"Independence of phonological features in speech perception"
Noah Silbert
Department of Linguistics, Department of Cognitive Science
Indiana University, Bloomington
Phonological features often correspond closely to pereptual
dimensions in speech perception, and evidence suggests that these
dimensions interact (i.e., that they are not independent). However,
this evidence is equivocal, and assumptions of independence are
widespread in models of word recognition. General Recognition Theory
(GRT) provides powerful tools for representing perceptual space and
for analyzing relationships between perceptual dimensions. An
introduction to GRT as a model of perceptual phonetics will be given,
and experimental applications of GRT to perception of two sets of
consonants will be presented. The results indicate that while
phonological dimensions capture the gross structure of perceptual
space for each set of consonants, deviations from a purely
feature-based structure (i.e., failures of independence) vary
systematically between the two sets. Implications for models of word
recognition will be discussed, and a number of extensions of the GRT
approach to perceptual phonetics will be proposed.
|
|
