Attempts at prehospital fluid replacement should not delay the patient's transfer to hospital. Before bleeding has been stopped, a strategy of controlled fluid resuscitation should be adopted. Thus, the risk of organ ischaemia is balanced against the possibility of provoking more bleeding with fluids. Once haemorrhage is controlled, normovolaemia should be restored and fluid resuscitation targeted against conventional endpoints, the base deficit, and plasma lactate. Initially, the precise fluid used is probably not important, as long as an appropriate volume is given; anaemia is much better tolerated than hypovolaemia. Colloids vary substantially in their pharmacology and pharmacokinetics and the experimental findings from one cannot be extrapolated reliably to another. We still lack reliable data to prove that any of the colloids reduce mortality in trauma patients. In the presence of SIRS, hydroxyethyl starch may reduce capillary leak. Hypertonic saline solutions may have some benefit in patients with head injuries although this has yet to be proven beyond doubt. It is likely that one or more of the haemoglobin-based oxygen carriers currently under development will prove to be valuable in the treatment of the trauma patient.