More adaptation of speech by nonspeech

Three experiments tested Samuel and Newport's (1979) hypothesis that the perceptual system sorts its input on the basis of its spectral quality (periodic vs. aperiodic). In Experiment 1, repeated presentation of a shaped white-noise segment (aperiodic) produced a labeling shift on a /ja-za/ continuum (primarily aperiodic); two periodic adaptors produced no effect, supporting Samuel and Newport's hypothesis. The second experiment replicated these results and showed that the nonspeech adaptor produced almost as much adaptation as the test series' endpoint /za). In addition, using several mixtures of periodic and aperiodic adaptors indicated that the aperiodic component dominates adaptation effects for /ja-za/. A final experiment, using a similarity rating task, confirmed that subjects group /za/ with unvoiced fricatives rather than with other voiced consonants. The results thus indicate that the perceptual system is sensitive to whether the input is primarily periodic or aperiodic, regardless of whether it is speech or nonspeech.