BACKGROUND An investigation into concerns about possible health effects of fire training practices at an Australian training facility recommended a study to investigate the risk of cancer and mortality of those with risk of chronic occupational exposure to several chemicals. AIMS To investigate mortality and cancer incidence in firefighters at the Country Fire Authority (CFA) Fiskville training facility, Victoria, Australia, between 1971 and 1999. METHODS CFA supplied human resources records, supplemented by self-reported information for a retrospective cohort, and allocated firefighters to low, medium or high groups based on probability of exposure. We linked the cohort to state and national cancer and mortality data. We calculated standardized mortality ratios and standardized cancer incidence ratios (SIRs). RESULTS The high group (n = 95) had a clearly increased risk of overall cancers SIR = 1.85 (95% CI 1.20-2.73), testicular cancer SIR = 11.9 (1.44-42.9) and melanoma SIR = 4.59 (1.68-9.99) relative to the population of Victoria. Brain cancer was significantly increased for the medium group (n = 256): SIR = 5.74 (1.56-14.7). Mortality was significantly reduced for all groups. CONCLUSIONS Dealing with supplied records can be problematic but despite the small numbers, we identified an increased risk of cancer for the high group. The mortality data suggested that there was under-ascertainment for the medium and low groups which underestimated risk and a possible reporting bias for brain cancer. Small cohorts can still provide statistically significant findings when investigating locations for cancer risk.