In this article I describe 3 conditions for understanding schizophrenia as a kind of brain disease. First, the disorder must cohere as a clinical entity that is distinct from other disorders. Second, schizophrenia must be linked with an identifiable neuropathology. Third, implicated brain systems must have behavioral functions that fit the psychological characteristics actually seen in schizophrenia. In each case, the evidence is weak or equivocal. However, this does not mean that neurogenic interpretations are untenable. Instead, the primacy of problems associated with the first condition, the need to parse schizophrenia and reduce heterogeneity, must be recognized. Progress in understanding schizophrenia depends on a neurobehavioral approach to resolving the heterogeneity problem.