Skip to main content

Processing Activates Specific Constraints for Language Acquisition

Final Report Summary - PASCAL (Processing Activates Specific Constraints for Language Acquisition)

PASCAL is a project to investigate the human ability to acquire language, and the underlying biological endowment that provides specific language learning mechanisms. Our studies of language acquisition during the first few months of life aim at fulfilling five objectives listed below together with the results we obtained so far. Objective 1: to explore perceptual and memory constraints in development - using NIRS, we have demonstrated that neonates are biased to encode items positioned at the edges of sequences, compared to items in the middle of sequence, indicating that this enhanced processing of edges is a constraint present from birth. We have also demonstrated that neonates can compute statistical information from words and use this information to extract and remember words from continuous speech. We have further shown that 4-month-old infants can recognize various objects and movements after a short exposure. Objective 2: to explore perception and production in development - we investigated the relation between production and perception, specifically whether experience with the vocal tract is necessary to be sensitive to constraints imposed by the sonority hierarchy. We found that newborns are already sensitive to these constraints. Objective 3: to explore prosodic processing from birth onwards - we investigated the sensitivity to the Iambic-Trochaic law (ITL). We carried out experiments on syllabic grouping on infants as well as adults and found that both group trochaically syllables that alternate in pitch, but only adults, and not 7-months old infants, group syllables iambically when they alternate in duration. We concluded that the sensitivity to duration – less salient then pitch in the signal is developed later. In fact it is by now known that 9-month-olds are sensitive to both. We also investigated how cross-linguistic differences may influence segmentation in a novel language. We investigated how people use statistical cues and how they extract rules, based on prosodic regularities and how they detect the boundaries between words and phrases in continuous speech. We also carried out experiments to see to what extent the use of these cues is determined by the native language of the listener, and to what extend it is universal. Objective 4: to explore sound categories from birth onwards - in past work we have shown that transition probabilities (TPs) are computed on syllables and on consonants (Cs) but not on vowels (Vs). We argued that the reasons for this asymmetry between the two segmental categories is that Cs are more distinctive than Vs cross-linguistically and that vowels are too variable because they carry prosody. Similar experiments were carried out with 12-months old infants and we found that they rely more on consonants for word extraction and on vowels for generalizations. We then investigated newborns using NIRS to see if they are more sensitive to one of the two types of segments, and found that contrary to adults and older infants, neonates better recognize the information carried by Vs than that carried by Cs. Objective 5: to explore bilingualism in the crib: from basic research to educational policies - we examined the cues that help infants (12-month-old) to distinguish between different sources of information in order to optimize later learning processes. We hypothesized that infants would be biased to pay more attention to, and to learn more from people with whom they share the same language.