To investigate the noradrenergic modulation of working memory in humans single doses of two beta-blockers [either 25 mg of propranolol (lipophilic) or 50 mg of atenolol (hydrophilic)] or placebo were administered to young healthy volunteers (16 subjects per drug condition) performing a numerical working memory task that requires either short-term maintenance or maintenance plus manipulation of visually presented four-number sequences. Higher manipulation costs (i.e. process-specific slowing of reaction times in the manipulation conditions compared to the control condition) were observed after propranolol but not after atenolol. The propranolol effect was mainly observed in subjects with low emotional arousal (i.e. low state anxiety rating at baseline). Because both beta-blockers induced a comparable decrease of blood pressure and pulse, the propranolol effect on the 'working component' of working memory is considered to be a central, presumably prefrontal one.