QT dispersion was originally proposed to measure spatial dispersion of ventricular recovery times. Later, it was shown that QT dispersion does not directly reflect the dispersion of recovery times and that it results mainly from variations in the T loop morphology and the error of QT measurement. The reliability of both automatic and manual measurement of QT dispersion is low and significantly lower than that of the QT interval. The measurement error is of the order of the differences between different patient groups. The agreement between automatic and manual measurement is poor. There is little to choose between various QT dispersion indices, as well as between different lead systems for their measurement. Reported values of QT dispersion vary widely, e.g., normal values from 10 to 71 ms. Although QT dispersion is increased in cardiac patients compared with healthy subjects and prognostic value of QT dispersion has been reported, values are largely overlapping, both between healthy subjects and cardiac patients and between patients with and without adverse outcome. In reality, QT dispersion is a crude and approximate measure of abnormality of the complete course of repolarization. Probably only grossly abnormal values (e.g. > or =100 ms), outside the range of measurement error may potentially have practical value by pointing to a grossly abnormal repolarization. Efforts should be directed toward established as well as new methods for assessment and quantification of repolarization abnormalities, such as principal component analysis of the T wave, T loop descriptors, and T wave morphology and wavefront direction descriptors.