Abstract:
Listeners can readily perceive both the linguistic message and the identity of the talker from an utterance that has been replicated with a few time-varying sinusoids. One puzzling outcome of studies of the identification of sine-wave talkers is that erroneous identifications did not cluster by sex. Males were often mistaken for females, and vice versa. A series of new experiments used frequency transpositions of sinusoidal replicas of speech to determine the acoustic attributes responsible for the identification of a talker's sex and a talker's identity. The central spectral tendency of a sinusoidal sentence was found to affect the perception of the sex of the talker; a sine-wave pattern derived from the formant frequencies of an individual talker seemed male, female, or neither when transposed to match the male, female, or overall average formant frequencies in our talker set, respectively. However, performance on a test of individual recognition was very good under acoustic conditions that reduced sex determination to chance levels. This suggests that recognizing a sine-wave talker does not depend on prior identification of sex, and begins to explain why listeners are likely to confuse sine-wave talkers without regard for the talker's sex. [Work supported by NIDCD.]