Abstract for Savita Bernal's talk

Syntactic bootstrapping of lexical acquisition : ERP investigation in adults and infants
Savita Bernal (LSCP-Ecole Normale Superieure, Paris)
April 14, 2005 (Thursday)
6:30 PM - 8:00 PM; Room 7102, The CUNY Graduate Center

The general purpose of my doctoral work is to investigate the link between word learning and syntax acquisition in infants, building on Lila Gleitman's hypothesis of 'Syntactic Bootstrapping'(SB). SB postulates that knowing some aspects of syntax might help in restricting the wide array of possible meanings for a given new word. For example, knowing that verbs generally refer to action and nouns to objects would induce the kid to think that in "the dax is blicketing", 'blicketing' refers to the action and 'dax' to the agent involved in the action performed in front of her. But how can infants know that a word is a verb or a noun in the first place? Function words (FW) are very good indicators of the syntactic structure of a sentence. Since the categories of FW map with the grammatical categories of the word they accompany, and since FW have been shown to be identified early in development, we postulated that they may be a good cue to word category, and thus, indirectly, to word meaning. We thus designed an ERP experiment to test this hypothesis. It was performed with adults to check on the methodology used. We first familiarized the subjects with a new pseudo-word, accompanied by three FW of the same category (e.g. : 'the blicket', 'this blicket', 'my blicket') and then presented the same content word with either a new FW from the same category (e.g. 'a blicket') or from another category (e.g. 'I blicket'). We believed that this incongruent situation should trigger a syntactic mismatch in the ERP as compared to the congruent situation. A control condition with known words was present. Contrary to our expectations, no significant syntactic mismatch was observed in the Pseudo-word condition, whereas this mismatch was present in the Known-word condition, in addition to a phonological expectation mismatch response. We believe that these results reflect the fact that three out-of-context repetitions of a new word are not enough to build a fully functional lexical entry. We will then present the design of a new 2-year-olds ERP experiment that is to begin soon, that uses real words in full sentences.