Current systems for analyzing ambulatory electrocardiograms (ECGs) are unable to distinguish precisely between sinus rhythm and atrial fibrillation (AF) episodes, and are unable to produce RR interval listings that distinguish AF from sinus rhythm on a beat-to-beat basis. We describe a method for obtaining such a computerized listing ("Composite Rhythm" file) from ambulatory recordings containing episodes of AF. The file lists the rhythm of each beat, its real time, and the QRS complex morphology. A visual inspection is made of a full printout of the recording to identify the precise time of onset and termination of each episode of AF. These times are entered into a computer and identified with the corresponding beats on a conventional RR interval file generated by Holter analysis. The method was validated using 1-hour segments from 20 ambulatory ECGs containing 145 episodes of AF. These were visually identified by four independent observers with a mean sensitivity of 99.1%. The first beat of AF was identified concordantly in 96% of episodes, with a discrepancy of < or = 3 beats in the other episodes. The times of 200 selected QRS complexes were then entered into the computer by each observer; 91.1% of these complexes were identified exactly and 100% were identified to within one beat. The Composite Rhythm files have several potential applications for testing AF detection algorithms and studying the mode of onset of AF.