Studies have been performed using an on-line computer system programmed for blood gucose control of insulin and dextrose infusion (artificial pancreas). The aim of these studies was to test performance of the artificial pancreas and to suggest directions for future optimisation. Blood glucose stabilisation studies of diabetic volunteers were extended throughout the day and included three main meals and light exercise periods. Monitoring of blood glucose profiles of the same diabetics after depot insulin were performed on a separate occasion for comparison. The presence of insulin antibodies did not impair operation of the artificial pancreas. Most of the insulin infused by the artificial pancreas was to initially correct hyperglycaemia with relatively little required to subsequently maintian euglycaemia. The afternoon intra-meal average infusion rate was 0-9 U/hr. It is suggested that correction of fasting hyperglycaemia and maintenance of euglycaemia in diabetics be treated as separate control problems for the artificial pancreas. The overall ability of the artificial pancreas to control blood glucose to a degree not attainable by conventional insulin therapy is confirmed, in this case under conditions which include patient activity.