A randomized prospective study was conducted to determine whether the insulin regimen for NIDDM subjects poorly controlled on oral therapy should be designed primarily to control basal metabolism or to control mealtime hyperglycaemia. Grossly obese subjects were excluded. After a 2-month run-in phase involving intensive education, subjects were randomized to therapy with twice daily isophane or three times daily soluble insulin. Both Protaphane and Actrapid brought about similar improvement in HbA1 (9.5 +/- 0.5 and 9.7 +/- 0.4%) compared with baseline (11.7 +/- 0.5%; p < 0.001). Diurnal blood glucose profiles showed that despite the good post-prandial control achieved by pre-meal soluble insulin, loss of control occurred overnight, resulting in higher fasting blood glucose levels compared with Protaphane therapy (8.0 +/- 0.8 vs 10.6 +/- 0.8 mmol l-1; p < 0.05). The overall rate of hypoglycaemia was 0.44 patient-1 year-1. Thirty-two mild hypoglycaemic episodes occurred on Protaphane therapy and 79 on Actrapid therapy. Using formal psychometric tests it was shown that insulin therapy was associated with improved treatment satisfaction and that this was greater on Protaphane therapy (p < 0.05). Overall well-being increased similarly in the two groups. All subjects wished to continue with insulin therapy after the conclusion of the study. The insulin regimen for moderately or poorly controlled non-insulin-dependent diabetes should primarily be designed to correct the basal insulin deficiency rather than to mimic normal meal-induced insulin secretion.