The VOT category boundary in word-initial stops: Counter-evidence against rate normalization in English spontaneous speech

Abstract

Some languages, such as many varieties of English, use short-lag and long-lag VOT to distinguish word- and syllable- initial voiced vs. voiceless stop phonemes. According to a popular view, the optimal category boundary location between the two types of stops moves towards larger values as articulation rate becomes slower (and speech segments longer), and listeners accordingly shift the perceptual VOT category boundary. According to an alternative view, listeners need not shift the category boundary with a change in articulation rate, because the same VOT category boundary location remains optimal across articulation rates in normal speech, although a shift in optimal boundary location can be induced in the laboratory by instructing speakers to use artificially extreme articulation rates. In this paper we applied rate-independent VOT category boundaries to word-initial stop phonemes in spontaneous English speech data, and compared their effectiveness against that of Miller, Green and Reeves's (1986) rate-dependent VOT category boundary applied to laboratory speech. The classification accuracies of the two types of category boundaries were comparable, when factors other than articulation rate are controlled, suggesting that perceptual VOT category boundaries need not shift with a change in articulation rate under normal circumstances. For example, Optimal VOT category boundary locations for homorganic word-initial stops differed considerably depending on the following vowel, however, when boundary location was assumed to be affected by the relative frequency of voiced vs. voiceless categories in each vowel context.