Listeners form average-based representations of individual voice identities

Abstract

Models of voice perception propose that identities are encoded relative to an abstracted average or prototype. While there is some evidence for norm-based coding when learning to discriminate different voices, little is known about how the representation of an individual's voice identity is formed through variable exposure to that voice. In two experiments, we show evidence that participants form abstracted representations of individual voice identities based on averages, despite having never been exposed to these averages during learning. We created 3 perceptually distinct voice identities, fully controlling their within-person variability. Listeners first learned to recognise these identities based on ring-shaped distributions located around the perimeter of within-person voice spaces – crucially, these distributions were missing their centres. At test, listeners’ accuracy for old/new judgements was higher for stimuli located on an untrained distribution nested around the centre of each ring-shaped distribution compared to stimuli on the trained ring-shaped distribution.

Are super‐face‐recognisers also super‐voice‐recognisers? Evidence from cross‐modal identification tasks

“…which may also explain the conservative response biases 2 . Nevertheless, as previously discussed, and despite potential floor effects, SF‐Identifiers still outperformed controls, with proportionally more achieving super‐voice‐recogniser status on this test, suggesting that they may develop stronger internal representations for faces (e.g., Bruce & Young, 1986; Valentine, 1991) and voices (e.g., Kreiman & Sidtis, 2011; Lavan, Knight, & McGettigan, 2019).…”

Deep neural network models reveal interplay of peripheral coding and stimulus statistics in pitch perception

“…Pitch is also believed to be used in the service of more complicated auditory tasks, such as voice recognition (Latinus and Belin, 2011; McPherson and McDermott, 2018; Lavan et al, 2019) or the perception of stress in spoken language (Shattuck-Hufnagel and Turk, 1996; Cutler et al, 1997). Moreover, we often must estimate the F0 of a sound amid other sounds that have their own F0s.…”

How does familiarity with a voice affect trait judgements?

“…Listeners were asked to listen carefully and try to memorize the voice and the name (cf. Lavan, Knight, Hazan & McGettigan, 2019; Lavan, Knight & McGettigan, 2019, for studies using a similar training paradigm). For the valenced training groups, an additional vignette was presented as text alongside the name on the screen during playback (e.g., ‘She helped an elderly man cross the road’ and ‘She is very patient when dealing with other people’s problems’ for positive valence; and ‘She is lying about her age in her dating profile’ and ‘She didn't apologise even though she knew she was in the wrong’ for negative valence).…”

“…Strikingly, the effects of within‐person variability on voice identity perception were much reduced when listeners were familiar with the voices: Familiar listeners were able to accurately perceive recordings of the same person as a single identity, despite the substantial within‐person variability. These differences in behaviour between the two groups have been ascribed to listeners having access to stable and robust representations of familiar voices, which enables them to link variable stimuli back to a single identity (e.g., Burton, Kramer, Ritchie & Jenkins, 2016; Lavan, Burton et al ., 2019; Lavan, Knight & McGettigan, 2019, for faces).…”