Cognitive development researchers have drawn conclusions about young children’s developing knowledge of number by studying their behavior, while at the same time acknowledging that behavior is an imperfect index of knowledge, e.g., it may be disputed whether a given behavioral task accurately measures, overestimates, or underestimates children’s knowledge. The texts of published research articles from these investigations are the focus of a discourse analysis described in the present article. The results of the discourse analysis suggest that claims about what a person knows are actually generalized descriptions of behavior. Therefore, in studying behavior on tasks to draw conclusions about participants’ conceptual knowledge, researchers are merely making behavioral generalizations, not investigating hidden cognitive or epistemic content.