How do children manage to identify number words in the input and acquire their meaning? One solution that has been proposed is that children benefit from attending to their relative distribution in a limited range of syntactic environments, and comparing this distribution with competing lexical items with overlapping distribution, such as quantifiers and adjectives. The strong version of this proposal has children obtaining the semantics of number words directly from these syntactic environments, while a weaker version maintains that these environments allow children to distinguish quantity-denoting words from others, but that other verbal cues such as the counting routine serve to highlight number word meaning specifically. In this talk, I present results from two complementary sources - a corpus analysis of child-directed speech and experimental evidence from child and adult participants - arguing against the strong version of this proposal. The results also remind us to be cautious when inferring a lack of underlying competence from less-than-perfect performance.