Investigating the consonant‐vowel boundary. II. Perceptual contributions of glimpse windows.

Abstract

Discrete segmental units do not occur in fluent speech because of coarticulation. However, separation of the meaningful sounds in language into the categories of consonants and vowels is one of the most fundamental principles of how sound is structured. Our previous research has shown that for sentence presentations, vowels have a distinct perceptual advantage over consonants in determining sentence intelligibility. TIMIT sentences were used to investigate perceptual contributions of consonants and vowels across the consonant‐vowel (CV) boundary, by shifting the CV boundary by specific proportions of the vowel, such that consonant duration increased while vowel duration decreased. Glimpse windows are defined as the speech signal preserved between noise replacements. The perceptual effect of windows either locked to specific segmental information or placed randomly was examined. Results from glimpse windows locked to segmental information confirmed a 2‐to‐1 vowel advantage for intelligibility at the traditional CV boundary and suggest that vowel contributions remain robust against deletions of the signal. When glimpses were presented randomly, summed duration of glimpses predicted performance. However, performance remained lower than when glimpse windows of equivalent duration were locked to vowels. Specific segmental information appears to differentiate performance between consonant and vowel conditions. [Work supported by the NIH.]