The methodology for evaluating medicinal products is now well established. It is based partly on scientific studies provided in support of marketing application, and also on less rigorous "real-life" studies conducted in a specific healthcare system. The gap between these two methodological perspectives needs to be reduced. In the case of antipsychotic drugs, what is needed most is a better definition of endpoints for efficacy. Recent studies show that symptomatic improvements may enhance patients' insight and, in turn, increase their expectations, with a resulting stagnation of their perceived quality of life. The results of randomized trials are difficult to extrapolate to everyday clinical practice. Epidemiological studies with strict methodologies and conducted by independent bodies should be encouraged.