Reliability of reported sunburn history was assessed. Fifty cases and 50 controls were randomly selected from a population-based case-control study of skin examination practices and lethal melanoma in Connecticut. Eighty-two percent of these participants were interviewed from January 15, 1987, to May 15, 1989, and reinterviewed in 1990. The kappa coefficients ranged between 0.37 and 0.57; the intraclass correlation coefficients were similar and ranged between 0.34 and 0.69. There were no significant differences found by case-control status or sex for any variables, and the degree of disagreement was similar for cases and controls and could be considered free of recall bias.