Controlled clinical trials are becoming increasingly frequent in neurology. A review of the literature indicates that several trials have serious flaws in study design and conduct that render the results questionable or uninterpretable. These reports, together with my own experience with the trial of plasmapheresis in the treatment of acute Guillain-Barré syndrome, have led me to conclude that while the formulation of the crucial research question and the definition of outcomes as a measure of efficacy are extremely important, the availability of a compliant patient population of appropriate size and the compliance by the study physicians are essential. Moreover, close cooperation with the statistician in planning the trial and the statistical strategy for analysis is also critical. Suggestions are made to aid the clinician in setting up the most efficacious trial and reporting the results.