The belief that there is total irregularity of the pulse in atrial fibrillation has been re-examined. In a computerised analysis of R-R intervals and pulse volumes, 100-500 (mean 237) consecutive cycles were examined in 74 patients with atrial fibrillation, of whom 36 were on digoxin and 38 were not taking any antiarrhythmic treatment. A Doppler ultrasound technique was used to assess pulse volumes, against which R-R intervals were correlated. Although the sequence of consecutive R-R intervals was random in 52 (70%), patients there was a significant correlation between consecutive intervals in 22 (30%), the correlation coefficient being negative in 11 and positive in 11. In 43 (58%) cases the sequence of consecutive pulse volumes was significantly non-random; 34 (46%) showed pulsus alternans, indicated by a negative correlation between consecutive volumes. The proportion of patients with a non-random sequence of R-R intervals or pulse volumes was the same whether or not they were taking digoxin. Thus patients with atrial fibrillation often have patterns of regularity of the pulse, with the ventricular rhythm being non-random in almost one third and the sequence of pulse volumes being non-random in over a half. Contrary to classic teaching, in many patients with atrial fibrillation the pulse is not irregularly irregular.