Evidence for syllable structure, stress and juncture from segmental durations

Abstract

This report examines the informational content of segmental durations with respect to three abstract suprasegmental aspects: stress, juncture and syllable structure. Tokens of the template C-V-Sibilant-Stop-V-C were embedded in 112 different sentences read by four male speakers of English. Average durations and standard deviations show that these data are comparable to those found in continuously read speech. Discriminant analysis was used to classify tokens into phonologically motivated categories on the basis of segmental durations. We find that lexical stress is predicted accurately if word boundary locations are already known. However, word boundaries cannot be reliably predicted from timing measures. A comparison of three different syllable models indicates that segmental durations convey more information about autosegmental and hierarchical syllable models than they do about traditional linear segmental models of the syllable.