Although the beneficial effects of ventricular rate (VR) slowing during atrial fibrillation (AF) are axiomatic, the precise relationship between VR and hemodynamics has not been determined. We hypothesized that selective atrioventricular node (AVN) vagal stimulation (AVN-VS) by varying the nerve stimulation intensity could achieve precise graded slowing and permit evaluation of an optimal VR during AF. The aims of the present study were the following: 1) to develop a method for computerized vagally controlled VR slowing during AF, 2) to determine the hemodynamic changes at each level of VR slowing, and 3) to establish the optimal anterograde VR during AF. AVN-VS was delivered to the epicardial fat pad that projects parasympathetic nerve fibers to the AVN in 14 dogs. Four target average VR levels, corresponding to 75%, 100%, 125%, and 150% of the sinus cycle length (SCL), were achieved by computer feedback algorithm. VR slowing resulted in improved hemodynamics and polynomial fit analysis found an optimum for the cardiac output at VR slowing of 87% SCL. We conclude that this novel method can be used to maintain slow anterograde conduction with best hemodynamics during AF.