Voice-Familiarity Benefit in Infant Phoneme Perception: Acoustic versus Exemplar-based Familiarity

Starting in utero, infants perceive voice information and preferentially attend to the maternal voice. In adults, speaker familiarity facilitates word and phoneme perception. The current study investigated whether a similar voice-familiarity benefit supports phoneme perception in early infancy, and whether this benefit is driven by acoustic familiarity with speaker voice, prior exposure to phoneme exemplars, or their combination. We tested 72 German-learning 3-month-olds, who were familiarized at home with two speakers. In the lab, infants received 5 minutes of training on a difficult native vowel contrast (/ɛ/–/ɪ/) from either one of the familiarized speakers or an unfamiliar speaker. Discrimination was tested with the electrophysiological mismatch response using three talkers: (1) the phoneme training speaker (familiarized or unfamiliar, depending on group), (2) the familiarized non-training speaker, and (3) a novel speaker (i.e., an unfamiliarized non-training speaker). This design dissociated acoustic familiarity (at-home familiarization) from phoneme-exemplar exposure (training tokens). Preregistered contrasts provided evidence against discrimination differences across test speakers when training was delivered by the unfamiliar speaker; this training might promote speaker-independent generalization, as predicted by abstractionist accounts of speech perception. By contrast, exploratory analyses revealed a voice-familiarity benefit in the group trained by the familiar speaker; this training may drive infants to rely more strongly on speaker-specific cues. Thus, the voice-familiarity benefit only arose when acoustic familiarity and exemplar-based familiarity were combined, suggesting that neither form alone was sufficient. Together, these results are consistent with hybrid accounts of speech perception in which exemplar-like and abstract representations coexist.