Genetic linkage studies of schizophrenia depend on accurate psychiatric diagnosis of relatives within multiply affected families. Each investigator makes a series of explicit or implicit decisions to define which relatives will be assumed to share a schizophrenia-related genotype, that is, who is an "affected relative." In this article we delineate issues that we believe should be considered in such studies and review the relevant literature. Issues include criteria for selecting probands; whether broader criteria should be used to select affected relatives; approaches to including or excluding diagnoses for which family study data suggest a relationship to schizophrenia or to affective disorders or other psychiatric disorders; clarification of diagnostic hierarchy; and issues related to substance abuse and neurological disorders. Also discussed are whether relatives without spectrum diagnoses should be considered unaffected or undiagnosed in linkage analyses, how bilateral familial affectedness should be defined, and provision for independent review of study diagnoses. As an illustration, the clinical model for the authors' schizophrenia linkage study is described.