This research concerns how children learn the distinction between substance names and object names. Quine (1969) proposed that children learn the distinction through learning the syntactic distinctions inherent in count/mass grammar. However, Soja et al. (1991) found that English-speaking 2-year-olds, who did not seem to have acquired count/mass grammar, distinguished objects from substances in a word extension task, suggesting a pre-linguistic ontological distinction. To test whether the distinction between object names and substance names is conceptually or linguistically driven, we repeated Soja et al.'s study with English- and Japanese-speaking 2-, 2.5-, and 4-year-olds and adults. Japanese does not make a count-mass grammatical distinction: all inanimate nouns are treated alike. Thus if young Japanese children made the object-substance distinction in word meaning, this would support the early ontology position over the linguistic influence position. We used three types of standards: substances (e.g., sand in an S-shape), simple objects (e.g., a kidney-shaped piece of paraffin) and complex objects (e.g., a wood whisk). The subjects learned novel nouns in neutral syntax denoting each standard entity. They were then asked which of the two alternatives--one matching in shape but not material and the other matching in material but not shape--would also be named by the same label. The results suggest the universal use of ontological knowledge in early word learning. Children in both languages showed differentiation between (complex) objects and substances as early as 2 years of age. However, there were also early cross-linguistic differences. American and Japanese children generalized the simple object instances and the substance instances differently. We speculate that children universally make a distinction between individuals and non-individuals in word learning but that the nature of the categories and the boundary between them is influenced by language.