ARCHIVAL REPORT

Time-Dependent Effects of Cortisol on theContextualization of Emotional Memories

Vanessa A. van Ast, Sandra Cornelisse, Martijn Meeter, Marian Joëls, and Merel Kindt

Background: The inability to store fearful memories into their original encoding context is considered to be an important vulnerabilityfactor for the development of anxiety disorders like posttraumatic stress disorder. Altered memory contextualization most likely involveseffects of the stress hormone cortisol, acting via receptors located in the memory neurocircuitry. Cortisol via these receptors inducesrapid nongenomic effects followed by slower genomic effects, which are thought to modulate cognitive function in opposite,complementary ways. Here, we targeted these time-dependent effects of cortisol during memory encoding and tested subsequentcontextualization of emotional and neutral memories.

Methods: In a double-blind, placebo-controlled design, 64 men were randomly assigned to one of three groups: 1) received 10 mghydrocortisone 30 minutes (rapid cortisol effects) before a memory encoding task; 2) received 10 mg hydrocortisone 210 minutes (slowcortisol) before a memory encoding task; or 3) received placebo at both times. During encoding, participants were presented withneutral and emotional words in unique background pictures. Approximately 24 hours later, context dependency of their memories wasassessed.

Results: Recognition data revealed that cortisol’s rapid effects impair emotional memory contextualization, while cortisol’s slow effectsenhance it. Neutral memory contextualization remained unaltered by cortisol, irrespective of the timing of the drug.

Conclusions: This study shows distinct time-dependent effects of cortisol on the contextualization of specifically emotional memories.The results suggest that rapid effects of cortisol may lead to impaired emotional memory contextualization, while slow effects of cortisolmay confer protection against emotional memory generalization.

Key Words: Consolidation, context, cortisol, emotion, hydrocortisone,memory, PTSD

Memories are more likely to be remembered when theretrieval context resembles the encoding context (1–3).Such contextual dependency of memories is highly

adaptive, as it can help to retrieve memories that are likely tobe appropriate in a specific context. Consequently, the ability tostore memories into their original encoding context (i.e., memorycontextualization) may protect against memory generalization.Since patients suffering from posttraumatic stress disorder(PTSD) display augmented memory generalization (4), contextu-alization of emotional memories seems to be compromised inPTSD (5–8). The hippocampus, which is supposed to subservecontext effects on memory (9–12), is a main target of the stresshormone cortisol (13). Through its effects on the hippocampus,cortisol may impair the contextual dependency of memories, butwhether this is indeed the case in healthy humans is presentlyunknown. In general, the literature on the potential role ofcortisol in (traumatic) memory formation is equivocal. Forinstance, one recent animal study showed that corticosteroids,

From the Department of Clinical Psychology (VAvA, MK), University ofAmsterdam, Amsterdam; Department of Neuroscience and Pharma-cology (SC, MJ), Rudolf Magnus Institute of Neuroscience, UniversityMedical Center Utrecht, Utrecht; Amsterdam Brain and Cognition (SC,MJ, MK), University of Amsterdam, Amsterdam; and Department ofCognitive Psychology (MM), VU University Amsterdam, Amsterdam,The Netherlands.

Address correspondence to Vanessa A. van Ast, M.Sc., University ofAmsterdam, Department of Clinical Psychology, Weesperplein 4, Room5.19, Amsterdam 1018 XA, Netherlands; E-mail: V.A.vanAst@uva.nl.

Received Nov 27, 2012; revised Jun 27, 2013; accepted Jun 29, 2013.

0006-3223/$36.00http://dx.doi.org/10.1016/j.biopsych.2013.06.022

injected after fear conditioning, reduce context-specific fearresponses the next day, causing PTSD-like symptoms (14).Conversely, another study showed that corticosteroid adminis-tration before acute stress could dampen subsequent behavioralcharacteristics of PTSD (15).

One possible explanation for such paradoxical effects ofcorticosteroids is that these hormones exert time-dependenteffects on neurobiological processes, with disparate net effectson behavior (16–20). After a stressful event, nongenomic cortico-steroid actions quickly enhance neural activity mediated byglutamate in mice, particularly in the amygdala (21,22). Inhumans, acutely elevated cortisol levels generally suppress hippo-campal activity (23–25). In interaction with noradrenergic activa-tion, corticosteroids (via rapid actions) enhance human amygdalaactivity (26–29) and facilitate instinctive, habitual behavior (30)and negative response biases (31,32), while impairing higherorder controlled executive processes (33). Together, these behav-iors may promote survival at the short term, helping the organismto select the most appropriate strategy immediately after stress,though at the cost of remembering contextual details. Somehours after stress, slower long-lasting genomic corticosteroidactions develop (16,34). The available data suggest that theseslow actions serve to restore homeostasis following stressfulperiods (20,35,36). In agreement, slow corticosteroid effects havebeen shown to enhance cognitive self-control (37), enhanceworking memory processing involving the dorsolateral prefrontalcortex (38), promote sustained attentional processing (39), andreduce amygdala activity (39,40). As such, slower genomiccorticosteroid effects may aid in remembering a certain eventin a more cognitively controlled, contextualized, manner.

Given these findings, we probed these two time domainsand tested the hypothesis that rapid cortisol effects [presum-ably through nongenomic pathways and in interaction witharousal-evoked central adrenergic release caused by the emo-tional words (41,42)] impair the contextual dependency

BIOL PSYCHIATRY 2013;74:809–816& 2013 Society of Biological Psychiatry

810 BIOL PSYCHIATRY 2013;74:809–816 V.A. van Ast et al.

specifically of emotional memories and that delayed (presum-ably gene-mediated) effects of cortisol enhance contextualdependency of subsequent emotional memories. To test this,we randomly assigned healthy young men to one of threegroups: 1) a group receiving placebo at 210 minutes and 10 mghydrocortisone at 30 minutes before encoding (rapid cort); 2) agroup receiving cortisol 210 minutes and placebo 30 minutesbefore encoding (slow cort); and 3) a group receiving placeboat both times. During encoding, participants were presentedwith neutral and emotional (i.e., of negative valence and higharousal) words in unique background pictures. Approximately24 hours later, memory contextualization was assessed; half ofthe emotional and neutral words were tested in intact contexts,while the other half of the words were tested in rearrangedcontext combinations (43,44). In addition to objective alter-ations in memory performance, we conducted exploratoryanalyses to test for changes in the subjective quality ofmemories (45).

Methods and Materials

ParticipantsIn total, 64 male subjects gave written informed consent.

Participant characteristics are given in the Supplementary Resultsin Supplement 1. The local ethical committee of the University ofAmsterdam approved the study. Inclusion criteria as assessed byself-report were no past or present psychiatric or neurologicalcondition, no diagnosis of dyslexia, and age between 18 and 35years. Men having any somatic or endocrine disease (e.g., acuteasthma) or taking any medication known to influence centralnervous system or endocrine systems were excluded fromparticipation. A final exclusion criterion constituted nonadherenceto the encoding instructions on day 1. Further, participants wereasked to refrain from taking any drugs 3 days before participationand to get a night of proper sleep, refrain from heavy exerciseand alcohol and caffeine intake 12 hours before participation, andto not eat, drink, smoke, or brush teeth 2 hours beforeparticipation. Subjects were rewarded for their participation withcourse credits or paid €65.

www.sobp.org/journal

Drug Administration and AssessmentHydrocortisone and placebo (albochin) treatments were

administered through identically appearing pills. A single doseof 10 mg of hydrocortisone was employed to elevate endogenouscortisol to a level equivalent to moderate acute stress (46).Salivary free cortisol concentrations were assessed with Salivettecollection devices (Sarstedt, Nümbrecht, Germany). Cortisol sam-ples were taken at 10 time points spread throughout the experi-ment (Figure 1) and subsequently stored at �251C. Uponcompletion of the study, samples were sent out to Dresden(Technische Universität, Germany) where salivary free cortisolconcentrations were measured using a commercially availablechemiluminescence immunoassay with high sensitivity of .16 ng/mL(IBL, Hamburg, Germany).

Memory MeasurementsEncoding. The encoding task was modeled freely after

Talamini et al. (43) and Tambini et al. (44). To induce emotionalversus neutral declarative memories, participants were shown 30neutral, low-arousing words and 30 negative, high-arousingwords on a small gray rectangle presented against color picturesof natural scenes or city landscapes (Figure 2). We utilized wordsbecause they are easier than pictures to match on a range ofdimensions (e.g., frequency or familiarity) that affect memoryperformance (47). Furthermore, memory-enhancing effects forarousing words are likely mediated by rapid arousal-evokedcentral adrenergic release from the locus coeruleus (48) mediatedby the amygdala (41,42). Subjects were instructed to vividlyimagine the meaning and content of each word in the back-ground 1) to promote deep encoding (49), 2) to create anassociation of the word with its unique context, and 3) to createa complete and rich episodic memory. Each individual word waspresented for 5 seconds together with a unique context inrandom order. Next, participants evaluated their mental imageon arousal and valence dimensions using self-assessment man-ikins (50) in a fixed time window of 4 seconds for each rating. Inbetween trials, a fixation cross was presented for 1 second. Wordswere concrete nouns drawn from a validated database (51),consisted of 5 to 10 letters, and contained no more than threesyllables. All words were balanced, so that the emotional and

Figure 1. Overview of the experimental design and salivarycortisol curves. Participants received a pill 210 minutes (pill1) and 30 minutes (pill 2) before memory encoding (t ¼ 0) atday 1 that could contain 10 mg hydrocortisone (Cort) orplacebo (albochin). Hydrocortisone administration in bothgroups significantly elevated salivary cortisol as comparedwith placebo but did not differ immediately before each pillintake or during baseline at day 2. Throughout the experi-ment, saliva samples were taken at 240, 210, 180, 150, 120,30, and 0 minutes before encoding; 30 and 60 minutes afterencoding; and before the surprise recall and recognition test24 hours later. Error bars represent standard error of themean (SEM). Significant Bonferroni corrected differenceswith placebo are depicted by ***p � .005. Rapid Cort;hydrocortisone 30 minutes before encoding, Slow Cort;hydrocortisone 210 minutes before encoding.

Figure 2. Experimental paradigm. To induce neutralversus emotional declarative memories, subjects wereshown 30 emotional and 30 neutral words on a smallgray rectangle (image not to scale), each in combinationwith unique color pictures of natural scenes or citylandscapes. During recognition testing the following day,participants were presented with the 60 old words thatwere presented during encoding on day 1, intermixedwith 60 foil words (30 emotional and 30 neutral) thatwere not presented before. Crucially, to assess contextdependency of memories, half of the old words werepresented in the same word-context combination(Intact) as at the first session, while the other half ofthe word-context pairs were rearranged to form newpairs. The foil items were presented against backgroundsviewed during the first session. Subjects indicatedwhether they recognized the word as having been seenduring the imagination task by responding old or new.

V.A. van Ast et al. BIOL PSYCHIATRY 2013;74:809–816 811

neutral words did not differ in terms of familiarity ratings, length,and/or amount of syllables but did significantly differ in terms ofvalence and arousal. Background pictures contained no distin-guishing objects; thus, distinctiveness of the contexts likely reliedon their unique spatial configuration. The task began with threepractice trials that at the same time functioned as a buffer forpossible primacy effects on subsequent memory. All memorytasks were presented using E-prime (version 2.0; PsychologySoftware Tools, Inc., Pittsburgh, Pennsylvania).

Surprise Memory Tests. See Supplementary Methods andMaterials in Supplement 1 for a description of the retrieval test.During the recognition test, participants were presented with the60 old words that were presented during memory encoding onday 1, intermixed with 60 foil words (30 emotional and 30 neutral)that were not presented before. Crucially, half of the old word-context pairs were presented in the same combination as at thefirst session, while the other half of the word-context pairs wererearranged to form new pairs. All of the foil words were combinedwith one of the old contexts. Consequently, each context waspresented twice in total. The test was self-paced. Subjects wereinstructed to indicate whether or not they recognized the word ashaving been seen during the encoding task, by responding old ornew, respectively. In addition, they were told that the backgroundimage may or may not be the same but that they should focus onthe words. Reaction times of the responses were recorded. SeeSupplementary Methods and Materials in Supplement 1 forfurther description of the recognition task.

Design and General ProcedureIn a between-subjects, placebo-controlled, double-blind study

design, participants were randomly assigned to either the rapidcort (hydrocortisone 30 min before testing), slow cort (hydro-cortisone 210 min before testing), or placebo group (seeexperimental outline in Figure 1). Testing took place between12:00 PM and 8:00 PM, when endogenous cortisol levels are stableand relatively low (52). Upon arrival, participants read theinformation brochure, were screened by means of an interviewto assess eligibility for participation, signed the informed consent,and filled out the trait scale of the State-Trait Anxiety Inventory(53) to probe potential individual differences in processing emo-tional material. Baseline self-reported mood states were assessedwith the State-Trait Anxiety Inventory state scale (STAI-S) (53) andthe Positive Affect and Negative Affect Schedule (PANAS) (54).A first baseline saliva sample was given as well. Then, we used theCube and Paper Test, a test of allocentric manipulation [i.e., the

ability to process configural cues among distal environmentalrelationships (55)] to control for possible individual differences inmemory contextualization processes. Directly following a secondbaseline sample, participants received their first pill (cortisol orplacebo). To ascertain that in the slow cort group 1) cortisol levelswould return to baseline, and 2) nongenomic effects would beabsent during encoding, a 3-hour waiting period was insertedduring which participants either read or studied. Four more salivasamples were obtained and participants ate lunch. Waiting tookplace in the same room as testing to prevent any (unwanted)additional context effects on memory. The second pill (cortisol orplacebo) was given 3 hours after the first. To reach peak plasmalevels (56) and activate nongenomic effects in the brain for therapid cort condition, a second resting period of 30 minutes wasinserted. Participants gave another saliva sample and again filledout the mood questionnaires (PANAS and STAI-S). After waiting,an eighth saliva sample was taken, followed by the encoding taskand two more samples. Time between the two testing sessionswas kept at 24 hours as much as possible. The session thefollowing day commenced with the mood questionnaires (STAI-S,PANAS), a saliva sample, and surprise retrieval and recognitiontests. A final postexperimental questionnaire assessed 1) motiva-tion and concentration to complete the encoding task, 2) whetherparticipants expected a memory test, and 3) insight into whichsubstance was received and at what time.

Data AnalysisStatistical analysis was performed using the SPSS statistical

software package (SPSS Inc., Chicago, Illinois). The effect ofhydrocortisone administration on salivary cortisol during the firsttesting day was assessed by means of a mixed analysis ofvariance (ANOVA) with the within-subjects factor sample (S1–S9) and between-subjects factor group. To assess participants'recognition memory performance as a function of context andemotion, hit rate (i.e., correct classification of previously presentedwords as old) was calculated per factor (i.e., neutral intact, neutralrearranged, emotional intact, emotional rearranged). False alarmrates (i.e., misclassification of new words as old) were calculatedas a function of emotion. Using the hit and false alarm rates, d-prime sensitivity index was calculated according to signaldetection theory. For this goal, hit rates >.975 were truncatedat .975, and false alarm �.025 rates were truncated at .025 (57).Recognition data were initially analyzed by means of an omnibusANOVA with the repeated measures factors emotion and context(intact, rearranged), while group was the between-subjects factor.

www.sobp.org/journal

812 BIOL PSYCHIATRY 2013;74:809–816 V.A. van Ast et al.

See Supplementary Methods and Materials in Supplement 1 for adescription of analysis of the retrieval data and additionalanalyses on the recognition data. A Greenhouse-Geisser proce-dure was used in case of violation of the sphericity assumption inANOVAs. Alpha level was set at .05 for all statistical analyses.

Results

Cortisol LevelsFigure 1 displays salivary cortisol levels for the three exper-

imental groups during the experiment. As expected, the ANOVA forsalivary cortisol levels showed a significant group � time inter-action (F8,472 ¼ 86.10, p � .001, ηp

2 ¼ .745). Planned comparisonswith placebo showed that in the slow cort group cortisol levelswere increased from 30 minutes after first pill intake until at least90 minutes later (S3–S5; all t59 �13.22, ps � .001). Right beforesecond pill intake, salivary cortisol levels of the slow cort group didnot differ from placebo (S6; t59 ¼ .68, ns). Further plannedcomparisons showed that in the rapid cort group, cortisol levelswere increased 30 minutes after second pill intake until the end ofthe first session (S7–S9; all t59 � 10.86, ps � .001). On day 2, therewere no differences in cortisol levels between the three groups(S10; F2,59 ¼ .05, ns). The exit interview showed that participantswere unable to identify the substance received (χ21 ¼ .180, ns).

Memory PerformanceSubjective ratings during encoding confirmed that the induc-

tion of emotional versus neutral mental images was successful (seeSupplementary Results and Figure S1 in Supplement 1). The cuedretrieval data revealed that memory retrieval was enhanced withintact contexts and that emotional memories were less context-dependent, but the timing of cortisol did not alter these processes.Overall, the rapid cort group retrieved less memories (for elabo-ration, see Supplementary Results and Figure S2 in Supplement 1).

The omnibus ANOVA on recognition d-prime data (Figure 3A)showed that context exerted a strong influence on memoryrecognition: memory was better for words that were presentedin their original context versus a rearranged context (F1,59 ¼ 178.44,p � .001, ηp

2 ¼ .752). Recognition of emotional versus neutralwords depended on context (context � emotion interaction; F1,59 ¼8.54, p ¼ .005, ηp

2 ¼ .126) as well. Furthermore, overall recognitionperformance was sensitive to the timing of cortisol (F2,59 ¼ 3.92, p ¼.025, ηp

2 ¼ .117), with the rapid group performing worse than boththe placebo (F1,59 ¼ 6.98, p ¼ .011, ηp

2 ¼ .106) and slow cort (F1,59¼ 4.76, p ¼ .033, ηp

2 ¼ .075) groups. This effect was caused by anenhanced false alarm rate in the rapid cort group (see Supplemen-tary Results in Supplement 1). Crucial to the hypothesis at hand,contextual dependency of emotional and neutral memories variedas a function of group (context � emotion � group interaction; F2,59¼ 3.26, p ¼ .045, ηp

2 ¼ .100).To decompose this three-way interaction, we separately ana-

lyzed neutral and emotional memories. Neutral memories werestrongly context-dependent (F1,59 ¼ 153.80, p � .001, ηp

2 ¼ .723),but context dependency was not modulated by cortisol (F2,59 ¼.06, p ¼ ns). Emotional memories were context-dependent as well(F1,59 ¼ 41.48, p � .001, ηp

2 ¼ .413). Importantly, this contextdependency was modulated by timing of cortisol (context �group; F2,59 ¼ 6.88, p ¼ .002, ηp

2 ¼ .189). To further decomposethis interaction, we entered difference scores of the rearrangedcondition minus the intact condition (i.e., contextualization; alarger contextualization index reflects greater contextual depend-ency of memories) in an ANOVA with the between-subjects factor

www.sobp.org/journal

group and within-subjects factor emotion (Figure 3B). Note thatthe difference scores within this analysis reflect the factor contextfrom the omnibus analysis. To investigate how exactly the timingof cortisol altered the contextual dependency of emotionalmemories, we used planned contrasts to directly compare con-textualization (i.e., the difference score) across groups. Theserevealed that compared with placebo, cortisol administrationseveral hours before encoding (slow cort) resulted in significantenhancement of the contextualization of emotional memories(t59 ¼ 2.01, p ¼ .049, d ¼ .621), while elevated cortisol levelsduring encoding (rapid cort) resulted in a marginally significantimpairment of the contextualization of emotional memories (t59 ¼ 1.85,p ¼ .069, d ¼ .561). Contrasting neutral and negative memorycontextualization within each group confirmed that negativememories tended to be decontextualized in the placebo group(F1,59 ¼ 3.51, p ¼ .066, ηp

2 ¼ .056), an effect that was present evenstronger in the rapid cortisol group (F1,59 ¼ 6.88, p ¼ .001, ηp

2 ¼.161). By contrast, neutral and negative memories were equallycontextualized in the slow cortisol group (F1,59 ¼ .41, ns).

The above relationships continued to exist when rerunningthe same analyses with subjective mood ratings from the first orsecond day as covariates, showing that the timing of cortisol didnot interact with mood to alter memory contextualization.Further, analysis of hit rates revealed generally the same patternof results as revealed by the d-prime analysis, although the effectswere only marginally significant (Supplementary Results andFigure S3 in Supplement 1). Finally, exploratory analyses of thesubjective quality of memories (remember-know judgments)indicated that recollection was unaffected by the timing ofcortisol, but slow cortisol enhanced familiarity sensitivity tocontext for emotional memories (Supplementary Results andFigure S4 in Supplement 1).

Discussion

A key question is under what circumstances cortisol may exertprotecting as opposed to detrimental effects on memory con-textualization. We tested the hypothesis that rapid cortisol effectsimpair the contextual dependency specifically of emotionalmemories and that delayed effects of cortisol enhance thecontextual dependency of subsequent emotional memories. Toprobe these two time domains, cortisol elevations were induceddirectly or some hours before memory encoding. In agreementwith our hypothesis, cortisol’s rapid effects impaired emotionalmemory contextualization, while cortisol’s slow effects enhancedthe contextualization of emotional memory. In contrast, thecontextualization of neutral memory remained unaltered bycortisol irrespective of the timing of the drug.

The rapid—presumably nongenomic—effects of cortisol showthat acute cortisol during encoding decreased contextualizationof emotional memories, though this was only marginally signifi-cant. Also, we found that acute cortisol during encodingenhanced the false alarm rate for the emotional words (Supple-mentary Results in Supplement 1), which explains why there wasa general memory impairment in the rapid group as measured byd-prime. The enhanced false alarm rate, however, cannot explainthe decreased contextualization in this group, since contextual-ization is the difference between memory performance in intactand rearranged context conditions, both calculated employingthe same false alarm rate (i.e., it is a constant). Together, theenhanced false alarm rate for emotional memories and reducedcontextualization by rapid cortisol corroborate previous studies,

Figure 3. Recognition performance. Graphs depictmemory performance indexed by D-prime as afunction of group (rapid cortisol [cort], slow cort, orplacebo), context (intact or rearranged), and emotion(neutral or emotional). Error bars represent standarderror of the mean. (A) Contextual dependency ofneutral and emotional memories. Difference scoresbetween D-prime values for the rearranged condi-tions subtracted from the intact conditions aredepicted. A higher difference score reflects greatercontextual dependency of memories. (B) As can beseen, cortisol did not in any way affect the con-textualization of neutral memories but pronouncedlyaffected the contextualization of emotional mem-ories. The significant planned contrast between theslow cort and placebo group is depicted with *p �.05 and the marginally significant planned contrastbetween the rapid cort group and placebo isdepicted with #p � .07.

V.A. van Ast et al. BIOL PSYCHIATRY 2013;74:809–816 813

all suggesting that stress may enhance consolidation of the gist ofan experience at the cost of detailed information (58–60).

On the other hand, slow—presumably genomic—effects ofcortisol during encoding enhanced contextualization of emo-tional memories. Theories of delayed effects of cortisol on thebrain suggest that these serve an adaptive function by promotinga variety of higher order cognitive functions (20,37–40). Ourresults are in line with this notion and demonstrate how slowcorticosteroid effects can drastically alter the way experiences areprocessed, stored, and subsequently remembered. Overall, thissuggests that delayed effects of cortisol not merely restorebaseline cognitive functioning (35) but rather lead to a redistrib-ution of resources toward superior executive functioning.

Taken together, the rapid and slow cortisol effects of cortisolappear to affect contextual processing in opposite directions. Asthe hippocampus is believed to subserve context effects onmemory (9–12) and is a main target of cortisol (13), it is likely thatthis structure mediated the present time-dependent effects ofcortisol. However, cortisol did not affect general contextualizationprocesses, because neutral information was not affected. Thus,apart from the hippocampus, our cortisol manipulations probablyaffected other brain areas as well. One likely area is the baso-lateral amygdala (BLA). The BLA is believed to be a site of storagefor memories of fearful or stressful experiences (61). In rodents,a single acute dose of corticosterone was sufficient to enhanceglutamatergic transmission through mineralocorticoid receptors(22) and facilitate noradrenergic actions on synaptic plasticity (62).Conversely, more than 1 hour after corticosterone administration,noradrenergic actions were gradually suppressed (62,63). In agree-ment, corticosterone administered before acute stress preventedsubsequent increases in anxiety; this was associated with reducedspinogenesis in the BLA (15). Perhaps, our cortisol manipulationsaffected the amygdala in a similar manner. This is possible, since

the encoding of emotional arousing words was shown to dependon an amygdala–hippocampal network (41) that can be amplifiedby stress (64). Remembrance of nonarousing words is supportedby more controlled processes mediated by a prefrontal cortex–hippocampal network (41). Taken together, we propose thatinteracting rapid cortisol effects—presumably in the hippocam-pal–amygdalar network—may have caused emotional memoriesto be consolidated in an isolated, context-independent, andgeneralized manner. Slower corticosteroid effects, on the otherhand, may have caused a shift in the brain toward morecognitively controlled processes, supported by a prefrontal cor-tex–hippocampal network, dealing with the emotional, arousingmemories in a comparable manner as the neutral ones.

Time-dependent effects of cortisol on contextualizationbecame apparent in the d-prime and hit rate recognition databut not in the retrieval data. Further exploratory analyses revealedthat recollection remained unaffected by cortisol, but slowcortisol enhanced familiarity sensitivity to context for emotionalmemories (though the effect was only marginally significant). Thislatter finding suggests that the enhanced contextualizationobserved for the slow cortisol group could be explained by anenhanced familiarity in intact contexts, rather than by alteredrecollection (Supplementary Results in Supplement 1). At firstsight, such a pattern of results may be consistent with the viewthat retrieval and recollection on the one hand and familiarity onthe other are subserved by separate, dissociable brain systems[e.g., (65)]. However, the fact that we found strong context andcortisol effects—both supposedly mediated by the hippocampus—on familiarity (or recognition) may be difficult to reconcile witha view in which the hippocampus is not involved in familiarity.Perhaps then, the current cortisol manipulation, and mostpronouncedly the slow effect, may have primarily affected weakmemories, strengthening their hippocampal components. This

www.sobp.org/journal

814 BIOL PSYCHIATRY 2013;74:809–816 V.A. van Ast et al.

would be in line with suggestions that weak memories are mostlikely to fit a familiarity profile [e.g., (66,67)]. Clearly, such ideasrequire more in-depth investigations.

Another unresolved, but related, question is whether theprocess of contextualization is associative in nature (i.e., compris-ing several separate constituents) or unitizing in nature (i.e.,representing separate stimulus components as a single com-pound) (68). Neuroimaging studies are particularly well suited toinvestigate which brain regions mediate cortisol effects on thesememory processes.

From a clinical perspective, our current findings may also shedlight on the question of how cortisol can sometimes contribute totraumatic memory formation (14,69) but in other cases can exertprotective effects (15). Here, we show that one and the samehormone can induce disparate effects on behavior: not only theabsolute levels of cortisol but also the delay between elevation incortisol levels and encoding (and perhaps also consolidation,reconsolidation, and retrieval) of traumatic events may play animportant role in modulating certain memory processes. This is inline with the temporal dynamics model (36), the emotionaltagging hypothesis (70), and theories by Joëls et al. (20), whichall predict differential effects by corticosteroids in the timedomain shortly after stress as opposed to several hours later(though directionality and region specificity of effects differamong these theories). This insight not only bears importantimplications for basic experimental research into the mechanismsof emotional memory modulation but also suggests that cortico-steroids might protect against traumatic memory formation whenadministered hours ahead of exposure to possible traumaticexperiences (for instance, in soldiers sent out to a war scene).Interestingly, cortisol has been suggested as a facilitating agent inexposure therapy (71–73). If genomic effects of cortisol indeedenhance the contextual dependency of memories, cliniciansshould take care to administer cortisol in close proximity ofexposure/extinction, as these are thought to be highly context-dependent processes (74). In such a way, maximum general-izability of extinction learning may be realized.

The present study has several limitations. First, we usedexogenous administration of hydrocortisone to probe the twotime domains of hormonal action. While this has the greatadvantage of allowing investigation of corticosteroid actions inisolation, it clearly differs from the situation after real-life stress.In line with the present findings of rapid cortisol effects, onestudy in humans showed that social stress impaired context-dependent memory of an object-location task but did notexplicitly relate this effect to stress-induced cortisol levels (75).Thus, it remains to be seen if disparate effects on contextualiza-tion of emotional memory are also seen with short versus longdelays between encoding and stress exposure and whetherthese effects indeed relate to stress-induced cortisol. Further,our design does not allow distinguishing between possibleeffects of cortisol on attention allocation (76) during encodingand effects on how information is subsequently maintained.Finally, besides timing, many other factors may moderate thecomplex relationship between cortisol and memory contextual-ization, such as hippocampal volume (55,77), the experience ofprevious trauma (78), chronic stress (79), amygdala reactivity (80),and genetic makeup (16). Further experimental research intothese modulatory mechanisms in healthy and clinical popula-tions is warranted.

In conclusion, this study shows distinct time-dependent effectsof cortisol on the contextualization of emotional memories. Theresults suggest that rapid effects of cortisol may lead to impaired

www.sobp.org/journal

emotional memory contextualization, while slow effects of cortisolmay confer protection against emotional memory generalization.

This work was supported by a TopTalent grant (VAvA,#021.002.103), a Vici grant (MK, #453-07-006), and a Vidi grant(MM, #452-09-007), all by the Netherlands Organization for ScientificResearch, and the priority program Brain and Cognition by theUniversity of Amsterdam.

We thank Clemens Kirschbaum, Ph.D., Technical University ofDresden, Germany, for analyzing the salivary cortisol samples.

The authors report no biomedical financial interests or potentialconflicts of interest.

Supplementary material cited in this article is available online athttp://dx.doi.org/10.1016/j.biopsych.2013.06.022.

1. Godden DR, Baddeley AD (1975): Context-dependent memory in twonatural environments: On land and underwater. Br J Psychol 66:325–331.

2. Tulving E, Thomson DM (1973): Encoding specificity and retrievalprocesses in episodic memory. Psychol Rev 80:352–373.

3. Smith SM, Vela E (2001): Environmental context-dependent memory:A review and meta-analysis. Psychon Bull Rev 8:203–220.

4. Elzinga BM, Bremner JD (2002): Are the neural substrates of memorythe final common pathway in posttraumatic stress disorder (PTSD)? JAffect Disord 70:1–17.

5. Acheson DT, Gresack JE, Risbrough VB (2012): Hippocampal dysfunc-tion effects on context memory: Possible etiology for post-traumaticstress disorder. Neuropharmacology 62:674–685.

6. Liberzon I, Sripada CS (2008): The functional neuroanatomy of PTSD: Acritical review. Prog Brain Res 167:151–169.

7. Ehlers A, Clark D (2000): A cognitive model of posttraumatic stressdisorder. Behav Res Ther 38:319–345.

8. Brewin C, Holmes E (2003): Psychological theories of posttraumaticstress disorder. Clin Psychol Rev 23:339–376.

9. Chun MM, Phelps EA (1999): Memory deficits for implicit contextualinformation in amnesic subjects with hippocampal damage. NatNeurosci 2:844–847.

10. Rasch B, Büchel C, Gais S, Born J (2007): Odor cues during slow-wavesleep prompt declarative memory consolidation. Science 315:1426–1429.

11. Davachi L (2006): Item, context and relational episodic encoding inhumans. Curr Opin Neurobiol 16:693–700.

12. Kalisch R, Korenfeld E, Stephan KE, Weiskopf N, Seymour B, Dolan RJ(2006): Context-dependent human extinction memory is mediated bya ventromedial prefrontal and hippocampal network. J Neurosci 26:9503–9511.

13. Joëls M, Baram TZ (2009): The neuro-symphony of stress. Nat RevNeurosci 10:459–466.

14. Kaouane N, Porte Y, Vallée M, Brayda-Bruno L, Mons N, Calandreau L,et al. (2012): Glucocorticoids can induce PTSD-like memory impair-ments in mice. Science 335:1510–1513.

15. Rao RP, Anilkumar S, McEwen BS, Chattarji S (2012): Glucocorticoidsprotect against the delayed behavioral and cellular effects of acutestress on the amygdala. Biol Psychiatry 72:466–475.

16. de Kloet ER, Joëls M, Holsboer F (2005): Stress and the brain: Fromadaptation to disease. Nat Rev Neurosci 6:463–475.

17. Joëls M, Fernández G, Roozendaal B (2011): Stress and emotionalmemory: A matter of timing. Trends Cogn Sci 15:280–288.

18. Schwabe L, Wolf OT, Oitzl MS (2010): Memory formation under stress:Quantity and quality. Neurosci Biobehav Rev 34:584–591.

19. Schwabe LL, Joëls MM, Roozendaal BB, Wolf OT, Oitzl MSM (2012):Stress effects on memory: An update and integration. NeurosciBiobehav Rev 36:1740–1749.

20. Joëls M, Sarabdjitsingh RA, Karst H (2012): Unraveling the timedomains of corticosteroid hormone influences on brain activity: Rapid,slow, and chronic modes. Pharmacol Rev 64:901–938.

21. Karst H, Berger S, Turiault M, Tronche F, Schütz G, Joëls M (2005):Mineralocorticoid receptors are indispensable for nongenomic mod-ulation of hippocampal glutamate transmission by corticosterone.Proc Natl Acad Sci U S A 102:19204–19207.

V.A. van Ast et al. BIOL PSYCHIATRY 2013;74:809–816 815

22. Karst H, Berger S, Erdmann G, Schütz G, Joëls M (2010): Metaplasticityof amygdalar responses to the stress hormone corticosterone. ProcNatl Acad Sci U S A 107:14449–14454.

23. Pruessner JC, Dedovic K, Khalili-Mahani N, Engert V, Pruessner M, BussC, et al. (2008): Deactivation of the limbic system during acutepsychosocial stress: Evidence from positron emission tomographyand functional magnetic resonance imaging studies. Biol Psychiatry 63:(7–7).

24. Lovallo WR, Robinson JL, Glahn DC, Fox PT (2010): Acute effects ofhydrocortisone on the human brain: An fMRI study. Psychoneuroen-docrinology 35:15–20.

25. Henckens MJAG Hermans EJ, Pu Z, Joëls M, Fernández G (2009):Stressed memories: How acute stress affects memory formation inhumans. J Neurosci 29:10111–10119.

26. van Stegeren AH, Wolf OT, Everaerd W, Scheltens P, Barkhof F,Rombouts SA (2007): Endogenous cortisol level interacts with nora-drenergic activation in the human amygdala. Neurobiol Learn Mem 87:57–66.

27. van Marle HJF, Hermans EJ, Qin S, Fernández G (2009): From specificityto sensitivity: How acute stress affects amygdala processing ofbiologically salient stimuli. Biol Psychiatry 66:649–655.

28. van Marle HJ, Hermans EJ, Qin S, Fernández G (2010): Enhancedresting-state connectivity of amygdala in the immediate aftermath ofacute psychological stress. Neuroimage 53:348–354.

29. van Stegeren AH, Roozendaal B, Kindt M, Wolf OT, Joëls M (2010):Interacting noradrenergic and corticosteroid systems shift humanbrain activation patterns during encoding. Neurobiol Learn Mem 93:56–65.

30. Schwabe L, Tegenthoff M, Hoffken O, Wolf OT (2010): Concurrentglucocorticoid and noradrenergic activity shifts instrumental behaviorfrom goal-directed to habitual control. J Neurosci 30:8190–8196.

31. Kukolja J, Schläpfer TE, Keysers C, Klingmüller D, Maier W, Fink GR,Hurlemann R (2008): Modeling a negative response bias in the humanamygdala by noradrenergic-glucocorticoid interactions. J Neurosci 28:12868–12876.

32. Enkel T, Gholizadeh D, Bohlen Und Halbach von O, Sanchis-Segura C,Hurlemann R, Spanagel R, et al. (2010): Ambiguous-cue interpretationis biased under stress- and depression-like states in rats. Neuro-psychopharmacology 35:1008–1015.

33. Elzinga BM, Roelofs K (2005): Cortisol-induced impairments of workingmemory require acute sympathetic activation. Behav Neurosci 119:98–103.

34. Wiegert O, Pu Z, Shor S, Joels M, Krugers H (2005): Glucocorticoid receptoractivation selectively hampers N-methyl-D-aspartate receptor dependenthippocampal synaptic plasticity in vitro. Neuroscience 135:403–411.

35. McEwen B (2007): Physiology and neurobiology of stress andadaptation: Central role of the brain. Physiol Rev 87:873.

36. Diamond DM, Campbell AM, Park CR, Halonen J, Zoladz PR (2007): Thetemporal dynamics model of emotional memory processing: A syn-thesis on the neurobiological basis of stress-induced amnesia,flashbulb and traumatic memories, and the Yerkes-Dodson law. NeuralPlast 2007:60803.

37. Oitzl MS, Reichardt HM, Joëls M, de Kloet ER (2001): Point mutation inthe mouse glucocorticoid receptor preventing DNA binding impairsspatial memory. Proc Natl Acad Sci U S A 98:12790–12795.

38. Henckens MJ, van Wingen GA, Joëls M, Fernández G (2011): Time-dependent corticosteroid modulation of prefrontal working memoryprocessing. Proc Natl Acad Sci U S A 108:5801–5806.

39. Henckens MJ, van Wingen GA, Joëls M, Fernández G (2012): Time-dependent effects of cortisol on selective attention and emotionalinterference: A functional MRI study. Front Integr Neurosci 6:66.

40. Henckens MJ, van Wingen GA, Joëls M, Fernández G (2010): Time-dependent effects of corticosteroids on human amygdala processing.J Neurosci 30:12725–12732.

41. Kensinger EA, Corkin S (2004): Two routes to emotional memory:Distinct neural processes for valence and arousal. Proc Natl Acad SciU S A 101:3310–3315.

42. Roozendaal B, Okuda S, de Quervain DJ-F, McGaugh JL (2006):Glucocorticoids interact with emotion-induced noradrenergic activa-tion in influencing different memory functions. Neuroscience 138:901–910.

43. Talamini LM, de Haan L, Nieman DH, Linszen DH, Meeter M (2010):Reduced context effects on retrieval in first-episode schizophrenia.PloS One 5:e10356.

44. Tambini A, Ketz N, Davachi L (2010): Enhanced brain correlations duringrest are related to memory for recent experiences. Neuron 65:280–290.

45. Yonelinas AP (2002): The nature of recollection and familiarity: Areview of 30 years of research. J Mem Lang 46:441–517.

46. Abercrombie HC, Kalin NH, Thurow ME, Rosenkranz MA, Davidson RJ(2003): Cortisol variation in humans affects memory for emotionallyladen and neutral information. Behav Neurosci 117:505–516.

47. Kensinger EA (2004): Remembering emotional experiences: Thecontribution of valence and arousal. Rev Neurosci 15:241–251.

48. Sara SJ (2009): The locus coeruleus and noradrenergic modulation ofcognition. Nat Rev Neurosci 10:211–223.

49. Craik FIM, Lockhart RS (1972): Levels of processing: A framework formemory research. J Verbal Learn Verbal Behav 11:671–681.

50. Bradley MM, Lang PJ (1994): Measuring emotion: The Self-AssessmentManikin and the Semantic Differential. J Behav Ther Exp Psychiatry 25:49–59.

51. Hermans D, de Houwer J (1994): Affective and subjective familiarityratings of 740 dutch words. Psychol Belg 34:115–139.

52. Pruessner JC, Wolf OT, Hellhammer DH, Buske-Kirschbaum A, Auer vonK, Jobst S, et al. (1997): Free cortisol levels after awakening: A reliablebiological marker for the assessment of adrenocortical activity. Life Sci61:2539–2549.

53. Spielberger CD, Gorsuch RL, Lusthene RE (1970): Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologist Press.

54. Watson D, Clark LA, Tellegen A (1988): Development and validationof brief measures of positive and negative affect: The PANAS scales.J Pers Soc Psychol 54:1063–1070.

55. Gilbertson MW, Williston SK, Paulus LA, Lasko NB, Gurvits TV, ShentonME, et al. (2007): Configural cue performance in identical twinsdiscordant for posttraumatic stress disorder: Theoretical implicationsfor the role of hippocampal function. Biol Psychiatry 62:513–520.

56. Czock D, Keller F, Rasche FM, Häussler U (2005): Pharmacokinetics andpharmacodynamics of systemically administered glucocorticoids. ClinPharmacokinet 44:61–98.

57. Stanislaw H, Todorov N (1999): Calculation of signal detection theorymeasures. Behav Res Methods Instrum Comput 31:137–149.

58. Payne JD, Nadel L, Allen JJB, Thomas KGF, Jacobs WJ (2002): Theeffects of experimentally induced stress on false recognition. Memory10:1–6.

59. Payne JD, Jackson ED, Ryan L, Hoscheidt S, Jacobs JW, Nadel L (2006):The impact of stress on neutral and emotional aspects of episodicmemory. Memory 14:1–16.

60. Mather M (2007): Emotional arousal and memory binding: An object-based framework. Pers Psychol Sci 2:33–52.

61. Schafe GE, Nader K, Blair HT, Ledoux JE (2001): Memory consolidationof Pavlovian fear conditioning: A cellular and molecular perspective.Trends Neurosci 24:540–546.

62. Liebmann L, Karst H, Joëls M (2009): Effects of corticosterone andthe beta-agonist isoproterenol on glutamate receptor-mediated syn-aptic currents in the rat basolateral amygdala. Eur J Neurosci 30:800–807.

63. Pu Z, Krugers HJ, Joëls M (2009): Beta-adrenergic facilitation ofsynaptic plasticity in the rat basolateral amygdala in vitro is graduallyreversed by corticosterone. Learn Mem 16:155–160.

64. Qin S, Hermans EJ, van Marle HJF, Fernández G (2012): Understandinglow reliability of memories for neutral information encoded understress: Alterations in memory-related activation in the hippocampusand midbrain. J Neurosci 32:4032–4041.

65. Brown MW, Aggleton JP (2001): Recognition memory: What are theroles of the perirhinal cortex and hippocampus? Nat Rev Neurosci 2:51–61.

66. Squire LR, Wixted JT, Clark RE (2007): Recognition memory andthe medial temporal lobe: A new perspective. Nat Rev Neurosci 8:872–883.

67. Kirwan CB, Wixted JT, Squire LR (2010): A demonstration that thehippocampus supports both recollection and familiarity. Proc NatlAcad Sci U S A 107:344–348.

68. Staresina BP, Davachi L (2010): Object unitization and associativememory formation are supported by distinct brain regions. J Neurosci30:9890–9897.

69. Cohen H, Liberzon I, Richter-Levin G (2009): Exposure to extremestress impairs contextual odour discrimination in an animal model ofPTSD. Int J Neuropsychopharmacol 12:291–303.

www.sobp.org/journal

816 BIOL PSYCHIATRY 2013;74:809–816 V.A. van Ast et al.

70. Richter-Levin G, Akirav I (2003): Emotional tagging of memoryformation–in the search for neural mechanisms. Brain Res Brain ResRev 43:247–256.

71. Soravia LM, Heinrichs M, Aerni A, Maroni C, Schelling G, Ehlert U, et al.(2006): Glucocorticoids reduce phobic fear in humans. Proc Natl AcadSci U S A 103:5585–5590.

72. de Quervain DJ, Margraf J (2008): Glucocorticoids for the treatment ofpost-traumatic stress disorder and phobias: A novel therapeuticapproach. Eur J Pharmacol 583:365–371.

73. de Quervain DJ, Bentz D, Michael T, Bolt OC, Wiederhold BK, Margraf J,Wilhelm FH (2011): Glucocorticoids enhance extinction-based psycho-therapy. Proc Natl Acad Sci U S A 108:6621–6625.

74. Bouton ME (2004): Context and behavioral processes in extinction.Learn Mem 11:485–494.

75. Schwabe L, Böhringer A, Wolf OT (2009): Stress disrupts context-dependent memory. Learn Mem 16:110–113.

www.sobp.org/journal

76. Putman P, Roelofs K (2011): Effects of single cortisol administrationson human affect reviewed: Coping with stress through adaptiveregulation of automatic cognitive processing. Psychoneuroendocrinol-ogy 36:439–448.

77. Gilbertson MW, Shenton ME, Ciszewski A, Kasai K, Lasko NB, Orr SP,Pitman RK (2002): Smaller hippocampal volume predicts pathologicvulnerability to psychological trauma. Nat Neurosci 5:1242–1247.

78. Resnick HS, Yehuda R, Pitman RK, Foy DW (1995): Effect of previoustrauma on acute plasma cortisol level following rape. Am J Psychiatry152:1675–1677.

79. McEwen BS (2003): Mood disorders and allostatic load. Biol Psychiatry54:200–207.

80. Admon R, Lubin G, Stern O, Rosenberg K, Sela L, Ben-Ami H, Hendler T(2009): Human vulnerability to stress depends on amygdala's predis-position and hippocampal plasticity. Proc Natl Acad Sci U S A 106:14120–14125.