We have studied the hyperglycaemic effect of the carbohydrate of glucose, sucrose, and honey equivalent to 20 g in twelve normal volunteers, eight patients with insulin-dependent diabetes mellitus (IDDM), and six patients with non-insulin-dependent diabetes mellitus (NIDDM). Honey produced an attenuated postprandial glycaemic response in normal volunteers (vs glucose p less than 0.005; vs sucrose p less than 0.05) and IDDMs (vs glucose p less than 0.005; vs sucrose p less than 0.05). The glycaemic index (GI) showed considerable variability within each subject group. Combined with a peak incremental index (PI), the two indices appear to be more valuable in predicting the glycaemic effects of carbohydrates rather than either one alone. We suggest that honey may prove to be a valuable sugar substitute in diabetics, and that both the GI and PI should be used in the analysis of food.