Use of sunscreens is widely advocated as a preventive measure against sun-induced skin cancers. However, to date, no epidemiologic study has reported a decreased melanoma risk associated with sunscreen use. We have conducted a case-control study aimed at evaluating the influence of sunscreen use on the occurrence of cutaneous malignant melanoma. In 1991 and 1992, 418 melanoma cases and 438 healthy controls were interviewed in Germany, France and Belgium. The questionnaire used differentiated between regular sunscreens, psoralen sunscreen (prepared with 5-methoxypsoralen, a tanning activator and photocarcinogen), and self-tanning cosmetics (which produce a tan without ultraviolet radiation). After adjusting for age, sex, hair colour and holiday weeks spent each year in sunny resorts, the melanoma risk was of 1.50 (95% Cl:1.09-2.06) for regular sunscreens, and of 2.28 (95% Cl: 1.28-4.04) for psoralen sunscreens. No melanoma risk was associated with use of self-tanning cosmetics. Among subjects with a poor ability to tan, psoralen sunscreen users displayed a melanoma risk of 4.45 (95% Cl: 1.25-15.8) when compared with regular sunscreen users. There was a significant negative interaction between regular sunscreen use and sunburns experienced in adulthood. Use of sunscreens, especially psoralen sunscreen, was associated with higher density of pigmented lesions of the skin. Although we cannot exclude the presence of an unknown confounding factor, our results support the hypothesis that sunscreens do not protect against melanoma, probably because of their ability to delay or avoid sunburn episodes, which may allow prolonged exposure to unfiltered ultraviolet radiation. Serious doubts are raised regarding the safety of sunscreens containing psoralens.