Dehydroepiandrosterone (DHEA) is a steroid that shows a marked age-related decline in humans. Previous research suggests potential for DHEA replacement in old age to enhance cognition and well-being. We conducted a clinical trial to test these hypotheses in a nonclinical sample of 46 men aged 62-76. Participants received either 50 mg DHEA daily for 13 weeks, followed by placebo for 13 weeks, or the reverse, in a randomised double-blind crossover trial design. Levels of salivary cortisol and DHEA were measured at 0800 h and 2000 h prior to each assessment session. Cognition was assessed with tests of speed, attention and episodic memory. Well-being was measured with questionnaires of mood and perceived health. Mood questionnaires were completed at the assessment session as well as concurrently with saliva sampling. A correlational analysis of baseline behavioural data with hormonal data, controlling for age, revealed that higher morning DHEA was assocated with lower confusion (r=-0.33; P=0.04), while higher evening DHEA was associated with lower anxiety (r=-0.35; P=0.03) and lower current negative mood in the morning (r=-0.37; P=0.03). Conversely, higher morning cortisol and a morning cortisol/DHEA ratio were assocated with higher anxiety (r=0.35; P=0.03), (r=0.46; P=0.004), general mood disturbance (r=0.32; P=0.046), (r=0.32; P=0.04) and higher current negative mood in the evening (r=0.37; P=0.03), (r=0.38; P=0.03). A higher morning cortisol/DHEA ratio was also associated with higher confusion (r=20.39; P=0.01) and lower visuo-spatial memory performance (r=-0.39: P=0.01). Unexpectedly, higher evening cortisol was associated with faster choice reaction time (r=-0.33: P=0.04). These findings are consistent with an impairing effect of high cortisol on episodic memory and mood in older men, which may be attenuated by DHEA. When treatment effects were analysed, no significant effects of DHEA were observed on any of the trial outcomes, providing no support for benefits of DHEA supplementation for cognition or well-being in normal older men in the shorter-term. (C) 2001 Elsevier Science Ltd. All rights reserved.