Dietary, clinical, and biochemical data from the Ten-State Nutrition Survey (1968 to 1970) and the Health and Nutrition Examination Survey I (1971 to 1974), have been reexamined by factor analysis to focus attention on eating patterns as a means of relating food intake to health. The seven statistically different eating patterns generated were characterized by disproportionate consumption of different food groups. The relationship between the combination of foods that people ate and the state of their nutritional health was examined for both samples in total, and for various age, sex, race, region, and income groups within the Health and Nutrition Examination Survey I sample. Significantly different associations between the seven eating patterns and the absence of clinical symptoms and biochemical deficiencies were found. Some eating patterns consistently stood out as being significantly better or worse in this regard (p less than 0.05). This food eating pattern model should prove useful for 1) examining the association between food consumption and the incidence of disease states, such as obesity, hypertension, cardiovascular disease, cancer, and periodontal disease for various large scale dietary-health surveys, 2) establishing food regulatory policies, 3) setting national dietary goals, and 4) educating the public on nutrition and health issues.