A canine sepsis model that simulates the human cardiovascular response to septic shock was produced in 10 conscious unsedated dogs by implanting an Escherichia coli-infected clot into the peritoneum, resulting in bacteremia. By employing serial, simultaneous measurements of radionuclide scan-determined left ventricular (LV) ejection fraction (EF) and thermodilution cardiac index (CI), the end-diastolic volume index (EDVI) was calculated (EDVI = stroke volume index divided by EF). By using three different methods of quantifying serial ventricular performance (EF, shifts in the Starling ventricular function curve using EDVI vs. stroke work index, and the ventricular function curve response to volume infusion), this study provides evidence (P less than 0.01) that septic shock produces a profound, but reversible, decrease in systolic ventricular performance. This decreased performance was not seen in controls and was associated with ventricular dilatation (P less than 0.01); the latter response was dependent on an adequate volume infusion. Further studies of EDVI and pulmonary capillary wedge pressure during diastole revealed a significant, though reversible, shift (P less than 0.001) in the diastolic volume/pressure (or compliance) relationship during septic shock.