Part of: Queer Voices
Abstract:
This study examines how gender presentation (expression of gender through behavior and appearance) and sexuality function as predictors of variation in vowel production among 45 assigned female at birth US English speakers from New England. This work builds on the minimal literature exploring correlations between sexuality and vowel production and further explores gender presentation as a predictor of variation (as proposed by Sulkin (2025)) through a large-scale spontaneous speech corpus. We find that queer participants have significantly higher vowels compared to straight participants and that sexuality and gender presentation interact in their correlations with vowel dispersion.
Previous literature links vowel production to gender and sexuality. Lowered formants (F1, F2) and contracted vowel spaces are associated with masculinity and higher formants and expanded vowel spaces are associated with femininity (e.g. Neel (2008), Simpson (2001)). Additionally, Pierrehumbert et al. (2004) and Munson et al. (2006) both found that, in the Chicago and St. Paul/Minneapolis metropolitan areas, respectively, queer women produced lower average F1 and F2 values for certain vowels compared to straight women; Pierrehumbert et al. found that straight women had more contracted vowel spaces than queer women, while Munson et al. found the opposite effect. In Canadian English, Rendall et al. (2008) found that queer women had lower formants and more contracted vowel spaces than straight women. More recently, however, a large-scale study (Holmes et al., 2024) on British English found no significant differences in vowel quality with respect to sexuality in structured spontaneous speech interviews, although they examined tokens from only three words (bed, cat and go).
The present data comes from sociolinguistic interviews conducted with 15 lesbian, 15 bisexual and 15 straight participants from New England. There were no significant differences in height between these three groups, and none reported regular smoking habits or any testosterone use. Each speaker was also asked to rate themselves on a gender presentation Likert scale, ranging from 1 (=hyper-feminine) to 5 (=hyper-masculine). The interlocutor was a feminine-presenting cisgender lesbian from New England. Recordings were made as 48kHz, 24bit .wav files using a Zoom H4n Pro with an AKG C520 condenser microphone. Segmentation was done using the Montreal Forced Aligner (McAuliffe et al., 2017) and manually corrected. F1 and F2 values were collected at the midpoint of each vowel in Praat, resulting in 2000 stressed vowel tokens per participant of the vowels [i, ɪ, e, ɛ, ʌ, æ, ɑ, o, u, ʊ]. Vowel space dispersion was calculated by finding the average Euclidean distance between each vowel and the calculated center of each speaker’s F1/F2 space.
We find that lesbian and bisexual speakers have significantly lower F1 and F2 compared to straight speakers (but do not differ significantly from each other). Additionally, feminine-presenting bisexual and straight speakers have significantly more dispersed vowel spaces than masculine-presenting bisexual and straight speakers, while feminine-presenting lesbians have significantly less dispersed vowel spaces compared to their masculine-presenting counterparts. These results partially support the results of previous studies and underscore the importance of large-scale analysis for understudied communities.
