2018 Friday Session A 1100

Friday, November 2, 2018 | Session A, East Balcony | 11am

French-learning infants use novel syntactic contexts to acquire the meanings of novel content words
M. Barbir, M. Babineau, A. Fiévet, A. Christophe

Syntactic bootstrapping studies have shown that infants use syntactic cues they already know to distinguish between categories of words, such as transitive and intransitive verbs. However, the question of how infants learn to map syntactic contexts onto categories of words has only been investigated using computational modeling [1]. One hypothesis relies on the fact that 6-month-olds already know the meanings of a few common content words [2] and 12-month-olds have already constructed some semantic categories (e.g. objects, actions, agents) [3] to suggest that infants may use these familiar words as a semantic seed, in order to learn which contexts correspond to which categories –e.g., “the bottle”, “the book”, therefore ‘the X’, X=object [4]. This hypothesis is supported by models that achieve high precision in categorization tasks when relying on a semantic seed [1, 5].

The current study investigates experimentally whether, by relying on a semantic seed, infants can learn to use novel syntactic cues to distinguish between categories of words. We presented French-learning infants with two novel syntactic contexts –the determiner “ko” before animates and “ka” before inanimates –alongside words they already know, the semantic seed (ko chien, ka poussette, ‘ko dog, ka stroller’). We selected animacy because it is marked morpho- syntactically in many of the world’s languages (although not in French), and is likely to be salient to infants [6, 7]. The novel determiners replace existent determiners (le, la, un, une, ‘the, a’). These contexts are presented in a short video in which a woman acts out stories, using “ko” and “ka” a total of 60 times (Fig. 1).

We tested 20-month-old infants (n=24) to see whether they could use these newly-learned contexts to categorize novel words. Before coming to the lab, infants watched the video three times at home, over three consecutive days. In the lab, they saw the video once more (training phase). Then, during the test phase, they saw two novel images side by side (one of an animate and one of an inanimate) and were asked to look at one of the two images (e.g., Regarde ko bamoule!, ‘Look at ko bamoule!’, Fig. 1). An analysis of overall looking-time revealed that infants looked significantly longer to the animate image when they heard a novel content word preceded by “ko” than preceded by “ka” (p < 0.05, Fig. 2a,b). A cluster-based permutation analysis confirmed that infants looked longer toward the animate image when hearing “ko”, between 300 and 2440 ms after hearing “ko” (p = 0.01, Fig. 3).

These results suggest that infants can use newly-acquired syntactic contexts to infer the plausible meaning of novel content words. Ongoing work aims at determining whether the capacity to learn novel syntactic contexts is present throughout development and into adulthood. The present findings provide experimental evidence in favor of the semantic seed as a sufficient cue for mapping syntactic contexts onto categories of words, and deepen our understanding of how infants go about stitching together their syntactic bootstraps.