The threshold concentrations responsible for circulatory collapse were experimentally investigated by intravenously infusing the drug at high rates (0.2 and 0.3 mg kg-1 min-1 over 20 min, and 0.4 mg kg-1 over 10 min) to 11 anaesthetized and ventilated dogs. Bupivacaine plasma concentrations at the time of haemodynamic measurements were, respectively, 4303 +/- 46, 5829 +/- 615, and 8930 +/- 689 ng ml-1 (means +/- SEM). Cardiac output appeared to be the first and the most affected of the haemodynamic variables studied. Its reduction was already significant with the 0.2 mg kg-1 min-1 bupivacaine infusion, whereas mean systolic blood pressure remained unchanged because of the compensatory increase in vascular resistance. The fall of cardiac output was enhanced by the rise in bupivacaine infusion rate, with simultaneous substantial decreases in left ventricular pressure and LV dP/dt max. At this stage, the increase in systemic vascular resistance was less marked than at low infusion rates, and was not sufficient to prevent hypotension. The variations of mean pulmonary blood pressure and pulmonary capillary wedge pressure did not reach statistical significance, and the absence of significant change in mean pulmonary blood pressure at the time cardiac output was reduced reflected the increase in pulmonary vascular resistance. These results suggest that high plasma concentrations of bupivacaine exert a depressant effect on cardiac contractions earlier than on arteriolar tone.