A study of 207 consecutive white patients older than 18 years of age with nonfamilial nodular and superficial spreading melanoma and 295 controls was conducted in southeast France. Controls were recruited in a public health center. They were matched for sex and age to the overall population older than 18 years of age according to the last census. Melanocytic nevi were counted over the entire body, and data on sun exposure and skin type were collected. A multiple logistic model was used to determine the variables best predicting the risk of melanoma. In order of entry into the model, these variables were number of nevi from 5 to less than 10 mm (N5/10), outdoor leisure per year, frequency of sunburn in the last years, depth of suntan, number of nevi less than 5 mm in diameter (N1/5), age, social level, and hair color. Nevi counts were shown to be indispensable to the construction of a good predictive model for nonfamilial melanoma. After adjustment, estimated relative risks for nonfamilial melanoma rose with increasing number of N1/5, N5/10, and clinically atypical nevi (CAN) on the whole body. Compared with baseline groups the presence of more than 120 N1/5 was associated with a RR of 19.6, the presence of at least five N5/10 with a RR of 10, and the presence of more than one CAN with a RR of 2.77. The number of nevi on the buttocks seemed to be a strong risk factor of melanoma and may provide a simple technique for mass screening. The association between CAN and nonfamilial melanoma could be explained largely by the association between this tumor and the number of large nevi (greater than or equal to 5 mm). It is suggested that when using number of nevi as risk markers, their size might be more important than the other features of clinical atypia. Phenotypic traits and sun exposure, on the one hand, and number of nevi, on the other, seem to be independent risk factors. On the basis of nevus count alone, high-risk subgroups can be identified, which represent only a small part of the overall population older than 18 years of age and from which a large proportion nonfamilial melanoma would theoretically derive.