Epidemiological data on occupational exposure and personal use of hair dyes was reviewed with specific focus on bladder cancer and lymphoid neoplasms. At least seven cohort and 11 case-control studies included data on occupational exposure to hair dyes by hairdressers, barbers and beauticians, and their subsequent bladder cancer risk. The relative risk (RR) estimate was 1.4 (183 observedvs129 expected) for cohort studies, and in several casecontrol studies the RRs were somewhat above unity. These results are compatible with some moderate association between past professional exposure to hair dyes and subsequent bladder cancer risk, but also with errors and biases in observational epidemiological studies, particularly since allowance for smoking was lacking or inadequate in most studies. An open question is whether current occupational exposure to modern hair dyes is still related to some excess bladder cancer risk. Five case-control studies included information on personal use of hair dyes and bladder cancer risk. There was no evidence of any association. Nine cohort and eight case-control studies considering occupational exposure to hair dyes and lymphoid neoplasms were reviewed. In the cohort studies, a total of 100 lympboid neoplasms was observed compared with 84.4 expected (RR 1.2). The RR estimates were 1.5 for non-Hodgkin's lympbomas (NHL, 17 observedvs11.2 expected) and 1.1 for multiple myeloma (MM, 19 observed casesvs16.8 expected). Interpretation of case-control studies of occupational exposure is seriously hampered by the small number of exposed cases. Five case-control studies considered personal use of hair dyes and the risk of lymphoid neoplasms. Of these, three reported some association, particularly with NHL and MM. However, the RR estimates were only moderately above unity, and inadequate allowance was made for potential confounding factors, including social class and greying hair, which could be correlates of both hair dye use and lymphoid neoplasms. Further, there is little information on the biodistribution and bioavailability of potential carcinogens in hair dyes, particularly their concentrations in lymphoid tissue. These findings, therefore, require further research, particularly since they may be influenced by selective publication of positive findings (publication bias). None of the other neoplasms extensively studied, including breast, skin and lung was related to hair dye use.