A case-control study was carried out in Geneva on the basis of an interview involving 88 new cases of pancreatic cancer and 336 controls. The percentage of histologically confirmed cases was 67 p. 100. Controls were drawn at random from the general population. An excess risk was observed with increased consumption of fatty foods. This excess, however, was only significant for butter; in contrast, margarine was associated with a reduction in risk (relative risk = 0.35, p less than 0.05). Meat intake did not modify the risk, except for lean pork, for which the risk was significantly reduced. For carbohydrates, an excess risk was observed, with a dose-response effect. Fresh fruit and vegetables led to a marked and significant reduction in risk. Among beverages, beer was the only one for which consumption entailed an increased risk (RR = 2.85, p less than 0.05). The results concerning tea and coffee were not significant, but there appeared to be a trend towards an increase in relative risk.