Five double-blind trials of the prophylactic treatment of migraine with beta-blockers were reviewed critically with regard to problem investigated, design, "blindness" of patient, side effects, statistical validity, and conclusions. The initial use of a "responder-finding" period may limit the validity of results from a subsequent double-blind crossover placebo-controlled study because it may impair "blindness". A d-propranolol versus racemic propranolol versus placebo study was reanalysed with conventional statistical methods, which showed that d-propranolol had no significant effect. In a multicentre study including 96 patients calculation of 95% confidence limits demonstrated that timolol and propranolol are equally effective in common migraine prophylaxis. Metopropolol and propranolol were apparently equally effective in 34 patients, but the calculation of confidence limits showed that a larger study is needed to confirm this conclusion. A group comparison study with three doses of nadolol and placebo lacked statistical evaluation of the results, thereby making it impossible to draw a conclusion about the efficacy of nadolol. It is concluded that there are considerable methodological problems in published papers on beta-blocker and migraine.