Proceedings of the Annual Meeting of the Cognitive Science Society

Children tend to produce words earlier when they are connected toa variety of other words along both the phonological and semanticdimensions. Though this connectivity effect has been extensivelydocumented, little is known about the underlying developmentalmechanism. One view suggests that learning is primarily drivenby a network growth model where highly connected words in thechild’s early lexicon attract similar words. Another view suggeststhat learning is driven by highly connected words in the externallearning environment instead of highly connected words in the earlyinternal lexicon. The present study tests both scenarios system-atically in both the phonological and semantic domains, and across8 languages. We show that external connectivity in the learningenvironment drives growth in both the semantic and the phonolog-ical networks, and that this pattern is consistent cross-linguistically.The findings suggest a word learning mechanism where childrenharness their statistical learning abilities to (indirectly) detect andlearn highly connected words in the learning environment.