Abstract:
Macmillan, Goldberg, and Braida [J. Acoust. Soc. Am. 84, 1262-1280 (1988)] show that when stimuli are presented along a continuum, instances of consonants that are near good exemplars are most memorable for labeling by a listener. However, for vowels (produced in isolation), instances which are near category boundaries are most memorable. Can these results be generalized to vowels produced in context? What is the effect of different consonant contexts on vowel perception? To answer these questions, this study tested high vowels along a front--back continuum produced in two CVC contexts (labial or alveolar). The results show that these vowels produced in context exhibit the patterns Macmillan et al. (1988) discovered for consonants, consistent with the idea that both vowels in context and consonants are perceived as linguistic entities, whereas isolated vowels are not. The results also show a difference in perception between the two consonant contexts. For vowels produced in the labial context, the boundary between front versus back perception was closer to the /u/ end of the continuum than for vowels produced in the alveolar context. This reflects compensation by listeners for the expected effects of coarticulation with the flanking consonants [Ohala and Feder, Phonetica 51, 111--118 (1994); Bradlow and Kingston, J. Acoust. Soc. Am. Suppl. 1 88, S56 (1990)]. [Work supported by NIH Grant No. 5-R29-DC01708.]