885095993

WEDNESDAY MORNING, 1 MAY 1991

INTERNATIONAL B, 8:00 A.M. TO 12:15 P.M.

Session 4SP

Speech Communication: Segmental Acoustic Properties

Mark Liberman, Chair

Linguistics Department. Uniuersity of Pennsyhania, Williams Hall, Philadelphia, Pennsyloania 19147

Contributed Papers

8:00

4SP1. Velar height and sentence length: Deciination? F. Bell-Berti (St. John’s Univ., Jamaica, NY 11439 and Haskins Labs., New Haven, CT 06511) and Rena A. Kraków (Tempie Univ., Philadelphia, PA 19122 and Haskins Labs., New Haven, CT 06511)

Considerable evidencc atfests to the phenomenon of /D deciination during the course of an utterance. Recently, the question of whether upper articulators might show declination-type eflfects has been raised for the jaw [E. V.-Bateson and C. A. Fowler, J. Acoust. Soc. Am. Suppl. 1 84, SI28 (1988)]. In previous studies, lower velar peaks were noted in sentence-final than -initial syllables, prompting a systematic investiga-tion of velar deciination. Thus peak velar height for early and late syllables in natural and reiterant-speech sentences of from three to nine syllables has been examined. Velar position and acoustic data were collected from three speakcrs of American English. DifTcrences in velar height were examined with respect to number of syllables and acoustic duration. Possible interactions between declination-type processes and factors such as stress were also studied, sińce strcss has been shown to have a strong influence on velar movement [R. A. Kraków, J. Acoust. Soc. Am. Suppl. 1 82, SI7 (1987); J. Vassiere, Phonetica 45, 122-139 (1988)]. (Work supported by NIH Grants DC-00121 and RR-05596 to Haskins Labs.]

8:15

4SP2. The F3 trajectories of American English /r/*s. Suzanne Boyce and Carol Y. Espy-Wilson (Res. Lab. of Electron., MIT, Cambridge, MA 02139)

It is well known that the coarticulatory influence of /r/ on preccding segments, as evidenced by a lower than expected F3 frequency, may be substantial. Researchers who have studied this phenomenon have found that this “spreading” occurs under a number of differcnt conditions (cf. Espy-Wilson, 1987). A unifying expłanation may be that the articula-tory gesture for /r/ has a relatively incompressible trajectory whose timing “<!lides” with r«sp<sct to other gcsturcs in a scqucncc. Tłiin hy-pothesis was tested acoustically by analyzing the time course of the f3 trajectory in several American English speakcrs’ production of the words “varam,” “vavram,” and “vavam.” In these data, previous find-ings of “.spreading” to the preceding segments were confirmed. In ad-dition, the /rPs in “varam” and “vavram” showed a nearly identical time course. As a consequence, F3 during the post-vocalic /v/ of “vavram” is considerably lower than was seen in the post-vocalic /v/ of “vavam.” This suggests that the exp!anation for spreading of the feature retroflex into the /v/ is due to a constrained time course for /r/. This hypothesis will be further exp!ored in words when /r/ cooccurs with /g/, a segment whose tongue-body requirement potentially conflicts with the articulatory requirement for /r/.

1916 J. Acoust. Soc. Am., Vol. 89, No. 4, Pt. 2, April 1991

8:30

4vSP3. Acoustic parameters for place of articulation in nasals and stops. Kathleen Mary Kurowski (Dept. of Cognitive and Linguistic Sci., Box 1978, Brown Univ., Providcnce, RI 02912)

A metric that identified labial and coronal place of articulation in English nasal consonants with over 80% accuracy had been developed by correlating distinctive pattems of rapid energy change at the nasal release with each place of articulation (K. M. Kurowski and S. E. Blumstein, J. Acoust. Soc. Am. 81, 1917-1927 (1987); K. M. Kurowski, unpublishcd dissertation, Brown Univ. (1990)]. This study attempted to address two questions. (1) Could the same metric be succcssfully applied across manner to classify place of articulation in homorganic voiced stops? (2) In the event that the metric failed here, to what extent did the burst onset alter the pattems of rapid energy change predicted by the metric? The same three speakers used in the nasal study (1990) produced CV syllables, consisting of [b,d] followed by [i,e,a,o,u]. The results showed that the classification scores across speakers for the voiced stops (labial = 64%; coronal = 83%) were lower than thosc for nasals (labial = 79%; coronal = 87%). Still, the generał patteming of the results was the same across manner: Labial scores remained lower than thosc of the coronals duc to the intractable problem of classifying labials in the cnvironment of [i]. Two reanalyses using different window sizes demonstrated that the aperiodicity of the initial portion of the burst inflated some targeted areas of energy change, adversely affecting the performance of the metric. Various approaches to the question of how to reconcile formal aspects of the present metric with the added complexity introduced by the voiced stops are discussed in the light of their ramifications for the theory of acoustic invariancc. [Work supported by NIH.)

8:45

4SP4. An acoustic phonetic description of the American English /d/ in two modes. Josephine Homa (Dept. of Linguistics/ISALS, New York Univ., 719 Broadway, New York, NY 10003)

Phoneticians traditionally describe American English /d/ as a voiced, nonsibilani, interdental Fncative. However, in flowing speech, /d/ often sounds stoplike. This study investigates the acoustic attributes of stoplike, as well as continuant, variants of /d/ in “the” occurring in various phonological conditions in text-reading and conversation modes. The results of spectrographic analysis allow for the definition of five degrees of constriction and two types of constriction release, burst and smooth formant transition into the following segment. The degree of constriction is a coarticulatory effect of the preceding phoneme. Smooth formant transitions tend to occur with less constricted degrees of constriction; a burst may occur with any degree of constriction and is prevalent across phonological conditions. Articulatory interpretations that explain these and other variations of /d / are forwarded. The results also show intermodal differences in /d / in that there is a higher

121st Meeting: Acoustical Society of America 1916