ABSTRACT

Two central problems for a theory of speech perception are those of segmentation and invariance. The segmentation problem is to partition the acoustic signal into the phone-sized segments reported by phoneticians and (literate) listeners. The invariance problem—in the aspect that interests us here—is to explain why acoustically distinct, apparently context-sensitive, versions of a phonetic segment may sound free of contextual influences to listeners. We call this the problem of "perceptual invariance."