Perner (1991) has claimed that the linguistic structures and reasoning tasks mastered by 4-year-olds share a requirement to handle metarepresentation. In contrast, de Villiers (2000) has argued that they share a requirement to handle misrepresentation. In the current study, a correlation is observed between success on false belief tasks and the acquisition of relative clause sentences. This correlation is not predicted by de Villiers's account because such sentences do not require the handling of misrepresentation, but it is consistent with Perner's account because such sentences do require the handling of metarepresentation. It is proposed that only an account that integrates the accounts of both de Villiers and Perner can explain extant data on language and cognition in 4-year-olds.