In atrial fibrillation (AF), beat-to-beat changes in left ventricular (LV) systolic performance are caused by variations in filling (preload), aortic pressure (afterload), and ventricular inotropic or contractile state. These factors are known to be influenced by the preceding diastolic or RR interval (RR1), but the independent impact of variations in the pre-preceding RR interval (RR2) on contractile state is not well defined. This aspect was studied in 10 patients with lone AF and 8 with coronary artery disease by measuring LV peak ejection velocity (V[pe] Doppler echocardiography) in 80 to 100 consecutive cardiac cycles. V(pe) was plotted against RR1 for beats with a short RR2 and for beats with a long RR2. Such function-interval plots indicate a direct relation between V(pe) and RR1 (for RR1 = 500 to 1,000 ms). In lone AF, the slope (linear fit) of V(pe) versus RR1 was similar for short and long RR2 (slopes = 46 and 50 s[-1]). V(pe), calculated from best linear fit and a common RR1, was consistently higher when RR2 was short than when it was long. At an RR1 = 750 ms, V(pe) (% of max) was 87 +/- 6% when RR2 was short versus 76 +/- 6% when RR2 was long, p <0.05. Results were similar in patients with coronary artery disease and the observed interval-dependent potentiation of contractile state was preserved in patients with a low ejection fraction. By comparing V(pe) at a common RR1, the effects of time-dependent changes in LV preload and afterload are minimized if not abolished. Thus, differences in V(pe) reflect differences in contractile state caused by variations in RR2. Data confirm interval-dependent alterations in contractile state that are likely an expression of the force-frequency relation. Studies of LV function in AF should incorporate a consideration of cycle length-dependent changes in LV contractile state.