‘Neuroprotection’ is a term used to describe the putative effect of interventions protecting the brain from pathological damage. In occlusive stroke, the concept of neuroprotection involves inhibition of a cascade of pathological molecular events occurring under ischaemia and leading to calcium influx, activation of free radical reactions and cell death. This article will summarize neuroprotection trials to date, some facts and fancies about neuroprotection, ischaemic pathophysiology and possible reasons for the apparent failure of human neuroprotective stroke trials. Facts: In the acute stage of occlusive stroke, moderate reduction of blood flow results in a ‘penumbra’ of brain cells, often surrounding a core infarct, in which brain cells survive for a few hours but gradually die if reperfusion is not established. Increased knowledge of the complex pathophysiology in acute ischaemic stroke has led to the development of a great number of candidates for neuroprotective interventions. Many neuroprotective agents have proven efficacious in animal models, but so far no human study has shown a statistically significant benefit in patients with acute ischaemic stroke on primary endpoint measures. Some neuroprotective agents show beneficial effects on post hoc analyses, and some studies are still ongoing. Fancies: In the early years of neuroprotective studies in stroke, it was thought that a drug with almost no adverse effects could be given by ambulance staff on the way to hospital and induce a clinically significant effect on outcome. Since there were only benefits and no risks, diagnostic skills by neurologists and neuroradiological evaluations would no longer be required. Why Have Neuroprotective Agents Failed in Human Stroke Trials? There are several possible explanations why neuroprotective trials have been unable to prove an effect in addition to the eventuality that the basic concept is wrong. The effects of neuroprotective agents on infarct size are time dependent, and treatment has often been initiated much later than in successful experimental stroke models. Insufficient doses of the drugs and slow availability of the drug at the target area may be other explanations. Too small sample sizes in trials and imbalance of prognostically important baseline variables are examples of shortcomings in trial methodology. What Can Be Done? Future New Approaches: In animal models, preclinical testing of neuroprotective candidates should be standardized. Conventional stroke models with young and healthy animals may be replaced by older animals with common co-morbidity such as atherosclerosis. Highly effective new neuroprotective agents need to be discovered, and combination therapies should be tried. In clinical trials, the greatest chances of success may be with neuroprotective concepts involving mechanisms in both ischaemic and reperfusion pathophysiology, in combination with a thrombolytic therapy protocol. Neuroprotective agents, possibly combinations of agents, should preferably approach several of these mechanisms. Treatments should be initiated early, at least within 3 h after stroke onset, by an intravenous route. The selected compound(s) should easily pass the blood-brain barrier. Neuroprotective agents shown to be highly effective in stroke models should be preferred, and doses used experimentally should be used also in the clinical setting. Trials should use randomization techniques, which reduce imbalances of prognostically important baseline variables, and the estimated sample size of a trial should be based on expectations of a modest clinical effect.