AUDITORY STREAM SEGREGATION OF VOWEL SEQUENCE DEPENDING ON SIZE DIFFERENCE: INVESTIGATING THE COHERENCE BOUNDARY AND THE FISSION BOUNDARY
A typical stream segregation is demonstrated using the alternation of two frequencies of sinusoids, A and B, in the galloping pattern (ABA-ABA-...). The purpose of the current study was to investigate whether segregation could occur based on the size difference imposed on each vowel in a sequence. The size difference was applied by dilating or shrinking the spectral response of each vocal tract characteristic without changing the fundamental frequencies. Five Japanese vowels, "a, e, i, o, u," were used as the base signals. For each segment of a vowel sequence, the vowel categories were selected randomly from the five types with the restriction that the same category as the two previous should not appear. Two types of boundaries were obtained on the plane of the size difference vs. the alternation speed: one is the coherent boundary; the other is the fission boundary. The results indicate that size-based segregation can occur although it is not as compelling as F0-based segregation.