This paper discusses applications of advances in psychiatric diagnostic practice to genetic research. Diagnostic misclassification and unreliability can lead to spurious conclusions about patterns of familial aggregation and co-aggregation of psychiatric disorders. Subject, occasion, information, observation and criterion variance are the major components of unreliability in diagnosis. Structured diagnostic criteria are useful for reducing errors; they may, however, lead to unjustified confidence due to their algorithmic and semi-quantitative format. Six misconceptions regarding such criteria are discussed. The choice of instruments and method of implementation must take into account the qualifications of interviewers and the need for blindness in the experimental design. A fundamental design decision in psychiatric genetic research is the choice between the family history or family study method. The costs and benefits of each method are examined. Finally, the psychiatric geneticist must face the problem of defining the ill phenotype. This is made difficult by the presence of phenocopies and the possibility that milder spectrum disorders and some symptom-free individuals may carry the genotype that underlies the phenotype of the more severe clinical form of the disorder.