1. Animal studies suggest that the heart-rate-lowering effect of vagal stimulation during atrial fibrillation is due to: (1) a direct depressant effect on atrioventricular node conductivity, (2) enhancement of concealed atrioventricular nodal conduction of atrial impulses through augmenting fibrillatory activity, thereby indirectly prolonging atrioventricular nodal refractoriness. The purpose of the present study was to analyse these effects in man. 2. Sixteen patients with chronic atrial fibrillation were studied. After administration of propranolol (0.2 mg/kg intravenously) baseline ventricular rhythm was recorded (500 R-R intervals). Recordings were repeated after methylatropine (0.02 mg/kg intravenously). The shortest R-R interval was taken to represent atrioventricular nodal refractoriness. The ratio of the longest to the shortest R-R interval and the coefficient of variation of R-R intervals were used as parameters of concealed conduction. 3. Methylatropine foremost shortened long R-R intervals: values for the mean, shortest and longest R-R intervals decreased from 834 to 685 ms (-18%) (P < 0.001), 573 to 498 ms (-13%) (P < 0.001) and 1228 to 924 ms (-25%) (P < 0.001), respectively. Accordingly, the ratio of the longest to the shortest R-R interval decreased: 2.12 to 1.89 (-11%) (P < 0.05). Also, the coefficient of variation decreased: 0.24 to 0.20 (-17%) (P < 0.05). 4. This study supports the contention that vagal stimulation lowers ventricular rate during atrial fibrillation both by exerting a direct effect on the atrioventricular node and by augmenting concealed conduction.