Little is certain about the principles by which the grouping of simultaneous and successive acoustic elements occurs in a speech signal. Neither is there much understood about the details of the perceptual tolerance for huge variability in the physical acoustic elements on which speech perception is based. To expose the logic of acoustic coherence pertinent to both issues, our review examines a candidate explanation deriving from perceptual research on the sinewave replication of utterances. Three empirical cases are reviewed to show how the diverse assortment of acoustic elements in a speech signal cohere perceptually. The first uses sinewave replicas of utterances to reveal the action of a perceptual sensitivity to time-critical properties of spectral variation; the second identifies an effect of time-critical sensitivity using synthetic sentences; and the third converging test shows that this perceptual susceptibility to the time-varying structure of utterances is independent of the apparent naturalness of the signals. These findings show the dual action of time-varying properties of speech signals and the listener's perceptual susceptibility to them: to establish the perceptual coherence of a set of acoustic elements, and, no less, to sustain phonetic perception despite the novelty of acoustic elements composing the signal.
