Vowel reproduction exhibits an intrinsic dynamic organization

Abstract

Speech perception and production are generally thought to be tuned by the speaker’s linguistic environment. If so, iterative reproduction of a vowel by a single speaker may amplify subtle biases introduced by the tuning. Four male monolingual speakers of American English were asked to reproduce 100 synthetic vowel-like stimuli, uniformly distributed in a two-formant acoustic space (F1, F2 varying, F0=120 Hz, F3=2500 Hz, duration=200 ms). Fifteen of the subject’s own productions were chosen for serial reproduction, as follows: Each original production was presented to the subject through headphones, and the subject immediately mimicked it. After a 0.5-s delay, the ‘‘mimic’’ was played back to the subject as the target for his next production. This second mimic was the target for the following production, and so on, for ten iterations. Subjects made systematic errors in reproducing their own vowel sounds, and sequences of serially reproduced vowels behaved differently in different parts of the vowel space. These results suggest that some vowel regions are preferred over other regions, indicating that the cognitive processes for perceiving and producing vowels have complex intrinsic dynamics. [Work supported by NIMH.]