The aim of this study was to compare the dependence of heart rate variability (HRV) on heart period (RR interval length) under different physiological and pathological states in order to detect changes in HR modulation. The dependence of HRV on the RR interval length in healthy elderly subjects, congestive heart failure (CHF) patients and one patient with a transplanted heart (T) was compared with healthy young subjects. Spectral powers, sample entropy (SampEn) and short-term fractal scaling exponent (alpha1) were determined from 24 h free-running recordings. For the same HR, HRV measures were different in different groups. In healthy subjects HRV measures depended on RR interval length and all spectral powers were highly correlated, although reduced in elderly subjects. SampEn at high HR was the most sensitive quantity to changes induced by aging. In disease, CHF and T, an achievable HR range was decreased, all spectral powers were reduced, but correlated, and the dependence of HRV measures on RR was lost. There was an evident difference in the dependence of nonlinear on linear measures between young subjects and all the other studied groups. In disease the reduction in autonomic control was associated with the decrease in short-range correlation and regularity in RR series. We have concluded that the analysis of HRV measures as functions of RR interval length can reveal important aspects of HR control that might be lost in averaging.