In order to assess the autonomic nervous system activity in normal subjects and in diabetic patients affected by different degrees of diabetic autonomic neuropathy, we applied autoregressive power spectral analysis to 35 diabetic patients and 15 normal controls. This analysis was then compared to conventional methods by CAN score tests. Each subject was placed on a bed and connected to electro-cardiographic electrodes. After 15-minute rest in a sitting position, the electrocardiogram (ECG) recorded 512 heartbeats in a supine position. Thereafter, the same measure was taken in a standing position. During the test, the subjects quietly breathed in synchronosity with a 15/min (0.25 Hz) metronome signal to obtain stationary respiratory activity without frequency change or phase drift. R-wave detection by fast peak detection algorithm and spectrum computation by Fast Fourier transform enabled the study of the power spectrum of heart rate fluctuations. The power of fluctuations at different frequencies was the result of sympathetic and vagal input into the sinoatrial node. The autoregressive power spectral density of RR interval variability contained 2 major components: a high frequency (0.25 Hz), which is a quantitative marker of cardiac vagal activity, and a low frequency (<0.15 Hz), which is a quantitative marker of sympathetic activity with vagal modulation. A marked reduction in HF spectral density was found in diabetic patients relative to the normal controls (p<0.001). When neural activity was provoked through standing, less decreased HF spectral density and less increased LF spectral density were noted in diabetic patients relative to the normal controls (p<0.01). When diabetic autonomic neuropathy was advanced, HF and LF spectral density changes were lost. This method is simple, repeatible, objective, and quantitative. It may facilitate the screening of diabetic patients for autonomic neuropathy and enable convenient quantitative fallow-up.