CTLF Corpus de textes linguistiques fondamentaux • IMPRIMER • RETOUR ÉCRAN
CTLF - Menu général - Textes

Stetson, Raymond. Motor Phonetics – T03

Appendix VI
Reduction of a language to phonemes

If a language is literate, the orthography is bound to
figure in the process of developing the scientific
notation of the syllables. The orthography may be
approximately phonetic, German, Italian, or it may
be traditional and at variance with the present
phonemes and syllables (English and French); or it
may have been adapted from another language, with
fairly adequate modification in case of the Russian
orthography, or rather poor results in case of the
Arabic characters used for Turkish and Malay.

If the language is non-literate, the natives know
word plays, are aware of assonances, alliterations,
and may use rime. If the verse of the language has
a regular meter, this will help to determine the syllable
and the permissible stress system.

For this essential phonetic analysis, Sweet, Saussure,
Passy, Trubetzkoy and the phonemicists, have
depended on a traditional phonetics often inexact,
employing now articulatory and now auditory criteria.
Trubetzkoy (100, p. 17), “must be physiological”;
Martinet (101, p. 273 f), “must have concrete
individuality”; Pos (101, p. 78), Trager and
Bloch propose a careful study of experimental data
concerning segmental, prosodic, and rhythmic features,
in the course of a discussion of “open juncture”.

There is some general recognition of the physiological
articulation as the fundamental event, but
most linguists depend on what they assume to be an
auditory impression. Experienced observation has
been the principal resource; sometimes the investigator
tries repeating the “sound” if it is in his own
language, or tries to imitate the “sound” if it is
foreign, to determine its characteristics, perhaps
with the help of a “small mirror”. Because, of the
practical difficulties there has been very little use
made of such exact experimental methods as are
now available.

An intelligent native of the non-literate language,
with some practice in writing a culture language, is
likely to do best in noting the essential articulations.
In any case, only the phonetic distinctions which are
significant for the meanings of the given language
are to be noted.

The International Phonetic Alphabet as it stands
is not available for the notation of a given language.
Of necessity it does not fit the specific phonemic
system of any language. Investigators modify the
I.P.A. radically, introduce new characters, modify
the stock characters, and set up various conventions
as to breath-group, stress, and possibly ‘tones’.

The analysis is of course to be in terms of the
given language. For practical purposes ‘phonemes’
are always identified by reference to standard
words of the given language. Phonetic treatises use
a notation based on well-known languages; and the
standard word lists for the “sounds” of pronouncing
dictionaries identify the phonemes — at some
risk — in terms of familiar words. It is the one way
in experimental work to set patterns for a subject,
and to indicate the actual syllables used, for other
investigators. It is dependent on uniform habits of
pronunciation. In psychological terms it is dependent
on stability of conditioning.

The number of syllables in the given language
will vary from a minimum of 40 or 50 to as many
as 5000 or more, depending on the permissible syllable
patterns of the language. And variants of each
syllable which involve traits which may be significant
in the language in question must be noted.
Vowels may diphthongize in stressed syllables, or
reduce to shwa in unstressed syllables; an arresting
consonant may change, shift to the next syllable, or
drop; a releasing consonant may fuse with the preceding
arresting consonant, or shift to arresting
position in a preceding open syllable as its own
syllable drops. In the final form of the corpus of
the various syllables of the given language, there may
be several variants of each basic syllable which are
150the various phonetic forms of the syllable. The
basic syllables are determined by reference to meaning
in morphone stems, affixes and infixes. If these
syllable types are put on cards for machine sorting,
a card will be punched for each variant of every
syllable. (If new variants or new syllables are discovered
in the course of the analysis, they are added,
of course.)

Analysis of the corpus of syllables to determine
the complete list of the phonemes —the
phoneme alphabet — not as “elements,
segments” but as the characterized factors
of the syllables in the breath group

It is convenient to designate the phoneme by the
form in which it appears in slow, careful utterance,
and usually as it occurs in a single syllable in the
breath group. The use of the syllables as conditioned
signals with meaning has moulded the syllables
to be analyzed; but in the process of analysis they
are treated as nonsense syllables.

Phonetic and phonemic alphabets

Since the corpus of syllables is designated by phonetic
signs for the specific articulations (characterized
factors of the syllable) it would seem that the
task of isolating the “sounds” is already accomplished.
A phonetic alphabet has already been set up,
composed of such different articulations as are
significant (conditioned to meanings) in the given

The phonetic alphabet, however, merely indicates
the “sounds” as they occur in each syllable without
reference to their relations, while the phoneme
alphabet designates each type articulation and identifies
its variants through all the changes, compounds,
shifts, dropping and reappearances of the
type articulation. The phoneme, then, designates
the set of reversible forms which the type articulation
assumes in the varying conditions of utterance
of the syllable in the breath group. When the
variant of the one phoneme comes to coincide phonetically
with the variant of another phoneme, the
phoneme sign designates which phoneme is actually
represented by the ambiguous articulation. In such
cases a change of rate and stress will cause the
ambiguous articulation to revert to the type form
of the phoneme. It is easy to apply the test: if the
utterance keeps the same linguistic (“intellectual”)
meaning through the changes caused by rate and
stress, the phoneme remains the same.
It is worth emphasizing how wide may be the
“phonic difference” in certain conditions without
changing the linguistic meaning of the utterance
in which it occurs. The minimum “phonic difference”
to produce a change in linguistic meaning
is not a fixed difference, but varies radically with
the conditions of utterance. In one situation the
difference of voicing of a consonant (aspect of a
syllable factor) may change the meaning; in another
situation the change of voicing, or the dropping
of the consonant articulation, or even of the entire
syllable, may not change the meaning. The characteristics
of the phoneme change with the changing
situation (in which rate and stress are primary).

Analysis of the relations of the phonemes in
this list drawn from the corpus of syllables;
classifications, differentiations, various
relations and groupings

The syllable is not a self-subsistent unit; it must be
part of a foot and a breath group. The types of
breath group and the position of the syllable in the
breath group have a profound influence on the
phoneme. Hence, for the phonemes of any language,

1) the types of breath group movement incorporating
the syllables, and

2) the types of syllable movement (pulses) permissible
in the given language, are of primary importance.

Certain traits of syllables and of breath groups
are correlated. The breath group with light and
heavy stresses and with varied feet, is the basic
movement for a series of syllables with frequent
arresting consonants, which drop, abut, and shift
with the reducing of vowels and the dropping of
syllables; such syllables are a part of such breath
groups. Cf. English and German.151

The breath group with light and rather variable
“word” stress and with little variation in the
grouping of the syllables into feet, is the basic
movement for a series of syllables with few arresting
consonants, slight differences in the syllable
duration, relatively little variation in the quality of
the individual vowel, and few or no repressed
syllables. Such a series is essential to the dynamic
structure of such a breath group. Cf. French.

Phonemes from language to language

1. Since the beginning of comparative linguistics
it has been assumed that “sounds”, later called
“phonemes”, are equivalent or identical from
language to language. The types of breath group
and the types of permissible syllables will seldom
be the same in two languages. It can hardly happen,
however, that the two languages have no syllable
factors in common (enphonic phonemes). A number
of syllable factors will be somewhat alike. Nevertheless,
the syllable types and the breath-group types
of the respective languages make the like syllable
factors unlike members of a phonetic system. If
the essence of the phoneme is in its specific unlikeness
to the other phonemes of the system peculiar
to the one language, the handling of the ‘phoneme’
common to two or more phonemic systems, each
in a different language, needs to be considered.
Saussure insists: “Ce qui importe dans le mot, ce
n'est pas le son luimême, mais les différences phoniques
qui permettent de distinguer le mot de tout
autres.” (75, p. 163)

“The phonic difference capable of differentiating
intellectual meanings in a given language” (98; 99,
p. 232) cannot be applied when there are no common
meanings. Apart from meanings and in terms
of nonsense syllables, it is certain that the phonemes
of one language, i.e. the characterized syllable factors,
cannot be equivalent to the phonemes of another
language however closely alike the articulations
may be. The English “do”, and the German
“du”, and the French “doux”, are all represented
by the I.P.A. /du/. But the English series, “How de
ye do, how de' do, how d'-do”, shows possibilities
quite foreign to the French. The French “doux”
has a consonant form for the d- unlike the German
and English releasing consonant, due to the difference
in breath pressure. (89, p. 83; 61, p. 115)

The unit of utterance in which the phoneme of
two languages is handled cannot be a “free form”,
for it is seldom that the same ‘form’ will occur in
the two (or more) languages. The like syllables of
the two languages, independent of morphology and
of any form of meaning, must be the unit of utterance
for the enphonic phoneme. In some sense
such an enphonic phoneme is only. a potential
phoneme; the I.P.A. is a set of possible phonemes.
When incorporated into specific syllables in breath
groups, the enphonic phonemes become phonemes
in the specific system of a language.

2. The development of the I.P.A. and of various
notations for different language groups has been in
response of the need of enphonic units. The processes
of comparing and identifying the ‘sounds’
of the different living languages which have yielded
the enphonic articulations for languages, both ancient
and modern, has been carried on by bi-lingual
speakers (or multi-lingual speakers). In the case of
non-literate languages, members of the culture
language have learned the non-literate language,
often as children and without an accent; and members
of the non-literate people often acquire the
culture language in childhood. The practical business
of devising an alphabet to write and print the non-literate
language forces the consideration of the
‘sounds’ in terms of the stereotyped orthography
of the literate language.

Where two peoples of different culture languages
are in contact there must be bi-lingual speakers.
Common religious-, business-, government, craft-,
and art- terms are bound to interpenetrate and to
be notated in the characters of both languages.
Approximate and actual equivalents are found, and
the unique ‘sounds’ noted. The brogue of the
neighbors becomes familiar to each people, and
picks out the minor differences of the ‘sounds’, if
not the more fundamental differences of the breath-group
and syllable-organization. There is the
152famous bit of practical phonetics in the “shibboleth”
incident of the Old Testament when a phonetic
variant meant life or death (Judges 12 : 6).

The recovery of the approximate values of the
early characters is one of the amazing achievements.
The Sanskrit grammarians undertook phonetic
descriptions of the ‘sounds’; the Greeks made
a less successful analysis. Since the early descriptions
and classifications of the ‘sounds’ could not
be exact, the value of the ancient sign has ultimately
to be worked down to the living language.

The ancient physiological descriptions must be
verified by present-day manipulations of the speech
apparatus. The ancient character may be traced
down to living languages in various ways. There
are many lines of evidence: a single alphabet has
often been adapted to various unlike languages,
hence many writing systems stem from the primitive
Semitic alphabet. The cuneiform syllabary was
extensively employed for languages of all sorts. On
occasion words occur written in two or more notations.
Hurrian and Akkadian are written in derivatives
of the Sumerian syllabary, and also in the
Ras-Shamra alphabet. Greek words are written in
the Greek alphabet and in the Cypriote syllabary.

The various living Semitic languages furnish the
clues for the values of the early signs. The vocalizing
of the Late Egyptians has been established
in some degree by a study of the Old Coptic
‘sounds’ determined in turn by reference to the
phonological history of Coptic. The continuous
history of Greek as a living language down to the
present day — with all its recorded changes — and
the comparative study of the Romance languages,
have yielded a fair idea of the earlier values of the
Greek and Latin at various epochs. And the two
languages have furnished the alphabets for the
notation of the European languages and have been
adapted for the speech of many non-literate peoples.
The Slavic languages are written in some cases in
an alphabet derived from the Greek; and in some
cases in an alphabet derived from the Latin. Arabic
has been used for many unrelated tongues: Turkish,
Persian, Urdu, Bantu, Malay. In general, the notation
of a common culture language is likely to
spread through its entire territory and may persist
through all changes for centuries.

Words persist through the vicissitudes of phonological
and lexical changes, and furnish the continuous
chain by which phonological states are held
together and the changes can be traced and defined.
The method of comparison of modern forms in
different languages and dialects has been made to
yield remarkable results as to ancient values.

Relations among the phonemes of different

1. Affiliation: The history of a given language
and of its collaterals, all derived from a common
source, demands an identification of common
‘sounds’ and a study of their changes. This is the
basis of the historical phonology of the comparative

The relations of the phonemes and of the syllable
and breath-group units of languages which are actually
connected are due to conditioning to meanings
from generation to generation. The word or
locution proves to be more stable than either the
lexical meaning or the phonetic factors of the syllable.
The word or locution survives, though the
breath-group and syllable organization may change
radically — as in the case of French from Vulgar
Latin. This makes it possible to trace the change
or persistence of the phonemes. However marked
the transformation in the course of centuries, even
in times of rapid change the users of the language
in each generation have found the variants equivalent.

Where the derivative languages have been isolated,
their parallel histories may be very unlike. But
all these changes have the common chain of utterances
conditioned to meanings; the phonemes have
been modified as they passed down, but the locutions
have kept the continuity.

Such relation of phonemes have had a large
share of the attention of linguists and may be called

2. Association: Contemporary languages are
153to be compared and grouped according to their
phonogical relations, whatever their connections
may have been.

The relations of. phonemes from language to
language may be due to regional diffusion among
languages which have few or no other connections.
Regional diffusion or morphological and
lexical influences may or may not occur where
enphonic diffusion is striking, as in the case of
American Indian languages, or the occurrence of
the Hottentot (and Bushman) ‘clicks’ in the Bantu-
Zulu. The process of such peculiar transfers is
uncertain and may vary from case to case. Such
cross influences due to propinquity may be termed

3. Parellelism: Independent groups of languages
which can never have had common influences
nevertheless present striking likenesses. Cf. Speiser
(80, 81) and Z. S. Harris' comparison of American
Indian Yokuts and Semitic (23, p. 210). The possible
types of breath groups and of correlated syllable
organization are limited; the possible types of syllables
are limited; and the characterized syllable
factors which the race finds available are limited.
Therefore such (apparent) relation is due to parallel

The (special) mechanism of each language must
be a specialized version of the general phonology
which can be stated in physiological terms, independent
of meanings. It is not surprising that there
are striking chance resemblances. A general phonology
was the hope of Trubetzkoy. Such chance
likeness may be called parallel.

Groups of languages may be set up in reference
to the phonological types. The correlations between
types of breath groups and types of syllables, and
the influence on the characterized syllable factors
have been noted. German, English, Scandinavian
and Dutch languages belong to a group with heavy,
fixed stresses in the breath group and frequent
closed syllables and varied feet. French, Japanese,
and the South Sea languages tend to the open-syllable
type which is correlated with a lighter and
often variable stress in the breath group while the
duration of the syllable is less varied.

Although the morphological units have had
extensive consideration, we have had, as yet, rather
little study of the breath-group types, and of the
correlated syllable types. E. Hermann (30) made
some interesting observations:

1) He established the status of the syllable in the
history of Greek phonology.

2) He adopted the “sonority” concept, but his
careful findings are in line. His use of “sonority”
and of the “mora” — postulated unit of
length within the syllable — does not vitiate
his findings.

3) He recognized the simple or compound releasing
consonant in a train of syllables.

4) He recognized the abutting group of arresting
and releasing consonant as basic for the prosodic
“length by position” of Greek grammarians.

5) He recognized the simple and compound consonants,
and the abutting pair.

6) He recognized the singling of the double consonant.

7) He compared the Greek alphabetic notation
with that of the Cypriote syllabary which
confirmed his findings.

8) He followed the study down through the
Latin and the derivative Romance languages,
and noted the general tendency to make the
consonants releasing, which reaches its ultimate
form in modern French.

No study is available of the Greek Septuagint
form of the Hebrew proper names and technical
terms, nor of the Greek transliteration of cuneiform
words. There is a mass of evidence to be mustered
on the points which Hermann stresses.

Metrical formulations often illuminate the breath-group
organization. If the discussion of “juncture”
in the form of “Grenzsignale” had reached the stage
of considering the reciprocal relations of the types
of “juncture” and the corresponding modifications
of the syllable frontier, it would have opened the
problem of types of breath-group organization. But
the “juncture” method deals with the apparent
boundaries and not with the organized movements.154