Five insulin treated diabetics were studied on three consecutive days. Overnight variable intravenous insulin infusions were used before each study to maintain normoglycaemia and to calculate the optimal basal insulin infusion rate (1.1 +/- 0.1 U/h) which ws then kept constant throughout the study day. A standard 400 kCal breakfast with 25 g xylose was given at 0800 h. When the blood glucose rose above 4.1 mmol/l, an external artificial pancreas was used to infuse either extra insulin (day INS) or somatostatin for either 3h (day som) or the entire 8h experimental period (day SOM). Peak post-prandial blood glucose values were similar on all three days. The blood glucose rebounded after the cessation of the somatostatin infusion on day som. Post-prandial blood xylose peaks were lowered by somatostatin on both days but rebounded after the cessation of the somatostatin infusion on day som. The area under the plasma and urinary xylose curves was lowered by somatostatin only on day SOM. Growth hormone and glucagon levels were not statistically different on all 3 days. Thus somatostatin, when added to an optimal insulin infusion, minimised the insulin requirements by slowing intestinal absorption, but led to rebound hyperglycaemia if not feed-back controlled.