We evaluated the feasibility and effectiveness of continuous subcutaneous insulin infusion therapy (CSII) as compared to conventional injection treatment (CIT) in an ordinary diabetic clinic in a one-year randomized crossover study of 65 type I diabetic patients. Home blood glucose levels were lower during CSII (8.6 +/- 0.2 mmol/l, mean +/- SEM) than during CIT (9.1 +/- 0.3 mmol/l, p less than 0.05). During the first six months, HbA1 fell on CSII therapy (from 10.6 +/- 0.4 to 9.7 +/- 0.3%, p less than 0.001), whereas no change occurred during CIT. After the crossover, HbA1 decreased again on CSII (p less than 0.05), but rose in patients shifted from CSII to CIT (p less than 0.05). The fall in glycosylated haemoglobin during CSII correlated with the initial HbA1 level (r = 0.54, p less than 0.001). Ketoacidosis was more common during CSII (16 vs. 2 verified episodes). Hypoglycaemia occurred infrequently, without difference between CSII and CIT. Fifty-six per cent of the patients preferred CSII after the study. In conclusion, while CSII slightly improves the metabolic control, the improvement in the unselected study population is less than previously reported among highly selected patients.