Abnormal electrical activation occurring during ventricular pacing reduces left ventricular (LV) pump function. Two strategies were compared to optimize LV function using ventricular pacing, minimal asynchrony and optimal sequence of electrical activation. ECG and hemodynamics aortic flowprobe, thermodilution cardiac output, LV pressure and its maximal rates of rise (LVdP/dtpos) and fall (LVdP/dtneg) were measured in anesthetized open-chest dogs (n = 7) with healthy hearts. The QRS duration (a measure of asynchrony of activation) was 47 +/- 5 ms during sinus rhythm and increased to 110 +/- 12 ms during DDD pacing at the right ventricular (RV) apex with a short AV interval. During pacing at the LV apex and LV base, the QRS duration was 8% +/- 7% and 15% +/- 7% (P < 0.05) longer than during RV apex pacing, respectively. Stroke volumes, LVdP/dtpos and LVdP/dtneg, however, were higher during LV apex (15% +/- 16%, 10% +/- 12% [P < 0.05], and 15% +/- 10%, respectively) and LV base pacing (11% +/- 12% [P < 0.05], 3% +/- 12%, and 3% +/- 11%, respectively) than during RV apex pacing. Systolic LV pressure was not influenced significantly by the site of pacing. Biventricular pacing (RV apex together with one or two LV sites) decreased the QRS duration by approximately 20% as compared with RV apex pacing, however, it did not improve stroke volumes, LVdP/dtpos and LVdP/dtneg beyond those during pacing at the LV apex alone. In conclusion, the sequence of electrical activation is a stronger determinant of ventricular function than the synchrony of activation. For optimal LV function the selection of an optimal single pacing site, like the LV apex, is more important than pacing from multiple sites.