Empirical Foundations of Socio-Indexical Structure: Inquiries in Corpus Sociophonetics and Perceptual Learning
Loading...
Date
2024-01-09
Authors
Gunter, Kaylynn
Journal Title
Journal ISSN
Volume Title
Publisher
University of Oregon
Abstract
Speech is highly variable and systematic, governed by the internal linguistic system and socio-indexical factors. The systematic relationship of socio-indexical factors and variable phonetic forms, referred to here as socio-indexical structure, has been the cornerstone of sociophonetic research over the last several decades. Research has provided mounting evidence that listeners track and exploit cross-talker variability during speech processing tasks. As one such example, recent work has demonstrated listeners’ sensitivity to talker characteristics via retuning phonetic categories (i.e., perceptual learning) in response to talker-specific patterns. Drawing on Bayesian models, researchers have argued that listeners’ perceptual learning is influenced by listeners’ prior experience with socio-indexical factors conditioning segmental variation. From experience listeners form beliefs about the underlying cause of variation to determine when to adapt to talker-specific forms and generalize to other similar talkers. However, theoretical work has over-simplified descriptions of socio-indexical structure, leaving open questions about the nature and range of phonetic variation that listeners track and exploit.This dissertation seeks to provide both theoretical and empirical foundations of socio-indexical structure at the intersection of individual talkers and geographic dialects drawing on mixed methods. Using large-scale datasets of American English vowel measurements, the corpus analyses probe different quantitative descriptions of socio-indexical structure under various scopes of socio-indexical granularity and internal organizations across the vowel space. The corpus analyses reveal an asymmetry in socio-indexical conditioning of the joint cue distributions (i.e., F1xF2) across several simulations whereby some categories (e.g., /eɪ/) are conditioned by dialect, while others are conditioned by individual talker identity alone (e.g., /ʊ/; Chapter 4). Additionally, analyses show that individual talkers diverge from their dialect areas less for dialect conditioned vowels compared to talker conditioned vowels, confirming talkers’ distributional patterns generally align with their communities. Additional analyses highlight how internal principles provide specificity to socio-indexical conditioning of variability, focusing on the acoustic overlap of vowel pairs and individual cue dimensions (Chapter 5). Such descriptions suggest acoustic overlap across some vowel pairs may be attenuated by socio-indexical information while other vowel pairs generally demonstrate stability across talkers and dialects (e.g., /æ/ and /a/). Finally, descriptions of individual cue dimensions demonstrate multimodal distributions both across and within talkers for some categories conditioned by dialects (e.g., /ɔ/; Chapter 5).
Following from Bayesian models of speech processing and causal inference, this dissertation tests whether a priori links to socio-indexical structure influence perceptual learning (Chapter 6). A lexically guided perceptual learning experiment tests whether the asymmetry of socio-indexical conditioning (dialect vs. talker) observed in the corpus analyses correlates with listeners’ learning and generalization behavior after exposure to novel shifts in one of two vowels (/eɪ/ and /ʊ/) in a female speaker’s voice. The results demonstrate learning a novel shift in /ʊ/ but not in /eɪ/, with generalization of post-test categorization to a novel male talker but not a novel female talker. These results suggest that the asymmetry of social conditioning alone may guide listeners’ behavior for these vowels and challenge our current understanding of listeners’ adaptation to vocalic variability and the role of socio-indexical structure in perceptual learning. Overall, this dissertation advances our understanding of socially conditioned variation across speech production and perception.
Description
Keywords
corpus phonetics, perceptual learning, regional dialects, sociophonetics, speech perception, vowels