To investigate whether measurements of cortisol responses to exercise are confounded by neglect of the hormone's circadian rhythm, we measured the serum and salivary cortisol responses of eight women to 40 min of 70% maximal oxygen consumption treadmill exercise beginning at 0800 and 2000. Responses were calculated relative to the usually employed preexercise concentrations and also to concentrations at the same times of another day while subjects were at rest. Compared with areas under response curves (AUCs) calculated relative to their circadian baselines, AUCs for serum and salivary cortisol calculated by reference to preexercise concentrations were underestimated (serum, P < 0.001; salivary, P < 0.01) by 93 and 84% in the morning and by 37 and 35% in the evening, respectively. Calculated by the usual preexercise baseline method, rises in serum and salivary cortisol were similarly underestimated. More accurately calculated relative to their circadian baselines, serum and salivary cortisol AUCs were similar (P = 0.63 and P = 0.37, respectively) in the morning and evening, as were their rises (P = 0.23 and P = 0.70, respectively). In future investigations of the existence and magnitude of cortisol responses, those responses must be calculated relative to the hormone's circadian baseline.