Twenty-six diabetic ketoacidotic patients were treated with 3 different intravenous insulin regimes. Group (A) received 50 U initially and at 2 h intervals. Groups (B) and (C) were given continuous infusions of 10 and 2 U per hour respectively without added albumin. In addition, Group (C) received a loading dose of 3 U. The dosages were reduced when serum glucose declined to 300 mg/100 ml. Criteria for admission to the study included a plasma glucose above 350 mg/100 ml, plasma bicarbonate less than 9 mmol/l, serum ketone-bodies detectable by nitroprusside test at 8-fold or greater dilution, and arterial pH less than 7.3. The rate of normalization of blood glucose, bicarbonate, ketone bodies, and pH did not differ between Group (A) and (B). In contrast, the changes in pH, glucose, and ketone-bodies were significantly slower in Group (C). Two patients of Group (C) had worsening of these biochemical parameters during the first 6 h. They were treated successfully with regimen A. At 2 h, plasma immunoreactive insulin concentrations were 47±15, 135±19, and 25±3 μU/ml in previously untreated patients in Groups (A), (B) and (C), respectively. Potassium requirements to maintain adequate blood levels were significantly higher in Group (A). The data demonstrate that 10 U/h infusion of insulin is as effective as 50 U administered intravenously every 2 h. The amount of insulin infused should be reduced to 5 U/h when plasma glucose has declined to 300 mg/100 ml. The recovery is slow, plasma insulin concentration is inadequate and treatment failure may occur with very low insulin doses (2 U/h).