High-frequency components of the QRS complex, including late potentials, can be analyzed by signal averaging (SA). However, this method may fail to detect transient changes as a result of cancellation. The wavelet transform, which has a superior time-frequency resolution, was used to analyze beat-to-beat changes of the QRS components in 50 normal subjects and 50 patients who showed positive late potentials. The transformed data, displayed in three dimensions and in color, were highly reproducible in each patient. Measurement of high-power duration at a frequency of 50 Hz (WD50) showed a significant correlation between WD50 and filtered QRS duration in both groups. When the mean ± SD of WD50 in normal subjects was defined as normal, 96% of patients with late potentials were out of the normal range. The wavelet signals in patients with late potentials were more inhomogeneous than those of normal subjects. It is concluded that this newly developed color display, three-dimensional wavelet transform system showed extremely good time-frequency resolution in analyzing every beat without signal averaging.