1. The secretion of vasopressin in response to haemorrhagic shock has been investigated in anaesthetized dogs.2. The changes in the plasma concentrations of vasopressin were followed over a period of 5 hr, during which the arterial blood pressure was kept constant at 40 mm Hg. It was found that vasopressin concentration in plasma rose to a high peak shortly after the onset of shock and gradually declined thereafter. Five hours later, it was still 3.5 times higher than control. Re-transfusion of blood was followed by a return to control levels.3. The clearance of vasopressin was calculated before and during shock in normal dogs and in dogs with experimental diabetes insipidus. Soon after the onset of shock, the clearance rate dropped to one quarter of its normal level but slowly recovered, returning to near control values at the fifth hour of shock. Clearance rates did not vary as a function of infusion rates, suggesting that there is no maximal transport rate for the removal of the hormone over the entire secretory range found in normal and hypotensive dogs.4. From the clearance rates and from the plasma concentrations of endogenously secreted vasopressin it has been possible to calculate the approximate secretory rates of the hormone in response to shock. Secretion rose to a very high level, some 40 times greater than control, at the onset of shock. This was followed by a fairly constant secretory plateau. At the fifth hour of shock secretion was 3.5 times higher than control.5. The half-life of vasopressin was measured in normal and hypotensive dogs. Control measurements confirm the generally accepted value of approximately 5 min. The half-life was significantly higher in the early stage of shock, but returned to control values in the later stage.6. Haemorrhage experiments performed in normal and diabetic dogs suggest that vasopressin may play a part in the development of irreversible haemorrhagic shock: all normal animals died within a few hours of retransfusion, whereas four out of eight diabetic dogs similarly treated survived a 24 hr observation period. In a separate set of experiments, eight diabetic dogs were subjected to the haemorrhage procedure while receiving a constant infusion of vasopressin: only two of these survived. Surviving dogs showed none of the characteristic lesions of irreversible haemorrhagic shock.