To delineate causal pathways for melanoma, it is essential to derive unbiased estimates of risk. Extant knowledge derives largely from case-control studies with potential for bias. In a population-based prospective study (QSkin, n = 38,854), we assessed melanoma risks associated with pigmentation characteristics and other phenotypes, and we explored additive interactions. We fitted Cox proportional hazards models to adjust for other factors to estimate the independent effects of each characteristic on melanoma risk. During a mean follow-up of 3.5 years, 642 (1.5%) participants developed melanoma (253 invasive, 389 in situ). The characteristics most strongly associated with invasive melanoma were self-reported nevus density at age 21 years (many vs. no moles hazard ratio [95% confidence interval] = 4.91 [2.81–8.55]), inability to tan (no tan vs. deep tan, hazard ratio [95% confidence interval] = 3.39 [1.85–6.20]), and red hair color (vs. black, hazard ratio [95% confidence interval] = 3.11 [1.50–6.43]). Propensity to sunburn was not associated with melanoma after tanning inability was adjusted for. People with both high nevus density and a history of multiple keratinocyte cancers had significantly higher melanoma risks than those with only one of those traits. We infer that melanoma risk is more strongly related to nevus density and inability to tan than susceptibility to sunburn.