Objective: The purpose of this study was to assess the mode of inheritance for obsessive-compulsive disorder (OCD) in families ascertained through pediatric probands. Methods: We ascertained 52 families (35 case and 17 control families) through probands between the ages of 10 and 17 years. Direct interviews were completed with 215 individuals. Family informant data were collected on another 450 individuals without direct interviews, forming two data sets with one contained within the other. Complex segregation analyses were performed using regressive models as programmed in REGTL in the S.A.G.E. package.All models used in the analyses included sex-specific age and type parameters. Results: All models that excluded a residual effect of an affected parent were rejected. With that parameter included, the environmental and sporadic models were rejected in comparisons with the most general model in both data sets (all p < 0.005). With the direct interview data, the general codominant Mendelian model was not rejected when compared with the most general model (p = 0.140). We could not distinguish between any of the simple Mendelian models using either data set. However, the dominant Mendelian model provided a somewhat better fit than the other Mendelian models to the direct interview data. Conclusions: The results provide evidence for a major susceptibility locus in families with OCD when age at onset is incorporated into the model. Mendelian factors at most partially explained the familial aggregation of the phenotype, and residual familial effects were necessary to fit the data adequately. The results support the importance of linkage efforts by suggesting that a major locus is segregating within a proportion of families with OCD ascertained through pediatric probands.