Mayor, j. 1 & Plunkett, K. 2
1 FPSE, University of Geneva, Switzerland
2 Dept. of Experimental Psychology, University of Oxford, UK
The TRACE model of speech perception (McClelland & Elman, 1986) is used to simulate a variety of findings from the recent infant word recognition literature, in an attempt to provide a unified, theoretical framework for interpreting these findings. For example, recent evidence suggests a symmetry in infant sensitivity to vowel and consonant mispronunciations of familiar words from early in the second year of life (Mani&Plunkett, 2007) whereas other researchers argue for a prominence of consonants in lexical processing and language acquisition. Our results, using TRACE, support the claim that both vowels and consonants constrain lexical access to familiar words in the infant lexicon. However, TRACE predicts that infants should become increasingly sensitive to onset mispronunciations (usually consonants in English) of familiar words as vocabulary develops, whereas their sensitivity to non-onset (often vowels) mispronunciations should remain relatively stable during the second year of life, an effect driven purely by changes in the structure and size of the lexicon. Furthermore, vocabulary structure may lead to an asymmetry in mispronunciation sensitivities for lexical items drawn from large neighbourhoods. We also use TRACE to simulate graded sensitivity to mispronunciations of familiar words as reported by White & Morgan (2008). Our simulations predict that phoneme or lexical competition may be absent in the mental lexicons of the 19-month-old infants tested experimentally. Finally, we use TRACE to mimic the impact of phonological neighbourhoods on early word learning reported by Swingley & Aslin (2007). We how how TRACE, a model that does not incorporate a learning component, provides an alternative account of these findings such that infant responding can be explained in terms of mispronunciations of lexical items rather than imputing word learning to the infant.