The study investigated the differences in five different formulae for heart rate correction of the QT interval in serial electrocardiograms recorded in healthy subjects subjected to graded exercise. Twenty-one healthy subjects (aged 37+/-10 years, 15 male) were subjected to graded physical exercise on a braked bicycle ergometer until the heart rate reached 120 beats/min. Digital electrocardiograms (ECG) were recorded on baseline and every 30 seconds during the exercise. In each ECG, heart rate and QT interval were measured automatically (QT Guard package, Marquette Medical Systems, Milwaukee, WI, USA). Bazett, Fridericia, Hodges, Framingham, and nomogram formulae were used to obtain QTc interval values for each ECG. For each formula, the slope of the regression line between RR and QTc values was obtained in each subject. The mean values of the slopes were tested by a one-sample t-test and the comparison of the baseline and peak exercise QTc values was performed using paired t-test. Bazett, Hodges, and nomogram formulae led to significant prolongation of QTc intervals with exercise, while the Framingham formula led to significant shortening of QTc intervals with exercise. The differences obtained with the Fridericia formula were not statistically significant. The study shows that the practical meaning of QT, interval measurements depends on the correction formula used. In studies investigating repolarization changes (e.g., due to a new drug), the use of an ad-hoc selected heart rate correction formula is highly inappropriate because it may bias the results in either direction.