iii) Levels, e.g. phonology versus grammar versus semantics, etc.

We have discussed (i) and (ii). With respect to (iii), Saussure has claimed that
the only relationship among both the signifieds and among the signifiers is
difference, by virtue of their participation in distinct signs. Yet (Saussure
1959:126):

... the association [i.e. associative relation, PWD] may spring from the analogy of the
concepts signified (enseignement, instruction, apprentisage, éducation, etc.); or
again, simply from the similarity of the sound images (e.g. enseignement and
justement ‘precisely’). Thus there is at times a double similarity of meaning and
form, at times similarity only of form or of meaning. A word can always evoke
everything that can be associated with it in one way or another.

Cf. Figure 4 from Saussure (1959:126). Such a statement gives a strong indi
cation that there is an associative pattern to be found within the signifieds and
within the signifiers, a pattern which follows after their establishment in
those respective capacities by their participation in the sign. Such patterns —
while common sense from our perspective — are only suggested within the
Saussurean theory.4 Notice that, from the examples given, the phonological

3 Does this mean that language is not ‘pure form’?

Looking forward to Transformational Generative Grammar and its triad of observational,
descriptive, and explanatory adequacy, we can see a (perhaps unintended) reflection of
Saussure’s criteria in the TGG criteria of observational and descriptive adequacy.

4 Cf. the discussion of ‘value’ and ‘contrast’ in Chapter 4.

 SAUSSURE: Uncertainties 7

pattern exists only to the point of recognizing syllables, e.g. mã in clément

and justement. Further resolution of these into m and ã is not considered.
Associative patterns within signifiers appear to hold only between signifiers

Figure 4: Pattern at more than one level.

of complex signs, i.e., signifiers of signs which participate in a syntagmatic

solidarity. Thus, the mon ‘my’ and monter ‘to climb’ may not enter into an
associative relation because of some “analogy” in their signifiers, mõ and
mõte. Perhaps our awareness as speakers does not extend that far. 5

3. Conclusion
In Davis (1973:36-37), I attempted to summarize the accumulated
observations on Saussure in a more formal way:

Primitives:
1. Nonidentity, value relationships
2. Linearity
3. Conjunction
4. Mutual implication

Definitions
1. Word system: Defined in terms of mutual implication holding
between conjunctions of two systems that are in turn defined

5 This peripheral presence of associative relationships among the signifieds and the signifiers
such that each is dependent first upon their involvement within the system of signs gives an
even more Saussurean cast to Transformational Generative Grammar (cf. footnote 3), which
similarly begins with the syntactic component (system of signs) and then interprets the output
(signifieds and signifiers) by mapping a part (signifieds) onto a semantic reading (concepts)
and another part (sound image) onto a phonetic transcription (phonic substance).
8 HALT

by conjunctions of nonidentity relationships.

2. Syntagmatic solidarity: Defined as the relationships of
mutual implication holding between n elements of the word
system in which n ≥ 3 and n -1 of the terms are linearly
related.
3. Complex word: Defined as the element of a syntagmatic
solidarity with no linear relation with the other elements of the
syntagmatic solidarity.
4. Simple words: Defined as (a) the linearly related elements of a
syntagmatic solidarity or (b) elements of the word system not
related by syntagmatic solidarity.
5. Analysis: Defined in terms of word system, syntagmatic
solidarity, complex word, and simple words.
6. Associative relationship: Defined in terms of (a) complex
words analyzed such that at least one of the simple words of
each is the same (e.g. burglary and burglarize), or (b) a
complex word and a simple word such that they are related by
syntagmatic solidarity (e.g. burglar and burglary or ry and
burglary), or (c) the relationship between simple signs if the
complexity of signifieds and signifiers is admitted.
7. Syntagmatic relationship: Defined in terms of simple words
related to the same complex word by syntagmatic solidarity
and to one another by linearity.

Within such a theory, an accounting will now consist of the following:

1. A definition of a word system

2. A definition of the complex and simple words
3. A definition of analysis holding between the complex and
simple words
4. A statement of the associative relationships of minimum
words holding between them as elements related to
complex words
5. A statement of the syntagmatic relationships of minimum
words holding between them as elements related to
complex words
6. An interpretation of statements (1) - (5) such that the
elements and statements given there predict the signs and
patterns, and only those, in the data
 SAUSSURE: Uncertainties 9

7. An evaluation of the extent to which (6) is met and a

reworking of (1) - (5) until (6) is met and the correct data
are predicted.

[Version: September 8, 2005]

 Chapter 6

Bloomfield & Saussure

Samenesses and Differences

1. Introduction
Bloomfield’s (1933:22-24) interpretation of the speech act is derived from
a ‘typical’ scenario in which two individuals engage in a short conversation
one requesting the other for an apple and the second complying ... not
responding linguistically ... with the request:

Suppose that Jack and Jill are walking down a lane. Jill is hungry. She sees an apple
in a tree. She makes a noise with her larynx, tongue, and lips. Jack vaults the fence,
climbs the tree, takes the apple, brings it to Jill, and places it in her hand. Jill eats the
apple.

Note that there is no record of what language Jill speaks ... or even whether

she speaks a language; she “makes a noise”. If language resides somewhere in

this experience, we see immediately see that we stand outside it. We hear
only ‘noise’ without any idea of what, or whether, the noise ‘means’. Unlike
Saussure, we must burrow our way into language and discover its presence,
rather than working our way out from within it. Saussure’s statement that it is
2 HALT

the point of view that creates the object is acted out again here ... but the point
of view is decidedly different.
There may be many reasons for this turn. It is common to cite the
experience of linguists in North America with the indigenous languages,
especially Franz Boas (1858-1942) and his students (notably Edward Sapir
[1884-1939] and Alfred Kroeber [1876-1960]), who each worked on several
indigenous languages.1 Boas worked on languages of the northwest coast, e.g.
Bella Coola (Salishan) in the 1870’s and later, Nootka (Wakashan). Sapir
worked on languages of Canada (Sarcee [Athabaskan] and Cree [Algonkian]),

1 There were a complex of forces which produced the concentrated investigation of American
Indian languages. In 1840, James Smithson bequeathed $500,000 to the United States
government to establish an institution “for the increase and diffusion of knowledge among
men” (Hinsley 1981:17). It was called the National Institute until Congress created the
Smithsonian Institution in 1846. Circumstance contrived to cast a significant portion of its
research on the American Indian. Expeditions by amateur scientists yielded collections that
demanded museums to house them. Second, ethnologists of the time focussed on what seemed
to make the country unique. In 1846, “the respected ethnologist” Henry Rowe Schoolcraft
advocated development of (Hinsley 1981:20):

... an American scientific and literary tradition: “No people can bear a true nationality,
which does not exfoliate, as it were, from its bosom, something that expresses the
peculiarities of its own soil and climate”. In constructing its “intellectual edifice”
America must draw “from the broad and deep quarries of its own mountains, foundation
stones, and columns and capitals, which bear the impress of an indigenous geognosy”.
The native American Indians had borne this distinctive “mental geognosy”, and the
present tribes, “walking statues” of their progenitors, were monuments far more worthy
of study than the antiquarian remains of the Old World.

An in justifying a recommendation to the Institute to publish a Dakota dictionary and

grammar, the secretary of the Institute, William W. Turner, replied in 1851 (Hinsley
1981:49):

Scientific study of the aboriginal tongues ... rewarded the comparative philologist by
showing not only analogies with other languages of the world but fascinating
peculiarities as well, by disclosing “new and curious phases of the human mind” ... the
study of Indian tongues, even without their literature, provided the same kind of
“delight and instruction” that the naturalist enjoyed from a new species of plant or
animal.

The Institution supported expeditions, notably those by John Wesley Powell to the West and
Southwest and George Gibbs to the Northwest. In 1879, Powell successfully lobbied
Congress to establish the Bureau of American Ethnology within the Smithsonian, and “the
BAE was the only institution in the country willing to underwrite and publish the work in
linguistics and mythology that Boas considered integral to a complete science of
anthropology” (Hinsley 1981:251). Boas, himself, had a varied career which began in the
United States in the mid-1880’s as geographical editor for Science. Before he finally settled
in 1895 in New York at Columbia University and the American Museum, he engaged in
fieldwork, taught at Clark University (1889-1992), did museum work in Chicago until spring
1894, and for the eighteen months before the Columbia appointment was apparently
unemployed (Hinsley 1981:250-51).
 BLOOMFIELD & SAUSSURE 3

in California (Yana), and in the southwest (S. Ute). And Kroeber, primarily

Frans Boas 1905

an anthropologist chairing the department at UCBerkeley, worked on

California languages. Sapir continued the tradition with his students (e.g.
Stanley Newman [1905-1984], Morris Swadesh [1909-1967]2 , Mary Haas
[1910-1996], etc.). Newman worked primarily on languages of the southwest
(e.g. Yokuts), but, like Boas, he also worked on Bella Coola (1934). Mary
Haas, who spent most of her career at UCBerkeley, worked on languages of
the northwest (Nitinat [with M. Swadesh]), southeast (Muskogean), and
California. During WWII, she worked extensively on Thai. Bloomfield
worked on Algonkian (central and eastern Canada, and north central and
northeast USA), esp. Menomini. During WWII, Bloomfield had primary
responsibility for Russian (under the punning pseudonym “I. M. Lesnin”). For
a person trained in comparative Germanic or Indo-European, contact with
these languages cannot fail to make the impression that they are ‘other’ and to

2 “Even before receiving his doctorate at the age of twenty-four, he had worked with several
American Indian languages (Nez Perce, Nitinat, Chitimacha) and had collaborated with Sapir
on a monograph ... The early 1950’s were painful years for Swadesh. In his brief
probationary appointment at CCNY (1948-49), he had embarrassed the administrative
powers by vigorously championing student demonstrators. Being a man of powerful
convictions, he was inclined to be as uncompromising in battle for social or political idea as
he was in advancing a linguistic theory. As a result of this episode and of other less
publicized ones, he became labeled unambiguously as a ‘leftist’ during the noisiest period of
the McCarthy Era, and university administrators were unwilling to take the risk of hiring
him” (Newman 1967.948, 949).
4 HALT

emphasize the fact that the linguist is approaching the task as an ‘outsider’. It
is this perspective which inclines one to take language as the external act of
speaking rather than the internal knowledge which enables it.
Another source of this alternative view lies in these remarks (Bloomfield
1933:34):

It is a mistake, for instance, to suppose that language enables a person to

observe things for which he has no sense organs, such as the workings of
his own nervous system.

Because we cannot expect to gain insight from the speaker’s knowledge of the
language (Saussure’s “awareness of the speaker” , the speaker’s “feeling”, and
Hammerström’s “intuition” of the linguist), there is a resultant emphasis upon
how we go about working our way into the language. Here, Bloomfield
becomes programmatic, and it is left to other workers in this model to
elaborate the techniques (Bloomfield 1933:78 & 79):

In the case of a strange language we have to learn such things [whether two instances
of forms are the same, PWD] by trial and error, or to obtain the meanings from some
one that knows the language ... [A] little practice [Emphasis mine, PWD] will
enable the observer to recognize a phoneme when it appears in different parts of
words.

Yet Bloomfield does establish a ‘policy’ concerning the gathering of data.

Unlike Saussure who saw only variation in the individual and therefore placed
language/langue in the ‘average’ of the society, Bloomfield extends the
tightness of fit ... the arbitrariness, which characterizes the Saussurean sign, to
the behavior of individuals in their performance (Bloomfield 1933:37):

... there is another and simpler way [in addition to the statistical] of studying
human action in the mass: the study of conventional actions [Emphasis mine,
PWD] ... Here the linguist is in a fortunate position: in no other respect are the
activities of a group as rigidly standardized as in the forms of language. Large
groups of people make up all their utterances out of the same stock of lexical
forms and grammatical constructions. A linguistic observer therefore can describe
the speech-habits of a community without resorting to statistics.

Such a collocation as “conventional actions” would appear strange within a

Saussurean model, for it is the actions themselves which are assumed to be
chaotic. Convention applies to language/langue, and behavior, the actions of
speaking/parole, are individual and ungoverned by convention. Bloomfield,
 BLOOMFIELD & SAUSSURE 5

however, must make such an assumption.

2. The act of speech

Given the prior position about the issue of an ‘entry’ into language and the
conclusion that we must work our way into it from without, such a direction
will result in isolated and unpatterned data without the assumption of
conventional actions.3 Without it, we cannot proceed as a science because
there will be no pattern. Yet Bloomfield must also recognize that the
correlation between speech-utterances and practical events is variable; he
persists in extending the (potential) constancy of pattern to this aspect of
language as well. And he does it by adhering to the LaPlacean view of
determinacy described in Hockett 1983 (Bloomfield 1933:32-33):

The mechanism which governs speech must be very complex and delicate. Even if
we know a great deal about a speaker and about the immediate stimuli which are
acting upon him, we usually cannot predict whether he will speak or what he will say
... The materialistic (or, better, mechanistic) theory supposes that the variability of
human conduct, including speech, is due only to the fact that the human body is a
very complex system ... We could foretell a person’s actions ..., only if we knew the
exact structure of his body at the moment ... at birth or before –– and then had a
record of every change in that organism, including every stimulus that had ever
affected the organism.

Bloomfield’s attitude leaves only certain portions of the interaction

between Jack and Jill as subject for linguistic study. Bloomfield segments this
experience into three portions (Bloomfield 1933:23):

A. Practical events preceding the act of speech

B. Speech.
C. Practical events following the act of speech.

Notice the difference between Bloomfield’s interpretation of the act of speech

and Saussure’s. Saussure’s concern lay in activity which would be completely
subsumed in the ‘B’ portion of Bloomfield’s act of speech. The increased

3 This ‘constructive’ approach to language is directly reflected in Bloomfield’s (1926:154)

‘Assumption 1’: “Within certain communities successive utterances are alike or partly alike”.
Note well the un-Saussurean emphasis upon sameness/identity and not upon difference or
distinctness. Initially, everything strikes the ear as distinct; and this approach succeeds only if
it can produce/recognize identities.
6 HALT

complexity in Bloomfield’s notion is required precisely because of his

orientation to language as an outsider. The association between a thought and
its phonic representation for Saussure is found in the heads of the speaker and
hearer, but for Bloomfield that cannot exist ... or it cannot exist in a way that it
can be accessed and made the object of study. In its place we find the
externalized ... and observable ... addenda of ‘A’ and ‘C’, i.e., “practical
events”. As outsiders to language, there is no way to avoid this alteration, for
it is here that we encounter language. Figure 1 depicts the relation between
Saussure’s notion of the act of speech and Bloomfield’s.

PRACTICAL PRACTICAL
EVENTS EVENTS

A B C

Figure 1: The act of speech for Saussure and Bloomfield.

Examining the events in Bloomfield’s representation, he notes that it is

possible for A and C to occur without B. The events of ‘A’ constitute the
stimulus for the act of speech but may prompt another act (Bloomfield
1933:24), i.e., Jill herself may retrieve the apple. And B may not appear. Such
a connection of A with C, without the possibility of B, is the condition of
speechless animals; “The lone Jill is in much the same position of the
speechless animal” (Bloomfield 1933:24). In this condition, the occurrence is
represented as in (1):

(1) S ––––> R

‘S’ is the practical stimulus for some activity and ‘R’ is the practical
reaction to that stimulus. But (Bloomfield 1933:24):

Instead of struggling with the fence and the tree, she made a few small movements
in her throat and mouth, which produced a little noise. At once, Jack began to
make the reactions for her; he performed actions that were beyond Jill’s strength
and in the end Jill got the apple. Language enables one person to make a reaction
(R) when another person has the stimulus (S).

The interposition of those ‘movements’ have great effect culminating in “the

whole working of society” (Bloomfield 1933:24, 26, & 28):
 BLOOMFIELD & SAUSSURE 7

In the ideal case, within a group of people who speak to each other, each person
has at his disposal the strength and skill of every person in the group. The more
these persons differ as to special skills, the wider a range of power does each one
person control. Only one person needs to be a good climber, since he can get fruit
for all the rest; only one needs to be a good fisherman, since he can supply the
other with fish. The division of labor, and, with it, the whole working of human
society, is due to language ... The gap between the bodies of the speaker and
hearer –– the discontinuity of the two nervous systems –– is bridged by the sound-
waves ... The term society or social organism is not a metaphor. A human social
group is really a unit of a higher order than a single cell. The single cells in the
many-celled animal co-operate by means of such arrangements as the nervous
system; the individuals in a human society co-operate by means of sound-waves.

Rather than language depending upon society for its character, in the manner
of Saussure, the relation is now reversed, and language is here the source of
the integrity of society. 4 The position between the two is altered so that it is
language which is prior to society (Bloomfield 1933:29):

A group of people who use the same system of speech-signals is a speech-

community.

Again, the relation between language and its context is just the reverse of
Saussure’s. With the interposition of speech, r · · · s, (1) is amplified into (2):

(2) S ––––> r · · · s ––––> R

in which ‘s’ is “speech (or substitute) stimuli” (Bloomfield 1933:25). Now R

may follow S or s. And because “the reaction mediated by speech can take
place in the body of any person who hears the speech” (Bloomfield 1933:26):

The gap between the bodies of the speaker and the hearer — the discontinuity of
the two nervous systems — is bridged by the sound-waves.

And this finally is where the linguist finds his/her data (Bloomfield 1933:32):

... the linguist deals only with the speech-signal (r · · · s); he is not competent to
deal with problems of physiology or psychology.

4 Compare Saussure's projection of language onto 'thought' and 'speech'. Bloomfield projects
it onto 'human society'.
8 HALT

Language, once more, is what knits society together; language is primary

here. The speech occurrence or speech event (r · · · s) by itself attracts no
attention, and “the normal human being is interested only in S and R; though
he uses speech, and thrives by it, he pays no attention to it”. We as linguists
are interested precisely in that event (Bloomfield 1933:27):

worthless in itself

But we are interested in it because (Bloomfield 1933:27):

it has a meaning

And that ‘meaning’ is constituted in the practical events of S and R

(Bloomfield 1933:27):

When anything unimportant turns out to be closely connected with more important
things, we say that it has, after all, a ‘meaning’; namely, it ‘means’ these more
important things’. Accordingly, we say that speech-utterance, trivial and unimportant
in itself, is important because it has a meaning: the meaning consists of the important
things with which the speech-utterance (B) is connected, namely the practical events
(A and C).

3. Conclusion
We come around to a position, finally, similar to Saussure’s, yet with a
clearly different Bloomfieldian character:

signified practical events

| |
signifier speech-utterance

Figure 2: The sign for Saussure and Bloomfield.

But in place of signs, Bloomfield has forms (Bloomfield 1926:155):

6. Def. The vocal features common to same or partly same utterances are forms; the
corresponding stimulus-reaction features are meanings.
Thus a form [sign, PWD] is a recurrent vocal feature [signifier, PWD] which has
meaning [signified, PWD], and a meaning is a recurrent stimulus-reaction feature
which corresponds to a form.
 BLOOMFIELD & SAUSSURE 9

And (Bloomfield 1933:27):

To study this co-ordination of certain sounds with certain meanings [i.e., forms,
PWD] is to study language.

[Version: September 8, 2005]

 Chapter 7

Bloomfield
The Shape of the Theory

1. Introduction
Similar to Saussure, Bloomfield (1926) sets out a series of statements
which identify an initial conception of language.1 His initial primitives derive
from an act of speech, which Bloomfield accepts from the fields of
psychology and anthropology.2 Even so, to begin with, we do not know
which acts represent speech and which do not. That is resolved only when
some of acts are determined to be forms and some are not.

2. The Creation of the Theory

In a series of assumptions and definitions, Bloomfield establishes the
existence of utterance, language, and form (Bloomfield 1926:154-55):

1. Definition. An act of speech is an utterance ...

Assumption 1. Within certain communities successive utterances are
alike or partly alike3 ...
3. Def. Any such community is a speech-community ...

1 Bloomfield (1926.154) wirtes, “I am indebted to Sapir’s book on Language, New York

1921, and to de Saussure’s Cours de linguistique générale,3 Paris 1922; both authors take
steps toward a delimitation of linguistics.”

2 Cf. Chapter 6.

3 In Bloomfield (1933), this assumption is stated more specifically as follows (p. 144):

In certain communities (speech-communities) some speech-utterances are alike as to

form and meaning.

And Bloomfield remarks (1933:145) that “our fundamental assumption implies that each
linguistic form has a constant and specific meaning”. It is the failure of this implication (i.e.,
the absence of a constant and specific meaning) which may be the most serious weak point
of this and similar theories. Bloomfield recognizes the limitations of this assumption as well:
“... our basic assumption is true only within limits, even though its general truth is
presupposed not only in linguistic study, but by all our actual use of language”.
2 HALT

4. Def. The totality of utterances that can be made in a speech-

community is the language of that speech-community4 ...
5. Def. That which is alike will be called same. That which is not same i s
different ...
6. Def. The vocal features common to same or partly same utterances are
forms; the corresponding stimulus-reaction features are meanings ...
Assumption 2. Every utterance is made up wholly of forms.

There are two points here. First, the relation of ‘same’ (né ‘alike’) and
‘different’ are assumed and not defined. There is no definition of ‘alike’;
it is a primitive applied to vocal features and to stimulus-reaction
features without instruction on how to recognize it. It is either there or it
is not. Since there are no techniques presented for introducing it, and it
must be assumed that we can recognize ‘alike’. The elaboration of the
missing techniques and their incorporation constitute much of the later
activity in developing this theory. Second, language is identified here
with the totality of utterances that can be made. It is not equated with
the patterns which the utterances exhibit, and which, when expanded,
account for (generate) those utterances. It is the forms themselves.
Language is not thought of as a system, pregnant with all the potential
forms of language. This is similar in outline to Saussure’s notion of a
system of signs in which all information is represented without removal
of redundancies.5

2.1 The pattern of morphemes and sememes

Having identified the substance of language, Bloomfield then proceeds
to deal with forms in two ways: ‘morpheme, word, phrase’ in section III and
‘phonemes’ in section V. In a series of definitions, Bloomfield (1926:155-56)
provides succinct characterizations of our now common terminology:

8. Def. A minimum X is an X which does not consist entirely of lesser

X’s ...

4 Note again that ‘community’ is constructed from (and depends upon) the constancy of
language (i.e., “successive utterances are alike or partly alike”) and note that this reverses the
progression that Saussure proposed, in which ‘community’ was prior.

5 Notice, also, that this conception of language (as equivalent to the totality of its possible
utterances) is consistent with the outsider’s approach to the phenomenon. This attitude again
mirrors the American encounter with languages alien to Indo-European. There is an implicit
emphasis on corpus and linearity.
 BLOOMFIELD: The Shape of the Theory 3

9. Def. A minimum form is a morpheme; its meaning is a sememe ..

10. Def. A form which may be an utterance is free. A form which is not
free is bound ...
11. Def. A minimum free form is a word ...
12. Def. A non-minimum free form is a phrase ...
13. Def. A bound form which is part of a word is a formative ...
Assumption 3. The forms of a language are finite in number.

Recalling that a form is a selection of vocal features, what this series of

definitions does is describe one organization of the act of speech, i.e. the
vocal features. It does this first by identifying minimum forms, bound and
free, and then by organizing the minimum free forms into phrases by an ‘is a’
relationship.6 Thus, a linear continuum of vocal features may have an
organization/pattern of segmentation projected upon it as we recognize
minimum forms (the initial segments of vocal features), and then the
relationships of bound and free, and the boundaries of non-minimum forms.
Figure 1 is a depiction of this relationship. The sine-wave shape may be taken
as the continuum of vocal features; and upon them is projected a segmentation
into forms by virtue of certain portions of them being a “recurrent vocal
feature which has a meaning” (Bloomfield 1926:155). The hierarchical
layering of Figure 1 — at this point — is a convenience of representation. As
stated in the above series of definitions, morpheme, word, and phrase may be
projected separately and independently upon some portion of the vocal
features, but they may overlap in those features. The phrase the grandsons
and the word grandson may then segment the same vocal features as the
morphemes grand and son. Consider the four layers of boxes to be conflated,
the second row (grandsons) on top of the bottom (grandson), the third row

6 Comparing this to Saussure, Bloomfield’s forms find an analogue in Saussure’s words.

Each is a variable in terms of its extent. Bloomfield’s forms may be minimal and be
morphemes, middling and be phrases, or maximum and be sentences. Saussure’s words may
be minimum and be signs or they be be more inclusive and be (fixed) phrases or whole
sentences.
The organization of Bloomfield’s vocal features and of Saussure’s sound-images also
have a similar basis. For Bloomfield, vocal features exist by virture of the act of speech and
for Saussure, the sound image exists “only if it supports an idea” (Saussure 1959.103). Sapir
(1925.37-38) echoes this. He considers the sound [φ] to be non-language when used to blow
out a match, but part of language when it assists in the pronunciation of an utterance. In the
former performance, “The production of the candle-blowing sound is a directly functional
act.” (38) The is its own meaning. “The candle-blowing wh means business,” whereas in
the performance of when, the sound “is merely a link in the construction of a symbol [i.e.
sign or form, PWD].” Where language is present, the relation between sound-symbol, vocal
features, and sound and thought, stimulus-reaction features, and meaning is mediated by
sign, form, or symbol.
 4 HALT

(the grandsons) on top of that, and so forth to yield a kind of Chinese-box

arrangement.

The grand son s climb ed down

Figure 1: Depiction of the projection of segmenting organizationupon an utterance.

2.2 The pattern of phonemes

In Figure 1, it is the projection of the morphemic meanings alone —
sememes or ‘recurrent stimulus-reaction features’ — upon the vocal features
that results in the pattern. Another pattern is identified by the following series
of assumptions and definitions in section V (Bloomfield 1926:157):

15. Assumption 4. Different morphemes may be alike or partly alike as to

vocal features ...
16. Def. A minimum same of vocal feature is a phoneme or distinctive sound ... 7
Assumption 5. The number of different phonemes in a language is a
small sub-multiple of the number of forms ...
Assumption 6. Every form [each of which by Def. 6 is “vocal features
common to same or partly same utterances”] is made up wholly of
phonemes ...

7 Emphasis on linearity concentrates American phonology on segments and distribution.

Terminologically, ‘structure’ may label the linear pattern, and ‘system’, the non-linear
pattern. Phonological features are the focus of those more concerned with system. The first
real (home grown) intrusion of (simultaneous) phonological features is Hockett’s (1947)
“Componential analysis of Sierra Popoluca”. Concern with distribution results in ‘long
components’ (Harris 1944). Only with Jakobson, Fant & Halle’s 1951 Preliminaries to
Speech Analysis do features start to become the way to do phonology in North America.
 BLOOMFIELD: The Shape of the Theory 5

Assumption 7. The number of orders of phonemes in the morphemes and

words of a language is a sub-multiple of the number of possible orders ...
20. Def. The orders which occur are the sound-patterns of the language ...
21. Def. Different forms which are alike as to phonemes are homonymous.

This series takes the morpheme as its domain, and then projects a second —
and different — segmentation upon the same vocal features that have been
organized into forms.8 Cf. Figure 2. Phonological organization presup-
poses the morpheme segmentation, for it works within the segmentations pro-

The grand son s climb ed down

g r æ n s e n z k l a y m d d a w n

Figure 2: The second organization of an utterance in terms of phonemes.

vided by Figure 1. But it operates independently on the identified domains of

vocal features and does not presuppose any fixed relationship between the
units of the grammatical pattern, i.e., morphemes, and the units of the
phonemic pattern, i.e., phonemes. Any relation between the two organizations
of form and phomeme is indirect and mediated through the vocal features,
which are simultaneously, but independently formed by each. One does not

8 By Def. 9, morpheme is equivalent to (a kind of) form; and by Def. 6, form is reduced to
vocal features. Therefore, phonemes organize a range of predelimited (morpheme ‘size’)
vocal features, but require nothing in addition to that.
6 HALT

work ‘though’ grammar to reach phonemics, nor does one work through
phonemics to reach grammar. There is, therefore, no necessary hierarchical
connection between the two patterns. Each is an autonomous projection of
pattern upon the same vocal features. When Bloomfield (1926:157) writes:

The morphemes of a language can thus be analyzed into a small number of

meaningless phonemes. The sememes on the other hand, which stand in one-to-one
correspondence with the morphemes, cannot be further analyzed by linguistic
methods.9

it is not necessary to assume that the relation is

Sememes
|
Morphemes
|
Phonemes

Figure 3: A hierarchical relation among the organizations of an utterance.

Morphemes are projections of content, i.e., sememes, upon the vocal features;
and phonemes are a second, parallel projection upon the same vocal features.
But nothing which Bloomfield writes requires that morphemes ‘pass through’
phonemes on their way to the data, the utterances.
If there is a one-to-one correlation of meaning (i.e., sememes) to forms
(i.e., morphemes), then the top half of Figure 2 (above the continuum of vocal
features) is simultaneously a projection of content (Saussure’s signifieds)
upon vocal features, and the bottom half of Figure 2 is a projection of the
shape of expression (Saussure’s signifiers) upon that same stream of vocal
features. The congruent intersection of the two projections now stands as the
Bloomfieldian equivalent of the Saussurean sign. The isomorphism between
content/sememes and form/morphemes compels us not to distinguish a
grammar from a semantics, for the one is the other. To describe the grammar
of minimum and non-minimum forms is to describe simultaneously the

9 If the sememes could be further analyzed, the units of that analysis would parallel (in the
domain of ‘stimulus-reaction features’) the simultaneous components of Hockett’s (1947)
‘componential analysis’ (in the domain of ‘vocal features’). Although Bloomfield’s position
recognizes the associative relation that Saussure saw between enseignment, clément,
justement, etc., there is nothing which recalls the associative relation suggested by
enseignment, apprentissage, éducation, etc.
 BLOOMFIELD: The Shape of the Theory 7

semantics of language. And at this point there seem to be only two kinds of
pattern within language. Only when the techniques for establishing the units
are worked out more explicitly will the patterns be hierarchicalized as in
Figure 3.

2.3 The presence of order in language.

Bloomfield goes on to recognize a pattern analogous to Saussure’s
associative pattern, but this pattern is founded on sequence rather than the
paradigmatic associative pattern of forms (or signs as in Saussure)
(Bloomfield 1926:157-60):

22. Assumption 8. Different non-minimum forms may be alike or partly alike

as to the order of the constituent forms and as to stimulus-reaction features
corresponding to this order.
23. Def. Such recurrent sames of order are constructions; the corresponding
stimulus-reaction features are constructional meanings ...
24. Def. The construction of formatives in a word is a morphologic
construction ...
25. Def. The construction of free forms (and phrase formatives) in a phrase is
a syntactic construction ...
26. Def. A maximum X is an X which is not part of a larger X ...
27. Def. A maximum construction in any utterance is a sentence ...
28. Assumption 9. The number of constructions in a language is a small sub-
multiple of the number of forms ...
29. Def. Each of the ordered units in a construction is a position ...
31. Def. The meaning of a position is a functional meaning ...
32. Def. The positions in which a form occurs are its functions ...
33. Def. All forms having the same functions constitute a form-class ...
34. Def. The functional meanings in which the forms of a form class appear
constitute the class-meaning ...
35. Def. The functional meanings and class-meanings of a language are the
categories of the language ...
36. Def. If a form-class contains relatively few forms, the meanings of these
forms may be called sub-categories ...
37. Def. A form-class of words is a word-class ...
38. Def. The maximum word-classes of a language are the parts of speech of
that language ...

Note first that Bloomfield begins with “the order of the constituent forms”.
For Saussure, it is the presence of associative patterns (recognized by the
repetition of signs) which places the syntagmatic presence into relief. But here
it is partial sames of order (not of forms, but of position. Cf. Def 29.)
corresponding isomorphically to sames of stimulus-reaction which is the basis
8 HALT

for this pattern ... just the reverse of Saussure. This alternative emphasis upon
the syntagmatic at the expense of the paradigmatic is characteristic of
American structuralism. But given the prior emphasis on the r ··· s portion of
the speech act (as well as the common experience with analyzing spoken —
sequentially represented — samples of unfamiliar languages), this bias is not
surprising. Throughout, each formal construct maintains its one-to-one
relation with meaning:

Form Meaning

construction constructional meaning

position functional meaning
form-class class-meaning

If we are able to describe the morphemes, constructions, positions, and form-

classes, we have automatically described the semantics of language. Grammar
and semantics are not distinct. They constitute different aspects of one
patterning, that of constructions.

3. Conclusion and an alternative organization

The theory of language that is described in Bloomfield’s postulates posits
three distinct kinds of pattern in language: that of forms, that of phonemes,
and that of positions. Each of these three patterns supplies its distinct
organization to the stream of vocal features. Although the patterns of
phonemes depends upon the prior segmentation of the continuum of vocal
features into morpheme-sized chunks, phonemic organization does not work
‘through’ morphemes. And the organization of positions into constructions,
depends upon the assumption that “non-minimum forms may be alike or
partly alike as to order of the constituent forms”, but constructions are not
orders of forms. They are not another aspect of the previous patterning of
forms. Patterns of forms were recognized by the corresponding sememes. This
is a pattern of positions recognized by their own functional meaning.
Assumption 8 directs us to perceive this pattern via forms (in the way
phonemes were recognized via forms), but it is a separate and independent
organization of the stream of vocal features. Figure 4 attempts to depict the
relation of forms, phonemes, and positions to each other and to vocal features.
 BLOOMFIELD: The Shape of the Theory 9

Forms Positions

Phonemes

Figure 4: The Relations of Form, Position, and Phoneme to

Each Other and to Vocal features.

The series of definitions from 23 - 38 suggest a grammar of the familiar

sort with hierarchy: i.e., ordered positions filled by forms which are in turn
constructions composed of ordered positions, etc. But that is not a necessary
interpretation of Bloomfield’s construction(s), and it is not one that is
confirmed by Bloomfield’s Language (1933), in which he presents a slightly
different view of this aspect of language, one which centers about the notion
of taxeme, “a simple feature of grammatical arrangement” (Bloomfield
1933:166). Bloomfield (1933:163f.) identifies four taxemes:

(i) Order. This is the same “order” of Assumption 8. It consists of “the

succession in which the constituents of a complex form are spoken”
(Bloomfield 1933:163).
(ii) Modulation. Modulation consists of the use of “secondary phonemes ...
of pitch ... of stress” (Bloomfield 1933:163) to alter the sense of an
utterance: 2You’ve got a 3headache3|| versus 2You’ve got a 3headache1#
(iii) Phonetic modulation. It is recognized as “a change in the primary
phonemes of a form” (Bloomfield 1933:163). Compare:
(a) Who do you want to drive? /wánt /
(b) Who do you wanna drive? /wán /
(iv) Selection. “The meaning of a complex form depends in part upon the
selection of the constituent forms ... the features of selection are usually
quite complicated with form-classes divided into sub-classes”
(Bloomfield 1933:165)

Each of these taxemes by itself is a grammatical feature (cp. vocal feature)

which has no meaning. But the taxemes may act together in meaning
10 HALT

combinations (minimal or not) to create a tactic form, which when combined

with its meaning, is a grammatical form.10 Now the “smallest meaningful
units of grammatical form may be spoken of as tagmemes, and their meanings
as episememes” (Bloomfield 1933:166). Compare, for example, SV versus SV
in English:

(1) (a) Oh, am I lonely!

(b) ?Oh, I am lonely!

(2) (a) Why was the thief cáught?

(b) Why the thief was cáught ...

(3) (a) The thief had gotten caught.

(b) Had the thief gotten caught ...

(4) (a) I shall never do that again.

(b) Never shall I do that again.
(c) ?Never I shall do that again.

The taxeme of selection (of the noun form-class and the verb form-class) plus
the taxeme of order combines to effect the contrasting tagmemes of SV and
VS. SV will have one episememe which contrasts with the episememe of VS.
Finally (Bloomfield 1933:184):

The free forms (words and phrases) of a language appear in larger free forms
(phrases), arranged by taxemes of modulation, phonetic modification, selection, and
order. Any meaningful recurrent of such taxemes is a syntactic construction.

Such a conception of syntax (or more broadly grammar) allows us to

understand its patterns without requiring the notion of hierarchy. Such a flat
view of language is not, however, the one which comes to dominate the
syntax of American Structuralism. 11

10 A tactic form will always be a grammatical form, for it is the presence of meaning which
delimits the taxemes as a tactic form. Without meaning, taxemes are like vocal features
without accompanying stimulus-response reactions. Neither is an utterance.

11 The issue re-emerges within Transformational Generative Grammar in terms of whether or

not there exist languages which have transformational rules which do not refer to
hierarchical structure (i.e., trees or portions of trees). The fact that some languages do appear
to have rules which make reference to hierarchy is a strong justification for TGG as it stands;
 BLOOMFIELD: The Shape of the Theory 11

[Version: September 14, 2005]

but if languages (some or all) do not work in this way, then TGG is weakened. The issue here
is expressed in terms of configurational languages and nonconfigurational ones. Hale
(1976) suggests that Warlpiri (Walbiri) may be a nonconfigurational language
 Chapter 8

American Strucutralism:
Psychological Reality

1. Introduction
Bloomfield (1926:157) characterizes a phoneme as “a minimum same of
vocal feature”, and in his 1933 book, Language, he illustrates with the
description of phonemes for English while referring to “a moderate amount of
experimenting” (Bloomfield 1933:78), which will “reveal” (79) “replaceable
parts in the word” (79). And then when the phonemes in one part of the word
are determined by this replacement, “a little practice will enable the observer
to recognize a phoneme even when it appears in different parts of words”
(Bloomfield 1933:79). Such a statement may suffice when working with one’s
own language, but one’s intuitions (or “a little practice”) will not be adequate
to discover the phonemes of a language such as Chitimacha (a now extinct
member of the Gulf languages, which included Tunica, Natchez, and Atakapa;
formerly spoken in southwestern Louisiana). A more explicit statement of
how ... practically ... to proceed will be necessary (Swadesh 1934:117):

In studying the phonemes of Chitimacha (an Indian language of Louisiana), I

knew of no single source from which I could learn to understand all the
phenomena I observed. There seemed to be a need for an adequate and complete
exposition of the phonemic principle including, especially, an account of how it
applies to the more marginal and difficult types of phenomena.

As we now expect, the initial impulse to make phonology more explicit is the
practical one of language description ... description of an American Indian
language. The techniques then turn back upon the theory to reshape it in light
of the application. The direction of influence, from description to theory, is
clear in Sapir’s (1933/1951:23) remarks:

In the course of many years of experience in the recording and analysis of

unwritten languages, American Indian and African, I have come to the practical
realization that what the native speaker hears is not phonetic elements but
2 HALT

phonemes.

2. The Character of the Phoneme

To begin with, how is the notion of phoneme to be conceived? It is a
minimum same of vocal feature, but is it truly possible for the phoneme be a
constant, exact replication of a sound (vocal feature) from one occasion to the
next? If a phoneme is recognized as objectively different from utterance to
utterance, yet somehow still the same, is the objective difference being
attributed to interference of some sort? Is it an articulatory (motor/muscular)
sameness? Bloomfield (1933:80):

The phonemes of a language are not sounds but merely features of sounds [emph.
mine, PWD] which the speakers have been trained to produce and recognize in the
current of actual speech-sound — just as motorists are trained to stop before a red
signal, be it an electric signal-light, a lamp, a flag, or what not, although there is
no disembodied redness apart from these actual sounds.

This treats phonemes as phonetic stuff. A portion of the noise a speaker

makes can be recognized ... apparently ... as constant. In the same way that
“each linguistic form has a constant and specific meaning” (Bloomfield 1933:
145), a phoneme will have (a) constant feature(s) of sound and be recognized
by that diagnostic presence.
Swadesh (1934:118) emphasizes another aspect of the phoneme:

The phonemes of a language are, in a sense, percepts to the native speakers of the
given language. If they hear a foreign tongue spoken, they still tend to hear in
terms of their native phonemes ... If the phonemes are percepts to the native
speakers of the language, they are not necessarily percepts that he experiences in
isolation. They occur ordinarily as the elements of words or sentences. Phonemes
are perceptive units [Emphases mine, PWD] in the sense that the native can
recognize as different, words different as to one of the component phonemes, e.g.
bid and hid or bid and bed or bid and bit. The phoneme is the smallest potential
unit of difference between similar words recognizable as different to the native.

There are two points on which Swadesh, Sapir, and some others differ with
Bloomfield. First, a ‘percept’ or ‘perceptive unit’ is not ‘feature(s) of sound’.
And second, it will make a difference whether we choose to restrict the ability
to hear a difference to the domain of morphemes, as Bloomfield (1926:157)
appears to do with his Assumption 4 followed by Def. 16. In practice,
Bloomfield (1933, 1935) operates with words, minimum free forms. 1 Swadesh

1 In describing the Central Western dialect of American English (Chicago), Bloomfield

 PSYCHOLOGICAL REALITY 3

allows the domain within which that sameness is perceived to be much larger.
The issue is context, how much and whether it is relevant to the recognition
that “morphemes may be alike or partly alike as to vocal features”
(Bloomfield 1926.127).
Swadesh’s teacher, Edward Sapir, exemplifies both of these alternatives:
phonemes are ‘percepts’ and the domain for the application of these percepts
is larger than the word. Note this statement concerning the difference between
phonetics and the phoneme (Sapir 1933/1951:22):

... no entity in human experience can be adequately defined as the mechanical sum
or product of its physical properties.2

3. Phonology at Work
The principle of complementarity is illustrated in these data from S. Paiute
(Sapir 1933/1951:25):3

(1) [ '] /papa·/

(2) [ A] /papa/
(3) [páp·'A] /pap·a/
(4) [pApá'] /pap·a·/

In Sapir’s analysis, there are two contrasting phonemes /p/ and /p·/ with these
variants:

/p/ [p ]
/p·/ [p· p]

Only /p/ occurs initially, and both /p/ and /p / occur in V –– V. For /p·/, [p·]
appears after a voiced vowel (and it is also aspirated before an unvoiced

(1935:98) claims that “The existence of these phonemes is established by 136 such pairs as
pit : pet, look : luck, cam : calm, bomb : balm, see : say”. They are all words and
simultaneously, morphemes.

2 Sapir (1933/1951:22) continues: “These physical properties are needed of course to give us
the signal, as it were, for the identification of the given entity as a functionally significant
point in a complex system of relatednesses; but how many of these physical properties are, or
may be, overlooked as irrelevant, how one particular property, possessing for the moment or
by social understanding an unusual sign value, may have a determinedness in the definition
of the entity that is all out of proportion to its ‘physical weight’”.

3 Sapir transcribes a voiceless [a] as [A].

4 HALT

vowel as in [3]); and [p] appears after an unvoiced vowel (i.e., [A] in [4]). For
/p/, [p] appears initially; and [ ] and [ ] medially, with the latter before
voiceless vowels, the former before voiced. A form such as [ ] ‘at the
water’, when taught to a native speaker “of average intelligence” was rendered
... unguided by Sapir ... as “ ” (Sapir 1933/1951:24):

Tony was not ‘hearing’ in terms of actual sounds (the voiced bilabial was
objectively very different from the initial stop) but in terms of an etymological
reconstruction: pa·- ‘water’ plus postposition *-pa' ‘at’ ... a theoretically real but
actually non-existent form ... *pa' does not actually exist as an independent
element but must always be actualized in one of three possible postvocalic forms.

Sapir (1933/1951:25-27) describes Sarcee (Athabaskan) data as follows:

(5) [dìní'] ‘this one’ /dìní/

(6) [dìní'] ‘it makes a sound’ /dìnít'/
(7) [dìná· ] ‘he who is this one’ /dìnái/
(8) [dìníla] ‘it turns out that he is /dìníla/
this one’
(9) [dìnít'í] ‘he who makes a sound’ /dìnít'i/
(10) [dìní a] ‘it turns out that it /dìnít'la/
makes a sound’

With regard to the apparent ‘homonyms’ of (5) and (6), Sapir (1933/51:26)
notes:

In the early stage of our work I asked my interpreter, John Whitney, whether the
two words sounded alike to him and he answered without hesitation that they were
quite different ... When I asked him what the difference was, he found it difficult
to say, and the more often he pronounced the words over to himself the more
confused he became as to their phonetic difference. Yet all the time he seem
perfectly sure that there was a difference .... The one tangible suggestion that he
himself made was obviously incorrect, namely, that the -ní of ‘it makes a sound’
ended with a ‘t’. John claimed that he “felt a t” in the syllable, yet when he tested
it over and over to himself, he had to admit that he could neither hear a ‘t’ nor feel
his tongue articulating one.

Discovery of the additional forms of (7) - (10) show that there is a basis for
the speaker’s perception of a difference.
Nootka (Sapir 1933/1951) has these data:
 PSYCHOLOGICAL REALITY 5

(11) [his·ik] ‘?’ /hisik/

(12) [ ] ‘?’ / isa·/
(13) [kwis·i a] ‘to do differently’ /kwis-si a/
(14) [t as·at ] ‘the stick takes an /t a-sat /
upright position on the
beach’
(15) [ ] ‘we went there only to /
speak’ -go.in.order.to-just-

A pattern of “lengthening of consonants after a short vowel when followed by

a vowel” (Sapir 1933/51:27) accounts for the [s·] for /s/ in (11) and (12). The
curious thing is that a speaker of Nootka with whom Sapir worked “[wrote]
tsi·q it 'assat ni ‘we went there only to speak’” with ss and not s as (11) and
(12) (Sapir 1933/1951:27-28):

In such cases the long consonant is not felt to be a mechanical lengthening of the
simple consonant but as a cluster of two identical consonants ... Here again we
have objectively identical phonetic phenomena which receive different phono-
logic interpretation [Emphases mine, PWD].

Trager (1934) offers a phonemic interpretation of Russian vowels which

also reflects the same kind of concern that Sapir and Swadesh have. Consider
these phonetic data:

(16) [ ] ‘to drink’ / /

(17) [é] ‘to sing’ / /
(18) [ ] ‘five’ / /
(19) [ úst/
(20) [g I I

The stressed vowels of this variety of Russian are:

/i/ [i ] /u/ [u ü]
6 HALT

/e/ [e ] /o/ [o ö]
/a/ [a ä]

The pretonic vowel system has a four-way contrast, and the atonic system has
three distinctive vowels:

Pretonic Vowels Atonic Vowels

/i u /i u
e a/ a/

The pretonic vowels are assigned as allophones to one of the above phonemic
vowels as indicated by the transcriptions below:

[ dá] ‘misfortune’ / edá/

[ t] ‘misfortune [gen.pl.]’ / ed/

[ ] ‘misfortune [gen.pl.]’ / e dé/

The vowel [ ] in (32) is analyzed in the same way as the [ ] in (30) and
different from the vowel [a] in (31). Even though (31) and (32) contain the
same morpheme, “this [ ] [of (32)] cannot be distinguished from the other [of
(30)]”.

According to the principles of ‘structural’ transcription, we have to write [bjed j'e]

and [zjvj ezj dj'e]. We cannot write [bjidj'e], etc. since the symbol [i] [i.e., the
phoneme /i/] in pretonic position indicates the sound [i] [i.e., the allophone [i]],
while here we have the sound [I ] [i.e., the allophone [I]]. We conclude that in
pretonic position [I ] is part of the [e] phoneme, and is phonemically
(‘structurally’) different from the [I ] of unstressed syllables considered above [“we
write ... blue (nom. sg. masc) ... [sj'in ji] and pronounce [sj'in j I]”] as part of the [i]
phoneme.4
Trager (1934:337-38) concludes:

4 Note that phonemic notation is surrounded by [square brackets] in the same way that
[phonetic notation] is. Only with Hockett (1942) are /solidi/ conventionally used for
identifying phonemic transcriptions.
 PSYCHOLOGICAL REALITY 7

In atonic position, orthographic e is always [I ], except when final after [j]

preceded by a vowel, where we have [ ]. We can analyze this situation only in the
following manner: final after [j], the etymologically expected phoneme [e] is
replaced by the phoneme [a] (which, being unstressed, here appears with the
sound [ ]). In other atonic positions, the expected [e] is replaced by the regular
unstressed form of the phoneme [i]. Here there is no way to distinguish between
[the phoneme] [i] and [the phoneme] [e] because the sound [I ] [i.e., the phoneme
/i/] represents both of them in the same structural situations, whereas in pretonic
position, the sound [I] [i.e, the allophone [I ]] can only come from [the phoneme]
[e]. The phoneme [e], then appears only in stressed and pretonic positions, while
[the phoneme] [i] alone can appear in unstressed [i.e., atonic] position. The
psychological validity of the two analyses just made is proved by the spelling of
untutored speakers of Russian; in atonic position we find constant confusion
between the letters i and e, but in pretonic position they are kept distinct
[Emphases mine, PWD. Trager quotes Sapir 1933 for support here].5

A similar problem holds for the consonants:

(33) [gór t] ‘city’ / /

(34) [g r dá] ‘cities’ /garadá/
(35) [ráp] ‘slave’ /ráb/
(36) [r bá] ‘slave [gen.sg.]’ /rabá/
(37) [drúk] ‘friend’ /drúg/
(38) [drúg ] ‘friend [gen. sg.]’ /drúga/
(39) [páp] ‘priest’ /páp/
(40) [p pá] ‘priest [gen.sg.]’ /papá/
(41) [gró ] ‘farthing’ /gró /
(42) [pót] ‘sweat [n.] /pót/
(43) [p tú] ‘sweat [loc.sg.]’ /patú/
(44) [lúk] ‘onion’ /lúk/
(45) [lúk ] ‘onion [gen.sg.]’ /lúka/
(46) [ ] ‘herring’ / a/
(47) [útk ] ‘duck’ /útka/
(48) [ I ] ‘herring [gen.pl.]’ / /
(49) [út k] ‘ducks [gen.pl]’ /útak/
(50) [ ] ‘herring [nondim.]’ / /
(51) [ ] ‘herring [nondim.gen.sg.]’ / a/

5 “Examples of atonic e are materi, gen. of matj mother, [m'atjirj i]; budjte, imperative pl. be,
[b'utjtji]; velikan giant [vj iljik'an], where the first [i] is pronounced [I], according to the
regular rules for atonic and pretonic [i] phoneme; etc.” (Trager 1935:338).
8 HALT

(52) [rastóf] ‘Rostov’ /rastóv/

(53) [ ] ‘Rostov [loc.]’ /rastó i/
(54) [r stófsk y] ‘Rostov [adj.]’ /rastófskay/
(55) [g r tskóy] ‘city [adj.]’ /garatskóy/
(56) [rápsk y] ‘slave [adj.]’ /rápskay/

The rule for voiced stops and spirants is this, then: the etymological voiced sounds
... retain their psychological identity and distinction from the corresponding
voiceless sounds in final position or before a voiceless sound in all words in
which at least one inflectional form retains the original sound, even though they
are, objectively, completely voiceless in the positions indicated; but in
derivations under the same conditions, where the original voiced sound does not
reappear in any inflected form, we have complete psychological identification
[Emphases mine, PWD] of the original voiced sound with the new, voiceless
sound, and their merging into the voiceless phoneme, despite the presence of the
voiced sound in the original of the derivative, or in some other derivative. (Trager
1934:341-42)

4. Practice and Theory

Swadesh (1934:123) expresses the relationship between technique and
theory in this way:

The phonemes of a language can be discovered only by inductive procedure. This

going from particular instances to general conception is as characteristic of the
unconscious process of a native acquiring his language as it must be of conscious
scientific study [Emphasis mine, PWD].

A methodology is evolving, and it is assumed that the psychology of the

speaker will mirror the method. The form of phonology is isomorphic with the
psychology in the same way that form of grammar is assumed to be
isomorphic with the stimulus-reactions of psychology, anthropology, etc. The
theory is one which claims to have a degree of realism; but it makes a further
strong claim that the psychology it imputes is to be equivalent to the
emerging techniques. But this is already an intrusion of the linguist beyond
the realm suggested by Bloomfield (1933:32):

... the linguist deals only with the speech-signal (r ··· s); he is not competent
[emph. mine, PWD] to deal with problems of physiology or psychology.

Swadesh (1934:340) has remarked:

... the paradigm exists in the mind of the speaker as a psychological reality ...
 PSYCHOLOGICAL REALITY 9

The dangers of equating methodology with psychological reality is

illustrated with this set of data from Maori:6

ACTIVE VERBS PASSIVE VERBS

(57) awhi awhitia ‘embrace’

(58) hopu hopukia ‘catch’
(59) aru arumia ‘follow’
(60) tohu tohu ia ‘point out’
(61) mau mauria ‘carry’
(62) wero werohia ‘stab’
(63) patu patua ‘strike’
(64) kite kitea ‘see/find’

Conventional descriptions of these data, accepting the relevance of the

‘paradigm’, will recognize that some roots are consonant final and some are
not:

Consonant Final Vowel Final

awhit patu
hopuk kite
arum
tohu
maur
weroh

The regularity in (57) - (64) is that there appears to be no consonant final

words in the language. That is, a speaker of Maori knows:

C –––> Ø/ ––#

And additionally:

V –––> Ø/ V –– V
But there is a causative marker in Maori, whaka, which when added to verbs

6 The following forms are taken from Kiparsky (1971:591-93).

10 HALT

requires the additional presence of what appears to be the passive marker (to
indicate the causee). And when this appears, we find a pattern distinct from
(57) - (64):

(65) (a) mau whakamautia ‘make carry’

(b) *whakamauria

(66) (a) patu whakapatutia ‘make strike'

(b) *whakapatua

Generalized throughout, the shape tia appears in the causative. The tia shape
also is used in making passives of borrowed words, and in compounds, and
when nouns are used as verbs, and finally in forgotten words. They all exploit
the shape tia. This suggests that the pattern in (57) - (64) is misleading and
that the two regularities which drop consonants finally and vowels
intervocally are mistaken. Rather than the neatness they express, there is a
pattern in which some verbs take tia, others take kia, still others mia, ria,
hia, etc. And some use a. Our techniques guide us to the simplest, most
regular solution; but the behavior (and psychology) of the speakers indicates
this is mistaken and that the less simple solution in which the matching or
affixes with roots is effected by grammatical/lexical information (on an
idiosyncratic morpheme by morpheme basis) and not by phonological shape.

5. Conclusion
Several issues emerge from the analysis of these examples. The over
arching concern is how one discovers the presence of phonemic identity or
recurrence. Given the initial encounter with an unknown language, the first
impression is one of almost endless phonetic variation. Little seems to recur,
and the first demand is to find repetition. There must be some pattern in this
chaos, and it will be discovered by recognizing samenesses in spite of the
objective differences. How that is accomplished produces distinct notions of
what phonemes are like and of the character of language.
First, is the constancy of phonemes to be lodged in objective, physical
noise (articulation)? If so, how do we (or can we) discover some physical
sameness where we have decided a sameness of phonemes exists? If it is not
possible to discover a constant physical same for a phoneme in each of its
occurrences, where then does that ‘same’ reside?7 Contra Bloomfield,

7 Bloomfield wanted to maintain the phoneme as a physical concept while Sapir, Swadesh,
and Trager proposed the phoneme as non-physical, a “percept”.
 PSYCHOLOGICAL REALITY 11

language must be something other than vocal features. Second, how do we

project the phonemic sameness? If ‘same’ is not a given (and it is not, or we
would not be confronting this as a problem in the first place), how is it
created? Practically, we are introduced to Chitimacha or S. Paiute facing its
variation, and to achieve same, we must somehow neutralize the variety. This
is interpreted to mean that we should find ways to predict the variation. If we
can, the we have discovered a sameness uniting the variety, a pattern. The
obvious place to look is the environment. I.e., if the occurrence of X 1
correlates with what is adjacent to X1, and if the occurrence of X2 correlates
with what is adjacent to X2, and if the two correlates are different, then we
may think that there is a single same X, which is observed to be in fact either
X1 or X2 dependent upon what occurs adjacent to X. But this tact raises
additional questions. How are we to interpret “environment”? Is X1 or X 2
predicted by looking at the objective phonetic environment alone? Or is it
relevant to examine non-phonetic environment?8 If we examine phonetic
environment for conditioning context, how much of it (adjacent, close, or
distant) is permitted to effect variants in phonemes? Morpheme, word, or
phrase? Responses to some of these issues are discussed in the following
chapter.

[Version: October 4, 2005]

8 Bloomfield confined the environment to morphemes, or words equivalent to morphemes,

and allowed the simple recognition of same or different morpheme to decide the question.
Sapir allowed the environment to be broader, e.g. the morpheme X in all its occurrences, not
just the occurrence we happen to be considering at the moment. Thus, John Whitney could
ponder [dìní'] as meaning ‘It makes a sound’ and perceive a [t] because this technique
makes [dìnít'í] part of the relevant ‘environment’ along with [dìní']. Bloomfield confined the
process to difference in meaning among utterances, and Sapir extended the process to allow
what the utterances meant, not just that they were different.
 Chapter 9

American Structuralism:
Responses to Psychological Reality

“The term phoneme has been used

by a few linguists to refer to what
I
should call the psychological
correlate of the phoneme, but this
type of definition has been losing
ground ...” (Swadesh 1935:248)

1. Introduction
Patterns which are manifest in phonetic variety are attributed to several
distinct portions of language. We recognize these now by the names:
morphophonemics, phonemics/phonology, and phonetics. Morpho-phono-
logy was first thought to include (Swadesh 1934:128)

... in addition the study of phonemic structure of morphemes, the study of

interchange between phonemes as a morphologic [Emphasis mine, PWD]
process.

There is a second interchange between phonemes which is not morpho-

phonological; they constitute word variants (Swadesh 1934:118-119).
Swadesh provides a taxonomy of such variants: free variants and conditional
variants (“determined by position in the sentence”).1 The latter may be
particular (e.g. English a ~ an) or general; and the general kind may be
phonetically or structurally conditioned (Swadesh 1934:119):

Structurally conditioned, e.g. Tunica disyllabic words of the form CV V have that
form only when spoken in isolation; in context they become CV as: ri i ‘house’,
context form ri.

1 Free variants include variation of the familiar sort, e.g. economics / / and
/ /.
 Swadesh (1934:123) provides the following as statement as the first of his
‘methods’ for inductively discovering the phonemes of a language:

1. The criterion of consistency of words. Except for word variants [such as the
Tunica example, PWD] different occurrences of the same word have the
same phonemes.

By this characterization, Tunica [ ] and [ri] represent two distinct phonemic

sequences. In contrast to the Tunica example, different occurrences of the
same word will have the same phonemes, e.g. Russian [ ] ‘seven’ and
[ ] ‘seven [gen.]’ where the variants ‘seven’, [ ] and [ ], are
conditioned by the position of stress, and also [drúk] ‘friend’ and [drúga]
‘friend [gen.sg.]’, where the variation [drúk] ~ [drúg] is conditioned by the
absence of a following vowel.
Where the conditioning contexts of sound shapes of the morphemes are
grammatical and within the word, the variation is within morphopho-
nology. The alternation of /f/ with /v/ in the plural shape of /liyf/ leaf
identifies this /f/ as distinct from the /f/ in / / cuff, which does not vary in
this way (Swadesh 1934:129):

Whether it is a convenient fiction or a true reflection of linguistic psychology,

morphological processes are usually described as having a definite order. Leaves
is taken to be a secondary formation from leaf, and in consequence v is the
mutation of f and not f of v. But f does not always change to v in the
morphological process of plural formation; thus, we have cuff, cuffs. The f of cuff
is therefore morphologically different [Emphasis mine, PWD] from the f of leaf,
though phonemically it is the same entity. Morphologically, we have two f’s so
that f1 : v :: f2 :f. Morphologically distinct phonemes are called morpho-
phonemes.2

2 With the introduction of “morphophonology” into the terminology, there is already some
indecision as to where it should apply. Cf. Figure 1. Swadesh (1934) follows Prague school
usage in using it to label such variation as /liyf/ ~ /liyv/, but he (1935:249) also classifies the
Sarcee example of [dìní'] /dìní/ ‘this one’ and [dìní'] /dìnít/ ‘it makes a sound’ as an instance
of morphophonemics, writing that it “seems actually to demonstrate a morpho-phonological
rather than a phonemic difference”. It is not morphophonological by Swadesh’s (1934)
characterization of morphophonology. If the example is not one of phonemics, it would
appear to parallel Swadesh’s Tunica example, so that [dìní'] ‘it makes a sound’ and [dìnít'i]
‘he who makes a sound’ constitute word variants. That is, in place of the representation being
/dìnít'/ and /dinít'i/, it is /dìní/ and /dinít'i/. The only difference between this example and one
such as the Russian /rab/ ‘slave’ is that one of the allophones of Sarcee /t'/ would be [ ] ...
silence.
 RESPONSES TO PSYCHOLOGICAL REALITY 3

We have with these four examples (Russian [ ], Russian [drúk]

‘friend’ and [drúga] ‘friend’, Tunica [ri i] ~ [ri], and English /liyf/ ~ /liyv/)
four instances of variation and four instances of pattern. For Swadesh, certain
of that variation will be ‘sub-phonemic’. That is, the variation will be
contained within phonology and described by the method of complementary
distribution (Swadesh 1934:123):

4. The criterion of complementary distribution. If it is true of two similar types

of sounds that only one of them normally occurs in certain phonetic
surroundings and that only the other normally occurs in certain other phonetic
surroundings, the two may be sub-types of the same phoneme.

The Russian ‘seven’ example and the Russian ‘city’ example fall into this
category. The Tunica ‘house’ and the English leaf/leaves examples exhibit
phonological distinctness, but the second also falls into the class of morpho-
phonology.3
Figure 1 summarizes that array of variation and the interpretation of it. It
is the adjustment of the boundaries within Figure 1 which become the subject
of debate. Where is the boundary of pattern attributable to allophonics or
phonemics? Where is the boundary between phonemics (same representations
of words or morphemes) and morphophonology (different representations of
words or morphemes)? The issues which figure in that debate are the
invocation of psychological (‘mentalistic’) data/criteria and the imposition
of purified technique upon the theory.

VARIETY TYPE OF PATTERN

Phonemic
Tunica [ i] and Ø Word Variants
English [f] and [v] Morphophonology

Russian [d] and [t] Phonology

Russian [é] and [ ] Phonology

Figure 1: The boundaries between types of pattern.

3 The Maori example of awhi ‘embrace [active]’ ~ awhitia ‘embrace [passive]’, although it
involves an alternation of [t] with Ø as the Tunica involves an alternation of [ i] with Ø,
would be considered differently. The Maori difference would be a phonological sameness
on the order of the Russian [d] ~ [t] in ‘city’ because the conditioning environment is not
“position in the sentence” (Swadesh 1934:119), but its phonetic position in the word, i.e., the
presence of following morphological (bound) material and not word (free) material.
4 HALT

2. Some Responses to Psychological Approaches to the Phoneme

At the same time that Sapir, Swadesh, and Trager are describing languages
with an appeal to perceptual units and invoking psychological reality, others
are labeling those approaches as invalid. Twaddell’s (1935a) reluctance to
accept a psychological interpretation of the phoneme lies in his belief that
they “fail to meet the requirements of methodological feasibility” (Twaddell
1935a:9). The same concerns — techniques — which prompted Swadesh to
the statement of the phonemic principle compel Twaddell as well to comment
on the problem; but rather than coming to the problem from the starting point
of language description, Twaddell (1935a:18) addresses the issue in terms of
the problem noted earlier in Bloomfield’s Assumption 1, namely, that
“Within certain communities successive utterances are alike or partly alike”:

... physical events [of the relevant sort] are the material of linguistic study,
however, and accordingly the first step in linguistic study must be the establishing
of criteria for determining which of the innumerable physical differences are also
significant linguistic differences. Until those criteria have been established, the
use of the terms ‘linguistically same’ and ‘linguistically different’ is unjustified.

Twaddell is more purely theoretical and comments (Twaddell 1935a:9)

critically upon Sapir’s (and others’) psychologizing:

Such a [‘mental’] definition is invalid because (1) we have no right to guess about
the linguistic workings of an inaccessible ‘mind’, and (2) we can secure no
advantage from such guesses.

The issue here can be examined from two perspectives:

(i) whether a speaker hears objectively distinct sounds as same or

different,
(ii) whether a speaker hears objectively same sounds as same or
different.

Sapir’s speaker of S. Paiute exemplifies the first sort (i.e., [p] and [ ]), and his
speakers of Sarcee and Nootka exemplify that of the second sort (i.e., Sarcee
[i'] as i or it' and Nootka [s·] as s or as ss).4 In the first case involving S. Paiute
[p] and [ ], Twaddell (1935a:11) is troubled because “Sapir is obliged to
present negative evidence”; that is, the speaker fails to distinguish the two.

4 Of course, ‘distinct as same’ and ‘same as distinct’ are the interesting cases and the ones
focussed on.
 RESPONSES TO PSYCHOLOGICAL REALITY 5

What Twaddell (1935a:12) would prefer is positive evidence of the identity

of [p] and [ ]:

If Sapir’s Paiute guide had been trained to record one variety of labial consonant
as [p]; if he had been requested to pronounce all other varieties of ‘the same
sound’; if he had then without hesitation and without duplication responded with
examples of all these other varieties –– we might then concede that he appeared to
have some mental concept of a unified [p]-phoneme which was variously
actualized in his speech.

But the ‘sameness’ requires precisely that a speaker cannot accomplish the
task Twaddell describes; that is what a ‘perceptual unit’ means.5 It is a unit
(unity). Twaddell’s reservations concerning the examples of (ii) — Sarcee and
Nootka — refer to the information which the speaker exploits for his response
of same-as-different (Twaddell 1935a:13):

In so far as this incident [the Sarcee speaker’s assuring Sapir that “dìní ‘celui-ci’
and dìní ‘cela fait du bruit’ ... ‘étaient totalement differents’”] may be interpreted
as evidence of any mental reality, it would appear to be a morphological class or
lexical [Emphasis mine, PWD] unit than any phonetic or quasi-phonetic class or
unit.

The Nootka example is similarly discounted (Twaddell 1935a:13):

Similarly, Alex Thomas’s transcription of his native Nootka distinguished

between long consonants as positional variants of short consonants and as the
product of combination of similar final and initial consonants of separate lexical
elements. This practice, too, appears more appropriately referred to lexical
[Emphasis mine, PWD] than to quasi-phonetic consciousness, in terms of mental
reality.

And in summary (Twaddell 1935a:14):

In short, until positive and unambiguous evidence of the mental reality of

phonemes can be adduced, it appears methodologically dangerous to define the
phoneme in terms of mental reality.

Notice that there is no evidence to contradict Sapir’s data. It is a matter of

5 The proposed test is one that could only be seriously entertained by someone who has never
engaged in ‘field linguistics’. Twaddell was trained in German and in linguistics (Ph.D. 1930
from Harvard). He seems never to have encountered a non-Indo-European language in a field
situation (Hill 1983).
6 HALT

whether those data in themselves suffice. Sapir is not wrong; he is not right
enough.
Twaddell is equally skeptical of Bloomfield’s attempt to capture the
notion of phonological sameness in terms of a minimum same of physical
phonetic feature (Twaddell 1935a:23 & 24):

If such features existed, the determination of the phonemes of a given language,

and the definition of the phoneme, would be achieved ... That we do not find any
such constant, characteristic fraction is of course a commonplace of experimental
phonetics ... and there is no reason to believe that it will be.

Twaddell is not completely negative since he proposes an alternative

characterization of the phoneme as a fiction justifying this as a way to provide
a “terminological convenience” while avoiding “the promissory notes of the
laboratory” (Twaddell [1935a:33] referring to Bloomfield) and avoiding also
“the embarrassment of having in our discipline a fundamental unit which is
undetermined and the nature of which is a matter of wide disagreement among
linguists” (Twaddell 1935a:34). In a series of eleven statements, Twaddell
(1935a:38ff) defines a microphoneme and a macrophoneme. The micro-
phoneme applies to paradigmatic contrasts within one position. Using a
supposed series of minimally different forms, and “The term of any minimum
phonological difference among forms is called a MICRO-PHONEME” (Twaddell
1935a:38). Should a second (or additional) class of forms be found such that
they bear a one-to-one relation to the first and also such that the “qualitative
articulatory differences among the corresponding phonetic events are similar”
(Twaddell 1935a:38), then “the sum of all similarly ordered terms (micro-
phonemes) of similar phonological differences among forms is called a
MACRO-PHONEME” (Twaddell 1935a:39). In order to establish the series of
forms which enter into the paradigm for comparison, Twaddell (1935a:41)
must move beyond the phonetic uniqueness of each utterance to reach a
sameness of form:6

The phonetic events ‘Light the lamp’ as produced by two different individuals are
objectively very different; in so far as those events evoke similar responses in

6 Utterances belong to speaking, and forms belong to language (Twaddell 1935a:40):

We must operate with the abstracted forms and their relations but these
forms are not themselves susceptible of operation. Accordingly we
observe the utterance-fractions which correspond to a form; we study the
abstracted form which corresponds to utterance-fractions.
 RESPONSES TO PSYCHOLOGICAL REALITY 7

similar social situations [Talk about promissory notes! PWD], the two events are
phonetically significantly alike.

3. The Consequences of Technique

The effect of this alternative is twofold. First, it creates for the first time
an abstraction, which does not necessarily exist in any piece of the data. It is
a descriptive convenience; it is neither phonetic (as Bloomfield would have it)
nor psychological (as Sapir and others would have it). The macrophoneme
(the closest to the notion of the phoneme) is “an abstraction: ... [a] sum”
(Twaddell 1935aa:39).7 Second, because the macrophoneme is a sum
(Twaddell 1935a:49):

What occurs is not a phoneme, for the phoneme is defined as the term of a
recurrent differential relation. What occurs is a phonetic fraction or a
differentiated articulatory complex correlated to a micro-phoneme. A phoneme,
accordingly, does not occur [Emphasis mine, PWD]; it ‘exists’ in the somewhat
peculiar sense of existence that a brother, qua brother, ‘exists’ — as a term of a
relation.8

Concern with practice, which initially colored the approach to language, has
become technique, a monster, taking the theory further from the data, ending
here with a kind of instrumentalism. Continued emphasis upon technique
results in its refinement and in its increased consistency and ‘logicalness’, but
often with unwanted results. Cf. Pa’d go below.
Beginning with what one can observe, i.e., phonetics and some
information concerning stimulus-response correlations, one begins a series of
abstractions. And one must be logical in this. A technique cannot employ
what it does not have available. Joos (1958:96) comments:
It was the present article by Bloch [Bloch 1941] that made clear, as it never had

7 This directly contradicts Sapir’s (1933/1951.22) assertion: “... no entity in human

experience can be adequately defined as the mechanical sum or product of its physical
properties”. In 1941, Bloch is also characterizing the phoneme as a class (Bloch 1941:278).
Hockett (1942:9) declares that “A phoneme is a class of phones determined by six criteria”.

8 This is not what Bloomfield had in mind. Compare Twaddell’s 1935a statement about
‘brother’ with Bloomfield’s then recent (1933.80) claim about ‘redness’:

The phonemes of a language are not sounds but merely features of sounds [emph.
mine, PWD] which the speakers have been trained to produce and recognize in the
current of actual speech-sound — just as motorists are trained to stop before a red
signal, be it an electric signal-light, a lamp, a flag, or what not, although there is no
disembodied redness apart from these actual sounds.
8 HALT

been before, that phonemics must be kept unmixed from all that lies on the
opposite side of it from phonetics: kept uninfluenced by the identities of the items
of higher rank [Emphases mine, PWD] (morphemes and so on) which the
phonemes ‘spell’, and hence free from all that their identities entail, such as their
meaning and their grammar.

And Hockett (1942:20 & 21) further narrows the purview of phonology:

No grammatical fact of any kind is used in making phonological analysis ... There
must be no circularity; phonological analysis is assumed for grammatical analysis,
and so must not assume any part of the latter. The line of demarcation between the
two must be sharp.

Why is it that “The line of demarcation between the two must be sharp”? Why
must phonemics “be kept unmixed from all that lies on the opposite of it”?
And how can it be that meaning and grammar are “on the opposite side” from
phonetics? What sense is there in that? Where did this idea of “opposite side”
come from? Clearly, the model implied by Bloomfield’s work has been
altered. It is all base technique driven to be “consistent”.
Circumstance forces us to consider phonetics first because that is what
we first encounter with an unknown language. Correct technique (‘scientific’
technique) forces us to describe phonetics and phonology without reference to
morphology, grammar, semantics, etc. The process of analysis is linear, one
step at a time; therefore language must be constituted to reflect this linearity,
and a hierarchy of sorts is created. Having done phonology, we can then
safely move to analysis of grammar, perhaps morphology. The constant and
“same” reactions of Sarcee speaking John Whitney and S. Paiute speaking
Tony are disregarded, and the practice of consistent analysis forces us to begin
with what we perceive ... phonetics ... and pretending that neither we nor the
speaker knows anything more than the phonetics, we must complete the
phonemic analysis before we proceed further. The conditions in which we
practice linguistics, our concern with description and technique, and our
concern with being consistent so that the results of our description can be
confirmed (or rejected) move us from the non-hierarchicalized concept of
language of Bloomfield (the patterns of form, construction, and phoneme) and
away from the psychological concept of Sapir to one which is abstract and
separate from either sort of our data. Cf. Figure 2.
The earlier, and more interesting, problems concerning the psychological
reality of the phoneme are now redrawn in a logico-formal mode. Unlike
Sapir marveling at John Whitney’s ability to hear differences which Sapir
cannot, Bloch (1941) worries whether there is a complete overlap in the anal-
 RESPONSES TO PSYCHOLOGICAL REALITY 9

Stimulus-response relations

{Morphemes}

/Phonemes/

Phonetics
Figure 2: The beginning of layering of patterns of language.

ysis. Cf. Figure 3. Partial overlap is permitted because the distinct phonetic
environments of the phones [x1] and [x 2] allow them to be distinguished

/A/ /B/

[x1] [x2]

Partial Overlap

/A/ /B/

[x1] [x2]
Complete Overlap

Figure 3: Partial and complete overlap.

without regard to grammar or meaning. There is no circularity in partial

overlap. Complete overlap has no such resolution, and thus must be
condemned. To resolve complete overlap, i.e., to assign [x1] and [x2 ] to non-
same entities, requires some way of recognizing when we have either [x1 ] or
[x2]. The phonetic environment is not available for this decision because it is
the same in both occurrences. Then we must rely upon non-phonetic criteria,
e.g. John Whitney’s Sarcee reliance on knowing when he was confronted with
utterances meaning ‘this one’ or it makes a sound’. And with the rejection of
10 HALT

complete overlap go Sapir’s analyses of Sarcee and Nootka, and Trager’s

analysis of the Russian consonants, all cases of complete overlap.

4. Some Outcomes
There are two further effects of note to follow from this change. First, the

VARIETY TYPE OF PATTERN

Tunica [ i] and Ø Morphophonology

English [f] and [v] Morphophonology

Russian [d] and [t] Morphophonology

Russian [é] and [ ] Phonology

Figure 4: A revision of types of pattern.

alignment of the patterns as drawn in Figure 1 are changed. Cf. Figure 4.

Second, the refinement in method has some unwanted results. The English
vowel system has regularly short and long allophones of the stressed vowels
based on the voiced quality of the following consonant (Bloch 1941:283-84):

(1) (a) cot /kat/ [ ]

(b) cod /kad/ [ ]

(2) (a) sat /s t/ [ ]

(b) sad /s d/ [ ]

(3) (a) nought /nót/ [ ]

(b) gnawed /nód/ [

/ [ ]
(b) bud /b d/ [ ]

and so forth. Bloch personally had in his speech a series of words

bomb, bother, and sorry
 RESPONSES TO PSYCHOLOGICAL REALITY 11

which differ vocalically from

balm, father, and starry

only by the fact that the vowel in the second group is longer than the vowel in
the first. The long vowel of balm, father, and starry reappears in

alms, palm, pa, star, and card

The difficulty arises in these utterances:

(5) The pod grows.

(6) Pa’d go (if he could).

Pod and Pa’d are both [ ], yet the previous statements incline us to write
pod as /pad/ (cf. [1] - [4]) and Pa’d as /pa:d/ (cf. alms, palm, pa, star, and
card): a case of complete overlap. What to do? Bloch’s solution appears to be
this (Bloch 1941:284):

We are left, then with the other alternative. By classifying the vowel of pod —
and consequently also the vowels of rob, nod, bog, fond, and the like — as
members of the phoneme of balm, we destroy the neat parallelism of the pairs bit
bid, bet bed, bite bide, pot pod: the words in the last pair, instead of exhibiting
shorter and longer allophones of the same phoneme, have totally different
phonemes. But by sacrificing this symmetry we are able to account for all the
facts of pronunciation, which is sure the more important requirement
[Emphases mine, PWD]. The resulting system is lopsided; but the classes it sets
up are such that if we start from the actual utterances of the dialect we can never
be in doubt of the class to which any particular fraction of utterance must be
assigned.9

Thus,

(7) (a) cot /kat/ [ ]

(b) cod /ka:d/ [ ]

(8) (a) sat / / [ ]

(b) sad / / [ ]

9 And we have become the linguistic equivalent of accountants.

12 HALT

etc.

Not only have we replaced the substance of language with methodology, we

have lost all sense of proportion!

[Version: October 4, 2005]

 Chapter 10

American Strucutralism:
Preeminent Methodology

1. Introduction
From the period of 1933 (Bloomfield and Sapir) to 1942 (culminating with
Hockett 1942), we have seen preoccupation with practical methodology and
with the derivation of theoretical notions such as the phoneme control the
discussion of language to the point where analytic technique now dominates.
Method is no longer the application (guide and refiner) of the theory; method
is the theory.

2. Method as Theory
I abstract here as PHONETIC TRANSCRIPTION , OBSERVATIONS, and
PHONOLOGICAL ANALYSIS what would have been considered ‘good’
linguistic practice with regard to phonology at the period following its
codification by Hockett (1942). We may call this the period of Classical
American Structuralism.1

PHONETIC TRANSCRIPTION. Objective. Elements are phones. Source:

the trained ear. Employs a phonetic alphabet. Indicated by square
brackets.

OBSERVATIONS about phonetic transcription, about pairs of phones.

Contrast: Two distinct phones are in contrast if (i) they occur in

utterances such that their respective phonetic environments are
similar (If we say ‘identical’, then we identify the two utterances

1 ... with respect to phonology. Other aspects, e.g. morphophonemics, morphology, and
syntax, follow later. Hockett (1942:9) provides a more schematic characterization in terms of
six criteria for determining whether phones constitute a phoneme: similarity, non-
intersection, contrastive and complementary distribution, completeness, pattern congruity,
and economy.
2 HALT

as constituting a minimal pair.) and (ii) the glosses of the

utterances are distinct.‘Similar’ means that we cannot
characterize the difference between the environments, although
we may be able to list the differences. Listing does not suffice.
Complementary distribution: Two distinct phones are in
complementary distribution if they occur in utterances such that
their respective phonetic environments are not similar; that is, we
can characterize the difference between the environments.
Free variation: Two distinct phones are in free variation if they
occur in utterances such that their respective phonetic
environments are identical, but the glosses of the utterances are
not distinct.
Phonetic similarity: Two distinct phones that share more phonetic
properties are more similar phonetically than two which share
fewer. You determine this by knowing phonetics, i.e., how sounds
are articulated. Phonetic similarity is a continuum of similarity.

PHONOLOGICAL ANALYSIS (conclusions based on the OBSERVATIONS).

Stage I. Any pair of phones that are related by free variation and/or
by complementary distribution and which are phonetically similar
are to be considered members (allophones) of the same class
(phoneme). Definition: Phoneme — a class of sounds/phones
which are in free variation and/or complementary distribution and
which are phonetically similar. The phonetic members of the
phoneme are allophones. Phonemic notation is by slant line
(solidi), i.e., / /; allophonic, like phonetic, notation is by square
bracket, i.e., [ ].
Any pair of phones related by contrast (in minimal pairs or not)
belong to separate phonemes, and the phonetic property(ies) that
distinguish(es) them is/are phonemic or distinctive. For
example, voice is phonemic/distinctive for English obstruents.
Stage II. If Stage I yields more than one possible grouping into
allophones, choose the grouping that yields the simplest statement
of phonotactics; that is, maximize pattern congruity.

The application of these techniques to a corpus of data from any language

whatsoever should result in the phonological description of them. Note ‘the’,
not ‘a’. The purpose of a scientific description is eliminate arbitrariness from
 PREEMINENT METHODOLOGY 3

the description, and whenever the analyst is required to select the better (or
‘the’) description of the language, thus identifying its structure, the theory
reveals itself as less that maximally ‘scientific’. Hockett (1942:9) notes “the
danger of arbitrary procedure” and introduces the last two of his criteria (cf.
footnote 1) to avoid arbitrariness. The concern with this is shared by both
sides of the psychological – material disagreement.2 Swadesh (1935:1) states
his position:

Is phonemic procedure arbitrary? It can be, but I submit that it need not be. If one
defines a phoneme as ‘one of an exhaustive list of classes of sound in a language’,
one admits an endless variety of treatments and the choice of one or another
treatment is arbitrary. The ideal of exhaustiveness is not in itself sufficient to
define a unique scientific procedure. But if we also take the ideals of simplicity
and self-consistency, we have the basis for a non-arbitrary method ... To attain this
ideal [Emphasis mine, PWD], it is necessary to consider always the totality of
phenomena in the given language.

And Twaddell (1935b:57):

... intuition may or may not be ‘correct’; the fact remains that an intuition has no
business in a science [Emphasis mine, PWD], if we can get along without it.

In that the techniques are now mechanical applications, they should determine
without intervention of the linguist what the description of the language is
like.

3. Evaluation Procedures
There are two evaluation procedures within the theory. They are pattern
congruity and simplicity (or economy). But arbitrariness was not only the
result of ‘incomplete’ techniques; it could also arise from the conflict between
procedures. An example from Swadesh (1935:246) will illustrate this conflict
briefly:

Open syllables Closed syllables

[i] [
]
[ ] [ ]

2 There was a series of exchanges between Twaddell and Swadesh: Twaddell (1935a),
Swadesh (1935), Twaddell (1935b,which includes a final note with Twaddell recording some
interim comments of Swadesh from personal correspondence), and Swadesh (1937).
4 HALT

The simplest description to these data will recognize three contrasting

phonemes, but if we take phonetic similarity literally, then phonetic identity is
the maximum degree of phonetic similarity. The simplest phonological
description for the above data ignores the absolute phonetic similarity
between [ ] in open syllables and the [ ] in closed syllables and proposes a
phonemic system of three vowels. The criterion of simplicity overrides the
criterion of phonetic similarity. If we reverse the importance of the two
criteria, the description will require four vowels: /i/ [i ], /e/ [ / [ ], and / /
[ ]. If we insist on maintaining the description with three phonemic vowels, it
will require that we ignore phonetic parallelism placing [e] and [ ] together as
allophones. Such problems as this lead to the interpretation of phonetic
similarity in a relative way, not an absolute way.3
The notion of pattern congruity appears as early as Swadesh (1934:124):

The criterion of pattern congruity. Particular formulations must be congruous with

the general phonemic pattern of the given language. Thus, although Navaho i
(occurring only after consonants) and y (occurring only before vowels) are
complementary in distribution, they are nevertheless independent phonemes
because of the fact that Navaho is generally characterized by a sharp distinction
between vowel and consonant. (As a matter of fact, any vowel would be found to
be in complementary distribution to almost any consonant.) In another language,
non-syllabic and syllabic i might be positional variants of the same phoneme.

For Hockett (1942:9), pattern congruity is one of six criteria for determining
whether a class of phonemes constitute a phoneme (“a class of phones”): 4

3 Recall Twaddell’s (1935a:38-39) use of ‘phonetic ordering’ of micro-phonemic classes and

the comparison of these phonetically ordered micro-phonemic classes with other to
determine the macro-phonemic sums. For Twaddell, the example from Swadesh would
exhibit phonetic ordering of ‘high’, ‘mid’, and ‘low’ for both micro-phonemic systems. The
relevant comparison would be between the ‘high’ vowels of each (i.e., [i] and [ ]), the ‘mid’
vowels of each (i.e., [e] and [ ]), and the ‘low’ vowels of each (i.e., [ ] and [ ]). “Only
similarities in the relations can be valid; not phonetic similarity of the units themselves”
(Twaddell 1935b:58).
In a final comment in Twaddell (1935b:59), Twaddell records a suggestion from Swadesh
proposed to him in correspondence:

He [Swadesh] proposes not absolute, but relative phonetic similarity as a phonemic

criterion, and now offers a more unambiguous statement: Micro-phonemes that, on the
basis of their phonetic nature, occupy a like place in comparable series [matched one-
to-one, PWD] are equivalent.

They agree on this.

4 The relation of α-phonetics and β-phonetics to each other and the other terms of language
can be represented as follows:
 PREEMINENT METHODOLOGY 5

Pattern congruity: two contrasting -segments (single -sounds or groups of

such sounds) which occur in similar -phonetic environments are to be analyzed
as having similar structures; either they are both unit phones, members of different
phonemes belonging to the same functional class, or else they are similar clusters
of two or more phones. This statement emphasizes the importance of structural
parallelism: like function, like structure.

Any evaluation is retrospective. Thus, while one is in the act of analyzing

data, it is not possible to employ that criterion. This aspect of the criterion
prompts some, first Twaddell in 1935, and then Haugen & Twaddell again in
1942, to condemn its use (Twaddell 1935:55):

In Swadesh’s defence of his ‘criterion of pattern congruity’ (248), I miss any

attempt to justify the assumption that there is an a priori phonemic system or
pattern. Until he establishes such a system as existent, and existent apart from the
phonemic relations, I must continue to regard his ‘criterion of pattern congruity’ a
dangerously circular one to apply in the determination of particular phonemic
relations.

and (Haugen & Twaddell 1942:235)

The additional concept of ‘phonetic interrelationship’ or ‘pattern analysis’ ...

seems to be little more than a covert appeal to the system that is to be established
and therefore a circular argument.

Like reference to grammatical identity, reference to not-yet-known phonemic

system smacks of illegitimate circularity. Most, however, ignore Haugen and
Twaddell’s objection (although they still reject the use of grammar) in their
analyses.
One of the best known applications of pattern congruity is in the analysis
of English vowels by Trager and Bloch (1941). As might be expected, there
were competing analyses of English phonology, most notably by Swadesh
(1947), and by Pike (1947). The phonetic facts involved in the discussion are
organized into three sets and presented below as THE PHONETICS:

Morphology
| <–– Morphophonemics
Phonology
| <–– β-Phonetics
α-Phonetics

β-phonetics is phonetics seen from the perspective of phonology; α-phonetics is phonetics

objectively and without the bias of a phonological perspective.
6 HALT

THE PHONETICS.
I. [ ] pit [ ] put
[ ] pet [ ] putt
[ ] pat [ ] bought
[ ] pot

II. [ ] beat [ ] boot

[ ] bait [ ] boat
[ ] bite [ ] bout

III. [ ] bee [ ] boo

[b ] bay [ ] bow
[ ] buy [ ] bow

PHONEMICIZATION I (Trager & Bloch 1941).

/i/ [ ] /u/ [ ]
/e/ [ ] /o/ [

/i/ [ ]
/e/ [ ] /o/ [

Swadesh (1947:146) questions whether the /y/ and /w/ of Trager and Bloch
behave like consonants in positions other than in the monosyllabic frame CVC

5 Bloomfield (1935:100-101) comes to a similar conclusion for his description of Central-

Western English.
 PREEMINENT METHODOLOGY 7

and he cites these data:6

... it is necessary to ask if the diphthongal second elements are like consonants in
still other positions. If the following syllable is unaccented and the preceding is
accented, it has a different treatment, one that had been described by Trager and
Bloch as ambisyllabic ... In any event there is a difference, which we can
symbolize by the placing of the hyphen, as between spi-túun, di-lém- , ti-r n- -
kl, si-nít-ik, d -kèid, hí-tàit, sé-rèit (spitton, dilemma, tyrannical, Sinitic, decade,
Hittite, serrate) and pép- r, wík- r, míl- t, pén-ii (pepper, wicker, millet, penny).
The diphthongal elements follow the second type of treatment (with preceding
syllable or indeterminate) even before an accented vowel: kei-áat-ik, mii- n-d r,
dai- , pou-ét-ik, duu-ét (chaotic, meander, Diana, poetic, duet). Thus they are
seen to divide in a manner that is clearly diferent from consonants.

They pattern syllabically unlike spi-ttoon, di-lemma, de-cade, and Hi-ttite.

Treating the data of PHONETIC DATA II (‘long vowels’) and PHONETIC DATA
III (‘diphthongs’) as unit vowels is also criticized (Swadesh 1947:147-48):

... It cannot be said that the unit treatment is basically wrong or inadequate. Yet
there is a difference, which can be described as one of simplicity. The analytic
treatment reduces the syllabics to a smaller number of primary units –– between
five and nine, instead of between fourteen and more than twenty. Moreover, the
basic units are relatively uniform instead of falling into two fundamentally
different classes (short-bound and long free).

PHONEMICIZATION III (Pike 1947).

/i/ [ ] /u/ [ ]
/e/ [ ] /o/ [

Pike’s alternative relies upon native speaker perception of the diphthongs and
the reproduction of pure long vowels by these speakers (Pike 1947:151):

... my colleagues and I have observed the following fact: It is relatively easy to
teach these students to notice that [a ] is phonetically composed of two parts –– a

6 Swadesh replaces the symbol / / with / /.

8 HALT

vowel somewhat similar to the [a] of father and an [ ] somewhat similar to the [ ]
of bit or the [i] of beet; [ ] and [ ] act like [a ] ...
On the other hand, practically without exception, the students have
considerable difficulty in learning to recognize two elements, or a glide, in the
[ ] of boat, toe, and similar words ... A similar situation exists for the [ ] of bait,
may, and the like, except that the difficulty is possibly a bit greater ... For [ ] and
[ ] as in beet and boot, even greater difficulty exists for the students, presumably
because the diphthongization is phonetically less pronounced than for [ ] and
[ ].

Additionally, in precontour (or pendant) position the diphthongization can

be lessened or lost and a pure vowel appear (Pike 1947:155): 7

(1) (a) [ ] It’s a «bike that I want.

3- ˚2- -4//
(b) [ ] The bike that «you have is the «best.
3- ˚2- -4-3 3- ˚2-4//

(2) (a) [ ] You say you saw the «bout?!

3- ˚2- -4//
(b) [ ] The bout for the «championship will come next «month.
3- ˚2 -4-3 3- °2-4//

(3) (a) [ ] It’s «bait that I want.

3- ˚2 -4//
(b) [e] The bait is «spoiled.
3- ˚2- -4//

Pike (1947:155) notes the following of such examples:

From this evidence we draw the following conclusions. (1) The set [ ], [ ], [ ]
acts differently from the set [ ], [ ], [ ], [ ] in that the first set retains its
strongly diphthongal character even in the rapid part of the intonational contour,
while the second set tends to lose most of its diphthongal character in such a
position. (2) In the set [ ], [ ], [ ] , each sound is a sequence of two units, of
which the second does not disappear even in rapid pronunciation; but in the set
[ ], [ ], [ ], [ ], each sound is structurally a single unit (phonetically

7 “Primary contours begin on ... syllables ... as indicated by the degree sign [i.e., º] under
them ... The numbers indicate relative pitch of the voice. Pitch 1 is the highest, pitch 4 is
lowest, pitches 2 and 3 are intermediate. Syllables which have no number beneath them need
not occur on one of these relative levels but are more or less evenly distributed between the
pitches preceding and following them” (Pike 1947.153).
 PREEMINENT METHODOLOGY 9

complex), which may be modified according to the intonational environment in

which it occurs. (3) The difference in students’ reactions to the two sets is
substantiated by the fact that these sets show different ranges of variation in the
degree to which they remain diphthongal in various intonation contours. (4) Since
the configurational reality of the difference between the two sets is attested by the
occurrence of different variants in similar contexts and by different native
reactions, it appears that the latter type of evidence [Emphases mine, PWD] can
be legitimately used to support other evidence for grouping sounds in a linguistic
system: namely, the manner in which speakers of one language react to the sounds
of a second language in their attempts to hear, pronounce, and record them, is
valuable and valid evidence to be considered in analyzing the phonemic system of
the first language.

This is just the kind of psychologizing that has been condemned by Twaddell
and others. The generally accepted system is that of Trager and Bloch.

4. Conclusion
By 1942, reliance on a speaker’s subjective reaction as earlier advocated
by Sapir, Swadesh and others is now not the accepted way to do phonology.
There is consensus on the use of the objective (hence, more scientific) notions
of contrast, complementary distribution, free variation, and phonetic
similarity. They are the way to go. However, a second or third, etc. application
of the techniques to the same data may produce alternative conclusions.
Assuming that only one can be “correct”, some have to be discarded.
Everyone agrees that it is not scientific to select arbitrarily among the
competing descriptions, and the debate turns on the elimination of
arbitrariness, i.e., how to finalize the description where the techniques yield
more than one. Refining the techniques, e.g. the idea of relative phonetic
similarity, is one way of reducing the number of possible descriptions.
Retrospective comparison of competing analyses with respect to some
evaluative criteria is another, and one that is standard in other sciences. The
fact that some linguists could react to evaluation as if it were “circular” is a
measure of the degree to which sanitized technique has substituted for an
understanding of language

[Version: October 5, 2005]