Four hundred and ninety pancreas cancer patients representative of confirmed cases in Los Angeles County residents of working age were compared to healthy controls individually matched by age, sex, race, and neighborhood. Home interviews were conducted on occupation, smoking, food and beverage consumption, and medical history. Cigarette smoking was a strong and consistent predictor of pancreas cancer occurrence; the effect disappeared after a decade of nonsmoking, and there was no increase in risk among current smokers as daily dose increased. There was no link between pancreas cancer and past consumption of tea, carbonated beverages, beer, or spirits; and an association with coffee consumption was inconsistent. A strong association between pancreas cancer and history of subtotal gastrectomy at any past time could not be explained by chance or any other factor. Pancreas cancer patients had experienced fewer allergies of any kind.