Humans' unique aptitude for reasoning about mental states, known as Theory of Mind (ToM), can help explain the unique character of human communication and social interaction. ToM has been studied extensively in children, but there is no clear account of the cognitive basis of ToM in adults. Evidence from functional imaging and neuropsychology is beginning to address this surprising gap in our understanding, and this evidence is often thought to favour a domain-specific or modular architecture for ToM. We present a systematic approach to this issue for the paradigmatic case of belief reasoning, and argue that neuropsychological data provide no clear evidence for domain-specificity or modularity. Progress in understanding ToM requires new tasks that isolate potentially distinct components of this complex ability.