Computational models have shown that purely statistical knowledge about words' linguistic contexts is sufficient to learn many properties of words, including syntactic and semantic category. For example, models can infer that "postman" and "mailman" are semantically similar because they have quantitatively similar patterns of…

This dissertation presents results from a series of experiments investigating adult learning of an artificial language and the effects that input frequency (high vs. low token frequency), frequency distribution (skewed vs. balanced), presentation mode (structured vs. scrambled), and first language (English vs. Japanese) have on such learning.…

According to "noun-cue" models, arbitrary linguistic categories, like those associated with case and gender systems, are difficult to learn unless members of the target category (i.e., nouns) are marked with phonological or semantic cues that reliably co-occur with grammatical morphemes (e.g., determiners) that exemplify the categories. "Syntactic…