A matched case-referent study was conducted to investigate the risk of gastrointestinal cancer in cement workers. All male cases of cancer in esophagus, stomach, colon and rectum in the period 1922-1988 from two parishes surrounding a cement plant were identified. For each case, four referents were chosen, matched to age and year of death. Employment in the cement plant was confirmed as from plant records or job title noted in the parish books. Of 25 cases with a definite colorectal tumour, 40% had been blue-collar cement workers for 25 years or more, compared with 20% of their referents [odds ratio (OR) 3.2; 95% confidence limits (CI) 1.1-9.4]. There were indications of a dose (time of employment)-response relationship. For 18 cases with an esophageal or stomach tumour, the OR for those ever employed as cement workers was 3.2 (CI 1.0-11), but short-term employment predominated, thus making a causal relationship less likely.