Visual biases from cis and trans judges in evaluation of voices with pitch and timbre manipulations.

When judging the voice of a person, one tends to integrate how this person looks into their voice qualities. This phenomenon is not only misleading but also unfair to many individuals who do not fit voice-body expectations, particularly those who are transgender. Interestingly, trans individuals (as judges) appear less prone to this visual bias, but it is unclear how they remain so. They might conceivably rely on certain characteristics of the voice-typically its pitch or its timbre-to operate in a more unimodal manner, effectively weighing down the influence of visual cues. This hypothesis was explored in two online studies, where cisgender and transgender participants were presented with 18 different actors (on a continuum from bass to soprano) singing or speaking short sentences. While the visual content of the stimuli was unchanged, the audio content was manipulated in fundamental frequency (F0) or in vocal tract length (VTL) without inducing any desynchronization between audio and video. Every participant provided voice type ratings for (1) Audio-only (A) stimuli to test the effect of the manipulation, (2) Video-only (V) stimuli to confirm that the manipulation had no role, and (3) combined Audio-Visual (AV) stimuli to see how much visual cues "contaminated" the audio evaluation in each manipulation. The manipulations worked as expected, lowering A-ratings for lower F0 and longer VTL, while having no impact on V-ratings. The AV-ratings were systematically shifted from A-ratings in the direction of V-ratings, but trans participants were less prone to this bias in experiment 2 (with VTL manipulations), not in experiment 1 (with F0 manipulations). We conclude that the trans advantage may partly originate from a better internalization of timbre cues, while pitch cues have more of a universal impact.