Six trained male cyclists and six untrained but physically active men participated in this study to test the hypothesis that the use of percentage maximal oxygen consumption (% VO2max) as a normalising independent variable is valid despite significant differences in the absolute VO2max of trained and untrained subjects. The subjects underwent an exercise test to exhaustion on a cycle ergometer to determine VO2max and lactate threshold. The subjects were grouped as trained (T) if their VO2max exceeded 60 ml.kg-1.min-1, and untrained (UT) if their VO2max was less than 50 ml.kg-1.min-1. The subjects were required to exercise on the ergometer for up to 40 min at power outputs that corresponded to approximately 50% and 70% VO2max. The allocation of each exercise session (50% or 70% VO2max) was random and each session was separated by at least 5 days. During these tests venous blood was taken 10 min before exercise (-10 min), just prior to the commencement of exercise (0 min), after 20 min of exercise (20 min), at the end of exercise and 10 min postexercise (+10 min) and analysed for concentrations of cortisol, [Na+], [K+], [Cl-], glucose, free fatty acid, lactate [la-], [NH3], haemoglobin [Hb] and for packed cell volume. The oxygen consumption (VO2) and related variables were measured at two time intervals (14-15 and 34-35 min) during the prolonged exercise tests. Rectal temperature was measured throughout both exercise sessions. There was a significant interaction effect between the level of training and exercise time at 50% VO2max for heart rate (f(c)) and venous [la-]. At 70% VO2max, VO2 and ventilation (V(E)) for the T group and VO2, V(E) and carbon dioxide production for the UT group increased significantly with time and there was a significant interaction effect for f(c), [la-], [Hb] and [NH3]. The change in body mass at 50% and 70% VO2max was significantly greater in the T group. The present study found that when two groups of male subjects with different absolute VO2max exercised at a similar percentage of VO2max some effector responses were significantly different, questioning the validity of selecting % VO2max as a normalising independent variable.