The study investigated interobserver and intrasubject reproducibility of QT interval duration and dispersion measured in standard 12-lead ECGs recorded at 25 mm/sec. Twenty-eight healthy volunteers were studied. Each underwent four ECG recordings, which were performed 1, 7, and 30 days apart. Two independent observers analyzed each ECG record. In each lead with a distinguishable T wave pattern, the RR interval, Q-peak of T interval, and Q-end of T interval were measured using a digitizing board with a 0.1-mm resolution. From each recording the following measures were derived: the maximum, minimum, and mean QT interval; maximum, minimum, and mean heart rate corrected QT interval (QTc); QT and QTc dispersion (the difference between the maximum and minimum QT interval among the 12 leads); and adjusted QT and QTc dispersion (dispersion divided by the square root of the number of leads measured). The interobserver and short-term (1 day) and long-term (1 week and 1 month) reproducibility of individual indices was assessed by computing the relative errors and comparing them by a standard sign test. In addition, the distributions of maximum and minimum QTc values among electrocardiographic leads, and the differences between QT-end and QT-peak based measurements were investigated. The results showed that: (1) the measurement of the QT interval from standard ECG recordings is feasible and not operator dependent (interobserver relative error < 4%); (2) the duration of the QT interval in healthy volunteers is stable and its short- and long-term reproducibility is high (intrasubject relative error < 6%); (3) parameters that characterize dispersion of the QT interval in the 12-lead ECG are highly nonreproducible, both between subsequent recording (relative error of 25%-35%) and between observers (relative error 28%-33%), the reproducibility of QT dispersion is significantly lower than that of QT duration (P < 0.01); and (4) the duration of the entire QT interval correlates only weakly with the duration of the Q-peak of T interval.