This sounds important: Electrophysiological responses reveal a dedicated learning mechanism to process salient...

Isolating relevant sounds in the auditory stream is a crucial feat accomplished by human infants and a pivotal ability for language acquisition. It is therefore reasonable to postulate the existence of early mechanisms that reorient attention toward salient acoustic stimuli. Previous studies suggest that infants tend to allocate more attention to consonant sounds because they resemble human vocalizations. However, systematic evidence on the neural processes underlying consonance tuning in newborns is still scarce. Here, we investigate newborns' learning of auditory regularities by computing mismatch responses (MMRs) and the trial-by-trial correlation between the neural signal and Bayesian surprise (a theoretical measure of learning). We present 22 healthy newborns (40.4 ± 15.8 h) with a roving paradigm (a pseudorandom sequence of deviant and standard events), composed of either consonant or dissonant sounds, while we record their electroencephalogram. Our results show that 1) newborns exhibit a neural encoding of auditory regularities for all sound types (consonant and dissonant), as demonstrated by the presence of MMRs and significant correlation of the neural signal with Bayesian surprise; 2) consonant and dissonant sounds elicited MMRs of opposite polarities, with consonant auditory stimulation evoking negative responses, reminiscent of adult-like MMRs. Overall, our findings suggest that newborns display a dedicated perceptual learning mechanism for consonance. We speculate that this mechanism might represent an evolutionarily acquired process to detect and learn novel auditory stimuli with acoustic features resembling human vocalizations.