Distributional learning and lexical category acquisition: What makes words easy to categorize?

In this study, results of computational simulations on English child-directed speech are presented to uncover what distributional properties of words make it easier to group them into lexical categories. This analysis provides evidence that words are easier to categorize when (i) they are hard to predict given the contexts they occur in; (ii) they occur in few different contexts; and (iii) their contextual distributions have a low entropy, meaning that they tend to occur more often in one of the contexts they occur in. This profile fits that of content words, especially nouns and verbs, which is consistent with developmental evidence showing that children learning English start by forming a noun and a verb category. These results further characterize the role of distributional information in lexical category acquisition and confirm that it is a robust, reliable, and developmentally plausible source to learn lexical categories.