Abstract
Introduction. In classical vocal competitions, judges reduce a performance to a single overall score, yet what that score reflects is unclear. Research has separately identified the perceptual attributes experts use and the acoustic features of trained singers. Few studies link these to expert evaluation, and those that do have used playback of sound-treated recordings and rarely tested for sex differences. Yet classical singing is judged live in halls, and male and female voices differ acoustically. We asked which attributes and acoustic features are associated with the overall score in a concert hall, and whether they differ by sex.Methods. Twenty-two classical singers (10 males, 12 females) performed Caro mio ben a cappella in a concert hall. Five singer-instructors rated each performance on a 100-point overall score and 7-point scales for six vocal attributes. Singing power ratio (SPR), loudness units full scale (LUFS), vibrato rate (VR), and vibrato extent (VE) were computed from annotated vowel segments of close-microphone recordings. Linear mixed-effects models related the overall score to the ratings or acoustic features, with sex interactions and a random judge intercept.Results. Higher ratings of vibrato, resonance, and expressiveness were associated with higher overall scores, whereas timbre, diction, and intonation were not. Higher SPR, greater VE, and lower VR were associated with higher overall scores, whereas LUFS was not. Sex interacted with VE and VR but not with SPR or LUFS: the VE association was stronger in females, the VR association in males.Conclusion. Overall scores were associated with perceived vibrato, resonance, and expressiveness, and with spectral prominence at 2–4 kHz. Vibrato associations were sex-dependent: a faster rate accompanied lower scores in males, a greater extent higher scores in females. These findings give expert evaluation a measurable basis under realistic conditions and indicate that vibrato should be interpreted with reference to singer sex.