Recently diabetic patients have been encouraged to increase their carbohydrate intake, but exact details of which foods to use are lacking. To determine whether sufficiently large differences existed to justify more specific dietary advice, we compared the glycaemic responses to 50 g carbohydrate portions of different foods, taken as breakfast test meals by groups of five to seven diabetic patients. Two- to threefold differences were seen amongst the 15 foods tested. The glycaemic responses for spaghetti, 'All-bran', rice and beans were significantly below those for bread, while 'Cornflakes' were above. Factors predicted to influence this were without effect, including: substituting wholemeal for white bread, increasing substantially the simple sugars (using 'All-bran' or bananas instead of wholemeal bread) and doubling meal protein by adding cottage cheese to bread. Paired comparisons of the glycaemic response to the five legumes with those of the seven other starchy foods (breads, spaghetti, rice, Cornflakes, oatmeal porridge and potatoes) showed that the mean peak rise in blood glucose concentration and mean area under the glucose curve after beans were 23 and 28% lower, respectively, than the mean for the other foods (p less than 0.001). Such results suggest a potentially valuable role for dried leguminous seeds in carbohydrate exchanges for individuals with impaired carbohydrate tolerance. These large differences in blood glucose response to different food cannot at present be predicted directly from tables of chemical composition. Nevertheless, physiological testing may both aid in understanding the factors responsible and help selection of the appropriate carbohydrate foods for the diabetic diet.