1. Acute haemorrhage was simulated in five unanaesthetized rabbits, by inflating a cuff on the inferior vena cava so that cardiac output fell by 8.3% of its resting level per minute. Simulated haemorrhage was performed after sham treatment, after graded doses of intravenous and intracisternal naloxone, and after cardiac nerve blockade with intrapericardial procaine. 2. After sham treatment, the haemodynamic response to simulated haemorrhage was biphasic. During the first phase, systemic vascular conductance fell steadily, heart rate rose steadily, and arterial pressure fell only slightly. A second decompensatory phase began abruptly when cardiac output had fallen to approximately 55% of its resting level. Vascular conductance rose steeply, heart rate fell slowly, and arterial pressure fell precipitately. 3. Treatment with naloxone (intravenous, 0.04‐0.4 mg kg‐1; intracisternal, 0.2‐2 micrograms kg‐1) did not affect either phase of the haemodynamic response to simulated haemorrhage. 4. After treatment with larger doses of naloxone (intravenous, 4‐8 mg kg‐1; intracisternal, 4‐69 micrograms kg‐1), the first phase was unaffected, but the second phase no longer occurred. Throughout simulated haemorrhage, systemic vascular conductance fell steadily, heart rate rose, and arterial pressure was well maintained. The dose of intracisternal naloxone which prevented the second phase was 90‐900 times less than the corresponding intravenous dose. The second phase was also prevented by cardiac nerve blockade. 5. We conclude that an endogenous opiate mechanism is responsible for the haemodynamic decompensation that occurs when cardiac output falls to a critical level. The mechanism is located within the central nervous system. It is triggered by a signal from the heart.