Foundations of Vocabulary: Does Statistical Segmentation of Events Contribute to Word Learning?

Abstract

This dissertation evaluates the untested assumption that the individuation of events into units matters for word learning, particularly the learning of terms which map onto relational event units (Gentner & Boroditsky, 2001; Maguire et al., 2006). We predicted that 3-year-old children’s statistical action segmentation abilities would relate to their verb comprehension and to their overall vocabulary knowledge (Research Question 1). We also hypothesized that statistical action segmentation would facilitate children’s learning of novel verbs (Research Question 2). Largely confirming our first prediction, children who were better able to statistically segment novel action sequences into reliable units had more sophisticated overall vocabularies and were quicker to select the correct referents of overall vocabulary items and verb vocabulary items; nevertheless, they did not have larger verb vocabularies. Unexpectedly, statistical action segmentation did not facilitate children’s learning of verbs for statistically consistent action units. However, children showed greater learning of verbs labeling statistical action part-units than verbs labeling statistical action non-units, providing some evidence for our second prediction. In sum, this dissertation takes an important step towards understanding how event segmentation may contribute to vocabulary acquisition.