1. In patients with chronic heart failure, heart rate variability is reduced with relative preservation of very-low-frequency power (< 0.04 Hz). Heart rate variability has been measured without acceptable information on its stability and the optimal recording periods for enhancing this reproducibility. 2. To this aim and to establish the optimal length of recording for the evaluation of the very-low-frequency power, we analysed 40, 20, 10 and 5 min ECG recordings obtained on two separate occasions in 16 patients with chronic heart failure. The repeatability coefficient and the variation coefficient were calculated for the heart rate variability parameters, in the time-domain (mean RR, SDRR and pNN50), and in the frequency-domain: very low frequency (< 0.04 Hz), low frequency (0.04-0.15 Hz), high frequency (0.15-0.40 Hz), total power (0-0.5 Hz). 3. Mean RR remained virtually identical over time (variation coefficient 8%). The reproducibility of time-domain (variation coefficient 25-139%) and of spectral measures (variation coefficient 45-111%) was very low. The stability of the heart rate variability parameters was only apparently improved after square root and after log transformation. 4. Very-low-frequency values derived from 5 and 10 min intervals were significantly lower than those calculated from 40 and 20 min intervals (P < 0.005). Discrete very-low-frequency peaks were detected in 11 out of 16 patients on the first 40, 20 and 10 min recording, but only in seven out of 16 when 5 min segments were analysed. 5. The reproducibility of both time or frequency-domain measures of heart rate variability in patients with chronic heart failure may vary significantly. Square root or log-transformed parameters may be considered rather than absolute units in studies assessing the influence of management on heart rate variability profile. Recordings of at least 20 min in stable, controlled conditions are to be recommended to optimize signal acquisition in patients with chronic heart failure, if very-low-frequency power in particular is to be studied.