Is some information perceptually primary?

Abstract

The existence of multiple correlates for phonological contrasts invites the hypothesis that one is perceptually primary and that the others enhance the primary correlate. (Stevens Acoustics Phonetics, MIT Press, 1998) particularly emphasizes the perceptual primacy of acoustic landmarks. Our experiments manipulate the quality or variability of the acoustic information in a landmark versus a neighboring nonlandmark to determine whether listeners rely on the landmark or the most reliable information source. Stimuli are synthetic stop‐vowel sequences where the stop’s place of articulation is conveyed by its burst spectrum—landmark—and the formant transitions to the following vowel—nonlandmark. Parameter values in the two intervals fall on opposite sides of, but close to, a category boundary and convey competing place information. Information is degraded in one interval by broadening the bandwidths of the spectral peaks or formants, or it is made more variable by varying the peaks frequencies or the following vowel quality. If landmark information is perceptually primary, then degrading or varying that information should shift place judgments more toward the category conveyed by the nonlandmark interval than vice versa. However, if the quality or consistency of the information matters, then manipulating either interval should produce comparable shifts in place judgments. [Work supported by NIH.]