Von Holzen, K. 1 , Nishibayashi, L. 2 & Nazzi, T. 1
1 Laboratoire Psychologie de la Perception, CNRS and Université Paris Descartes
2 University of Ottawa
Using ERPs, we examined the neural basis of word segmentation together with the segmental detail with which 8-month-old infants process newly segmented word forms. In each block, infants (n = 32) were first presented with eight sentences, each containing the same (familiarized) CVC word (e.g. /kav/). This was followed by presentation of isolated words: the familiarized word, a 1-feature mispronunciation (e.g. /gav/), and a phonologically unrelated control word (e.g. /r?g/). Mispronunciations were consonant initial, vowel medial, or consonant final. If newly segmented words are phonologically detailed, ERP responses for mispronounced words should pattern with control words; if they are not phonologically detailed, then mispronounced words should pattern with familiarized words.
Overall, ERP responses in the anterior region to the target and mispronounced words were more negative than control words, revealing evidence of segmentation but not sensitivity to mispronunciations. A linear regression analysis of polarity of response to target and control words (negative vs. positive responders; Kooijman et al., 2013) revealed that when accounting for variability in infants? familiarity response, infants were sensitive to mispronunciations and this did not vary by type of mispronunciation. Furthermore, infants with more negative responses produced more words at 16 months, suggesting a developmental advantage.