ED. /PSYCH.
LIB.
P
221
J21f
cop, JAKOBSON
FUNDAMENTALS OF
LANGUAGE
THE LIBRARY
OF
THE UNIVERSITY
OF CALIFORNIA
LOS ANGELES
:cl/Psych
.ib.
ROMAN JAKOBSON AND
MORRIS HALLE
O
Fundamentals
of
Language
MOUTON & CO - 'S-GRAVENHAGE
';!?T
FUNDAMENTALS OF LANGUAGE
JANUA LINGUARUM
STUDIA MEMORIAE
NICOLAI VAN WIJK DEDICATA
cdenda curat
CORNELIS H.VAN SCHOONEVELD
LEIDEN
NR.l
1956
MOUTONi&CO •
S-GRAVENHAGE
FUNDAMENTALS OF
LANGUAGE //
by
ROMAN j£AKOBSON
HARVARD UNIVERSITY
and
MORRIS HALLE
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
1956
MOUTON&CO •
'S-GRAVENHAGE
Printed in the Netherlands by
Mouton & Co., Printers, The Hague
Liorary
FOREWORD
The Gate of Languages (Janua linguarum) is indeed an appropriate
title for a series of essays seeking the key to the laws that govern
language and its relationship with other social institutions. This
name appeals to me, moreover, as a link that connects the modern
search with the writings of Johann Amos Comenius, the great
humanist thinker in the science of language. His works, like many
Greek and Latin treatises from the Stoa to the Cartesian epoch,
carry numerous fruitful ideas which now again capture the attention
of linguists.
The title of the series refers, furthermore, to the recent past of our
science. Nicolaas van Wijk, whose name heads this set of essays,
was one of the outstanding pioneers in the inquiry into the structure
of language and into the principles of its evolution. The subtitle of
his book Phonologie - "een hoofdstuk uit de structurele taalwetenschap"
(a chapter of structural linguistics) - may be applied to his
whole life's work. In 1902, as a twenty-two year old student at
Leipzig, he offered a bold contribution 'Zur relativen Chronologic
der urgermanischen Lautgesetze', published in Paul-Braune J?e//rflge
zur Geschichte der deutschen Sprache und Literatur, XXVIII, in
which he displays a clear insight into the coherence of sound
patterns and their mutations, and some twenty years later he took up
and elaborated these views in his first original work in comparative
phonology, 'Een phonologiese parallel tussen Germaans, Slavics
en Baltics', Mededeelingen der Koninklijke Akademie van Wetenschappen,
Afd. Letterkunde, deel 77-79, serie A (1934-5). Van
Wijk, and there lies his main strength, never sacrificed the manifold
empirical data in favor of a speculative theory, nor did his amazing
mastery of the concrete philological material conceal from him the
theoretical corollaries.
I am particularly glad to inaugurate the series of essays dedicated
to the memory of this eminent Dutch linguist, since twenty five
years ago it was he who, along with Antoine Meillet, encouraged
1W7':'14
\ I
FORFWORD
in> first, modest attempts to grasp the structural laws of language
with respect to the factors of time and space {De nieune taalgids,
WIV. \XV). It is again to the author o^ Phonologic (1939) that
I feel deep gratitude for the first support of my initial efforts to
dissolve language into its ultimatecomponents,thedyadicdistinctive
features.
When a quarter of a century separates us from the Prague
International Conference, which broke the ground for general
phonology, it is appropriate to survey the main problems of this
discipline in its present stage. On the other hand, it was tempting
to explore, forty years after the publication of Saussure's Cours
with its radical distinction between the "syntagmatic" and "associative"
plane of language, what has been and can be drawn
from this fundamental dichotomy.
Leiden. October 1955 ROMAN JAKOBSON
TABLE OF CONTENTS
Foreword V
PARTI
PHONOLOGY AND PHONETICS
I The Feature Level of Language .... 3
1 . 1 Distinctive features in operation .... 3
L2 Structure of distinctive features .... 4
1.3 Opposition and contrast 4
L4 Message and code 5
1.5 Ellipsis and explicitness 5
II The Variety of Features and their Treatment
IN Linguistics 7
2.1 Phonology and phonemics 7
2.2 The "inner approach" to the phoneme in relation
to sound 8
2.3 Types of features 8
2.4 The "outer" approaches to the phoneme in relation
to sound: A. The mentalist view ... 11
2.42 B. The code-restricting view 12
2.43 C. The generic view 12
2.44 D. The fictionalist view 13
2.441 "Overlapping" of phonemes .... 14
2.45 E. The algebraic view 15
2.5 The cryptanalyst's and decoder's devices as two
complementary techniques 17
III—The Identification of Distinctive Features . 20
3.1 Syllable 20
3.2 Two kinds of distinctive features .... 22
3.3 Classification of prosodic features ... 22
3.31 Tone features 22
VllI TABLE OF CONTENTS
3.32 Force features
3.33 Quantity features
3.34 The interconnection between stress and length
3.4 Comparison of prosodic and inherent features
3.5 General laws of phonemic patterning
3.51 Restriction in the over-all inventory of distinctive
features
3.6 The two classes of inherent features .
3.61 Sonority features
3.62 Tonality features
3.7 Stages of the speech event
3.71 The use of different stages in the study of distinctive
features
3.72 Nomenclature of distinctive features
IV Phonemic Patterning
4.11 Stratification: nuclear syllable . . . .
4.12 The role of the nasal consonant .
4. 1 The primary triangle
4.14 The split of the primary triangle into two triangles,
'
consonantal and vocalic
4. 1 Patterning of oral resonance features
4.16 Sonority features in relation to the optimal consonant
and vowel
4.2 The dichotomous scale
4.3 The spatio-temporal pattern of phonemic opera-
tions
23
24
24
25
26
27
28
29
31
31
33
36
37
37
38
38
39
40
41
44
49
PART II
TWO ASPECTS OF LANGUAGE AND
TWO TYPES OF APHASIC DISTURBANCES
I Aphasia as a Linguistic Problem .... 55
II The Twofold Character of Language 58
III Similarity Disorder 63
IV Contiguity Disorder 7!
V The Metaphoric and Metonymic Poles. . 76
TABLE OF CONTENTS IX
SUPPLEMENT
Selected List of Studies in General Phonology
(1931—1955) 83
The first study is an expanded version of the paper to appear in the Handbook
ofPhonetics, initiated by the International Committee for Phonetic Sciences and
prepared for publication by the North Holland Publishing Company, Amsterdam.
The second study is based on the author's essay written for a collective
volume At the Beginning Was the Word, Harper, New York, and on a few
passages of his paper "Aphasia as a Linguistic Problem" from the symposium
On Expressive Language, Clark University Press, Worcester, Mass., 1955.
PART I
PHONOLOGY AND PHONETICS
BY
ROMAN JAKOBSON
AND
MORRIS HALLE
THE FEATURE LEVEL OF LANGUAGE
LI Distinctive features in operation. Family names such as Bitter,
Chitter, Bitter, Fitter, Gitter, Hitter, Jitter, Litter, Mitter, Fitter,
Ritter, Sitter, Titter, Witter, Zitter, all occur in New York. Whatever
the origin of these names and their bearers, each of these
vocables is used in the EngUsh of New Yorkers without colliding
with their linguistic habits. You had never heard anything about
the gentleman introduced to you at a New York party. "Mr.
Ditter," says your host. You try to grasp and retain this message.
As an English-speaking person you, unaware of the operation,
easily divide the continuous sound-flow into a definite number of
successive units. Your host didn't say bitter /bits/ or dotter /data/
or digger /digs/ or ditty /diti/ but ditter /dita/. Thus the four
sequential units capable of selective alternation with other units
in Enghsh are readily educed by the listener: /d/+/i/+/t/+/9/.
Each of these units presents the receiver with a definite number
of paired alternatives used with a differentiating value in EngUsh.
The family names, cited above, differ in their initial unit; some
of these names are distinguished from each other by one,
single alternative, and this minimal distinction is common to
several pairs, e.g. /nits/ r/dits/ = /mita/ : /bits/ = nasalized vs. nonnasalized;
/tit9/:/dit3/ = /sita/i/zits/ = /pits/ :/bit9/ =/kit3/:/git9/ =
tense vs. lax. Such pairs as /pita/ and /dits/ offer an example of
two concurrent minimal distinctions : grave vs. acute together with
tense vs. lax. The pair bitter /bita/ and detter /deta/ presents two
successive minimal distinctions : grave vs. acute followed by diff'use
vs. compact. ! For an acoustic and motor definition of the cited
distinctions, see 3.61 and 3.62.
1.2 Structure of distinctive features. Linguistic analysis gradually
breaks down complex speech units into morphemes as the
4 PHONOLOGY AND PHONETICS
ultimate constituents endowed with proper meaning and dissolves
these minutest semantic vehicles into their ultimate components,
capable of differentiating morphemes from each other. These
components are termed distinctive features. Correspondingly,
two levels of language and linguistic analysis are to be kept apart:
on the one hand, the semantic level involving both simple and
complex meaningful units from the morpheme to the utterance and
discourse and, on the other hand, the feature level concerned
with simple and complex units which serve merely to differentiate,
cement and partition or bring into relief the manifold meaningful
units.
Each of the distinctive features involves a choice between two
terms of an opposition that displays a specific differential property,
diverging from the properties of all other oppositions. Thus grave
and acute are opposed to each other in the listener's perception by
sound-pitch, as relatively low-pitched and high-pitched; in the
physical aspect they are correspondingly opposed by the distribution
of energy at the ends of the spectrum and on the motor level by
the size and shape of the resonating cavity. In a message conveyed
to the listener, every feature confronts him with a yes-no decision.
Thus he has to make his selection between grave and acute, because
in the language used for the message both alternatives occur in
combination with the same concurrent features and in the same
sequences: /bits/—/dita/, /fits/—/sita/, /bil/—/biil/. The listener is
obliged to choose either between two polar qualities of the same
category, as in the case of grave vs. acute, or between the presence
and absence of a certain quality such as voiced vj-. voiceless,
nasalized vs. non-nasalized, sharp V5. plain.
1.3 Opposition and contrast. Since in the listener's hesitation "Is
it /bits/ or /dita/?" only one of the two logically correlated alternatives
belongs to the actual message, the Saussurian term
opposition is suitable here, whereas the term contrast is rather
to be confined to cases where the polarity of two units is brought
into relief by their contiguity in sensory experience as, for instance,
the contrast of grave and acute in the sequence /pi/ or the same
THE FEATURE LEVEL OF LANGUAGE 5
contrast, but with a reversed order of features, in the sequence /tu/.
Thus opposition and contrast are two different manifestations of
the polarity principle and both of them perform an important
role in the feature aspect of language (cf. 3.4).
1.4 Message and code. If the listener receives a message in a
language he knows, he correlates it with the code at hand and this
code includes all the distinctive features to be manipulated, all
their admissible combinations into bundles of concurrent features
termed phonemes, and all the rules of concatenating phonemes
into sequences - briefly, all the distinctive vehicles serving
primarily to differentiate morphemes and whole words. Therefore,
the unilingual speaker of English, when hearing a name like /zita/
identifies and assimilates it without difficulty even if he had never
heard it before, but either in perception or reproduction he is prone
to distort, and to distrust as alien, a name such as /ktita/ with its
unacceptable consonantal cluster, or /xita/ which contains only
familiar features but in an unfamiliar bundle, or, finally, /myta/,
since its second phoneme has a distinctive feature foreign to
English.
1.5 Ellipsis and explicitness. The case of the man faced with
family names of people entirely unknown to him was deliberately
chosen because neither his vocabulary, nor his previous experience,
nor the immediate context of the conversation give him any clues
for the recognition of these names. In such a situation the hstener
can't afford to lose a single phoneme from the message received.
Usually, however, the context and the situation permit us to
disregard a high percentage of the features, phonemes and sequences
in the incoming message without jeopardizing its comprehension.
The probability of occurrence in the spoken chain varies for different
features and likewise for each feature in different texts. For
this reason it is possible, from a part of the sequence, to predict
with greater or lesser accuracy the succeeding features, to reconstruct
the preceding ones, and finally to infer from some features in a
bundle the other concurrent features.
6 PHONOLOGY AND PHONETICS
Since in various circumstances the distinctive load of the
phonemes is actually reduced for the listener, the speaker, in his
turn, is relieved of executing all the sound distinctions in his
message: the number of effaced features, omitted phonemes and
simplified sequences may be considerable in a blurred and rapid
style of speaking. The sound shape of speech may be no less elliptic
than its syntactic composition. Even such specimens as the slovenly
/tem mins sem/ for 'ten minutes to seven', quoted by D. Jones, are
not the highest degree of omission and fragmentariness encountered
in familiar talk. But, once the necessity arises, speech that is
elliptic on the semantic or feature level, is readily translated by the
utterer into an explicit form which, if needed, is apprehended by
the listener in all its explicitness.
The slurred fashion of pronunciation is but an abbreviated
derivative from the explicit clear-speech form which carries the
highest amount of information. For many American English
speakers /t/ and /d/ are ordinarily not distinguished between a
stressed and unstressed vowel but can be produced distinctively
when there is danger of a confusing homonymity: "Is it Mr. Bitter
/bits/ or Bidder /bids/?" may be asked with a slightly divergent
implementation of the two phonemes. This means that in one type
of American English the code distinguishes the inter-vocalic /t/
and /d/, while in another dialectal type this distinction is totally
lost. When analyzing the pattern of phonemes and distinctive
features composing them, one must resort to the fullest, optimal
code at the command of the given speakers.
II
THE VARIETY OF FEATURES
AND THEIR TREATMENT IN LINGUISTICS
2.1 Phonology and phonemics. The question of how language
utilizes sound matter, selecting certain of its elements and adapting
them to its various ends, is the field of a special linguistic discipHne.
In Enghsh this discipline is often called phonemics (or, puristically,
phonematics) since among the functions of sound in language
the primary one is to serve as a distinctive vehicle and since the
basic vehicle for this function is the phoneme with its components.
The prevailingly continental term phonology (launched in 1923
and based on the suggestions of the Geneva school),^ or the circumlocution
functional phonetics is to be preferred however,
although in English the label "phonology" frequently designated
other domains and especially served to translate the German Lautgeschichte.
The advantage of the term "phonology" might be
its easier application to the whole variety of linguistic functions
performed by sound, whereas "phonemics" willy-nilly suggests a
confinement to the distinctive vehicles and is an appropriate
designation for the main part of phonology dealing with the
distinctive function of speech sounds.
While phonetics seeks to collect the most exhaustive information
on gross sound matter, in its physiological and physical properties,
phonemics, and phonology in general, intervenes to apply strictly
linguistic criteria to the sorting and classification of the material
registered by phonetics. The search for the ultimate discrete
differential constituents of language can be traced back to the
sphota-docix'\n& of the Sanskrit grammarians"^ and to Plato's conception
of oxoixeiov, but the actual linguistic study of these
^ R. Jakobson, O ceSskom stixe (Berlin, 1923), pp. 21ff.
^ Cf. J. Brough, 'Theories of general linguistics in the Sanskrit Grammarians,'
Transactions of the Philosophical Society (1951).
8 PHONOLOGY AND PHONETICS
invariants started only in the 1870's and developed intensively after
World War 1, side by side with the gradual expansion of the principle
of in variance in the sciences. After the stimulating international
discussion of the late twenties and early thirties, the first attempts
to sum up the basic results of the research, Trubetzkoy's and van
Wijk's outlines of general phonology, appeared in 1939.^ The subsequent
theoretical and practical achievements in the structural
analysis of language required an ever more adequate and consistent
incorporation of speech sounds into the field of linguistics with its
stringent methodology; the principles and techniques of phonology
improve and its scope becomes ever wider.
2.2 The "inner" approach to the phoneme in relation to sound.
For the connection and delimitation of phonology (especially
phonemics) and phonetics, the crucial question is the nature of the
relationship between phonological entities and sound. In Bloomfield's
conception, the phonemes of a language are not sounds
but merely sound features lumped together "which the speakers
have been trained to produce and recognize in the current of speech
sounds - 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 signals."^
The speaker has learned to make sound-producing movements in
such a way that the distinctive features are present in the sound
waves, and the listener has learned to extract them from these waves.
This so-to-speak inner, immanent approach, which locates the
distinctive features and their bundles within the speech sounds, be
it on their motor, acoustical or auditory level, is the most appropriate
premise for phonemic operations, although it has been
repeatedly contested by outer approaches which in different ways
divorce phonemes from concrete sounds.
2.3 Types offeatures. Since the differentiation of semantic units
is the least dispensable among the sound functions in language,
* N. Trubetzkoy, 'Grundziige der Phonologic' = Tiavaux dii Cercle
Linguistique de Prague, VII (1939); N. van Wijk, Phonologic: een hoofdstuk
uit de structurele taalwetenschap (The Hague, 1939).
* L. Bloomfield, Language (New York, 1933), p. 79f.
THE VARIETY AND TREATMENT OF FEATURES 9
Speech participants learn primarily to respond to the distinctive
features. It would be deceptive, however, to believe that they are
trained to ignore all the rest in speech sounds. Beside the distinctive
features, there are, at the command of the speaker, also other types
of coded information-bearing features that any member of a speech
community has been trained to manipulate and which the science
of language has no right to disregard.
Con figurative features signal the division of the utterance
into grammatical units of different degrees of complexity, particularly
into sentences and words, either by singhng out these units and
indicating their hierarchy (culminative features) or by delimiting
and integrating them (demarcative features).
Expressive features (or emphatics) put the relative emphasis
on different parts of the utterance or on different utterances
and suggest the emotional attitudes of the utterer.
While the distinctive and the configurative features refer to
semantic units, these two types of features, in turn, are referred to
by the redundant features. Redundant features help to identify
a concurrent or adjoining feature, either distinctive or configurative,
and either a single one or a combination. The auxiliary role of
redundancies must not be underestimated. Circumstances may
even cause them to substitute for distinctive features. Jones cites
the example of the English /s/ and /z/ which in final position differ
from each other solely in the degree of breath force. Although
"an Enghsh hearer will usually identify the consonants correctly,
in spite of their resemblance to one another," the right identification
is often facilitated by the concomitant difference in the length of
the preceding phoneme: pence /"pensy -pens [pen:zj.^ In French,
the difference between voicelessness and voicing ordinarily accompanies
the consonantal opposition tense/lax. Martinet notes that
in an energetic shout the lenis /b/ matches the fortis /p/ in energy
so that a strong bis! differs from pisse! only through the normally
redundant feature voiceless/voiced.* Conversely, in Russian, the
difference between lax and tense is a redundant feature accompany"
D. Jones, The Phoneme: its nature and use (Cambridge, 1950), p. 53.
• Word, XI (1955), p. 115. Cf. R. Jakobson, C. G. M. Fant, M. Halle,
10 PHONOLOGY AND PHONETICS
ing the distinctive opposition voiced/voiceless, while under the
special conditions of whispering only the redundant feature
remains and takes over the distinctive function.
If the distinctive function of speech sounds is the only one under
analysis, we use the so-called "broad" or phonemic transcription,
that notes nothing but phonemes. In a Russian specimen
/pil,il/ '(one) spread dust', /i/ is an unstressed phoneme that includes,
furthermore, two distinctive features: in traditional articulatory
terms, /i/ is opposed to /a/ of /pal,il/ '(one) fired' as
narrow to wide and to /u/ of /pul,dl/ '(orie) took a pot shot' as
unrounded to rounded. The information load of the vowel analyzed
is, however, far from confined to its distinctive features, notwithstanding
their paramount relevance in communication.
The first vowel of /pil,il/ is a velar [m] in contradistinction to
the palatal [i] of /p,il,il/ '(one) sawed' and this difference between
back and front is a redundant feature pointing to the distinctive
opposition of the preceding unpalatalized (plain) and palatahzed
(sharp) consonant: cf. Russian /r,ap/ 'pitted' - /r,ap,/ 'ripple'.
If we compare the sequences /krugom pil,il/ '(one) spread dust
all around' and /ispompi l,il/ '(one) poured from a pump', we
observe that the syllable /pi/ in the second specimen contains a
^ore obscure (tending toward a brief, mid-central articulation)
variety of vowel than in the first sample. The less obscure variety
appears only immediately before the stressed syllable of the same
word and thus displays a configurative feature: it signals that no
word boundary follows immediately.
Finally, /pil,il/ may be uttered with a prolongation of the first,
pretonic vowel [ui:] to magnify the narrated event, or with a
prolongation of the second, accented vowel [i :] to imply a burst of
emotion.
The velarity in the first vowel of /pil,il/ denotes the antecedent
plain feature; the unreduced, less obscure character denotes that
no word boundary follows ; the vowel lengthening denotes a certain
Preliminaries to speech analysis, third printing (Massachusetts Institute of
Technology, Acoustics Laboratory, 1955), p. 8.
THE VARIETY AND TREATMENT OF FEATURES 1
kind of emphasis. Possession of a single specific denotation unites
the redundant features with the configurative and expressive
features and separates them from the distinctive features. Whatever
distinctive feature we tackle, the denotation is always identical:
any such feature denotes that the morpheme to which it pertains
is not the same as a morpheme having another feature in the
corresponding place. A phoneme, as Sapir remarked, "has no
singleness of reference." ' All phonemes denote nothing but mere
otherness. This lack of individual denotation sets apart the
distinctive features, and their combinations into phonemes, from all
other linguistic units.
The code of features used by the Hstener does not exhaust the
information he receives from the sounds of the incoming message.
From its sound shape he extracts clues to identify the sender. By
correlating the speaker's code with his own code of features, the
listener may infer the origin, educational status and social environment
of the sender. Natural sound properties allow the identification
of the sex, age, and psychophysiological type of the sender and,
finally, the recognition of an acquaintance. Some ways to the
exploration of these physiognomic indices were indicated in
Siever's Schallanalyse,^ but their systematic study still remains on
the agenda.
2.41 The "outer" approaches to the phoneme in relation to sound:
A. The mentalist view. An insight into the complexity of the informational
content of speech sounds is a necessary premise for
the discussion of the various outer approaches to the phoneme in
its relation to sound. In the oldest of these approaches, going
back to Baudouin de Courtenay and still surviving, the phoneme
is a sound imagined or intended, opposed to the emitted sound as
a "psychophonetic" phenomenon to the "physiophonetic" fact.
It is the mental equivalent of an exteriorized sound. The unity of
' E. Sapir, 'Sound patterns in language,' Selected Writings (Berkeley and
Los Angeles, 1949), p. 34.
* See especially E. Sievers, 'Ziele und Wege der Schallanalyse,' Festschrift
fur W. Streitberg (Heidelberg, 1924).
1 PHONOLOGY AND PHONETICS
the phoneme, as compared with the variety of its implementations,
is seen as a discrepancy between the internal impetus aiming at the
same pronunciation and the involuntary vacillation in the fulfill-
ment.
This conception is based on two fallacies: we have no right to
presume that the sound correlate in our internal speech or in our
speech intention is confined to the distinctive features to the exclusion
of the configurative, or redundant features. On the other hand, the
multiplicity of contextual and optional variants of one and the same
phoneme in uttered speech is due to the combination of this
phoneme with diverse redundant and expressive features; this
diversity, however, does not hamper the extraction of the invariable
phoneme from among all these variations. Thus the attempt to
overcome the antinomy between invariance and variability by
assigning the former to the internal and the latter to the external
experience distorts the two forms of experience,
2.42 B. The code-restricting view. Another attempt to locate the
phoneme outside the uttered sounds confines the phonemes to the
code and the variants to the message. A rejoinder to this view
would be that the code includes not only the distinctive features,
but also the redundant and configurative features which induce
contextual variants as, well as the expressive features which underhe
optional variations : the users of a language have learned to eff"ect
and Apprehend them in the message. Thus phoneme and variants
ahke are present, both in the code and in the message.
A cognate tenet, advanced especially in Russia, opposed the
phoneme to the variants as social value to individual behavior.
This is hardly justifiable since not only the distinctive features but
all the coded features are equally socialized.
2.43 C. The generic view. Phoneme has frequently been opposed
to sound as class to specimen. It has been characterized as a family
or class of sounds related through a phonetic resemblance. Such
definitions, however, are vulnerable in several respects.
First, the vague and subjective search for resemblance must be
replaced by the extraction of a common property.
THE VARIETY AND TREATMENT OF FEATURES 13
Second, both the definition and the analysis of the phoneme must
take into account the logical lessbn that "classes can be defined by
properties, but it is hardly possible to define properties by classes." '
In fact, when operating with a phoneme or distinctive feature we
are primarily concerned with a constant which is present in the
various particulars. If we state that in English the phoneme /k/
occurs before /u/, it is not at all the whole family of its various
submembers, but only the bundle of distinctive features common
to all of them, that appears in this position. Phonemic analysis is
a study of properties, invariant under certain transformations.
Third, when deaUng with a sound that in a given language
figures in a definite position, under definite stylistic conditions, we
are again faced with a class of occurrences and their common
denominator, and not with a single, fleeting specimen. Whether
studying phonemes or contextual variants ("allophones"), it is
always, as the logician would say, the "sign-design" and not the
"sign-event" that we define.
2.44 D. The fictionalist view. According to the opinion most
effectively launched by Twaddell in 1935,^" but latently tinging
writings of various authors, phonemes are abstractional, fictitious
units. As long as this means nothing more than that any scientific
concept is a fictional construct, such a philosophical attitude cannot
affect phonemic analysis. Phoneme, in this case, is a fiction, in
the same way as morpheme, word, sentence, language, etc. If,
however, the analyzer opposes the phoneme and its components
to sound as a mere contrivance having no necessary correlate in
concrete experience, the results of the analysis will be distorted
through this assumption. The belief that the choice among
phonemes to which we assign the sound might, upon occasion, be
made arbitrarily, even at random, threatens the objective value of
phonemic analysis. This danger may, however, be avoided by the
• R. Carnap, Meaning and necessity (Chicago, 1947), p. 152.
^" W. F. Twaddell, 'On defining the phoneme' = Supplement to Language,
XVI (1935); cf. M. J. Andrade, 'Some questions of fact and policy concerning
phonemes,' Language, XII (1936).
14 PHONOLOGY AND PHONETICS
methodological demand that any distinctive feature and, consequently,
any phoneme treated by the hnguist, have its constant
correlate at each stage of the speech event and thus be identifiable
at any level accessible to observation. Our present knowledge of
the physical and physiological aspects of speech sounds is sufficient
to meet this demand. The sameness of a distinctive feature throughout
all its variable implementations is now objectively demonstrable.
Three reservations, however, must be made.
First, certain features and combinations of features may be
obliterated in the various kinds of phonemic ellipsis (cf. 1.5).
Second, features may be masked by abnormal, distorting conditions
of sound production (whispering, shouting, singing, stammering),
transmission (distance, filtering, noise) or perception
(auditory fatigue).
Third, a distinctive feature is a relational property so that the
"minimum same" of a feature in its combination with various
other concurrent or successive features lies in the essentially
identical relation between the two opposite alternatives. However
the stops in tot may differ from each other genetically and acoustically,
they are both high-pitched in opposition to the two labials in
pop, and both display a diffusion of energy, as compared to a
greater concentration of energy in the two stops of cock. How
the sameness of a phoneme in two divergent contextual variants is
sensed % the speakers, may be illustrated by such onomatopoetic
sound reduplications as cack, kick, tit, peep, poop.
2.441 ^^
Overlapping^' of phonemes. The so-called overlapping of
phonemes confirms the manifestly relational character of the
distinctive features. A pair of palatal vowel phonemes, genetically
opposed to each other through relative wideness and narrowness
and, acoustically, through a higher and lower concentration of
energy (compact/diffuse), may in some languages be implemented
in one position as [?t] - [q] and in another position as [t] - [\],
so that the same sound [q] in one position implements the diffuse,
and in another, the compact term of the same opposition. The
relation in both positions remains identical. Two degrees ofaperture
THE VARIETY AND TREATMENT OF FEATURES 15
and, correspondingly, of concentration of energy - the maximal
and the minimal - oppose each other in both positions.
The focusing of selective operations upon relational properties
is typical not only of human, but even of animal behavior. In
W. Koehler's experiment, chickens were trained to pick grain from
a gray field and to leave the grain untouched on the adjacent
darker field ; when, subsequently, the pair of fields, gray and dark,
was replaced by a pair, gray and light, the chickens looking for
their food left the gray field for its lighter counterpart. Thus "the
chicken transfers its response to the relatively brighter area."^^ It
is first of all through the relational rules that the listener guided by
the linguistic code apprehends the message.
2.45 E. The algebraic view. The approach one might call "algebraic"
aims at the maximal estrangement between phoneme and
sound or, correspondingly, between phonemics and phonetics. The
champion of this trend, Hjelmslev, calls on linguistics to become
"an algebra of language, operating with unnamed entities, i.e.
arbitrarily named entities without natural designation".^^ Particularly,
the "expression plane" of language, as he christened the
aspect named signans in Stoic and Scholastic tradition, and in the
work of its reviver Ferdinand de Saussure, is to be studied without
any recourse to phonetic premises.
Each venture, however, to reduce language to its ultimate invariants,
by means of a mere analysis of their distribution in the text
and with no reference to their empiric correlates, is condemned to
failure. The comparison of two English sequences - /ku/ and /uk/ will
yield no information on the identity of the first segment in one
of these samples with the second segment in the other sample,
unless we bring into play sound properties common to initial and
final /k/ and those common to /u/ in both positions. The con^^
See H. Werner, Comparative psychology of mental development (New
York-Chicago-Los Angeles, 1940), p. 216f.
" L. Hjelmslev, 'Prolegomena to a theory of language' = Indiana
University Publications in Anthropology and Linguistics, VIII (1935), p. 50;
cf. the objective criticism of this approach by B. Siertsema, A study ofglossematics
('s-Gravenhage, 1954), chapters VI, XI, and by F. Hintze, 'Zum Verhaltnis
der sprachlichen "Form" zur "Substanz",' Studia Linguistica, III (1949).
16 PHONOLOGY AND PHONETICS
frontation of the syllables /ku/ and /ki/ does not authorize us
to assign both initial segments to one phoneme /k/ as two variants
appearing to their mutual exclusion before two different
vowels, unless we have identified the common features, unifying
the retracted and advanced variety of the phoneme /k/ and
differentiating it from all other phonemes of the same language.
Only through such a test are we able to decide whether the retracted
fk-] in/ku/ implements the same phoneme as the advanced /'k +7 in
/ki/ and not the advanced [g+] in /gi/. Therefore, despite the
theoretical requirement of an analysis totally independent of the
sound substance, in practice "on tient compte de la substance a
toute etape de Tanalyse", as Eh Fischer-Jorgensen exposes the
troubling discrepancy.^^
As to the theoretical requirement itself, it arose from the assumption
that in language form is opposed to substance as a
constant to a variable. If the sound substance were a mere variable,
then the search for linguistic invariants would indeed need to
expunge it. But the possibility of translating the same linguistic
form from a sound substance into a graphic substance, e.g. into a
phonetic notation or into an approximately phonemic spelling
system does not prove that the sound substance, Hke other "widely
different expression substances", is a mere variable. In contradistinction
to the universal phenomenon of speech, phonetic or
phonemic writing is an occasional, accessory code that normally
impHes the ability of its users to translate it into its underlying
sound code, while the reverse ability, to transpose speech into
letters, is a secondary and much less common faculty. Only after
having mastered speech does one graduate to reading and writing.
There is a cardinal difference between phonemes and graphic units.
Each letter carries a specific denotation - in a phonemic orthography,
it usually denotes one of the phonemes or a certain limited
series of phonemes, whereas phonemes denote nothing but mere
otherness (cf. 2.3). Graphic signs that serve to interpret phonemes
or other linguistic units stand for these units, as the logician would
" E. Fischer-Jergensen, 'Remarques sur les principes de I'analyse phonemique,'
Travaux du Cercle Linguistique de Copenhague, V (1949), p. 231.
THE VARIETY AND TREATMENT OF FEATURES 17
say. This difference has far-reaching consequences for the cardinally
dissimilar patterning of letters and phonemes. Letters never, or
only partially, reproduce the different distinctive features on which
the phonemic pattern is based and unfailingly disregard the
structural relationship of these features. There is no such thing
in human society as the supplantation of the speech code by its
visual replicas, but only a supplementation of this code by parasitic
auxiliaries, while the speech code constantly and unalterably
remains in effect. One could neither state that musical form is
manifested in two variables - notes and sounds - nor that linguistic
form is manifested in two equipollent substances - graphic and
phonic. And just as musical form cannot be abstracted from
the sound matter it organizes, so form in phonemics is to be
studied in relation to the sound matter which the linguistic code
selects, readjusts, dissects and classifies along its own lines.
Like musical scales, phonemic patterning is an intervention of
culture in nature, an artifact imposing logical rules upon the
sound continuum.
2.5 The cryptanalyst's and decoder's devices as two complementary
techniques. The addressee of a coded message is supposed to be in
possession of the code and through it he interprets the message.
Unlike this decoder, the cryptanalyst comes into possession
of a message with no prior knowledge of the underlying code and
must break this code through dexterous manipulations of the
message. A native speaker responds to any text in his language as
a regular decoder, whereas a stranger, unfamiliar with the language,
faces the same text as a cryptanalyst. A linguist, approaching a
totally unknown language, starts as a cryptanalyst until through a
gradual breaking of its code he finally succeeds in approaching any
message in this language like a native decoder.
The native or naturalized user of a language, when trained
linguistically, is aware of the functions performed by its different
sound elements and may utilize this knowledge to resolve the
sound shape into its manifold information-bearing elements. He
will employ various "grammatical pre-requisites to phonemic
1 PHONOLOGY AND PHONETICS
analysis" as aids to the extraction of destinctive, configurative and
expressive features. ^^
On the other hand, the question raised by Bloch as to the
applicabihty of the cryptanalyst's technique to the inquiry into
phonemic structure has great methodological importance: to
what extent might a sufficient sample of accurately recorded speech
enable a linguist to work out "the phonemic system without
knowing what any part of the sample meant, or even whether any
two parts meant the same thing or different things". ^^ Under such
conditions, the extraction of redundant features is, in many instances,
laborious but feasible. More difficult is the isolation of
the expressive features but, in this regard also, the record may
yield some information, given the difference between the markedly
discrete, oppositional character of distinctive features and the more
continuous "grading gamut" characterizing most of the expressive
features.^^ Even a hybrid - bilingual or multilingual - message,
as for instance, the sentences combined of Russian, French and
English words or phrases, as used in the conversation of the
Russian aristocracy in the late nineteenth century, could be, through
the comparison of their heterogeneous phonetic make-up, roughly
divided into monolingual sections: ''On se reunit le matin au
break/east et puis vsjakij delaet cto xocet'' /"osa neyni bmatS
objekfast epui fs,ak3J d,ebit Jtox^oJitJ, as Tolstoj reproduces the
colloquial speech of his milieu in Anna Karenina.
An even less manageable problem would be the cryptanalytical discrimination
between distinctive and configurative features, especially
word border signals, e.g. it would hardly be possible to
discover that in such Russian pairs of samples as /danos/ ^danosj
'denunciation' - /da nos/ /^danos/ 'and the nose too', /pagar,el,i/
Ap9gar,el,i7 '(they) burned up' - /pagar,e 1,1/ ^p9gar,el,i7 'whether
along a mountain', /jixida/ ^jix,id97 'spiteful person' - /jix ida/
^* K. L. Pike, 'Grammatical prerequisites to phonemic analysis,' Word,
III (1947), and 'More on grammatical prerequisites,' Word, V1II(1952).
" B. Bloch, 'A set of postulates for phonemic analysis'. Language, XXIV
(1948).
'* Cf. Jakobson, Fant, Halle, Preliminaries . . ., p. 15.
THE VARIETY AND TREATMENT OF FEATURES 19
[jixidoj 'their Ida', the difference between [a.] and the obscure [^J,
the close [ej and the open [^] or the palatalized [x,] and the nonpalatalized
[xj is not a distinction of two phonemes but only a
word border signal. Here a cryptanalytical technique runs the
risk of multiplying the number of Russian phonemes and distinctive
features as compared to their actual stock.
Ill
THE IDENTIFICATION OF
DISTINCTIVE FEATURES
3.1 Syllable. The distinctive features are aligned into simultaneous
bundles called phonemes; phonemes are concatenated
into sequences; the elementary pattern underlying any grouping
of phonemes is the syllable." The phonemic structure of the
syllable is determined by a set of rules and any sequence is based
on the regular recurrence of this constructive model. A free
form (a sequence, separable by means of pauses) must contain
an integral number of syllables. Obviously, the number of different
syllables in a language, is a small submultiple of the number of free
forms, just as the number of phonemes is a small submultiple of
the number of syllables, and the number of distinctive features,
a submultiple of the number of phonemes.
The pivotal principle of syllable structure is the contrast of
successive features within the syllable. One part of the syllable
stands out from the others. It is mainly the contrast vowel vs.
consonant which is used to render one part of the syllable more
prominent. There are languages where every syllable consists of a
consonant and a succeeding vowel (CV): in such a case it is possible
from any point of the sequence to predict the class of phonemes
that is to follow. In a language with a greater variety of syllable
types, the recurrence of a phonemic class presents different degrees
of probabiUty. In addition to CV, other schemes may be used:
^' E. Polivanov was the first to draw attention to the "phonemic syllable",
he labelled syllabeme, as the basic constructive cell in the speech sequence:
see his and A. Ivanov's Grammatika sovremennogo kitaiskogo jazyka (Moscow,
1930). Cf. A. Sommerfelt, 'Sur Timportance generale de la syllabe,' Travaux du
Cercle Linguistique de Prague, IV (1931); A. W. de Groot, 'Voyelle, consonne et
syllabe,' Archives neerlandaises de phonetique experimentale, XVII (1941);
J. Kurylowicz, 'Contribution a la theorie de la syllabe,' Bulletin de la Societe
Polonaisede Linguistique, VIII (1948) ; J. D. O'Connor and J. L. M. Trim, 'Vowel,
consonant, and syllable - a phonological definition,' Word, IX (1953).
THE IDENTIFICATION OF DISTINCTIVE FEATURES 21
CVC, V, VC. In contradistinction to C, the part V can neither
be omitted, nor figure twice in the syllable.
The contrast vowel/consonant is either unique or merely
predominant: it can be sporadically substituted by other cognate
contrasts. Both part C and part V may contain more than
one phoneme. The phonemes constituting parts V and C of
the syllable are termed crest phonemes and slope phonemes
respectively. If the crest contains two or more phonemes,
one of them, termed the peak phoneme (or syllabic), is raised
over the others by a contrast compact vs. diffuse or vowel vs.
sonorant.
The motor correlate of the phonemic syllable has been most
adequately described by Stetson^^ as "a puff of air forced upward
through the vocal channel by a compression of the inter-costal
muscles". According to this description, every syllable invariably
consists of three successive factors: release, culmination, and
arrest of the pulse. The middle one of these three phases is the
nuclear factor of the syllable while the other two are marginal.
Both marginal factors - initiation and termination - are effected
either by the mere action of the chest muscles or by speech sounds,
usually consonants. If both marginal factors are effected by the
action of the chest muscles alone, the nuclear phase of the syllable
is the only audible one; if, however, the release and/or the arrest is
effected by speech sounds, the nuclear phase of the syllable is the
most audible. In other words, the nuclear part of the syllable is in
contrast to its marginal part as the crest to the slopes.
In the acoustic aspect, the crest usually exceeds the slopes in
intensity and in many instances shows a higher fundamental
frequency. Perceptually, the crest is distinguished from the slopes
by a greater loudness, which is often accompanied by a heightened
voice-pitch. As a rule, the crest phonemes are inherently louder
than the slope phonemes of the same syllable: ordinarily the crest
is formed by vowels, while the slopes contain the other phonemes;
less frequently the contrast of crest and slope phonemes is displayed
" R. H. Stetson, Motor phonetics (Amsterdam, 1951).
22 PHONOLOGY AND PHONETICS
by liquids vs. pure consonants; or by nasals vs. oral consonants, and
in exceptional cases by constrictives V5. stops (cf. 4.16). If a
slope is constituted by a whole cluster and if within such a cluster
there is an inherently louder phoneme in less loud surroundings, its
loudness is noticeably reduced to preserve the unity of the syllable:
e.g. Czech /jdu/, /jsem/, /rti/, /Ipi/, or Polish monosyllable /krvi/ vs.
Serbocroatian disyllabic /krvi/.^^
3.2 Two kinds of distinctive features. The distinctive features are
divided into two classes: 1) prosodic and 2) inherent. A
prosodic feature is displayed only by those phonemes which form
the crest of the syllable and it may be defined only with reference
to the relief of the syllable or of the syllable chain, whereas the
inherent feature is displayed by phonemes irrespective of their role
in the rehef of the syllable, and the definition of such a feature does
not refer to the rehef of the syllable or of the syllable chain.
3.3 Classification ofprosodicfeatures. The three types of prosodic
features, which, following Sweet, we term tone, force, and
quantity, correspond to the three attributes of sensation - voicepitch,
voice-loudness and subjective duration. The dimensions of
frequency, intensity and time are their closest physical correlates.
Each of these three sub-classes of prosodic features presents two
varieties : according to its frame of reference a prosodic feature may
be either intersyllabic or intrasyllabic. In the first case, the
crest of one syllable is compared with the crests of other syllables
within the same sequence. In the second case, an instant pertaining
to the crest may be compared with other instants of the same crest
or with the subsequent slope.
3.31 Tone features. In the intersyllabic variety of tone features
the level feature, different syllable crests within a sequence are
contrasted by their register: higher and lower. The level feature
may be spht in two : either a neutral register is contrasted with an
*' See particularly A. Abele, 'K voprosu o sloge,' Slavia, III (1924).
THE IDENTIFICATION OF DISTINCTIVE FEATURES 23
elevated register, on the one hand, and with a lowered one, on the
other, or, finally, each of the two opposite registers, high and low,
may appear in two varieties, raised and diminished. When the
Jabo people transpose these four levels from speech into drum
signals, they use for the two underlying oppositions two different
pairs of terms: the opposites high and low are called "little bird"
and "big bird", while the opposites raised and diminished are
termed "smaller" and "larger", so that the four signals are
distinguished - "smaller little bird", "larger little bird", "smaller
big bird", and "larger big bird".^" The voice-tone mechanism has
been closely investigated by Farnsworth, who states that the motion
of the vocal cords, more complex at low frequencies of vibration,
becomes simplified as the rate is raised, until at the highest
frequencies of vibration, only the edges of the cords nearest the
glottis are seen to vibrate.^^
The intrasyllabic variety of tone features, the modulation
feature, contrasts the higher register of one portion of a phoneme
with a lower register of another portion of the same phoneme, or
the higher register of one component of a diphthong with the
lower register of its other components, and this distribution of
registers within the crests of the syllable is opposed to the reverse
distribution, e.g. a rising modulation to a falling one, or both of
them to an even intonation.
3.32 Force features. The intersyllabic variety of the force features,
the stress feature, is a contrast of a louder, stressed crest to the
less loud, unstressed crests of other syllables within the same
sequence, a difference produced by the sublaryngeal mechanism,
in particular by the abdomen-diaphragmal movements, as Sievers
and Stetson attempt to prove. ^^
^ See G. Herzog, 'Drum signaling in West African tribes,' Word, I (1945).
** D. W. Farnsworth, 'High-speed motion picture of the human vocal
cords,' Bell Laboratories Record, V (1940).
" E. Sievers, 'Neues zu den Rutzschen Reaktionen,' Archiv fiir experimentelle
und klinische Phonetik, I (1914); R. H. Stetson, I.e. Cf. W. F. Twaddell,
'Stetsons's model and the "supra-segmental phonemes",' Language, XXIX
(1953).
24 PHONOLOGY AND PHONETICS
In the intrasyllabic variety of the stress features, the so-called
St OSS ton (stod) feature, two contiguous fractions of the stressed
phoneme are compared with each other. To an even distribution
of loudness throughout the phoneme, another type is opposed:
the initial portion of the phoneme presents the peak of loudness,
whereas in the final portion the loudness decreases. According
to S. Smith's analysis of the Danish stod,^^ the decline of amplitude,
often accompanied by a decrease of the fundamental frequency,
is due to an abruptly decreasing innervation of the expiratory
muscles. A ballistic movement of the expiratory muscles, opposed
to a more even movement, produces a similar prosodic feature,
e.g., in Latvian, Lithuanian dialects and Livian.
3.33 Quantit} features. The intersyllabic variety of quantity
features, the length feature, contrasts a normal, short, unstretchable
phoneme within the crest of the syllable with the long,
sustained phonemes of the other syllables in the same sequence,
and/or a normal, short but steady phoneme with a punctual,
reduced, transient one.
The second variety of the quantity features, the contact
feature, is based on a different distribution of duration between
the vowel and the subsequent consonant: in the case of the socalled
close contact {scharf geschnittener Akzent), the vowel is
abridged in favor of the following, arresting consonant, whereas
at the open contact (schwach geschnittener Akzent), the vowel
displays its full extent before the consonant starts.
3.34 The interconnection between stress and length. Wherever
there is a contrast of stressed and unstressed syllables, stress is
always used as a configurative, namely culminative feature, whereas
length never assumes this function. The culminative function of
the stress is regularly combined either with the other variety of
configurative functions, the demarcation (cf. 2.3), or with the
distinctive function. Languages where both length and stress
appear as distinctive features are quite exceptional, and if the stress
^^ S. Smith, 'Contributions to the solution of problems concerning the
Danish stod,' Nordisk T'uhskrift for Tale og Stemine, VHI (1944).
THE IDENTIFICATION OF DISTINCTIVE FEATURES 25
is distinctive the latter is mostly supplemented by a redundant
length.
The observation of force and quantity features in their intersyllabic
variety seems to indicate that the prosodic distinctive
features utilizing intensity and those utilizing time tend to merge.
3.4 Comparison of prosodic and inherent features. Any prosodic
feature is based primarily on the contrast between two variables
within one and the same time sequence: the relative voicepitch,
voice-loudness or duration of a given fraction is determined
with respect to preceding and/or succeeding fractions. As Herzog
has pointed out concerning the tone features, "the actualizations
of the contrasts - given by successive distances between tone
levels or by successive tone movements - do shift all the time."-^
Tone level, or tone modulation, stress degrees or its decrescendo
(stosston), are always purely relative and highly variable in their
absolute magnitudes from speaker to speaker, and even from one
utterance to another in the usage of the same speaker. Also the
quantity of a vowel may be established only in relation to the *
quantity of the other vowels within the context or to the subsequent
consonants (contact feature), while the absolute duration of the
long or short vowels in the given language presents a considerable
vacillation in speed, depending upon the speech-habits of the
speaker and his expressive variations of tempo. A long vowel
must be ceteris paribus longer than the surrounding short vowels.
Similarly, the only thing required of a stressed vowel is to be
uttered in a louder voice than the unstressed vowels of the same
chain; and the high register vowels must be of a higher voice-tone
than the neighboring low register vowels. But the high register
vowels of one, e.g. bass speaker, may be even deeper than the low
register vowels of another, e.g. soprano speaker, and in the speech
of one and the same person there may be passages with relative
expressive lowering of both high and low register phonemes.
A prosodic feature involves two coordinates: on the one hand,
^*
G. Herzog, review of K. L. Pike, Tone languages, in Iniernalional Journal
of American Linguistics, XV ( 1 949).
26 PHONOLOGY AND PHONETICS
polar terms such as high and low register, rising and falling pitch,
or long and short, all may appear, ceteris paribus, in the same
position in the sequence, so that the speaker selectively uses and
the listener selectively apprehends one of the two alternatives and
identifies the chosen alternative in relation to the rejected one.
These two alternatives, the one present and the other absent in the
given unit of the message, constitute a veritable logical opposition
(cf. 1.3). On the other hand, both polar terms are fnlly.
recognizable only when^hoth of tjiem are present in the gjyen
sequence, so that the speaker effects and the listener perceives
their contrast. Thus both alternatives of a prosodic feature coexist
in the code as two terms of an opposition and, moreover,
co-occur and produce a contrast within the message. If the
message is too brief to include both contrasting units, the feature
may be inferred from the substitutive clues offered by the sequence,
e.g. the quantity of a vowel in a monosyllabic message may be I
inferred from the relative duration of the surrounding consonants, j
and the register of a monophonemic message, from the modulation
span in the onset and/or decay of the vowel.
The recognition and definition of an inherent feature is based only
on/ the choice between two alternatives admissible in the same
position within a sequence. No comparison of the two polar terms
co-occurring within one context is involved. Hence, both alternatives
of an inherent feature co-exist in the code as two terms of an
opposition, but^o noLie^mreji contrasting juxtaposition within
onemessage^ince the inherent feature is identified only through
the comparison of the alternative present in the given position with
the absent alternative, the implementation of an inherent feature
in a given position admits less variabiUty than that of the prosodic
features.
3.5 General laws of phonemic patterning. The comparative
description of the phonemic systems of diverse languages and
their confrontation with the order of phonemic acquisitions by
infants learning to speak, as well as with the gradual dismantling
of language and of its phonemic pattern in aphasia, gives us
I
THE IDENTIFICATION OF DISTINCTIVE FEATURES 27
important insights into the interrelation and classification of the
distinctive features. The hnguistic, especially phonemic progress
of the child and the regression of the aphasic obey the same laws
of implication. If the child's acquisition of distinction B implies
his acquisition of distinction A, the loss of A in aphasia implies
the absence of B, and the rehabilitation of the aphasic follows the
same order as the child's phonemic development. The same laws
of implication underhe the languages of the world both in their
static and dynamic aspects. The presence of B implies the presence
of A and, correspondingly, B cannot emerge in the phonemic
pattern of a language unless A is there; Hkewise, A cannot disappear
from a language as long as B exists. The more limited the number
of languages possessing a certain phonemic feature or combination
of features, the later is it acquired by the native children and the
earlier is it lost by the native aphasics.
3.51 Restrictions in the over-all inventory ofdistinctive features.
The progressing phonemic investigation of infants and aphasics,^*
along with the ever increasing number of detected laws, moves
into the foreground the problem of the universal rules underlying
the phonemic patterning of languages. In view of these laws
of implication and stratification, the phonemic typology of languages
is becoming an ever more feasible and urgent task. Every
step in this direction permits us to reduce the list of distinctive
features used in the languages of the world. The supposed multiplicity
of features proves to be largely illusory. If two or more
allegedly different features never co-occur in a language and if they,
*° Cf. R. Jakobson, 'Kindersprache, Aphasie und allgemeine Lautgesetze,'
Uppsala Universitets Arsskrift (1942); H. V. Velten, 'The growth of phonemic
and lexical patterns in infant language,' Language, XIX (1943); W. F. Leopold,
Speech development of a bilingual child, II (Evanston, 1947); A. Gvozdev,
Usvoenie rebenkom zvukovoj storony russkogo jazyka (Moscow, 1948);
K. Ohnesorg, Foneticka studie o detske feci (Prague, 1948); L. Kaczmarek,
Ksztaltowanie si^ mowy dziecka (Poznan, 1953); P. Smoczynski, Przyswajanie
przez dziecko podstaw systemu jqzykowego (Lodz, 1955). - Th. Alajouanine,
A. Ombredane, M. Durand, Le syndrome de disintegration phonetique dans
raphasie (Paris, 1939); A. Luria, Travmatiieskaja afazija (Moscow, 1947);
K. Goldstein, Language and language disturbances (New York, 1948).
28 PHONOLOGY AND PHONETICS
furthermore, yield a common property distinguishing them from all
other features, then they are to be interpreted as different implementations
of one and the same feature, each occurring to the
exclusion of the others and, consequently, presenting a particular
case of complementary distribution. The study of invariances
within the phonemic pattern of one language must be supplemented
by a search for universal invariances in the phonemic patterning of
language.
Thus no language simultaneously displays two autonomous
consonantal oppositions - pharyngealized/non-pharyngealized
and rounded/unrounded. The back orifice of the mouth resonator
(pharyn.x) is involved in the first instance and the front orifice (hps)
in the second, but in both cases a narrowed orifice of the mouth
resonator, producing a downward shift in the resonances, is
opposed to the absence of narrowing. Hence these two processes
(narrowed back slit and narrowed front slit) are to be treated as two
variants of one and the same opposition which on the motor level
may be defined as narrowed vs. adjusted slit (cf. 3.62). The relation
of retroflex to the dental consonants proves to be a mere variety of
the opposition of pharyngealized and non-pharyngealized dentals.
Four consonantal features listed by Trubetzkoy (I.e., pp. I32f.)
- the tension feature, the intensity or pressure feature, the aspiration
feature and the pre-aspiration feature - also turn out to be complementary
variants of one and the same opposition which by virtue
of its common denominator may be termed tense/lax.
Double stops (particularly clicks) with closures in rapid succession,
followed by two distinct releases in the same order, appear to
the exclusion of other types of clusters in the same positions and
present simply a different implementation of ordinary consonantal
sequences.-*^
3.6 The nro elasses of inherent features. The inherent distinctive
features which have so far been discovered in the languages of the
world and which, along with the prosodic features, underlie their
^* Cf. C. M. Doke, 'Notes on a problem in the mechanism of the Zulu
clicks,' Bantu Studies, II (1923).
THE IDENTIFICATION OF DISTINCTIVE FEATURES 29
entire lexical and morphological stock, amount to twelve oppositions,
out of which each language makes its own selection. All
the inherent features are divided into two classes that might
be termed sonority features and tonality features, the
former akin to the prosodic force and quantity features and the
latter to the prosodic pitch features. The sonority features utilize
the amount and concentration of energy in the spectrum and in
time. The tonality features involve the ends of the frequency
spectrum.
3.61 Sonority features
I. Vocalic/non-vocalic:
acoustically - presence vs. absence of a sharply defined formant
structure;
genetically - primary or only excitation at the glottis together
with a free passage through the vocal tract.
II. Consonantal/non-consonantal:
acoustically - low {vs. high) total energy;
genetically - presence vs. absence of an obstruction in the vocal
tract.
Vowels are vocalic and non-consonantal; consonants are
consonantal and non-vocalic; liquids are vocalic and consonantal
(with both free passage and obstruction in the oral cavity and the
corresponding acoustic effect); glides are non-vocalic and non-
consonantal.
III. Compact/diffuse:
acoustically - higher {vs. lower) concentration of energy in a
relatively narrow, central region of the spectrum, accompanied
by an increase {vs. decrease) of the total amount of energy;
genetically - forward-flanged v.y. backward-flanged. The difference
lies in the relation between the volume of the resonance chamber
in front of the narrowest stricture and behind this stricture.
The ratio of the former to the latter is higher for the forwardflanged
phonemes (wide vowels, and velar and palatal, including
30 PHONOLOGY AND PHONETICS post-alveolar,
consonants) than for the corresponding backwardflanged
phonemes (narrow vowels, and labial and dental, including
alveolar, consonants).
IV. Tense/lax:
acoustically - higher (vs. lower) total amount of energy in conjunction
with a greater (vs. smaller) spread of the energy in the spectrum
and in time;
genetically - greater (v5. smaller) deformation of the vocal tract
- away from its rest position. The role of muscular strain affecting
the tongue, the walls of the vocal tract and the glottis requires
further examination.
V. Voiced/voiceless:
acoustically - presence vs. absence of periodic low frequency
excitation
genetically - periodic vibrations of the vocal cords vs. lack of
such vibrations.
VI. Nasal/oral (nasalized/non-nasalized):
acoustically - spreading the available energy over wider (vs.
narrower) frequency regions by a reduction in the intensity of
certain (primarily the first) formants and introduction of additional
(nasal) formants;
genetically - mouth resonator supplemented by the nose cavity vs.
the exclusion of the nasal resonator. ,
VII. Discontinuous/continuant:
acoustically - silence (at least in frequency range above vocal cord
vibration) followed and/or preceded by spread of energy over a
wide frequency region (either as burst or as a rapid transition of
vowel formants) vs. absence of abrupt transition between sound
and such a silence;
genetically - rapid turning on or off of source either through a
rapid closure and/or opening of the vocal tract that distinguishes
plosives from constrictives or through one or more taps that
differentiate the discontinuous liquids like a flap or trill /r/ from
continuant liquids like the lateral /I/.
THE IDENTIFICATION OF DISTINCTIVE FEATURES 3
VIII. Strident/mellow:
acoustically - higher intensity noise vs. lower intensity noise;
genetically - rough-edged vs. smooth-edged: supplementary obstruction
creating edge effects (Schneidenton) at the point of articulation
distinguishes the production of the rough-edged phonemes
from the less complex impediment in their smooth-edged counter-
parts.
IX. Checked/unchecked:
acoustically - higher rate of discharge of energy within a reduced
interval of time vs. lower rate of discharge within a longer interval;
genetically - glottalized (with compression or closure of the glottis)
vs. non-glottalized.
3.62 Tonality features.
X. Grave/acute:
acoustically - concentration of energy in the lower {vs. upper)
frequencies of the spectrum
genetically - peripheral vs. medial : peripheral phonemes (velar and
labial) have an ampler and less compartmented resonator than the
corresponding medial phonemes (palatal and dental).
XI. Flat/plain:
acoustically - flat phonemes in contradistinction to the corresponding
plain ones are characterized by a downward shift or
weakening of some of their upper frequency components;
genetically - the former (narrowed slit) phonemes in contradistinction
to the latter (wider slit) phonemes are produced with a
decreased back or front orifice of the mouth resonator, and a
concomitant velarization expanding the mouth resonator.
XII. Sharp/plain:
acoustically - sharp phonemes in contradistinction to the corresponding
plain ones are characterized by an upward shift of some
of their upper frequency components;
genetically - the sharp (widened slit) v.y. plain (narrower slit)
phonemes exhibit a dilated pharyngeal pass, i.e. a widened back
32 PHONOLOGY AND PHONETICS
orifice of the mouth resonator; a concomitant palatalization
restricts and compartments the mouth cavity.
3.7 Stages of the speech event. Each of the distinctive features
has been defined above both on its acoustical and articulatory
level. The communication network, however, comprises a higher
number of stages. The initial stage in any speech event - the
intention of the sender - is not yet open to a precise analysis.
The same may be said of the nerve impulses sent from the brain to
the effector organs. The work of these organs - the motor stage
of the speech event - is at present quite accessible to observation,
especially with the progress of X-rays and other tools that uncover
the activities of such highly important parts of the speech apparatus
as the pharyngeal, laryngeal and sub-laryngeal mechanisms. The
status of the message between the bodily pathways of the speaker
and listener, the transmitted vibrations in the air, are ever more
adequately mastered, owing especially to the amazing advance of
modern acoustics.
The translation of the physical stimulus, first into aural and
then into neural processes, is about to be charted.-' The search
for^the models of distinctive features used by the auditory system
is a timely task. As to the transformation of speech components by
the nervous system, we can, for the time being, at best only hazard
what psychophysiologists have intimated as "a mere speculative
assertion":'^ sonority features seem to be related to the amount,
density and spread of nervous excitation, while the tonality features
relate to the location of this excitation. The present development
of research on the neural responses to sound stimuli promises,
however, to supply a differential picture of distinctive features on
this level as well.
The psychological study of sound perception has endeavored
-"
For tentative moves in this direction, see J. C. R. Licklider, 'On the
process of speech perception," Journal of the Acoustical Society of America,
XXIV (1952); H. Mol and E. M. Uhlenbeck, 'The analysis of the phoneme
in distinctive features and the process of hearing,' Lingua, IV (1954).
-' S. S. Stevens and H. Da\is, Hearing (New York, 1938), p. 164.
THE IDENTIFICATION OF DISTINCTIVE FEATURES 33
to isolate the diverse subjective attributes of sound and to determine
the discriminatory capacity of the listeners for each of the dimensions
of the stimulus. The expansion of this investigation to
speech sounds is likely to illuminate the perceptual correlates of the
diverse distinctive features in view of their phenomenal autonomy.
The initial experiments on English consonants transmitted with
frequency distortion and with random masking noise have actually
confirmed that the perception of each of these features is relatively
independent of the perception of the others, as if "separate, simple
channels were involved rather than a single complex channel. "^^
To a psychologist, each attribute is defined through a differential
reaction to a stimulus by a listener under a particular SQt (Aufgabe).
In application to speech sounds this set is determined by the
decoding attitude of the listener to the message received and to
each of its constituents. The listener correlates the incoming
message with the code common to himself and the speaker. Thus
the role of sound components and combinations in the linguistic
pattern is implicit in the perception of speech sounds. To find
out what motor, acoustic and perceptual elements of sounds are
utilized in a given language, we must be guided by its coding rules:
an efficacious physiological, physical and psychological analysis of
speech sounds presupposes their linguistic interpretation.
3.71 The use of different stages in the study of distinctive features.
In order to decode the message, its receiver extracts the distinctive
features from the perceptual data. The closer we are in our
investigation to the destination of the message, the more accurately
2* G. A. Miller and P. E. Nicely, 'An analysis of perceptual confusions
among some English consonants,' Journal of the Acoustical Society of America,
XXVII (1955). A fruitful check of the distinctive features on the perceptual
level may also be expected from the experiments in progress at the Haskins
Laboratories (New York) on the perception of synthetic speech sounds. Furthermore,
a cautious study of synesthetic associations between phonemic features
and color attributes should yield clues to the perceptual aspect of speech sounds.
There seems to be a phenomenal affinity between optimal chromaticity (pure red)
and vocalic compactness, attenuated chromaticity (yellow—blue) and vocalic
diffuseness, optimal achromaticity (black—white) and consonantal diffuseness^
attentuated achromaticity (grayed) and consonantal compactness; moreover
between the value axis of colors (dark—light) and the tonality axis in language,
34 PHONOLOGY AND PHONETICS
can we gauge the information conveyed by the sound-chain. This
determines the operational hierarchy of levels in their decreasing
pertinence: perceptual, aural, acoustical and motor (the latter
carrying no direct information to the receiver except for the
sporadic help of lip-reading). The auditory experience is the only
aspect of the encoded message actually shared by the sender and
the receiver since the speaker normally hears himself.
In the process of communication there is no single-valued
inference from a succeeding to a preceding stage. With each
successive stage, the selectivity increases ; some data of an antecedent
stage are irrelevant for any subsequent stage, and each item of the
latter stage may be a function of several variables from the former
stage. The measurement of the vocal tract permits an exact
prediction of the sound wave, but one and the same acoustical
effect may be attained by altogether different means. Similarly,
the same attribute of the auditory sensation may be the result of
different physical stimuli.
The theoretically unlikely surmise of a closer relationship
between perception and articulation than between perception and
its immediate stimulus finds no corroboration in experience: the
kinaesthetic feedback of the listener plays a very subordinate and
incidental role. Not seldom do we acquire the ability to discern
foreign phonemes by ear without having mastered their production,
and in a child's learning of language an auditory discrimination
of adults' phonemes often precedes the use of these phonemes in
his own speech.
The specification of distinctive oppositions may be made with
respect to any stage of the speech event, from articulation to
perception and decoding, on the sole condition that the invariants
of any antecedent stage be selected and correlated in terms of the
subsequent stages, given the evident fact that we speak to be heard
and need to be heard in order to be understood.
The distinctive features have been portrayed only on the motor
and on the acoustic level, because these are the only two aspects
for which we so far possess detailed information. Either of these
two patterns must give the complete picture of all the ultimate,
THE IDENTIFICATION OF DISTINCTIVE FEATURES 35
further irreducible distinctions. But since articulation is to acoustic
phenomenon as means to effect, the classification of motor data
must be made with reference to the acoustic patterns. Thus, the^
difference among four articulatory classes of consonants - velar,
palatal, dental and labial - dissolves itself on the acoustic level into
two binary oppositions : on the one hand, labials and velars concentrate
their energy in the lower frequencies of the spectrum in
contradistinction to dentals and palatals, which concentrate their
energy in the upper frequencies - the grave/acute opposition. On
the other hand, velars and palatals are distinguished from labials
and dentals by a greater concentration of energy - the compact/
diffuse opposition. The gravity of the labials and velars is generated
by a larger and less divided mouth cavity, and the acuteness of
dentals and palatals, by a smaller and more compartmented cavity.
jThus, on the motor level, the decisive difference is between the
/stricture in a medial region of the mouth - dental or palatal - and
a stricture in a peripheral region - labial or velar. An identical
articulatory difference opposes the velar to palatal vowels (back front)
as acoustically grave vs. acute. A larger volume of the
resonating cavity in front of the point of articulation and a smaller
volume of the cavity behind this point distinguishes velar from
labial consonants and palatal from dental consonants and engenders
the compactness of velars and palatals. The same articulatory
factor determines the compactness of the wide vowels vs. the
diffuseness of the narrow vowels. It would have been much more
difficult to extract the common denominator of the distinctions
between labial and dental consonants and velar and palatal
consonants or vowels, as well as the common denominator of the
distinctions between velars and labials, palatals and dentals, wide
and narrow vowels, if the striking acoustical and perceptual
oppositions grave/acute and compact/diffuse were not taken into
account.
Although it was evident to observers that among plosives, the
labio-dental, alveolar (hissing), post-alveolar (hushing) and uvular
affricates are opposed by their noisy friction to the bilabial, dental,
palatal and velar stops, nonetheless a similar opposition between
36 PHONOLOGY AND PHONETICS
the corresponding constrictives was usually overlooked, notwithstanding
that all these affricates and the homorganic constrictives
are distinguished by a special kind of turbulence due to forcing the
air stream over a supplementary barrier (the edge of the teeth or
uvula). In the spectrogram, the random distribution of black areas
in these strident consonants, as compared with the considerably
more regular patterns in the mellow consonants, is the only
differentiating clue for all such pairs, and this clue, common to all
the pairs in question, reveals a distinct binary opposition,
3.72 Nomenclature ofdistinctivefeatures. Traditional terminology
resorted indiscriminately to different stages of the speech event:
terms such as nasal, palatalized, rounded, glottalized, referred to
the motor level; other labels (voiced, high, falHng, pitch, lenis,
liquid) referred partly to the acoustical, partly to the perceptual
aspect, and even when a figurative term was used, it had some basis
in phenomenal experience. Insofar as the feature we define has a
traditional term, we use the latter regardless of the stage of the
speech event to which it relates, e.g. nasal/oral, tense/lax, voiced/
voiceless, stressed/unstressed. A traditional articulatory term is
retained as long as it points to an important criterion of division
with respect to the sound transmitted, perceived and decoded. In
several cases, however, there is no current phonetic term to cover
the feature we define. For such features we take over terms from
acoustics or psycho-acoustics. But since each of these features is
definable and has actually been defined both on the acoustic and
on the motor level, any of them could with equal right bear a
newly-coined articulatory designation such as forward-flanged/
backward-flanged instead of compact/diffuse, rough-edged/
smooth-edged instead of strident/mellow, peripheral/medial
instead of grave/acute, narrowed slit/wider slit instead of
flat/plain and widenend slit/narrower slit instead of
sharp/plain.
We are not concerned with substituting an acoustic classification
for an articulatory one but solely in uncovering the most productive
criteria of division valid for both aspects.
IV
PHONEMIC PATTERNING
4. 1 Stratification: nuclear syllable. Ordinarily child language begins,
and the aphasic dissolution of language preceding its complete loss
ends, with what psychopathologists have termed the "labial stage",
in this phase speakers are capable only of one type of utterance,
which is usually transcribed as /pa/. From the articulatory point
of view the two constituents of this utterance represent polar
configurations of the vocal tract: in /p/ the tract is closed at its
very end while in /a/ it is opened as widely as possible at the front
and narrowed toward the back, thus assuming the horn-shape of a
megaphone. This combination of two extremes is also apparent
on the acoustic level: the labial stop presents a momentary burst
of sound without any great concentration of energy in a particular
frequency band, whereas in the vowel /a/ there is no strict limitation
of time, and the energy is concentrated in a relatively narrow region
of maximum aural sensitivity. In the first constituent there is an
extreme limitation in the time domain but no ostensible limitation
in the frequency domain, whereas the second constituent shows no
ostensible limitation in the time domain but a maximum limitation
in the frequency domain. Consequently, the diffuse stop with its
maximal reduction in the energy output offers the closest approach
to silence, while the open vowel represents the highest energy
output of which the human vocal apparatus is capable.
This polarity between the minimum and the maximum of energy
appears primarily as a contrast between two successive units the
optimal consonant and the optimal vowel. Thus the elementary
phonemic frame, the syllable, is established. Since many languages
lack syllables without a prevocalic consonant and/or with a postvocalic
consonant, CV (Consonant -f Vowel) is the only universal
model of the syllable.
38 PHONOLOGY AND PHONETICS
4.12 The role of the nasal consonant. The choice between /pa/
and /a/ and/or /pa/ and /ap/ may become the first carrier of
meaning in the very early stages of child language. Usually, however,
the infant preserves for a time a constant syllable scheme
and splits both constituents of this syllable, first the consonant
and later the vowel, into distinctive alternatives.
Most frequently, the oral stop, utilizing a single closed tract,
obtains a counterpart in the nasal consonant, which combines a
closed main tract with an open subsidiary tract and thereby supplements
the specific traits of a stop with a secondary vocalic characteristic.
Before there appeared the consonantal opposition nasal/oral,
consonant was distinguished from vowel as closed tract from open
tract. Once the nasal consonant is opposed to the oral as presence
to absence of the open tract, the contrast consonant/vowel is
revalued as presence vs. absence of a closed tract.
Various further oppositions, modifying and attenuating the
primary optimal contrast of consonant and vowel follow. All
these later formations reshape in some way the mouth resonator,
while nasalization merely adds a secondary resonating cavity to
the mouth resonator without changing its volume and shape.
The opposition of nasal and oral consonant, which belongs to
the earliest acquisitions of the child, is ordinarily the most resistant
cofisonantal opposition in aphasia and it occurs in all the languages
of the world except for some American Indian languages.
4.13 The primary triangle. The opposition nasal vs. oral stop,
however, may be preceded by the split of the stop into two
opposites, labial and dental. After the appearance of the contrast
CV, founded upon one attribute of sound, loudness, the utilization
of the other basic attribute, pitch, is psychologically inferable.
Thus the first tonahty opposition is instituted: grave/acute, in
other words, the concentration of energy in the lower vs. upper
frequencies of the spectrum. In /p/ the lower end predominates,
while in /t/ the upper end is the stronger one. It is quite natural
that the first tonality feature should affect not the vowel /a/, with its
maximal concentration of energy in a narrow central region of the
PHONEMIC PATTERNING 39
Spectrum, but the consonant /p/, with its maximal diffusion of
energy over a wide frequency band.
At this stage the pole of high and concentrated energy /a/
contrasts with the low energy stops /p/ and /t/. Both stops are
opposed to each other by a predominance of one or the other end
of the frequency spectrum, as the gravity and acuteness poles.
These two dimensions underlie a triangular pattern of phonemes
(or at least of oral phonemes if the nasality feature has already
emerged)
a
4. 1 4 The split of the primary triangle into two triangles, consonantal
and vocalic. The rise of the consonantal tonality feature is followed
by the first vocalic split. The polarity of two successive units, CV,
based on the contrast of reduced and full energy is supplemented
by a polarity of two alternative vowels, founded on the opposition
of lower and higher concentration of energy. The single compact /a/
finds its opposite in a diffuse vowel. Henceforth, both the
consonantal and the vocalic section of the primary triangle construct
each its own linear pattern - the grave/acute consonantal axis
and the compact/diffuse vocalic axis.
The consonants duplicate this originally vocalic opposition and
the consonantal base-line of the over-all triangle proves to be
complemented by a consonantal apex - the velar stop that Grimm
had already justly defined as the "fullest of all producible conso-
nants".
The tonality opposition, originally consonantal, may in turn
be extended to the vocalic pattern: it is naturally the diffuse
vowel that sphts into grave and acute, complementing the vocalic
40 PHONOLOGY AND PHONETICS
apex of the over-all triangle by a /u/ - /i/ base-line. In this way the
originally single primary triangle is split into two autonomous
two-dimensional patterns - the consonantal and the vocalic
triangle.
4.15 Patterning of oral resonance features. Both the vocalic and
the consonantal pattern may subsequently pass from the triangular
to the quadrangular pattern by superimposing the distinction
between velar and palatal upon the wide vowels and/or upon the
consonants. In this way the grave/acute feature spreads to the
compact vowels and/or consonants. In the languages of the world,
however, the triangular pattern prevails over the quadrangular for
vowels and even more so for consonants - it is the minimum model,
both for the vocalic and for the consonantal patterns, with the very
ra^e exceptions when either the vocalic or the consonantal pattern but
never both - is linear. In the rare cases of a linear patterning,
the vowels are confined to the feature compact/diffuse and the
consonants, almost unfailingly, to the tonality feature. Thus no
language lacks the oppositions grave/acute and compact/diffuse,
whereas any other opposition may be absent.
The alternation in volume and shape of the mouth resonator
is used for the grave/acute opposition. In the early stages of
child language, in the advanced stages of aphasia and in numerous
languages of the world, this alternation is reinforced by the
variation in the size of one or both orifices of the mouth cavity.
The restriction of the back and front orifices, together with an
expanded and unified oral cavity, serve to lower the resonance
frequencies, whereas the combined action of the dilated orifices
PHONEMIC PATTERNING 41
and of a restricted and compartmented cavity raise the resonance
frequencies. But the change in the size of each of these orifices
may achieve an autonomous status and set in operation secondary
tonahty features (flatting and/or sharping).
The development of the oral resonance features in child language
presents a whole chain of successive acquisitions interhnked by
laws of implication. We tentatively tabulate this temporal series
in the following chart, using for the distinctions acquired the
traditional articulatory terms and designating each of these
acquisitions by a sequence of numbers preceded by 0., i.e. writing
each sequence as a decimal fraction. The sequences were composed
in such a way that if sequence S^ is assigned to distinction A and
sequence Sg to distinction B, and S^ is an initial subsequence of Sg
(i.e. Sj is an initial subsequence of S., if the first digits of Sg are
identical with Si; e.g. Si==0.l9 and S. = 0.195), then the
acquisition of distinction B implies that of A. The numerical
values of the digits and their number have no other significance.
It is obvious that only those distinctions are acquired by the child
that are present in the language beiqg learned.
Consonants: dental vs. labial 0.1
Vowels: narrow vs. wide 0.11
Narrow vowels: palatal vs. velar 0.111
Wide vowels: palatal vs. velar 0.1111
Narrow palatal vowels: rounded vs. unrounded . . 0.1112
Wide palatal vowels: rounded vs. unrounded 0.111,21
Velar vowels: unrounded vs. rounded 0.1113
Consonants: velopalatal vs. labial and dental. . . 0.112
Consonants: palatal vs. velar 0.1121
Consonants: rounded vs. unrounded or
pharyngealized vs. non-pharyngealized . 0.1122
Consonants: palataHzed vs. non-palatalized . . .0.1123
4.16 Sonority features in relation to the optimal consonant and
vowel. The reduced concentration of energy in the diffuse vowel
moves it away from the optimal, compact vowel in the direction of
42 PHONOLOGY AND PHONETICS
the consonants and, conversely, the reduced spread of energy in
the compact consonant diverts it from the optimal, diffuse
consonant in the direction of the vowel.
In the nasal consonants the addition of the new, open resonator
superimposes sharply-defined nasal formants upon the spectrum of
the oral stop. Nasal resonance brings consonants closer to vowels
and, on the other hand, when superimposed upon a vocalic spectrum,
damps the other formants and deflects the vowel from its
optimal pattern.
The optimal, stop consonant finds its opposite in the constrictive
that attenuates the consonantal reduction of energy. Stops are an
earlier acquisition of children and a later loss of aphasics than
constrictive phonemes. There are in the world several languages
without constrictives but no languages without stops.
The appearance of liquids which combine the clear-cut formant
structure of a vowel with the consonantal reduction of energy,
changes the contrast consonant/vowel into two autonomous oppositions,
consonantal/non-consonantal and vocalic/non-vocalic.
While the consonantal feature, reduction of energy, is optimally
represented by the stop, which tends toward a single pulse, the nonvocalic
feature, absence of sharply-defined formant structure, is
optimally manifested by the strident consonant which tends toward
white noise. Therefore, the emancipation of the two features from
each other, discontinuous/continuant on the one hand, and strident/
mellow on the other, implies the acquisition of a liquid that
combines two autonomous features, the vocalic and the consonantal.
Actually, mellow constrictives, opposed to strident constrictives, or
strident plosives (affricates) opposed to mellow plosives (stops
proper) do not appear in child language before the emergence of
the first liquid, and under aphasia, vanish when the liquids are lost.
Strident plosives, in contradistinction to mellow plosives,
attenuate the consonantal reduction of energy. The mellow constrictives
deviate from the non-vocalic optimum embodied in the
strident constrictives, namely from their markedly noisy pattern.
One and the same split of the consonantal feature, on the one hand,
and of the non-vocalic feature, on the other, is manifested both in
PHONEMIC PATTERNING 43
the appearance of the liquids and of the strident stops. This
explains the "strange but widespread" interchangeability of strident
stops and liquids, especially laterals, that Bouda notes in ManchuTungus
and in Paleosiberian languages.^"
Since nasality, by superimposing a clear-cut formant structure
upon the consonantal pattern, brings consonants closer to vowels,
and since liquids combine the consonantal with the vocalic feature,
it is advantageous to range these two related classes of phonemes
under a common heading of sonorants. The consonantal
character of these two classes is again reinforced in such relatively
rare phonemes as the discontinuous nasals (the so-called prenasalized
stops) and the strident liquids (the sibilant laterals or
vibrant).
The oral phonemes with an obstructed vocal tract have a noise
source at the obstruction and may use voice - if ever - only as a
supplementary source, whereas for the phonemes with an open
tract, the voice is the main source. While the optimal consonant is
voiceless and the optimal vowel voiced, the voicing of consonants
or, in very rare instances the unvoicing of vowels, may be utilized as
one of the various attenuations of the maximum contrast CV.
Since the consonant is primarily characterized by reduction
of energy, the optimal consonant is lax but may be subsequently
opposed by a tense consonant that attenuates the contrast between
consonant and vowel. Normally, however, the voiced consonant
is of lower energy than the voiceless and therefore, in the opposition
of tense and lax consonants, the laxness is frequently accompanied
by voicing and the tenseness by voicelessness, so that the consonant,
optimal in one respect - reduction oFenergy - deviates from the
consonantal optimum in another way, the presence of voice.
If both oppositions act autonomously in a language, the doubly
optimal consonant, lax and voiceless, is opposed by two phonemes,
one, a voiceless tense and the other, a voiced lax, both of which,
in different ways, shift the structure of the consonant toward that
of the vowel. A further move in this direction is a consonant
30 K. Bouda, 'Lateral and Sibilant,' Zeitschrift fur Phonetik, I (1947).
44 PHONOLOGY AND PHONETICS
endowed with the distinctive features of tenseness and voicing,
such as /d7 in some languages of India.
Normally, the total energy of a vowel increases along with
concentration of energy (compactness), but in a tense vowel,
as compared with the corresponding lax vowel, the total energy
increases, whereas the concentration of energy decreases. This
reversal separates the tense vowels from the vocalic optimum.
While reducing the time, the checked consonants increase their
energy and thus attenuate the consonantal optimum. If a language
possesses the two oppositions, checked/unchecked and tense/lax,
then the optimal consonant, lax and unchecked, is opposed by two
phonemes, the one checked (glottalized), the other tense. A double
attenuation of the consonantal optimum may be further presented
by the rare combination of two distinctive features, tense and
checked within one and the same phoneme, such as the Avar /K'/.
Thus, all the inherent distinctive features actually rest upon two
axes. On the one hand, the oppositions bearing upon the
sonority axis display various fissions and attenuations of the
primary contrast between optimal consonant and optimal vowel,
and thus give rise to more minute and specific distinctions. On
the other hand, those oppositions that involve the tonality
axis, perpendicular to the sonority axis, emerge originally as
the counterpart and corollary of the contrast, "optimal vowel vs.
optimal consonant" and, subsequently, as the corollary of the
opposition, "optimal, compact vowel vs. attenuated, diffuse vowel"
or "optimal, diff'use consonant vs. attenuated, compact consonant".
4.2 The dichotomous scale. In their recent, quite autonomous
development, phonemic analysis and the mathematical theory of
communication arrived at fundamentally similar and mutually
complementary conclusions, making possible a most productive
cooperation on both sides. ^^ Any spoken message presents the
*^ For procedures of the communication theory utiiizable in phonemic
analysis, see especially C. E. Shannon and W. Weaver, The mathematical theory
of communication (Urbana, 1949); C. E. Shannon, 'The redundancy of English,'
Cybernetics, Transactions of the Seventh Conference (New York, 1951);
D. M. Mackay, 'In search of basic symbols,' Cybernetics, Transactions of the
PHONEMIC PATTERNING 45
listener with two complementary alignments of information: on
the one hand, the chain of phonemes yields sequentially encoded
information, on the other hand, every phoneme is composed of
several distinctive features. The totality of these features is the
minimum number of binary selections necessary for the specification
of the phoneme. In reducing the phonemic information
contained in the sequence to the smallest number of alternatives, we
find the most economical and consequently the optimal solution:
the minimum number of the simplest operations that would suffice
to encode and decode the whole message. When analyzing a
given language into its ultimate constituents, we seek the smallest
set of distinctive oppositions which allow the identification of each
phoneme in the messages framed in this language. This task
requires an isolation of distinctive features from concurrent or
adjoining redundant features.
If in a language, one and the same phoneme is implemented as
a palatal stop before /i/, as a post-alveolar affricate before /e/ and
as a velar stop in all other positions, the invariant must be determined
as a compact (forward-flanged) consonant, distinct from
the diffuse (backward-flanged) consonants /p/ and /t/ of the same
language. While in such a case the redundant features are conditioned
by the diverse distinctive features of the following phoneme,
a striking example of redundant features linked to concurrent
distinctive features may be found in the French consonantal
pattern. Here, the compactness of the consonant is implemented
by a velar articulation when lumped with plosiveness in /k/ and
/g/, by a palatal articulation when lumped with nasality in /ji/,
and by a post-alveolar articulation when lumped with constrictiveness
in /J/ and jzlSuch
a delimitation of distinctive and redundant features not
only permits an identification of all the phonemes involved but is
Eighth Conference (New York, 1952); D. Gabor, 'Lectures on communication
theory,' M.I.T., Research Laboratory of Electronics, Report, No. 238 (1953);
E. C. Cherry, Human communication (to be published by Wiley and Sons,
together with The Technology Press). Cf. L Pollack, 'Assimilation of sequentially
encoded information," American Journal of Psychology, LXVI (1953).
46 PHONOLOGY AND PHONETICS
the unique solution, since any different analysis of these five
phonemes deviates from the optimal solution. All fifteen French
consonant phonemes in the proposed test require only five binary
decisions: nasal/oral, and if oral then continuant/discontinuous,
and tense/lax; compact/diffuse, and if diffuse then grave/acute.
Each French consonant contains from two (compact nasal) to five
distinctive features.
If one deemed the point of articulation distinctive, and the
difference between constrictive and stop redundant, then the six
French voiceless consonants - velar /k/, post-alveolar /J/, alveolar
/s/, dental /t/, labiodental /f/ and bilabial /p/^^ - would require,
for their identification, fifteen distinctions instead of three,
according to the elementary mathematical formula cited by
Twaddell (1935): "If x is the maximum number of significant
phonological differentiations within a given articulatory range in
a language, then 2x =/?(«- 1), where n is the maximum number
of phonemes in that range." Some of the minute differences in the
point of articulation have, moreover, the disadvantage of being
acoustically hardly recognizable by themselves. Finally, such
distinctions as /s/ vs. /f/ and /t/ vs. /p/ present an identical differential
criterion, namely the opposition of an acute and grave consonant
due to the same difference in the size and shape of the mouth
resonator. Also /k/ vs. /t/ and /J/ vs. /s/ display (acoustically as
well as genetically) one and the same opposition, due to a parallel
relation of the front and back resonators, so that the operation
with the two pairs as if they were distinguished by two separate
features introduces superfluous redundancies.
The reduction of language into distinctive features must be
consistent. If, for instance, the Czech /!/, which can occur in
identical positions with each of the 32 other phonemes of the
language, is declared "an unanalyzable distinctive unit", its
distinction from the other 32 phonemes would require 32 unanalyzable
relations, whereas the dissolution of the /I/ bundle into its
three features - vocalic, consonantal and continuous - reduces
'- See L. E. Armstrong, The phonetics of French (London, 1932).
\
PHONEMIC PATTERNING 47
its relation to all other phonemes of the pattern to three binary
selections.
The maximum elimination of redundancies and the minimum ,
amount of distinctive alternatives is a principle that permits an
affirmative answer to the focal problem raised by Chao in 1934
on whether the task of breaking down a given language into its
ultimate components yields a unique solution. ^^ Not less crucial
is his more recent question (1954), whether the dichotomous
scale is the pivotal principle which the analyzer can profitably
impose upon the linguistic code or whether this scale is inherent in
the structure of language.^* There are several weighty arguments in
favor of the latter solution.
First, a system of distinctive features based on a mutually/
implicating relation between the terms of each binary opposition!
is the optimal code and it is unwarranted to assume that the
speech participants in their encoding and decoding operations use
a more complicated and less economic set of differential criteria.
Recent experiments have disclosed that multidimensional auditory
displays are most easily learned and perceived when "binary-
coded".^^
Second, the phonemic code is acquired in the earliest years of
childhood and, as psychology reveals, in a child's mind the pair
is anterior to isolated objects. ^^ The binary opposition is a child's
first logical operation. Both opposites arise simultaneously and
force the infant to choose one and to suppress the other of the two
alternative terms.
Third, almost all of the distinctive features show an unquestion^^
Y. R. Chao, 'The non-uniqueness of phonemic solution of phonetic
systems,' Academia Sinica, Institute of History and Philology, Bulletin, IV
(Shanghai, 1934).
'"
Y. R. Chao, review of Jakobson, Fant, Halle, Preliminaries ..., in
Romance Pliilology, VIII (1954).
^^ I. Pollack and L. Picks, 'Information of elementary multi-dimensional
auditory displays,' Journal of the Acoustical Society oj America, XXVI (1954).
^^ See H. Wallon, Les origines cle la pensee chez Fenfant, I (Paris, 1945).
For the pivotal role of gradual binary fissions in child development, cf. T.
Parsons and R. F. Bales, Family, socialization ami interaction process {G\encoe,
1955).
48 PHONOLOGY AND PHONETICS
ably dichotomous structure on their acoustical and, correspondingly,
on their motor level. Among the inherent features, only the
vocalic distinction compact/diffuse often presents a higher number
of terms, mostly three. For instance, /ae/ is to /e/ as /e/ is to /i/:
the geometric mean /e/ is non-compact in relation to jxl and nondiffuse
in relation to /i/. Psychological experiments that obtained
/e/ through the mixture of jxj and /i/ confirm the peculiar structure
of this vocalic feature. ^^ Parallel experiments in mixing vowels
situated on the tonality axis showed that grave and acute vowels,
when sounded simultaneously, are not perceived as a single vowel:
/u/ and I'll do not merge into /y/. The feature grave/acute is a
patently binary opposition. Since the second formant in /y/ is
higher than in /u/ and lower than in /i/ and since in cavity length /y/
occupies a middle position in relation to /u/ with its longest, and /i/
with its shortest resonator, attempts were made to get along with
one dimension for all three vowels. ^^ But the chief genetic distinction
is cardinally dilTerent: the disparity in the size of the lip orifice
is primarily responsible for the relation jyj vs. /i/, and the disparity
in the size and shape of the resonator itself for the relation /y/ vs. /u/.
_0n the acoustical level, the distinction of grave from acute in
vowels is manifested in a relative proximity of the first and second
formants which has, as a consequence, a very striking weakening
of the upper formants, whereas the distinction of flat from plain is
mairtly due to the lowering in the second formant. ^^
'"
See K. Huber, 'Die Vokalmischung und das Qualitatensystem der
Vokale,' Archiv fur Psychologie, XCI (1934).
'* See, e.g., F. Delattre, The physiological interpretation of sound spectrograms,'
PMLA, LXVI (1951).
** Cf. Jakobson, Fant, Halle, Preliminaries . . ., p. 48; H. K. Dunn, 'The
calculation of vowel resonances, and an electrical vocal tract,' Journal of the
Acoustical Society of America, XXII (1950), p. 650; K. N. Stevens and A. S.
House, 'Development of a quantitative description of vowel articulation,'
ibidem, XXVII (1955); detailed data are presented by Fant and Halle in the
first volume of the series The Structure of Contemporary Standard Russian
(to be published by Mouton & Co.. The Hague).
•" See especially L. Barczinski and E. Thienhaus, 'Klangspektren und
Lautstarke deutscher Sprachlaute,' Archives neerlandaises de phonetique
experimentale, XI (1935).
PHONEMIC PATTERNING 49
Similarly, the effort to project the vocalic oppositions tense/lax
and compact/diffuse upon one and the same line is hampered
by the salient difference between their physical essence,^" by the
dissimilar parts they play in linguistic structure and by the considerable
disadvantages which their unidimensional treatment imposes
upon the analysis.
Finally, the application of the dichotomous scale makes the
stratified structure of the phonemic patterns, their governing laws
of implication, and the conclusive typology of languages so
transparent that the inherence of this scale in the linguistic system
is quite manifest.
4.3 The spatio-temporal pattern of phonemic operations. If there
is a difference betv/een the linguistic patterns of two speech
communities, interlocution between members of the two communities
demands an adjustment of the listener to the speaker and/or
of the speaker to the listener. This adjustment may involve all
aspects of language or only a few of them. Sometimes the phonemic
code is the only one affected. Both on the listener's and on the
speaker's side there are different degrees of this adjustment process,
neatly called code switching by the communication engineers.
The receiver in trying to understand the sender, and/or the sender in
trying to make himself understood, concentrate their attention on
the common core of their codes. A higher degree of adjustment is
represented in the effort to overcome the phonemic differences by
switching rules, heightening the intelligibility of the message for
its addressee. Having found these clues, the interlocutor may try
to use them not only as a listener, but also more actively, in
adapting his own utterances to the pattern of his addressee.
The phonemic adjustment may cover the whole lexical stock or the
imitation of the neighbor's phonemic code may be confined to a
certain set of words directly borrowed from the neighbor or at least
particularly stamped by his use of them. Whatever the adjustments
are, they help the speaker to increase the radius of communication,
and if often practiced, they are likely to enter into his
everyday language. Under favorable circumstances they may
50 PHONOLOGY AND PHONETICS
subsequently infiltrate into the general use of the speech community,
either as a particular speech fashion or as a new pattern fully
substituted for the former norm. Interdialectal communication
and its influence on intradialectal communication must be analyzed
from a linguistic and, particularly, from a phonemic point of
view,^^
The problem of bridging space stops neither at the borders of
distant and highly differentiated dialects, nor at the boundaries of
related or even unrelated languages. Mediators, more or less
bilingual, adapt themselves to the foreign phonemic code. Their
prestige grows with the widening radius of their audience and may
further a diffusion of their innovations among their monolingual
tribesmen.
Not only the interdialectal, but also the interlingual adjustment
may affect the phonemic code without limitation to borrowed
words or even without any lexical borrowing. In all parts of the
world, linguists have been surprised, as Sapir confesses, to observe
"the remarkable fact that distinctive phonetic features tend to be
distributed over wide areas regardless of the vocabularies and
structures of the languages involved. "^^ This far-reaching phenomenon
still awaits systematic mapping and study in connection with
the equally urgent inquiry into the typology of phonemic patterns.
The other possibility of phonemic adjustments to a different
dialect or foreign language is a partial or total preservation of its
phonemic structure in borrowed words. As noted repeatedly in
phonemic literature, and closely examined by Fries and Pike, "the
speech of monohngual natives of some languages is comprised of
more than one phonemic system. "^^ Such a coexistence of two
systems within one language is due either to a phonemic difference
betv/een the original vocabulary and unassimilated loanwords, or
*^ See 'Results of the Conference of Anthropologists and Linguists,'
Indiana University Publications in Anthropology and Linguistics, VIII (1953),
pp. 16 f., 36ff.
" S. Sapir, 'Language,' Selected Writings (Berkeley and Los Angeles, 1949),
p. 25.
*' C. C. Fries and K. L. Pike, 'Coexistent phonemic systems,' Language,
XXV (1949).
I
PHONEMIC PATTERNING 51
to the use of two patterns, one native and the other imitative, as
different styles of speech. Thus phenomena of space, namely
interdialectal or interUngual isoglosses, especially isophones, may
be projected into the framework of a single dialect, individual or
social.
The same statement, mutatis mutandis, may be made about the
time factor in language, particularly in the phonemic field. Any
sound change in progress is a synchronic fact. Both the start and
the finish of a change coexist for a certain length of time. If the
change differentiates the younger generation from the older, there
is always some intercourse between the two generations, and the
receiver belonging to one is accustomed to recode messages from
a sender of the other. Furthermore, the initial and the final stage
may co-occur in the use of one and the same generation as two
stylistic levels: on the one hand, a more conservative and solemn,
on the other, a more fashionable way of talking. Thus synchronic
analysis must encompass linguistic changes and, vice versa,
linguistic changes may be comprehended only in the light of
synchronic analysis.
The decisive factor in phonemic changes and in the diffusion
of phonemic phenomena is the shift in the code. The interpretation
of events in time and space is primarily concerned with the question
in what respect the structure of the code is affected by such shifts.
The motor and physical aspects of these innovations cannot be
treated as self-sufficient agents, but must be subordinated to the
strictly linguistic analysis of their role in the coding system.
PART II
TWO ASPECTS OF LANGUAGE
AND TWO TYPES OF APHASIC
DISTURBANCES
BY
ROMAN JAKOBSON
I
i
I
APHASIA AS A LINGUISTIC PROBLEM
If aphasia is a language disturbance, as the term itself suggests
then any description and classification of aphasic syndromes must
begin with the question of what aspects of language are impaired
in the various species of such a disorder. This problem, which
was approached long ago by Hughhngs Jackson,^ cannot be solved
without the participation of professional linguists familiar with
the patterning and functioning of language. To study adequately
any breakdown in communications we must first understand the
nature and structure of the particular mode of communication
that has ceased to function. Linguistics is concerned with language
in all its aspects - language in operation, language in drift,^
language in the nascent state, and language in dissolution.
At present there are psychopathologists who assign a high importance
to the linguistic problems involved in the study of language
disturbances;^ some of these questions have been touched upon in
the best recent treatises on aphasia.^ And yet, in most cases, this
valid insistence on the linguists' contribution to the investigation
of aphasia is still ignored. For instance, a new book dealing to a
* Hughlings Jackson, Papers on affections of speech (reprinted and
commented by H. Head), Brain, XXXVIIl (1915).
^ E. Sapir, Language (New York, 1921), Chapter VII: 'Language as a
historical product ; drift.'
^ See, for instance, the discussion on aphasia in the Nederlandsche Vereeniging
voor Phonetische Wetenschappen with papers by the linguist J. van
Ginneken and by two psychiatrists, F. Grewel and V. W. D. Schenk, Psychiatrische
en Neurologische Bladen, XLV (1941), p. 1035ff. ; cf. furthermore,
F. Grewel, 'Aphasie en linguistiek,' Nederlandsch Tijdschrift voor Geneeskunde,
XCIII (1949), p. 726ff.
* A. R. Luria, Travmaticeskaja afazija (Moscow, 1947); Kurt Goldstein,
Language and language disturbances (New York, 1948); Andre Ombredane,
Uaphasie et Velaboration de la pensee expHcite (Paris, 1951).
56 TWO ASPECTS OF LANGUAGE
great extent with the complex and intricate problems of infantile
aphasia, calls for a coordination of various disciphnes and appeals
for cooperation to otolaryngologists, pediatricians, audiologists,
psychiatrists, and educators; but the science of language is passed
over in silence, as if disorders in speech perception had nothing
whatever to do with language. ^ This omission is the more deplorable
since the author is Director of the Child Hearing and Aphasia
Clinics at Northwestern University, which counts among its
linguists Werner F. Leopold, by far the best American expert in
child language.
The Hnguists are also responsible for the delay in undertaking a
joint inquiry into aphasia. Nothing comparable to the minute
linguistic observations of infants of various countries has been
performed with respect to aphasics. Nor has there been any
attempt to re-interpret and systematize from the point of view of
linguistics the multifarious clinical data on diverse types of aphasia.
That this should be true is all the more surprising in view of the
fact that, on the one hand, the amazing progress of structural
linguistics has endowed the investigator with efficient tools and
methods for the study of verbal regression and, on the other, the
aphasic disintegration of the verbal pattern may provide the
linguist with new insights into the general laws of language.
The application of purely linguistic criteria to the interpretation
and classification of aphasic facts may substantially contribute to
the science of language and language disturbances, provided that
linguists remain as careful and cautious when dealing with psychological
and neurological data as they have been in their traditional
field. First of all, they should be familiar with the technical terms
and devices of the medical disciplines dealing with aphasia, then
they must submit the clinical case reports to thorough linguistic
analysis, and, further, they should themselves work with aphasic
patients in order to approach the cases directly and not only through
a re-interpretation of prepared records which have been quite
differently conceived and elaborated.
* H. Myklebust, Auditory disorders in children (New York, 1954).
APHASIA AS A LINGUISTIC PROBLEM 57
There is one level of aphasic phenomena where amazing agreement
has been achieved during the last twenty years between those
psychiatrists and linguists who have tackled these problems, namely
the disintegration of the sound pattern.* This dissolution exhibits
a time order of great regularity. Aphasic regression has proved to
be a mirror of the child's acquisition of speech sounds, it shows the
child's development in reverse. Furthermore, comparison of child
language and aphasia enables us to establish several laws of
implication. This search for the order of acquisitions and losses
and for the general laws of implication cannot be confined to the
phonemic pattern but must be extended also to the grammatical
system. Only a few tentative trials have been made in this direction,
and these efforts deserve to be continued.'
' The aphasic impoverishment of the sound pattern was observed and
discussed by the linguist Marguerite Durand together with the psychopathologists
Th. Alajouanine and A. Ombredane (in their common work Le syndrome
lie desintegration phonetique dans Faphasie, Paris, 1939) and by R. Jakobson
(the first draft presented to the International Congress of Linguists at Brussels
in 1939 - see N. Trubetzkoy, Principes de plwnologie, Paris, 1949, pp. 367-79 was
later developed into an outline, 'Kindersprache, Aphasie und allgemeine
Lautgesetze,' Uppsala Universitets Arsskrift, 1942 : 9) and has been further
studied in his Sound and Meaning (to be published by Wiley and Sons, together
with The Technology Press). Cf. K. Goldstein, p. 32ff.
' A joint inquiry into certain grammatical disturbances was undertaken
at the Bonn University Clinic by a linguist, G. Kandler, and two physicians,
F. Panse and A. Leischner: see their report, Klinische und sprachwissenschaftliche
Untersucliungen zum Agrammatismus (Stuttgart, 1952).
II
THE TWOFOLD CHARACTER OF LANGUAGE
Speech implies a selection of certain linguistic entities and their
combination into linguistic units of a higher degree of complexity.
At the lexical level this is readily apparent: the speaker
selects words and combines them into sentences according to the
syntactic system of the language he is using; sentences are in their
turn combined into utterances. But the speaker is by no means a
completely free agent in his choice of words; his selection (except
for the rare case of actual neology) must be made from the lexical
storehouse which he and his addressee possess in common. The
communication engineer most properly approaches the essence of
the speech event when he assumes that in the optimal exchange of
information the speaker and the listener have at their disposal
more or less the same "filing cabinet of prefabricated representations":
the addresser of a verbal message selects one of these
"preconceived possibihties" and the addressee is supposed to make
an identical choice from the same assembly of "possibilities already
foreseen and provided for".^ Thus the efficiency of a speech event
demands the use of a common code by its participants.
" 'Did you say pig or figT said the Cat. 'I said pig' replied
Alice."^ In this peculiar utterance the feline addressee attempts
to recapture a linguistic choice made by the addresser. In the
common code of the Cat and Alice, i.e. in spoken English, the
difference between a stop and a continuant, other things being
equal, may change the meaning of the message. Alice had used
the distinctive feature "stop vs. continuant", rejecting the latter
and choosing the former of the two opposites ; and in the same act
* D. M. MacKay, 'In search of basic symbols,' Cybernetics, Transactions
of the Eighth Conference (New York, 1952), p. 183.
' Lewis Carroll, Alice's adventures in Wonderland, Chapter VI.
THE TWOFOLD CHARACTER OF LANGUAGE 59
of speech she combined this solution with certain other simultaneous
features, using the gravity and the tenseness of /p/ in
contradistinction to the acuteness of /t/ and to the laxness of /b/.
Thus all these attributes have been combined into a bundle of
distinctive features, the so-called phoneme. The phoneme /p/ was
then followed by the phonemes /i/ and /g/, themselves bundles of
simultaneously produced distinctive features. Hence the concurrence
of simultaneous entities and the concatenation of
successive entities are the two ways in which we speakers combine
linguistic constituents.
Neither such bundles as /p/ or /f/ nor such sequences of bundles
as /pig/ or /fig/ are invented by the speaker who uses them. Neither
can the distinctive feature "stop versus continuant" nor the phoneme
/p/ occur out of a context. The stop feature appears in combination
with certain other concurrent features, and the repertory
of combinations of these features into phonemes such as /p/, /b/,
/t/, /d/, /k/, /g/, etc. is limited by the code of the given language.
The code sets limitations on the possible combinations of the
phoneme /p/ with other following and/or preceding phonemes; and
only a part of the permissible phoneme-sequences are actually
utilized in the lexical stock of a given language. Even when other
combinations of phonemes are theoretically possible, the speaker, as
a rule, is only a word-user, not a word-coiner. When facing with
individual words, we expect them to be coded units. In order to
grasp the word nylon one must know the meaning assigned to this
vocable in the lexical code of modern English.
In any language, there exist also coded word-groups called
phrase-words. The meaning of the idiom how do you do cannot
be derived by adding together the meanings of its lexical constituents;
the whole is not equal to the sum of its parts. Those
word-groups which in this respect behave like single words are a
common but nonetheless only marginal case. In order to comprehend
the overwhelming majority of word-groups, we must be
familiar only with the constituent words and with the syntactical
rules of their combination. Within these limitations we are free to
set words in new contexts. Of course, this freedom is relative, and
60 TWO ASPECTS OF LANGUAGE
the pressure of current cliches upon our choice of combinations is
considerable. But the freedom to compose quite new contexts is
undeniable, despite the relatively low statistical probability of
their occurrence.
Thus in the combination of linguistic units there is an ascending
scale of freedom. In the combination of distinctive features into
phonemes, the freedom of the individual speaker is zero; the code
has already established all the possibilities which may be utilized
in the given language. Freedom to combine phonemes into words
is circumscribed, it is limited to the marginal situation of wordcoinage.
In the forming of sentences out of words the speaker
is less constrained. And finally, in the combination of sentences
into utterances, the action of compulsory syntactical rules ceases
and the freedom of any individual speaker to create novel contexts
increases substantially, although again the numerous stereotyped
utterances are not to be overlooked.
Any linguistic sign involves two modes of arrangement.
1) Combination. Any sign is made up of co'nstituent signs
and/or occurs only in combination with other signs. This means
that any linguistic unit at one and the same time serves as a context
for simpler units and/or finds its own context in a more complex
linguistic unit. Hence any actual grouping of linguistic units binds
them into a superior unit: combination and contexture are two
faces of the same operation.
2) Selection. A selection between alternatives impUes the
possibility of substituting one for the other, equivalent to
the former in one respect and different from it in another.
Actually, selection and substitution are two faces of the same
operation.
The fundamental role which these two operations play in language
was clearly realized by Ferdinand de Saussure. Yet from the
two varieties of combination - concurrence and concatenation it
was only the latter, the temporal sequence, which was recognized
by the Geneva linguist. Despite his own insight into the phoneme
as a set of concurrent distinctive features (elements differentiels des
phonemes), the scholar succumbed to the traditional belief in the
THE TWOFOLD CHARACTER OF LANGUAGE 61
linear character of language "qui exclut la possibilite de prononcer
deux elements a lafois"".^^
In order to delimit the two modes of arrangement which we have
described as combination and selection, F. de Saussure states that
the former "is in presentia: it is based on two or several terms
jointly present in an actual series," whereas the latter "connects
'
terms in absentia as members of a virtual mnemonic series". That
is to say, selection (and, correspondingly, substitution) deals with
entities conjoined in the code but not in the given message, whereas,
in the case of combination, the entities are conjoined in both or only
in the actual message. The addressee perceives that the given
utterance (message) is a combination of constituent parts
(sentences, words, phonemes, etc.) selected from the repository
of all possible constituent parts (code). The constituents of a
context are in a status of contiguity, while in a substitution set
signs are linked by various degrees of similarity which fluctuate
between the equivalence of synonyms and the common core of
antonyms.
These two operations provide each linguistic sign with two sets of
interpretants, to utilize the effective concept introduced by Charles
Sanders Peirce:^^ there are two references which serve to interpret
the sign - one to the code, and the other to the context, whether
coded or free; and in each of these ways the sign is related to
another set of linguistic signs, through an alternation in the
former case and through an alignment in the latter. A given
significative unit may be replaced by other, more explicit signs of
the same code, whereby its general meaning is revealed, while its
contextual meaning is determined by its connection with other
signs within the same sequence.
The constituents of any message are necessarily linked with the
code by an internal relation and with the message by an external
relation. Language in its various aspects deals with both modes of
"* F. de Saussure, Cours de linguistiqite generate, 2nd ed. (Paris, 1922),
pp. 68f. and 170f.
'1
C. S. Peirce, Collected Papers, II and IV (Cambridge, Mass., 1932,
1934) - see Index of subjects.
62 TWO ASPECTS OF LANGUAGE
relation. Whether messages are exchanged or communication
proceeds unilaterally from the addresser to the addressee, there must
be some kind of contiguity between the participants of any speech
event to assure the transmission of the message. The separation
in space, and often in time, between two individuals, the addresser
and the addressee, is bridged by an internal relation: there must
be a certain equivalence between the symbols used by the addresser
and those known and interpreted by the addressee. Without such
an equivalence the message is fruitless - even when it reaches the
receiver it does not affect him.
i
Ill
SIMILARITY DISORDER
It is clear that speech disturbances may affect in varying degrees
the individual's capacity for combination and selection of linguistic
units, and indeed the question of which of these two operations is
chiefly impaired proves to be of far-reaching significance in describing,
analyzing and classifying the diverse forms of aphasia.
This dichotomy is perhaps even more suggestive than the classical
distinction (not discussed in this paper) between emissive and
receptive aphasia, indicating which of the two functions in
speech exchange, the encoding or the decoding of verbal messages,
is particularly affected.
Head attempted to classify cases of aphasia into definite groups,^^
and to each of these varieties he assigned "a name chosen to
signify the most salient defect in the management and comprehension
of words and phrases" (p. 412). Following this device, we
distinguish two basic types of aphasia - depending on whether
the major deficiency lies in selection and substitution, with relative
stability of combination and contexture; or conversely, in combination
and contexture, with relative retention of normal selection
and substitution. In outlining these two opposite patterns of
aphasia, I shall utilize mainly Goldstein's data.
For aphasics of the first type (selection deficiency), the context
is the indispensable and decisive factor. When presented with
scraps of words or sentences, such a patient readily completes
them. His speech is merely reactive: he easily carries on conversation,
but has diflftculties in starting a dialogue; he is able to reply
to a real or imaginary addresser when he is, or imagines himself to
be the addressee of the message. It is particularly hard for him to
" H. Head. Aphasia and kindred disorders of speech, I (New York. 1926).
64 TWO ASPECTS OF LANGUAGE
perform, or even to understand, such a closed discourse as the
monologue. The more his utterances are dependent on the context,
the better he copes with his verbal task. He feels unable to utter
a sentence which responds neither to the cue of his interlocutor
nor to the actual situation. The sentence "it rains" cannot be
produced unless the utterer sees that it is actually raining. The
deeper the utterance is embedded in the verbal or non-verbalized
context, the higher are the chances of its successful performance
by this class of patients.
Likewise, the more a word is dependent on the other words of
the same sentence and the more it refers to the syntactical context,
the less it is affected by the speech disturbance. Therefore words
syntactically subordinated by grammatical agreement or government
are more tenacious, whereas the main subordinating agent
of the sentence, namely the subject, tends to be omitted. As long
as any start is the main obstacle for the patient, it is obvious that
he will fail precisely at the starting point, the cornerstone of the
sentence-pattern. In this type of language disturbance, sentences
are conceived as elliptical sequels to be supplied from antecedent
sentences uttered, if not imagined, by the aphasic himself, or
received by him from the other partner of the colloquy, actual if not
imaginary. Key words may be dropped or superseded by abstract
anaphoric substitutes.^^ A specific noun, as Freud noticed, is
replaced by a very general one, for instance machin, chose in the
speech of French aphasics.^^ In a dialectal German sample of
"amnesic aphasia" observed by Goldstein (p. 246ff.), Ding 'thing'
or Stiickle 'piece' were substituted for all inanimate nouns, and
iiberfahren 'perform' for verbs which were identifiable from the
context or situation and therefore appeared superfluous to the
patient.
Words with an inherent reference to the context, hke pronouns
and pronominal adverbs, and words serving merely to construct
the context, such as connectives and auxiliaries, are particularly
^^ Cf. L. Bloomfield, Language (New York, 1933), Chapter XV: Sub-
stitution.
" S. Freud, On aphasia (London, 1953), p. 22.
SIMILARITY DISORDER 65
prone to survive. A typical utterance of a German patient, recorded
by Quensel and quoted by Goldstein (p. 302) will serve as illu-
stration:
„Ich bin doch hier unten, na wenn ich gewesen bin ich wees nicht, we
das, nu wenn ich, ob das nun doch, noch, ja. Was Sie her, wenn ich,
och ich weess nicht, we das hier war ja. .
."
Thus only the framework, the connecting links of communication,
is spared by this type of aphasia at its critical stage.
In the theory of language since the early Middle Ages, it has
repeatedly been asserted that the word out of context has no
meaning. The validity of this statement is, however, confined to
aphasia, or more exactly to one type of aphasia. In the pathological
cases under discussion an isolated word means actually nothing
but "blab". As numerous tests have disclosed, for such patients
two occurrences of the same word in two different contexts are
mere homonyms. Since distinctive vocables carry a higher amount
of information than homonyms, some aphasics of this type tend
to supplant the contextual variants of one word by different terms,
each of them specific for the given environment. Thus Goldstein's
patient never uttered the word knife alone, but according to its use
and surroundings, alternately called the knife pencil-sharpener,
apple-parer, bread-knife, knife-and-fork (p. 62); so that the word
knife was changed from a free form, capable of occurring alone,
into a bound form.
"I have a good apartment, entrance hall, bedroom, kitchen,"
Goldstein's patient says. "There are also big apartments, only in
the rear live bachelors." A more explicit form, the word-group
unmarriedpeople, could have been substituted for bachelors, but this
univerbal term was selected by the speaker. When repeatedly
asked what a bachelor was, the patient did not answer and was
"apparently in distress" (p. 270). A reply like "a bachelor is an
unmarried man" or "an unmarried man is a bachelor" would
present an equational predication and thus a projection of a substitution
set from the lexical code of the English language into the
context of the giveh message. The equivalent terms beco ne two
66 TWO ASPECTS OF LANGUAGE
correlated parts of the sentence and consequently are tied by
contiguity. The patient was able to select the appropriate term
bachelor when it was supported by the context of a customary
conversation about "bachelor apartments", but was incapable of
utilizing the substitution set bachelor = unmarried man as the topic
of a sentence, because the abihty for autonomous selection and
substitution had been affected. The equational sentence vainly
demanded from the patient carries as its sole information:
"«bachelor» means an unmarried man" or "an unmarried man is
called «bachelor»".
The same difficulty arises when the patient is asked to name an
object pointed to or handled by the examiner. The aphasic with a
defect in substitution will not supplement the pointing or handhng
gesture of the examiner with the name of the object pointed to.
Instead of saying "this is [called] a pencil," he will merely add an
eUiptical note about its use: "To write." If one of the synonymic
signs is present (as for instance the word bachelor or the pointing
to the pencil) then the other sign (such as the phrase unmarried
man or the word pencil) becomes redundant and consequently
superfluous. For the aphasic, both signs are in complementary
distribution: if one is performed by the examiner, the patient will
avoid its synonym: "I understand everything" or "Ich weiss es
schon" will be his typical reaction. Likewise, the picture ofan object
will cause suppression of its name : a verbal sign is supplanted by a
pictorial sign. When the picture of a compass was presented to a
patient of Lotmar's, he responded: "Yes, it's a ... I know what
it belongs to, but I cannot recall the technical expression . . . Yes . .
direction ... to show direction ... a magnet points to the north. "^^
Such patients fail to shift, as Peirce would say, from an index or
icon to a corresponding verbal symbol.^^
Even simple repetition of a word uttered by the examiner seems
1° F. Lotmar, 'Zur Pathophysiologic der erschwerten Wortfindung bei
Aphasischen,' Schweiz. Archiv fiir Neurologic und Psychiatrie, XXXV (1933),
p. 104.
" C. S. Peirce, 'The icon, index and symbol,' Collected papers, II (Cambridge,
Mass., 1932).
SIMILARITY DISORDER 67
to the patient unnecessarily redundant, and despite instructions
received he is unable to repeat it. Told to repeat the word "no,"
Head's patient repHed "No, I don't know how to do it." While
spontaneously using the word in the context of his answer ("No,
I don't . .
.
"), he could not produce the purest form of equational
predication, the tautology a = a : «no» is «no».
One of the important contributions of symbolic logic to the
science of language is its emphasis on the distinction between
object language and metalanguage. As Carnap states, "in
order to speak about any object language, we need a metalanguage.
"^'^
On these two different levels of language the same linguistic stock
may be used; thus we may speak in English (as metalanguage)
about English (as object language) and interpret English words
and sentences by means of English synonyms, circumlocutions
and paraphrases. Obviously such operations, labeled metalinguistic
by the logicians, are not their invention: far from
being confined to the sphere of science, they prove to be an integral
part of our customary linguistic activities. The participants in a
dialogue often check whether both of them are using the same code.
"Do you follow? Do you see what I mean?", the speaker asks,
or the listener himself breaks in with "What do you mean?" Then,
by replacing the questionable sign with another sign from the same
linguistic code, or with a whole group of code signs, the sender of
the message seeks to make it more accessible to the decoder.
The interpretation of one hnguistic sign through other, in some
respect homogeneous signs of the same language, is a metahnguistic
operation which also plays an essential role in child language
learning. Recent observations have disclosed what a considerable
place talk about language occupies in the verbal behavior of preschool
children. Recourse to metalanguage is necessary both for the
acquisition of language and for its normal functioning. The
aphasic defect in the "capacity of naming" is properly a loss of
metalanguage. As a matter of fact, the examples of equational
predication sought in vain from the patients cited above, are meta^'
R. Carnap, Meaning and necessity (Chicago, 1947), p. 4.
68 TWO ASPECTS OF LANGUAGE
linguistic propositions referring to the English language. Their
explicit wording would be: "In the code that we use, the name of
the indicated object is «pencil»" ; or "In the code we use, the word
«bachelor» and the circumlocution «unmarried man» are equiva-
lent."
Such an aphasic can neither switch from a word to its Synonyms
and circumlocutions, nor to its heteronyms, i.e. equivalent expressions
in other languages. Loss of a polyglot ability and confinement
to a single dialectal variety of a single language is a
symptomatic manifestation of this disorder.
According to an old but recurrent bias, a single individual's way
of speaking at a given time, labeled idiolect, has been viewed as
the only concrete linguistic reality. In the discussion of this
concept the following objections were raised
Everyone, when speaking to a new person, tries, deliberately or
involuntarily, to hit upon a common vocabulary: either to please or
simply to be understood or, finally, to bring him out, he uses the terms
of his addressee. There is no such thing as private property in language:
everything is socialized. Verbal exchange, like any form of intercourse,
requires at least two communicators, and idiolect proves to be a somewhat
perverse fiction. ^^
This statement needs, however, one reservation: for an aphasic
who has lost the capacity of code switching, his "idiolect" indeed
becomes the sole linguistic reality. As long as he does not regard
another's speech as a message addressed to him in his own verbal
pattern, he feels, as a patient of Hemphil and Stengel expressed
it: "I can hear you dead plain but I cannot get what you
say ... I hear your voice but not the words ... It does not pronounce
itself."^^ He considers the other's utterance to be either
gibberish or at least in an unknown language.
As noted above, it is the external relation of contiguity which
unites the constituents of a context, and the internal relation of
^* 'Results of the Conference of Anthropologists and Linguists,' Indiana
University Publications in Anthropology and Linguistics, VIII (1953), p. 15.
" R. E. Hemphil and E. Stengel, 'Pure word deafness,' Journal ofNeurology
and Psychiatry, III (1940), pp. 251-62.
SIMILARITY DISORDER 69
similarity which underlies the substitution set. Hence for an aphasic
* '
with impaired substitution and intact contexture, operations
involving similarity yield to those based on contiguity. It could be
predicted that under these conditions any semantic grouping would
be guided by spatial or temporal contiguity rather than by similarity.
'
Actually Goldstein's tests justify such an expectation: a female
patient of this type, when asked to list a few names of animals,
'
disposed them in the same sequence in which she had seen them
'
in the zoo; similarly, despite instructions to arrange certain objects
according to color, size and shape, she classified them on the basis
!
of their spatial contiguity as home things, office materials, etc. and
! justified this grouping by a reference to a show window where
"it does not matter what the things are," i.e. they do not have to
i be similar (pp. 61f., 263ff.). The same patient was wiUing to name
I
the primary hues - red, blue, green and yellow - but declined to
\i
extend these names to the transitional varieties (p. 268f.), since, for
I
her, words had no capacity to assume additional, shifted meanings
'
associated by similarity with their primary meaning.
One must agree with Goldsiein's observation that patients of
I
this type "grasped the words in their literal meaning but could not
I
be brought to understand the metaphoric character of the same
words" (p. 270). It would, however, be an unwarranted generaliza-
I
tion to assume that figurative speech is altogether incomprehensible
'
to them. From the two polar figures of speech, riietaphor and
metonymy, the latter, based on contiguity, is widely employed by
. aphasics whose selective capacities have been affected. Fork is
substituted for knife, table for lamp, smoke for pipe, eat for toaster.
A typical case is reported by Head
When he failed to recall the name for "black", he described it as "What
you do for the dead"; this he shortened to "dead" (I, p. 198).
Such metonymies may be characterized as projections from the
line of a habitual context into the line of substitution and selection:
a sign (e.g. fork) which usually occurs together with another sign
(e.g. knife) may be used instead of this sign. Phrases like "knife
and fork", "table lamp", "to smoke a pipe", induced the metonymies
70 TWO ASPECTS OF LANGUAGE
fork, table, smoke; the relation between the use of an object
(toast) and the means of its production underhes the metonymy
eat for toaster. "When does one wear black?" - "When mourning
for the dead" : in place of naming the color, the cause of its traditional
use is designated. The escape from sameness to contiguity is
particularly striking in such cases as Goldstein's patient who
would answer with a metonymy when asked to repeat a given word
and, for instance, would say glass for window and heaven for God
(p. 280).
When the selective capacity is strongly impaired and the gift
for combination at least partly preserved, then contiguity
determines the patients' whole verbal behavior, and we may
designate this type of aphasia similarity disorder.
I
IV
CONTIGUITY DISORDER
From 1864 on it was repeatedly pointed out in Hughlings Jackson's
pioneer contributions to the modern study of language and language
i
disturbances:
(
j
It is not enough to say that speech consists of words. It consists of
i words referring to one another in a particular manner; and, without
I
a proper interrelation of its parts, a verbal utterance would be a mere
I
succession of names embodying no proposition (p. 66).^"
[
Loss of speech is the loss of power to propositionize. . . Speechlessness
1. does not mean entire wordlessness (p. 1 14). ^^
The impairment of the ability to propositionize, or generally
speaking, to combine simpler linguistic entities into more complex
units, is actually confined to one type of aphasia, the opposite of
the type discussed in the preceding chapter. There is no wordlessness,
since the entity preserved in most of such cases is the
word, which can be defined as the highest among the linguistic
units compulsorily coded, i.e., we compose our own sentences and
utterances out of the word stock supplied by the code.
This contexture-deficient aphasia, which could be termed
contiguity disorder, diminishes the extent and variety of
sentences. The syntactical rules organizing words into a higher
unit are lost; this loss, called agrammatism, causes the degeneration
of the sentence into a mere "word heap", to use Jackson's
image.'^^
Word order becomes chaotic; the ties of grammatical
^" H. Jackson, 'Notes on the physiology and pathology of the nervous
system' (1868), Brain, XXXVIII (1915), pp. 65-71.
^' H. Jackson, 'On affections of speech from disease of the brain' (1879),
Brain, XXXVIII (1915), pp. 107-29.
^"^
H. Jackson, 'Notes on the physiology and pathology of language'
(1866), Brain, XXXVIII (1915), pp. 48-58.
72 TWO ASPECTS OF LANGUAGE
coordination and subordination, whether concord or government,
are dissolved. As might be expected, words endowed with purely
grammatical functions, like conjunctions, prepositions, pronouns,
and articles, disappear first, giving rise to the so-called "telegraphic
style", whereas in the case of similarity disorder they are the most
resistant. The less a word depends grammatically on the context,
the stronger is its tenacity in the speech of aphasics with a contiguity
disorder and the sooner is it dropped by patients with a similarity
disorder. Thus the "kernel subject word" is the first to fall out of
the sentence in cases of similarity disorder and, conversely, it is the
least destructible in the opposite type of aphasia.
The aphasia affecting contexture tends to infantile one-sentence
utterances and one-word sentences. Only a few longer, stereotyped,
"ready made" sentences manage to survive. In advanced cases of
this disease, each utterance is reduced to a single one-word sentence.
While contexture disintegrates, the selective operation goes
on. "To say what a thing is, is to say what it is like," Jackson
notes (p. 125). The patient confined to the substitution set (once
contexture is deficient) deals with similarities, and his approximate
identifications are of a metaphoric nature, contrary to the metonymic
ones familiar to the opposite type of aphasics. Spyglass for
microscope, or fire for gaslight are typical examples of such quasimetaphoric
expressions, as Jackson christened them, since, in
contradistinction to rhetoric or poetic metaphors, they present no
deliberate transfer of meaning.
In a normal language pattern, the word is at the same time both
a constituent part of a superimposed context, the sentence, and
itself a context superimposed on ever smaller constituents,
morphemes (minimum units endowed with meaning) and
phonemes. We have discussed the effect of contiguity disorder
on the combination of words into higher units. The relationship
between the word and its constituents reflects the same impairment,
yet in a somewhat different way. A typical feature of agrammatism
is the abolition of inflection: there appear such unmarked
categories as the infinitive in place of diverse finite verbal forms,
and in languages with declension, the nominative instead of all
CONTIGUITY DISORDER 73
I the oblique cases. These defects are due partly to the elimination
of government and concord, partly to the loss of ability to dissolve
words into stem and desinence. Finally, a paradigm (in particular
a set of grammatical cases such as he - his - hun, or of tenses such
as he votes - he voted) presents the same semantic content from
different viewpoints associated with each other by contiguity; so
there is one more impetus for aphasics with a contiguity disorder
to dismiss such sets.
Also, as a rule, words derived from the same root, such as grant grantor
- grantee are semantically related by contiguity. The
patients under discussion are eitlier inclined to drop the derivative
words, or the combination of a root with a derivational suffix and
even a compound of two words become irresolvable for them.
Patients who understood and uttered such compounds as Thanksgiving
or Battersea, but were unable to grasp or to say thanks and
giving or batter and sea, have often been cited. As long as ti':3
sense of derivation is still alive, so that this process is still used for
creating innovations in the code, there can be observed a tendency
toward oversimplification and automatism: if the derivative word
constitutes a semantic unit which cannot be entirely inferred from
the meaning of its components, the Gestalt is misunderstood.
Thus the Russian word mokr-ica signifies 'wood-louse', but a
Russian aphasic interpreted it as 'something humid', especially
'humid weather', since the root mokr- means 'humid' and the suffix
-ica designates a carrier of the given property, as in nelepica
'something absurd', svetlica 'light room', temnica 'dungeon'
(literally 'dark room').
When, before the Word War II, phonemics was the most
controversial area in the science of language, doubts were expressed
by some linguists as to whether phonemes really play an autonomous
part in our verbal behavior. It was even suggested that
the meaningful (significative) units of the hnguistic code, such
as morphemes or sooner words, are the minimal entities with
which we actually deal in a speech event, whereas the merely
distinctive units, such as phonemes, are an artificial construct
to facilitate the scientific description and analysis of a language.
74 TWO ASPECTS OF LANGUAGE
This view, which was stigmatized by Sapir as "the reverse of
reahstic",-^ remains, however, perfectly valid with respect to a
certain pathological type: in one variety of aphasia which sometimes
was labeled ''atactic'*, the v»ord is the sole linguistic unity
preserved. The patient has only an integral, indissolvablc image of
any famihar word, and all other sound-sequences are either alien
and inscrutable to him or he merges them into familiar words by
disregarding their phonetic aberrations. One of Goldstein's
patients "perceived some words, but . . . the vowels and consonants
of which they consisted were not perceived" (p. 218). A French
aphasic recognized, understood, repeated, and spontaneously
produced the word cate 'coffee' or pave 'roadway', but was unable
to grasp, discern, or repeat such nonsensical sequences as feca,
fake, kefa, pafe. None of these difficulties exists for a normal
French-speaking hstener as long as the sound-sequences and their
components fit the French phonemic pattern. Such a listener may
even apprehend these sequences as words unknown to him but
plausibly belonging to the French vocabulary and presumably
different in meaning, since they differ from each other either in the
order of their phonemes or in the phonemes themselves.
If an aphasic becomes unable to resolve the word into its
phonemic constituents, his control over its construction weakens,
and perceptible damages in phonemes and their combinations easily
follow. The gradual regression of the sound pattern in aphasics
regularly reverses the order of children's phonemic acquisitions.
This regression involves an inflation of homonyms and a decrease
of vocabulary. If this twofold - phonemic and lexical - disablement
progresses further, the last residues of speech are one-phoneme
- one-word - one-sentence utterances: the patient relapses into the
initial phases of infants' linguistic development or even to their
pre-lingual stage : he faces aphasia universalis, the total loss of the
power to use or apprehend speech.
The separateness of the two functions - one distinctive and the
other significative - is a peculiar feature of language as compared
^^ E. Sapir, 'The psychological reality of phonemes,' Selected writings
(Berkeley and Los Angeles, 1949), p. 46ff.
CONTIGUITY DISORDER 75
to Other semiotic systems. There arises a conflict between these
two levels of language when the aphasic deficient in contexture
exhibits a tendency to abolish the hierarchy of linguistic units and
to reduce their scale to a single level. The last level to remain is
either a class of significative values, the word, as in the cases
touched upon, or a class of distinctive values, the phoneme. In
the latter case the patient is still able to identify, distinguish and
reproduce phonemes, but loses the capacity to do the same with
words. In an intermediate case, words are identified, distinguished
and reproduced; according to Goldstein's acute formulation, they
"may be grasped as known but not understood" (p. 90). Here
the word loses its normal significative function and assumes the
purely distinctive function which normally pertains to the phoneme.
THE METAPHORIC AND METONYMIC POLES
The varieties of aphasia are numerous and diverse, but all of them
oscillate between the two polar types just described. Every form
of aphasic disturbance consists in some impairment, more or less
severe, either of the faculty for selection and substitution or for
combination and contexture. The former affliction involves a
deterioration of metalinguistic operations, while the latter damages
tl-.e capacity for maintaining the hierarchy of linguistic units. The
relation of similarity is suppressed in the former, the relation of
contiguity in the latter type of apliasia. Metaphor is alien to the
similarity disorder, and metonymy to the contiguity disorder.
The de\'elopmenl of a discourse may take place along two
d ifferent semantic hnes : one topic may lead to another either through
their similarity or through iheir contiguity. The metaphoric
way would be the most appropriate term for the first case and tiie
metonymic way for the second, since they find their most
condensed expression in metaphor and metonymy respectively. In
aphasia one or the other of these two processes is restricted or
totally blocked - an effect which makes the study of aphasia
particularly illuminating for the linguist. In normal verbal behavior
both processes are continually operative, but careful observation
will reveal that under the influence of a cultural pattern, personality,
and verbal style, preference is given to one of the two processes
over the other.
In a well-known psychological test, children are confronted with
some noun and told to utter the first verbal response that comes
into their heads. In this experiment two opposite linguistic predilections
are invariably exhibited: the response is intended either
as a substitute for, or as a complement to the stimulus. In the
latter case the stimulus and the response together form a proper
THE METAPHORIC AND METONYMIC POLES 77
syntactic construction, most usually a sentence. These two types
of reaction have been labeled substitutive and predicative.
To the stimulus hut one response was burnt out; another, is a
poor little house. Both reactions are predicative; but the first
creates a purely narrative context, while in the second there is a
double connection with the subject hut : on the one hand, a positional
(namely, syntactic) contiguity, and on the other a semantic
similarity.
The same stimulus produced the following substitutive reactions:
the tautology hut; the synonyms cabin and hovel; the antonym
palace, and the metaphors den and burrow. The capacity of two
words to replace one another is an instance of positional similarity,
and, in addition, all these responses are linked to the stimulus by
semantic similarity (or contrast). Metonymical responses to the
same stimulus, such as thatch, litter, or poverty, combine and
contrast the positional similarity with semantic contiguity.
In manipulating these two kinds of connection (similarity and
contiguity) in both their aspects (positional and semantic) selecting,
combining, and ranking them - an individual exhibits
his personal style, his verbal predilections and preferences.
In verbal art the interaction of these two elements is especially
pronounced. Rich material for the study of this relationship
is to be found in verse patterns which require a compulsory
parallelism between adjacent lines, for example in Biblical poetry
or in the West Finnic and, to some extent, the Russian oral traditions.
This provides an objective criterion of what in the given speech
community acts as a correspondence. Since on any verbal level morphemic,
lexical, syntactic, and phraseological - either of these
two relations (similarity and contiguity) can appear - and each in
either of two aspects -, an impressive range of possible configurations
is created. Either of the two gravitational poles may prevail.
In Russian lyrical songs, for example, metaphoric constructions
predominate, while in the heroic epics the metonymic way is
preponderant.
In poetry there are various motives which determine the choice
between these alternants. The primacy of the metaphoric process
78 TWO ASPECTS OF LANGUAGE
in the literary schools of romanticism and symbolism has been
repeatedly acknowledged, but it is still insufficiently realized that it
is the predominance of metonymy which underlies and actually
predetermines the so-called "reaHstic" trend, which belongs to an
intermediary stage between the decline of romanticism and the
rise of symbolism and is opposed to both. Following the path of
contiguous relationships, the reahstic author metonymically
digresses from the plot to the atmosphere and from the characters
to the setting in space and time. He is fond of synecdochic details.
In the scene of Anna Karenina's suicide Tolstoj's artistic attention
is focused on the heroine's handbag; and in War and Peace the
synecdoches "hair on the upper lip" or "bare shoulders" are
used by the same writer to stand for the female characters to whom
these features belong.
The alternative predominance of one or the other of these two
processes is by no means confined to verbal art. The same oscillation
occurs in sign systems other than language.^* A salient example
from the history of painting is the manifestly metonymical orientation
of cubism, where the object is transformed into a set of
synecdoches; the surrealist painters responded with a patently
metaphorical attitude. Ever since the productions of D. W.
Griffith, the art of the cinema, with its highly developed capacity
for changing the angle, perspective and focus of "shots", has
broken with the tradition of the theater and ranged an unprecedented
variety of synecdochic "close-ups" and metonymic "set-ups"
in general. In such pictures as those of Charlie Chaplin, these
devices in turn were superseded by a novel, metaphoric "montage"
with its "lap dissolves" - the filmic similes. ^^
The bipolar structure of language (or other semiotic systems),
and, in aphasia, the fixation on one of these poles to the exclusion
** I ventured a few sketchy remarks on the metonymical turn in verbal
art ('Pro realizm u mystectvi,' Vaplite, Kharkov, 1927, No. 2; 'Randbemerkungen
zur Prosa des Dichters Pasternak,' Slavische Rundschau, VII, 1935),
in painting ('Futurizm,' Iskusstvo, Moscow, Aug. 2, 1919) and in motion
pictures ('Upadek filmu,' Listy pro umeni a kritiku, I, Prague, 1933), but the
crucial problem of the two polar processes awaits a detailed investigation.
" Cf. B. Balazs, Theory of the film (London, 1952).
THE METAPHORIC AND METONYMIC POLES 79
of the other require systematic comparative study. The retention
of either of these alternatives in the two types of aphasia must be
confronted with the predominance of the same pole in certain
styles, personal habits, current fashions, etc. A careful analysis
and comparison of these phenomena with the whole syndrome of
the corresponding type of aphasia is an imperative task for joint
research by experts in psychopathology, psychology, linguistics,
poetics, and s e mi o t i c, the general science of signs. The dichotomy
here discussed appears to be of primal significance and consequence
for all verbal behavior and for human behavior in general, ^^
To indicate the possibilities of the projected comparative research,
we choose an example from a Russian folktale which
employs parallelism as a comic device: "Thomas is a bachelor;
Jeremiah is unmarried" (Fomd xolost; Erjoma nezendt). Here
the predicates in the two parallel clauses are associated by
similarity: they are in fact synonymous. The subjects of
both clauses are masculine proper names and hence morphologically
similar, while on the other hand they denote two contiguous
heroes of the same tale, created to perform identical
actions and thus to justify the use of synonymous pairs of predicates.
A somewhat modified version of the same construction occurs in
a famihar wedding song in which each of the wedding guests is
addressed in turn by his first name and patronymic: "Gleb is a
bachelor; Ivanovid is unmarried." While both predicates here
are again synonyms, the relationship between the two subjects is
changed: both are proper names denoting the same man and are
normally used contiguously as a mode of polite address.
In the quotation from the folk tale the two parallel clauses refer
to two separate facts, the marital status of Thomas and the similar
status of Jeremiah. In the verse from the wedding song, however,
the two clauses are synonymous: they redundantly reiterate the
2' For the psychological and sociological aspects of this dichotomy see
Bateson's views on "progressional" and "selective integration" and Parsons'
on the "conjunction-disjunction dichotomy" in children's development:
J. Ruesch and G. Bateson, Communication, the social matrix of psychiatry
(New York, 1951), pp. 183ff.; T. Parsons and R. F. Bales, Family, socialization
and interaction process (Glencoe, 1955), pp. 119f.
80 TWO ASPECTS OF LANGUAGE
celibacy of the same hero, splitting him into two verbal hypostases.
The Russian novelist Gleb Ivanovic Uspenskij (1840-1902) in
the last years of his life suffered from a mental illness involving a
speech disorder. His first name and patronymic, Gleb Ivanovic,
traditionally combined in polite intercourse, for him split into two
distinct names designating two separate beings : Gleb was endowed
with all his virtues, while Ivanovic, the name relating the son to
the father, became the incarnation of all Uspenskij 's vices. The
linguistic aspect of this spUt personaUty is the patient's inabihty to
use two symbols for the same thing, and it is thus a similarity
disorder. Since the similarity disorder is bound up with the
metonymical bent, an examination of the hterary manner
Uspenskij had employed as a young writer takes on particular
interest. And the study of Anatolij Kamegulov, who analyzed
Uspenskij's style, bears out our theoretical expectations. He
shows that Uspenskij had a particular penchant for metonymy, and
especially for synecdoche, and that he carried it so far that "the
reader is crushed by the multiplicity of detail unloaded on him in
a hmited verbal space, and is physically unable to grasp the whole,
so that the portrait is often lost."^^
To be sure, the metonymical style in Uspenskij is obviously
prompted by the prevailing Hterary canon of his time, late nineteenth-century
"realism" ; but the personal stamp of Gleb Ivanovic
made his pen particularly suitable for this artistic trend in its
extreme manifestations and finally left its mark upon the verbal
aspect of his mental illness.
A competition between both devices, metonymic and metaphoric,
is manifest in any symbolic process, either intrapersonal or sociaL
^' A. Kamegulov, StW Gleba Uspenskogo (Leningrad, 1930), pp. 65, 145.
One of such disintegrated portraits cited by the monograph: "From underneath
an ancient straw cap with a black spot on its shield, there peeked two
braids resembling the tusks of a wild boar; a chin grown fat and pendulous
definitively spread over the greasy collars of the calico dicky and in thick layer
lay on the coarse collar of the canvas coat, firmly buttoned on the neck. From
below this coat to the eyes of the observer there protruded massive hands with
a ring, which had eaten into the fat finger, a cane with a copper top, a significant
bulge of the stomach and the presence of very broad pants, almost of muslin
quality, in the broad ends of which hid the toes of the boots."
THE METAPHORIC AND METONYMFC POLES 81
Thus in an inquiry into the structure of dreams, the decisive question
j
is whether the symbols and the temporal sequences used are based
on contiguity (Freud's metonymic "displacement" and synecdochic
j
"condensation") or on similarity (Freud's "identification and
I
symbolism").-^ The principles underlying magic rites have been
i resolved by Frazer into two types : charms based on the law of
j
similarity and those founded on association by contiguity. The
! first of these two great branches of sympathetic magic has been
'
called "homoeopathic" or "imitative", and the second, "contagious
magic". ^® This bipartition is indeed illuminating. Nonetheless,
for the most part, the question of the two poles is still neglected,
despite its wide scope and importance for the study of any symbolic
behavior, especially verbal, and of its impairments. What is the
main reason for this neglect?
Similarity in meaning connects the symbols of a metalanguage
with the symbols of the language referred to. Similarity connects
a metaphorical term with the term for which it is substituted.
Consequently, when constructing a metalanguage to interpret
tropes, the researcher possesses more homogeneous means to
handle metaphor, whereas metonymy, based on a different principle,
easily defies interpretation. Therefore nothing comparable
to the rich literature on metaphor^^ can be cited for the theory of
metonymy. For the same reason, it is generally realized that
romanticism is closely linked with metaphor, whereas the equally
intimate ties of realism with metonymy usually remain unnoticed.
Not only the tool of the observer but also the object of observation
is responsible for the preponderance of metaphor over metonymy
in scholarship. Since poetry is focused upon sign, and pragmatical
prose primarily upon referent, tropes and figures were studied
mainly as poetical devices. The principle of similarity underlies
poetry; the metrical parallelism of lines or the phonic equivalence
of rhyming words prompts the question of semantic similarity and
^''
S. Freud, Die Traumdeutung, 9th ed. (Vienna, 1950).
"' J. G. Frazer, The golden bough: A study in magic and rehgion. Part I,
3rd ed. (Vienna, 1950), chapter III.
*" C. F. P. Stutterheim, Het begrip metaphoor (Amsterdam, 1941).
82 TWO ASPECTS OF LANGUAGE
contrast; there exist, for instance, grammatical and anti-grammatical
but never agrammatical rhymes. Prose, on the contrary, is
forwarded essentially by contiguity. Thus, for poetry, metaphor,
and for prose, metonymy is the line of least resistance and,
consequently, the study of poetical tropes is directed chiefly toward
metaphor. The actual bipolarity has been artificially replaced in
these studies by an amputated, unipolar scheme which, strikingly
enough, coincides with one of the two aphasic patterns, namely
with the contiguity disorder.^^
I
**
Thanks are due to Hugh McLean for his valuable assistance and to
Justinia Besharov for her original observations on tropes and figures.
SUPPLEMENT
1
j
SELECTED LIST OF STUDIES IN GENERAL PHONOLOGY
(1931-1955)
i
•
M. J. Andrade, 'Some questions offact and policy concerning phonemes,'
Language, XII (1936).
O. AxMANOVA, Fonologija (Moscow, 1954).
C. E. Bazell, 'The choice of criteria in structural linguistics,' Word, X
(1954).
B. Bloch, 'A set of postulates for phonemic analysis,' Language, XXV
(1948).
L. Bloomfield, Language (New York, 1933). Chapters V-VIII.
E. BuYSSENS, 'Mise au point de quelques notions fondamentales de la
phonologic,' Cahiers Ferdinand de Saussure, VIII (1949).
J. Camara, Principios de Linguistica Geral (Rio de Janeiro, 1954).
Chapters II-III.
Y. R. Chao, 'The non-uniqueness of phonemic solution of phonetic
systems,' Academica Sinica, Institute of History and Philology,
Bulletin, IV (Shanghai, 1934).
E. C. Cherry, M. Halle, R. Jakobson, 'Toward the logical description
of languages in their phonemic aspect,' Language, XXIX (1953).
E. CosHRiu, W. Vasques, Para la unificacion de las sciencias fonicas
(Montevideo, 1953).
E. DiETH, Vademecuni der Phonetik (Bern, 1950). Chapter III C.
B. Faddegon, 'Phonetics and phonology, 'Meded. Kan. Nederl. Akad.
Wetensch., Afd. Letterkunde, II (1938).
E. Fischer-Jorgensen, 'Phonologic,' Archiv fiir vergleichende Phonetik,
V (1941); 'On the definition of phoneme categories on a distributional
basis,' Acta Linguistica, VII (1952).
H. Frei, 'Langue, parole et differenciation,' Journal de Psychologie (1952).
C. C. Fries, K. L. Pike, 'Coexistent phonemic systems,' Language, XXV
(1949).
A. W. de Groot, 'Neutralisation d'oppositions,' Neophilologus, XXV
(1940).
M. Halle, 'Tiie strategy of phonemics,' Word, X (1954).
Z. S. Harris, Methods in structural linguistics (Chicago, 1951); 'From
phoneme to morpheme,' Language, XXXI (1955).
L. Hjelmslev, 'Ober die Beziehungen der Phonetik zur Sprachwissenschaft,'
Archiv fiir vergleichende Phonetik, II (1938).
C. F. HocKETT, 'A manual of phonology' = Indiana University Publications
in Anthropology and Linguistics, XI (1955).
R. Jakobson, 'The phonemic and grammatical aspects of language in
86 SUPPLEMENT
their interrelations,' Acres du Sixieme Cungres International des
Linguistes (Paris, 1949).
R. Jakobson, C. G. M. Fant, M. Halle, Preliminaries to speech analysis,
third printing (Massachusetts Institute of Technology, Acoustics
Laboratory, 1955).
D. Jones, The phoneme: its nature and use (Cambridge, 1950).
A. G. JuiLLAND, 'A bibliography of diachronic phonemics,' Word, IX
(1953), pp. 198-208.
J. M. KorInek, Vvod do jazykospytu (Bratislava, 1948). Chapter II.
E. Kruisinga, 'Fonetiek en fonologie,' Taal en Leven, VI (1943).
J. V. Laziczius, 'Probleme der Phonologie,' Ungarische Jahrbiicher,
XV (1935).
E. A. Llorach, Fonologia Espanola (Madrid, 1954). Primera parte:
fonologia general.
A. Martinet, Phonology as functional phonetics (London. 1949); 'Oil en
est la phonologic?' Lingua, I (1949).
K. L. Pike, Phonemics: a technique for reducing languages to writing
(Ann Arbor, 1947); Tone languages (Ann Arbor, 1948); 'Grammatical
prerequisites to phonemic analysis,' IVord, III (1947); 'More on
grammatical prerequisites," Word, Will (1952); Language in relation to
a unified theory of the structure of human behavior, II (Glendale. Cal.,
1955).
M. PoLAK, 'Fonetiek en fonologie,' Levende Talen (1940)
H. J. Pos, ''PhonoXogiQ.QnbQXtkQmsAeQx'' = Mcdedeelingen der Koninklijke
Nederlandsche Akademie van Wetenschappen. Afd. Letterkunde. NR,
No. 13 (1938).
E. Sapir, Selected writings (Berkeley and Los Angeles, 1949), pp. 7-60.
E. Seidel, Das Wesen der Phonologie (Bucharest-Copenhagen, 1943).
A. SoTAVALTA, 'Die Phonetik und ihre Beziehungen zu den Grenzwissenschaften'
= Annales Academiae Scientiarum Fennicae, XXXI,
No. 3 (1936).
R. H. Stetson, Motor phonetics (Amsterdam, 1951).
M. SwADESH. 'The phonemic principle,' Language, X (1934).
B.TRNKA,'Urcovanifonem.u,'/1c/a Universitatis Carolinae (Prague, 1954).
N. Trubetzkoy, Principes de phonologie (Paris, 1949). German text:
'Grundzuge der Phonologic' = Travaux du Cercle Linguistique de
Prague, VII (1939).
W. F. TwADDELL, 'On defining the phoneme' = Supplement to Language,
XVI (1935); 'Stetson's model and the <supra-segmental phonemes>,'
Language, XXIX (1935).
N. VAN WiJK, Phonologic: een hoofdstuk uit de structurele taalwetenschap
(The Hague, 1939). A substantially revised version is-being prepared
by A. Reichling.
SUPPLEMENT 87
[E. Zwirner, 'L'opposition phonologique et la variation des phonemes,'
Archivfiir vergleichende Phonetik, II (1938) ; 'Phonologic und Phonetik,'
Acta Linguistica, I (1939).
DISCUSSIONS
Akademija Nauk SSSR, Otdelenie literatury i jazyka, Izvestija, XI (1952)
and XII (1953) - Diskussija po voprosam fonologii.
International Congresses of Phonetic Sciences, Proceedings, I-III (1933,
1935, 1938).
Journal of the Acoustical Society of America, XXII (1950) - Proceedings
of the Speech Communication Conference at MIT; XXIV (1952) Conference
on Speech Analysis.
Travaux du Cercle Linguistique de Copenhague, V (1949) - Recherches
structurales.
Travaux du Cercle Linguistique de Prague, IV (1931) - Reunion phonologique
Internationale tenue a Prague; VIII (1939) - Etudes phonologiques.
dediees a la memoire de N. S. Trubetzkoy.
"u€?ff«.
i\
i
RddM^SL
U.l University of California Library
Los Angeles
This book is DUE on the last date stamped below.
JAN
Phone Renemfafs
310/82^^
mC'D LD-URL
4I^S38«
MAR 7 1998
JSN22'* 4 WK APR 5 19^7
tu/HSY(
QL JUN
REC
APK2
tu/pj
w mf ^ 1 2001
n-" 71998
,;J
l^iwv
IPNl
UBJ
w.
19^