Vous êtes sur la page 1sur 24

Linguistic Society of America

Saigon Phonemics
Author(s): Laurence C. Thompson
Source: Language, Vol. 35, No. 3 (Jul. - Sep., 1959), pp. 454-476
Published by: Linguistic Society of America
Stable URL: http://www.jstor.org/stable/411232 .
Accessed: 04/04/2014 08:52

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.

http://www.jstor.org

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS
LAURENCEC. THOMPSON
University of Washington
[The phonological analysis of Saigon Vietnamese presents problems of indeter-
minacy, multiple analyses, and asymmetry. The phonetic structure of the dialect
is described here in detail and certain of these problems are discussed. Special
attention is called to the general problem of asymmetry in phonological analysis,
and a principle involving it is proposed for general theory. A specific procedure is
outlined for applying the principle, and the application is illustrated by examples
from the description of Saigonese.]

Long the economic and cultural center of the southern plains of Viet Nam,
and more recently also the political capital, Saigon has become perhaps the
most cosmopolitan of all Vietnamese cities. Native sons and daughters, however,
continue to speak with the local 'accent', which appears to be typical of the entire
.area. It was this language that I subjected to detailed study during my residence
in Viet Nam (1951-3).1
H. Masp6ro, in the most authoritative of the efforts to classify the Vietnamese
dialects, has divided them into two main groups: Tonkinese-Cochinchinese and
Haut-Annam dialects.2 This means that the speech of the far north of the
country (Tonkinese) and that of the south (Cochinchinese) are more closely
related to one another than either is to the set of dialects spoken along the
narrow north-central coastal strip. Unfortunately, as Maspero points out, studies
of the Haut-Annam dialects and Cochinchinese speech have been scanty and
lacking in precision.3
1 J am grateful for the financial assistance which made this field work possible: to Yale
University and the American Council of Learned Societies for their support from July 1951
to June 1953;to the Ford Foundation for its support from July 1952to June 1953;and to the
Human Relations Area Files, Inc., for assisting with transportation. I am deeply indebted
to Franklin Edgerton, who originally encouraged me to undertake this study and supported
me wholeheartedly throughout; to Murray B. Emeneau, who furnished me with numerous
materials from his work on the northern dialects; and to Bernard Bloch, Kun Chang, Isidore
Dyen, Mary R. Haas, Fang-Kuei Li, and William A. Smalley, with whom I have discussed
various problems of the analysis.
A full grammar of the dialect was presented as my doctoral dissertation at Yale Univer-
sity in 1954.This described mainly the speech of Saigonese who had received fairly extensive
formal education in Vietnamese (the level of educational attainment in French is not a
useful criterion for these purposes), and outlined the phonetic variations of popular speech
and the conventions adopted by learned speakers. The present paper constitutes a consider-
able revision of the section on phonology in the larger work, so far as it pertains to educated
speech. The analysis presented here is based on the speech of Mr. Lam-quang-Hong, Miss
Le-thi-Bai, Miss Nguyen-thi-Cut, Dr. Nguyen-van-Hien, and Miss Truong-thi-Danh, all
of Saigon. I owe thanks especially to Miss Le-thi-Bai and Mr. Huynh Sanh Thong for help
with some of the examples. The dialect is henceforth referred to as Saigonese.
2 Etudes sur la phonetique historique de la langue annamite: Les initiales, Bulletin de
.l'EcoleFrangaise d'Extrdme-Orient(hereafter BEFEO) 12.1-3 (1912).
3 Even after nearly half a century, this is still true. For the Haut-Annam dialects, forms
which Maspero himself collected appear most reliable; the southern forms that he quotes
454

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 455

Investigation of the phonetic reality of both northern and southern speech


reveals patterns which present opportunities for varied interpretation. The
choice among alternate phonemic solutions is problematic in various ways:
the search for a sound basis for the choice leads at times to the consideration
of articulatory and acoustic questions for which I have found few definitive
answers; the principles of economy and symmetry in phonemic inventory and
patterning here seem in frequent conflict with faithful representation of the
phonetic reality. Vietnamese thus provides a further example for the principle
of multiple analyses enunciated over two decades ago by Y. R. Chao.4 I shall
first present a detailed analysis of the data according to one system; then examine
other possible systems that are suggested by the raw phonetic data; and finally
discuss the implications for general descriptive theory.5
Basic terms are for the most part used as defined by B. Bloch, and the ex-
position is modeled to some extent on his treatment of Japanese phonemics.6
In addition I have adopted certain terms from C. F. Hockett and some of the
conceptual framework of his presentation of phonemic systems.7 The unmodified
terms INITIALand FINALare used throughout to mean syllable-initial and syllable-
final.

1. GENERAL
In Saigonese a fraction of utterance which begins with an onset of stress and
ends immediately before the next onset of stress or before pause is a SYLLABLE.
No phone is construed to extend over a syllable boundary; an intersyllabic long;
consonant or vowel is taken to be two separate phones, the second beginning with
the new onset of stress.
Syllables are segmented into phones, which are identified by their auditory
qualities (described below in articulatory terms). One general aspect of allegro

seem at some variance with those in my material. He also refers to P. Cadiere, Phonetique
annamite, dialecte du Haut-Annam (Paris, 1902); and Le dialecte du Bas-Annam, BEFEO
11.67-100 (1911). He cites the latter as evidence that Bas-Annam speech (spoken in the
southern part of the central coastal strip) is very close to that of the Saigon area. For
another early account of the southern dialect, see Maurice Grammont, Recherches experi-
mentales sur la prononciation du cochinchinois, Memoires de la SocUete de Linguistique de
Paris 16.69-86 (1909-10). A summary statement of pronunciation of Saigon speech in pho-
nemic terms is presented in R. B. Jones and Huynh Sanh Thong, Introduction to spoken
Vietnamese 1-7 (Washington, 1957). A modern phonemic analysis of northern speech is
presented in M. B. Emeneau, Studies in Vietnamese (Annamese) grammar 5-43 (Berkeley
and Los Angeles, 1951).
4 The nonuniqueness of phonemic solutions of phonetic systems, Bulletin of the Institute
of History and Philology, Academia Sinica (hereafter BIHPAS) 4.363-97 (1934).
5I am indebted to Samuel E. Martin for encouragement to discuss at greater length the
possible alternative analyses, as well as for specific suggestions concerning certain points
in these analyses, and especially for corroboration of the obvious similarities between prob-
lems in Saigonese and those in Mandarin which he treated in his Problems of hierarchy and
indeterminacy in Mandarin phonology, BIHPAS 29.209-29 (1957).
6 A set of postulates for phonemic analysis, Lg. 24.3-46 (1948); Studies in colloquial
Japanese IV, Phonemics, Lg. 26.86-125 (1950).
7 A manual of phonology (Baltimore, 1955).

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
456 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

speech invites mention at this point: before pause and before a voiceless initial
in the succeeding syllable, a final phone is partially or wholly unvoiced.
Phones are grouped into forty-five phonemes: eight defined by features of
stress, five tones, twenty-one nonsyllabics, and eleven syllabics.

2. STRESS

Two aspects of stress are phonemically distinctive: the relative intensity of


stress onsets and the contour of intensity of stress over the syllable.
2.1. Relative stress defines four phonemes.8
/"/ extra-loudstress
/'/ loud stress
medium stress (unmarked)
/?/ weak stress
The symbol for a relative stress phoneme is placed at the beginning of its syllable.
(In citations involving more than one syllable, a space separates each syllable
from the next.)
/?XAWf '"pihk'./ '[I1 don't know.'
/toj xAwQ "' tij !/ 'I'm not going!'
Loud and extra-loud stresses are HEAVY;weak and medium stresses are LIGHT.
Syllables vary in length in proportion to the loudness of stress; those with /?/
are very short.
A series of syllables accompanied by the phoneme of medium stress is fre-
quently (but not always) characterized by an alternation of slightly louder
syllables with less loud syllables such that the final syllable of the series is (a)
louder if it is followed by pause or a weakly stressed syllable /?/, (b) less loud
if it is followed immediately by a heavy stress / " '/.9
2.2. Stress contours define four phonemes. Whatever its relative intensity,
each syllable has also a distinctive contour. Most syllables are characterized
by a stress which begins relatively loud and decreases in intensity gradually at
a more or less constant rate: this is DIMINUENDO;
it is written with a comma, but
only when it occurs at the end of a pause group. Other contours deviate from
this pattern:
/./ MORENDO (rapid decrease in intensity)
/?/ SOSTENUTO(sustaining of the original intensity of the syllable, followed
by a gradual decrease)
/ !/ CRESCENDO
(increasing intensity after the onset of stress, followed by a
gradual decrease)
8 The present analysis of stress is unsatisfactory from several points of view; it is in-
cluded here because other attempted analyses, while containing germs of a more basic and
meaningful description at some points, are impossible to state as overall systems without
at other points distorting the phonetic facts or adducing statements which lie at least for
the present in the realm of physical indeterminacy. The most significant progress in under-
standing the accentual system has resulted from a study of the changing rates of speed in
allegro speech.
9 Jones and Thong treat this feature presumably at a morphophonemic or possibly a
morphemic level (Introduction6-7); they regard it as the fundamental pattern, indicating
exceptions as they occur. My material shows vastly more numerous exceptions in the flow
of allegro speech; I have therefore not adopted this pattern as the underlying one.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 457

The symbol for a stress contour is placed at the end of its syllable.
/AWf ?kA' '?pihk'! xAwU./ 'Do you know [or just surmise]?'
/toj xAwu ' tij !/ 'I'm not going!'
/'tij xAwI)?/ 'Are [you] going [or not]?'
/xij naw' AWfw'?tij, .../ 'Whenever you go, ...'
Syllables accompanied by /!/ are considerably longer than those accompanied
by other stress contours.

3. TONES

Five phonemes are defined by a combination of relative pitch heights and


contours. In order to specify these distinctions it is necessary and sufficient to
recognize seven divisions of the normal speaking-voice range, numbered 1
through 7 from low to high. The symbol for a tone phoneme is placed at the end
of its syllable just before its stress contour symbol.
In general, tones have somewhat lower allophones with /./ than with /,/,
somewhat higher allophones with /?/, and with /!/ allophones with a greater
range between the lowest and highest parts of the contour. One tone, however,
has special allophones in these environments (see next paragraph).
High nonrising pitches (unmarked):
with /,/ and heavy stresses [54]
with /,/ and light stresses [5]
with /./ and heavy stresses [41]
with /./ and light stresses [4]
with /?/ [ss]
with / !/ [61]
/xij naw' Aw| '?tij,/ 'Whenever you go, ...'
/xij naw' 'toj ?tij,/ 'Whenever I go, ...'
/toj xAwu ' tij./ 'I'm not going.'
/AwU ?kA' '?pihk'! xAwn./ 'Do you know [or just surmise]?'
/tij 'xAwu?/ 'Are [you] going [or not]?'
/toj XAWI '?tij !/ 'I'm not going!'
In a sequence of several phones distinguished by the quality [i], a phone ac-
companiedby /'/ or /"/ is slightly higherthan the others.
/toj ?xAwfl '?tij ??pAj jag'./ 'I'm not going now.'
/'/ High rising pitches [67]. In a sequence of several phones distinguished by
this quality, each succeeding phone is slightly higher than the preceding, so that
the last is the highest of all. When a heavy stress occurs medially in such a
sequence, it is accompanied by the highest [67] in the sequence, and succeeding
phones are slightly lower but successively higher in their own series.
/no' noj' tihn' 'fap'./ 'He's speaking French [of a child].'
/tihn' fap' 'xa' noj' ?cag'./ 'On the contrary, French is hard to speak.'
/1/ Mid rising pitches [35]. With weak stress, the rise is somewhat shorter and
more abrupt.
/hiw7/ 'understand', /?kuw^ 'han'/ 'onion'
/v/ Low nonfalling pitches. All syllables accompanied by this tone end in a
weakly articulated glottal stop:
with syllables accompanied by /?/ and those ending in/p t k/: [2].

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
458 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

with other syllables: [23].


/?jv 'ka'./ 'Yes, thereis [polite].'
/?tepV/ 'beautiful', /metv/ 'tired', /?tyhkv/ 'possible', /nair/ 'heavy',
/sagV/'be afraid'
/'/ Low falling pitches [21].
/lkm'/ 'do, make'
4. NONSYLLABICS
The Saigonese nonsyllabic system involves six positions and four manners of
articulation; the inventory is shown in Table 1.

o o
N

Z- Err
? Xo

z <
N
0d! Z U2
-<P-i P4 O 0 JO
a < < . Q

FOBTIS STOP p t t C k ?
FORTISCONTINUANT f 8 S j X10 -
LENIS W t' r 1 g h
NASAL m n 1 1J -
TABLE 1. NONSYLLABICS.

The two apical positions merge in the nasal articulation; /n/ has both retroflex
and nonretroflex allophones. In final position nonsyllabics show fewer distinctions
of both position and manner; they are reduced to a 4-by-3 system: labial /p w m/,
frontal/t j n/, dorsal/k g n/, and glottal /? h/. In such codas, the distribution
of fortis and lenis oral continuants is predictable in the following terms: fortis
postsyllabic oral continuants occur in syllables accompanied by /v/ and in those
accompanied by /'t/ and heavy stresses; elsewhere in codas only lenis al-
lophones occur. Oral continuants which occur in codas are SEMIVOWELS;
other
nonsyllabics are CONSONANTS.
All initial nonsyllabics are weakly labialized before /w u o/. Final stops are
unreleased; final nasals and semivowels are unreleased when they precede a
homorganic onset. Final nasals have freely varying allophones with slight oral
stop onset.
4.1. Labials.
/p/ Simple" fortis (bi)labial stops.
After initial /?/: [b] imploded or plain in free variation; imploded phones are
more common in syllables accompanied by heavy stresses.
10/x/ has an allophone with stop onset.
11The term SIMPLE is used to distinguish these phonemes from dorsovelar phones with
coarticulated full labial closure, which belong to dorsal phonemes (cf. ?4.4 /k n/).
In the definitions of phonemes, phonetic details which are not essential to the definitions
but are of descriptive interest are enclosed in parentheses.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 459

Final: [p].
/?panv/ 'friend', /tApv/ 'to practice'
/f/ Fortis voiceless labial (labiodental) continuants.
/fap'/ 'France'
/w/ Lenis labial oral nonsyllabics.
Initial before /j/: [v] palatalized.
/wja/ 'and'
Before /y/, and after /i y u A/: nonsyllabic [u].
/wyr/ 'Huan [family or given name]', /twy)'/ 'week'; /?iw/ 'cherish',
/kyw'/ 'sheep', /iuw~/ 'sleep', /?tAw/ 'unspecified location', /^Aw/
'gentleman'
Before syllabics other than /y/, and after /e a/: nonsyllabic [u].
/kwen/ 'forget', /kwa/ 'to pass', /waj'/ 'outside'; /new'/ 'if', /?aw/
'bee'
After /e a/: nonsyllabic [o].
/hew/ 'pig', /?taw/ 'sick'
/m/ Simple (bi)labial nasals.
/mam'/ 'fish sauce'
4.2. Apicals.
/t/ Fortis apical nonretroflex stops.
After initial /?/: [d] alveolar, imploded or plain in free variation; imploded
phones are more common in syllables accompanied by heavy stresses.
Final before /t/ in the following syllable: [t] alveolar.
Elsewhere (i.e. initial and final before pause or before phonemes other than
/t/): [t] dental.
/?tij/ 'go'; /?it' 'ta'/ 'little tea'; /toj/ 'I', /9it7/ 'little [quantity]', /sat'
'^Awf/ 'your book', /?it' 'syhr/ 'little milk'
/s/ Fortis (voiceless) apical nonretroflex spirants.
/se/ 'vehicle'
/t'/ Lenis (voiceless aspirated) apical nonretroflex (dental) stops.
/ttoj/ 'stop'
/t/ Fortis (voiceless slightly affricated) apical retroflex (alveolar) stops.
/ta/ 'tea'
/s/ Fortis (voiceless) apical retroflex (alveolar) spirants.
/syhl/ 'milk'
/r/ Lenis (voiced) apical retroflex (alveolar) oral continuants. In free vari-
ation: [r] flap or trill, accompanied or not by spirantization; [z] retroflex.
/ra/ 'go out'
/n/ Apical nasals.
Final before /t/ in the following syllable: [n] alveolar retroflex.
Final before /?t/ in the following syllable: [n] alveolar.
Elsewhere (i.e. initial and final before pause or before onsets other than
/t ?t/): [n] dental.
/?an 'tca/ 'you pay [speaking to male equal]'; /?an ' tij/ 'you're going';
/new'/ 'if', /?an/ 'elder brother', /?an 'sagV/ 'you're afraid'
4.3. Frontals are characterized by the acoustic effect of the front of the

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
460 LANGUAGE,VOLUME35, NUMBER3 (1959)

tongue touching or approaching the alveolar ridge without accompanying lip


closure.'2
Ic/ Fortis (voiceless slightly affricated) laminoalveolar stops.
/cijv/ 'elder sister', /cyh/ 'not yet'
/j/ Relatively high front (voiced unrounded) semivowels.
Before /i y/, and after /i y u/: nonsyllabic [i].
/jijv '?twao/ 'superstition', /jyDI/ 'to offer', /jyj'/ 'underneath', /tuj7/
'years old'
Before vowels other than /i y/, and after /a o a A/: nonsyllabic [I].
/jej7/ 'easy', /wjej'/ 'return home', /juw/ 'wander', /jo'/ 'wind', /jAjv/
'get up'; /taj'/ 'arrive at', /toj/ 'I', /haj/ 'two', /nAj'/ 'this'
After /o/: nonsyllabic [e].
/hcj7/ 'ask'
/1/ (Lenis [e]-colored) laterals.
/lyht/ 'fire', /lij/ 'drinking glass'
/Ji/ Laminoalveolar nasals.
/pa'/ 'house', /Jiw'/ 'much'
4.4. Dorsals.
/k/ Fortis unaffricated dorsal (mediovelar voiceless) stops.
Final after /u w/: [k] with simultaneous labial closure.
Final after /o/: [k] with simultaneous labial closure (more common) in free
variation with [k] strongly labialized.
Final after /o/: [k] weakly labialized.
Elsewhere (i.e. initial and final except after /u w o o/): [k].
/luk'/ 'time [as specified periodi', /hawk/ 'to study'; /tok'/ 'good'; /nokv/
'sweet'; /kam/ 'orange', /nyhk'/ 'water'
/x/ Fortis dorsal (mediovelar voiceless) spirants and affricates.
Initial after /k / in the preceding syllable: [kx](stop articulation very weak
after /u/).
Initial except after /k i/: [x].
/?0yk' 'xwe7/ 'health', /hai' 'xAwU/ 'air line', /xAwfl/ 'not', /Jliw' 'xij/
'many times', /t'it' 'xAwU?/ 'Do [you] like [it] or not?'
/g/ Lenis voiced dorsal (mediovelar) oral nonsyllabics.
Initial after /k / in preceding syllable: [g].
Final: nonsyllabic high back unrounded vowel, lower after /3/ than after /y/.
Elsewhere (i.e. initial after pause or phonemes other than /k !/): [T].
/mokv 'gawk'/ 'one corner', /?kog 'gi'/ 'the hen'; /tyg/ 'fourth', /sagv/
'be afraid'; /gar 'kyh'/ 'knock at the door', /?kAj' 'gej'/ 'the chair'
/j/ Dorsal (mediovelar) nasals.
Final after /u w/: [o] with simultaneous labial closure.
Final after /o/: [ul with simultaneous labial closure (more common) in free
variation with [U]strongly labialized.
Final after /o/: [u]weakly labialized.
Elsewhere (i.e. initial and final except after /u w o a/): [u].
12 This makes clear that
palatalized [vl, an allophone of /w/, belongs with the labials,
not here.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 461

/?tui'/ 'correct', /?Awu/ 'gentleman'; /?pon'/ 'four'; /run/ 'good-tasting';


/?paV'/ 'table', /?tehn/ 'black'
4.6. Glottals.
/?/ Glottal stops. /?/ is in free variation with zero initially before vowels; in
addition, as stated in ?3 above, all syllables accompanied by /v/ end in a weakly
articulated glottal stop. I adopt the convention of not writing /?/ where its
occurrence is always predictable (i.e. at the end of syllables accompanied by
/v/), and of writing it where it is in free variation with its absence, as well as
before /p t/, where it is never predictable.
/sagv/ 'be afraid'; /?an/ 'elder brother', /?pa/ 'three', /?tij/ 'go'
/h/ Lenis glottal continuants: nonsyllabic modified anticipations and pro-
longations of syllabic nuclei.
Initial: voiceless anticipation of following vowel.
/hiw~/ 'understand', /het'/ 'be completed', /hew/ 'pig', /hyg/ 'be spoiled',
/han/ 'more', /huk'/ 'suck up', /hopv/ 'box', /hajj/ 'ask', /hAj/ 'interest-
ing', /han'/ 'onion', /hak'/ 'sing'
Postsyllabic: voiced centralized prolongation of preceding syllabic-

AFTER/i Y U/ AFTER/e/
BEFORE/k/ [h] [h2]
BEFORE/n/ [h2] [h3]
FINAL [h3] _

[h], higher mid central semivowel.


/?pihk'/ 'know', /?tyhkv/ 'possible', /t'uhk'/ 'medicine'
[2], mean mid central semivowel.
/tihj'/ '[spoken] language', /?tyhV'/ 'road', /?puhj'I 'sad'; /xehk'/ 'having
odor of burning substance'
[h3],lower mid central semivowel.
/nihl/ 'idea', /myh/ 'rain', /muh'/ 'season'; /?tehn'/ 'light, lamp'

5. SYLLABICS

The syllabic system is conveniently described as consisting of a main system


of higher vowels and two subsystems of lower vowels.
The main system has a two-by-three pattern:

FRONT-CENTRALBACKUNROUNDED BACKROUNDED
HIGH i y U
MID e 0o

The vowels of this system occur in diphthongs before their homorganic higher
semivowels (i.e. /ij ej yg ag uw ow/); they do not occur finally.13
The first subsystem has three vowels which occur long in final position; they
13 Phonetic observation suggests that there may be a morphophonemic alternation in

which these sequences of vowel plus homorganic higher semivowel are replaced by simple
vowels under weak stress, but the facts are not clear.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
462 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

FRONT CENTRAL BACK BACK


UNBOUNDED UNROUNDED UNBOUNDED ROUNDED

U
UPPER HIGH i Y
is y2 U2

ia
LOWER HIGH ya
i2 Us

o
UPPER MID e
e2 et a 02

a2
MEAN MID
e

LOWER MID 82 e8 03
A A2 3

02
UPPER LOW a aa

a a,
LOWER LOW
a2 a, a4

FIGURE 1. SYLLABIC ALLOPHONES.

are front unrounded, back rounded, and low (unrounded):


e 0

a
The second subsystem has two vowels which do not occur long;L4they are
characterized by the opposition of lip-spreading (upper low nonback phones) to
its absence (lower mid nonfront phones):
a A

Allophones of syllabic phonemes are located on the traditional vowel quadri-


lateral as shown in Fig. 1. With nonhigh tones (i.e. mid rising, low falling, and
low nonfalling), /a/ has allophones which are slightly farther back than those
shown; in the same positions, allophones of other vowels are slightly lower than
shown.
The syllabics /a A/ occur nasalized before final /wi/.
/nawI'/ 'hot', /?AWU/ 'gentleman'
14 In citation forms these vowels are regularly shorter than all other vowels, but in con-
nected allegro speech this difference in length is frequently absent.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 463

A syllabic occurring in the environment /m-w/ occurs slightly nasalized or


not, in free variation.
/mew'/ 'cat', /mcw/ 'fast'
The forms /hA/ 'isn't that right?' and /hew/ 'pig' were recorded with free
variants containing nasalized syllabics.
The relative length of vowel allophones is predictable with respect to tones
and the types of syllable ends. With high tones vowels are relatively short, with
low tones somewhat longer, and with mid tone considerably longer. With respect
to codas, they are longer in syllables ending in semivowels and/or nasals than in
those endings in stops, and are longest in syllables without codas.
5.1. The main system. In syllables accompanied by heavy stresses, the high
syllabics are noticeably longer than the mid syllabics before labial codas.
5.11. High syllabics.
/i/ High nonback syllabics.
After /c j Jp/ before /p w m j h/: [i].
/jip'/ '[vehicle] spring', /ciwv/ 'endure', /Jlim'/ 'porcupine', /cij/ 'some-
thing', /cih/ 'divide'
After /c j pj/ before /t n/: [i2].
/cin'/ 'nine', /wjitv/ 'duck', /Jinv/ 'abstain'
Except after /c j j1/ before /p w m j h/: [ia].
/kipv/ 'arrive on time', /hiw7/ 'understand', /tim'/ 'to search', /mij~/
'America', /ihn/ 'peaceful'
Except after /c j p/ before /t n/: [i4].
/?it'/ 'little [quantity]', /?pin'/ 'container for liquids'
/y/ High back unrounded syllabics.
After /c j jp/ before /k D h/: [y].15
/jpyk'/ 'throbbing pain', /jy^/ 'to offer', /cyh/ 'not yet'
After /c j p/ before /j g/;16 except after /c j jp/ before /p m k n h/: [y2].
/jyj'/ 'underneath', /cyg7/ '[written] word', /jpyg/ 'similar to', /?yp'/ 'to
perfume', /lymv/ 'gather' /syk'/ 'strength', /lyj/ 'spine', /?tyh '/ 'road'
Except after /c j pi/ before /w j g/: [y3].
/kyw'/ 'save', /myj'/ 'ten', /sygv/ 'matter'
/u/ High back rounded syllabics.
Before /p m/: [u].
/hupv/ 'sink in water', /pium'/ 'a pinch of [a substance]'
Before /w j h/: [u2].
/uuw/ 'to sleep', /wjuj/ 'enjoy oneself', /muh'/ 'season'
Before /k /: [u3].
/fukv/ 'clothes', /?vtun'/ 'correct'
5.12. Mid syllabics.
/e/ Upper mid nonback syllabics.
Before/p w j/: [e].
/repv/ 'bedbug', /kew/ 'to call', /wjej'/ 'concerning'
15 This allophone would be expected before /p m/ also, but no examples occur.
16This allophone would be expected after /c j jn/ before /w/ also, but no examples occur.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
464 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

Before /m/: [e2].


/mem'/ 'soft, pliable'
Before /t n/: [e3].
/cet'/ 'die', /len/ 'ascend'
/a/ Mid back unrounded syllabics.
Before/p m k n/: [a].
/lap'/ '[school] class', /?pamv/ 'thief', /ak'/ '[green or red] pepper', /tafl/
'audacious'
Before /j g/: [a2].
/taj'/ 'arrive at', /agV/ 'be afraid'
/o/ Mid back rounded syllabics.
Before/e/: [o].
/QoI/ '[spoken] word'
Before /w m j k/: [02].
/wjow/ 'enter', /tom/ 'shrimp', /loj~/ 'error', /mok?/ 'one'
Before /p/: [03].
/hopV/ 'box'
6.2. First subsystem. In syllables accompanied by stresses other than /0/,
/e o/ are noticeably shorter than /a/ before labial codas; /e/ is shorter than
/a a/ before final /k/.
/e/ Lower mid front syllabics.
Before /p k h/: [r].
/tepv/ 'beautiful', /sca '?tek'/ 'Sa Dec [a province of South Viet Nam]',
/?tehn/ 'black'
Final and before /w/: [e2].
/0e/ 'listen', /m?w'/ 'cat'
Before /m/: [83].
/?em/ 'younger sibling'
/o/ Low back rounded syllabics.
Before /j n/: [o].
/noj'/ 'speak', /hon'/ 'island'
Before /k/: [02].
/gokV/ 'sweet'
Final and before /pm/: [33].
/pjo/ 'small', /kopv/ 'tiger', /hom'/ 'coffin'
/a/ Lower low syllabics.
AFTER
/t cj Jy ? h/17 ELSEWHERE
BEFORE /p k/ [a] [a2]
FINAL AND BEFORE /W j )/ [a2] [a3]
BEFORE /m/ [c3] [c4]

17 Note that here and elsewhere the notation 'after /?/' includes syllabics in initial posi-

tion in the absence of the glottal stop.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 465

[a]
/tIak'/ 'waterfall',/cap'/ 'last month of year', /jap'/ 'armor', /jlakv/
'music',/?up'/ 'oppress',/haCk'/'sing'
[l2]
/t'a13'/ 'month', /caw'/ 'greet', /ja'/ 'price', /jpaw/ 'reciprocally',/a'/
'Asia',/haj/ 'two'; /tacp'/ 'to answer',/lakv/ 'lackingin flavor'
[a3]
/t'am/ 'be greedy',/cam'/ 'Cham[a once powerfulpeople of south central
Viet Nam]', /jam'/ 'dare',/Jiam'/ 'bored',/?am'/ 'darken',/ham'/ 'jaw';
/ka'/ 'fish', /gawv/ 'husked rice', /maj/ 'tomorrow', /?pauv/ 'friend'
[04]
/nom/ 'south', /kam/ 'orange'
6.3. Second subsystem.
/a/ Nonback upper low syllabics, accompaniedby lip-spreading.
Before /k i/: [a].
/mak'/ 'expensive',/?aU/ 'eat'
Before /p w m t n/: [a2].
/gapv/ 'meet', /?awU/ 'bee',/lam'/ 'to great extent', /sat'/ 'book',/manV/
'strong'
/A/ Lowermid nonfrontsyllabics,without lip-spreading.
BEFORE /j/ FINAL AND BEFORE /j/ ELSE-
WITH HEAVY STRESSES WITH LIGHT STRESSES WHERE

AFTER/pfWmCj1j1xgij/18 [A] 1
} [A] [A2]
AFTER /t st' t e r n k ? h/ [A2]j

[A]
/'?pAj/ 'close relation to speaker', /'kwAj'/ 'disturb', /'mAj'/ 'unspecified
number', /xAwU cawu' ?'tij' 'cAj'/ 'sooner or later', /'jAj'/ 'paper', /'lAj'/
'take', /'flAjV 'viscous, gluey', /'xAj-/ 'excite' [infrequent], / 'gAj/ 'cane',
/')AJ/ 'stupid'; /?cA/ 'for benefit of', /??tA'/ 'location near hearer', /?Aj'/
'just mentioned', /kAj/ 'stick [wood]', /hAj/ 'interesting'
[A2]
/'tAj/ 'west',/"?tAj/ 'location nearspeaker',/'sAj/ 'build',/t'Aj '/ 'teacher',
/'tAJ'/ 'scratched', /'SAJ'/ 'dry [by heat]', /'rAj/ 'sift [flour]', /'nAj'/
'this', /'kAj/ 'stick [wood]', /'Aj'/ 'just mentioned', /'hAj/ 'interesting';
/0tApp/ 'to practice', /'jAw'/ 'oil', /'nAm'/ 'mushroom', /fAU/ 'manure', etc.

6. DISTRIBUTION
OF PHONEMES
6.1. Syllables and syllable types. Distributional facts are most conveniently
stated in terms of the syllable, the smallest unit of Saigonese speech which
occurs in isolation. Each syllable contains the followingelements,presentedin
the conventional order in which they are written (optionally means that the
18/f/ does not occur before final /A/ or /AJ/.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
466 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

VP (?)t(w) &(w) C(W) k(w) ?


f 8 (W) S(W) j(W) X(W)
w(j) t'(w) r 1(w) g(w) h
m n J(w) i(W)
TABLE 2. ONSETS

element occurs only in some syllables; elements not so labeled occur in all syl-
lables): (a) one relative stress; (b) an onset-a single nonsyllabic or cluster;19
(c) one vowel; (d) optionally a coda-a single nonsyllabic or cluster; (e) one
tone; and (f) one stress contour.
Syllables may be meaningfully classified according to their terminations:
Type I. Without coda /ca/'father'
Type II. With coda
A. Semivowel /maw/ 'fast'
B. Nasal /kcm/ 'orange', /?awf/ 'gentleman'
C. Stop /mokV/ 'one', /hawkv/ 'to study'
6.2. Onsets are shown in Table 2. Note that aside from /?p ?t wj/ all clusters
have /w/ as second element, and that the only three-phoneme cluster is /?tw/.
/?pacUv/ 'friend', /faw '/ 'room', /wa/ 'flower', /wja'/ 'and', /mokv/
'one'; /toj/ 'I', /?tuw^/ 'enough', /twyg'/ 'week', /jij '?twaU/ 'supersti-
tion', /se/ 'vehicle', /swaj'/ 'mango', /t'oj/ 'stop', /t'wij^/ 'water [as one
of the elementsl',/Iay 'tea', /twihu'/ 'transmit', /sagv/ 'be afraid', /swaUV/
'prepare', /roj'/ 'already',/new'/ 'if'; /cyh/ 'not yet', /cwihBv/ 'conversa-
tion topic', /jug'/ 'to use', /jwihg/ 'destiny', /lAw/ 'long [time]', /lwj '/
'species', /piiw'/ 'much', /jiwaj'/ 'exhausted'; /ka'/ 'fish', /kwa/ 'cross
over', /xJ'/ 'difficult', /xwaj/ 'potato', /gej'/ 'chair', /gwa'/ 'be left spouse-
less', /loIu/ 'good-tasting', /wcj '/ 'outside'; /?an/ 'elder brother', /haj7/
'ask'.
The clusters /wj kw/ tend to disappear in less careful speech; the resulting
forms are the same as the corresponding popular forms. Since these clusters
are distinguished in the official orthography but are not present in the popular
speech of the dialect area, it is reasonable to suggest that in the educated dialect
of Saigon they represent artificial restorations under the influence of the written
language.
/wj/ is replaced by /j/ /ja'/ 'and'
/kw/ is replaced by /w/ /wa/ 'cross over'
The cluster /gw/ is exceedingly rare: only the form /gwa'/ 'be left spouseless'
was recorded. Dictionaries list a few forms which would indicate a cluster */nw/,
but none of these occur in the speech of the informants used.
6.3. Syllabics.
Back rounded syllabics /u o a/ do not occur after /w/; /A/ occurs after /w/
only before/j/.
/kwAj'/ 'disturb'

19In one sense the onset is an optional element; but since the option consists of the occur-
rence or absence of /?/ and since all syllables without onset have alternates with the onset
/?/, it is more economical to consider the presence of an onset as constant.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 467

i- y- u- e- a- o- e- a- o- a- A-

- - - - - - e o - A
-j ii yj uj ej aj oj - aj oj - AJ
-w iw yw uw ew - ow ew aw - awk/q Aw(k/l)
- - - - - - -
-g- yg og
-h ih(k/n) yh(k/nJ) uh(k/n) - - - ehk/q - - - -
-t it - - et - - - - - at -
-n in - - en - - - - - an -
-p ip yp up ep ap op ep ap op ap ap
-m im ym um em am om em am am am am
-k - yk uk - ok ok ek ak ok ak ak
-3 - yn un - an on cx a0 01 an A1
TABLE 3. COMBINATIONSOF SYLLABICSAND CODAS
The notation /Aw(k/n)/ should be read: /A/ occurs with a coda /w/ and with the
complex codas /wk wn/; the notation /awk/n/ should be read: /a/ occurs with the
complex codas /wk wn/ but not with the simple coda /w/; and similarly for similar
notations.

In Type I syllables (without coda) only /e a a/ occur commonly.


/se"/ 'future action', /Joif/ 'small', /kCa'/ 'fish'
/A/ occurs as a morphophonemic alternant replacing final /o/ in syllables ac-
companied by /?/; it also was recorded with louder stresses in two forms.
/co/ varying with /?cA/ 'for benefit of'; /?^'/ 'informal address particle',
/hA'?/ 'isn't that right?'
6.4. Codas are limited to the oral continuants (i.e. semivowels) /w j g h/, the
nasals /m n u/, the stops /p t k/, and the clusters /wk wi hk hD/.20
/maw/ 'fast', /haj/ 'two', /jpyg/ 'similar to', /myh/ 'rain'; /ncm/ 'south"'
/?an/ 'elder brother', /?ai/ 'eat';/ fap'/ 'France', /sat'/ 'book', /makv/
'face'; /hawkV/ 'to study', /?Aw4j/ 'gentleman', /?pihk'/ 'know', /tihi'I
'money'
Syllables with codas involving stops occur only with the tones /' v/.
/?aU 'kap'/ 'steal', /tapt/ 'general merchandise', /sat'/ 'book', /metv/
'tired', /nyhk'/ 'water', /makv/ 'face'
6.6. Combinations of syllabics and codas are shown in Table 3.
/g/ occurs only after the homorganic syllabics /y a/; /h/ occurs only after
the high syllabics /i y u/, except that in the clusters /hk hg/ it also follows /e/.21
/hyg/ 'spoiled', /mog/ 'to dream'; /uihj/ 'idea', /myh/ 'to rain', /muh'/
'season'; /?pihk'/ 'know', /tihi'1 'money', /nyhk'/ 'water', /tyhif/ 'think',
/cuhkv/ 'rat', /?uhl'/ 'to drink', /xehk'/ 'having odor of burning sub-
stance', /?tgh)/ 'black'
/w/ does not occur after /o o/, but only after /i y u e o e a/ (and after /a/
in the clusters /wk wi/; see below).
/piw'/ 'much', /kyw'/ 'sheep', /Uuw^/ 'to sleep', /new'/ 'if', /kow/ 'young
lady', /hew/ 'pig', /maw/ 'fast'
20The predictable occurrence of a weakly articulated glottal stop at the end of syllables
accompanied by /V/ is ignored in the statement of phoneme distributions.
21 Martin
feels that [e'k] should be interpreted /ek/, parallel to /ok/. Note, however, the
four occurrences of [ek] which were recorded (listed below, toward the end of this section).

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
468 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

/j/ does not occur after /e a/, but only after /i y u e a o a A/.
/?tij/ 'go', /yj'/ 'person', /muj'/ 'salt', /wjej'/ 'return home', /maj /
'invite', /toj/ 'I', /hacj/ 'two', /haj7/ 'ask', /nAj/ 'this'
/t n/ occur only after /i e a/.
/?it'/ 'little [quantity]', /lin'/ 'soldier', /het'/ 'completed', /nen/ 'there-
fore', /$at'/ 'book', /?an/ 'elder brother'
/k n/ do not occur after /i e/, but only after /y u a o a a a A/.
/mykv/ 'ink', /?tyi'/ 'don't', /luk'/ 'time [as specific period]', /tun'/
'correct', /?ak'/ 'pepper', /lan'/ 'large', /tok'/ 'good', /UoD/ '[spoken]
word', /hak'/ 'sing', /Sac'/ 'shine', /akV/ 'sweet', /nOn/ 'good-tasting',
/mak'/ 'expensive', /naUv/ 'heavy', /mAk'/ 'lose', /IAU'/ 'time, instance'
/n/ does not occur after /e/, but four occurrences of /ek/ were recorded:
/?Ek' '?ck'/ 'pig's cry [onomatopoetic]', /?ehn '?ck'/ id., /sa~ '?tek'/
'name of a province of South Viet Nam [of probable Cambodian origin]',
/mek'/ 'announce' (variant of /mat'/ id.)
/wk wU/ occur only after /a A/.
/hawkv/ 'to study', /nawi'/ 'hot', /?Awk'/ 'snail', /^Aw/ 'gentleman'
6.6. Stress contours.
/./ and /?/ occuronly in pause-groupfinal.
/toj xAwn '?tij./ 'I'm not going.'
/'?tij XAWI?/ 'Are [you] going?'
/ !/ occurs only with heavy stresses.
/?Awfl ?kA' '?pihk'!xAwn./ 'Do you know [or just surmise]?' /toj XAWf
"?tij!/ 'I'm not going!'
7. SAMPLETEXT
As an example of a connected text in the dialect, a short narrative is trans-
scribed below; the story is a personal experience related by Dr. Nguyen-van-Hien.
Each sentence in the transcription is numbered, and these numbers correspond
to numbered sentences in the free translation which follows.
/1. luk' 'tyhk',?toj ?kA' haj?kan 'ca', t'AkV 'nowv,?wja' jejv 't'yhn. 2.?mokv
?kan?cA' 'mykv, wjc' "mokv 0kan)?cA' 'fehn'. 3. t'yhn' 'lejv, toj 'jAn" ?cun'
?nA' ?tijjawV 'caj,?moj~ buj^ 'ciw', taj' 'tok'. 4. 0mokv 'hom, tojjAnF ?cun'
'?ten' 0mok gawk' ry'. 5. ?cpac kwa mokv ?0oj jia' 'kih, 0t'in' 'lin', ?haj
?kan 'ca' t'ihkv 'la' ?wja' t'Akv 'jyg~, ?tow~ ?uh' 'ra, ?uh" ?Am' 'len,?wja'
?rianran?muhn' 'kan'?cun' toj. 6. ?tokv nokv 'kwa'! 7. toj lihn' kuj'?suhn' ?om
'cAm' ?tyhkv 0kag?cA' mykv, ?tan?luhn' 'kuhn' 0?ag"cABI toj. 8. kan'?kaB
?cA' 'fehn' wjij'
sagv waci, nen 'tAwn len 'tuhn cajV 'ton'. 9. haj?kao ?CA' 'lan'
?tuj^ few saw. 10. xij wan"lajv,t'ij' ?kan 'fehn' ?kuh" toj??tac ?pihg' mAk',
tawi)Stam' 'ka^, ?jyj' tihi)' 'nuj'. 11.tai)^Baj' 'vta', ?^wja' kwa homn'aaw, ?toj
kim' 'maj^, kim' 'luhn, xAWn '?tyhkv. 12. ?ten' "jia' 'nAj', ?ten' ?"pa' 'kih,
haj" 'ttam, "ma'XAWBaj 't Aj' fehB' ^ag^ ?tAw.13. Vtit 't Ao, toj
k)
wjow naj tawn 'ryn', tim' 'kim','pjyn,xAwB t'Aj' tam 'jacn "kanca' naw' het'.
14. t'Ak' wjawg v,?toj 'tag' wjej', ?An hAnv?wjow 'kun'.15.?toj nij^ loj~ "tajc
'toj,
toj tfjaj 'nyh'?luk' ?pAk' tak' ?Aj'. 16.?kan ?CA' mykvta^ 'wje~, t'AkV
'?puhn ', ?wjij' mAk' '?painv. 17. toj?kun- '?puhn', moj~ xij 'niji ?ten' na'./

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGONPHONEMICS 469

1. I used to have two very cute, lovable dogs. 2. A black one and a brown one.
3. I used to take them out for a walk every nice afternoon. 4. One day I took them
to a certain part of the woods. 5. As we were passing one house, suddenly two big,
vicious dogs rushed out, barking and showing their teeth as if they wanted to
bite us. 6. It was very sudden and unexpected. 7. I immediately bent over and
managed to pick up the black dog cowering at my feet. 8. But the brown one,
very much frightened, looked up at me, then crowded past and ran away. 9. The
two big dogs set out in pursuit. 10. When I turned to look, my brown dog had
disappeared in a field at the bottom of the hill. 11. All that day and the next I
kept looking and looking, but it was no use. 12. I went from house to house in-
quiring, but no one had seen the brown dog anywhere. 13. I went myself right
into the forest to search, but found no trace of any dog at all. 14. Discouraged,
I returned home, feeling very sorry. 15. I felt at fault: I should have taken pre-
cautions against such an eventuality. 16. The black dog looked very sad, for he
had lost his companion. 17. I feel sad, too, whenever I think about her.

8. ALTERNATIVEANALYSES

Although numerous possibilities for alternative analyses appeared in the


collation of this Saigonese material, only a few of the more interesting and sig-
nificant ones are discussed here.
8.1. Nonsyllabic systems. In considering the nonsyllabic structure of the
dialect it is important to note that the system presented in ?4 above was arrived
at only after a rather intensive search for the features which are distinctive
throughout the system. Employing the usual criteria of position, manner, and
voiced-voiceless distinctions, we find the pattern of phones shown in Table 4.
Limitation to initial or final position is shown by hyphens. There appear to be
few symmetrical features, and those that are evident seem difficult to handle.
Two of the plain voiceless stops have limited distributions, while the other two
have not. There is only one aspirated stop. One of the affricated stops, [kx],
must clearly be classed with its homorganic voiceless spirant, and one of the
voiced stops, [g], with its homorganic voiced spirant. The other two voiced stops
share the feature of preglottalization.
Attempting to follow the phonetic reality closely, we might easily set forth

Labial Apical Retroflex a i Dorsal Glottal


alveolar
Stops, plain voiceless -p t k ?
Stops, voiceless aspirated t'-
Stops, voiceless affricated t- c- kx-
Stops, voiced ?b- Sd- g-
Spirants, voiceless f- s- - x- h-
Spirants, voiced v- ?- 7-
Nasals (voiced) m n -n Jo q
Laterals (voiced) 1-
Trills (voiced) r-
Semivowels (voiced) w j -i -A
TABLE 4. NONSYLLABICPHONES

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
470 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

ONSETS CODAS

t'(w)
t(w) t (w) c(w) k(w) t (w/h)k
b b
d(w)
g(w) g
r
f(j)
s(w) s(w) x(w)
m n ji(w) i (w) m n (w/h)3
l(w)
w j(w) h w j h
TABLE 5. FIRST NONSYLLABIC REANALYSIS: ONSETS AND CODAS

t[h(w)]
t[r(w)]
tj (w)]
t(w) k(w)
b
d(w)
g(w)
r
f(j)
s[r(w)] x(w)
s(w)
m n[j(w)] l;(w)
l(w)
w j (w) h
TABLE 6. SECOND NONSYLLABIC REANALYSIS: ONSETS.

an analysis extracting [?] as predictable. Table 5 shows the occurring onsets and
codas in such a system. Here we have put [-p] and [b-] together in /b/, [f-] and
[v-] (occurring only before /j/) together in /f/.22 And we have placed [i] in /g/
and [A] in /h/ as before. The appearance now tempts us to introduce clusters:
we can place retroflex articulation in the /r/ phoneme, laminoalveolar articulation
in /j/, and aspiration in /h/. We have an option here of interpreting [c p] as
22 Here Martin favors putting [v] with
[b], and presumably [p f] together in another
phoneme. This analysis would make a happy correspondence to certain idiolects in which
speakers pronounce [bj] in positions where most Saigonese say [vj] or simply [j]; the speakers
in question are few, and with many of them this feature disappears in allegro speech. Since
these individuals seem always to have had extended affiliations with the Vietnamese edu-
cational system, the phenomenon has the earmarks of a hyperurbanizing tendency. It is
interesting that Maspero mentions this pronunciation in quite a different context: '...
l'inaptitude des populations chames annamitisees a prononcer certains sons annamites a
donn6 naissance, en cochinchinois, A des sons nouveaux: p' pour f, et bvpour v' (Etudes 3).
It is especially interesting to note that the [b] involved in this cluster is phonetically quite
different from the preglottalized phone of the regular educated dialect (my /?p/): it is
lenis and never preglottalized or imploded. To return to Martin's suggestion: although
presumably this argument would fall into the category of his 'off-hand notions of "phonetic
similarity" ' (Problems 229), it seems relevant to point out that [p b] are stops, fortis,
bilabial, whereas [b v] have in common only a general labial position and voicing-[v] is
lenis, labiodental, and a spirant.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 471

either /tj nj/ or as /kj ij/, although the definition of /j/ for the latter alternative
is difficult. (We might also interpret [1]as /rj/, although it would prove difficult
to define /r/.) We then have a considerably more symmetrical inventory. Table 6
shows the resulting onsets with the laminoalveolars interpreted as /tj nj/; the
codas are the same as those shown in Table 5.
8.2. Syllabic systems. The positing of eleven syllabic phonemes, even in such
a balanced system as that shown in ?5 above, will displease advocates of symbol
economy. At first glance variations in length seem to provide an opportunity
to reduce the number: the official spelling system implies by the symbol a that
our /a/ is the short variety of /a/, and our /A/ has likewise for some time been
considered the short variety of /a/.23
Other differences of length are equally noticeable. For northern speech A. G.
Haudricourt has suggested that we should recognize six short and long pairs
(each pair presumably constituting one phoneme).24For Saigonese, however, a
problem of indeterminacy arises immediately: so far as I have been able to ob-
serve, the differences of length which are regularly present in syllables accom-
panied by heavy stresses disappear entirely under weak stress; with medium
stress there is considerable variation; but under all stress conditions the vowel
qualities remain distinct.25 But the phonetic reality of weakly stressed syllables
is extremely elusive; future investigation may reveal length distinctions in them.
Leaving aside for a moment the question of these indeterminate lengths in
weakly stressed syllables, we may return to a consideration of other aspects of
the nine-vowel inventory achieved by this reanalysis. Having extracted length
as distinctive in the case of /a A/ we have syllable terminations without con-
sonantal codas as shown in Table 7.
Utilizing the other differences of length apparent in heavily stressed syllables,
we arrive at the reanalysis shown in Table 8. In this analysis we have eliminated
the qualitative distinction between /e o/ and /e o/, reducing the basic syllabic
inventory to seven vowels: /i y u e A o a/.
In addition to the difficulty of apparently lacking length distinctions under
23 It has not been possible to trace the beginning of this treatment, but it is significant

that Vietnamese school texts teach children in spelling to read the printed symbol d (my
/A/) as /a/ with high rising tone /'/, parallel to the reading of a (for my /a/) as /a'/. Recent
phoneticians, such as Le-van-Ly (Le parler vietnamien [Paris, 1948]) and Le-ngoc-Tru
(Chanh-ta Viet-Ngu [Saigon, 1951]), recognize these sounds as related in a way which would
suggest that they are short and long varieties of the same phonemes. Jones and Thong (1-2)
analyze my /a/ as my /a/ plus length, but do not link the other two vowels; Martin favors
extracting either length or shortness in both cases.
24 Les voyelles breves du vietnamien, BSLP 48.90-3 (1952). For his phonetic information
he cites Le-van-Ly 17-44; Emeneau; and Nguyen Bat-tuy, Chuva van Viet khoa-hoc(Saigon,
1951). It should be emphasized that Haudricourt's analysis concerns only an apparent
subsystem of the Tonkinese syllabic pattern, and some of his distinctions are exemplified
by only a very few forms; it is thus quite a different matter from the reanalysis we are about
to consider.
25Except for the weakly stressed alternate of /u/, which I hear as [A]; some native
speakers have insisted that the vowel here is (phonemically) distinct from /A/ and should
be assigned rather to /o/. It should be noted that the native speaker may be influenced by
the orthography, which is based on the stressed form; a similar problem arises with the
morpheme /kaj '/, 'general classifier', under weak stress, which I hear as clearly /?kAj'/.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
472 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

INVENTORY NO CODA FINAL /j/ FINAL /W/

iy u - - - i yj uj iw yw uw
e A(A) o - A - - ej Aj Aj oj ew AW - ow
e a(a-)o e- a - - aj oj ew - aw -
FINAL /g/ FINAL /h/
yg ih yh uh
Ag -

TABLE 7. NINE-VOWEL REANALYSIS: SYLLABLE TERMINATIONS WITHOUT


CONSONANTAL CODAS

NO CODA /j/ CODAS /W/ CODAS /g/ CODAS

- - - ij - uj iw - uw yg
- A - ej AJ oj ew AW(k/I) ow Ag
- - awk/ -
APICAL CODAS
- - - - - - y*w
yj
it/n - -
e - o - Aj oj e-w -
-
et/n
a a-j a-w
at/n

/h/ CODAS LABIAL CODAS DORSAL CODAS

ih(k/A) yh(k/j) uh(k/j) ip/m - up/n a - yk/lj uk/A


ehk/j - ep/m Ap/m op/n n ek Ak/A ok/A
ap/m ak/r
ip/m yp/m u p/i mn - - _
- A'p/m - - A-k/l o-k/
a-p/m a?k/q
TABLE 8. SEVEN-VOWEL REANALYSIS: SYLLABICS AND CODAS

weak stress, this analysis has an added problem: the distinctions implied between
/ip im up um/ and /iPp im u p u m/ presume the assignment of phones to
/i u/ which are assigned in ?5 to /e o/ before /p m/; these phones are the same
as or lower than phones which must remain assigned to /e o/ in other positions.
Unless re-examination of the phonetic reality reveals different criteria for defining
/i u/ in these positions, the analysis must therefore be abandoned.26
8.3. Reinterpretations of semivowels. It seems important to mention two other
possible ways of handling the combinations of syllabic and nonsyllabic vowel
phones. The first partly follows the pattern implied in the official orthography,
adopted also by Jones and Thong, as well as by Emeneau for Tonkinese (where
the circumstances are somewhat different) :27it involves the assignment of vowels
with their homorganic off-glides to one phoneme for each span. That is, instead
of /ij ej yg ag uw ow/ we have simply final /i e y a u o/ respectively. An especially
attractive feature of this reanalysis is the elimination of the high back unrounded
semivowel coda with its limited distribution. Column A of Table 9 shows the
resulting syllable terminations without codas and with codas involving semi-
vowels.
26 In addition, tape-recorded materials leave considerable doubt whether the /i p i m/
of this analysis always represent vowels which are distinctively longer than /ip im/, even
with the louder stresses.
27 Jones and Thong 1-2; Emeneau 5, 10. The phonetic descriptions of Jones and Thong
seem to imply different phonetic facts in these cases: diphthongization is specifically ruled
out in the case of /e/ and not mentioned in the other cases.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGONPHONEMICS 473

A B
i y u NO CODAS - - -
e a o
e a o
- A

- yi Uj /j]/ CODAS ii yi ui
aj oj ei ai oi
- ai oe
- _ Aj
A] - Ai

iw yw - /W/ CODAS iu yu uu
ew - _ eu - ou

EW aw - Eo ao -

- AW AU

/g/ CODAS yy
ay

ih yh uh /h/ CODAS iA yA UA

ihk yhk uhk CLUSTER CODAS iek yek uek


ihj yhq uh3 iLA yAI UAI
ehk/i eAk/q
awk/n Awk/n auk/A Auk/I

TABLE 9. DISTRIBUTION OF REINTERPRETED POSTSYLLABIC SEMIVOWELS.


A. Homorganic semivowels assigned to preceding vowel phonemes;
B. Semivowels assigned to vowel phonemes.

The second possibility interprets all postsyllabic semivowels as nonsyllabic


allophones of corresponding vowel phonemes; thus our /-j/ becomes /-i/ in all
cases except after /o/, where its allophone must be assigned to /e/, and so on.
Syllable terminations are shown in Column B of Table 9. This treatment suggests
a similar reinterpretation of presyllabic semivowels, which would result in the
removal of /j w/ from our system, the removal of the oral continuant manner
from our table of codas, and the establishment of a new phoneme /v/ replacing
/w/, necessary in order to distinguish such forms (in our original system) as
/wjAu'/ 'to rime' and /wihi'/ 'name of the low falling tone symbol'-now
/viAU'/ and /uiAU'/ respectively.

9. THEORETICALIMPLICATIONS

Structural analysis is concerned with perceiving and describing significant


order in the superficial chaos of languages. In phonology we may say that this
order has typically the form of symmetrical patterns in phonemic inventory and
distribution, and takes account of parallelisms in the phonetic data. It is gen-
erally agreed, however, that asymmetrical features appear in most or all lan-
guages. To date the structuralist has concerned himself with discovering the
symmetry, the pattern, and has tended to regard the asymmetry, the lack of
pattern, as 'interesting' (with an overtone of 'quaint, exotic'), regrettable, or
irrelevant.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
474 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

In phonological descriptions some asymmetrical features appear at the level


of phonetic description, others in the phonemic inventory, others only in dis-
tributional statements, and still others are not treated at all. The last category
is a troublesome one, since it includes not only what the analyst has observed
and has decided to regard as irrelevant, but also features which have escaped his
observation. It may be that at some future time we will have a need-and,
let us hope, also a method-for dealing with this level. For the time being it is
rather the other categories which concern us.
It is the intention of this paper to propose a procedure for dealing with asym-
metry, not primarily because it exists, or even because it is 'interesting', but
because through it we may test the validity of our basic symmetrical system.
The application of this procedure will frequently give us a choice among phonemic
solutions that otherwise seem equally valid; in many cases it will reflect a more
rigid and tangible basis for a choice which might otherwise seem based rather on
'esthetic sense' or 'intuition'. It has already often-I might even say usually-
been applied; but so far as I know, it has not been stated as a principle, and its
implications have not been studied.
The principle is a simple one; it assumes the elimination of asymmetrical
features from the phonemic inventory and the distributional table wherever pos-
sible, and their treatment on the level of phonetic description. Systems thus
evolved are compared: that one is preferable which contains the fewest asym-
metrical aspects not directly representing phonetic reality. To apply the principle
we ask the following questions:
(1) Do apparently missing items from the inventory or the distributional
table represent actual absences on the phonetic level?
(2) Do phonemes which fall outside the symmetrical pattern of the inventory
lack all features which might place them in some vacant position in the system
or suggest a different set of defining criteria for the overall arrangement?
(3) Do anomalous sequences represent occurring sequences on the phonetic
level?
(4) Do similar anomalous inventory items and sequences represent similar
anomalies on the phonetic level? Do contrasting anomalous inventory items and
sequences represent similar contrasts on the phonetic level?
In comparing various possible systems, the one that returns the fewest negative
answers to these questions is deemed best. Since the questions are suggestive as
well as critical, they are also useful in bringing out possible reorientations of
the material.
9.1. Empty positions in the pattern. The first reanalysis in ?8.1 (cf. Table 5)
shows no voiced spirant and no semivowel in the apical nonretroflex column, no
voiced stop or semivowel in the retroflex column, no voiced stop and no spirants
in the laminoalveolar column: these lacks all correspond to nonoccurrences on
the phonetic level (cf. Table 4). Two absences (dorsal semivowel, glottal voice-
less spirant) are only apparent; the phonemes /g h/ cover these positions. But
the absence of a glottal stop does not represent the phonetic reality. On the dis-
tributional level, we find no /w/ clusters with /b f m r n h/; all these absences
correspond to phonetic facts.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
SAIGON PHONEMICS 475

9.2. 'Leftover' phonemes. The leftovers in the first system of ?8.1 are /t' 1/.
Our new point of view suggests redefining the voiced spirant manner as voiced
oral continuant manner, so that /1/ may be added in the laminoalveolar column,
although it will also occur to us that /j/ could fill the same position. At first
/t'/ seems to defy any reanalysis in these terms; but the fact that /t'/ alone
among the stops is lenis suggests that we re-examine the list of lenis phones. It
was precisely this re-examination which led to the revised overall system pre-
sented in ?4.
A similar approach to the vowel system suggests the interpretation of /a A/
as the short varieties of /CIa a/, as described in ?8.2. Since other factors are also
involved here, I shall reserve the discussion for ?9.4; but it must be noted here
that the posing of this question is frequently suggestive in another direction, even
if it results in no reanalysis: it may call attention to the existence of subsystems
to which some or all of the leftovers belong.
9.3. Anomalous sequences. It is this consideration which brings out the ab-
surdity of the second analysis of nonsyllabics proposed in ?8.1. Of the anomalous
sequences /fj th tr tj sr nj/ only /fj/ and possibly /th/ represent actually oc-
curring sequences on the phonetic level. The occurrence of all these sequences
before /w/ in onset clusters, like those involving single nonsyllabics, is pointed
up by this approach; and we are prompted to notice the further anomaly that
*/rw hw nw/ do not occur on either the phonemic or the phonetic level, while
we have posited /trw srw thw njw/. The reanalysis involving /kj nj/ instead of
/tj nj/ presents similar problems. Note that this question actually amounts to
an expansion of the general principle of cluster patterning in phonemic analysis,
although the point of view is somewhat different.
9.4. Parallels and contrasts implied by anomalous items. A further weakness
of the second reanalysis of ?8.1 appears when we note that the anomalous se-
quence /fj/, justifiable in terms of Question 3, is apparently parallel to /tj nj/,
although the latter represent quite dissimilar and nonparallel phonetic facts.
In the reanalysis of /a A/ (cf. ?8.2 and Table 7), the implied length contrasts
between /a- A / and /i e u o/, as well as /A/, are generally justifiable on the basis
of phonetic facts; but the apparent contrasts with /y e 3/ are spurious. Similarly,
the parallels in length suggested by /A/ compared with /i e u o/ are valid; those
implied between /A/ and /y ? o/ are not. Examination of the rest of the distribu-
tional table for this analysis multiplies the examples, and shows that /a/ is
involved in similar anomalies.
These considerations led us to assign phonemic status to all differences in
length, and resulted in the system presented next in ?8.2 (cf. Table 8). Although
we find this analysis untenable for other reasons, it is useful to consider here an
interesting aspect of the distributional pattern, in order to demonstrate a more

ih yh uh - yhj - - yhw - ihp/m yhp/m uhp/m ihk/3 yhk/ij uhk/l)


eh - oh - 9hj ohj ehw - - - ahp/m - ehk/q ahk/q ohk/j
ah ahj ahw ahp/m ahk/q
TABLE 10. VOWEL LENGTH REINTERPRETED AS /h/: /h/ CODAS.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions
476 LANGUAGE, VOLUME 35, NUMBER 3 (1959)

intricate working of the principle under discussion. In this analysis we note that
the diphthongs with postsyllabic /h/ complement the long vowels almost exactly
in final position and before /k n/. This suggests including vowel length in the /h/
phoneme, and yields the syllable terminations shown in Table 10.
In this system, Question 4 would not be posed for the clusters of /h/ with dor-
sals or for final /h/, because the distributions show symmetrical patterns. But the
asymmetry which appears in the clusters of /h/ with /j w/ prompts a considera-
tion of parallels and contrasts. In the asymmetrical distributions with /j w/
there is no conflict: the clusters shown represent always the same phonetic
reality-long vowel plus coda-and the vacancies indicate the real lack of such
sequences. The same consistency is shown before labial codas. But the apparent
parallel /yh: yhj/ represents a different reality from the parallel /ah: ahj/;
similarly, while /ah: ahp/ represents the phonetic facts directly, /ih: ihp,
yh: yhp, uh: uhp/ suggest phonetically nonexistent parallels. This case suggests
a kind of divided rating for such a system: the distortion resulting in this system
is clearly less disturbing than that of the system discussed at the beginning of
this section. Generalization on the basis of one case is undesirable, but this
example will encourage us to examine others and perhaps arrive at a gradation of
credit for varying degrees of distortion.
9.5. Consideration of the nonsyllabic and syllabic systems presented in ?4
and ?5 in the light of this principle shows that their asymmetrical aspects, both
in inventory and distribution, represent the phonetic reality quite directly. It
is on this basis that they were chosen. Note in this connection that the choice
of the best interpretation of semivowels rests not only on the symmetry of the
nonsyllabic system, but also on the fact that the pattern of syllable terminations
displays more symmetry in terms of this interpretation than the second solution
discussed in ?8.3 (Table 9, Column B).
If we had more meaningful criteria for identifying asymmetry (or symmetry)
in stress and tone systems, the same procedure might be applied with profitable
results to these systems also.

This content downloaded from 86.164.127.110 on Fri, 4 Apr 2014 08:52:09 AM


All use subject to JSTOR Terms and Conditions

Vous aimerez peut-être aussi