The purpose of this paper is to evaluate errors among group regressions predicting oxygen consumption (Vo2) by heart rate (HR) method, and to obtain the fittest regression for each respective age groups and possibly a single regression for the entire subject population. Twenty male adults participated in the study and were divided into three age groups. Body height, weight, surface area (SA), lean body mass (LBM) and body cell mass (BCM) were determined by the standard anthropometry and the measurement of 40K by a whole body (human) counter. At the submaximal work test, Vo2 and HR were calculated at 4 to 6 grades of work loads on a bicycle ergometer. All correlation coefficients of Vo2 to HR for groups were over 0.885 (p<0.001), while regression showed differences in slopes and intercepts between age groups. After correction applying indices of body com position and the ratio to resting HR (HRR), correlation coefficients of the old age group became greater, whereas correlation coefficients for the other two groups were almost unchanged when Vo2 were corrected by SA, LBM or BCM and related to HR. The age-difference in the group regressions disappeared in the relation between Vo2 corrected by LBM and HR. Among ten sets of relation of corrected Vo2 and HR (or HRR), only three indicated the age-difference clearly. These facts suggest the possibility of a single regression for all groups. Tentative regression equation for all the study-subjects showed the highest value of correlation coefficient, 0.936 (p<0.001) in case that Vo2 corrected by LBM or BCM was related to HRR. It was concluded that the group regression should be constructed by adopting HR (for young age groups) or HRR (for old age groups) with the correction of Vo2 by SA, LBM or BCM (for both young and old age groups). We can use additionally body weight (for only old age groups); and that a single group regression, covering an age-range up to 50 years old, is possibly applied in the relation between corrected Vo2 by LBM or 631 632 T. INAOKA and T. SUZUKI