Delayed effects of cortisol enhance fearmemory of trace conditioning

Sandra Cornelisse a,b,1,**, Vanessa A. van Ast c,d,1,*,Marian Joels a,b,1, Merel Kindt c,b,1

aDepartment of Neuroscience and Pharmacology, Rudolf Magnus Institute of Neuroscience, UMC Utrecht,Utrecht, The Netherlandsb Priority Program Brain and Cognition, University of Amsterdam, The NetherlandscDepartment of Clinical Psychology, University of Amsterdam, Amsterdam, The NetherlandsdDonders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen, Kapittelweg 29, 6525,EN The Netherlands

Received 22 March 2013; received in revised form 31 October 2013; accepted 18 November 2013

Psychoneuroendocrinology (2014) 40, 257—268

KEYWORDSFear conditioning;Cortisol;Memory;Trace and delay;Fear potentiated startle;PTSD

Summary Corticosteroids induce rapid non-genomic effects followed by slower genomic effectsthat are thought to modulate cognitive function in opposite and complementary ways. It is presentlyunknown how these time-dependent effects of cortisol affect fear memory of delay and traceconditioning. This distinction is of special interest because the neural substrates underlying thesetypes of conditioning may be differently affected by time-dependent cortisol effects. Delayconditioning is predominantly amygdala-dependent, while trace conditioning additionally requiresthe hippocampus. Here, we manipulated the timing of cortisol action during acquisition of delay andtrace fear conditioning, by randomly assigning 63 men to one of three possible groups: (1) receiving10 mg hydrocortisone 240 min (slow cort) or (2) 60 min (rapid cort) before delay and trace acquisi-tion, or (3) placebo at both times, in a double-blind design. The next day, we tested memory for traceand delay conditioning. Fear potentiated startle responses, skin conductance responses and uncon-ditioned stimulus expectancy scores were measured throughout the experiment. The fear potenti-ated startle data show that cortisol intake 240 min before actual fear acquisition (slow cort) uniquelystrengthened subsequent trace conditioned memory. No effects of cortisol delivery 60 min prior tofear acquisition were found on any measure of fear memory. Our findings emphasize that slow,presumably genomic, but not more rapid effects of corticosteroids enhance hippocampal-dependentfear memories. On a broader level, our findings underline that basic experimental research and

* Corresponding author at: Donders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen, Kapittelweg 29, 6525 EN, TheNetherlands. Tel.: +31 0 24 3610981.** Corresponding author at: Department of Neuroscience and Pharmacology, Rudolf Magnus Institute of Neuroscience, University MedicalCenter Utrecht, Universiteitsweg 100, 3584 CG Utrecht, The Netherlands. Tel.: +31 6 5312 3688; fax: +31 8875 69032.

E-mail addresses: sandra.cornelisse@gmail.com (S. Cornelisse), v.vanast@donders.ru.nl (V.A. van Ast).1 These authors contributed equally to this work.

Available online at www.sciencedirect.com

ScienceDirect

j our na l h omepa g e: www.e l se v ie r.c om/l oca te/ psyne ue n

0306-4530/$ — see front matter # 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.psyneuen.2013.11.013

clinically relevant pharmacological treatments employing corticosteroids should acknowledge thetiming of corticosteroid administration relative to the learning phase, or therapeutic intervention.# 2013 Elsevier Ltd. All rights reserved.

258 S. Cornelisse et al.

1. Introduction

Corticosteroids are released after stress and modulate learn-ing and memory processes via mineralocorticoid and gluco-corticoid receptors. These receptors are abundantlyexpressed in limbic brain areas, like the hippocampus (Reuland de Kloet, 1985). In humans, effects of cortisol on memoryhave predominantly been investigated for non-associativeand distinct emotional stimuli (De Quervain et al., 2009).

Stress is considered to be an important vulnerabilityfactor for anxiety disorders (Korte, 2001). It is generallyassumed that anxiety disorders originate from a learnedassociation between a previously neutral or ambiguous event(conditioned stimulus; CS) and an anticipated aversive event(unconditioned stimulus; US) (Mineka and Oehlberg, 2008).Thus, associative fear learning (i.e., discriminative fearconditioning) seems a suitable experimental model to deline-ate the mechanism by which corticosteroids contribute to thedevelopment of anxiety disorders.

Earlier animal and human studies have shown that stressand/or corticosteroids can indeed alter associative fear learn-ing (Rodrigues et al., 2009; Wolf et al., 2012). However, theeffects of stress hormones on fear acquisition seem equivocal,even when differences in the employed paradigms, dependentvariables, and sex are taken into account: Studies investigatingrelationships between cortisol and delay fear conditioning asmeasured by skin conductance reported enhanced (Jacksonet al., 2006; Zorawski et al., 2005, 2006), impaired (Van Astet al., 2012; Stark et al., 2006; Wolf et al., 2009), or unalteredacquisition (Merz et al., 2012a, 2013a) in men. In women, bothimpairing effects (Merz et al., 2013a; Wolf et al., 2009) and noeffect of cortisol on fear acquisition (Van Ast et al., 2012; Merzet al., 2012a; Stark et al., 2006; Tabbert et al., 2010) havebeen reported. Only two studies investigated the relationshipbetween post acquisition cortisol and fear retention, but thesestudies did not reveal any relationships between cortisol andretention in either men or women (Zorawski et al., 2005,2006). On the neural level, cortisol seemed not to influenceinstructed fear conditioning (Merz et al., 2012a, 2013a)neither in men nor in women, but when fear learning wasinvolved (either in a learned aware or an unaware sample),men displayed reduced neuronal fear responses after cortisolapplication, whereas women taking oral contraceptives exhi-bit enhanced fear responses on the neuronal level (Merz et al.,2010, 2012b, 2013b; Stark et al., 2006; Tabbert et al., 2010).

Studies investigating the relationship between cortisoland trace fear conditioning (or occasion setting) in menreported impairing (Wolf et al., 2012) or enhancing (Kuehlet al., 2010) effects on fear acquisition as measured by %eyeblink responses, enhancing effects as measured by fear-potentiated startle (Van Ast et al., 2012), or no effects asmeasured by skin conductance (Van Ast et al., 2012). Onestudy suggested impairing effects of cortisol on trace con-ditioning, but here sex effects were not investigated (Nees

et al., 2008). Taken together, the high variety of paradigms(e.g., delay or trace conditioning), dependent measurements(e.g., skin conductance responses vs. startle responses), andparticipants (men vs. women) used across studies, precludesvalid comparisons. In addition, the majority of these studiesonly focused on fear acquisition, that is, they did not examinelong-term memory aspects that are exquisitely sensitive tocorticosteroids. Revealing the effects of cortisol on retentionof fear might be very relevant to understanding maintenanceof fear in anxiety disorders, as opposed to cortisol effects onfear learning or mere fear expression (i.e., in instructedconditioning paradigms). For these reasons, we aimed totarget the expression of fear memory as assessed 1 day afterfear acquisition.

Corticosteroids are known to affect neurobiological pro-cesses in a time-dependent manner (Diamond et al., 2007;Joels et al., 2006); this may have added to seemingly incon-sistent cortisol effects across studies. Shortly after stress,corticosteroids interact with noradrenaline to synergisticallypromote rapid increases in neuronal activity (Karst et al.,2005, 2010). This effect is most pronouncedly sustained inthe basolateral amygdala (BLA) through a nongenomic path-way (Karst et al., 2005, 2010). In humans, such rapid corti-costeroid effects have indeed been described for emotion-and arousal-related brain areas, such as the amygdala (VanMarle et al., 2010). They promote habitual, reflex-like beha-viour (Schwabe et al., 2010) and attention (Vedhara et al.,2000), at the expense of goal-directed behaviour (Schwabeet al., 2010) and higher cognitive functioning (Elzinga andRoelofs, 2005). Indeed, shortly after stress a vigilance net-work including the amygdala is activated, as demonstratedwith fMRI in humans (see Hermans et al., 2011). Duringmemory encoding after stress, hippocampal and prefrontalcortex activity is generally suppressed (e.g. Qin et al., 2009;Van Stegeren et al., 2010). All in all, most observations inanimals and humans agree that rapid cortisol effects —presumably through nongenomic pathways and in interactionwith arousal-evoked central adrenergic release — enhanceamygdala activity while reducing hippocampal and PFC activ-ity. This may help the organism to focus and subsequentlyremember the most significant aspects of an event (Roozen-daal et al., 2006), at the cost of the more complex, cognitiveaspects (see also Joels et al., 2011; Karst et al., 2010).

By contrast, some hours after stress, slower long-lastinggenomic corticosteroid actions develop (De Kloet et al.,2005; Wiegert et al., 2005). Such delayed effects of cortisolon the brain are thought to restore homeostasis followingstressful periods (Diamond et al., 2007; Joels et al., 2006).Although slower genomic effects have not been investigatedin fear conditioning studies, other studies indicated thatgenomic effects promote consolidation (Barsegyan et al.,2010), cognitive self-control (Oitzl et al., 2001), enhanceworking memory (Henckens et al., 2011), promote sustainedattentional processing (Henckens et al., 2012), and

Cortisol effects on trace and delay fear memory 259

strengthen connectivity between the PFC and amygdala(Henckens et al., 2010). As such, slower genomic corticos-teroid effects may facilitate remembering a certain event ina more cognitively controlled manner.

Here, we examined the potential role of these time-domains in corticosteroid effects on discriminative humanfear conditioning, with special interest in the slow effectssince these have not been addressed so far. Hereto, wedesigned a within-subjects delay- and trace-conditioningparadigm, concurrently measuring fear potentiated startleresponses (FPS), skin conductance responses (SCR) and sub-jective shock expectancies (EXP) during fear acquisition and,importantly, a fear memory retention phase. In delay con-ditioning, the US and the CS co-terminate, whereas in traceconditioning a stimulus free period (the trace interval) passesbetween the offset of the CS and the US delivery. Thisdistinction is of special interest because different neuralsubstrates are believed to underlie delay and trace condi-tioning. Both animal and human research has identified theamygdala as predominantly involved in delay conditioning,while trace conditioning requires the dorsal hippocampus andthe prefrontal cortex in addition to the amygdala (Burmanet al., 2006; Knight et al., 2004). By including delay and traceconditioning in a within-subjects paradigm, we modelledrelatively simple, amygdala-dependent features and morecomplex, hippocampal- and prefrontal cortex-dependentfeatures of learning and memory, respectively.

We targeted time-dependent cortisol effects by rando-mizing participants into one of three experimental groups:(1) receiving 10 mg hydrocortisone either 240 or (2) 60 minbefore delay- and trace acquisition, or (3) placebo at bothtimes. Approximately 24 h later, fear memory expression(extinction and reinstatement) was tested. Trace and delayconditioning were assessed by a within-subjects design. Dur-ing acquisition, the delay conditioned stimulus (CSdel+) wasdirectly followed by the US (a shock), whereas for the traceconditioned stimulus (CStr+) a time gap was introducedbetween the end of the CS and the start of the US. A controlstimulus (CS�) was never followed by the US. During extinc-tion and reinstatement, these stimuli were presented unrein-forced (CSdel�, CStr�, CS�). Concerning time-dependenteffects of cortisol on subsequent fear memory expression(i.e., retention), we hypothesized that cortisol administra-tion 240 min before acquisition (slow cort) would enhancememory of trace conditioning, in line with the idea that slowgenomic corticosteroid actions promote consolidation ofmore complex, hippocampal- and prefrontal cortex-depen-dent stress-related memories for future use. By contrast,cortisol administration 60 min before acquisition (rapid cort)would enhance more simple, amygdala-dependent memoryof delay conditioning (Zorawski et al., 2005, 2006), whereastrace conditioning memory would be impaired, in line withthe idea that the most significant, aspects of an event will beremembered, at the cost of the more complex aspects.

2. Methods

2.1. Participants

Sixty-three male participants (mean age 21.4 years(SD = 2.9 years); mean body mass index 22.2 (SD = 2.9))

gave written informed consent. The local ethical commit-tee of the University of Amsterdam approved the study.Inclusion was conditional on having no past or presentpsychiatric or neurological condition, assessed by self-report. In addition, participants having any somaticor endocrine disease (e.g., acute asthma), or taking medica-tion known to influence central nervous system or endocrinesystems were excluded from participation. Participants wereasked not to eat, drink or smoke 2 h before participation.Participants were rewarded for participation with eithercourse credits or a monetary reward of s65. Due to technicalerrors, extinction and reinstatement data from one partici-pant were missing, as well as reinstatement data fromanother.

2.2. Physiological measures

2.2.1. Drug administration and assessmentA single dose of 10 mg hydrocortisone was employed toelevate cortisol to a level equivalent to acute stress. Hydro-cortisone and placebo (albochin) pills looked identical. Toassess salivary free cortisol concentrations for each partici-pant, Salivette collection devices (Sarstedt, Numbrecht,Germany) were employed. After testing, salivettes werestored at �25 8C. Upon completion of the study, sampleswere sent to Dresden (Technische Universitat, Dresden, Ger-many) for biochemical analysis.

2.2.2. Fear-potentiated startleStartle probes to induce fear-potentiated startle reflexeswere 104 dB, 40 ms bursts of white noise with a near instantrise time and delivered binaurally through headphones(Sennheiser, model HD25-1II). Sound pressure and dB levelwere calibrated using a sound level meter (Rion, NA-27,Japan). Startle reflexes were measured through electromyo-graphy (EMG) of the left orbicularis oculi muscle. Hereto,two 6 mm Ag/AgCl electrodes filled with a conductive gel(Signa, Parker) were placed approximately 1 cm under thepupil and 1 cm below the lateral canthus (Fridlund andCacioppo, 1986). A ground electrode was placed on theforehead, 1 cm below the hairline. The EMG amplifier con-sisted of two stages. The input stage or pre-amplifier had aninput resistance of 10,000 V. The EMG signal was set at afrequency response of DC-1500 Hz and was then amplified by200. A 50 Hz notch filter was used to reduce interferencefrom the mains noise. For the second stage, the signal wasamplified with a variable amplification factor of 0—100times. Finally, the EMG signal was digitized at a rate of1000 S/s.

2.2.3. Skin conductance responseElectrodermal activity was measured by two curved Ag/AgClelectrodes of 20 by 16 mm that were attached with adhesivetape to the medial phalanges of the first and third fingers ofthe left hand. The in-house built amplifier applied a sine-shaped excitation voltage (1 V peak-peak) of 50 Hz derivedfrom the mains frequency to the electrodes in order to detectchanges in the electrodermal activity. The signal from theinput device was led through a signal-conditioning amplifier.The analogue output was digitized at 1000 S/s by a 16-bit AD-converter (National Instruments, NI-6224).

260 S. Cornelisse et al.

2.2.4. US-expectancy ratingsUS-expectancies were measured continuously throughoutacquisition, extinction, and reinstatement, thus enablingus to collect ratings during the CS presentation and CS traces.Ratings were given by sliding a lever on a box (custom madefrom a joystick) that in turn operated a cursor on a scale thatshowed at the bottom of the computer screen. The scale wascontinuous, ranging from ‘certainly no electrical stimulus’through ‘certainly an electrical stimulus’. Expectancy datawere sampled at 1000 S/s. Startle responses, electrodermalactivity and US-expectancy ratings were recorded with thesoftware program VSSRP98 v6.0 (Versatile Stimulus ResponseRegistration Program, 1998; Technical Support Group of theDepartment of Psychology, University of Amsterdam).

2.3. Subjective measures

Participants filled out the Trait and State Anxiety Inventory(Spielberger et al., 1970) and the Positive Affect and NegativeAffect Schedule (Watson et al., 1988). In a post experimentalquestionnaire Unpleasantness ratings of the US and startleprobes were assessed on 9-point scales ranging from ‘‘notunpleasant’’ (1) to ‘‘extremely unpleasant’’ (9). We alsoprobed knowledge of which substance (placebo or hydrocorti-sone) was administered at the two time points. To assesscontingency awareness, participants indicated which CSs werefollowed by the US. Participants were classified as ‘‘aware’’ ifthey correctly recognized that the CSdel+ and the CStr+ werefollowed by a shock, while the CS� was not followed by a shock.

2.4. Experimental task

Three pictures (Langner et al., 2010) of male neutral facesserved as conditioned stimuli. Assignment of the three faces asCSdel+, CStr+ and CS� was counterbalanced across the three

Figure 1 Trace and delay conditioning procedure. CSs were presenwere presented 9 s after CS onset. In case of the CSdel+, the CS prob(denoted with: ‘trace probe’) were presented 9 s after CS offset. In0.5 s. Inter trial intervals (ITI) were 30 (�5) s. In case a noise alone (Nafter 10 (�5) s.

experimental groups. During the preconditioning phase, par-ticipants were presented a habituation phase consisting of 8startle probes (‘‘noise alone’’ trials, NA) with an inter-probeinterval of 20 � 5 s. Then, the CSdel+, CStr+ and CS�, werepresented unreinforced once in random order, along with oneNA trial. The acquisition phase commenced with a habituationphase, that consisted of 8 NA trial presentations. Duringacquisition, the four stimulus types (CSdel+, CStr+, CS�,NA) were randomized within blocks. These blocks were pre-sented 10 times in total. During acquisition, the CSdel+ and theCStr+ were followed by the US at all times, except for the firsttrial. We did so in order to obtain a measure of possible time-dependent effects of cortisol on responses to the stimulithemselves (i.e., from pre-conditioning to first presentationsof each trial during acquisition). CSs were presented for 10 s.Startle probes were presented during CSs (9 s after CS onset)and during traces (9 s after CS offset). In case of the CSdel+,the CS probe was followed by the US after 0.5 s. In case of theCStr+, the trace probe was followed by the US after 0.5 s. Intertrial intervals (ITI) were 30 (� 5) s. In case a noise alone (NA)probe was presented during an ITI, it followed the trace probeafter 10 (� 5) s (see Fig. 1).

During extinction, all stimulus types were again presented10 times, but CSs were never followed by the US. Fifteenseconds after the last extinction ITI, three unsignalled USs(respectively intermitted by 40 and 30 s time lag) werepresented. The reinstatement test phase started 30 s afterthe last reinstating US, during which each stimulus type wasagain presented four times. Trial timing was similar for allexperimental phases.

2.5. Design and procedure

In a between-subjects, placebo-controlled, double blindstudy design, participants were randomly assigned to either

ted for 10 s. Startle probes during CSs (denoted with: ‘CS probe’)e was followed by the US after 0.5 s. Startle probes during traces

case of the CStr+, the trace probe was followed by the US afterA) probe was presented during an ITI, it followed the trace probe

Figure 2 Overview of the experimental design and salivary cortisol levels during the experiment. Participants received a pill 240 and60 min prior to delay- and trace fear acquisition on day 1 (Pill 1 and Pill 2). The pills contained either hydrocortisone (10 mg) or placebo(albochin). Two baseline saliva samples were taken at the beginning of the experiment (t = �270 and t = �240), six more beforeacquisition (t = �210, �180, �150, �60, �30, 0), one directly after acquisition (t = 30) and a last sample was taken before theextinction procedure on day 2. In the slow cort condition cortisol levels were increased from 30 min after pill intake until 180 min laterand in the rapid cort condition cortisol levels were increased from 30 min after pill intake until the end of the first session. Error barsrepresent standard error of the mean (S.E.M.). Significant Bonferroni corrected differences with placebo are depicted by **p < 0.01;***p < 0.001.

Cortisol effects on trace and delay fear memory 261

the ‘slow cort’ (hydrocortisone 240 and placebo 60 min priorto acquisition), ‘rapid cort’ group (placebo 240 and hydro-cortisone 60 min prior to acquisition) or placebo group (pla-cebo at both 240 and 60 min prior to acquisition), resulting in21 participants per group. Testing took place in between12 am and 8 pm, when endogenous cortisol levels are stableand relatively low (Pruessner et al., 1997). A schematicoverview of the experiment is depicted in Fig. 2. Participantsfilled out the STAI-T, PANAS and STAI-S to assess baseline self-reported mood states and a first saliva sample (S1) was taken.To acquire physiological baseline responding to the startleprobes and the different CS stimuli, a preconditioning phasefollowed. A shock workup procedure was completed to estab-lish a shock level that was ‘‘unpleasant, but not painful’’.Participants were told that this phase involved baselineresponse assessment to the stimuli, and no shocks wouldbe administered. Then, directly following sample S2, parti-cipants received their first pill (hydrocortisone or placebo).The second pill (hydrocortisone or placebo) was given180 min later. While waiting, participants read or studied,and they were allowed to eat lunch. During this period,samples S3—S6 were taken. After Pill 2, participants gavesample S7 and again filled out the PANAS and STAI-S. Prior toconditioning, participants were told they would see threefaces, one of which would never be followed by the US, whilethe other two could be followed by the US at different timepoints. They were told to learn to predict whether and whenthey would receive an electrical stimulus. These explicitinstructions were given because awareness is a necessarycondition to acquire conditioned responses in hippocampusdependent tasks such as trace conditioning (Weike et al.,

2007). In addition, paradigms using explicit instructions con-cerning the CS� US relationships are best suited to investi-gate subsequent memory effects (e.g., Kindt et al., 2009;Merz et al., 2012a,b), as they reduce variability on day 1 thatcould perhaps explain subsequent variability in memory testson day 2. Sixty minutes after second pill intake, sample S8was taken and conditioning started. At the end of the firstexperimental day, participants gave a final saliva sample(S9). Time between the two testing sessions was kept at24 h, in order to substantiate consolidation of fear memories.The following day, participants were told that they would seethe same faces again and tested for memory of what waslearned the day before. A final set of mood questionnaires(STAI-S, PANAS) was filled out and sample S10 was taken,followed by extinction and reinstatement phases. Theexperiment was completed by the post-experimental ques-tionnaire. The experiment described here was part of a largerstudy into time-dependent cortisol effects on cognitive func-tion. The other task (i.e., a 5-min delay discounting task(Berns et al., 2007)) took place approximately 20 min prior tofear acquisition. Results on this task will be reported else-where.

2.6. Data reduction

Raw EMG data were conditioned to a band-pass between 28and 500 Hz. Relative to startle probe onset the latencywindow for the blink reflex was 0—120 ms and maximumpeak amplitude was determined within a window of 20—150 ms. Electrodermal responses during the CSs wereobtained by subtracting the baseline (1 s average before

262 S. Cornelisse et al.

CS onset) from the maximum absolute SCR score obtainedfrom a window of 1—9 s following CS onset. Electrodermalresponses for the traces were obtained by subtracting thebaseline (1 s average before CS onset) from the maximumabsolute SCR score obtained from a window of 1—9 s follow-ing trace onset. Raw SCR and FPS scores were standardizedacross all experimental phases and converted to T-scores(T = (z � 10) + 50). To obtain US-expectancy ratings for theCSs and the traces, 1 s averages were calculated immediatelybefore each startle probe onset.

2.7. Data analysis

In order to assess group differences in sample characteristics,univariate ANOVAs with the between subject factor Condition(slow cort, rapid cort, placebo) were employed.

Cortisol levels showed a skewed distribution with theShapiro—Wilk test of normality and were log transformed.The effect of hydrocortisone administration on salivary cor-tisol during the first session was assessed by means of a mixedANOVA with the within subject factor Time (S1—S9) andbetween subjects factor Condition. Bonferroni-correctedpost hoc comparisons were used to detect significant groupdifferences at all sample points.

Analyses to assess whether the trace and delay condition-ing paradigm was successful are described in the Supple-mentary Methods.

To check for cortisol effects on baseline FPS responses,FPS on baseline habituation trials and habituation trialsbefore acquisition were entered into a mixed ANOVA withPre—Post (before vs. after cort manipulation) and Trial aswithin subject factors and Condition as between subjectfactor. Further, to check for possible time-dependenteffects of cortisol on responses to the stimuli themselves,FPS and SCR data from the pre-conditioning phase and thefirst (non-reinforced) trials of acquisition were entered intomixed ANOVAs with Conditioning type (trace vs. delay con-ditioning), CS type (CSdel+, CStr+, CS�) and Pre-Post aswithin subject factors and Condition as between subjectfactor. To assess slow and rapid corticosteroid effects onacquisition, extinction and reinstatement, FPS, SCR and EXPdata were entered into mixed ANOVAs with Conditioningtype (trace vs. delay conditioning), CS type (CSdel+, CStr+,CS�) and Trial as within subject factors and Condition asbetween subject factor. If this overall analysis revealed aninteraction effect between Conditioning type, CS type andCondition, trace and delay conditioning were further ana-lyzed separately. Thus, for analysis of delay conditioning,data obtained during presentation of the CSs were enteredinto mixed ANOVAs with CS type and Trial as within subjectfactors and Condition as between subject factor. To assesstime-dependent corticosteroid effects on trace condition-ing, data obtained from the traces of the CSs were entered ina similar analysis. Planned comparisons to assess groupdifferences were performed on difference scores of therelevant CSs (i.e., traces of CStr vs. CS� for trace condition-ing and CS responses of CSdel vs. CS� and possibly CStr vs.CS� for delay conditioning). A Greenhouse—Geisser proce-dure was used in case of violation of the sphericity assump-tion in ANOVAs. Alpha level was set at 0.05 for all statisticalanalyses.

3. Results

3.1. Participant characteristics

The slow cort, rapid cort and placebo groups did not differ interms of age, body mass index, trait anxiety and anxietysensitivity (all Fs < 1.98, n.s.). Shock intensity ranged from 6to 56 mA (M = 22.42, SD = 11.74), and did not differ betweenthe three experimental groups (F = 1.11, n.s.). The subjec-tive evaluation of the US and the startle probe did not differeither (all Fs < 1.89, n.s.). Furthermore, groups did notdiffer on subjective mood (STAI-S, PANAS) (all Fs < 2.34,n.s.). Three participants were classified as unaware of theCS� US contingencies (one participant in each group). Thosethree participants were excluded from analyses. Inclusion ofall subjects led to similar results.

3.2. Manipulation check

Fig. 2 displays salivary cortisol levels for all groups (seesupplementary Table 1 for the raw cortisol data). Asexpected, the ANOVA for salivary cortisol levels showed asignificant Time � Condition interaction (F16,456 = 116.79,p < 0.001, h2 = 0.804). In addition, a significant main effectof Time (F8,456 = 24.20, p < 0.001, h2 = 0.302) and Condition(F2,57 = 25.41, p < 0.001, h2 = 0.471) emerged. Bonferroni-corrected post hoc comparisons showed that in the slow cortcondition cortisol levels were increased from 30 min after pillintake until 180 min later (S3—S6; ps < 0.002) but hadreturned back to baseline before and during the acquisitionphase (S7—S9). In the rapid cort condition, cortisol levelswere increased from 30 min after pill intake until the end ofthe first session (S7—S9; ps < 0.001). On day 2, cortisol levelsdid not differ between the three groups (S10; F = 0.102,n.s.). Participants were unable to identify the substancereceived during the exit interview (x2(1) = 0.185, n.s.; intotal 8 subjects correctly identified both pills they received;1 subject in the rapid cortisol group, 4 subjects in the slowcortisol group and 3 subjects in the placebo group).

3.3. Delay and trace conditioning paradigm

As described in the Supplementary Results and illustrated inSupplementary Figure 1, successful acquisition and extinc-tion was demonstrated for EXP, FPS and SCR.

3.4. Rapid and slow corticosteroid effects

3.4.1. Fear potentiated startleFig. 3 displays the fear potentiated startle responses duringthe acquisition, extinction and reinstatement phases indelay- and trace conditioning for the three experimentalgroups. Cortisol did not affect baseline FPS responses tohabituation NA trials before vs. after the cortisol manipula-tion (Pre-Post), as evidenced by the absence of a Condi-tion � Pre-Post � Trial interaction effect (F = 1.11; n.s.) ora Condition � Pre-Post interaction effect (F = 0.130; n.s.).Cortisol did also not affect FPS responses during the CSs whenpre-conditioning trials before pill intake were compared withthe first (un-reinforced) trials of acquisition, as evidenced by

Figure 3 Fear potentiated startle data acquired during the CS pictures (A, C, E) and during the CS traces (B, D, F) of all three trialtypes (i.e., CSdel+, CStr+ and CS�) for the placebo (A, B), rapid cort (C, D) and slow cort (E, F) groups. Note that participants in the slowcort group uniquely demonstrate enhanced startle responses during early extinction, suggesting that cortisol, given 4 h beforeacquisition, enhances fear memory of the trace stimulus 24 h later. Error bars represent standard error of the mean (S.E.M.). Significantpost hoc differences with placebo are depicted by *p < 0.05. Error bars represent standard error of the mean (S.E.M.).

Cortisol effects on trace and delay fear memory 263

the absence of a Condition � Pre-Post � CS Type � Condi-Conditioning type interaction (F = 2.00; n.s.), nor werethere any other significant interaction effects involving Con-dition and Pre-Post (Fs < 2.38, n.s.).

Acquisition: Analysis of acquisition data did not show anyinteraction effects involving Condition, CS type and Condi-tioning type (Fs < 1.23, n.s.), nor a main effect of Condition(F = 0.644, n.s.). This suggests that the cortisol manipulationdid not affect delay- or trace acquisition.

Extinction: For extinction, we did observe a significantinteraction effect between Conditioning type, Condition andCS type (F4,112 = 4.32, p = 0.003, h2 = 0.134). To further dis-entangle this effect, we analyzed delay and trace condition-ing separately. For delay conditioning, the cortisolmanipulation did not affect extinction (Fs < 1.98, n.s.).For trace conditioning however, we did observe a significantCS type � Condition interaction (F4,112 = 3.04, p = 0.020,

h2 = 0.026). If this reflected a memory retention effect,we expected it to originate mainly from the first part ofextinction. Indeed, we found a significant CS Type � Condi-Condition interaction during the first part (F4,112 = 3.42,p = 0.011, h2 = 0.109), but not the second part of extinction(F = 0.83, n.s.). Planned comparisons revealed that the slowcort group showed more differentiation of the CStr� vs. theCS� compared to the placebo group ( p = 0.027) and margin-ally more differentiation of the CStr� vs. the CS� comparedto the rapid cort group ( p = 0.058). Differentiation of theCStr� vs. the CS� did not differ between the placebo groupand the rapid cort group (n.s.). Also, differentiation betweenthe CSdel� and the CS� did not differ between any of thegroups (n.s.). Analysis of the last half of acquisition (i.e., thelast 5 trials) vs. the first half (i.e., the first 5 trials) ofextinction also revealed a Phase (last half acquisition vs.first half of extinction) � CS type � Condition effect

264 S. Cornelisse et al.

(F2,56 = 5.519, p = 0.003, h2 = 0.187). Follow-up analysesshowed that the difference between the differentialresponding from the end of acquisition to the beginning ofextinction was enhanced in the slow cortisol group as com-pared to the placebo group (F1,37 = 10.81, p = 0.002,h2 = 0.226), while the rapid and placebo group did not differ(F = 3.76, n.s.). This effect was not caused by altered base-line (NA) FPS responses during extinction (F = 0.56, n.s.).These same effects were significant when the raw (untrans-formed) data were used in analyses. Notably, although theslow cort group showed more differentiation of the CStr� vs.the CS�, both the placebo and the rapid cort group showedsignificantly higher startle responses on the CStr� vs. theCS� during the first half of extinction (Placebo: F1,19 = 6.55,p = 0.019, h2 = 0.256; Rapid Cort: F1,19 = 5.388, p = 0.031,h2 = 0.212). Together, this suggests that cortisol, given 4 hbefore acquisition, enhanced fear memory of the tracestimulus 24 h later.

Reinstatement: There were neither any interactioneffects involving Condition, CS type and Conditioning type(Fs < 0.27, n.s.), nor a main effect of Condition (F = 0.10,n.s.) suggesting that the cortisol manipulation did not affectdelay- and trace reinstatement.

3.4.2. Skin conductance responses and expectancyratingsAnalysis of the EXP and SCR data did not show any effectsof the experimental manipulation (see SupplementaryResults).

4. Discussion

A first aim of the present study was to develop a paradigm inwhich delay and trace conditioning can be tested withinsubjects, concurrently measuring FPS, SCR and EXP. We showthat participants are able to acquire delay and trace con-ditioning simultaneously, and retain the same pattern ofconditioned responses the following day on all dependentmeasures. Although delay and trace conditioning were testedwithin-subjects before (Cheng et al., 2008; Knight et al.,2004), the present paradigm adds in important ways tocurrent test models of associative fear memory, as this isthe first human trace conditioning study including FPS, and amemory retention test one day later. On a broader level, theinclusion of FPS and a retention test renders this paradigmmore relevant to understand the root and maintenance offear-related disorders in humans (Mineka and Oehlberg,2008).

Further, we tested time-domain effects of cortisol actionduring acquisition on subsequent memory. FPS data showedthat cortisol intake 240 min before fear acquisition — focus-ing on genomic actions — uniquely bolstered trace, but notdelay, fear memory. The hippocampus plays an additionalrole in trace vs. delay conditioning (Burman et al., 2006;Knight et al., 2004) and successful trace conditioning relies toa large extent on working memory (Carter et al., 2003) andselective attention (Han et al., 2003). These processes wererecently found to be enhanced by slow corticosteroid actions(Henckens et al., 2011, 2012). Thus, perhaps by means ofboosted executive processing after acquisition, we hereprovide the first evidence that slow (presumably gene-

mediated) corticosteroid effects during fear acquisitioncan strengthen subsequent fear memory in humans. Mostlikely though, this effect on trace conditioning is mediated bygenomic effects during fear conditioning and (early) conso-lidation (Joels et al., 2012) of trace memories that were notyet present in the rapid cort group. Because cortisol levelswere too low for non-genomic actions to occur, and therewere sufficient hours in between cortisol administration andencoding, likely genomic effects were at play. Such a sloweffect on consolidation fits with earlier findings in humansand rodents on the role of gene-mediated corticosteroidactions in consolidation (Van Ast et al., 2013; Oitzl et al.,2001). The most explicit example is provided by Oitzl et al.(2001) who showed that a mutant mouse in which GR cannotbind to the DNA (which prohibits any genomic actions) isimpaired with respect to its consolidation of spatial informa-tion. We are not aware of any other studies in humansinvestigating exogenous cortisol effects on the retention offear memory (as opposed to fear acquisition). Our findingssuggest that slow effects of cortisol do not merely restorebaseline functioning, but may actually lead to a redistribu-tion of neural resources towards superior executive function-ing (Henckens et al., 2011, 2012) that can enhancesubsequent fear memory as well.

Since cortisol levels during fear acquisition in the slow cortgroup already had gone back to baseline, not only theabsolute levels of cortisol per se may play an important rolein modulating certain memory processes, but also the timelag that exists between cortisol level enhancement andencoding. This is in line with the temporal dynamics model(Diamond et al., 2007), the emotional tagging hypothesis(Richter-Levin and Akirav, 2003), and theories by Joels andcolleagues (Joels et al., 2006), which all predict differentialeffects by corticosteroids in the time-domain shortly afterstress as opposed to several hours later. At the cellular level,rapid activation, followed by inhibition of neuroplasticity hasbeen observed in the hippocampus in response to severaltypes of stress manipulations (Richter-Levin and Akirav,2003), threat (Diamond et al., 2007), or cortisol administra-tion (Joels et al., 2011). How exactly these cellular effectsrelate to the overall efficiency of network function andconcomitant behavioural performance is not easy to predict.

We did not observe effects of cortisol administered 60 minbefore task onset on either delay or trace memory. This is inline with two other studies that did not find a relationshipbetween post-acquisition endogenous cortisol and fearretention (Zorawski et al., 2005, 2006). The delay betweenthe onset in rise of cortisol level (certainly in the brain) andthe start of the behavioural task (<45 min) was too short toallow gene-mediated actions to develop. However, fearacquisition took on average 25 min, so that by the timesubjects had completed the task cortisol possibly no longerexclusively exerted non-genomic corticosteroid actions.Because cortisol levels were still high at the beginning ofconsolidation, non-genomic effects may have been at work atthe same time as genomic effects. Perhaps, these processescancel each other out, but we can only speculate about this.Alternatively, we may not have had enough power to find aneffect, or the hydrocortisone dose that we utilized couldhave been too low (i.e., 10 mg). Also, cortisol by itself maynot be sufficient to strengthen cued fear memory as thisprocess may require strong noradrenergic activation; e.g.,

Cortisol effects on trace and delay fear memory 265

manipulations targeting the adrenergic system affect cuedfear memory to a great extent (Soeter and Kindt, 2011).Finally, the fact that the rapid group was only marginallydifferent from the slow group (as measured during the traces)may reflect that similar effects of cortisol occurred in theslow and rapid groups, but that these were just more clearlyexpressed in the slow group. This would fit with the putativepartial onset of genomic actions in the rapid group during theconsolidation phase. Taken together, we found no support forthe expected rapid effects on delay conditioned memory andin light with the raised issues mentioned before it is hard tomake any claims concerning changes in consolidation in thisgroup.

As described in Section 1, earlier studies into effects ofcortisol on acquisition of delay or trace acquisition have beenequivocal. It is important to note that the nature of theinstructions concerning the conditioning contingencies pre-vent us from drawing strong conclusions on fear learning(Merz et al., 2012a; Tabbert et al., 2011). Rather, the acqui-sition data reflect fear expression. Such paradigms are bestsuited to investigate subsequent memory effects (e.g., Kindtet al., 2009; Merz et al., 2012a,b), in which we were foremostinterested. Here we did not observe altered fear expressionon day 1 in either delay or trace conditioning, irrespective oftiming of corticosteroid treatment. The absence of cortisoleffects on psychophysiological measures due to cortisol is inline with other delay conditioning studies using explicitinstructions that found no effects of cortisol either on psy-chophysiological expression of fear, or cortisol inducedalterations of fear learning on the neural level (Merzet al., 2012a, 2013a). Taken together, we conclude thatcortisol does not exert time-dependent effects on theexpression of fear in delay and trace conditioning. In orderfor cortisol to exert strong effects on psychophysiologicalexpression of fear, less explicit prior conditioning instructionsmay be required.

The absence of an effect in the slow cort group at day 1seems in apparent contrast with the enhancing memoryeffects on trace conditioning the next day. This corroboratesthe idea of a dissociation between fear expression duringacquisition learning and subsequent long term fear memory.Such a dissociation has been convincingly illustrated bystudies in which pharmacological manipulations, adminis-tered before acquisition, affected the extinction processseveral days later while leaving the acquisition process intact(Soeter and Kindt, 2011, 2012). Post-learning processes mayaccount for this dissociation, as they induce the structuralchanges underlying the stabilization of a memory trace afterits acquisition (McGaugh, 1966). Thus, the implementation ofa test phase of associative fear memory after consolidation ismost appropriate to assess long-term fear memory.

Some considerations regarding our results should be men-tioned. First, effects of the experimental manipulation wereonly apparent on the startle measure. This corroborates theidea that SCR and FPS variables reflect rather distinct aspectsof conditioned responses, and can display opposite effectsafter certain experimental manipulations (Hamm and Weike,2005; Soeter and Kindt, 2011). At the same time, the absenceof strong (time-dependent) effects of cortisol on any otherdependent variable also suggests that the present findingsare preliminary and require replication. Second, this studyused a within-subjects discrimination procedure that likely

recruited abilities over and beyond those that are needed fordelay or trace conditioning alone (Knight et al., 2004). Wecannot exclude the possibility that delay and trace condi-tioning may influence each other. Further, we only testedmen, to exclude unwanted effects of the menstrual cycle; forinstance, women are known to display different HPA axisreactivity than men (Kajantie and Philips, 2006). Further,interactions between sex and stress hormone levels can haveimportant consequences for fear learning and its laterexpression (Merz et al., 2010; Milad et al., 2010). Indeed,women in the follicle or luteal phase seem to respond tocortisol manipulations in comparable ways as men, while oralcontraceptive use enhances differential fear learning on theneural level, in an implicit fear-conditioning paradigm (Merzet al., 2012b; Wolf et al., 2009). It is important to realize thatwithin the context of etiological models of anxiety disorderssex specific sensitivity to stressful events has been repeat-edly associated with the higher prevalence of mood andanxiety disorders in women (Cahill, 2006; Kessler et al.,1995). Consequently, the omission of women in this studyis an important limitation to how these data may be trans-lated to women, or the clinic. Also, this is the first humanstudy to include an extinction and reinstatement phase in awithin-subjects trace-delay (or trace alone) conditioningparadigm. Therefore, it is not known what a trace extinctioncurve typically looks like under basal (e.g., placebo) circum-stances. However, analyses of the difference between theCStr and the CS� startle data do show fear retention onstartle responses during the first half of extinction in theplacebo as well as both cortisol groups. Further, we must notethat a full reinforcement schedule during acquisition wasused (after one initial unreinforced trial) during acquisition,which may have fastened extinction processes. However, aspilot testing showed that including a partial reinforcementrate hampered fear acquisition itself, we utilized a fullreinforcement acquisition schedule. Concerning reinstate-ment, the present study showed differential reinstatementfor both trace and delay conditioning as measured by the US-expectancies, but non-differential reinstatement as mea-sured by FPS or SCR. Non-differential reinstatement hasearlier been found in even simpler paradigms (e.g., Dirikxet al., 2009). Thus, most likely, due to the complicatedparadigm, the reinstatement procedure was not sensitiveenough to detect effects of our subtle manipulation. Finally,a dose of 10 mg hydrocortisone was used, while in this field ofresearch a typical dose of 30 mg is being used (Merz et al.,2012a; Stark et al., 2006; Tabbert et al., 2010). Severalstudies have shown that experimental effects can alter (oreven flip over to the opposite side) depending on the dose(Buchanan et al., 2001). Hence, dose-response studies maybe necessary to get a more complete view on the effects ofcortisol on fear memory.

In summary, we show that slow, presumably genomic,effects of cortisol enhance memory for more complex, hip-pocampus-mediated fear learning, but not for simple amyg-dala-mediated fear learning. The present findings emphasizethat corticosteroids can affect associative fear memory in atime-dependent manner, adding to earlier findings showingtime-dependent cortisol effects on other neurobiologicalprocesses (Henckens et al., 2010, 2011, 2012). The insightthat cortisol exerts time-dependent effects on associativefear memory has implications for experimental human

266 S. Cornelisse et al.

research into corticosteroid effects on associative fear mem-ory, and perhaps more generally, also on declarative memoryresearch, fields that are often characterized by disparatefindings. The majority of studies have typically tested mem-ory performance 30—120 min after cortisol administration.Since gene-mediated transcriptional changes are discerniblealready 1 h after cortisol exposure (Morsink et al., 2006), themajority of current human experimental research has testedin a time window where both genomic and non-genomicprocesses are active, complicating a straightforward inter-pretation of results. Clinically, this is very relevant for thedevelopment of pharmacological treatments, especiallysince cortisol has been suggested as a pharmacological addon to cognitive-behavioural intervention in anxiety disorders(De Quervain and Margraf, 2008).

Role of funding source

The funding sources did not play a role in carrying out theresearch described herein, the writing of the review, or in thedecision to submit the paper for publication.

Conflict of interest statement

None declared.

Acknowledgements

This work was supported by the priority program Brain andCognition by the University of Amsterdam. Vanessa van Ast issupported by a TopTalent grant (#021.002.103) and MerelKindt is supported by a VICI grant (#453-07-006) by theNetherlands Organization for Scientific Research. We thankBert Molenkamp for technical assistance and Clemens Kirsch-baum, Ph.D., Technical University of Dresden, Germany, foranalyzing the salivary cortisol samples.

Appendix A. Supplementary data

Supplementary material related to this article canbe found, in the online version, at http://dx.doi.org/10.1016/j.psyneuen.2013.11.013.

References

Van Ast, V.A., Vervliet, B., Kindt, M., 2012. Contextual control overexpression of fear is affected by cortisol. Front. Behav. Neurosci.11, 6—67.

Van Ast, V.A., Cornelisse, S., Meeter, M., Joels, M., Kindt, M., 2013.Time-dependent effects of cortisol on the contextualization ofemotional memories. Biol. Psychiatry 74, 809—816.

Barsegyan, A., Mackenzie, S.M., Kurose, B.D., McGaugh, J.L., Roo-zendaal, B., 2010. Glucocorticoids in the prefrontal cortex en-hance memory consolidation and impair working memory by acommon neural mechanism. Proc. Natl. Acad. Sci. U. S. A. 107,16655—16660.

Berns, G.S., Laibson, D., Loewenstein, G., 2007. Intertemporalchoice — toward an integrative framework. Trends Cogn. Sci.11, 482—488.

Buchanan, T.W., Brechtel, A., Sollers, J.J., Lovallo, W.R., 2001.Exogenous cortisol exerts effects on the startle reflex independent

of emotional modulation. Pharmacol. Biochem. Behav. 68, 203—210.

Burman, M.A., Starr, M.J., Gewirtz, J.C., 2006. Dissociable effects ofhippocampus lesions on expression of fear and trace fear condi-tioning memories in rats. Hippocampus 16, 103—113.

Cahill, L., 2006. Why sex matters for neuroscience. Nat. Rev. Neu-rosci. 7, 477—484.

Carter, R.M., Hofstotter, C., Tsuchiya, N., Koch, C., 2003. Workingmemory and fear conditioning. Proc. Natl. Acad. Sci. U. S. A. 100,1399—1404.

Cheng, D.T., Disterhoft, J.F., Power, J.M., Ellis, D.A., Desmond, J.E.,2008. Neural substrates underlying human delay and trace eye-blink conditioning. Proc. Natl. Acad. Sci. U. S. A. 105, 8108—8113.

Diamond, D.M., Campbell, A.M., Park, C.R., Halonen, J., Zoladz,P.R., 2007. The temporal dynamics model of emotional memoryprocessing: a synthesis on the neurobiological basis of stress-induced amnesia, flashbulb and traumatic memories, and theYerkes-Dodson law. Neural Plast. 2007, 60803.

Dirikx, T., Vansteenwegen, D., Eelen, P., Hermans, D., 2009. Non-differential return of fear in humans after a reinstatementprocedure. Acta Psychol. (Amst.) 130, 175—182.

Elzinga, B.M., Roelofs, K., 2005. Cortisol-induced impairments ofworking memory require acute sympathetic activation. Behav.Neurosci. 119, 98—103.

Fridlund, A.J., Cacioppo, J.T., 1986. Guidelines for human electro-myographic research. Psychophysiology 23, 567—589.

Hamm, A.O., Weike, A.I., 2005. The neuropsychology of fear learningand fear regulation. Int. J. Psychophysiol. 57, 5—14.

Han, C.J., O’Tuathaigh, C.M., van Trigt, L., Quinn, J.J., Fanselow,M.S., Mongeau, R., Koch, C., Anderson, D.J., 2003. Trace but notdelay fear conditioning requires attention and the anterior cin-gulate cortex. Proc. Natl. Acad. Sci. U. S. A. 100, 13087—13092.

Henckens, M.J.A.G., van Wingen, G.A., Joels, M., Fernandez, G.,2010. Time-dependent effects of corticosteroids on human amyg-dala processing. J. Neurosci. 30, 12725—12732.

Henckens, M.J.A.G., van Wingen, G.A., Joels, M., Fernandez, G.,2011. Time-dependent corticosteroid modulation of prefrontalworking memory processing. Proc. Natl. Acad. Sci. U. S. A. 108,5801—5806.

Henckens, M.J.A.G., van Wingen, G.A., Joels, M., Fernandez, G.,2012. Time-dependent effects of cortisol on selective attentionand emotional interference: a functional MRI study. Front. Integr.Neurosci. 6, 66.

Hermans, E.J., van Marle, H.J.F., Ossewaarde, L., Henckens,M.J.A.G., Qin, S., van Kesteren, M.T.R., Schoots, V.C., Cousijn,H., Rijpkema, M., Oostenveld, R., et al., 2011. Stress-relatednoradrenergic activity prompts large-scale neural network recon-figuration. Science 334, 1151—1153.

Jackson, E.D., Payne, J.D., Nadel, L., Jacobs, W.J., 2006. Stressdifferentially modulates fear conditioning in healthy men andwomen. Biol. Psychiatry 59, 516—522.

Joels, M., Pu, Z., Wiegert, O., Oitzl, M.S., Krugers, H.J., 2006.Learning under stress: how does it work? Trends Cogn. Sci. (Regul.Ed.) 10, 152—158.

Joels, M., Fernandez, G., Roozendaal, B., 2011. Stress and emotionalmemory: a matter of timing. Trends Cogn. Sci. (Regul. Ed.) 15,280—288.

Joels, M., Sarabdjitsingh, R.A., Karst, H., 2012. Unraveling the timedomains of corticosteroid hormone influences on brain activity:rapid, slow, and chronic modes. Pharmacol. Rev. 64, 901—938.

Karst, H., Berger, S., Turiault, M., Tronche, F., Schutz, G., Joels, M.,2005. Mineralocorticoid receptors are indispensable for nonge-nomic modulation of hippocampal glutamate transmission bycorticosterone. Proc. Natl. Acad. Sci. U. S. A. 102, 19204—19207.

Karst, H., Berger, S., Erdmann, G., Schutz, G., Joels, M., 2010.Metaplasticity of amygdalar responses to the stress hormonecorticosterone. Proc. Natl. Acad. Sci. U. S. A. 107, 14449—14454.

Cortisol effects on trace and delay fear memory 267

Kessler, R.C., Sonnega, A., Bromet, E., Hughes, M., Nelson, C.B.,1995. Posttraumatic stress disorder in the National ComorbiditySurvey. Arch. Gen. Psychiatry 52, 1048—1060.

Kindt, M., Soeter, M., Vervliet, B., 2009. Beyond extinction: erasinghuman fear responses and preventing the return of fear. Nat.Neurosci. 12, 256—258.

De Kloet, E.R., Joels, M., Holsboer, F., 2005. Stress and the brain:from adaptation to disease. Nat. Rev. Neurosci. 6, 463—475.

Kajantie, E., Philips, D.I.W., 2006. The effects of sex and hormonalstatus on the physiological response to acute social stress. Psy-choneuroendocrinology 31, 151—178.

Knight, D.C., Cheng, D.T., Smith, C.N., Stein, E.A., Helmstetter, F.J.,2004. Neural substrates mediating human delay and trace fearconditioning. J. Neurosci. 24, 218—228.

Korte, S.M., 2001. Corticosteroids in relation to fear, anxiety andpsychopathology. Neurosci. Biobehav. Rev. 25, 117—142.

Kuehl, L.K., Lass-Hennemann, J., Richter, S., Blumenthal, T.D., Oitzl,M., Schachinger, H., 2010. Accelerated trace eyeblink condition-ing after cortisol IV-infusion. Neurobiol. Learn. Mem. 94, 547—553.

Langner, O., Dotsch, R., Bijlstra, G., Wigboldus, D.H.J., Hawk, S.T.,van Knippenberg, A., 2010. Presentation and validation of theRadboud Faces Database. Cogn. Emot. 24, 1377—1388.

Van Marle, H.J.F., Hermans, E.J., Qin, S., Fernandez, G., 2010.Enhanced resting-state connectivity of amygdala in the immedi-ate aftermath of acute psychological stress. Neuroimage 53,348—354.

McGaugh, J.L., 1966. Time-dependent processes in memory storage.Science 153, 1351—1358.

Merz, C.J., Tabbert, K., Schweckendiek, J., Klucken, T., Vaitl, D.,Stark, R., Wolf, O.T., 2010. Investigating the impact of sex andcortisol on implicit fear conditioning with fMRI. Psychoneuroen-docrinology 35, 33—46.

Merz, C.J., Tabbert, K., Schweckendiek, J., Klucken, T., Vaitl, D.,Stark, R., Wolf, O.T., 2012a. Neuronal correlates of extinctionlearning are modulated by sex hormones. Soc. Cogn. Affect.Neurosci. 7, 819—830.

Merz, C.J., Tabbert, K., Schweckendiek, J., Klucken, T., Vaitl, D.,Stark, R., Wolf, O.T., 2012b. Oral contraceptive usage alters theeffects of cortisol on implicit fear learning. Horm. Behav. 62,531—538.

Merz, C.J., Stark, R., Vaitl, D., Tabbert, K., Wolf, O.T., 2013a. Stresshormones are associated with the neuronal correlates ofinstructed fear conditioning. Biol. Psychol. 92, 82—89.

Merz, C.J., Wolf, O.T., Schweckendiek, J., Klucken, T., Vaitl, D.,Stark, R., 2013b. Stress differentially affects fear conditioning inmen and women. Psychoneuroendocrinology.

Milad, M.R., Zeidan, M.A., Contero, A., Pitman, R.K., Klibanski, A.,Rauch, S.L., Goldstein, J.M., 2010. The influence of gonadalhormones on conditioned fear extinction in healthy humans.Neuroscience 168, 652—658.

Mineka, S., Oehlberg, K., 2008. The relevance of recent develop-ments in classical conditioning to understanding the etiology andmaintenance of anxiety disorders. Acta Psychol. (Amst.) 127,567—580.

Morsink, M.C., Steenbergen, P.J., Vos, J.B., Karst, H., Joels, M., DeKloet, E.R., Datson, N.A., 2006. Acute activation of hippocam-pal glucocorticoid receptors results in different waves of geneexpression throughout time. J. Neuroendocrinol. 18, 239—252.

Nees, F., Richter, S., Lass-Hennemann, J., Blumenthal, T.D., Scha-chinger, H., 2008. Inhibition of cortisol production by metyraponeenhances trace, but not delay, eyeblink conditioning. Psycho-pharmacology (Berl.) 199, 183—190.

Oitzl, M.S., Reichardt, H.M., Joels, M., de Kloet, E.R., 2001. Pointmutation in the mouse glucocorticoid receptor preventing DNAbinding impairs spatial memory. Proc. Natl. Acad. Sci. U. S. A. 98,12790—12795.

Pruessner, J.C., Wolf, O.T., Hellhammer, D.H., Buske-Kirschbaum, A.,von Auer, K., Jobst, S., Kaspers, F., Kirschbaum, C., 1997. Freecortisol levels after awakening: a reliable biological marker forthe assessment of adrenocortical activity. Life Sci. 61, 2539—2549.

Qin, S., Hermans, E.J., van Marle, H.J.F., Luo, J., Fernandez, G.,2009. Acute psychological stress reduces working memory-relat-ed activity in the dorsolateral prefrontal cortex. Biol. Psychiatry66, 25—32.

De Quervain, D.J.-F., Margraf, J., 2008. Glucocorticoids for thetreatment of post-traumatic stress disorder and phobias: a noveltherapeutic approach. Eur. J. Pharmacol. 583, 365—371.

De Quervain, D.J.-F., Aerni, A., Schelling, G., Roozendaal, B., 2009.Glucocorticoids and the regulation of memory in health anddisease. Front. Neuroendocrinol. 30, 358—370.

Reul, J.M., de Kloet, E.R., 1985. Two receptor systems for cortico-sterone in rat brain: microdistribution and differential occupa-tion. Endocrinology 117, 2505—2511.

Richter-Levin, G., Akirav, I., 2003. Emotional tagging of memoryformation — in the search for neural mechanisms. Brain Res. BrainRes. Rev. 43, 247—256.

Rodrigues, S.M., LeDoux, J.E., Sapolsky, R.M., 2009. The influence ofstress hormones on fear circuitry. Annu. Rev. Neurosci. 32, 289—313.

Roozendaal, B., Okuda, S., de Quervain, D.J.-F., McGaugh, J.L.,2006. Glucocorticoids interact with emotion-induced noradren-ergic activation in influencing different memory functions. Neu-roscience 138, 901—910.

Schwabe, L., Schachinger, H., de Kloet, E.R., Oitzl, M.S., 2010.Corticosteroids operate as a switch between memory systems.J. Cogn. Neurosci. 22, 1362—1372.

Soeter, M., Kindt, M., 2011. Noradrenergic enhancement of associa-tive fear memory in humans. Neurobiol. Learn. Mem. 96, 263—271.

Soeter, M., Kindt, M., 2012. Stimulation of the noradrenergic systemduring memory formation impairs extinction learning but not thedisruption of reconsolidation. Neuropsychopharmacology 37,1204—1215.

Spielberger, C., Gorsuch, R., Lusthene, R., 1970. Manual for theState-Trait Anxiety Inventory. Psychologist Press, Palo Alto, CA.

Stark, R., Wolf, O.T., Tabbert, K., Kagerer, S., Zimmermann, M.,Kirsch, P., Schienle, A., Vaitl, D., 2006. Influence of the stresshormone cortisol on fear conditioning in humans: evidence for sexdifferences in the response of the prefrontal cortex. Neuroimage32, 1290—1298.

Van Stegeren, A.H., Roozendaal, B., Kindt, M., Wolf, O.T., Joels, M.,2010. Interacting noradrenergic and corticosteroid systems shifthuman brain activation patterns during encoding. Neurobiol.Learn. Mem. 93, 56—65.

Tabbert, K., Merz, C.J., Klucken, T., Schweckendiek, J., Vaitl, D.,Wolf, O.T., Stark, R., 2010. Cortisol enhances neural differentia-tion during fear acquisition and extinction in contingency awareyoung women. Neurobiol. Learn. Mem. 94, 392—401.

Tabbert, K., Merz, C.J., Klucken, T., Schweckendiek, J., Vaitl, D.,Wolf, O.T., Stark, R., 2011. Influence of contingency awareness onneural, electrodermal and evaluative responses during fear con-ditioning. Soc. Cogn. Affect. Neurosci. 6, 495—506.

Vedhara, K., Hyde, J., Gilchrist, I.D., Tytherleigh, M., Plummer, S.,2000. Acute stress, memory, attention and cortisol. Psychoneur-oendocrinology 25, 535—549.

Watson, D., Clark, L.A., Tellegen, A., 1988. Development and vali-dation of brief measures of positive and negative affect: thePANAS scales. J. Pers. Soc. Psychol. 54, 1063—1070.

Weike, A.I., Schupp, H.T., Hamm, A.O., 2007. Fear acquisitionrequires awareness in trace but not delay conditioning. Psycho-physiology 44, 170—180.

Wiegert, O., Pu, Z., Shor, S., Joels, M., Krugers, H., 2005. Glucocorti-coid receptor activation selectively hampers N-methyl-D-aspartate

268 S. Cornelisse et al.

receptor dependent hippocampal synaptic plasticity in vitro. Neu-roscience 135, 403—411.

Wolf, O.T., Minnebusch, D., Daum, I., 2009. Stress impairs acquisitionof delay eyeblink conditioning in men and women. Neurobiol.Learn. Mem. 91, 431—436.

Wolf, O.T., Bauser, D.S., Daum, I., 2012. Eyeblink conditional dis-crimination learning in healthy young men is impaired after stressexposure. Psychophysiology 49, 164—171.

Zorawski, M., Cook, C.A., Kuhn, C.M., LaBar, K.S., 2005. Sex, stress,and fear: individual differences in conditioned learning. Cogn.Affect. Behav. Neurosci. 5, 191—201.

Zorawski, M., Blanding, N.Q., Kuhn, C.M., LaBar, K.S., 2006. Effectsof stress and sex on acquisition and consolidation of human fearconditioning. Learn. Mem. 13, 441—450.