Beta-adrenergic blocking drugs are a widely used, well tolerated and effective treatment for a variety of cardiovascular and noncardiovascular disorders. Over the years, beta-blockers have been associated with an incidence, albeit low, of CNS side effects. The question of interest, however, is whether the incidence is the same for all members of the class or whether other properties, such as hydrophilicity, have a bearing on the incidence of this type of side effect? This article addresses this question. In pharmacokinetic terms the lipophilic beta-blockers have been shown, both in animals and man, to readily cross the blood-brain barrier in contrast to hydrophilic beta-blockers. This is thought to have possible clinical relevance with respect to the relative incidence of CNS side-effects. To clarify the situation every published clinical paper, in which the beta-blockers propranolol (highly lipophilic, nonselective, no intrinsic sympathomimetic activity (ISA)), pindolol (moderately lipophilic, nonselective, moderate ISA), metoprolol (moderately lipophilic, beta 1-selective, no ISA) and atenolol (hydrophilic beta 1-selective, no ISA) were compared, was assessed for information pertaining to CNS side effects. This comprehensive review of the literature has shown, with few exceptions, that the incidence of CNS side effects such as sleep disturbances, dreaming, nightmares and hallucinations following clinically accepted doses of the four beta-blockers under scrutiny is generally low and that effects on short-term memory are minimal or absent. However, within this group of four drugs the incidence of these side effects is lowest with hydrophilic atenolol and generally highest with pindolol and propranolol. Metoprolol occupies an intermediate position. This order is in agreement with the pharmacokinetic observation that the more hydrophilic the molecule, the less is found in the brain tissue of both animals and man, although in the case of pindolol other factors may be important. The clinical relevance of studies involving psychometric testing is not clear.