The blood glucose of 10 stable unstressed uncomplicated insulin-dependent diabetics (age: 39 ± 4 years; duration of diabetes: 11 ± 1 years) was controlled by a preprogrammed open-loop system of i.v. insulin infusion during two randomized consecutive days. Individual constant basal insulin rates were determined by adapting the rate of the pump to obtain normoglycemia during an overnight infusion. The post-prandial doses of insulin, given as square waves (12 U/h) or boluses (120 U/h), were calculated on one day (type I programme) from the fractional individual insulin requirements during a previous external artificial pancreas run, and on the other day (type II programme), from the formula: constant basal rate (U/h) × amount of carbohydrates (g) × 0.4 if for breakfast, on × 0.15 if for midday and evening meal. Blood glucose control, which was similar for the two methods, was identical to that provided by the artificial pancreas closed-loop system for correction of post-prandial glycemic excursions (mean peak values: 7.0 ± 0.4, 7.2 ± 0.1, 7.0 ± 0.1 mmol/l for type I, type II and closed-loop programmes). Basal blood glucose was significantly higher with the open-loop system (7.2 ± 1.1, 6.8 ± 0.8, 4.8 ± 0.3 mmol/l for type I, type II and closed-loop programmes) because of a drift of the values during the first half of the night. Boluses gave a slightly better post prandial control than square waves but produced a late hyperglycemic rebound. Daily insulin doses required by type I, type II and closed-loop programmes were respectively 75 ± 7, 64 ± 5 and 80 ± 7 U. Individual basal rate was uniform (1.1 ± 0.1 U/h). From these results it is concluded: (1) the quality of blood glucose control obtained from open-loop and closed-loop systems of insulin infusion is remarkably similar; (2) our formula is adequate for preprogramming insulin for the prandial periods, but careful individual determination of the basal rate by use of a closed-loop system is required.