Intelligibility benefit for familiar voices does not depend on better discrimination of fundamental frequency or vocal tract length

Speech is more intelligible when it is spoken by familiar than unfamiliar people. Two cues to voice identity are glottal pulse rate (GPR) and vocal tract length (VTL): perhaps these features are more accurately represented for familiar voices in a listener’s brain. If so, listeners should be able to discriminate smaller manipulations to perceptual correlates of these vocal parameters for familiar than unfamiliar voices. We recruited pairs of friends who had known each other for 0.5–22.5 years. We measured thresholds for discriminating pitch (correlate of GPR) and formant spacing (correlate of VTL; ‘VTL-timbre’) for voices that were familiar (friends) and unfamiliar (friends of other participants). When a competing talker was present, speech was substantially more intelligible when it was spoken in a familiar voice. Discrimination thresholds were not systematically smaller for familiar compared to unfamiliar talkers. Although, participants detected smaller deviations to VTL-timbre than pitch uniquely for familiar talkers, suggesting a different balance of characteristics contribute to discrimination of familiar and unfamiliar voices. Across participants, we found no relationship between the size of the intelligibility benefit for a familiar over an unfamiliar voice and the difference in discrimination thresholds for the same voices. Also, the intelligibility benefit was not affected by the acoustic manipulations we imposed on voices to assess discrimination thresholds. Overall, these results provide no evidence that two important cues to voice identity—pitch and VTL-timbre—are more accurately represented when voices are familiar, or are necessarily responsible for the large intelligibility benefit derived from familiar voices.

Highly accurate and robust identity perception from personally familiar voices

Previous research suggests that familiarity with a voice can afford benefits for voice and speech perception. However, even familiar voice perception has been reported to be error-prone in previous research, especially in the face of challenges such as reduced verbal cues and acoustic distortions. It has been hypothesised that such findings may arise due to listeners not being “familiar enough” with the voices used in laboratory studies, and thus being inexperienced with their full vocal repertoire. By extension, voice perception based on highly familiar voices – acquired via substantial, naturalistic experience – should therefore be more robust than voice perception from less familiar voices. We investigated this proposal by contrasting voice perception of personally-familiar voices (participants’ romantic partners) versus lab-trained voices in challenging experimental tasks. Specifically, we tested how differences in familiarity may affect voice identity perception from non-verbal vocalisations and acoustically-modulated speech. Large benefits for the personally-familiar voice over less familiar, lab-trained voice were found for identity recognition, with listeners displaying both highly accurate yet more conservative recognition of personally familiar voices. However, no familiar-voice benefits were found for speech comprehension against background noise. Our findings suggest that listeners have fine-tuned representations of highly familiar voices that result in more robust and accurate voice recognition despite challenging listening contexts, yet these advantages may not always extend to speech perception. Our study therefore highlights that familiarity is indeed a continuum, with identity perception for personally-familiar voices being highly accurate.

Speech-evoked brain activity is more robust to competing speech when it is spoken by someone familiar

People are much better at understanding speech when it is spoken by a familiar talker—such as a friend or partner—than when the interlocutor is unfamiliar. This provides an opportunity to examine the substrates of intelligibility and familiarity, independent of acoustics. Is the familiarity effect evident as early as primary auditory cortex, or only at later processing stages? Here, we presented sentences spoken by naturally familiar talkers (the participant’s friend or partner) and unfamiliar talkers (the friends or partners of other participants). We compared multivariate activity in speech-sensitive regions of cortex between conditions in which target sentences were presented alone and conditions in which the same target sentences were presented at the same time as a competing sentence. Using representational similarity analysis (RSA), we demonstrate that the pattern of activity evoked by a spoken sentence is less degraded by the presence of a competing sentence when it is spoken by a friend or partner than by someone unfamiliar; the results cannot be explained by acoustic differences since familiar and unfamiliar talkers were nearly identical across the group. This familiar-voice advantage is most prominent in nonprimary auditory cortical areas, along the posterior superior and middle temporal gyri. Across participants, the magnitude of the familiar-unfamiliar RSA difference correlates with the familiar-voice benefit to intelligibility. Overall, our results demonstrate that experience-driven improvements in intelligibility are associated with enhanced patterns of neural activity in nonprimary auditory cortical areas.Significance statementSpeech is a complex signal, and we do not yet fully understand how the content of a spoken sentence is encoded in cortex. Here, we used a novel approach based on analysing multivariate activity: we compared activity evoked by highly intelligible sentences presented alone and by the same sentences presented with a competing masker. The distributed pattern of activity in speech-sensitive regions of the brain was more similar between the alone and masker conditions when the target sentence was spoken by someone familiar—the participant’s friend or partner—than someone unfamiliar. This metric correlated with the intelligibility of the familiar voice. These results imply that the spatial pattern of activity in speech-sensitive regions reflects the intelligibility of a spoken sentence.

Toward a unified account of person perception from familiar and unfamiliar voices

When we hear a voice, we instantly form rich impressions of the person it belongs to – whether we are familiar with this voice or whether we are hearing it for the first time. Despite the rich impressions we can form of both familiar and unfamiliar voices, current models of voice processing primarily focus on familiar voice identity perception only and do not explicitly account for the processing of unfamiliar voices. Where unfamiliar identity processing is described, it tends to be in the context of specific identity perception tasks, such that the extant literature is largely built on a distinction between familiar voice recognition and unfamiliar voice discrimination. We argue that the current focus of the literature is too narrow in its strong emphasis on identity-specific perception, and does not adequately reflect person perception from voices beyond experimental tasks. Here, we propose a broader, unified account of person perception from both familiar and unfamiliar voices. We suggest that listeners routinely perceive all person characteristics from voices via common recognition processes, based on representations – be those of a specific identity, speaker sex, accent, or a perceived personality trait. While explicit discrimination processes may still be used to disambiguate percepts, they are likely to play a smaller role in perception in naturalistic settings. We offer discussions of how this representation-centred person perception from voices may work, in terms of the nature of representations, their specificity and interactions of different kinds of representation.

Using spatial release from masking to estimate the magnitude of the familiar-voice intelligibility benefit

The ability to segregate simultaneous speech streams is crucial for successful communication. Recent studies have demonstrated that participants can report 10–20% more words spoken by naturally familiar (e.g., friends or spouses) than unfamiliar talkers in two-voice mixtures. This benefit is commensurate with one of the largest benefits to speech intelligibility currently known—that gained by spatially separating two talkers. However, because of differences in the methods of these previous studies, the relative benefits of spatial separation and voice familiarity are unclear. Here, we directly compared the familiar-voice benefit and spatial release from masking, and examined if and how these two cues interact with one another. We recorded talkers speaking sentences from a published closed-set “matrix” task and then presented listeners with three different sentences played simultaneously. Each target sentence was played at 0° azimuth, and two masker sentences were symmetrically separated about the target. On average, participants reported 10–30% more words correctly when the target sentence was spoken in a familiar than unfamiliar voice (collapsed over spatial separation conditions); we found that participants gain a similar benefit from a familiar target as when an unfamiliar voice is separated from two symmetrical maskers by approximately 15° azimuth.

The benefit to speech intelligibility of hearing a familiar voice

Previous experience with a voice can help listeners understand speech when a competing talker is present. Using the Coordinate-Response Measure (CRM) task (Bolia, 2000), Johnsrude et al. (2013) demonstrated that speech is more intelligible when either the target or competing (masking) talker is a long-term spouse than when both talkers are unfamiliar (termed ‘familiar-target’ and ‘familiar-masker’ benefits, respectively). To better understand how familiarity improves intelligibility, we measured the familiar-target and familiar-masker benefits in older and younger spouses using a more challenging matrix task, and compared the benefits listeners gain from spouses’ and friends’ voices. On each trial, participants heard two sentences from the Boston University Gerald (Kidd et al., 2008) corpus (“ ”) and reported words from the sentence beginning with a target name word. A familiar-masker benefit was not observed, but all groups showed a robust familiar-target benefit and its magnitude did not differ between spouses and friends. The familiar-target benefit was not influenced by relationship length (in the range of 0.5–52 years). Together, these results suggest that the familiar-target benefit can develop from various types of relationships and that it reaches a ceiling within several months of meeting a new friend or partner.