In atrial fibrillation, cardiac performance is dependent on both preceding RR (RR-1) and prepreceding RR (RR-2) intervals. However, relative contributions were not well defined. Left ventricular outflow peak ejection velocity (Vpe) was measured by echocardiography from 21 patients. The relation between RR-1 and Vpe could be divided into two zones; steep slope in short RR-1 intervals (<or=0.5 sec) and plateau in long RR-1 intervals (> 0.5 sec). RR-2 had a weak negative association with Vpe. The mean squared correlation coefficient (r2 ) between RR-2 and Vpe was 0.15 +/-0.13 and improved to 0.29 +/-0.21 (p<0.001), when coordinates with RR-1 <or=0.5 sec were excluded. The RR-1 was positively associated with Vpe. The mean r2 between RR-1 and Vpe was 0.52 +/-0.17 and improved to 0.72 +/-0.11 (p<0.001), when adjusted by RR-2. Simple linear regression analysis showed that mean RR interval, age, fractional shortening (FS), and mean peak velocity were negatively correlated with modified r2 between RR-2 and Vpe. Multiple stepwise regression analysis revealed that mean RR interval (r2 =0.32) and FS (r2 =0.16) were significant. In summary, simple modification could improve the relationship of both RR-1 and RR-2 with cardiac performance. RR-2 might play a more role in cardiac performance than previously expected, and when cardiac function was impaired.