We have examined the sample sizes required to detect linkage using the affected-sib-pair (ASP) method for major psychiatric disorders that are characterized by population prevalences of 1-7%, decreased penetrance, phenocopies, and heterogeneity. In addition, the nature of these illnesses makes large, multigenerational pedigrees difficult to collect. We calculated the sample sizes needed to have 80% power of finding linkage (with a type I error rate of 5%) under dominant and recessive models with incomplete penetrance and allowing for recombination rates of up to 10% between the disease gene and marker gene. We have assumed that the identical-by-descent (IBD) status of ASPs is known exactly. For a disease like schizophrenia (1% population prevalence), if 50% of families are linked to a marker locus at 10% recombination, then 60 and 120 pairs are needed under recessive and dominant inheritance, respectively. For a disorder such as major affective disorder (7% population prevalence), the sample size is similar if the inheritance is recessive, but larger (160 pairs) if the inheritance is dominant. We conclude that this method may be a reasonable alternative for psychiatric disorders, especially to confirm that a linkage found in a specific pedigree or population isolate is also present in the general population.