How many voices did you hear? Natural variability disrupts identity perception in unfamiliar listeners.

Abstract

Within-person variability is a striking feature of human voices: our voices sound different depending on the context (laughing vs. talking to a child vs. giving a speech). When perceiving speaker identities, listeners therefore need to not only "tell people apart" (perceiving exemplars from two different speakers as separate identities) but also "tell people together" (perceiving different exemplars from the same speaker as a single identity). In the current study, we investigated how such natural within-person variability affects voice identity perception. Using voices from a popular TV show, listeners, who were either familiar or unfamiliar with the show, sorted naturally-varying voice clips from 2 speakers into clusters to represent perceived identities. Across three independent participant samples, unfamiliar listeners perceived more identities than familiar listeners and frequently mistook exemplars from the same speaker to be different identities. These findings point towards a selective failure in "telling people together". Our study highlights within-person variability as a key feature of voices that has striking effects on (unfamiliar) voice identity perception. Our findings not only open up a new line of enquiry in the field of voice perception but also call for a re-evaluation of theoretical models to account for natural variability during identity perception.

Sorting through the impact of familiarity when processing vocal identity: Results from a voice sorting task

“…First, familiar listeners were significantly more able than unfamiliar listeners to tell different instances of the same speaker together, as indicated by both the number of resultant identity clusters following sorting, and the withinidentity clustering scores. This outcome met with expectations following the use of the sorting task with faces (Andrews et al, 2015;Jenkins et al, 2011) and the use of the sorting task with voices (Lavan et al, 2018;. Furthermore, it cements the conclusion that familiarity with an identity helps the perceiver to map different instances of that identity together despite inherent variability between one instance and the next.…”

“…Comparison by means of Mann-Whitney U test indicated that the two groups differed significantly in the number of intrusion errors (U = 5.28, p < .001), suggesting that familiarity with the speakers improved the ability to tell voices apart rather than confuse them into mixed-identity clusters. This pattern of errors contrasted with that in the face sorting task (Andrews et al, 2015;Jenkins et al, 2011) and the voice sorting task (Lavan et al, 2018; where the number of intrusion errors was very low for both listener groups. This perhaps reflected the relative difficulty of the face discrimination and voice discrimination tasks per se, but is a point that is considered further in the Discussion.…”

“…When the same task was applied in the voice domain, the results were remarkably consistent to those obtained with faces. Lavan et al (2018) used three pairs of female voices from the TV series "Orange is the New Black." Familiarity was manipulated by recruiting participants who had either watched the TV series or who had not.…”

Flexible voices: identity perception from variable vocal signals

“…In these speaker-discrimination studies, within-person variability therefore consistently affects listeners' performance: when unfamiliar with a voice, listeners can only generalize to a limited extent across differences between two vocal signals to accurately extract a stable percept of speaker identity. With no robust perceptual representation of the voice (also referred to as familiar voice patterns; Kreiman & Sidtis, 2011; Sidtis & Kreiman, 2012) being available for unfamiliar voices, within-person variability is therefore likely to be mistaken for between-person variability (see Lavan et al, 2018).…”

No clear advantage for high variability training during voice identity learning

“…In the absence of a person-specific representation of a voice, unfamiliar listeners perceive multiple variable examples of a single person's voice as having been produced by multiple speakers, thus failing to "tell people together". Familiar listeners, on the other hand, tend to outperform unfamiliar listeners in identity perception tasks: they can access person-specific representations, enabling them to perceive within-person variability appropriately and to thus succeed in "telling people together" (Lavan, Burston & Garrido, 2018b;Lavan, Burston, Ladwa, Merriman, Knight & McGettigan, 2018c;Lavan, Merriman, Ladwa, Burston, Knight & McGettigan, 2018d).…”