Dietary information from a case-control study of pancreatic cancer conducted in Metropolitan Toronto between 1983 and 1986 is reported. A total of 249 cases and 505 population-based controls completed quantitative diet histories from which total caloric intake and the intake of a number of nutrients were estimated. A positive association with total caloric intake was observed with a relative risk of 2.39, 95% confidence interval 1.18-4.83 (highest versus lowest quartile), due primarily to the intake of carbohydrates. Inverse associations were seen with fibre from fruit, vegetable and cereal sources, with a relative of risk of 0.42, 0.22-0.78 (highest versus lowest quartile), for total fibre intake.