Semantic and structural factors in sentence processing and word learning

This work presents two studies of language processing and development in children. The first study focuses on passive sentence comprehension in 4-5-year-old children with developmental language disorder (DLD) and same-age peers with typical development (TD). We explore the effect of animacy, morphosyntactic, vocabulary, and event probability cues on children’s offline comprehension and online processing of passive sentences using an eye-tracked looking-while-listening design. The children were first exposed to short videos of agents doing characteristic actions (e.g., hard physical activities or passively observant activities). The children then engaged in an eye-tracked online processing task in which they heard reversible and nonreversible passive sentences describing events that matched or did not match the characteristics set up in the exposure videos. During these sentences, images on-screen were displayed that corresponded to the potential interpretations of the sentence. Online processing data was collected using eye tracking. After each sentence, the children were asked to point to the image corresponding to their interpretation to measure their offline comprehension. The offline comprehension data indicated that compared to the children with TD, the children with DLD were less likely to correctly interpret the passive sentences and made comprehension errors that suggested poorer attention to and integration of potentially informative sentence cues. The eye-tracked online processing data was examined in two ways. First, we analyzed the online processing data to determine to what extent the children’s processing was consistent with the use of the sentence cues. We found that the children in the two groups were just as likely to demonstrate looking patterns consistent with the use animacy cues but children with DLD were less likely to use morphosyntactic, vocabulary, and event probability cues. We then analyzed the online processing data in correctly interpreted sentences only to examine how the sentence cues were integrated over the course of the sentence. We found that in correctly interpreted sentences, children with DLD demonstrated a slower, less robust response to most of the informative cues in the sentences but quicker and less linguistically mediated use of event probability cues. Finally, we examined the relationship between the children’s use of event probability cues and their stimuli-specific vocabulary knowledge but found no strong associations.

The second study focuses on the semantic network structure of the vocabularies of young 18-30-month-old children and its influence on noun and verb learning. Prior work had examined how noun semantic network structure affects noun learning. Here, we extended that work to ask how noun and verb semantic network structures differ in their influence on noun and verb learning. We examined vocabulary network structure at the word, semantic neighborhood, and lexicon levels in a large sample of child vocabulary checklist data using semantic features. We analyzed the data in three ways. First, we charted the relationship between verb and noun semantic network structure and vocabulary size across children. We found that early-learned nouns tended to have strong network relationships with other nouns and other verbs across network levels. We also found that early-learned verbs tended to have strong network relationships with other nouns but, in contrast, were unlikely to have strong relationships with other verbs. Next, we examined patterns of normative vocabulary development, asking whether the cross-sectional patterns seen in the first analysis influenced the time at which nouns and verbs tended to be learned. We found similar patterns. Nouns with strong semantic connections to other nouns and other verbs tended to be learned early. Verbs with strong semantic connections to other nouns tended to be learned early but verbs with strong semantic connections to other verbs tended to be learned later. Finally, in an effort to determine how the very earliest vocabulary knowledge sets the stage for later word learning, we examined how word knowledge gaps formed and were filled by nouns and verbs in normative vocabulary development. These gaps corresponded to structured “cavities” in the semantic networks. We found that nouns tended to form semantic cavities that were later filled by nouns and verbs, reinforcing the findings from the first analyses that early noun knowledge is a prerequisite for later verb learning.