Power spectrum analysis of the R-R interval was used in 20 controls and in two groups of type I (insulin dependent) diabetics (27 patients) to detect changes in total power or in its components (low frequency and high frequency) that might be considered an early evidence of impairment of cardiac autonomic nervous system control. A significant difference between controls and severe diabetics (with autonomic involvement) was found in all components. In the early stage of diabetes without evidence of autonomic involvement, an absolute reduction of the low frequency component in the standing position and a significant reduction of the percentage increase compared with the lying position, was found to discriminate diabetics from controls. The occurrence of somatic neuropathy was unrelated to changes in autonomic function. These data indicate that: (1) power spectrum analysis is sensitive enough to detect cardiac autonomic neuropathy in diabetics, where standard methods fail; (2) power spectrum analysis is the method of choice in the early stages; (3) in severe type I dependent diabetes there is a reduction of power spectrum analysis total power and a defective response to standing up; (4) cardiac autonomic neuropathy develops independently from somatic neuropathy.