A group of 35 normal weight patients with secondary failure of sulphonylurea therapy (fasting plasma glucose greater than 8.0 mmol l-1 on maximal dose of sulphonylurea) were randomly assigned to receive either a single injection of a basal insulin supplement (human ultralente insulin, n = 16) or three or four injections of a preprandial insulin supplement (human unmodified insulin, n = 19). Patients performed self-monitoring of capillary blood glucose and adjusted their insulin doses in an effort to achieve fasting and preprandial capillary glucose concentrations of less than 7.0 mmol l-1. Blood glucose control after 16 weeks of insulin therapy was improved to a similar extent by both regimens (HbA1 basal insulin group 12.5 +/- 1.2 (+/- SD) falling to 10.7 +/- 2.2%; preprandial group 12.0 +/- 1.6 falling to 9.5 +/- 1.6%). Preprandial insulin gave better control of daytime blood glucose levels but fasting plasma glucose did not differ between the two regimens (basal group 10.6 +/- 3.6, preprandial group 11.1 +/- 3.6 mmol l-1). Insulin dose was greater in the preprandial group (44.1 +/- 17.9 U day-1) than in the basal group (26.7 +/- 12.5 U day-1 (p less than 0.005), but there was no difference in the frequency or severity of hypoglycaemia between the two treatments. Only the preprandial therapy group showed significant weight gain (2.7 +/- 3.0 kg). While both regimens led to improvement of blood glucose control, these results suggest that neither basal nor preprandial insulin alone can achieve ideal blood glucose control through 24 h in patients with fairly severe failure of control on sulphonylurea therapy.