Can mergers-in-progress be unmerged in speech accommodation?

Molly Babel,Graham Haber,Michael Mcauliffe

doi:10.3389/fpsyg.2013.00653

Abstract

This study examines spontaneous phonetic accommodation of a dialect with distinct categories by speakers who are in the process of merging those categories. We focus on the merger of the NEAR and SQUARE lexical sets in New Zealand English, presenting New Zealand participants with an unmerged speaker of Australian English. Mergers-in-progress are a uniquely interesting sound change as they showcase the asymmetry between speech perception and production. Yet, we examine mergers using spontaneous phonetic imitation, which is phenomenon that is necessarily a behavior where perceptual input influences speech production. Phonetic imitation is quantified by a perceptual measure and an acoustic calculation of mergedness using a Pillai-Bartlett trace. The results from both analyses indicate spontaneous phonetic imitation is moderated by extra-linguistic factors such as the valence of assigned conditions and social bias. We also find evidence for a decrease in the degree of mergedness in post-exposure productions. Taken together, our results suggest that under the appropriate conditions New Zealanders phonetically accommodate to Australian English and that in the process of speech imitation, mergers-in-progress can, but do not consistently, become less merged.

Highlights

IntroductionThis variability is conditioned in part by the multiple degrees of freedom involved in the highly complex act of producing speech: an individual will never produce a word exactly the same way twice, though perceptual constancy ensures that the perception of variable productions remains relatively constant
A mixed effects logistic regression model predicting the proportion of shadowed judgments selected by listeners as more similar-sounding to the model was fit with Block (Shadowed 1, Shadowed 2, Post-task), diphthong Category (NEAR, SQUARE), and Condition (Positive, Negative) as predictor variables
The most merged female participant starts out with an almost completely overlapping distribution in her baseline productions (Pillai score = 0.07), and she remains fully merged throughout her shadowed productions (Pillai scores = 0.11, 0.00) and Post-task productions (Pillai score = 0.07). While her lexical sets do not become less merged, one can see in Figure 6 that the absolute position of www.frontiersin.org the vowel cloud moves around the vowel space throughout the task; the distribution on the whole becomes more near-like in Shadowed 1, and moves to a higher F1 and lower F2 space in Shadowed 2 and the Post-task

Summary

Introduction

This variability is conditioned in part by the multiple degrees of freedom involved in the highly complex act of producing speech: an individual will never produce a word exactly the same way twice, though perceptual constancy ensures that the perception of variable productions remains relatively constant This within-speaker variability along with the physiological differences between speakers and the co-mingling of speakers from multiple dialect and language backgrounds in urban settings, attest to the great phonetic variability in spoken language. Classic papers like Peterson and Barney (1952) and Hillenbrand et al (1995) showcase the massive amount of overlap of phonetic categories seen when describing vowel systems even within a single speech variety Despite this significant overlap and constantly variable signal, speakers and listeners successfully map an utterance onto linguistically meaningful categories. Listeners’ task of making sense of the signal is made even more challenging by the social and indexical factors which condition variation within and across speakers (e.g., Labov, 1963)

Methods

Results

Discussion

Conclusion