Categorical tendencies in imitating self-produced isolated vowels

An earlier experiment requlrlng literal imitation of synthetic isolated vowels from [uJ-[iJ and [iJ-[a=J continua (Repp & Williams, 1985) was replicated using as stimuli vowels produced by the subj ects themselves. Even though imi tation accuracy was much improved, the responses deviated from the stimuli in ways similar to those observed previously with synthetic stimuli. That is, categorical tendencies (nonlinear stimulus-response mappings of formant frequenci es, nonuniform response vari abil i ty across each continuum, and peaks in formant frequency distributions) were obtained even with stimuli that matched the subjects' articulatory capabilities. This rules out one possible explanation of the observed categorical tendencies, viz., that they arise in the perceptual translation of synthetic stimuli into a talker's production space. Introduction In a recent study (Repp & Williams, 1985), we investigated the claim that SUbjects' vocal imitations of isolated, steady-state vowels follow a categorical pattern Two subj ects (the authors) imitated synthetic vowel s from 12-member [u J-[i J and [i J-[a= J cont i nua at three different temporal delays, which had li ttle effect on response patterns. The functions relating stimulus and (average) response formant frequencies across each vowel continuum exhibited local changes in slope, response standard deviations varied, and the distri butions of response formant frequencies showed distinct peaks and valleys. The response patterns thus showed categorical tendencies, but few instances of strictly categorical responses (Le., identical responses to different stimuli representing the same vowel category). Where do these categorical tendencies in imitation come from? There are at least four independent (but not mutually exclusive) possibilities, some perhaps more plausible than others. The tendencies could originate either in the SUbjects' perception of the stimulus vowels or in their production of the imitations. On the perceptual side, there are two possibilities: (1) Perceptual nonlinearities might arise when the stimuli are synthetic and/or not well matched to the subject's production capabilities.