A group of 14 healthy elderly subjects was submitted to a non- stressful (attentional task) and a stressful (public speaking task) condition. Declarative (conscious recollection of learned information) and nondeclarative (retrieved information without conscious or ex- plicit access) memory as well as salivary cortisol levels were measured before and after each condition. The results showed that the stressful condition significantly decreased declarative memory performance, whereas the nonstressful condition did not. Nondeclarative memory performance was not affected by either condition. Further analyses separating the subjects into responders and nonresponders in terms of stress-induced cortisol change revealed a very different pattern of cortisol secretion and declarative memory performance in both pop- ulations. We showed that the responders presented increased cortisol levels 60 min before the actual stressor, whereas the nonresponders presented increased cortisol levels 25 min before the actual stressor. Although the responders did not differ from the nonresponders in declarative memory performance before and after the nonstressful condition, they presented a lower declarative memory performance when measured before and after the stressful condition. The early increase in cortisol levels observed in the responder group suggests that the anticipation of the stress, rather than the actual stressor per se, may have played a more significant role in the stress-induced declarative memory deficits observed in this subgroup. Together, these results show that the cortisol response to anticipation of stress and/or to stress in the elderly specifically affects those memory func- tions that are dependent on hippocampal activity. They also suggest that an altered cortisol responsivity to acute and/or chronic stress, with its detrimental effects on memory, could be an important factor explaining the genesis of memory deficits in aged populations. (J Clin Endocrinol Metab 82: 2070 -2075, 1997)