All deaths with a mention of leukaemia on the death certificate, in men employed over a period of 25 years of 8 oil refineries in the U.K. were identified. The potential benzene exposure of these cases was compared with that of two sets of controls selected from the total refinery population. One set of controls was matched for refinery and year of birth, the other set was matched for refinery, year of birth nd length of service. No information was available on measurement of benzene in the work environment but a job history was obtained from refinery personnel records for all cases and controls. This was used to allocate each man to a benzene exposure level of "low", "medium", or "high". There was no overall excess of deaths from leukaemia when compared with the expectation from national rates. There was also no excess of cytological types of leukaemia which have been shown to be particularly associated with benzene exposure. However, the risk for those men with medium or high exposure relative to the risk for those with low benzene exposure approached a significance (P=0.05) when length of service was taken into account. If there were an increased risk of leukaemia due to benzene exposure, it could have only been one that affected a very small proportion of men within the refinery workforce.