Breathing rates during physical exercise suggest that, during these conditions, the high-frequency (HF) bandwidth of heart rate variability (HRV) analysis should be extended beyond conventional guidelines. However, there has been little investigation of the most appropriate choice of HF bandwidth during exercise. HRV analysis was performed in 10 males and six females during progressive bicycle exercise. Cardiac cycle (RR) interval and breath-by-breath respiratory data were simultaneously recorded. HRV powers were determined for the band-limited ranges 0.04-0.15 Hz [low-frequency (LF)], 0.15-0.4 Hz (HF 0.4) and 0.15-bf Hz (HF bf, where bf represents maximum breathing frequency). Mono-exponential functions described the relationship between HRV and work rate for each bandwidth (r=0.92-0.95) and were used to calculate the "HRV decay constant" (work rate associated with a 50% reduction in HRV power). The HRV decay constants for each bandwidth were linearly related to maximal work rate (r>0.71; P<0.001) and were substantially greater in males than in females (P<0.001). There was a significant difference between the HRV decay constants for HF 0.4 and HF bf (P<0.005) in both genders. The HRV decay constants for the LF and HF bf bandwidths appear to provide an indication of work capacity from submaximal exercise, without prior assumption regarding heart rate and its relationship with work rate.