When presented with a pair of objects, one familiar and one unfamiliar, and asked to select the referent of a novel word, children reliably demonstrate the disambiguation effect and select the unfamiliar object. The current study investigated two competing word learning accounts of this effect: a pragmatic account and a word learning principles account. Two-, three- and four-year-olds were presented with four disambiguation conditions, a word/word, a word/fact, a fact/word and a fact/fact condition. A pragmatic account predicted disambiguation in all four conditions while a word learning principles account predicted disambiguation in the word/word and fact/word conditions. Results indicated that children disambiguated in word/word and fact/word conditions and two-year-olds disambiguated at above chance levels in the word/word condition but at "below" chance levels in the fact/fact condition. Because disambiguation varied both as a function of age and condition these findings are presented as challenges to a pragmatic account of the disambiguation effect.