Intravenous fluid is life-saving in hypovolemic shock, but fluid sometimes aggravates the bleeding. During the past 25 years, animal models have helped our understanding of the mechanisms involved in this unexpected effect. A key issue is that vasoconstriction is insufficient to arrest the bleeding when damage is made to a major blood vessel. ‘Uncontrolled hemorrhage’ is rather stopped by a blood clot formed at the outside surface of the vessel, and the immature clot is sensitive to mechanical and chemical interactions. The mortality increases if rebleeding occurs. In the aortic tear model in swine, hemorrhage volume and the mortality increase from effective restoration of the arterial pressure. The mortality vs. amount of fluid curve is U-shaped with higher mortality at either end. Without any fluid at all, irreversible shock causes death provided the hemorrhage is sufficiently large. Crystalloid fluid administered in a 3 : 1 proportion to the amount of lost blood initiates serious rebleeding. Hypertonic saline 7.5% in 6% dextran 70 (HSD) also provokes rebleeding resulting in higher mortality in the recommended dosage of 4 ml/kg. Uncontrolled hemorrhage models in rats, except for the ‘cut-tail’ model, confirm the results from swine. To avoid rebleeding, fluid programs should not aim to fully restore the arterial pressure, blood flow rates, or blood volume. For a hemorrhage of 1000 ml, computer simulations show that deliberate hypovolemia (−300 ml) would be achieved by infusing 600–750 ml crystalloid fluid over 20–30 min or 100 ml of HSD over 10–20 min in an adult male