Children infer the meaning of novel verbs through morphological bootstrapping

Abstract

Children are incredibly efficient at detecting and extracting structure from input. Such skills are evident in the efficiency with which children learn language; for example, children in their second year of life already rely on syntactic context to infer the meaning of words, i.e., syntactic bootstrapping. However, the syntactic bootstrapping hypothesis was originally developed using data from English-learning children, and tested mostly on Indo-European languages. Yet, languages vary in how they mark information. Here, we expand the idea of bootstrapping words’ meaning to include morphology - a linguistic feature which is lean in English but substantial cross-linguistically. Hebrew-speaking 3- to-4-year-olds (N = 29) viewed dialogues containing novel verbs constructed in familiar morphological templates: an intransitive template (hitCaCeC), and a template that tends to be transitive (CiCeC). Each dialogue was followed by a test phase, in which children viewed side-by-side a transitive video (a child performing a novel action on an adult), and an intransitive video (the child performing a novel action alone) - and heard a sentence containing the novel verb. Children looked more than chance towards the transitive video in transitive trials, but not on intransitive trials, suggesting they were able to use the transitive but not intransitive template to infer word meaning. This is consistent with previous results on syntactic bootstrapping. We discuss possible reasons why children might not have succeeded in the intransitive condition, and suggest that children rely on distributional information, instead of only hard rules, to infer word meaning.