Rational treatment of diabetic polyneuropathy depends upon establishing its cause, which is at present unknown. A number of animal models of diabetes have been examined and although abnormalities are detectable in the peripheral nervous system they do not duplicate the degenerative neuropathy encountered in the human. The relevance of these abnormalities is therefore uncertain, although they may reflect the earlier changes in man. For human neuropathy, it is likely that vascular lesions or an abnormal susceptibility to mechanical injury are responsible for focal neuropathies. The evidence that ischaemia and hypoxia are responsible for the diffuse sensory neuropathy and autonomic polyneuropathy is still equivocal and it is often difficult to establish whether the vascular changes are primary or secondary. Metabolic explanations, such as sorbitol accumulation in nerve, have not so far been adequately validated by responses to treatment. The manifestations of diabetic neuropathy are complex and a single explanation should not be sought.