Despite a well-established prognostic value in cardiac patients, HRV (heart rate variability) indexes have been used little in the clinical setting. Poor reliability of the measurements might be a possible explanation for this. In the present study, we assessed the reliability of short-term HRV indexes in post-MI (myocardial infarction) patients. We studied 61 MI patients [50 males; age, 59+/-8 years; and LVEF (left ventricular ejection fraction), 46+/-6%; values are means +/-S.D.],who underwent a 5+5 min ECG recording during spontaneous and paced breathing on two consecutive days. Standard time-domain [SDNN (S.D. of RR interval values) and RMSSD (root- mean-square of successive RR interval differences)] and frequency-domain [LF (low-frequency) and HF (high-frequency) power, and LF/HF] indexes of HRV were computed. Absolute and relative reliability were assessed by the 95% limits of random variation and by the ICC (intra-class correlation coefficient). The agreement between the two measurements in classifying patients at low or high risk, according to different cut-points, and the sample size needed to detect a clinically relevant change, were also assessed. During spontaneous breathing, individual changes in test-retest measurements ranged from -41 to + 61% (SDNN; best case) and from -76 to + 316% (LF/HF; worst case). The ICC ranged from 0.72 to 0.81. Most patients (79-90 %) were assigned to the same class by the two measurements. Paced breathing did not improve reliability. In conclusion, short-term HRV parameters in MI patients may have large day-to-day variations, making the detection of treatment effects in individual patients difficult; however, the ICC values and the analysis of the consistency of classification between repeated tests indicate that HRV measurements fulfill the criteria required to be used for diagnostic or classification purposes.