Dietary factors are widely studied as risk factors for colorectal cancer, with much information from case-control studies. We evaluated the validity of dietary data from a retrospective case-control study of diet and colorectal cancer. As part of the alpha-Tocopherol, beta-Carotene Cancer Prevention Study, diet was assessed at baseline and after diagnosis for colorectal cancer cases and at baseline and regularly during the trial for a random control group. The dietary assessment referred to the previous 12 months (in cases before diagnosis). In the two dietary assessments, the cases reported a greater increase in consumption of fruits and dairy products and a decrease in consumption of potatoes. Accordingly, relative risks for colorectal cancer by baseline dietary data differed markedly from odds ratios from case-control data; e.g., relative risk for a 652-mg increase in calcium intake was 0.79 (95% confidence interval = 0.48-1.30) in case-cohort analysis vs. an odds ratio of 1.57 (95% confidence interval = 1.06-2.33) for case-control analysis. The most likely explanation is the influence of current diet on recall of prediagnosis diet and effects of occult cancer on diet in the year before cancer diagnosis, which have implications for interpretation of case-control studies in evaluating associations between diet and colorectal cancer.