The present study was undertaken to test the hypothesis that the human RR-QT relationship during dynamic exercise differs markedly from that during the recovery phase. Fourteen subjects from the age of 16 to 71 years exercised on a treadmill according to the Bruce protocol. Electrocardiograms were recorded continuously on a magnetic tape, from 1 minute before exercise to 10 minutes into recovery. An exponential formula, proposed by us earlier, closely represented the exercise RR-QT data. However, it was not appropriate for the often S-shaped recovery curves which invariably deviated from the exercise curves, exhibiting hysteresis. Initially, all recovery QT intervals were shorter than the exercise values, but later in the recovery, some crossed the exercise curves from below, resulting in longer QT intervals. The recovery data were fitted by a third degree polynomial, and the hysteresis was calculated as the area between the exercise and recovery curves within a 150 ms range of the RR interval starting from its minimum value. The mechanisms for the occurrence of hysteresis are likely to involve the sympatho-adrenal activity in the early post-exercise period and the time course of QT interval adaptation to rapid changes in the RR interval.