A hospital-based case-control study of oesophageal cancer was carried out in the Heilongjiang Province, a low-risk area for oesophageal cancer in China. From May 1985 to May 1989, 196 histologically confirmed cases and 392 controls with other (non-neoplastic) diseases were personally interviewed in the wards of 5 major hospitals. Information was obtained about usual consumption in the early 1980s of 32 major contributors to the diet in the province, socio-demographic status, smoking and alcohol consumption. Odds ratios (OR) were obtained from logistic regression models, and confounding was controlled by means of multivariate models. Smoking and alcohol consumption were major risk factors for oesophageal cancer in this population. Smokers of handmade cigarettes exhibited a particularly high risk. A near multiplicative synergism was found between smoking and alcohol consumption. There was a significant inverse dose-risk trend for combined consumption of vegetables and fruits; a 300-g increase per day lowered risk by 35%. Vitamin C intake was negatively associated with risk; a 100-mg increase per day lowered risk by 39%. Our data suggest a modifying effect of vitamin C and beta-carotene on risk associated with smoking, but the power of analyses was low. Salt, salt-preserved foods and pickled vegetables were not associated with increased risk. High temperature of meals and drinks was a strong risk indicator in this population. The strength of tea and overall tea consumption were independent determinants of the risk.