Visual gender biases in English stop voicing perception

Abstract

Listeners leverage visual information, such as perceived speaker gender, during speech perception. While work has shown visual gender biases on fricative (Johnson & Strand, 1996) and vowel perception (Johnson et al., 1999), its effects on stop voicing perception are understudied. We present an identification task where visual gender primes (male/female portraits) were used to investigate perceived speaker gender effects on English stop voicing. Subjects (n = 22) identified ‘pa’/’ba’ syllables from an 11-step VOT continuum (0–50 ms) in 5 pitch levels (high: 250 Hz, 230 Hz, mid: 170 Hz, and low: 130 Hz, 110 Hz). High and low pitch tokens were always preceded by a female and male portrait, respectively. Mid pitch tokens were preceded by female or male images in separate blocks. Given that f0 is higher following voiceless obstruents (Lehiste & Peterson 1961; Mohr 1971), if listeners utilize visual gender cues, acoustically ambiguous tokens paired with a male portrait should elicit more voiceless "pa" responses than those paired with a female portrait. Preliminary results pattern in the expected direction; mid tokens are perceived as "pa" more when paired with a male (15–25 ms = 79% "pa" response) than a female face (73% "pa" response); however, this effect was not statistically significant.