Flege ASA 97 Boston 1980

8/3/2019 Flege ASA 97 Boston 1980

1/5

,,

Speech Communication PapersPresented at the 97th Meeting of the

Acoustical Society of America

Massachusetts Institute of TechnologyCambridge, Massachusetts

12-16 June 1979

Edited by Jared J. Wolf and Dennis H. Klatt

Copyright (C) 1979 Acoustical Society of America, 335 E. 45th Street, New YorNY 10017

To order additional copies:Send $15 per copy (check or money order, no purchase orders) to:Ms. Betty GoodfriendAcoustical Society of America335 E. 45th StreetNew York, NY 10017


2/5

L3TEMPORAL CORRELATES OF [voice] IN ARABIC-ACCENTED ENGLISH

James E. Flege, Linguistics, Indiana U., Bloomington IN

Temporal correlates of [voice] for initial andfinal stops in Arabic and English are examined. Crosslanguage differences in phonetic implementation of[voice] are found to have direct effects on the [voice]contrast produced by Saudi Arabians speaking English.

English speakers are known to use a number of different acoustic dimensionsto contrast /p,t,k/ and /b,d,g/ (Lisker, 1978). The phonetics literature of thepast 20 years has made it abundantly clear that there is more to the English[voice] contrast than just vocal fold vibration, and that the acoustic correlatesused to contrast sounds like /p,t,k/ and /b,d,g/ differ across languages. Amongthe acoustic dimensions used to contrast voiced and voiceless post-stress stopsin English are two important temporal correlates: preceding vowel duration (Chen,1970) and consonant closure duration (Li~ker, 1957). Closure duration may alsodistinguish pre-stress voiced and voiceless stops (Stathapoulos and Weismer,1979).At this time it is unclear to what extent this pattern of temporal correlates of[voice] is unique to English ..

In this study we investigated temporal phonetic correlates of stop voicingin American English and Saudi Arabian Arabic to determine if these two languagesdiffer in their use of duration to implement the voicing contrast. We also wishedto learn whether any differences in phonetic implementatfon of [voice] would affect production of the English voicing contrast by Saudi Arabians. Most previousstudies of phonological interference in second language acquisition have dealtsimply with replacement of one speech sound by another rather than consideringdetails of the phonetic implement~tion of phonological contrasts in the targetlanguage (Tarone, 1976).

Our first hypothesis was that there is direct phonetic interference in second language acquisition. According to this hypothesis, Arabs use only Arabic phonetic correlates of [voice]when contrasting English /p,t,k/ and /b,d,g/ ratherthan adopting the correlates of [voice] used by Americans. Our second hypothesiswas that differences in phonological inventories between Arabic and English wouldaffect implementation of [voice] in Arabic-accented English. Teachers of Englishas a Second Language report that Arabs often cannot pronounce English /p/, whichseems to follow from the fact that Arabic lacks a phoneme /p/. If our second hypothesis is correct; it would follow that Arabs should be less able to produce acontrast between /p-b/ in a completely English-like way than one between /t-d/ or/k-g/. Finally,"we wished to learn to what extent several years of residence in anEnglish-speaking environment would render phonetic implementation of [voice] morelike that of native English speakers.

Methods. Two similar production studies were carriedout, one in Arabic and the.other in English. ArabicCVC test words (Table 1) differing in pl.ce of articulation and voicing ob both pre- and post-stressstops were read in an Arabic carrier sentence by 8Saudi students at Indiana University. In the EnglishExperiment similar CVC English words were read in acomparable English carrier sentence by three populations: (a) 8 native speakers of Midwestern AmericanEnglish (b) 8 Saudi students who had lived less than

TabLe 1. Test Words. 1 year in the U.S. (Group Ar-l) and (c) 8 Saudi stu-dents who had lived in the U.S. for more than 2 years (Group Ar-2). Preliminaryresults of 6 speakers in each population are reported here. Wide-band spectrogramswere made of 6 tokens of each test word in both experiments, and measurements were

[IiGlIStiUP.AliC UP.

.......... ,.....

TIST WOROS

."IoU

tAO~.'"'"IT)lAD.,co..

171


3/5

made of several acoustic intervals to the nearest 5 msec. Analyses of variance weperformed on a number. of the measured intervals. We report here on only consonantand vowel durations.

Results. These experiments reveal that a number of differences in the phonetiimplementation of [voice] exist between Arabic and English. First, .we found thatArabic, unlike English, pre-stress /t-d/ and /k-g/ are distinguished on the basisof closure duration (Fig. 1). The phonologically voiceless stops are significantlYlonger than the voiced stops for both places of

articulation (p '

.-,

ARABIC UP.

N

,-,

Finol Consonant DurationE NGl .I SH EXP .

IpIIbI Nidi NJqISiQNf.

'.Ii""

I ..

u.E


4/5

Finally. we found that use of vowel durationas a correlate of post-stress stop voicing also dis

Figupe 4. Dupation diffep- ti~guishes native English frbm Arabic-accented Engences between post-stpess 1ish. In Arabic we found a small (6 msec.) but nonvoiced and voiceless stops. significant effect of consonant voicing on the duration of preceding vowels. In Arabic-accented English we found that the differ-ence in duration of vowels before /k/ vs. /g/ and

/t/ vs. /d/ is far smaller that that produced by Americans. The vowel duration'difference produced bythe two Saudi groups is. however. significant inthree of these four cases(pc.01). We also discoveredthat neither group of Saudis produced a significantvowel duration contrast before /p/ vs. /bj. This isinteresting since the more experienced group of Saudis did produce a contrast,for /p-b/ closure.

VO_" bekx. Voiced Con5Oi'lG'lt.minu.s V O_ 's b el rx e V Oi ce ln . C on so na nl .

(pc.01). For both groups. however. the difference induration between final voiced and voiceless stops ismuch less than that produced by the Americans. Anexception to this is the /t-d/ pair which. as mentioned above. is largely neutralized by flapping bythe Americans. The more experienced group of Saudis(Group Ar-2) makes /t/ significantly longer than /d/(pc.01).

.-,-b

VOICe'''' milly. Voiced StOf)Closur. o...,o/ionin Word F inol Po. it ion

..

:s.,..

Figupe 5. Dupation diffepences between vowels ppeceding voiced vs. voiceless consonants.

Discu~sion and Summary. We have seen that Arabstend to use Arabic temporal correlates of [voice]when speaking English. thus producing a duration contrast for initial but not final stops. and virtuallyno' use of vowel duration as a correlate of poststress stop voicing. We now wish to consider how

two other factors - speakers' experience with English as a second language. andthe fact that Arabic lacks /p/ - affect production of the English voicing contrast.

lOp-lOb bo t - bo d bocll-bo9

We see some indication that length of residence in an English-speaking environment affects phonetic implementation. The less experienced group produced smallbut significant differences in vowel duration as a function of the voicing of afollowing velar or alveolar stop. This same group, however, failed to produce aclosure duration correlate of [voice] for post-stress stops. The more experiencedSaudi speakers, on the other hand. seem to have acquired a durationa1 contrastfor post-stress stop voicing, a finding which has also been noted for the moreexperienced of two groups of Finns learning English by Suomi (1976). The seemingpreference of a consonant duration correlate over a vowel duration correlate of[voice] by the more experienced group can be seen in the production of the minimal pair bat-.E...l.The six speakers of Group Ar-1 produced a significant vowelduration contrast, but no closure duration contrast; whereas the more experiencedgroup. Ar-2, produced no vowel duration contrast but did make /t/ significantlylonger than /d/ (pc.01). This last finding is especially interesting when we consider that the American speakers iQ the surrounding community tend to neutralizeany /t-d/ durationa1 contrast by flapping. These findings suggest that the moreexperienced Arab speakers of English have generalized post-stress stop closureduration as a correlate of [voice] from the labial and dental places of articulation. Gen~ra1ization of a sub-featura1 voicing correlate may even extend topre-stress position: we see that the more experienced group also produces a larger duration contrast for intia1 stop voicing that that which exists in Arabic.

Contrary to expectation, the lack of /p/ in Arabic does not seem to greatlyaffect phonetic implementation of [voice] in Arabic-accented English, at leastinsofar as the temporal parameters discussed here are cpncerned. We do see somewhat more of a non-English use of duration to contrast /p-b/ in initial positionthan for other stop pairs; and we also see somewhat less use made of vowel dura-'tion before /p-b/ than before /t-d/ and /k-g/. But these differences are small.and are consistent with a larger pattern of temporal inteference.

The frequently reported production of [b] for intended /p/ seems tobe

173


5/5

more importantly the result of incorrect presence of glottal vibration during stopclosure. In our production data we found that 28% of pre-stress {pis and 61%of post-stress {pis were inappropriately voiced, as far a5 can be obs~rved on asound spectrogram.

In summary, we have found that Arabic and English differ in their use of temporal correlates of stop voicing: Arabic contrasts pre-stress stops differing in[voice] by duration, but not post-stress stops. Just the opposite is true of English. These differences between English and Arabi'c result in temporal interferencein Arabic-accented English. We found that Arab speakers of English continue touse Arabic phonetic correlates of [voice] when speaking English, and that thisphonetic interference does not seem to be greatly affected by either severalyears of experience with English in an English-speaking environment or by differences ifl the phonological inventories of Arabic and English.

ReferencesChen, M. (1970) "Vowel length variation as a function of the voicing of

the consonant environment,"Phonetica, 22,129-159.Lisker~ L. (1957) "Closure duration and the intervocalic voiced-voiceless

distinction in English," Language, 33, 42-49.Lisker, L. (1978) "Rapid vs. Rabid: A catalogue of acoustic features which

may cue the distinction," Haskins Labs Status Reports, SR-54.Port, R. (1976) The influence of speaking tempo on duration of stressed

vowels and medial stops in English trochee words, unpublished Ph.D. dissertation, Univ. of Connecticutt, available from the I.U. LinguisticsClub, Bloomington, IN.

Stathapoulos, E. and Weismer, G. (1979) "Duration of Stop Consonants,"this volume.Suomi, K. (1976) English voiceless and voiced stops as produced by

native and Finnish speakers, Jyv~skyla Contrastive Studies, 2,Jyv~skyl~ University, Finland.

Tarone, E. (1976) "Some influences on Interlanguage phonology,"Working Papers ~n Bilingualism, 8, Ontario Institute for Studies in Education, Bilingual Education Project.

174

Documents

Flege ASA 97 Boston 1980