To establish normal ranges for assessment of autonomic dysfunction, a battery of cardiovascular reflex tests was performed in 120 healthy subjects aged 15-67 years using a computer-based technique. Tests of heart rate variation (HRV) included 8 measures at rest: coefficient of variation (CV), root mean squared successive difference (RMSSD), spectral analysis of HRV in the low frequency, mid frequency, and high frequency bands in the supine and standing postures; 5 measures during deep breathing: CVb, RMSSDb, Expiration-Inspiration (E-I) difference, E/I ratio, and mean circular resultant of vector analysis; Valsalva ratio, and max/min 30:15 ratio. In addition, the change in systolic and diastolic blood pressure in response to standing and the diastolic blood pressure response to sustained handgrip were determined. The results of all measures, the blood pressure tests excepted, declined significantly with increasing age (r = -0.16 to -0.59; p less than 0.05). Moreover, RMSSD, RMSSDb, and E-I difference decreased considerably with increasing heart rate (r = -0.37 to -0.52; p less than 0.001). The longest and shortest R-R intervals in response to standing were distributed within beats 21-39 and 6-24, respectively. All tests were independent of sex. Log transformation was used to define the age-related lower limits of normal at the 2.3 centile for all tests of HRV, except for the E/I, Valsalva, and max/min 30:15 ratios. The results of these tests had to be analysed using a log(y-1) transformation. The intra-individual reproducibility determined on two consecutive days in 20 healthy subjects and 21 diabetic patients indicated that there were no major differences between the two groups regarding the day-to-day variation of test results, which was highest for the Valsalva ratio. We conclude that: (1) all indices of spectral and vector analyses of HRV are age-dependent and have the advantage of being independent of heart rate; (2) RMSSD, E-I difference, and the 30:15 ratio as it was used previously are not suitable for evaluation of autonomic dysfunction in diabetes; (3) log(y-1) transformation is required to determine age-dependent normal ranges and reproducibility for the three ratios.