Collocational patterns of disfluencies in native and non-native speech: Evidence from the Wildcat Corpus

Abstract

While disfluencies are generally considered a natural part of spontaneous speech, patterns of disfluencies in non-native (NN) speech could contribute to making that speech sound less proficient or less “fluent”. NN speech has been described as having more frequent and longer pauses, pauses at within-clause boundaries, shorter mean length of runs, and slower speech rate, compared to native (N) speech (e.g., Riggenbach, 1991; Trofimovich & Baker, 2006). However, in order to understand what makes NN disfluency patterns unique, it is important to examine such phenomena locally by examining the contexts of different types of disfluency. In the present study, we describe what features collocate with disfluencies in spontaneous English dialog. We examine speech from the Wildcat Corpus (Van Engen et al., 2010) which includes English dialogue produced by various pairings of N and NN speakers (N-N, N-NN, or NN-NN). Specifically, we extract turn-internal disfluencies of several types (e.g., silent pause, filled pause, repetition), and examine lexical (e.g., part-of-speech) and phonetic (e.g., vowel length) features of speech surrounding these disfluencies. Further, we examine whether these disfluency patterns change depending on speaker pairings. Differences between N and NN disfluency patterns are discussed in terms of their potential sources (e.g., speaker- vs. listener-oriented).