The basis for acquisition of categories in child language was investigated. The early encoding of the distinction between mass and count nouns was examined to determine whether children categorize them on the basis of semantic type or syntactic regularities. An experiment was designed in which semantic and syntactic cues were in competition: objects were inappropriately paired with mass noun frames and substances with count noun frames. A total of 44 children aged 3 to 5 were taught nonsense words in either a count or mass noun syntactic context. To test for categorization, the subjects were required to complete a sentence. Nearly all subjects categorized on the basis of syntactic context. However, the appropriate pairing of semantic and syntactic cues did cause a significant number of the younger subjects to produce inconsistent patterns. Thus, while there is no evidence for a semantic encoding of the mass/count distinction, the results for the younger subjects were inconclusive. (RW)