tre

ua Center for Studies on Human Stress, Douglas Hospital Research Center, McGill University, Montreal, Que., Canada

b

discuss the inverse negative correlations reported between hippocampal volume and memory for young and older adults and suggest thatthese inverse correlations could be partly due to the eects of contextual stress in young and older adults, as a function of age-related

therapies have been put forward to decrease stress andthus, promote health. However, there is a great paradox

sure. We feel stressed when we do not have the time to per-form the tasks we want to perform within a given period oftime. This time pressure usually triggers a set of physiolog-ical reactions that give us the indication that we arestressed. Although this denition is certainly accuratein terms of one component of the stress response, it is

* Corresponding author. Fax: +1 514 888 4064.E-mail address: sonia.lupien@mcgill.ca (S.J. Lupien).

Available online at www.sciencedirect.com

Brain and Cognition 65 (2dierences in hippocampal volume. 2007 Published by Elsevier Inc.

Keywords: Stress; Glucocorticoids; Catecholamines; Memory; Aging; Hippocampus

1. Introduction

Stress is a popular topic these days. A week seldompasses without hearing or reading about stress and its del-eterious eects on health. Given this negative impact ofstress on human health, many types of stress management

in the eld of stress research, and it relates to the fact thatthe popular denition of stress is very dierent from the sci-entic denition of stress. This has left a multitude of peo-ple and experts talking about, and working on, verydierent aspects of the stress response.

In popular terms, stress is mainly dened as time pres-Mood and Anxiety Disorders Program, Emotional Development and Aective Neuroscience Branch,

National Institute of Mental Health, NIH, Bethesda, MD 20892, USAc University of British Columbia, Centre for Community Child Health Research, 480 Oak Street, L408, Vancouver, BC, Canada V6H 3V4

Accepted 21 February 2007Available online 26 April 2007

Abstract

In this review, we report on studies that have assessed the eects of exogenous and endogenous increases in stress hormones on humancognitive performance. We rst describe the history of the studies on the eects of using exogenous stress hormones such as glucocor-ticoids as anti-inammatory medications on human cognition and mental health. Here, we summarize the cases that led to the diagnosisof glucocorticoid-induced steroid psychosis in human populations and which demonstrated that these stress hormones could thus crossthe bloodbrain barrier and access the brain where they could inuence cognition and mental health. We then summarize studies thatassessed the eects of the exogenous administration of glucocorticoids on cognitive performance supported by the hippocampus, thefrontal lobes and amygdala. In the second section of the paper, we summarize the eects of the endogenous release of glucocorticoidsinduced by exposure to a stressful situation on human cognition and we further dissociate the eects of emotion from those of stress onhuman learning and memory. Finally, in the last section of the paper, we discuss the potential impact that the environmental context towhich we expose participants when assessing their memory could have on their reactivity to stress and subsequent cognitive performance.In order to make our point, we discuss the eld of memory and aging and we suggest that some of the age-related memory impairmentsobserved in the literature could be partly due to increased stress reactivity in older adults to the environmental context of testing. We alsoThe eects of stress and scognition: Implications for th

S.J. Lupien a,*, F. Maheu b, M. T0278-2626/$ - see front matter 2007 Published by Elsevier Inc.doi:10.1016/j.bandc.2007.02.007ess hormones on humaneld of brain and cognition

c, A. Fiocco a, T.E. Schramek a

www.elsevier.com/locate/b&c

007) 209237

(Dickerson & Kemeny, 2002) (Table 1).

ble, for example, a public speaking task; for a completereview of these concepts, see Lupien et al., 2006). Thebodys response to absolute stressors is adaptive in nature.Being in or witnessing an accident, confronting a danger-ous animal, and being submitted to extreme cold or heatare all examples of absolute stressors that will necessarilylead to a stress response in the majority (if not the totality)of individuals when they are rst confronted with it. Theseextreme and particular situations constitute absolute stress-ors in that, due to their aversive nature, a stress responsehas to be elicited for ones survival and/or well-being. Inour western societies, absolute stressors are rare, but arenonetheless those that elicit the greatest physiologicalresponse.

Conversely, relative stressors are those events or situa-tions that will elicit a stress response only in a certain pro-portion of individuals. Moreover, this response may bemild or pronounced (Lupien et al., 2006). For example,having to unexpectedly deliver a videotaped speech may

Cimportant to acknowledge that in scientic terms, stress isnot equivalent to time pressure. If this were true, every indi-vidual would feel stressed when pressured by time. However,we all know people that seek time pressure in order to per-form adequately and others that are extremely stressed bytime pressure. This shows that stress is a highly individualexperience that does not depend on a particular event suchas time pressure, but rather, it depends on specic psycho-logical determinants that trigger a stress response.

2. What is stress?

Prior to becoming part of our day-to-day conversations,the term stress was used by engineers to explain forcesthat can put strain on a structure. For example, one couldplace strain on a piece of metal in such a way that it wouldbreak like glass when it reached its stress level. In 1936,Hans Selye (reproduced in Selye, 1998) borrowed the termof stress from the eld of engineering and talked aboutstress as being a non-specic phenomenon representingthe intersection of symptoms produced by a wide varietyof noxious agents. For many years, Selye tested variousconditions (e.g., fasting, extreme cold, operative injuries,and drug administration) that would produce morphologi-cal changes in the body that were representative of a stressresponse, such as enlargement of the adrenal glands, atro-phy of the thymus, and gastric ulceration. Selyes view ofthe concept of stress was that the determinants of the stressresponse are non-specic, that is, many unspecic condi-tions can put strain on the organism and lead to disease,the same way that many unspecic conditions can putstrain on a piece of metal and break it like glass.

Not all researchers agreed with Selyes model, particu-larly with the notion that the determinants of the stressresponse are non-specic. The reason for this was simple.While Selye spent his entire career working on physicalstressors (e.g., heat, cold, and pain), we all know that someof the worst stressors we encounter in life are psychologicalin nature, and are induced by our interpretation of events.For this reason, a physician named John Mason (1968)spent many years measuring stress hormone levels inhumans subjected to various conditions that he thoughtwould be stressful in order to describe the psychologicalcharacteristics that would make any condition stressful,to anyone exposed to it. By summarizing the results ofstudies measuring the circulating levels of stress hormonesbefore and after individuals were exposed to various jobsor situations that were deemed to be stressful (e.g., air-traf-c controllers or parachute jumping), Mason (1968) wasable to describe three main psychological determinants thatwould induce a stress response in any individual exposed tothem. Using this methodology, he showed that in order fora situation to induce a stress response by the body, it has tobe interpreted as being novel, and/or unpredictable, and/orthe individual must have the feeling that he/she does not

210 S.J. Lupien et al. / Brain andhave control over the situation. Although this work led toa general debate between Selye and Mason (Selye, 1975a,2.1. The relativity of stress

Stress can be absolute (a real threat induced by an earth-quake in a town, leading to a signicant stress response inevery person facing this threat) or it can be relative (animplied threat induced by the interpretation of a situationas being novel, and/or unpredictable and/or uncontrolla-1975b), further studies conrmed that the determinantsof the stress response are highly specic, and therefore,potentially predictable and measurable. More recently, ameta-analysis conrmed the importance of these character-istics, and added that the presence of a social evaluativethreat to a situation constitutes the fourth characteristicthat leads to physiological stress reactivity in humans

Table 1The four grades of steroid psychosis as described by Rome and Bracelandin 1952

Grade 1 Mild euphoriaLessened fatigueImproved concentrationElevated mood

Grade 2 Heightened euphoriaFlight of ideasImpaired judgmentInsomniaIncreased appetiteMemory impairment

Grade 3 AnxietyPhobiaRuminationHypomaniaDepression

Grade 4 Psychosis

ognition 65 (2007) 209237be very stressful for a given individual, and not at allfor another. Large inter-individual variations in the

stress-response to psychological challenges have been fre-quently reported (Hellhammer, Buchtal, Gutberlet, & Kirs-chbaum, 1997; Kirschbaum & Hellhammer, 1989;Kirschbaum, Klauer, Filipp, & Hellhammer, 1995a,1995b, Kirschbaum, Kudielka, Gaab, Schommer, & Hell-hammer, 1999; Kudielka, Buske-Kirschbaum, Hellham-mer, & Kirschbaum, 2004a; Kudielka, Schommer,Hellhammer, & Kirschbaum, 2004b; Lupien et al., 1997;Pruessner, Hellhammer, Pruessner, & Lupien, 2003; Rohle-der, Wolf, & Kirschbaum, 2003; Roy, Steptoe, & Kirsch-baum, 1998). It may be argued that absolute stressors areclosely linked to physiological systems due to their life-or self-threatening nature. On the contrary, relative stress-ors, because they are milder or because they necessitate acognitive interpretation in order to elicit a response, willnot necessarily lead to a physiological response, and thepresence or absence of a physiological response will dependon the outcome of the cognitive analysis.

S.J. Lupien et al. / Brain and CoThe stressor is the event itself, such as the earthquake inthe case of the absolute stressor, or the public speech in thecase of the relative stressor. The stress response is thebodys reaction to the event (Selye, 1975a, 1975b, 1998),and it is this bodys response to stress that is the foundationfor the studies that determined the impact of stress on cog-nitive function. The reason for this is that the stress hor-mones that are secreted in response to an absolute orrelative stressor are steroids that can easily cross thebloodbrain barrier and access the brain, where they caninuence learning and memory by binding to receptorslocalized in various brain regions known to be involvedin learning and memory.

2.2. Stress hormones

As presented in Fig. 1, when a situation is interpreted asbeing stressful, it triggers the activation of the hypotha-

Fig. 1. Schematic representation of the hypothalamic-pituitaryadrenal(HPA) axis. Following the perception of a stressor, the hypothalamusreleases CRF, which activates the pituitary and leads to secretion ofACTH. The levels of ACTH are detected by the adrenal cortex which then

secretes glucocorticoids and catecholamines (illustration: JasonBlaichman).lamicpituitaryadrenal (HPA) axis whereby neurons inthe hypothalamus, a brain structure often termed themaster gland, releases a hormone called corticotropin-releasing hormone (CRH). The release of CRH triggersthe subsequent secretion and release of another hormonecalled adrenocorticotropin (ACTH) from the pituitarygland, also located in the brain. When ACTH is secretedby the pituitary gland, it travels in the blood and reachesthe adrenal glands, which are located above the kidneys,and triggers secretion of the so-called stress hormones.

There are two main classes of stress hormones, the glu-cocorticoids (called corticosterone in animals, and cortisolin humans), and the catecholamines (adrenaline and nor-adrenaline). The acute secretion of glucocorticoids and cat-echolamines in response to a stressor constitutes theprimary mediators in the chain of hormonal events trig-gered in response to stress. When these two hormones aresecreted in response to stress, they act on the body to giverise to the ght-or-ight response whereby one would, forinstance, experience an increase in heart rate and bloodpressure (for a review, see Lupien et al., 2006).

Glucocorticoids have a variety of dierent eects in tar-get systems throughout the organism, which can be sum-marized as aiming to increase the availability of energysubstrates in dierent parts of the body, and allow for opti-mal adaptations to changing demands of the environment.While the activation of the HPA axis can be regarded as abasic adaptive mechanism in response to change, pro-longed activation of this system presents a health risk tothe organism. The highly catabolic glucocorticoids antago-nize insulin and increase blood pressure, thus increasing therisk for developing diabetes, hypertension, and arterial dis-ease. Also, growth and tissue repair are impaired. Further-more, activation of the HPA axis suppresses immunefunctions, which in a chronic state can be considered harm-ful for the organism, since it is associated with increasedrisk of infection (for a general review, see McEwen, 1998,2000).

Given their liposoluble characteristics, the glucocorti-coids can easily cross the bloodbrain barrier and accessthe brain where they bind to receptors. Three of the mostimportant brain areas containing glucocorticoid receptorsare the hippocampus, amygdala, and frontal lobes, whichare brain structures known to be involved in learning andmemory. Although adrenaline does not readily access thebrain, it can still impact the brain through its actions onthe sensory vagus outside of the bloodbrain barrier, withinformation transmitted into the brain via the nucleus ofthe solitary tract. The most important brain area contain-ing adrenergic receptors is the amygdala, which has beenshown to play an important role in fear processing, andmemory for emotionally relevant information.

Because of their actions on brain structures known to beinvolved in fear detection and memory for emotionally rel-evant information, the stress mediators enhance the forma-

gnition 65 (2007) 209237 211tion of so-called ashbulb memories of events associatedwith strong emotions, including fear but also positive emo-

Ctions. This process involves the amygdala, and the pathwayfor encoding these memories involves the interactionbetween neurotransmitters in the amygdala and in relatedbrain areas such as the hippocampus along with circulatingstress hormones (McGaugh, 2000; Roozendaal, Hahn,Nathan, de Quervain, & McGaugh, 2004; Roozendaal,McReynolds, & McGaugh, 2004; van Stegeren, Everaerd,Cahill, McGaugh, & Gooren, 1998). The importance ofthese stress mediators on memory for emotionally relevantinformation has been recently conrmed by studies inwhich blockade of either glucocorticoids (Maheu, Joober,Beaulieu, & Lupien, 2004) or noradrenaline (Cahill, Prins,Weber, & McGaugh, 1994) activity impaired the recall ofemotionally relevant information. Consequently, secretionof these primary stress mediators is necessary for the ade-quate encoding of emotionally relevant information. Thisenhancement of memory for stimuli inducing stressfuland/or emotional responses may be essential for speciessurvival. A recent study published by Zorawski and collab-orators goes along with this suggestion. They had subjectsparticipate in a fear conditioning task and assessed theassociation between fear-induced increase in cortisol, andthe consolidation of the memory. Results showed that par-ticipants whose fear learning was accompanied by high cor-tisol levels presented a better consolidation of this memory.Interestingly, this eect was more important in men(Zorawski, Blanding, Kuhn, & LaBar, 2006). Similar gen-der dierences in the pattern of cortisol activation inresponse to fear have recently been observed using func-tional brain imaging (Stark et al., 2006).

While short-term responses of the brain to novel andpotentially threatening situations may be adaptive andresult in new learning and acquired behavioral strategiesfor coping, as may be the case for certain types of fear-related memories, repeated stress can cause both cognitiveimpairments, and structural changes in the hippocampus,mainly through the actions of glucocorticoids. Conse-quently, in this paper, we will concentrate our eorts ondescribing the eects of glucocorticoids on human cogni-tive function and their potential impact on studies of brainand cognition.

2.3. Important characteristics of glucocorticoids

Under basal conditions, glucocorticoid secretion exhib-its a 24-h circadian prole in which glucocorticoid concen-trations present a morning maximum in humans (thecircadian peak), and slowly declining levels in the late after-noon, evening and nocturnal period (the circadian trough),and an abrupt elevation after the rst few hours of sleep(see Fig. 2).

Circulating glucocorticoids bind with high anity totwo receptor subtypes; the mineralocorticoid (hereaftercalled Type I) and glucocorticoid (hereafter called TypeII) receptors. Although both receptor types have been

212 S.J. Lupien et al. / Brain andimplicated in mediating glucocorticoid feedback eects,there are two major dierences between Type I and TypeII receptors. First, Type I receptors bind glucocorticoidswith an anity that is about 610 times higher than thatof Type II receptors. This dierential anity results in astriking dierence in occupation of the two receptor typesunder dierent conditions and time of day. Thus, duringthe circadian trough (the PM phase in humans and theAM phase in rats), the endogenous hormone occupiesmore than 90% of Type I receptors, but only 10% of TypeII receptors. However, during stress and/or the circadianpeak of glucocorticoid secretion (the AM phase in humansand the PM phase in rats), Type I receptors are saturated,and there is occupation of approximately 6774% of TypeII receptors (see Fig. 2).

The second major dierence between these two receptortypes is related to their distribution in the brain. The Type Ireceptor is exclusively present in the limbic system, with apreferential distribution in the hippocampus, parahippo-campal gyrus, entorhinal, and insular cortices. The TypeII receptor, however, is present in both subcortical (para-ventricular nucleus and other hypothalamic nuclei, the hip-pocampus and parahippocampal gyrus) and corticalstructures, with a preferential distribution in the prefrontalcortex. As we will see in the following sections, the impactof glucocorticoids on cognitive function can be best under-stood in terms of the dierential eects of Type I and TypeII receptor activation.

3. Eects of exogenous glucocorticoids on cognition

In an attempt to present the reader with a clear andcomplete view of the eects of glucocorticoids on humancognition, our background will be historical, as we willpresent the various models of glucocorticoid-eects onhuman cognition as a function of new approaches andmodels that have been described from the nineteenth tothe twenty-rst century. We use this approach for twomain reasons. First, history will teach us that our viewof glucocorticoid eects on human cognition remainedstable from 1970 to 1999, after which time new dataobtained in both animals and humans dramatically mod-ied our views. Second, because the history of the searchfor glucocorticoid eects on cognitive function took placein many dierent laboratories across the world, sometimessimultaneously, our historical report of the search for glu-cocorticoid eects on human cognition will describe oneof the best attributes of this eld of research, i.e. its richmultidisciplinary nature.

3.1. The era of clinical descriptions

The study of the eects of glucocorticoids on humanbehavior has always been an emotional one, lled withdebate, reconciliations, and the advancement of science.The story starts in 1855, when Thomas Addison describedfor the rst time a dark skin disease, a pigmentary char-

ognition 65 (2007) 209237acteristic that he found to be associated with pathologicalmodications of the adrenal glands (Addison, 1855). One

e (8

Typctiv

m7p

n abereivat

d Coyear later, Brown-Sequard (1856) showed the importanceof these small capsules for an animals life and conrmedAddisons observations. At about the same time, a debateexploded at the Societe de Biologie de Paris, when Dr.Bouillaud publicly accused Brown-Sequard of only doingsome amusing physiology, which is said to have pro-foundly insulted Dr. Brown-Sequard. In order to clearthe name of his colleague, Dr. Armand Trousseau cameto the rescue of Brown-Sequard at the Societe de Biologiede Paris and defended both his study and the clinical obser-vations of Dr. Addison. He further proposed to name thesyndrome described by Addison, Addisons disease, adisease that later became a textbook clinical description(see Olmsted, 1946). In 1889, after Brown-Sequard hadsucceeded Claude Bernard as professor of medicine at theColle`ge de France, he reported another controversial nd-ing to the Societe de Biologie de Paris (Brown-Sequard,1889). In experiments performed on himself, he had shownthe rejuvenating eects of testicular extracts from healthyyoung guinea pigs (Brown-Sequard, 1889). By 1890, anestimated 12,000 physicians were giving testicular extractsto their patients, despite the skepticism and embarrassmentexpressed by many medical journals (see Olmsted, 1946).

0

08a

m9a

m

10Cor

tisol 20

30

Tim

Type I/Type II receptors activation

a

10am

11am

12pm 1pm

2pm

3pm

4pm

5pm

6p

Fig. 2. Example of a circadian rhythm of serum cortisol levels. There is aacross the day. At the time of cortisol peak (early AM phase in humans), ththe time of the cortisol trough (PM phase in humans), there is mainly act

S.J. Lupien et al. / Brain anThis was the birth of organotherapy (see Borell, 1976a;Borell, 1976b).

Twelve years later, in 1901, Harvey Cushing, known asthe father of neurosurgery, misdiagnosed a patient withheadaches, visual troubles and sexual immaturity as hav-ing a cerebellar tumor and operated three times in thislocation (Cushing, 1913). The patient did not survive. Acouple of months later, Frohlich described the case of ayoung boy with headaches, visual troubles and sexualimmaturity in whom he successfully diagnosed a pituitarytumor. The patient did survive (see Fulton, 1946). Thedisease is now named Frohlich syndrome. It is said thatCushing never really accepted this defeat and alwaysremained very alert for pituitary tumors and their mani-festations (Fulton, 1946). This vigilance paid o since hedescribed in 1932, at the end of his career, the basophiltumors of the pituitary, leading to hypersecretion of glu-cocorticoids (Cushing, 1932). The disease is now namedCushings disease. However, 19 years earlier, in 1913,Cushing had already described cases of basophil tumorsof the pituitary who presented psychic disturbances(Cushing, 1913). In his biography, Dr. Cushing reportsthat his task had been facilitated by the fact that one ofhis rst patients was in an asylum, due to the psychic dis-turbances associated with this endocrinological disorder(Fulton, 1946).

3.2. Glucocorticoids and steroid psychosis

In 1949, a century after Addisons observation, occurredwhat some authors have named the most cataclysmicevent in the history of glucocorticoid endocrinology(Munck, Guyre, & Holbrook, 1984), that is, the discovery,by Hench, Kendall, Slocumb, and Polley (1949), of thetherapeutic eects of glucocorticoids on inammatory dis-eases such as rheumatoid arthritis and asthma. Regardedby many scientists and clinicians as the wonder drugs,glucocorticoids soon became very popular for the treat-ment of various inammatory diseases. Besides beingemployed in hormone replacement therapy in Addisonsdisease or after adrenalectomy, they have also been used

am-7am)

e Iation

m8p

m9p

m10

pm11

pm12

am 1am

2am

3am

4am

5am

6am

7am

rupt elevation a few hours before awakening, and slowly declining levelsis activation of both Type I and Type II glucocorticoid receptors, while ation of the high anity, Type I glucocorticoid receptors.

gnition 65 (2007) 209237 213in treating rheumatoid arthritis, ulcerative colitis, asthma,Hodgkins disease, systemic lupus erythematosus, and var-ious dermatological disorders (Munck et al., 1984). How-ever, no more than 2 years after their introduction asanti-inammatory drugs, the enthusiasm engendered byglucocorticoids was dampened by the nding that the ther-apeutic use of glucocorticoids was followed by several sideeects, particularly on aect and cognition. The rst casewas published in 1951 by Borman and Schmallenbergwho reported suicide following cortisone treatment (Bor-man & Schmallenberg, 1951). A year later, three paperswere published reporting severe mental disturbances inpatients under glucocorticoid therapy (Brody, 1952; Clark,Bauer, & Cobb, 1952; Rome & Braceland, 1952).

The mental side eects of glucocorticoid therapy consti-tuted a full spectrum of psychotic disorders. Most often,a vitalization eect was observed, particularly in theaged patient (Kountz, Ackermann, & Kheim, 1953) or in

Cindividuals with reduced vitality due to the underlyingillness (von Zerssen, 1976 cited in von Zerssen, 1976). Inthese cases, the patient displayed an elevation of mood,varying in degree from a feeling of well-being to abnormaldegrees of euphoria, psychomotor activation, increasedappetite and reduced sleep. However, these changes ofmood and behavior were generally observed in the rst daysor weeks of glucocorticoid therapy (Brody, 1952; Dordick &Gluck, 1955; Rees, 1953), and it was not clear whether thesesigns were related to glucocorticoid actions or to the relieffrom severe physical complaints and handicap related tothe underlying disease treated with glucocorticoids (Lidz,Carter, Lewis, & Surratt, 1952; Rees, 1953).

In a large proportion of patients, euphoria was not pres-ent in the rst days or weeks of treatment. Rather, tension,irritability, and sleeplessness were present. Both the eupho-ric and dysphoric states could gradually increase to a full-blown manic episode in which the patient would presentmarked euphoria or dysphoria, pronounced self-assertive-ness, hyperactivity, logorrhea, and ight of ideas (Cobb,Quarton, & Clark, 1954). These are the mental symptomsthat have led some scientists and clinicians to name the sideeects of glucocorticoid therapy a steroid psychosis(Clark et al., 1952; Rome & Braceland, 1952).

There were other types of aberrant behaviors that couldappear with glucocorticoid therapy and these later behav-iors closely resembled those observed in Cushings diseasepatients (Trethowan & Cobb, 1952). These mood andbehavioral changes covered a wide range of symptomsand included feelings of weakness, fatigue and drowsiness,lack of concentration, apathy, anxiety, depression, andsometimes suicide as reported by Borman and Schmallen-berg (1951). These behavioral changes were also associatedwith changes in brain activity as measured by EEG. In oneman, glucocorticoid therapy sometimes led to epileptic sei-zures and even to status epilepticus (Loewenberg, 1954;Stephen & Noad, 1951). Similarly, some EEG abnormali-ties were also found in patients with Cushings disease(Glaser, Kornfeld, & Knight, 1955; Plotz, Knowlton, &Ragan, 1952). On the whole, the clinical descriptions ofcases of steroid psychosis following glucocorticoid therapywere strikingly similar to the mental disturbances associ-ated with Cushings disease, leading to the idea that gluco-corticoids might be the underlying causes of steroidpsychosis.

Despite the adverse mental reactions that were reportedto occur in glucocorticoid-treated patients in 1951, it isinteresting to note that at about the same time, these com-pounds were also being tested for their possible psychotro-pic properties in the treatment of mental disorders (Rees &King, 1952). The rationale for testing the possible positiveeects of glucocorticoids on mental disturbances in psychi-atric patients was the fact that these compounds couldinduce euphoric or dysphoric states, and that by doingso, could potentially reverse mental states in depressed

214 S.J. Lupien et al. / Brain andand schizophrenic patients respectively. However, theresults of these trials were disappointing since neither inneurotic nor in endogenous psychotic disorders was thereany striking amelioration of the symptoms. In fact, thereverse was observed in both schizophrenia and depression(Rees & King, 1952).

Today, however, glucocorticoid-induced mood distur-bances are recognized and classied in the DSM-IV as sub-stance-induced mood disorders, with an associatedspecication of depressive, manic or mixed features,throughout history certain authors questioned whether glu-cocorticoids really could cause psychiatric adverse eects(Mitchell & Collins, 1984). A meta-analysis of randomizedcontrolled trials has provided rm conrmation that theycan indeed (Conn & Poynard, 1994). In general, prednisonehas been most frequently implicated in causing psychiatricside eects (for a review, see Hall, Popkin, Stickney, &Gardner, 1979), but other less widely used steroids suchas methylprednisone (Greeves, 1984; Perry, Tsuang, &Hwang, 1984), dexamethasone (Bick, 1983), and eveninhaled beclomethasone (Annett, Stansbury, Kelly, &Strunk, 2005; Kreus, Viljanen, Kujala, & Kreus, 1975)have also been reported to induce mental disturbances.Most patients exhibiting side eects are between 21 and60 years of age (Ling, Perry, & Tsuang, 1981), but adversemental changes are also being reported in children (Bender& Milgrom, 1995; de La Riva, 1958; Milgrom & Bender,1995; Sullivan & Dickerman, 1979), and older individuals(Varney, Alexander, & MacIndoe, 1984). Finally, femalesappear to be at somewhat greater risk than males, evenafter controlling for diseases that are more common towomen (Lewis & Smith, 1983). Importantly however, arecent study showed that in most patients, the cognitiveimpairments induced by high doses of exogenous cortico-steroids are reversible (Brunner et al., 2006).

3.3. Glucocorticoids, the brain, and vigilance

Armed with a large quantity of clinical and psychiatricdata revealing mental disturbances after long-term gluco-corticoid treatment in patients, researchers then took overthe eld of human glucocorticoid research, and started toperform psychopharmacological studies, measuring theeects of acute administrations of glucocorticoids in nor-mal individuals.

In 1970, Kopell and collaborators were the rst to studythe eects of exogenous administrations of glucocorticoidsin human subjects. Averaged event-related potentials(ERPs) were introduced at approximately the same time(1968), and based on the rationale that glucocorticoid-ther-apy could induce changes in EEG activity (Loewenberg,1954; Stephen & Noad, 1951), Kopell, Wittner, Lunde,Warrick, and Edwards (1970) measured whether exoge-nous administrations of glucocorticoids could have a sig-nicant impact on brain activity. Averaged ERPs reectelectrical activity produced by the brain in response to sen-sory stimulation and/or activity associated with the execu-

ognition 65 (2007) 209237tion of specic cognitive tasks. Using this paradigm, Kopelland collaborators found that after glucocorticoid

d Cotreatment, ERP amplitudes to visual stimuli were signi-cantly decreased. They concluded from these results thatglucocorticoids may have decreased the participants abilityto attend to the stimuli and thus reected a state of hypo-vigilance. This suggestion corroborated a clinical observa-tion made by Henkin and collaborators 3 years earlier(Henkin, McGlone, Daly, & Bartter, 1967) in Addisonspatients who present very low levels of circulating gluco-corticoids. These authors described how sensory acuity isstrangely elevated in Addisons patients and that treatmentof the hypocorticolism with steroids returns sensory acuityto normal. This led Henkin and collaborators to suggestthat glucocorticoids act by inhibiting the central nervoussystem, possibly leading to a state of hypovigilance.

In 1987, Born and collaborators (1987) performed astudy in which they infused glucocorticoids for 2 h, andmeasured ERPs to auditory stimulation. They conrmedthe hypovigilance hypothesis and showed a signicantreduction in ERP amplitude after glucocorticoid treatmentwithout changes in behavioral performance. In a secondstudy a year later, however, the same authors found thatglucocorticoid treatment had enhanced ERP amplitudesto auditory stimuli (Born, Hitzler, Pietrowsky, Pauschin-ger, & Fehm, 1988) and improved behavioral performance.Naturally, such a result would suggest that glucocorticoidsinstead increased vigilance, which contradicted the hypo-vigilance hypothesis described in their earlier paper (Born,Kern, Fehm-Wolfsdorf, & Fehm, 1987), and others (Kopellet al., 1970).

Although dierences in methodology could in partaccount for these contradictory results (e.g., slight dier-ences in stimulus intensity) one rm conclusion reachedby the authors was that glucocorticoids had a signicantimpact on vigilance (Born et al., 1988). In addition, the dif-ferential eects observed on behavioral performance andERP amplitudes led them to allude to the existence of amore complex relationship between glucocorticoids andcognitive processing that could be akin to the notion pro-posed by Yerkes and Dodson in 1908 (Yerkes & Dodson,1908) stating that an inverted-U shape curve describesthe relationship between vigilance and cognitive eciency.Optimal states of vigilance should be related to the optimalstate of cognitive eciency, while signicant decrease orincrease in vigilance should lead to impaired cognitiveeciency.

3.4. Glucocorticoids, the hippocampus, and declarativememory

At about the same time, and in very dierent laborato-ries, researchers were intrigued by the fact that glucocorti-coid treatment could lead to psychiatric symptoms. Indeed,for many years, endocrinologists and neuroscientiststhought that hormones, which are biological products,secreted by peripheral glands, did not access the brain

S.J. Lupien et al. / Brain anand acted mainly at the level of the peripheral nervous sys-tem. However, in the early 1960s, the discovery of neuro-peptides as substances having not only classical endocrineeects, but also aecting brain and behavior, signicantlyextended our view of hormones and opened the door tonew possibilities of hormonal actions on the brain (for acomplete historical background, see de Kloet, 2000). Theobservation that glucocorticoid treatment could lead to ste-roid psychosis suggested that the excessive concentrationsof the steroid may access the brain and exacerbate, perpet-uate or modify the presentation of mental symptoms asso-ciated with glucocorticoid administration. Such ahypothesis was appropriate since in 1943, Harris had com-pleted a series of landmark anatomical studies that clearlyestablished that the central nervous system regulates theHPA axis (Harris, 1972).

The search for brain receptors able to recognize periph-eral hormones began. In 1968, it culminated with BruceMcEwens seminal Nature paper showing that the rodentbrain was indeed able to recognize glucocorticoids (McE-wen, Weiss, & Schwartz, 1968). The story then took a veryimportant detour when McEwen and collaboratorsreported that the brain region showing the highest densityof receptors for glucocorticoids was the hippocampus, abrain region signicantly involved in learning and memory.

At this point in history, one observes a tremendousswitch in scientic studies of glucocorticoids from ERPstudies to memory-related studies. The rationale was basedon the fact that the largest amount of glucocorticoid recep-tors is found in the hippocampus, a structure now knownto be involved in specic types of memory functions inhumans. Scoville and Milner (1957) were the rst to reportthat the hippocampus is essential for declarative memory,while it is not essential for non-declarative memory (Squire,1992). The former underlies the conscious acquisition andrecollection of facts and event, while the latter holds infor-mation regarding processes and procedures for completinghighly practiced tasks such as riding a bike, (Scoville &Milner, 1957). Thus, this somewhat specialized role of thehippocampus served as the basis for specic hypothesesregarding the relation between increased cortisol secretionand impaired cognitive function in humans.

What is very interesting with the history of the searchfor glucocorticoid eects on human cognition is that fromnow on, scientists will be interested in showing thatglucocorticoids have specic eects on human declarativememory function that cannot be explained by glucocorti-coid-induced changes in vigilance or attention. In general,the majority of human studies that have measured theimpact of glucocorticoids on cognitive function reportimpaired declarative memory function after acute adminis-trations of synthetic glucocorticoids (for a complete review,see Lupien & McEwen, 1997).

The rst study performed on the acute eects of gluco-corticoids on human memory process was a doseresponsestudy that showed that the eects of hydrocortisone onhuman memory performance depend upon the dose admin-

gnition 65 (2007) 209237 215istered (Beckwith, Petros, Scaglione, & Nelson, 1986). Onlythe highest doses of hydrocortisone (40 mg) enhanced

Crecall when subjects were presented with lists of words.Hydrocortisone administration in the morning (at the timeof cortisol peak) impaired declarative memory function,while it had no eect on cognitive performance whenadministered at night (Fehm-Wolfsdorf, Reutter, Zenz,Born, & Lorenz-Fehm, 1993). Kirschbaum, Wolf, May,Wippich, and Hellhammer (1996) showed that 60 min afterthe oral administration of 10 mg of hydrocortisone, declar-ative memory performance was signicantly impaired whilenon-declarative memory performance remained intact,thus supporting the view that glucocorticoids aect hippo-campal-dependent cognitive functions.

More recently, de Quervain, Roozendaal, Nitsch,McGaugh, and Hock (2000) tested the impact of an acuteincrease in glucocorticoids as a function of the nature ofmemory processing. High doses of synthetic glucocorti-coids were administered either before the acquisition of alist of words, immediately after or just before the retrievalof the list. The results revealed signicant impairments inmemory when the drug was administered just before retrie-val, thus suggesting specic eects of glucocorticoids on theretrieval of previously learned information (see alsoDomes, Rothscher, Reichwald, & Hautzinger, 2005).

A specic eect of acute glucocorticoid elevations onretrieval processes in humans has recently been replicatedby Wolkowitz et al. (1990). Young and older men weregiven a medium dose of synthetic glucocorticoids after hav-ing learned a list of 10 words. A second word list waslearned and recalled after drug administration. Resultsshowed that glucocorticoids impaired recall of the word listlearned before treatment in both groups but did not inu-ence recall of the list learned after treatment. These resultsagree with previous data showing that acute exogenousadministrations of glucocorticoids have impairing eectson retrieval processes (de Quervain et al., 2000).

The in vivo demonstration of glucocorticoid eects onmemory retrieval processes was recently performed by thegroup of de Quervain et al. (2003) using positron emissiontomography (PET). Young subjects were given a mediumdose of synthetic glucocorticoids 24 h after learning variousdeclarative memory tasks. Brain activation was measuredby PET 1 h after drug administration. Results showed thatglucocorticoids induced a large decrease in regional cere-bral blood ow in the right posterior medial temporal lobecoupled with impaired cued recall of word pairs learned24 h earlier. These results were the rst to provide anin vivo demonstration that acutely elevated glucocorticoidlevels can impair declarative memory retrieval processesthat are related to measurable changes in medial temporallobe function. A similar impairment of retrieval functionwas recently reported by Buss, Wolf, Witt, and Hellham-mer (2004). These authors administered a small dose ofsynthetic glucocorticoids to young adults, and measuredretrieval of past events in their life (autobiographical mem-ory). Results showed that when compared to placebo, glu-

216 S.J. Lupien et al. / Brain andcocorticoids signicantly impaired retrieval of pastpersonal events.Besides acute actions, delayed eects of glucocorticoidswere reported in several memory studies in human subjects.Impaired memory performance was observed in normaladults following 5 days administration of high doses ofprednisone (80 mg p.o. daily), but normal memory perfor-mance in another group of subjects following a more acuteadministration of 1 mg of dexamethasone (Wolkowitzet al., 1990). Similarly, a 4-day administration procedurewith 0.5,1, 1, 1 mg/day of dexamethasone in normal con-trols produced impaired declarative memory performance(acquisition and recall) on the fourth day of treatment only(Newcomer, Craft, Hershey, Askins, & Bardgett, 1994).Similar results were obtained by the same group usinghydrocortisone (Newcomer et al., 1999). In both studies,no immediate or delayed eects of dexamethasone wereobserved on non-declarative memory or on selective atten-tion performance. Similar results were obtained using a 4-day regimen of 160 mg of prednisone. Taken together,these results were in accordance with a hippocampalinvolvement in glucocorticoid-related cognitive decitsand argued against a nonspecic eect of the steroid onattention and arousal (Schmidt, Fox, Goldberg, Smith, &Schulkin, 1999).

In summary, the majority of studies performed inhuman populations tend to conrm the rodent literaturereporting acute negative eects of glucocorticoids on hip-pocampal-dependent forms of memory (for a recentmeta-analysis, see Het, Ramlow, & Wolf, 2005). Alto-gether, the rodent and human data strengthened the viewthat stress hormones have a specic impact on the hippo-campus (for a complete critical review of the glucocorti-coid-hippocampus link, see Lupien & Lepage, 2001).

3.4.1. Interim: Lessons from history

Let us stop time and go back to Bruce McEwen et al.s,1968 discovery of glucocorticoid receptors in the rodenthippocampus (McEwen et al., 1968). It is clear from the lit-erature summarized above that the presence of glucocorti-coid receptors in the rodent hippocampus served as thebasis for the hypothesis that glucocorticoids should signif-icantly and specically impair declarative memory functionin humans. However, and although the hypothesis was sim-ple, economical, and easy to test, a closer look at historyteaches us that the glucocorticoid-hippocampus link mightnot be the best hypothesis to fully explain glucocorticoid-induced cognitive changes in humans (for a completereview on this topic, see Lupien & Lepage, 2001). Here iswhat history has to tell us.

In their 1968 paper, McEwen and collaboratorsdescribed the retention of corticosterone, a naturally occur-ring glucocorticoid, in the rodent brain (McEwen et al.,1968). The rats were rst adrenalectomized (the adrenalglands were removed and the animal was kept alive withphysiological doses of corticosterone) in order to depletethe rats system of any endogenous circulating glucocorti-

ognition 65 (2007) 209237coids, and then corticosterone was injected and the reten-tion of this naturally occurring glucocorticoid was

declarative memory function (see above), many studies per-formed in rodents reported that the ratio of Type I/Type IIoccupation is a major determinant of the direction of glu-cocorticoid-induced cognitive changes (for a review, seede Kloet, Oitzl, & Joels, 1999). For example, long-termpotentiation (LTP), a proposed neurobiological substrateof memory formation, has been shown to be optimal whenglucocorticoid levels are mildly elevated, i.e., when theratio of Type I/Type II occupation is high (see Diamond,Bennett, Fleshner, & Rose, 1992). In contrast, signicantdecreases in LTP are observed after adrenalectomy, whenType I occupancy is very low (Dubrovsky, Liquornik,Noble, & Gijsbers, 1987; Filipini, Gijsbers, Birmingham,& Dubrovsky, 1991), or after exogenous administrationof synthetic glucocorticoids (Bennett, Diamond, Fleshner,& Rose, 1991; Pavlides, Watanabe, & McEwen, 1993),which activate Type II and deplete glucocorticoids, againresulting in low occupancy of Type I.

d Cognition 65 (2007) 209237 217assessed. Using this method, they showed that corticoste-rone was highly retained by the hippocampus.

Researchers then sought to determine whether othersynthetic glucocorticoids would be retained by the hippo-campus. Notable among these was dexamethasone. deKloet, Wallach, and McEwen (1975) found the compoundto be very poorly retained by the rodent hippocampus, irre-spective of the route of administration (peripheral vs. intra-cerebroventricular). Later, it was shown that the smallamount of dexamethasone that did penetrate the brainwas retained in a regional pattern that was distinct fromthat of corticosterone (McEwen, 1976). In the same study,another steroid (cortisol) that is not a natural glucocorti-coid in the rat (although it is in humans) was very poorlyretained in the rodent brain, while corticosterone and aldo-sterone (another naturally occurring glucocorticoid) werehighly retained by the rodent hippocampus and surround-ing limbic structures.

The dierent modes of action of dexamethasone andcorticosterone on the rodent brain indicated that thesetwo compounds might bind to dierent types of glucocorti-coid receptors. This idea was conrmed six years later bythe presence of mineralocorticoid (Type I) and glucocorti-coid (Type II) receptors in the rodent hippocampus (Vel-dhuis, van Koppen, van Ittersum, & de Kloet, 1982). Thetrace amounts of corticosterone that were previouslyretained so abundantly by the rodent hippocampus wereactually bound to Type I and not to Type II receptors(Reul & De Kloet, 1985). In fact, these authors showedthat anity of hippocampal Type II for corticosterone inthe rat brain was actually too low for any signal to bedetected. At this point in time, history taught us one impor-tant fact about Type I and Type II receptors, i.e. that thereexists a tremendous dierence in the two receptor types interms of anity.

We now know that Type I receptors bind glucocorti-coids with an anity that is about 6- to 10 times higherthan that of Type II (Reul & De Kloet, 1985). This dier-ential anity results in a striking dierence in occupationof the two receptor types under dierent conditions andtime of day. Specically, during the circadian trough (thePM phase in humans), the endogenous hormone occupiesmore than 90% of Type I receptors, but only 10% of TypeII receptors. However, during stress and/or the circadianpeak of glucocorticoid secretion (the AM phase inhumans), Type I receptors are saturated, and there is occu-pation of approximately 6774% of Type II receptors (Reul& De Kloet, 1985). This dierential anity of Type I andType II opened the door to a brand new hypothesis aboutglucocorticoid eects on cognitive function, i.e. that gluco-corticoids could also have positive eects on cognitivefunction.

3.5. Positive eects of glucocorticoids on human cognition

S.J. Lupien et al. / Brain anIn contrast to human studies in which glucocorticoidswere consistently shown to have detrimental eects onIn a review of these issues, de Kloet et al. (1999) re-inter-preted the eects of glucocorticoids on cognitive perfor-mance in line with the Type I/Type II ratio hypothesis,suggesting that cognitive function can be enhanced whenmost of the Type I and only part of the Type II are acti-vated (top of the inverted-U shape function; increasedType I/Type II ratio; see Fig. 3). However, when circulat-ing levels of glucocorticoids are signicantly decreased orincreased (extremes of the inverted-U shape function; lowType I/Type II ratio), cognitive impairments will result.The authors suggested that the negative view of glucocorti-coid actions on human cognitive function could be partlyexplained by limitations in previous human experimentaldesigns, which did not allow for the dierential manipula-tion of Type I and Type II levels. In order to do this, suchstudies should measure cognitive function when glucocorti-coid receptors occupancy is decreased (rather thanincreased), thus allowing for functional measures of TypeI/Type II occupancy on learning and memory. One way

Mem

ory

Perfo

rman

ce

Type

I

Type

II

Type I saturated

50% Type II Occupied

Type II saturated

Circulating Levels of GCs

Fig. 3. The Type I/Type II glucocorticoid ratio hypothesis of theassociation between circulating levels of glucocorticoids, and memoryperformance (de Kloet et al., 1999). The gure shows occupancy of GCreceptors as a function of circulating levels of GCs and resultingmodulation of memory. When Type I glucocorticoid receptors aresaturated and there is partial occupancy of Type II glucocorticoidreceptors, there is maximization of memory, while when both Type I andType II glucocorticoid receptors are not occupied (left side of the inverted-

U shape function) or are saturated (right side of the inverted-U shapefunction), there is an impairment in memory performance.

to test such a hypothesis is the use of a hormone removalreplacement protocol.

In a hormone removalreplacement protocol, the behav-ior resulting from the absence of the hormone of interest isrst measured, and then baseline hormonal levels arerestored to normal values and the same behavior is mea-sured once again. If the hormone of interest has a realimpact in the behavior tested, then this behavior shouldbe restored to normal value after hormonal replacement(see Brown, 1998).

In order to test this suggestion, our group performed ahormone removalreplacement study in a population ofyoung normal controls (Lupien et al., 2002). In this proto-col, we used a within-subject double-blind experimentalprotocol in which we rst pharmacologically lowered glu-cocorticoids levels by administrating metyrapone, a potentinhibitor of glucocorticoid synthesis, we then restored base-line circulating glucocorticoid levels by infusing hydrocor-tisone, a synthetic glucocorticoid. Memory performance of

218 S.J. Lupien et al. / Brain and Cparticipants under each of these conditions was comparedto that measured on a placebo day. The results showed thatwhen compared to placebo, the pharmacological decreaseof circulating levels of glucocorticoids induced by metyra-pone signicantly impaired memory performance. Mostimportantly, we showed that this impairment was com-pletely reversed after hydrocortisone replacement (seeFig. 4 for a schematic representation of the results). Theseresults showed that glucocorticoids can modulate memoryfunction, and most importantly, they showed that theabsence of circulating glucocorticoid levels is as detrimen-tal for human memory function as is a signicant increasein glucocorticoids.

We have suggested that this modulation can happenthrough a dierential activation of Type I and Type IIreceptors. Indeed, during the metyrapone condition, TypeI occupancy was low, given the signicant decrease of glu-

MemoryPerformance

CIrculating Levels of Glucocorticoids

facilitation inhibition

Decrease (full line) and restoration (dashed line) in Cortisol levels

HormoneRemoval-ReplacementProtocol

8% decreaseIn memory

Restoration ofBaseline MemoryPerformance

Fig. 4. Schematic representation of the modulation of memory perfor-mance by pharmacological inhibition of cortisol secretion (full arrow), andby hydrocortisone replacement (dashed arrow) in young human partic-ipants. After corticol decrease, we observed a 8% decrease in declarativememory performance that was restored to placebo level after restoration

of circulating levels of cortisol by pharmacological infusion of hydrocor-tisone (see Lupien et al., 2002).cocorticoid secretion induced by metyrapone. At this point,impairment in memory was observed. In contrast, duringthe hydrocortisone replacement condition, glucocorticoidlevels were restored to the those typically found in theAM phase, i.e. leading to a saturation of Type I, with par-tial occupancy of Type II. This dierential occupation thusled to an increased Type I/Type II ratio, and a restorationof baseline cognitive performance.

In a second study (also in Lupien et al., 2002), we tookadvantage of the circadian variation in circulating levels ofglucocorticoids and tested the impact of a bolus injectionof glucocorticoids in the late afternoon, at a time of verylow glucocorticoid concentrations. In a previous study withyoung normal controls, we injected a similar dose of gluco-corticoids in the morning, at the time of the circadian peak,and reported detrimental eects of glucocorticoids onmemory (Lupien, Gillin, & Hauger, 1999). Here, whenwe injected a similar dose of hydrocortisone in the after-noon, at the time of the circadian trough, we observed thatalthough glucocorticoids did not change memory perfor-mance they nonetheless had a positive impact on cognitiveeciency and vigilance by signicantly decreasing reactiontimes on a recognition memory task when compared to theplacebo condition.

Data obtained by Oitzl and de Kloet in 1999 led them topropose that each glucocorticoid receptor type contributesto dierent aspects of cognitive processing. They foundthat Type I receptor activation is important for behavioralreactivity in response to environmental cues aecting vigi-lance and attention. Type II activation on the other handis essential for the consolidation of events/items inmemory.

In this view, the delayed memory impairment observedafter Metyrapone administration (i.e. due to the lack ofType II activation) in our rst experiment is consistent witha Type II receptor-mediated eect on consolidation pro-cesses. The signicant decreases in reaction times observedin our second experiment are in line with a Type I-mediatedeect on behavioral reactivity and vigilance. Takentogether these ndings oer further support for the notionthat the Type I/Type II occupancy ratio is an importantfactor in determining the direction and magnitude of gluco-corticoids eects on cognitive processing and that eachreceptor type plays a distinct yet complimentary role at dif-ferent stages of processing.

3.5.1. Interim: Lessons from historyNow, let us stop time again, and go back to Reul and De

Kloets, 1985 work on Type I and Type II receptors. Thestudy of Type I and Type II activity in the brain showedthat there exist large dierences in terms of anity for TypeI and Type II, a nding that led to the view that glucocor-ticoids are not only destructive (de Kloet et al., 1999).

However, the search for Type I and Type II receptors inthe brain led to another important discovery, i.e. the dier-

ognition 65 (2007) 209237ential distribution of Type I and Type II receptors in therodent brain. Following the work by Reul and De Kloet

d Co(1985), it was established that in the rodent brain, the TypeI is present exclusively in the limbic system, with a prefer-ential distribution in the hippocampus, parahippocampalgyrus, entorhinal, and insular cortices. In contrast, theType II receptor is present in both subcortical (paraventric-ular nucleus and other hypothalamic nuclei, the hippocam-pus and parahippocampal gyrus) and cortical structures,with a preferential distribution in the prefrontal cortex(Diorio, Viau, & Meaney, 1993; McEwen, De Kloet, &Rostene, 1986; McEwen et al., 1968; Meaney, Sapolsky,Aitken, & McEwen, 1985). Still, in the rodent brain, thelargest concentration of both Type I and Type II receptorswas found in the hippocampus, which again led to the glu-cocorticoid-hippocampus link described before.

However, in 2000, two papers were published whichdescribed the distribution of Type I and Type II receptorsin the primate brain, more closely related to the humanbrain in terms of neocortical development. These tworecent studies mapping both Type I and Type II receptordistribution in the primate brain strongly suggested thatextrapolation from rat brain to primate brain may be mis-leading when discussing the impact of glucocorticoids onthe hippocampus.

The rst study reported that, in contrast to its well estab-lished distribution in the rat brain, Type II mRNA is onlyweakly detected in the dentate gyrus and Cornu Ammonisof the macaque hippocampus (Sanchez, Young, Plotsky, &Insel, 2000). In contrast, Type II mRNA is strongly detectedin the pituitary, cerebellum, hypothalamic paraventricularnucleus and prefrontal cortices. In a second study, it wasreported that Type II receptors were well-expressed in thehippocampus, but were more prominently found in the pre-frontal cortex (Patel et al., 2000). Thus, Type I receptors arepresent in large quantities in the hippocampus and limbicstructures in the primate brain, while Type II receptors arepresent in all these structures and additionally in frontalregions. This latter nding suggested that in humans, gluco-corticoids should not only aect the hippocampus, but alsothe frontal lobes. This has been recently supported.

3.6. Glucocorticoids, the frontal lobe, and working memory

Studies in nonhuman primates (Goldman-Rakic, 1987,1995), and humans (Owen, Downes, Sahakian, Polkey, &Robbins, 1990; Petrides & Milner, 1982) show that lesionsof the dorsolateral prefrontal cortex (DLPFC) give rise toimpairments in working memory. Working memory is thecognitive mechanism that allows us to keep a small amountof information active for a limited period of time (seeBaddeley, 1995). In these working memory tasks, a tempo-ral gap is introduced between a stimulus and a response,which creates the need to maintain the stimulus in tempo-rary memory storage. Data obtained in monkeys showedthat cells in the lateral prefrontal cortex become particu-larly active during delayed response tasks, suggesting that

S.J. Lupien et al. / Brain anthese cells are actively involved in holding on to the infor-mation during the delay (Goldman-Rakic, 1990, 1995).Neuropsychological evidence suggests that humans withDLPFC damage are impaired in working memory (Fuster,1980; Luria, 1966). These patients are also highly suscepti-ble to cognitive interference and they perform poorly onneuropsychological tests that require response inhibitionsuch as the Wisconsin Card Sorting Test (Shimamura,1995; Stuss et al., 1982). Moreover, neuroimaging datashow a signicant relation between working memory pro-cessing and the activations observed in the prefrontal cor-tex (Smith, Jonides, Marshuetz, & Koeppe, 1998; butalso see Ungerleider, Courtney, & Haxby, 1998).

Working memory appears to be more sensitive thandeclarative memory to the eects of acute and short-termadministrations of glucocorticoids. Young, Sahakian,Robbins, and Cowen (1999) administered glucocorticoidsfor 10 days to young normal male volunteers and mea-sured various cognitive functions in a randomized, pla-cebo control, crossover, within-subject design. Theyshowed that this regimen of glucocorticoids led to decitsin cognitive function sensitive to frontal lobe dysfunction(working memory), while it did not impact on cognitivefunction sensitive to hippocampal damage (declarativememory).

Similar results were obtained by our group using anacute doseresponse neuroendocrine protocol (Lupienet al., 1999). In this study, 40 young subjects were infusedfor 100 min with either glucocorticoids or placebo, anddeclarative and working memory were tested during theinfusion period. Performance on the working memorytask decreased signicantly whereas performance on thedeclarative memory task remained the same followingan acute elevation of glucocorticoids. Curve t estima-tions revealed the existence of a signicant quadraticfunction (U-shape curve) between performance on theworking memory task and changes in glucocorticoids lev-els after hydrocortisone infusion. The results of these twostudies suggested that in young individuals, workingmemory is more sensitive than declarative memory toan acute elevation of glucocorticoids, supporting the sug-gestion that glucocorticoids have a signicant impact onfrontal lobe functions in humans.

Similar results were obtained by Hsu, Garside, Mas-sey, and McAllister-Williams (2003). In their study,twenty healthy subjects were treated with a high doseof synthetic glucocorticoids or placebo orally, in a dou-ble-blind, two-way crossover study. It was found that glu-cocorticoids impaired performance on an attentional task(Stroop), while they did not impair performance on adeclarative memory task. Other studies, however, reportan opposite pattern of ndings. Monk and Nelson(2002) for instance showed that working memory task(n-back) performance was unaected by a moderate doseof exogenous glucocorticoids but that declarative memoryperformance (intentional face recognition) suered as aresult of the increase in glucocorticoids. These data may

gnition 65 (2007) 209237 219in part be explained by the fact that face encoding isnot an entirely hippocampus-dependent process and that

Cit in fact recruits frontal regions (Sergerie, Lepage, &Armony, 2005).

3.7. Glucocorticoids and emotional memory

In humans, the frontal lobes have been shown to be sig-nicantly involved in emotional processing (for a review,see Damasio, 1995), and given the impact of glucocorti-coids on human frontal lobe function (Lupien et al.,1999; Young et al., 1999), it might be asked whether gluco-corticoids also impact on human emotional memory. In arecent study, Buchanan and Lovallo (2001) exposed youngparticipants to pictures varying in emotional arousal afterthey received a small dose of synthetic glucocorticoids(Buchanan & Lovallo, 2001). During acquisition, subjectswere not aware that their memory for the pictures wouldbe tested a week later (incidental memory). Resultsrevealed that glucocorticoid elevations during memoryencoding enhanced the delayed recall performance of emo-tionally arousing pictures while it had no impact on thedelayed recall of the neutral pictures.

Similarly, Abercrombie, Kalin, Thurow, Rosenkranz,and Davidson (2003) tested the eects of an exogenousadministration of two doses of synthetic glucocorticoidson emotional memory using a doseresponse study. Youngmen were presented with emotionally arousing and neutralstimuli after receiving either a placebo or a low or mediumdose of synthetic glucocorticoids. Free recall of the stimuliwas assessed 1 h after drug administration and recognitionmemory of the stimuli was assessed two evenings later.Results showed that glucocorticoid elevations decreasedthe number of errors committed on the free-recall tasks(increased performance). More importantly, the authorsshowed that when tested for recognition two evenings later,when cortisol levels were no longer manipulated, recogni-tion performance presented an inverted-U quadratic curve.Glucocorticoid-induced enhancements of recognitionmemory were only observed in the low-dose condition. Incontrast to the data obtained by Buchanan and Lovallo(2001), these results showed benecial eects of syntheticglucocorticoids on both emotionally arousing and neutralmaterial.

4. Stress, emotion, and cognition

Now that we have described the exogenous eects ofglucocorticoids on cognition, it is important to turn ourattention to the fact that endogenously released glucocorti-coids in the face of stress can also impact cognitive perfor-mance. Indeed, most of the previous literature covered usedexogenous administrations of synthetic glucocorticoids inorder to delineate the eects of glucocorticoids on humancognitive function. Yet, and as presented in Fig. 1, gluco-corticoids are natural substances that are secreted in theface of a challenge. Interestingly, glucocorticoids that are

220 S.J. Lupien et al. / Brain andsecreted endogenously in the face of a challenge still havethe ability to cross the bloodbrain barrier and impact cog-nitive performance through binding to Type I and Type IIglucocorticoid receptors in the brain. However, beforedescribing the eects of stress and related stress hormoneson cognitive performance, it is important to dissociate theeects of emotions from those of stress on humancognition.

4.1. Stress versus emotion

Emotionally arousing and stressful experiences are oftencited as the cause of many psychological and physical prob-lems. Many of us have experienced emotionally arousingand stressful experiences at one point or another in life,and noted that these experiences can have important eectson our memory. One can have forgotten an importantmeeting or anniversary due to work overload, or else,one can have a vivid recollection of a car accident or anyother emotionally arousing experience. Because of theirimpact on our lives, we have a tendency to pay more atten-tion to the negative eects of stress on our memory, and toforget that under certain conditions, emotionally arousingand stressful experiences can also have a positive impacton memory function.

Emotion and stress share many characteristics. A stress-ful experience will often cause a particular emotion (e.g.,surprise, fear, joy, etc.), and particular emotions can createstressful situations (e.g., blushing due to extreme timiditycan cause a stressful situation for an individual). Moreover,an emotion possesses many of the properties of a stressor.First, it often has an identiable source. Second, it is usu-ally brief and leads to an intense and conscious experienceof short duration. Finally, an emotion creates bodily reac-tions (e.g., increase in heart rate, perspiration, etc.) that aresimilar to those induced by a stressor, and both states actby increasing arousal. Because of these similarities betweenemotion and stress, most of the literature on emotion,stress and memory intermixes the eects of emotion andthose of stress upon memory function. However, emotionand stress are two dierent entities. Although a stressfulexperience will almost always trigger a specic emotion, aparticular emotion does not always elicit a stress reaction(for a complete review on the dierence between a stressand an emotion, see Lupien & Brie`re, 2000).

As far as laboratory settings are concerned, emotion andstress dier in the way they are induced and thus, in theway they inuence memory in humans. Hence, emotionsare usually induced by the presentation of emotionalwords, lms or pictures, while stress is usually induced byputting the individual in a social situation known to createstress (e.g., a public speaking task). Because of this impor-tant dierence between the experimental paradigms used tomeasure the eects of emotion and stress on memory, dif-ferent questions have been asked. Induction of emotionhas been used to measure memory for emotionally arousingevents, while induction of stress has been used to measure

ognition 65 (2007) 209237the specic eects of stress on subsequent memoryfunction.

d Co4.1.1. Mechanisms underlying memory for emotionally

arousing events

It is well known that what we encode and rememberfrom an event depends primarily on the attention that isdevoted to this event and its components. If you do notpay attention to what you are reading right now, there isless chance for you to remember it at a later time than ifyou give all your attention to your reading. This is becausethe more attention given to an event, the higher the prob-ability that this event will be elaborated (relating the infor-mation from this event to other situations and relatedconcepts in memory) at the time of encoding.

The level of attention devoted to an event at the time ofencoding will greatly depend on the emotional salience ofthis event. Most of us remember what we were doing andwith whom we were at the time we learned about the WorldTrade Center attacks, but the majority of us may have dif-culty remembering what we were doing and with whomwe were 13 days before or after the attacks. This ashbulbmemory phenomenon may be explained by the fact thatthe emotions (e.g., surprise, anger, fear etc.) that were trig-gered by the announcement of the World Trade Centerattacks directed the totality of our attention to the event,leading to a deeper elaboration and thus, to an optimiza-tion of our memory for this event.

Studies with trauma victims have reported how vividlythe traumatic event is recalled, in the absence of any mem-ory for information surrounding the traumatic event. Aclosely analogous situation appears in the eld of lawenforcement and describes the weapon focus phenome-non. Witnesses to violent crimes demonstrate a weaponfocus eect in which the weapon captures most of the vic-tims attention, resulting in a reduced ability to recall otherdetails of the scene and to recognize the assailant at a latertime (see Christianson, 1992). This phenomenon has beenexplained by Easterbrooks (1959; see Christianson 1992)cue utilization theory which suggests that emotionallyarousing events narrow subjects attention and lead themto attend only to the center of an event, and to excludemore peripheral information. In general, laboratory exper-iments have conrmed this hypothesis (Christianson,1992). Recently, Cahill, Gorski, Belcher, and Huynh(2004) pushed this analysis further and reported gender-related inuences in the recall of central vs. peripheralinformation from an emotional story (Cahill et al., 2004).Men and women with high male-related traits on theBem Sex-Role Inventory show better memory for centralaspects of an emotional story, relative to peripheral details.On the other hand, women and men with high female-related traits on the Bem Sex-Role Inventory have a bettermemory for peripheral details of an emotional story, asopposed to central information (Cahill et al., 2004; Cahill,Gorski, & Le, 2003).

Studies have examined the relation between emotion,retention intervals and memory. What these studies have

S.J. Lupien et al. / Brain anshown so far is that emotionally arousing events delay for-getting. In a highly cited study, Kleinsmith and Kaplan(1963) asked participants to learn neutral and emotionallyladen (e.g., rape, mutilation) word pairs. Both short-term (2 min) and long-term (1 week later) declarative mem-ory for the words pairs was assessed. The results showedthat emotional word pairs were better recalled after a longretention delay (Kleinsmith & Kaplan, 1963; Kleinsmith,Kaplan, & Tarte, 1963). These results have been replicatedmany times and the majority of studies have shownenhanced memory for emotionally arousing material, pro-vided that memory is tested at longer retention intervals.

There are extensive data showing that the observed ret-rograde enhancement of long-term memory for emotion-ally arousing events is related to the hormones releasedduring these experiences. Emotionally arousing events giverise to the secretion of the peripheral catecholamines adren-aline and noradrenaline by the adrenal medulla, centralnoradrenaline secretion by the locus coeruleus and to glu-cocorticoids secretion by the adrenal glands. Glucocorti-coids readily access the brain and can thus act directly toimpact emotional memory processing (Roozendaal, 2002;Roozendaal, Brunson, Holloway, McGaugh, & Baram,2002). The peripheral catecholamines on the other hand,enhance memory by reaching the vagus nerve, the nucleusof the solitary tract and the locus ceruleus. Central nor-adrenaline is then secreted by the locus ceruleus, activatingnoradrenergic neurons throughout the brain (Roozendaal,2002). Importantly, central noradrenaline is also triggeredas soon as an emotionally arousing event occurs, indepen-dently of peripheral catecholamines (McGaugh & Roo-zendaal, 2002; Roesler, Roozendaal, & McGaugh, 2002;Roozendaal, 2002). Specically though, hormonal eectson emotional memory processing are achieved via theinteractions between adrenergic hormones and glucocorti-coids and their respective receptors in the amygdala (Roo-zendaal, 2002). This then results in a modulation ofhippocampal activity and ultimately, enhanced long-termmemory for emotional material (Roozendaal, 2002; Roo-zendaal et al., 2002; Roozendaal, Quirarte, & McGaugh,2002).

Animal and human studies have conrmed the role ofthese hormones in the memory-modulating eects ofemotionally arousing events. In rodents, post-learningstimulation of the noradrenergic system enhances, whereaspost-learning blockade inhibits, long-term declarativememory of an inhibitory avoidance task (McGaugh,2000). Likewise, post-training injections of moderate dosesof synthetic glucocorticoids enhance, and pre-training glu-cocorticoid synthesis inhibition impairs, long-term expres-sion of inhibitory avoidance in animals (Roozendaal,2002; Sandi, 1998). In humans, pre-learning blockade ofcentral b-adrenergic receptors or pre-learning glucocorti-coid synthesis inhibition impairs long-term declarativememory for emotionally arousing material (Cahill et al.,1994; Maheu et al., 2004), whereas pre-learning or post-learning stimulation of the noradrenergic or glucocorticoid

gnition 65 (2007) 209237 221systems enhances it (Abercrombie et al., 2003; Buchanan &Lovallo, 2001; Cahill & Alkire, 2003). Although psycholog-

Cical and biological explanations of the arousal hypothesiswere used to explain the positive eects of emotion onmemory function, research performed to this day showsthat arousal is signicant as an intervening variable onlywhen the source of the arousal (in this case, the emotionallyarousing event) is directly related to the information to beremembered (Gore, Krebs, & Parent, 2006; Kuhlmann &Wolf, 2006). However, when the source of emotion is notdirectly related to the information to be remembered, otherpsychological and biological mechanisms come into playand have a stronger impact on memory function than arou-sal itself. Consistent with this suggestion, Rimmele,Domes, Mathiak, and Hautzinger (2003) reported that cor-tisol administration increased memory for details of neutralpictures, while it impaired memory for details of emotionalpictures (Rimmele et al., 2003), showing the intricate rela-tionship existing between stress, emotion, and memory.

4.2. Eects of stress on human memory

Remembering with great accuracy a particular emotion-ally arousing event is very dierent from performing vari-ous tasks involving memory in a day-to-day situationwhen one is faced with stress. We may remember with greataccuracy the events in our lives surrounding the WorldTrade Center attacks, but we may also have forgotten animportant appointment due to work stress.

Interestingly, rodent studies have shown that, when alaboratory stressor (e.g., tailshocks, water immersion orrestraint stress) is administered at various time-pointsbefore or after learning, as well as before recall, stress-induced elevations in glucocorticoid levels modulatedeclarative memory according to an inverted U-shapedfunction, a nding that is equivalent to the observed eectsof exogenous glucocorticoids on human cognition. Optimaldeclarative memory for material unrelated to the stressor(e.g., inhibitory avoidance protocols or spatial water-mazetasks) occur at moderate levels of stress and stress-inducedincreases in circulating glucocorticoid levels, whereas lower(i.e., boredom or drowsiness) or higher stress levels andstress-induced increases in circulating glucocorticoid levelsare less eective or may even impair declarative memoryperformance on these tasks (Roozendaal, 2002; Sauro, Jor-gensen, & Pedlow, 2003).

In humans, when a laboratory stressor (e.g., a publicspeaking task or a public mental arithmetic task) is admin-istered before learning or retrieval, high glucocorticoid lev-els following these stressors are associated with memoryimpairments for material unrelated to the stressor such asneutral words lists (see Jelici, Geraerts, Merckelbach, &Guerrieri, 2004; Lupien, Buss, Schramek, Maheu, & Pru-essner, 2005; Lupien, Fiocco, Wan, Maheu, Lord, Schram-ek, et al., 2005; Sauro et al., 2003; Takahashi et al., 2004;cf. Domes, Heinrichs, Reichwald, & Hautzinger, 2002;Wolf, Schommer, Hellhammer, Reischies, & Kirschbaum,

222 S.J. Lupien et al. / Brain and2002). Recently, studies measuring the inuence of stresson memory for emotional material unrelated to the stressorreported more heterogeneous ndings. Thus, when a labo-ratory stressor was presented before learning or retrieval ofemotional and neutral information unrelated to the stres-sor, high glucocorticoid levels following stress were associ-ated with memory impairments for emotional information(whether positive or negative), while they had no inuenceon memory for neutral material (Abercrombie, Speck, &Monticelli, 2006; Domes, Heinrichs, Rimmele, Reichwald,& Hautzinger, 2004; Kuhlmann, Piel, & Wolf, 2005;Maheu, Collicutt, Kornik, Moszkowski, & Lupien, 2005;but see Elzinga, Bakker, & Bremner, 2005). However,two other studies showed that stress administered before(Jelici et al., 2004) or after (Cahill et al., 2003) learningenhanced memory for emotional material, while it had noimpact (Cahill et al., 2003), impaired (Buchanan, Tranel,& Adolphs, 2006) or increased (Andreano & Cahill,2006) subsequent memory for neutral information (Jeliciet al., 2004).

Altogether, these results show that stress-related eleva-tions in glucocorticoids can have dierent eects on subse-quent memory for material unrelated to the stressor. Theeects of emotionally arousing and/or stressful events ondeclarative memory vary according to the nature of theto-be-remembered material, with elevated levels of gluco-corticoids enhancing memory for the emotionally arousingevent itself but leading, more often than not, to poor mem-ory for material unrelated to the source of stress/emotionalarousal.

The time of day (morning vs. afternoon) and levels ofcirculating glucocorticoids at the time of testing could alsobe important factors inuencing the eects of stress-relatedelevations in glucocorticoids on subsequent memory formaterial unrelated to the stressor. Glucocorticoid receptorsdier in terms of their anity for circulating levels of glu-cocorticoids. As we have discussed previously, Type Ireceptors have a 6- to 10-times higher anity for glucocor-ticoids than Type II receptors. A wealth of evidence nowdemonstrates that activation of Type I receptor is manda-tory for successful acquisition of environmental cues neces-sary to encode information, whereas activation of Type IIreceptors is necessary for long-term memory consolidationof this information (Oitzl & de Kloet, 1992). Endogenouslevels of glucocorticoids and thus, activation of Type Iand Type II receptors will vary across the day, with higherendogenous levels of glucocorticoids in the AM phase com-pared to the PM phase. Consequently, the addition of astressful event in the AM or PM phase, which by itself willtrigger a signicant increase in endogenous levels of gluco-corticoids, should have a dierential impact on activationof Type I and Type II receptor as a function of time ofday, and consequently, on memory performance. As indi-cated earlier, in the AM phase, most of the Type I receptorsand about half of the Type II receptors are activated, whilein the PM phase, most of the Type I receptors and about atenth of the Type II receptors are activated. If one applies a

ognition 65 (2007) 209237stressor in the AM phase, the endogenous increase in glu-cocorticoid levels that will be induced by the stressor will

d Coact by saturating Type II receptors, while the same stressorapplied in the PM phase will act by activating about half ofthe Type II receptors. Since stress-induced elevations inglucocorticoid levels have been shown to modulate declar-ative memory for material unrelated to the stressor accord-ing to an inverted U-shaped function, the dierentialactivation of Type I and Type II receptors at dierent timesof the day thus implies that a stressor applied in the morn-ing should impair memory function (right hand-side of theinverted U-shaped curve), while the same stressor appliedin the PM phase should increase or have no impact onmemory (left-hand-side or top of the inverted U-shapedcurve; see Lupien, Fiocco et al., 2005).

Finally, very few studies have measured the inuence ofthe central noradrenergic system, activated followingstress, on subsequent memory for material unrelated tothe stressor. Animal studies report that stress-inducedelevations in noradrenaline levels (following footshocks)modulate declarative memory according to an invertedU-shaped function, with optimal levels enhancing, andlower or higher levels of noradrenaline, impairing memoryfor information unrelated to the source of stress (e.g., pas-sive aversive conditioning; Gold & McGaugh, 1975; Gold,van Buskirk, & McGaugh, 1975). In humans, one recentstudy showed that blockade of peripheral and central nor-adrenergic b-receptors before the administration of a stres-sor did not impair memory for material unrelated to thesource of stress (Maheu et al., 2005). These results suggestthat the b-adrenergic system is not implicated in the eectsof stress on subsequent declarative memory function, con-trasting with the well-established role of this system duringthe memorization of events that are emotionally arousingin nature. Further studies in humans will be needed todetermine the exact role played by the central noradrener-gic system in the eects of stress on subsequent memory forinformation unrelated to the stressor.

5. Stress, memory, and the testing environment

In previous sections, we have summarized the studiesshowing that both exogenous and endogenous increasesin stress hormones can impact on cognitive function. Wehave also shown that the memory-enhancing eects ofemotions are mainly sustained by the catecholaminergicsystem, while the memory-impairing eects of stress onneutral memory are sustained by the glucocorticoid system.Now, can these studies showing impairing eects of stresson neutral memory have any implications for the eld ofbrain and cognition? We would like to end this large reviewof the literature on the eects of stress and stress hormoneson cognition by arguing that this eld of research is ofgreat importance for research on brain and cognition,and particularly for neuropsychology. The reason for thislies in the potential eects that the environmental contextin which we test our study participants may have on their

S.J. Lupien et al. / Brain anstress response, and what this stress response could inducein terms of neuropsychological performance. In order tomake our point, we chose to discuss potential stress eectsof the testing environment in young versus older adults andthe impact that testing-induced stress could have on theobtained results.

5.1. When we test, do we stress?

In the early 1990s, when most models of human memoryfunction were developed, they were based on the computer-model of information processing (Schacter, 1992; Squire,1992). These models were tested in both animals andhumans, most of the time without taking into consider-ation the physiological and/or neural mechanisms underly-ing memory function. However, in the last decade, newdata emerged showing that memory performance in ani-mals can be acutely modulated by manipulations of thetesting environment itself. More often than not, the para-digms used to study learning and memory in animals areaversive in nature, involving either shock, water immer-sion, restraint, or challenge to the homeostasis of theorganism through food restriction or deprivation. All theseparameters have been shown to activate, during the train-ing procedure, the physiological stress response (Cordero,Kruyt, Merino, & Sandi, 2002; Sandi, Loscertales, &Guaza, 1997; Sandi & Rose, 1997).

Given that circulating levels of glucocorticoids change inresponse to various stressors, it has been postulated thatchanges in circulating levels of glucocorticoids, inducedby the nature of the learning procedure used in animal pro-tocols, might be one of the most important factors inuenc-ing the strength of the memory trace obtained in thesestudies (Cordero et al., 2002; de Kloet et al., 1999; Sandi& Rose, 1997). This hypothesis has been conrmed bystudies showing that training conditions that result inlong-term memory formation are the same training condi-tions that induce a release of glucocorticoids.

For example, it has been shown that in the MorrisWater maze, in which a rat is immersed in water and hasto nd a platform using spatial cues in the room, the watertemperature is a potent modulator of the rate of acquisitionof the task. Rats that are trained at 19 C learn faster anddisplay better long-term retention than those trained at25 C (Sandi et al., 1997). Glucocorticoid levels are signif-icantly higher in rats in the 19 C group relative to those inthe 25 C group (cold water is a stressor). The role of glu-cocorticoids in the strength of memory consolidation hasfurther been conrmed using pharmacological protocols.So, when one combines a stimulus that leads to an endog-enous release of glucocorticoids (e.g. water temperature)with a pharmacological dose of glucocorticoids, this leadsto a modulation of the consolidation process. In the Morriswater maze, rats that are trained at 25 C (low endogenousrelease of glucocorticoids), and given synthetic glucocorti-coids show good long-term retention, while rats trainedat 19 C (high endogenous release of glucocorticoids),

gnition 65 (2007) 209237 223and given synthetic glucocorticoids (same dose) showimpaired long-term retention (Sandi & Rose, 1997).

stress and memory in aged humans were successful atshowing equivalent glucocorticoid levels before exposureto stress.

In 1997, we published a study assessing the eects of apsychosocial laboratory stressor on memory performancein acclimatized older adults (Lupien et al., 1997). Theresults revealed similar glucocorticoid levels before expo-sure to stress, and also showed that older participantswho presented an early increase in glucocorticoid levelsin anticipation of the stressor had poor memory perfor-mance before and after being confronted with the stressor,while older participants not showing any change in gluco-corticoid levels in anticipation of the stressor did not pres-ent changes in memory performance before or after thestressor. These results showed that increases of glucocorti-coids in response to environmental changes are potent pre-dictors of memory performance in older adults.

Now, a close look at the studies that reported impaired

CThese results again reveal the presence of a biphasicmodulation of memory formation by glucocorticoids,whereby an increase of glucocorticoids up to an optimumlevel should lead to enhanced consolidation, while gluco-corticoid levels above this optimum level should lead to adecrease in the consolidation process. These results con-rmed the presence of an inverted-U shaped function relat-ing the circulating levels of glucocorticoids induced by theexperimental context and the memory performanceobtained on the task.

Now, let us be frank. Although most rodent studies useaversive tasks to assess learning and memory, most humanstudies assess learning and memory with tools that are notthought to be stressful by nature, such as word lists andstory recall. Consequently, many would say that the proto-cols used in humans are non-aversive and non-stressful bynature.

However, this may not be the case, as the testing envi-ronment itself may lead to a dierent stress response inyoung and aged humans. In 1968, Mason published a sem-inal literature review in which he described t