Stop consonant perception: A comparison of several perceptual theories

Abstract

Acoustic cues to the identity of a plosive differ depending on aspects of phonetic environment such as features of a following vowel or consonant, properties of the previous segment, and stress. A complete perceptual theory should not only describe the perceptual strategy of an adult when confronted with this variability (and variability seen across talkers, transmission mediums, and speaking rates), but also account for the acquisition of these skills. In this paper we review the relative merits of burst invariance theories, relational invariance theories, the motor theory, analysis by synthesis, the locus theory, a theory based on the equivalent second formant F2′, and two new models called SCRIBER and LAFS [D. H. Klatt, J. Phonetics 7, (2) (1979)]. The advantage of SCRIBER and LAFS is that they can account for the differential use of various cues in different phonetic contexts by a conceptually simple decoding network structure. In its simplest form, the network represents all possible phonetic transitions of a language by distinct sequences of spectra. Perceptual abstractions such as [p]-ness or labialness are learned through association with the phonemic representations for words that are developed for speech production purposes (with the aid of a certain amount of natural acoustic similarity that exists for different realizations of a stop or of a distinctive feature). [Supported by NINCDS.]