To assess the role of surveillance bias in the observed threefold excess of cutaneous malignant melanoma (CMM) at the Lawrence Livermore National Laboratory (LLNL) in California, we examined the thickness of CMMs among all 20 laboratory employees who were members of a large prepaid health plan and whose CMM was diagnosed from 1970 through 1984. For comparison, we reviewed slides of 36 other members of the same health plan matched (usually 2:1) to the laboratory case by age, sex, facility, and year of diagnosis. Three expert dermatopathologists read the slides using a multiheaded microscope to reach a consensus; they were blind to the laboratory employment status of the subjects. We found that from 1970 to 1976, before there was widespread publicity about the excess incidence of CMM at LLNL, lesion thickness was greater for non-LLNL employees (mean difference =1.5 mm; 95% confidence interval 0.1–2.9). From 1977 through 1984, however, there was no appreciable difference [mean difference = −0.3 mm; 95% confidence limits (CL) = −1.4, 0.9]. Dropping the matching to adjust for histologic type of melanoma as well as gender, year, and age at diagnosis yielded substantially the same results. These data are compatible with an effect of surveillance bias up to around 1976, but in this health plan population, they do not support a role for surveillance bias in the continuing excess incidence observed since that time.