The vowel-sequence illusion: intrasubject stability and intersubject agreement of syllabic forms

Earlier studies have found that listeners presented with a loud and clear repeating sequence of brief steady-state vowels typically report hearing two voices with distinctly different timbres repeating different syllables that either are English words or occur in English words. One of the simultaneous voices is generally based upon frequencies below, and the other above, the "crossover frequency" at approximately 1500 Hz that divides normal speech into regions contributing equally to intelligibility. It has been hypothesized that the lack of linguistic content halts the processing of vowel sequences at the syllabic level, and that the spectral splitting corresponding to the concurrent voices reflects a mechanism for independent processing of different frequency regions that can lead to increased intelligibility under difficult listening conditions. The present study employed twelve randomly selected arrangements of the same six 70-ms vowels, and it was determined that: (1) individuals reported the same perceptual organizations the following week; (2) insertion of a brief silent gap between restatements of a sequence resulted in reports of similar (and occasionally identical) syllables by different listeners hearing the same sequence; and (3) when two listeners' responses differed, they could nevertheless identify the particular vowel sequences corresponding to each other's verbal forms. Spectrograms of vowel sequences were compared with time-aligned spectrograms of a speaker's synchronous production of the forms as they were being heard, and some common features of the acoustic patterns were noted. It is suggested that vowel sequences provide a reliable and useful tool for probing aspects of the perceptual organization of speech sounds that are normally obscured by additional linguistic processing.