A systematic meta-analysis of observational studies of melanoma and family history, actinic damage and phenotypic factors was conducted as part of a comprehensive meta-analysis of all major risk factors for melanoma. Following a systematic literature search, relative risks were extracted from 60 studies published before September 2002. Fixed and random effects models were used to obtain pooled estimates for family history (RR = 1.74, 1.41-2.14), skin type (I vs. IV: RR = 2.09, 1.67-2.58), high density of freckles (RR = 2.10, 1.80-2.45), skin colour (Fair vs. Dark: RR = 2.06, 1.68-2.52), eye colour (Blue vs. Dark: RR = 1.47, 1.28-1.69) and hair colour (Red vs. Dark: RR = 3.64, 2.56-5.37), pre-malignant and skin cancer lesions (RR = 4.28, 2.80-6.55) and actinic damage indicators (RR = 2.02, 1.24-3.29). Sub-group analysis and meta-regression were carried out to explore sources of between-study variation and bias. Sensitivity analyses investigated reliability of results and publication bias. Latitude and adjustment for phenotype were two study characteristics that significantly influenced the estimates.