From birth or after a few months onwards, infants are already capable of discriminating subtle differences in speech sounds (Jusczyk & Derrah, 1987; Kuhl, 1983) and of overriding the acoustic variability produced by changes in talkers voice, speaking rate and coarticulation (Eimas & Miller, 1980; Hillenbrand, 1984; Mehler et al., 1988) to categorize speech sounds. To explore what neural representation supports these early perceptive abilities for speech, we measured the brain responses of 3-month-old infants (N=30) listening to a set of natural syllables, using a 128-channel EEG system. The syllables were either a consonant followed by a vowel or vice-versa and varied along two orthogonal consonantal phonetic features -voicing and place of articulation- and two vocalic phonetic features – height and backness.

Using multivariate pattern analysis (MVPA), we extend previous results by showing that preverbal infants encoded all phonetic features, no matter how and by whom they were pronounced. More importantly, we demonstrate that these phonetic features are encoded independently from the place of the phoneme within a syllable, suggesting that alike adults, pre-verbal infants possess a position-invariant code for phonetic content. Next, we examine how this phonetic representation evolved across time and show that infants parallelly extracted and combined phonetic features, enabling them to identify phonemes. Finally, we also ran the same experiment in neonates (N=25) and compared the content and the dynamics of the phonetic encoding. Preliminary results suggest that already at birth, the brain encodes vocalic phonetic features independently from the context. Overall, this study sheds light on the neural representation behind infants’ early perceptual abilities concerning speech and contributes to a better understanding of what encoding mechanisms support rapid language acquisition in the first months of life.