Our aim was to estimate the association between smoking history, alcohol and tobacco smoking and tumours of the large intestine. Associations were studied at an early stage of colorectal cancer in order to avoid bias in the information. In order to estimate the link between adenoma and cancer the exposures were analysed separately for cancer and adenoma patients. The study was conducted as a case-control study within a randomized trial for colorectal cancer among males and females aged 45–74 years. Cases initially included all individuals with a positive Haemoccult-II test in three screens and an age- and sex-matched reference group was selected from the test negatives. Subsequent colonoscopy defined the final case group, which consisted of 49 colorectal cancer patients, 171 with adenoma and 177 test positives with no diagnosis or with non-adenomatous polyps or haemorrhoids. Controls were 362 age- and sex- matched test negatives. Data were collected by blind telephone interviewing before the first clinical examinations of test positives. Smoking history, coffee or alcohol intake were not statistically significantly associated with colorectal cancer. For adenomas, the odds ratios (OR) were between 2.0 and 2.7 in all smoking categories. For smokers with>40 years duration OR=2.7 195% confidence interval (Cl) : 1.6–4.7). Coffee consumption showed a clear protective ef fect. Consumers of 4–7 cups per day had an OR of 0.5 (95% Cl: 0.3–0.8) and heavy consumers of 8 cups had an OR of 0.3(95% Cl: 0.1–0.6). Neither tea nor alcohol consumption was related to adenoma risks. The differences in risk fac tors between colorectal cancers and adenomas may be taken as indirect evidence for a partly different or larger set of component causes necessary for the development of cancer, or that even very early and mainly asymptornatic cancers modify common lifestyle factors. The findings could, however, also be explained by differential selection of the case groups due to a possible link between the exposures and occult bleeding from the lesions.