Research on children's theory of mind often restricts conceptually meaningful talk about knowledge to instances where know references a corresponding mental state. This article offers a reappraisal of that view. From a social-pragmatic perspective, even nonreferential talk is meaningful when appropriately embedded in social routines. A synthesis of corpus data suggests children's early talk about knowledge routinely occurs in question–answer contexts. It is argued that the influence of interrogative contexts is evident in children's over-attributions of knowledge when someone is only (...) guessing. This influence is taken as evidence for the role of linguistic practices in shaping the concept of knowledge. (shrink)