Using a systematically ascertained sample of 57 families, each having 2 or more members with a consensus diagnosis of schizophrenia (DSM-III-R criteria), we have carried out linkage studies of 520 loci, covering approximately 70% of the genome for susceptibility loci for schizophrenia. A two-stage strategy based on lod score thresholds from simulation studies of our sample identified regions for further exploration. In each region, a dense map of highly informative dinucleotide repeat polymorphisms (heterozygosity greater than .70) was analyzed using dominant, recessive, and "affected only" models and nonparametric sib pair identity-by-descent methods. For one region, 8p22-p21, affected sib-pair analyses gave a P value = .0001, corresponding to a lod score approximately equal to 3.00. For 8p22-p21, the maximum two-point lod score occurred using the "affected only" recessive model (ZMAX = 2.35; theta M = theta F); allowing for a constant sex difference in recombination fractions found in reference pedigrees, ZMAX = 2.78 (theta M/theta F = 3). For a second region, 3p26-p24, the maximum two-point lod score was 2.34 ("affected only" dominant model), and the affected sib-pair P value was .01. These two regions are worthy of further exploration as potential sites of susceptibility genes for schizophrenia.