Using recent data from cancer incidence surveys and measures of UVB exposure levels at seven geographic locations within the United States, we estimate the dose-response relation between UVB and skin melanoma incidence. Mathematical models used information from general population interview studies conducted in these locations to adjust for potentially confounding factors such as age, skin color, ancestry, eye color, hair color, sunburn sensitivity, prevalence of moles, freckles, and hours spent outdoors, use of sunscreen/lotion, and other variables. The effect of geographic UVB exposure on incidence was found to be statistically significant (p less than 0.01) after adjusting for each variable and certain combinations of these variables. We found that incidence rates for those skin melanomas arising in the face, head, neck, or upper extremities (i.e, the most exposed sites) were more sensitive to UVB increases than the incidence rates for those lesions occurring in the ordinarily less exposed sites of the trunk and lower extremities.