Heart rate (HR) as an estimator of oxygen consumption (VO(2) ) usually requires HR to be individually calibrated in a separate test. This study examined the validity of a new HR - and HR variability-based method (Firstbeat PRO heartbeat analysis software) in the estimation of VO(2) in real-life tasks. The method takes into account the respiration rate determined from HR variability and the differences in the on/off dynamics of HR and VO(2) , and no calibration tests are needed. Ten men and nine women performed 25 tasks representing different types of daily activities. Portable devices were used to measure R-to-R intervals (ECG), VO(2) and respiration rate. In pooled regression analysis, the estimated VO(2) accounted for 87% of the variability in the actual VO(2) , SEE 3·5 ml min(-1) kg(-1) (1 MET). At group level, the method underestimated slightly the measured VO(2) (mean difference - 1·5 ml min(-1) kg(-1) or - 0·4 METs). Some of the values at low exercise intensities were markedly underestimated, but the agreement was better during light and heavy activities. The limits of agreement for the data were from -8·4 to 5·4 ml min(-1) kg(-1) or from -2·4 to 1·5 METs. At individual level, the average deviations of the predicted VO(2) ranged from -1·0 to 0·6 METs and R(2) from 0·77 to 0·94, respectively. The present data indicate that the prediction method may be considered sufficiently accurate to determine the average VO(2) in field use, but it does not allow precise estimation of VO(2) .