A data reduction technique is presented which enables the patterns of cardiac rhythm inherent in a long-term ECG recording to be described in a concise fashion. The method consists in combining an R-R interval difference histogram, which gives information about the detailed structure of cardiac rhythm patterns, with an R-R interval histogram, which concisely describes the trends in heart rate over a period. Examples of the histogram patterns corresponding to some common cardiac rhythms are presented. The histograms can be treated mathematically to yield information which is useful in assessing the long-term effects of drugs and in classifying disorders of cardiac rhythm.