Response-surface regression analysis was used to study dietary levels of fiber, carbohydrate, lipid and protein to minimize serum and liver cholesterol and triglyceride levels and maximize serum high-density lipoprotein-cholesterol levels of male weanling rats. Because the dietary components were not statistically independent, they were studied in combinations of three variables. The three-variable combinations were the most useful in locating the desired maximum or minimum lipid responses in terms of the proportions of the dietary components. These analyses indicated that dietary carbohydrate, lipid and protein were better than dietary fiber for predicting the serum and liver lipid response levels. Response-surface contours and three-dimensional plots were developed for each lipid response except serum triglycerides, which were not predictable. The contours and three-dimensional plots were used to help determine those combinations of the diet components that would produce the desired maximum or minimum lipid responses. The statistical analyses indicated that the desired lipid response levels could be attained with a diet consisting of 3-5% neutral detergent bran fiber, 6-10% lipid, 54-55% carbohydrate, 26-30% protein and 4.7% vitamins and minerals.