End-systolic pressure-volume relations (ESPVRs) were analyzed in 10 closed-chest autonomically blocked dogs before and after volume loading that restored end-diastolic volume to its value measured in the control conscious state. Dogs had been previously instrumented with a left ventricular pressure micromanometer and ultrasonic crystals for measurements of major, anteroposterior, and septum-free wall diameters. Left ventricular volume was calculated with an ellipsoidal model in the left ventricular cavity. ESPVRs obtained during caval occlusion after volume loading were curvilinear as shown by the division of the relation into two parts. The initial part of the relation had a significantly smaller ESPVR slope (Ees, 12.0 +/- 1.8 mm Hg/ml) and ESPVR volume-axis intercept (Vd, - 3.5 +/- 0.8 ml) than the final part of the relation (19.5 +/- 3.1 mm Hg/ml and 0.0 +/- 0.6 ml, respectively, p less than 0.01). The end-diastolic volume-peak dP/dt relation showed a similar curvilinearity when end-diastolic volumes were larger than 1.5-1.7 times the minimal end-diastolic volume reached during caval occlusion. ESPVRs were not different during aortic constriction and caval occlusion when end-diastolic volume was small. In contrast, with large end-diastolic volumes, Ees and Vd were significantly smaller during caval occlusion than during aortic constriction. The final part of ESPVR (with small end-diastolic volume) had the same slope and intercept as that during aortic constriction. We conclude that preload produces a curvilinearity of ESPVR that significantly modifies derived indexes when the range of preload changes is large.