We studied the effects of cardiac rehabilitation on the sympathovagal control of heart rate variability in 30 patients after a first, uncomplicated myocardial infarction. Twenty-two patients completed 8 weeks of endurance training (trained), while eight decided not to engage in the rehabilitation programme for logistical reasons, and were taken as untrained controls. Age, site of infarction, ejection fraction, ventricular diameter and stress test duration were similar in the two groups at baseline. Heart rate variability was evaluated 4 weeks after infarction before starting rehabilitation, and repeated 8 weeks and one year later in both trained and untrained patients. Measures of heart rate variability, obtained from both time- and frequency-domain analysis of a 15 min ECG recording in resting conditions, were as follows: mean RR interval and its standard deviation (RRSD), the mean square successive differences (MSSD), the percent of RR intervals differing > 50 ms from the preceding RR (pNN50), the low and high frequency components of the autoregressive power spectrum of the RR intervals and their ratio (LF/HF). At baseline, heart rate variability was similar in trained and untrained patients. In the short term (8 weeks after infarction), training increased RRSD by 25% (P < 0.01), MSSD by 69% (P < 0.01), pNN50 by 120% (P < 0.01), and reduced LF/HF ratio by 30% (P < 0.01). The effects persisted after one year in trained patients. In untrained patients, the autonomic control of heart rate variability did not change 8 weeks after myocardial infarction and was only slightly modified by time. Thus, exercise training, performed for 8 weeks after a myocardial infarction, modifies the sympathovagal control of heart rate variability toward a persistent increase in parasympathetic tone, known to be associated with a better prognosis. This may partly account for the favourable outcome of patients who undergo rehabilitation.