Large interindividual variation in maximal heart rate (HRmax) may not be accounted for by age alone. In order to evaluate a novel method in the prediction of HRmax, this study examined the profile of HR variability (HRV) during exercise in 40-67 years old men (n = 74). R-R intervals were recorded during supine rest and during a graded exercise test by cycle ergometry until exhaustion. A third-degree polynomial function was fitted to the HRV data recorded during exercise to represent the HRV profile of each subject. The instantaneous beat-to-beat R-R interval variability (SD1), high (HF) and low frequency power decreased between all consecutive exercise intensities (P < 0.033). The relationship between HRmax and a variable illustrating the declining rate of HF (HRHF50%) was stronger (r = 0.50, P < 0.001) than between HRmax and age (r = -0.36, P < 0.01). The regression analysis showed that a more accurate estimation of HR(max) was attained when HRV was used in the equation in addition to age: HRmax = 160.633-0.603(age) + 0.441(HRHF50%) (SEE = 9.8 bpm vs. 11.6 bpm in the equation based on age alone). The decline of HRV during incremental exercise seems to be useful for accurate estimation of maximal heart rate.