To assess whether screening people at high risk of malignant melanoma would be effective in reducing the mortality from the disease data from 400 case-control pairs in a study of cutaneous malignant melanoma conducted in Western Australia during 1980-1 were used to predict the risk of melanoma in the remaining 111 pairs. All variables previously shown to be associated with a decrease or increase in the incidence of melanoma were considered for inclusion in a single conditional logistic regression model of the incidence of melanoma in the randomly chosen subset of 400 case-control pairs. Five of these variables--number of raised naevi on the arms, arrival in Australia before 10 years of age, history of non-melanocytic skin cancer, time spent outdoors in summer from the age of 10 to 24, and family history of melanoma--provided good discrimination between patients and controls in this sample and the 111 other case-control pairs. Among the 222 subjects in these other case-control pairs a group defined as being at high risk of melanoma by a risk score derived from these five variables contained 60 (54%) of the patients with melanoma but only 18 (16%) of the controls. These data suggest that in Western Australia more than half of all new patients with melanoma arise in an identifiable subpopulation constituting less than one fifth of the whole population. Identifying this subpopulation and screening it regularly for cutaneous malignant melanoma could be cost effective in reducing mortality from this disease.