Perception of Vowel Features in Temporally-Segmented Noise Portions of Stop-Consonant CV Syllables

Abstract

The ability of listeners to identify vowel features, given only segments of the aperiodic portion of CVs, was investigated. Segments of the aperiodic portions of stop consonant CVs, increasing in duration in 10-msec steps from onset, were identified by 18 listeners. The responses were analyzed for the correct identification of vowel features. Coarticulatory effects of the vowel on the aperiodic portion were found to (1) occur early in the aperiodic portion, (2) vary with consonant and vowel, and (3) vary with vowel feature. In general, however, tongue advancement for the vowel was identified correctly most often, tongue height next most often, and the tense versus lax distinction least often. For some CVs, sufficient clues for above chance level of identification of a feature were present during the shortest segments of the aperiodic portion. The findings lend support to the point-of-view that listeners may be able to narrow the choice of the vowel in an unvoiced-stop-consonant CV to a small number of alternatives prior to the beginning of voicing.

Subcategorical phonetic mismatches slow phonetic judgments

“…Vowels can be identified much better than chance from the friction of a coarticulated fricative by itself (Yeni-Komshian & Soli, 1981). The vowel information in release bursts is generally poor, even for bursts of longer duration than the ones used here (Cullinan & Tekieli, 1979;Kewley-Port, 1980). Thus, any delay caused by inappropriate vowel information may actually be due to the burst's being taken as appropriate to a stop not among the choices in the task.…”

“…Since the bursts were necessarily chosen for their minimal place information (so that the transitions would override the burst), their lack of a slowing effect is not too surprising. The stop can be identified to some extent from the burst alone (Kewley-Port, 1980;Tekieli & Cullinan, 1979;Winitz, Scheib, & Reed, 1971), but that does not tell us whether that information would be overriding in different contexts.…”

Identification of vowel features from French stop bursts

“…It is worth recalling that /k/ is velar when followed by the velar vowel /u/, and post-palatal when followed by the palatal vowel /i/. The high identification rates of the vowel /u/ from /k/ and of the vowel /i/ is in good agreement with results of previous experiments [9,12,3]. The vowel /u/ was very poorly identified from /t/, the places of articulation of the stop and of the following vowel being probably too far from each other to allow the vowel articulation to be achieved at the consonant release.…”

“…With regard to the identification of vocalic features, these low confusion rates confirmed the very high intelligibility of the front-back feature, already noted in previous studies, especially in [3].…”

“…The vowel /i/ was the most easily identified. Cullinan and Tekieli [3] found that vocalic cues were also present in the first milliseconds of natural bursts. They showed that listeners had more than by chance ability to identify the front-back feature 10 ms after the release, and the vowel height feature 20 ms after the release.…”