Quantifying Qualitative Listener Assessments of Gender Ambiguous Speakers: Coding Lexical Data to Measure Social Distance
Résumé
Previous research suggests that speakers perceived to be non-gender prototypical in terms of the masculinity/femininity of their voices are also unlikely to be perceived to be straight (Munson, McDonald, DeBoe, & White, 2006). Less well researched are the social effects of sounding gender ambiguous (GA). One avenue of research is to determine the social distance – the degree of affiliation between two or more interlocutors – between listeners and the speakers they assess. Social distance can be framed in various ways, however, most views tend to require spoken and/or discourse level data as well as context. Our research question makes use of mixed methods: How can we quantify isolated lexical descriptions of voices in meaningful ways that capture social distance distinctions? Our prediction is that there will be greater social distance between listeners and GA voices vs. non-gender ambiguous voices.
A dataset of words, developed using experimental methods resulted in over 700 unique words to describe the nine speakers. A variety of coding protocols endeavouring to capture social distance were developed for the words: Semantic Type (Dixon, 1977), Polarity, Syntactic Type (Heylighen & Dewaele, 2002), and Syllable Length (Biber, Conrad & Reppen, 1998). Preliminary statistical analyses of main effects suggest that adjectives and negative and neutral words are more probable when the speaker is GA. Words describing human attributes (e.g., proud), making evaluations (e.g., irritating) or capturing voice quality (e.g., vocal fry) are also more likely to be used in describing GA voices. This suggests that social distance can be quantitatively assessed.
Origine : Fichiers produits par l'(les) auteur(s)