In our three-stage questionnaire study we investigated patterns of twin and familial aggregation of osteoarthritis (OA) for commonly affected joints. The baseline questionnaire study of the Finnish Twin Cohort was performed in 1975. In 1990, 4095 twin pairs of the same gender born 1930-1957 responded to a questionnaire and reported whether they had OA diagnosed by a physician. In 1996 both twins of 266 pairs of which at least one had reported OA in 1990 responded to a detailed questionnaire on joint-specific OA, including family history of OA. In male pairs shared (non-genetic) familial effects accounted for 37% of the total variance in liability to OA and unshared environmental effects for 63%. In female pairs additive gene effects explained 44% of the variance in liability to OA, and unshared environmental effects for 36%. Familial aggregation of finger and knee OA was clearly higher than that of hip OA. Twin-pair discordance for OA was, to some extent, associated with body-mass index, occupational loading and trauma. Our results indicate that genetic effects may be modulated by sex or by environmental factors distributed differently between men and women. Based on our joint-specific data finger and knee joints are the most optimal targets for studies of genetic factors predisposing to the development of OA.