We examined the reproducibility of the measurement of sun exposure in a cohort study of nonmelanocytic skin cancer in Geraldton, Western Australia. Two analyses were undertaken: a comparison of cutaneous sun damage with sun exposure reported at interview, and an analysis of test-retest reproducibility of reported exposure. Skin cancers and cutaneous indicators of sun damage (cutaneous microtopography and solar elastosis of the neck) were recorded at a survey in 1987. A case-control study was conducted in 1988 in which subjects were interviewed about their lifetime sun exposure. All subjects had European ancestry. A subset of these subjects was reinterviewed using the same interview schedule in 1993. The comparison of reported exposure with skin damage was restricted to 201 cases of basal cell carcinoma and 700 controls, all of whom were born in Australia and had no southern European ancestors. The analysis of test-retest reproducibility included 62 cases with basal cell carcinoma and 162 controls. After adjustment for the skin's sensitivity to sunlight, cutaneous microtopography explained 7% and solar elastosis of the back of the neck explained 13% of the variance in the reported time spent outdoors. The intraclass correlation between time spent outdoors reported in the two interviews was 0.77 [95% confidence interval (CI), 0.71-0.83], whereas for exposure to a specific anatomical site, it was 0.65 (95% CI, 0.55-0.73). The reported site-specific exposure was lower on the second occasion in controls but higher in cases. The hours of exposure on vacations and the proportion of exposure that occurred on nonworking days had poor reproducibility. Furthermore, cases reported a more intermittent pattern of weekly exposure on the first occasion than on the second, whereas the controls showed little difference in their pattern on the two occasions. The weighted kappa statistic for lifetime painful sunburns was 0.53 (95% CI, 0.41-0.66), and for lifetime number of blistering sunburns, it was 0.54 (95% CI, 0.44-0.65). Thus, the reported sun exposure showed only moderate agreement with biological markers of sun damage. Total sun exposure and, to a lesser extent, site-specific exposure showed good agreement on the two occasions. However, indicators of intermittent sun exposure had poor agreement, and sunburn had only fair agreement.