Twelve C-peptide deficient Type 1 (insulin-dependent) diabetic patients with abnormal peripheral nerve function were randomly assigned to continuation of conventional insulin therapy (CIT) or to continuous subcutaneous insulin infusion (CSII). There were no statistically significant differences at entry to the study between the two treatment groups in nerve function assessed by neurologic disability score, computer assisted sensation examination and measurements of amplitudes, distal latencies, F-wave latencies and somatosensory evoked potential latencies over the spine and conduction velocities of motor and sensory fibers of ulnar, median, peroneal, tibial, plantar and sural nerves. In addition, mean plasma glucose from 24 h profiles (12.5 vs 10.6 mmol/l, respectively) and HbA1 (11.0 vs 11.6%, respectively) did not differ significantly between the two treatment groups at entry. Despite improved glycaemia from CSII in 5 patients (one dropped out of the study after 2 months) contrasted to CIT in 6 patients (5.3 vs 9.9 mmol/l, respectively, p = 0.002) and HbA1 (8.5 vs 10.7%, respectively, p = 0.002), there were no significant differences in measurements of peripheral nerve function after 4 months. After 8 months of improved glycaemia (4.4 vs 10.2 mmol/l, p = 0.004) and improved HbA1 (8.3 vs 10.5%, p = 0.002), nerve conduction (p = 0.03) and vibratory sensation threshold (p = 0.002) were significantly better in patients treated with CSII than those who received CIT. The improvements in nerve function, although small, provide further evidence that some clinical endpoints of neuropathy are favorably influenced by improved control of glycaemia.