Cancer incidence in the Dalgety Bay area of Fife, Scotland, was examined following the detection of radium-226 particles by routine radiation monitoring. The study was confounded by rapid population growth, demographic change and the relatively high socioeconomic status of the Dalgety Bay population. Health Board Primary Care Division records were used to calculate population estimates and Carstairs deprivation score was used to adjust for socioeconomic characteristics. In the period 1975-90, 211 residents were registered as having cancer compared with 214.21 expected from Scottish national rates. Of specific cancers possibly associated with radiation, the incidence of stomach, liver, lung, bone, prostate, bladder and kidney cancer and lymphoma were lower than expected while colon, rectum, pancreas, skin, breast and thyroid cancer and multiple myeloma and leukaemia were higher. There were three cases of childhood leukaemia compared with 1.22 expected. The only statistically significant differences observed were for pancreas (11 cases, O/E 2.28), lung (25 cases, O/E 0.65) and non-melanoma skin (36 cases, O/E 1.50). Stomach cancer was of borderline statistical significance (four cases, O/E 0.40). Adjustments for socioeconomic factors accounted for the apparently low incidence of stomach and lung cancer and, to a lesser extent, skin cancer, which remained of borderline statistical significance. Results in relation to pancreas cancer were unchanged. The observations of raised incidence of pancreas and skin cancer arose in the context of a survey of 17 cancer sites, from which the finding of two or more statistically significant results is not unusual (P = 0.21), and the numbers of cases involved were small. The epidemiological evidence for an association between radiation exposure and pancreas cancer risk is weak. Stronger evidence exists for an association with skin cancer. In the present study the anatomical distribution of the 36 cases was similar to that found elsewhere in Scotland.