Objective. To investigate the inheritance of generalized osteoarthritis (OA). Methods. OA was identified on hand and knee radiographs obtained from members of the Framingham Study cohort (the parents) in 1967-1970 and 1992- 1993, and from their adult children in the Framingham Offspring Study in 1993-1994. All hand and knee radiographs evaluated for OA were graded using the Kellgren and Lawrence (K/L) scale. A measure of generalized OA was defined as the count of the number of hand and knee joints affected, as determined by the proportion of joints with a K/L grade ≤2. The OA count, treated as a continuous variable, was adjusted for age, body mass index, and a measure of physical activity for each joint area (hand or knee). Calculations were made separately for each generation and each sex, and correlations were analyzed against the standardized residual of OA. Segregation analysis was used to test whether OA aggregated in families, and if its transmission fit a Mendelian pattern. Results. A total of 337 nuclear families with 2 parents and at least 1 biologic offspring were studied. In parents, the mean age was 61.2 years at the time of hand radiographs and 72.8 years at the time of knee radiographs, which were mostly obtained at a later examination. The mean age at the time of radiographs in offspring was 53.9 years. Using standardized residuals, parent-offspring and sibling-sibling correlations ranged from 0.115 to 0.306. In segregation analyses, models testing the hypotheses of no familial aggregation, no familial transmission, or a Mendelian gene alone were all rejected (P < 0.001 for each of these models). The best-fitting models were mixed models with a Mendelian mode of inheritance and a residual multifactorial component. The Mendelian recessive model provided the best fit. Conclusion. These analyses support a significant genetic contribution to OA, with evidence for a major recessive gene and a multifactorial component, representing either polygenic or environmental factors.