Diabetic cardiovascular autonomic neuropathy (CAN) carries an increased risk of mortality. The early detection and characterization of CAN has traditionally been based on the results of autonomic reflex tests (AFTs). A variety of different measures to quantify 24-hour heart rate variability (HRV) have recently been introduced, but their normal ranges, reliability, and validity in patients with CAN have not been adequately studied. We established the normal ranges of statistical (SDNN index, CV, SNN50, RMSSD), geometric (triangular index (TI), triangular interpolation (TINN), top angle index [TAI]), frequency domain (spectral power in the VLF, LF, and HF bands, LF/HF ratio, LF in normalized units [NU]), and non-linear measures (CV1 and CV2 of the Poincaré plot) of 24-hour HRV in 94 healthy control subjects. Day-to-day reproducibility was evaluated on two occasions in 17 healthy subjects and 9 diabetic patients. The parameters of HRV were computed over time periods representing the day (6:00-24:00 hours), night (00:00-6:00 hours), and 24 hours in total. The results of all indexes, except for the LF/HF ratio and LF-NU, declined significantly with increasing age (p<0.05), but were independent of sex and BMI. The statistical, geometric, and non-linear measures (p<0.05), but not the frequency-domain parameters decreased significantly with increasing heart rate. Since the HRV data showed log normal distribution, log transformation was used to define the age-related lower limits of normal at the 2.5th centile. Intraindividual reproducibility was highest for the geometric measures. The nonlinear and statistical parameters also showed high reliability, except for the SNN50. The repeatability of the frequency domain measures was somewhat lower but still satisfactory. Reproducibility was lower in the diabetic than in the control group, higher during the day than during the night, and better than that reported previously for the AFTs. In conclusion, in healthy subjects the measures of 24-h HRV are not related to sex or BMI, but strongly dependent on age and heart rate, the latter except for the frequency domain measures. The majority of the HRV measures, in particular the geometric parameters, show a relatively high intraindividual reproducibility which underlines their suitability for the use in prospective studies.