Complex segregation analysis was performed on 467 nuclear families ascertained through febrile-convulsion probands. The probands were identified as having their first febrile convulsion while residents of Rochester, MN, during the years 1935-64. Parents and first- and second-degree relatives of probands were identified through the Olmsted County, MN, record-linkage system. Diagnoses of convulsive activity were made from review of medical records. The genetic models investigated included both single-major-locus and polygenic models, with likelihoods computed jointly on children and parents as well as being conditioned on parental phenotype. Possible heterogeneity was investigated by means of analyses of frequency of febrile convulsions in the proband. Analyses of the entire data set indicated that the single-major-locus models could be rejected. The most parsimonious model for these data was the pure polygenic (or common familial environment) model with a large heritable component (68% +/- 7%). However, when families were partitioned on the basis of frequency of febrile convulsions in the proband, significant heterogeneity was present. Our results indicated that the polygenic model was strongly corroborated in families of probands with a single febrile convulsion. In families of probands with multiple febrile convulsions, evidence was consistent with a single-major-locus model with nearly dominant seizure susceptibility.