Since the first report on dual chamber pacing for congestive heart failure (CHF) in 1991, a number of investigators have explored the topic with conflicting results. These conflicts may arise from an incomplete understanding of the mechanisms by which pacing improves cardiac function. Potential mechanisms include: (1) increase in filling time; (2) decrease in mitral regurgitation; (3) optimization of left heart mechanical atrioventricular delay (left heart MAVD); and (4) normalization of ventricular activation. One or more of these mechanisms may be operative in an individual patients, implying that patients may require individual optimization. Acute pacing studies were conducted on nine CHF patients, NYHA Class II-III to Class IV. Measurements of conduction times in sinus rhythm revealed: (1) normal interatrial conduction times (59 +/- 5 ms) in all patients, with wide variations in interventricular conduction times (range, -5-105 ms); and (2) a wide range of left heart MAVD (range, 97-388 ms). While pacing the right, left, or both ventricles, measurement of high fidelity aortic pressure and mitral and aortic velocities revealed the following: (1) 6 of 9 patients increased mean pulse pressure over sinus during RV or LV pacing at an optimal AV delay; (2) the maximum aortic pulse pressure was achieved when the atrium was not paced: an 8% increase over sinus pulse pressure with paced RV versus a 5% decrease for paced atrium and RV at optimum AV delay (paired Student's t-test, P = 0.01), and a 0% increase over sinus with paced LV versus 7% decrease for paced atrium and LV at optimum AV delay, P < 0.05; (3) significant dependence on pacing site was noted, with 4 patients doing best with RV pacing, 3 patients achieving a maximum with LV pacing, and 2 patients showing no preference; and (4) 2 of 4 patients with restrictive filling patterns were converted to nonrestrictive patterns with optimum pacing. Patient hemodynamics appear to benefit acutely from individually optimized pacing. Increases in filling time, optimization of left heart MAVD, and normalization of intraventricular activation are the most significant mechanisms. Atrial pacing is inferior to atrial sensed modes if the patient has a functional sinus node.