We systematically reviewed cohort studies on the effect of nutrient and food intake (except for alcohol) on the incidence of type 2 diabetes, which had been published in English as of lay 2004. Using the MEDLINE (PubMed) database as well as reference lists of searched papers, 15 individual cohort studies (a total of 31 papers) were identified. The number of subjects (n=895-85, 060), follow-up length (5.9-23 y), the number of diabetes cases (n=74-4, 085), dietary assessment method used (simple food questionnaire, food frequency questionnaire, food frequency interview, diet history interview, and 24-h recall), and method of case ascertainment (questionnaire, oral glucose tolerance test, fasting glucose level, death certificate, and nationwide registry) varied among studies. For nutrients, intakes of vegetable fat, polyunsaturated fatty acid, dietary fiber (particularly cereal fiber), magnesium, and caffeine were significantly inversely correlated and intakes of traps fatty acid and heme-iron, glycemic index, and glycemic load were significantly positively correlated with the incidence of type 2 diabetes in several papers. For foods and food groups, several papers showed significantly decreased risk for type 2 diabetes with the higher consumption of grain (particularly whole grain) and coffee, and significantly increased risk with processed meat consumption. Because all the studies were carried out in western countries, however, research in non-western countries including Japan is needed.