The effect of semantic density and sound density on early word learning

Early lexical acquisition is affected by biases and constraints within learners, but also by patterns and statistical regularities within a learner's environment. Much of the previous work examining the effect of statistical regularities on word learning has been directed at phonological regularities. Particularly, research has focused on the similarity between word forms, referred to as phonological neighborhood density or phonological density. In the current work we examine the effect of both phonological density and semantic density on word learning.;A set of regression analyses predicting early comprehension and production with semantic density, phonological density and frequency, were conducted. Results show that semantic density was a significant predictor of both comprehension and production. This illuminates the significant role that semantic similarity plays in lexical acquisition and reinforces the importance of examining regularities among the concepts onto which early words map. Results also show that effects of both phonological and semantic similarity are moderated by a word's frequency. Thus early comprehension and production can both be facilitated and hindered by similarity, depending on existing knowledge and experience.;To further examine word learning processes, a computational model was developed. The model mapped between phonological and semantic representations to simulate early lexical acquisition. Modeling results simulated important benchmark phenomena in word learning, such as the word spurt and comprehension preceding production. Next, several behavioral findings from word learning studies were simulated and explored, including mutual exclusivity, instances of similarity facilitating acquisition, and instances of similarity inhibiting acquisition. These effects were examined with respect to the model's hidden layer representations. Findings were discussed in terms of represented similarity and distinctiveness, and how these underlying representations are influenced by similarity of phonological and semantic representations of words'.