In intact chronically instrumented dogs, left ventricular dynamics were studied during cardiopulmonary resuscitation (CPR). Electromagnetic flow probes measured cardiac output and coronary blood flow, ultrasonic transducers measured cardiac dimensions, and micromanometers measured left ventricular, right ventricular, aortic, and intrathoracic pressures. The dogs were anesthetized with morphine, intubated, and fibrillated by rapid ventricular pacing. Data were obtained during manual external massage with dogs in the lateral and supine positions. Force of compression was varied from a peak intrathoracic pressure of 10 to 30 mm Hg, and compression rate was varied from 60 to 150/ min. Increasing force of compression increased stroke volume up to a peak intrathoracic pressure of approximately 20 mm Hg, beyond which stroke volume remained constant or declined. Stroke volume appeared to result primarily from direct transmission of manual compression force to the heart rather than from positive intrathoracic pressure because peak cardiac or vascular pressures or the change in these pressures were consistently two to four times greater than the corresponding intrathoracic pressures during manual compression. With increasing compression rate, stroke volume remained relatively constant, and total cardiac output increased significantly: 425 + 92 ml/min at 60/min, 643 130 ml/min at 100/min, and 975 ± 219 ml/min at 150/min (p < .05). Left ventricular dimensions decreased minimally at higher manual compression rates. In four patients undergoing CPR, systolic and diastolic arterial blood pressure increased with faster compression rates, correlating well with data obtained in the dog. Dynamic coronary blood flow in canine experiments decreased to zero or negative values during compression. Antegrade coronary flow occurred primarily during noncompression periods and seemed to be related to diastolic aortic perfusion pressure; coronary flow at a compression rate of 150/min averaged 75% of control. Therefore stroke volume and coronary blood flow in this canine preparation were maximized with manual chest compression performed with moderate force and brief duration. Increasing rate of compression increased total cardiac output while coronary blood flow was well maintained. Direct cardiac compression appeared to be the major determinant of stroke volume during manual external cardiac massage.