Familiarity and Voice Representation: From Acoustic-Based Representation to Voice Averages

Abstract

The ability to recognize an individual from their voice is a widespread ability with a long evolutionary history. Yet, the perceptual representation of familiar voices is ill-defined. In two experiments, we explored the neuropsychological processes involved in the perception of voice identity. We specifically explored the hypothesis that familiar voices (trained-to-familiar (Experiment 1), and famous voices (Experiment 2)) are represented as a whole complex pattern, well approximated by the average of multiple utterances produced by a single speaker. In experiment 1, participants learned three voices over several sessions, and performed a three-alternative forced-choice identification task on original voice samples and several “speaker averages,” created by morphing across varying numbers of different vowels (e.g., [a] and [i]) produced by the same speaker. In experiment 2, the same participants performed the same task on voice samples produced by familiar speakers. The two experiments showed that for famous voices, but not for trained-to-familiar voices, identification performance increased and response times decreased as a function of the number of utterances in the averages. This study sheds light on the perceptual representation of familiar voices, and demonstrates the power of average in recognizing familiar voices. The speaker average captures the unique characteristics of a speaker, and thus retains the information essential for recognition; it acts as a prototype of the speaker.

Highly accurate and robust identity perception from personally familiar voices

“…In contrast, mutual confusions of the newly-learned and unfamiliar voices were frequent, and increased with the degree of acoustic manipulation. We interpret this finding in line with our prediction that listeners may be more attuned to their romantic partner's voice and the dynamics of their vocal system, owing to a more robust representation (Lavan et al, 2016;Fontaine, Love, & Latinus, 2017). Hence, if voice acoustics are altered sufficiently, listeners hearing an acoustically deviant version of a personally-familiar voice may be less willing to accept these voice tokens because they are not compatible with stored representations.…”

“…Our data also suggest that brief exposure to a voice identity through training may be sufficient to distinguish a voice from other identities, but this ability appears vulnerable to interference from other voices (see Figure 2). Labtrained identities may therefore be associated with less robust stored representations, reducing the degree to which a listener can recognise a speaker with limited cues to vocal identity (Fontaine, Love, & Latinus, 2017). In contrast, for personally-familiar voices, for which a robust representation has been formed through extensive and varied exposure, we observe highly accurate identity perception.…”

“…Despite a general consensus that familiarity is advantageous for voice identity perception, the magnitude and nature of the familiarity advantages appear to vary. For example, Fontaine, Love, and Latinus (2017), found entirely different profiles of results in a voice identity perception task for lab-trained versus celebrity voices -in their study, listeners showed better voice identification performance for sounds created by averaging many exemplars of a voice, but this was only seen for famous voices. Such findings give a first indication that the type and degree of familiarity is likely to have an impact on how well listeners can perceive identity from voices.…”

Obligatory and facultative brain regions for voice-identity recognition

“…We speculate that highly familiar voices can be readily recognized as the incoming voice is compared to a previously established voice-reference pattern. In contrast, for newly-learned unfamiliar voices such a voice-reference pattern may not yet be established ( Fontaine et al , 2017 ). We suppose that right posterior/mid temporal lobe structures may support the acoustical voice-identity feature analysis that is necessary to establish a new voice-specific reference pattern and/or may be critical for recognition when such a reference pattern is not yet available.…”

Cognitive and Psychophysical Determinants of Voice Recognition

“…Another theory, not exclusive to the prototype model, is that voice patterns are stored as an average of all heard instances of a voice. To test this, Fontaine et al (2017) hypothesized that subjects would be better at recognizing voice averages, i.e. combined vocal morphs of varying numbers of vowel sounds, than singular vowels.…”