The effect of physical training on the circadian pattern of heart rate variability (recorded over 24 h in relation to both time and frequency) was assessed in 12 chronic heart failure patients randomized, in a cross-over design, to 8 weeks training or detraining, and compared with 12 age-matched normals. Training improved heart rate variability indices: all R-R interval 5 min standard deviations increased by 17.6%, the root mean square of the differences of successive R-R intervals by 34.9%, the percentage difference between adjacent normal R-R intervals > 50 ms by 112.5%, total power by 58.3%, high frequency by 128.5% and low frequency by 65.0%. Compared with controls, circadian variations in autonomic parameters were maintained in chronic heart failure. Training-induced changes were observed at different time intervals throughout the day: the highest values were at 0100 h-0700 h (detraining: low frequency 361 +/- 83 ms2, high frequency 126 +/- 47 ms2; training: low frequency 535 +/- 202 ms2, high frequency 227 +/- 115 ms2, P < 0.01) and the lowest at 1300 h-1900 h (detraining: low frequency 91 +/- 23 ms2, high frequency 39 +/- 14 ms2; training: low frequency 154 +/- 42 ms2, high frequency 133 +/- 67 ms2, P < 0.05). In chronic heart failure, training maintains and improves circadian variations in heart rate variability measures.