To determine the most accurate method based on spectral analysis of heart-rate variability (SA-HRV) during an incremental and continuous maximal test involving the upper body, the authors tested 4 different methods to obtain the heart rate (HR) at the second ventilatory threshold (VT2). Sixteen ski mountaineers (mean +/- SD; age 25 +/- 3 y, height 177 +/- 8 cm, mass 69 +/- 10 kg) performed a roller-ski test on a treadmill. Respiratory variables and HR were continuously recorded, and the 4 SA-HRV methods were compared with the gas-exchange method through Bland and Altman analyses. The best method was the one based on a time-varying spectral analysis with high frequency ranging from 0.15 Hz to a cutoff point relative to the individual's respiratory sinus arrhythmia. The HR values were significantly correlated (r(2) = .903), with a mean HR difference with the respiratory method of 0.1 +/- 3.0 beats/min and low limits of agreements (around -6 /+ 6 beats/min). The 3 other methods led to larger errors and lower agreements (up to 5 beats/min and around -23/+ 20 beats/min). It is possible to accurately determine VT2 with an HR monitor during an incremental test involving the upper body if the appropriate HRV method is used.