Abstract
Gender-affirming voice care is organized around speech, yet many people also want to change how they sing. Whether the two rest on one concept of vocal gender is unknown. Participants (N = 2,536) described what makes a speaking voice and a singing voice sound feminine and masculine, rated their current and ideal voices in each modality, and named people whose speaking and singing voices they would like to have. I embedded the descriptions with a sentence encoder and analyzed them, together with word frequencies, in exploratory analyses. The feminine–masculine direction in speaking descriptions was nearly identical to the direction in singing descriptions (cosine = .967), and a classifier trained on one modality separated the other with almost no loss (area under the curve = .960 and .979). The shared direction was carried by register, softness, and depth words; singing descriptions added range, notes, and power, and speaking descriptions added softness, roughness, and monotony. Individuals’ own contrasts aligned only modestly across modalities (mean cosine = .362). Wanting a more feminine speaking voice predicted choosing a female singer but not a more acoustically feminine singer within gender category. Vocal gender appears to be a shared, category-level abstraction expressed in modality-specific vocabulary.