Thrombolysis improves stroke outcome, but is applicable to a limited number of patients. Neuroprotection has the prospect to be universally offered, either alone or in combination with thrombolysis. Potential drug targets include elements of the excitotoxic glutamate cascade, calcium entry, intracellular protease activation, free radical damage, the inflammatory response and membrane repair. Clinical trials with many agents have so far been disappointing, but hindsight reveals flaws in the choice of compound, the dose that was administered or trial design. A further crop of trials has recently been initiated or completed, with results expected from 2000 to 2003. More selective, potent or better tolerated neuroprotective strategies are still being developed for clinical use, and approaches to trial conduct are advancing: increased use of computer randomisation algorithms or diffusion-weighted magnetic resonance imaging should improve trial power. The prospects for a safe and effective treatment to improve stroke outcome remain good.