Two groups of pregnant diabetic women, fifteen with type I and fourteen with type II diabetes, were randomly assigned either to CSII or to ICT and the subgroups compared with respect to glycaemic control, insulin requirement and perinatal out-come. Ten non-diabetic pregnant women served as controls for the variations in the metabolic parameters considered (24-hour mean blood glucose and glycosylated hemoglobin). Strict glycaemic control was achieved and maintained by both regimens before week 13 in all patients with type I and in 57.1% of patients with type II diabetes. The mean insulin requirements in the type I group increased up to week 34-36 and then stabilized to term in patients receiving CSII and rose progressively to term in those receiving ICT. In the type II group insulin requirements rose up to week 36 in patients receiving CSII and up to week 32 in those receiving ICT, stabilizing thereafter on both regimens. No significant differences in mean insulin requirement at the different stages of gestation were found between the patients receiving CSII and those receiving ICT of either group. Perinatal outcome was satisfactory in both groups, although control of foetal growth was better with ICT than with CSII. CSII is a practical, safe and effective method of maintaining maternal normoglycemia in pregnancy but for the present we cannot consider it superior to ICT in the treatment of pregnant diabetic women.