Psychoneuroendocrinology (2015) 58, 91—103

Available online at www.sciencedirect.com

ScienceDirect

j ourna l h om epa ge : www.elsev ier .com/ locate /psyneuen

Does cortisol influence core executivefunctions? A meta-analysis of acute cortisoladministration effects on working memory,inhibition, and set-shifting

Grant S. Shieldsa,∗, Joseph C. Bonnera, Wesley G. Moonsb

a University of California, Davis, CA 95616, USAb Moons Analytics, San Diego, CA 92101, USA

Received 7 January 2015; received in revised form 4 April 2015; accepted 22 April 2015

KEYWORDSCortisol;Executive function;Meta-analysis;Working memory;Inhibition;Set-shifting;Executive attention;Response inhibition;Cognitive inhibition;Interference control

Summary The hormone cortisol is often believed to play a pivotal role in the effects of stresson human cognition. This meta-analysis is an attempt to determine the effects of acute cortisoladministration on core executive functions. Drawing on both rodent and stress literatures, wehypothesized that acute cortisol administration would impair working memory and set-shiftingbut enhance inhibition. Additionally, because cortisol is thought to exert different nongenomic(rapid) and genomic (slow) effects, we further hypothesized that the effects of cortisol woulddiffer as a function of the delay between cortisol administration and cognitive testing. Althoughthe overall analyses were nonsignificant, after separating the rapid, nongenomic effects ofcortisol from the slower, genomic effects of cortisol, the rapid effects of cortisol enhancedresponse inhibition, g+ = 0.113, p = .016, but impaired working memory, g+ = −0.315, p = .008,although these effects reversed over time. Contrary to our hypotheses, there was no effect

of cortisol administration on set-shifting. Thus, although we did not find support for the ideathat increases in cortisol influence set-shifting, we found that acute increases in cortisol exertdifferential effects on working memory and inhibition over time.© 2015 Elsevier Ltd. All rights reserved.

∗ Corresponding author at: Department of Psychology, Universityof California Davis, One Shields Avenue, Davis, CA 95616,USA. Tel.: +1 5303026608.

E-mail address: gsshields@ucdavis.edu (G.S. Shields).

1

1

Wfc

http://dx.doi.org/10.1016/j.psyneuen.2015.04.0170306-4530/© 2015 Elsevier Ltd. All rights reserved.

. Introduction

.1. Executive function

hen studying complex cognition in humans, executiveunction—–the construct that underlies our ability to flexiblyontrol our thoughts and actions—–comes to the forefront.

9

Ei2ffaecmm(a(st

tittufltbpriFwblldt

1

Ptm(teti2i

aotalctcatail

mmtecftbbsmTt

rcfbrHfifOmtIopb2titsn

aittepcetvc

2

2

Aof

2

xecutive function is a general ability comprised of threenterrelated core processes (Miyake et al., 2000; Diamond,013). To enhance clarity, we will use the term ‘‘executiveunction’’ only to reference the general factor of executiveunction (that is, the latent ability facilitating performancecross all executive function tasks) and use the name ofach core executive function when referencing that spe-ific function. The first core executive function is workingemory, which allows people to integrate new and old infor-ation. Second, inhibition enables both cognitive inhibition

the ability to inhibit irrelevant information and selectivelyttend to goal-relevant information) and response inhibitionthe ability to inhibit a prepotent response). Third, set-hifting allows people to flexibly shift between modes ofhought.

Core executive functions are assessed using multipleasks (cf. Diamond, 2013). For example, a common work-ng memory task is the n-back, which requires participantso indicate whether a given stimulus was the same stimulushey were shown n trials previously—–thus requiring constantpdating of working memory. A common inhibition task is theanker task, which requires a participant to report the direc-ion of an arrow in the center of the screen, which is flankedy irrelevant arrows either pointing in the same direction orointing in the opposite direction as the target. The cost ineaction time when the target is flanked by arrows pointingn the opposite direction is an inverse index of inhibition.inally, a typical set-shifting task is the trail-making test,hich requires participants to draw a line connecting num-ered circles (e.g., 1-2-3-4-5-6) in one part of the task and aine connecting circles that alternate between numbers andetters (e.g., 1-A-2-B-3-C) in the second part of the task. Theifference in time taken to complete the different parts ofhe task indicates a participant’s mental flexibility.

.2. Stress, cortisol, and executive function

revious research consistently indicates that stress tendso impair performance on tasks that make use of workingemory (e.g., Schoofs et al., 2008, 2009) and set-shifting

e.g., Alexander et al., 2007; Plessow et al., 2011), althoughhe effects of stress on inhibition are less clear (Scholzt al., 2009; Schwabe et al., 2013). One promising explana-ion of stress-induced influences on core executive functionss found in the stress hormone cortisol (i.e., Butts et al.,011). Thus, a discussion of the system governing cortisol isn order.

The stress response occurs the moment the brain detects physical or perceived threat, resulting in the initiationf ‘‘allostasis,’’ or the maintenance of bodily stabilityhrough change (McEwen, 2004). Activating one pathway ofllostasis, stressors upregulate activity in the paraventricu-ar nucleus (PVN) of the hypothalamus, which then secretesorticotropin releasing hormone (CRH); CRH then acts onhe pituitary gland and promotes the release of adreno-orticotropin hormone (ACTH); ACTH in turn acts on thedrenal gland to stimulate the synthesis and release of

he cortisol (Ulrich-Lai and Herman, 2009). This is knowns the hypothalamic—pituitary—adrenal (HPA) axis, andt is primarily through this system that cortisol is regu-ated.

rdon

G.S. Shields et al.

Acute increases in glucocorticoids—–the class of hor-ones to which cortisol belongs—–function primarily toobilize the body’s resources in order to combat or evade

he stress-provoking stimulus. Receptors for glucocorticoidsxist throughout the brain and, notably, they are con-entrated within brain regions supporting core executiveunctions (Reul and de Kloet, 1985). The effects of glucocor-icoids are variable, however, as glucocorticoids can exertoth rapid-acting, nongenomic effects—–effects of cortisolrought about without modulation of gene expression—–andlow, genomic effects—–effects of cortisol brought about byodulation of gene expression—–(cf. Joëls et al., 2011).hus, cortisol can act through multiple pathways to enablehe body to combat or evade the stress-inducing stimulus.

Despite the interconnection of stress, cortisol, and neu-al activity, it is unclear whether cortisol actually influencesore executive functions. Previous research has indeedound correlations between cortisol and working memory,oth at baseline (Li et al., 2006; Franz et al., 2011) and inesponse to stress (Oei et al., 2006; Taverniers et al., 2010).owever, these data are inconsistent: some studies have

ound an inverse relationship between cortisol and work-ng memory (e.g., Oei et al., 2006), whereas others haveound a positive relationship (e.g., Stauble et al., 2013).ne potential reason for these discrepancies is that cortisolay interact with other components of the stress response

o exert effects on cognition (e.g., Schwabe et al., 2012).n contrast, the mediating effects of cortisol in the effectsf stress on inhibition are slightly more established, as onerevious study found that blocking a receptor for cortisollocked the effects of stress on inhibition (Schwabe et al.,013). Nonetheless, it is unclear from the prior study if cor-isol is both necessary and sufficient to improve inhibition orf it is simply necessary (and thus not a true cause). Finally,he impairing effects of stress on set-shifting are related toalivary cortisol (Plessow et al., 2011), but the correlationalature of these data preclude causal inferences.

Experimentally manipulating cortisol through exogenousdministration is a useful method for determining if acutencreases in cortisol actually influence core executive func-ions. Because endogenous cortisol is synthesized outsidehe brain and readily crosses the blood-brain barrier,xogenously administered cortisol should influence neuralrocesses in the same way as endogenously synthesizedortisol. Indeed, a recent meta-analysis demonstrated thatxogenously administered cortisol significantly impairs long-erm memory retrieval (Het et al., 2005), illustrating thealidity of this methodology for uncovering the effects ofortisol on cognitive processes.

. Current research

.1. Main hypotheses

number of studies have already investigated the effectsf cortisol administration on one or more core executiveunctions. Thus, our goal in this study is to aggregate these

esults using meta-analytic techniques in an attempt toetermine the true effect of acute cortisol administrationn each of the core executive functions. Given the heteroge-eous nature of the stress and working memory or inhibition

RitDetaow

snirw‘

3Of(sttsoatcte

3Ot

3

Tasks that make use of executive function were coded as oneof the three core executive functions based upon previousempirical or theoretical literature suggesting that a given

2 Inclusion criterion (1) was chosen to avoid confounding acuteincreases in cortisol and likely interactions with extant glucocorti-coid receptor resistance. We chose criterion (2) because we wereinterested in the effects of exogenous cortisol administration, notendogenous changes due to other factors. We chose criterion (3)because this allows direct comparison of cortisol administrationalone to a control condition. We chose criterion (4) because men-tal and physical illness has been linked to glucocorticoid receptordysregulation. We chose inclusion criterion (5) because executivefunction in animals, which is an extensively trained ability, may

Cortisol and core executive functions

literatures as well as the nascent stress and set-shiftingor inhibition literatures, one could argue that no stronglysupported a priori hypotheses could be made regarding theeffects of stress on core executive functions. Nonetheless,drawing on both rodent and stress literature (Schoofs et al.,2009; Butts et al., 2011; Plessow et al., 2011; Schwabe et al.,2013), we hypothesized that acute cortisol administrationwould impair working memory and set-shifting but enhanceinhibition.

2.2. Proposed moderators

Whether or not cortisol evidences an effect on coreexecutive functions may depend on moderating factors.Importantly, the timing of cortisol administration relative tocognitive testing should be considered. As previously men-tioned, cortisol is thought to have differential fast and sloweffects; as such, analyses exploring the effects of cortisolshould consider the delay between cortisol administrationand cognitive testing. We also considered a number of addi-tional moderators in this meta-analysis including cortisoldose, participant age and sex, mode of administration,whether the task included an affective component,1 time ofday for cognitive testing, outcome type (time or accuracy-based), and study design.

We only included studies that employ a methodology ofacute cortisol administration to participants. This is in partbecause effects of acute increases in cortisol likely differqualitatively from effects of chronic elevations of cortisol(e.g., McEwen, 2004). Consequently, any inferences madefrom the findings reported here should be constrained toacute increases in cortisol, not chronically high levels ofcortisol.

3. Method

3.1. Study selection and inclusion criteria

3.1.1. Literature reviewOur search terms for obtaining studies relevant to thismeta-analysis consisted of ‘‘hydrocortisone’’, ‘‘cortisoltreatment’’, ‘‘cortisol administration’’, or ‘‘exogenouscortisol’’ and ‘‘executive function’’, ‘‘working mem-ory’’, ‘‘updating’’, ‘‘response inhibition’’, ‘‘set-shifting’’,‘‘attention’’, ‘‘cognitive’’, or ‘‘cognition.’’ We queried andperformed an exhaustive search of the following databases:PsycINFO, PsycArticles, PubMed, Proquest Theses and Dis-sertations: Social Sciences, Web of Science, University ofAlabama Theses and Dissertations, and U.C. Davis Thesesand Dissertations. PubMed returned 1,684 results. ProQuestwas used to index PsycINFO, PsycArticles, ProQuest Thesesand Dissertations, and U.C. Davis Theses and Disserta-

tions, which collectively returned 1,643 results. Similarly,Web of Science returned 218 results, while the Universityof Alabama Theses and Dissertations returned 64 results.

1 There was an insufficient number of studies assessing detec-tion/activation but not suppression of emotional material todetermine differences of cortisol administration on detection vs.suppression of emotional material.

bw

2

cpw

r

93

eferences from relevant articles were reviewed, and stud-es that were potentially relevant were examined fromhose references. For all articles considered, we followedickerson and Kemeny (2004) in reviewing abstracts andxamining full texts whenever an article had the potentialo include a relevant effect (that is, whenever a study withinn article administered cortisol to participants, the full-textf the article was reviewed to ensure that a relevant effectas not omitted from the abstract and thus this analysis).

We also posted announcements to a number of list-ervs and reviewed conference proceedings for an extensiveumber of conventions. Corresponding authors of all stud-es with unpublished relevant results were emailed with aequest for unpublished statistics. Authors who respondedith these unpublished statistics are listed in Section

‘Acknowledgments’’.

.1.2. Inclusion criteriaur seven inclusion criteria for this study were as

ollows: studies had to (1) acutely, (2) administer,3) hydrocortisone—–without coadministration of anotherubstance—–and/or a placebo to, (4) healthy, (5) human par-icipants, who then completed, (6) a task utilizing one ofhe core executive functions. (7) If a study used a within-ubjects, crossover design, performance on the task utilizingne of the core executive functions had to be separated byt least one day (to reduce confounding the effects of cor-isol on learning and practice from the effects of cortisol onore executive functions). We chose these inclusion criteriao best isolate the relationship between cortisol and corexecutive functions.2

.1.3. Selected studiesur search and study inclusion criteria led to the incorpora-ion of 30 studies,3,4 27 of which were published.5

.2. Coding of covariates and moderators

e influenced in different ways than executive function in humans,hich is a natural ability.3 One study was reported in two separate papers (Henckens et al.,011, 2012) and was treated as one study for this analysis.4 One study (Newcomer et al., 1999) explored the effects ofhronic cortisol administration. However, we used only the dataresented for the effects of cortisol on day 1, presented in a tableithin the paper.5 A study was considered ‘‘unpublished’’ if the effect was not

eported in a published paper.

9

tbScptoi�oe

tteowpaeaidnctefgim

3

SmHaiespwuIrttdaM

vo

dtwut

mgtmotwac(new(csbtSibthi

trioaddar

For all of the following analyses, a positive effect sizeindicates that cortisol improved performance on the out-come relative to placebo, whereas a negative performance

-measures studies in which we had not obtained the correlationcoefficients. The estimated correlation between all tasks utilizing acore executive function under cortisol with all tasks utilizing a coreexecutive function under placebo was .69, 95% CI [.50, .81]. Weconducted sensitivity analyses by using the upper and lower boundsof the 95% confidence interval as the correlation between repeatedmeasures for which we had not obtained the actual correlation.None of the results were significantly from those using the point

4

ask loaded on one of the factors (working memory, inhi-ition, or set-shifting) primarily. See Table S3 in the onlineupplementary Material for a complete description of taskoding. Tasks were considered including an affective com-onent if the task employed affective characteristics or ifhe task incorporated faces as stimuli. The covariate of dosef hydrocortisone required some conversion between stud-es, as studies using an intravenous drip provided doses ing/kg rather than the more common single dosages given inral form. We thus converted doses given in �g/kg to thequivalent mg dose for typical individuals.

To test for differences among moderators, most poten-ial moderators were dummy-coded, including whetherhe task included an affective component, whether theffect size was derived from repeated measures, modef cortisol administration, and whether the outcomeas a reaction-time or performance-based measure. Oneotential moderator was contrast-coded—–time of cortisoldministration—–because none of the three groups (morning,arly afternoon, and late afternoon) could suitably serve as

reference for the others. Finally, average age of partic-pants, percentage of males, dose of hydrocortisone, andelay in minutes from hydrocortisone administration to cog-itive testing6 were treated as continuous variables andentered their respective lowest obtained values, makinghe intercept (which is the effect size estimate at the low-st value of the covariate) interpretable while controllingor the covariate. If the average participant age was notiven in the article, the median participant age was used ift was reported; if neither of these statistics were listed, theidpoint of the reported participant age range was used.

.3. Analytic strategy

tandardized mean differences were used as the effect sizeeasure of interest. For effect size estimates, we usededges’ g rather than Cohen’s d, given that the former is

relatively unbiased estimate of the population standard-zed mean difference effect size while the latter is a biasedstimate. To obtain Hedges’ g, we calculated the effectize from the means, standard deviations, and sample sizesresented in the article. If means and standard deviationsere not reported and the design was between-studies, wesed t or one-way F statistics to calculate the effect size.f none of these statistics were reported, we emailed cor-esponding authors for these statistics. If we were unableo contact the corresponding authors for necessary statis-ics for a study, that study was excluded. For within-studies

esigns, we converted effect size estimates and their vari-nces into the between-study effect size metric followingorris and DeShon (2002).7

6 Adding 15 min (or similar values) to studies employing intra-enous rather than oral cortisol administration did not alter anyf the findings presented here.7 We emailed the authors of studies using a repeated-measuresesign with a request for the correlation of the outcomes under cor-isol with the outcomes under placebo. Multiple authors respondedith these statistics, but not all. We used RVE to estimate the pop-lation correlation from the coefficients emailed to us and usedhe estimate as the correlation between measures for the repeated

e

esmtwswt

ecfv

G.S. Shields et al.

Given the multifaceted nature of executive function,ost studies often report more than one outcome for any

iven task that makes use of a core executive function.8 Mul-iple outcomes are a problem for conventional meta-analyticethods, as averaging effect sizes within studies with-

ut accounting for their correlations can alter or obscurerue effect size estimates (Scammacca et al., 2014). Thus,e employed the meta-analytic technique of robust vari-nce estimation, a random-effects meta-regression thatan account for dependence between effect size estimatesHedges et al., 2010; Tanner-Smith et al., 2013). This tech-ique robustly estimates effect size weights and standardrrors for the given effects, allowing for multiple outcomesithin studies (Hedges et al., 2010). We employed the robu

) function of the robumeta package in R, version 3.1.1, toonduct these analyses, using the small sample correctionsuggested by Tipton (2014) and the correlated weights giveny Hedges et al. (2010). To account for dependency betweenhe effect sizes, � was set to the recommended .80 (Tanner-mith and Tipton, 2014).9 Because we were more interestedn understanding factors that influenced the relationshipetween cortisol administration and core executive func-ions than we were in understanding factors contributing toeterogeneity, covariate analyses do not separate covariatesnto within- and between-study covariates.

Degrees of freedom for all analyses were estimated usinghe Satterwaite approximation, where df = 2/cv2 and cvepresents the coefficient of variation, as simulation stud-es have indicated that this method of estimating degreesf freedom is most analytically valid with the RVE meta-nalytic technique (Tipton, 2014). Because of how theegrees of freedom are estimated, if the degrees of free-om are less than four, then there is a heightened risk of

Type I error and the analysis results cannot be trusted toepresent population values (Tipton, 2014).

stimate.8 Some might suggest that instead of incorporating all reportedffect sizes we should select only the critical effect size from eachtudy. However, this approach is only valid if a strong case can beade for doing so (Scammacca et al., 2014). In our meta-analysis

his case cannot be made; many of these studies found effectsithin different subgroups, and many of these effect sizes reflect a

tudy’s use of multiple measures assessing core executive functionsith multiple outcomes. As such, a justification cannot be given for

he use of a single effect size from a given study.9 Because of how the RVE meta-analytic method estimatesffect size weights, the designation of � at .80 is largely non-onsequential; indeed, sensitivity analyses with values of � rangingrom 0 to 1.0 evidenced no change greater than 0.0001 in absolutealue in estimates.

95

Figure 1 Forest plot of working memory study-average effectsizes by weight. The grand effect was nonsignificant, g+ = −0.04,p = .50. Numbers on the Y-axis correspond to the studies listedas follows: 1Breitberg et al. (2013); 2Entringer et al. (2009);3Henckens et al. (2011, 2012); 4Kuhlmann and Wolf (2005);5Kuhlmann et al. (2005); 6Kumsta et al. (2010); 7Lupien et al.(1999); 8Monk and Nelson (2002); 9Oei et al. (2009); 10Porteret al. (2002); 11Symonds et al. (2012); 12Terfehr et al. (2011);13 14 15

1

(

g−aidili(cb7wei

i

Cortisol and core executive functions

indicates that cortisol impaired performance on the out-come relative to placebo. In addition, because the outcomein these analyses is the standardized mean differencebetween groups (the effect size), a significant covariatemeans that the effect size estimate depends upon levels ofthat covariate. In other words, if a covariate is significant,it means that as the covariate increases or decreases, themean difference between the effect of cortisol compared toplacebo on core executive functions increases or decreases.

4. Results

4.1. Preliminary analyses

4.1.1. Study characteristicsThe final sample consisted of 30 studies, each of which isrepresented by m. Twenty-seven of these studies were pub-lished. There were 260 total effect sizes,10 each of whichis represented by k. Of these studies, 18 assessed workingmemory, with k = 119. Similarly, 23 of these studies assessedinhibition, with k = 118. Finally, 6 of these studies assessedset-shifting, with k = 23.

4.1.2. Assessment of publication biasTo assess publication bias, we first compared effect sizesfrom unpublished work (m = 3, k = 11) to those of publishedwork (m = 27, k = 249). This analysis indicated that unpub-lished studies were of the same magnitude as publishedstudies, t(2.5) = −1.49, p = .25. Second we conducted Egger’stest (Egger et al., 1997) for funnel plot asymmetry on boththe overall study set and each core executive function indi-vidually. Egger’s test returned nonsignificant results whenlooking at the overall study set, t(28) = −0.17, p = .87, work-ing memory, t(16) = 0.04, p = .97, inhibition, t(21) = −0.45,p = .66, and set-shifting, t(4) = 0.61, p = .57, indicating a lackof bias. These results therefore indicate that any effectsobserved in this meta-analysis are unlikely to be due topublication bias.

4.2. Primary analyses

4.2.1. Working memoryThe overall effect of cortisol administration on tasksprimarily utilizing working memory (m = 18, k = 119) wasnonsignificant, g+ = −0.043, t(15.4) = −0.69, p = .50, 95% CI[−0.175, 0.090] (Fig. 1). There was substantial hetero-geneity within studies, I2 = 67.91, albeit with relatively lowbetween-study heterogeneity, �2 = 0.08. This heterogeneityillustrates that the effects exhibited within-study discrepan-cies that did not translate into between-study differences,as the effect was relatively homogenous between studies.

Despite a nonsignificant overall effect, however, afterseparating the genomic effects of cortisol from its nonge-

nomic effects by controlling for the delay between cortisoladministration and cognitive testing, the non-genomiceffects of cortisol significantly impaired working memory,

10 Although this may seem like a large number of effects per study,the number of effect sizes per study we obtained is relatively com-mon in social science research (Scammacca et al., 2014).

itcBhiat

Tollenaar et al. (2009); Tops et al. (2006); Vaz et al. (2011);6Wingenfeld et al. (2011); 17Wolf et al. (2001); 18Yehuda et al.2007).

+ = −0.315, t(3.8) = −5.15, p = .008, 95% CI [−0.488,0.142]. Although the degrees of freedom in the previousnalysis do not quite reach 4.0, the likelihood of mak-ng a Type I error only increases at most twofold withf < 4 (Tipton, 2014), and given that p < .025 (half of .05)n the previous analysis, inferences are applicable. Simi-arly, a longer delay between cortisol and cognitive testingmproved working memory, B = 0.005, t(1.8) = 6.07, p = .03cf. Fig. 2); thus, the unstandardized slope estimate indi-ates that 74 min after cortisol administration, cortisolegins to improve working memory, whereas from 15 to3 min following cortisol administration, cortisol impairsorking memory. In the previous significance test, how-ver, the restricted degrees of freedom preclude definitivenference—–given an increased likelihood of a Type I error.

Notably, however, the effects of cortisol dose did notnfluence the effect size of cortisol administration on work-ng memory, B = −.001, t(2.1) = −0.78, p = .51. Similarly,here was no evidence for a quadratic relationship betweenortisol dose and the effects of cortisol on working memory,

< −.001, t(9.3) = −0.15, p = .88. This lack of association

eld even when controlling for the time of cortisol admin-stration. Moderator analyses determined that cortisoldministration impaired accuracy relative to reactionime outcomes, t(9.9) = 2.22, p = .051, as accuracy was

96 G.S. Shields et al.

Table 1 Covariate effects on the relation between cortisol and core executive functions.

Variable B ˇ g+ (SE) controlling for covariate t df p

Working memoryPercent male participants .001 .02 0.37 11.4 .72Range: 0—100 −.07 (.06) −1.24 6.6 .24

Delay between cortisol and test .005 .12 6.07 1.8 .03Range: 15—240 −.32 (.06) −5.15 3.83 .008

Quadratic delay <.001 −.05 0.76 2.1 .52−.24 (.13) −1.86 2.6 .17

Cortisol dose −.001 −.03 −0.78 2.1 .51Range: 3.567—120 −.01 (.08) −0.17 10.8 .87

Quadratic cortisol dose <−.001 −.04 −0.15 9.3 .88−.03 (.14) −0.21 7.2 .84

Participant age −.006 −.05 −2.2 2.6 .13Range: 20.4—75.5 .05 (.09) 0.57 12.6 .58

InhibitionPercent male participants <.001 <.01 0.14 11.1 .89Range: 0—100 .05 (.05) 0.86 4.9 .43

Delay between cortisol and test −.001 −.12 −4.49 1.5 .08Range: 15—540 .11 (.04) 2.92 9.5 .02

Quadratic delay <.001 .10 0.73 3.6 .51.14 (.07) 2.10 8.5 .07

Cortisol dose −.003 −.08 −1.89 7.0 .10Range: 10—100 .11 (.04) 2.85 7.6 .02

Quadratic cortisol dose <−.001 −.10 −0.26 3.5 .81.10 (.05) 2.12 5.1 .09

Participant age .002 .02 0.97 3.4 .40Range: 20.1—75.5 .03 (.05) 0.51 16.1 .62

Set-shiftingPercent male participants .003 .13 0.95 2.4 .43Range: 0—100 −.20 (.27) −0.75 1.4 .56

Delay between cortisol and test <.001 .06 0.56 1.2 .66Range: 60—540 −.04 (.11) −0.36 3.4 .74

Quadratic delay <.001 NAa 1.62 2.2 .24.14 (.10) 1.37 1.0 .40

Cortisol dose −.004 −.13 −2.54 1.8 .14Range: 13.33—120 .11 (.06) 1.87 2.4 .18

Quadratic cortisol dose <−.001 −.25 −0.40 2.1 .73.06 (.14) 0.45 2.5 .69

Participant age .001 .01 0.19 2.7 .86Range: 22.2—75.5 −.03 (.19) −0.13 3.0 .90

Notes: B: unstandardized slope; ˇ: standardized slope; g+: effect size; SE: standard error of the effect size; t: t test statistic for testdetermining whether the effect size differs from zero; df: degrees of freedom for t test; p: p value for t test. Marginal or significant pvalues are in boldface. If df < 4, there is up to a 10% risk of Type I error, given how df are estimated. Linear associations are reported

sgtt

without controlling for quadratic effects.a Entails that the model was unsolvable.

ignificantly impaired after cortisol administration,+ = −0.12, t(12.3) = −2.91, p = .01, while reactionimes were not significantly affected, g+ = 0.13,(7.5) = 0.56, p = .60. However, even when only considering

aiBe

ccuracy-based outcomes, the dose of cortisol did notnfluence the effect size of cortisol on working memory,

= .001, p = .44, nor was there any evidence of a quadraticffect, B < .001, p = .58, indicating that cortisol lacked

Cortisol and core executive functions 97

Table 2 Moderator analyses of the effects of cortisol on executive function.

Variable g+ SE df p m k

Working memoryEmotive task

Nonemotive −.03 .05 15.0 .57 18 107Emotive .11 .56 2.0 .86 3 12

Reaction time versus accuracya

Reaction time .13 .23 7.5 .60 9 74Accuracy −.12* .04 12.3 .01 15 45

Study designRepeated measures −.05 .03 3.9 .15 12 38Between groups .11 .24 5.0 .67 6 81

Mode of administrationIntravenous/injection −.15 .09 3.1 .19 5 68Oral −.01 .08 11.1 .86 13 51

Time of treatmentMorning −.12 .11 3.4 .33 5 53Mid-afternoon .09 .17 4.9 .60 6 38Late afternoon −.06 .09 5.3 .52 7 28

InhibitionEmotive task

Nonemotive .01 .05 10.5 .89 14 72Emotive .12† .05 6.0 .06 11 46

Reaction time versus accuracyReaction time .07 .04 6.6 .13 15 58Accuracy −.03 .05 11.5 .58 15 60

Study designRepeated measures .07† .04 6.9 .07 14 73Between groups −.05 .10 7.5 .64 9 45

Mode of administrationIntravenous/injection −.02 .06 2.4 .82 4 42Oral .06 .04 11.0 .13 19 76

Time of treatmentMorning .03 .03 1.2 .42 4 16Mid-afternoon −.05 .10 5.2 .63 7 41Late afternoon .08 .05 7.7 .13 12 61

Set-shiftingEmotive task

Nonemotive −.02 .09 3.1 .86 5 7Emotive <.01 .07 1.0 .96 1 16

Reaction time versus accuracyReaction time .04 .18 1.9 .83 3 10Accuracy −.08 .11 3.6 .49 5 13

Study designRepeated measures −.08 .11 2.8 .54 4 20Between groups .22 .06 3.9 .16 2 3

Mode of administrationIntravenous/injection <.01 <.01 1.0 .56 2 17Oral −.02 .15 2.5 .91 4 6

Time of treatmentMorning .04 .04 1.0 .51 2 2Mid-afternoon <.01 .07 1.0 .96 1 16Late afternoon −.04 .24 1.9 .88 3 5

Notes: g+: effect size of the respective group; SE: standard error of the effect size; t: t test statistic for test determining whether theeffect size differs from zero; df: degrees of freedom for t test; p: p value for t test; m: number of studies in the analysis, k: number ofeffect sizes in the analysis. If df < 4, there is up to a 10% risk of Type I error, given how df are estimated.† p < .10.* p < .05.a Indicates that the two groups differ at p = .051.

98 G.S. Shields et al.

Figure 2 Association between the study-average, non-centered delay of cortisol administration and the study-averageeffect size for working memory. Size of point represents theeffect size weight. Cortisol administration impaired workingmd

awrm

4Tig(Iis

sfispittsvieae1ieitTtnaloa(

Figure 3 Forest plot of inhibition study-average effect sizesby weight. The grand effect was nonsignificant, g+ = 0.05,p = .14. Numbers on the Y-axis correspond to the studies listedas follows: 1Abercrombie et al. (2003); 2Bertsch et al. (2011);3Breitberg et al. (2013); 4Buss et al. (2004); 5Carvalho Fer-nando et al. (2013); 6Henckens et al. (2011); 7Hsu et al. (2003);8Kuhlmann and Wolf (2005); 9Kuhlmann et al. (2005); 10Monkand Nelson (2002); 11Newcomer et al. (1999); 12Porter et al.(2002); 13Putman and Berling (2011); 14Putman et al. (2007);15Putman et al. (2010); 16Schlosser et al. (2013); 17Taylor et al.(2011); 18Tollenaar et al. (2009); 19Tops et al. (2006); 20Vasaet al. (2009); 21Vaz et al. (2011); 22Wolf et al. (2001); 23Yehudaet al. (2007).

Figure 4 Association between the study-average, non-centered delay of cortisol administration and the study-averageeffect size for inhibition. Size of point represents the effectsize weight. Cortisol administration enhanced inhibition at shortdelays but impaired it over long delays. Removing the outlying

emory at short delays but enhanced working memory over longelays.

dose—response effect for influencing accuracy-basedorking memory performance. Further analyses did not

eveal any additional significant covariates (Table 1) oroderators (Table 2) with df > 4.

.2.2. Inhibitionhe effect of cortisol administration on tasks primar-

ly utilizing inhibition (m = 23, k = 118) was nonsignificant,+ = 0.055, t(12.9) = 1.6, p = .138, 95% CI [−0.020, 0.130]Fig. 3). There was low heterogeneity within the effects,2 = 30.80, and low between-study heterogeneity, �2 = 0.01,ndicating that effect sizes were fairly consistent betweentudies.

Despite a nonsignificant overall effect, however, aftereparating genomic from nongenomic effects by controllingor the delay between administration and inhibition test-ng, the fast, nongenomic effects of cortisol administrationignificantly improved inhibition, g+ = 0.113, t(9.5) = 2.92,

= .016, 95% CI [0.026, 0.201]. Conversely, a longer delayn minutes between cortisol administration and inhibi-ion testing marginally impaired inhibition, B = −0.001,(1.5) = −4.49, p = .080 (cf. Fig. 4). The unstandardizedlope estimate is centered at 15 min (the lowest obtainedalue for the delay between cortisol administration andnhibition testing), and it represents the change in theffect size for each minute increase between cortisoldministration and inhibition testing. As such, this slopestimate indicates that cortisol improves inhibition from5 to 135 min post-administration, but cortisol begins tompair inhibition after 136 min post-administration. How-ver, the restricted degrees of freedom preclude definitivenferences regarding the effect of the delay between cor-isol and cognitive testing—–given an increased likelihood ofype I error. Notably, however, cortisol dose did not influencehe effect size estimate, B = −.003, t(7.0) = −1.89, p = .10,or was there a quadratic relation between cortisol dosend the effect size, B < −.001, t(3.5) = −0.26, p = .81. This

ack of association held even when controlling for the timef cortisol administration. Further analyses did not revealny additional significant covariates (Table 1) or moderatorsTable 2) with df > 4.

study did not influence the results; this outlying study thus pro-vides a nice illustration of the accuracy of the slope estimate.

Cortisol and core executive functions

Figure 5 Forest plot of set-shifting study-average effect sizesby weight. The grand effect was nonsignificant, g+ = −0.01,p = .91. Numbers on the Y-axis correspond to the studies listedas follows: 1Breitberg et al. (2013); 2Newcomer et al. (1999);

vscce

etesTtmiortfmtmdi

5

Dnttiiadiditfstitftiartiittimer

3Porter et al. (2002); 4Vaz et al. (2011); 5Wingenfeld et al.(2011); 6Yehuda et al. (2007).

4.2.3. Set-shiftingAnalyzing only the effect sizes related to set-shifting(m = 6, k = 23) revealed a negligible effect, g+ = −0.011,t(4.2) = −0.12, p = .91, 95% CI [−0.266, 0.244] (Fig. 5) withlow heterogeneity between effects, I2 = 31.53, �2 = 0.03,indicating that the effects of cortisol administration was rel-atively consistent both within and between studies. Therewere no significant covariates (Table 1) or moderators(Table 2) of the effect size.

4.3. Supplementary analyses

Analyses exploring the effects of cortisol administration overall tasks utilizing one of the core executive functions arepresented within the online Supplementary Material.

5. Discussion

This is the first comprehensive meta-analysis of the effectsof cortisol administration on core executive functions. Ashypothesized, we found that acute increases in cortisolimpaired working memory but enhanced inhibition, andthese effects reversed over time. Contrary to our hypothe-sis, though, there was no effect of cortisol administrationon set-shifting. There was little heterogeneity between

effect sizes in analyses of inhibition, and although therewas moderate heterogeneity between effect sizes in anal-yses of working memory and set-shifting, most of thisheterogeneity was due to within-study (not between-study)

m

ts

99

ariability—–likely reflecting the variety of tasks used withintudies—–illustrating that the observed effects were largelyonsistent across studies. Thus, we observed remarkableonsistency in the effects of cortisol administration on corexecutive functions across various studies.

We found that the rapid, nongenomic effects of cortisolnhanced inhibition, but the slow, genomic effects of cor-isol impaired inhibition. Similarly, the rapid, nongenomicffects of cortisol impaired working memory, whereas thelow, genomic effects of cortisol enhanced working memory.hese effects coincide with stress research, which has linkedhe rapid effects of acute increases in cortisol to enhance-ents in inhibition (Schwabe et al., 2013) and impairments

n working memory (Schoofs et al., 2008). Additionally, thebservation that the effect of cortisol on these processeseverses over time may also help to explain some inconsis-encies in stress literature, given that some studies haveound stress-induced cortisol predicts enhanced workingemory (Stauble et al., 2013) or that stress impairs inhibi-

ion (Scholz et al., 2009). Thus, the overlap of our workingemory and inhibition results with stress literature vali-ates the assumption in much of this literature that cortisols causally related to the effects observed in those studies.

.1. Discussion of moderators

espite the overlap with stress research, there was no sig-ificant effect of cortisol dose. Although the reasons forhis lack of dose-dependent effect are unclear, one poten-ial explanation is that the lack of dose-dependent effectss due to a lack of dose—response studies (with only fivencluded in the above analyses); however, a meta-analysisllows comparisons of effects across studies with differentoses of cortisol. Thus, the lack of a dose—response effects likely as reliable as other potential moderators, such aselay between cortisol administration and cognitive test-ng. An alternative explanation is that the dosages given inhis study do not approximate endogenous cortisol increasesollowing exposure to a stressor. This explanation may haveome validity. Although a dose of 10 mg produces an increasehat approximates a mild stressor (Taylor et al., 2011)—–llustrating that the lower range of cortisol doses obtained inhis meta-analysis index relatively typical endogenous levelsollowing mild stress—–the upper range of dosages given inhis meta-analysis are much larger than dosages that approx-mate moderate to severe stress (i.e., 30—40 mg). Indeed,fter a post hoc restriction of cortisol dose to the upperange approximating a moderate to severe stressor (lesshan 40 mg), marginally significant linear or quadratic (i.e.,nverted U shaped) effects emerged for inhibition and work-ng memory, respectively (data not shown). Thus, althoughhis distinction was post hoc, it may be the case that cor-isol exerts dose-dependent effects to the extent that thencrease is relatively typical of an endogenous increase, and,oreover, these dose-dependent effects are eliminated with

normous cortisol increases, potentially due to aberrantecruitment of additional processes to dampen the enor-

ous cortisol response.Other proposed moderators that were not related to

he effects of cortisol on core executive functions thistudy included age, sex, or time of day during cortisol

1

ahoswaFdmw

5

Oelit(tilwccptiwo(wtsA(otimoomiett2mtoeo

ccon2ai

Sepvhtr1ave

cprCumtscwaagde

5

Thoetcbwcptefdtt3Hrthp2licp

00

dministration. This lack of association was surprising, givenow many of these variables influence the effects of cortisoln cognition (e.g., Het et al., 2005). It is tempting to offerpeculative explanations for this (e.g., age may interactith health status to alter the association between cortisolnd cognition), but these explanations go beyond our data.uture research could address this finding by attempting toetermine additional moderators influencing another givenoderator’s effects on the association between cortisol andorking memory or inhibition.

.2. Caveats

ne important caveat in interpreting these results is that theffects of cortisol in this meta-analysis are considered in iso-ation, whereas in the typical stress response cortisol exertsts effects in tandem with alterations in catecholamines,he sympathetic nervous system, and the immune systemAllen et al., 2014). Thus, it may be the case that cor-isol interacts with other biological processes to furthernfluence core executive functions. Indeed, there is pre-iminary support for the idea that cortisol may interactith stress-induced increases in noradrenergic (one type ofatecholamine) activity to exert its effects on cognitive pro-esses (Roozendaal et al., 2006; cf. Lupien et al., 2007),erhaps due to effects of noradrenergic activity on atten-ion (cf. Robbins and Arnsten, 2009). For example, studiesnvestigating the conjoint effects of cortisol administrationith a noradrenergic agonist have found synergistic effectsf administration of these substances on decision-makingSchwabe et al., 2012), although similar interactive effectsere not observed for memory encoding or, notably, inhibi-

ion (Vasa et al., 2009; van Stegeren et al., 2010). Similarly,tress also increases dopaminergic activity (Robbins andrnsten, 2009), and dopamine both interacts with cortisolMizoguchi et al., 2004) and follows an inverted U to enhancer impair working memory (Robbins and Arnsten, 2009). Putogether, these studies provide preliminary support for thedea that the effects of cortisol on core executive functionsay be more complex in vivo than is captured in studies

f cortisol administration, and therefore the true effectsf cortisol on core executive functions may be stronger inagnitude than effects of cortisol administration given in

solation. Thus, it is possible that cortisol may indeed influ-nce set-shifting when considered within the context of theypical stress response. Nonetheless, it is important to notehat many studies (i.e., Oei et al., 2006; Plessow et al.,011) have found associations between cortisol and perfor-ance on tasks that make use of core executive functions in

he absence of noradrenergic activity. Moreover, the effectsbtained in our meta-analysis show that cortisol can exertffects on working memory and inhibition independent ofther interactive effects.

It is also important to note that the effects of cortisolonsidered here are specific to acute increases. Chroni-ally elevated levels of cortisol contribute to dysregulationf the immune system, multiple hormonal systems, and

eural function (McEwen, 2004; Silverman and Sternberg,012). This multiple-system dysregulation is thought to bringbout structural alterations in neural systems contribut-ng to cognitive impairments (McEwen, 2004; Silverman and

tmsf

G.S. Shields et al.

ternberg, 2012). Thus, it is quite likely that chronicallylevated cortisol may influence core executive functions,resumably through different pathways than acute ele-ations in cortisol. Currently, only two studies exist thatave considered the effects of chronic cortisol administra-ion in healthy individuals, and they have presented mixedesults regarding effects on working memory (Young et al.,999; Newcomer et al., 1999). Thus, future research shouldttempt to examine whether sustained stress-related ele-ations in cortisol might contribute to impairments in corexecutive functions.

Finally, altered performance on tasks that make use ofore executive functions does not necessarily imply thaterformance on tasks that make use of global executiveesources, such as planning, playing chess, dieting, or goinghristmas shopping will be altered. It is interesting to spec-late how the differential influences of cortisol on workingemory and inhibition may contribute to performance in

asks common to everyday life. For example, our data coulduggest that cortisol may produce a more phenotype of moreontrol (enhanced inhibition) but less capacity (impairedorking memory), resulting in perhaps a more cautiousction orientation, which may increase success in things suchs dieting but decrease success in things such as winning aame of chess. However, these speculations go beyond ourata. Future research could thus extend these findings byxamining effects of cortisol on tasks common to daily life.

.3. Limitations

his meta-analysis has limitations. First, the substantialeterogeneity of within-study effect sizes in working mem-ry and set-shifting analyses coupled with the paucity ofvidence for significant covariates or effect size modera-ors indicates that important covariates may have not beenonsidered. Although both the small heterogeneity in inhi-ition analyses and small between-study heterogeneity inorking memory and set-shifting analyses alleviates the con-ern of this limitation inasmuch as the inferences made areertinent to this meta-analysis, this limitation highlightshe need to understand additional factors that may influ-nce the relationships of cortisol and the core executiveunctions of working memory and set-shifting. Second, theiversity of tasks that make use of core executive functionshroughout the various studies may have obscured poten-ial effects on a specific task. Third, the study set size of0 might be thought to lack power to detect an effect.owever, the low between-study heterogeneity illustratesemarkable consistency in the observed effects (indicatinghat more studies would not produce an effect unobservedere), and a study set size of 30 is relatively large in com-arison to other similar meta-analyses (e.g., Het et al.,005; Steptoe et al., 2007). As such, while 30 studies isess than ideal, it represents a sizeable number of stud-es that can be used to make inferences about effects ofortisol administration. Fourth, the neural substrates under-inning core executive functions may be more sensitive to

he effects of cortisol administration than are tasks thatake use of core executive functions, and these neural

ubstrates may evidence an effect that is hidden by per-ormance on broad cognitive tasks—–given that the neural

tmU

mt4

Sco2

[‘t

fAd2

Weo8

GfU

Gts

Wop4

rc

ooL

Eor4

mah

te

Cortisol and core executive functions

substrates are closer to the source of potential modula-tion. Fifth, this meta-analysis could not assess potentialdifferences in detection versus suppression of emotionalmaterial following cortisol administration due to an insuf-ficient number of studies assessing detection of emotionalinformation without also assessing suppression of that infor-mation; thus, this meta-analysis cannot determine whethercortisol differentially modulates detection versus suppres-sion of emotional information. Sixth, some might suggestthat different core executive functions should have beenchosen for analysis, since it is possible a different analyticstrategy would have returned different results. However,the core executive functions chosen for analysis are eachand every core executive function discussed in a majorrecent review of executive function (Diamond, 2013; seealso Miyake et al., 2000); as such, while a different strategymay have produced different results, the strategy chosen foranalysis is the most theoretically supported strategy. Finally,the effects examined in this study were restricted to acuteadministrations of cortisol given in isolation. As discussedpreviously, cortisol may interact with other components ofthe stress response to modulate cognitive function, which isnot something this meta-analysis can address.

6. Conclusion

We found that acute increases in cortisol impair workingmemory and enhance inhibition, and these effects reverseover time. Contrary to our hypothesis, however, there wasno effect of cortisol administration on set-shifting. Thesefindings inform future research by suggesting a timescale inwhich experiments should assess working memory and inhi-bition following stress, depending upon the effect of cortisoldesired.

Role of the funding

The funding source was in no way a part of this research.

Conflict of interest statement

The authors declare no conflict of interest in this work.

Acknowledgments

This research was supported by a U.C. Davis Provost’sFellowship to Grant S. Shields and a Hellman FoundationFellowship to Wesley G. Moons.

The authors wish to thank Heather Abercrombie, Clau-dia Buss, Mark Ellenbogen, Dirk Hellhammer, Robert Kumsta,Sonia Lupien, Christopher Monk, Peter Putman, CatherineSymonds, Veronique Taylor, Marieke Tollenaar, Oliver Wolf,and Stefan Wüst for providing us with requested unpublishedor additional statistics from their work.

Studies included in the meta-analysis

Abercrombie, H.C., Kalin, N.H., Thurow, M.E., Rosenkranz,M.A., Davidson, R.J., 2003. [Unpublished data referenced in

c

Cm

101

he published paper, ‘‘Cortisol variation in humans affectsemory for emotionally laden and neutral information’’].npublished raw data.

Bertsch, K., Böhnke, R., Kruk, M.R., Richter, S., Nau-ann, E., 2011. Exogenous cortisol facilitates responses

o social threat under high provocation. Horm. Behav. 5,28—434.

Breitberg, A., Drevets, W.C., Wood, S.E., Mah, L.,chulkin, J., Sahakian, B.J., Erickson, K., 2013. Hydro-ortisone infusion exerts dose- and sex-dependent effectsn attention to emotional stimuli. Brain Cogn. 81,47—255.

Buss, C., Wolf, O.T., Witt, J., Hellhammer, D.H., 2004.Unpublished data referenced in the published paper,‘Autobiographic memory impairment following acute cor-isol administration’’]. Unpublished raw data.

Carvalho Fernando, S., Beblo, T., Schlosser, N., Ter-ehr, K., Wolf, O.T., Otte, C., . . . Wingenfeld, K., 2013.cute glucocorticoid effects on response inhibition in bor-erline personality disorder. Psychoneuroendocrinology 38,780—2788.

Entringer, S., Buss, C., Kumsta, R., Hellhammer, D.H.,adhwa, P.D., Wüst, S., 2009. Prenatal psychosocial stressxposure is associated with subsequent working mem-ry performance in young women. Behav. Neurosci. 123,86—893.

Henckens, M.J., van Wingen, G.A., Joëls, M., Fernández,., 2011. Time-dependent corticosteroid modulation of pre-

rontal working memory processing. Proc. Natl. Acad. Sci..S.A. 108, 5801—5806.

Henckens, M.J., van Wingen, G.A., Joëls, M., Fernández,., 2012. Time-dependent effects of cortisol on selec-

ive attention and emotional interference: a functional MRItudy. Front. Integr. Neurosci. 6, 66.

Hsu, F.C., Garside, M.J., Massey, A.E., McAllister-illiams, R.H., 2003. Effects of a single dose of cortisol

n the neural correlates of episodic memory and errorrocessing in healthy volunteers. Psychopharmacology 167,31—442.

Kuhlmann, S., Wolf, O.T., 2005. Cortisol and memoryetrieval in women: influence of menstrual cycle and oralontraceptives. Psychopharmacology 183, 65—71.

Kuhlmann, S., Kirschbaum, C., Wolf, O.T., 2005. Effectsf oral cortisol treatment in healthy young women on mem-ry retrieval of negative and neutral words. Neurobiol.earn. Mem. 83, 158—162.

Kumsta, R., Entringer, S., Koper, J.W., van Rossum,.F.C., Hellhammer, D.H., Wüst, S., 2010. Working mem-ry performance is associated with common glucocorticoideceptor gene polymorphisms. Neuropsychobiology 61,9—56.

Lupien, S.J., Gillin, C.J., Hauger, R.L., 1999. Workingemory is more sensitive than declarative memory to the

cute effects of corticosteroids: a dose—response study inumans. Behav. Neurosci. 113, 420—430.

Monk, C.S., Nelson, C.A., 2002. The effects of hydrocor-isone on cognitive and neural function: a behavioral andvent-related potential investigation. Neuropsychopharma-

ology 26, 505—519.

Newcomer, J.W., Selke, G., Melson, A.K., Hershey, T.,raft, S., Richards, K., Alderson, A.L., 1999. Decreasedemory performance in healthy humans induced by

1

s5

2f1

Oo6

sP

jof7

sa2

Oopc

Jof

RoP

Oipl

WnN

[‘i

CynP

pmc

Sfli3

EeN

Stw

A

Sfj

R

A

A

B

D

D

E

F

H

H

J

L

L

M

02

tress-level cortisol treatment. Arch. Gen. Psychiatry, 56,27—533.

Oei, N.Y., Tollenaar, M.S., Spinhoven, P., Elzinga, B.M.,009. Hydrocortisone reduces emotional distracter inter-erence in working memory. Psychoneuroendocrinology 34,284—1293.

Porter, R.J., Barnett, N.A., Idey, A., McGuckin, E.A.,’Brien, J.T., 2002. Effects of hydrocortisone administrationn cognitive function in the elderly. J. Psychopharmacol. 16,5—71.

Putman, P., Berling, S., 2011. Cortisol acutely reduceselective attention for erotic words in healthy young men.sychoneuroendocrinology 36, 1407—1417.

Putman, P., Hermans, E.J., Koppeschaar, H., Van Schi-ndel, A., Van Honk, J., 2007. A single administrationf cortisol acutely reduces preconscious attention forear in anxious young men. Psychoneuroendocrinology 32,93—802.

Putman, P., Hermans, E.J., Van Honk, J., 2010. Corti-ol administration acutely reduces threat-selective spatialttention in healthy young men. Physiol. Behav. 99,94—300.

Schlosser, N., Wolf, O.T., Fernando, S.C., Terfehr, K.,tte, C., Spitzer, C., . . . Wingenfeld, K., 2013. Effectsf acute cortisol administration on response inhibition inatients with major depression and healthy controls. Psy-hiatry Res. 209, 439—446.

Symonds, C.S., McKie, S., Elliott, R., William Deakin,.F., Anderson, I.M., 2012. Detection of the acute effectsf hydrocortisone in the hippocampus using pharmacologicalMRI. Eur. Neuropsychopharmacol. 22, 867—874.

Taylor, V.A., Ellenbogen, M.A., Washburn, D., Joober,., 2011. The effects of glucocorticoids on the inhibitionf emotional information: a dose—response study. Biol.sychol. 86, 17—25.

Terfehr, K., Wolf, O.T., Schlosser, N., Fernando, S.C.,tte, C., Muhtz, C., Wingenfeld, K., 2011. Hydrocortisone

mpairs working memory in healthy humans, but not inatients with major depressive disorder. Psychopharmaco-ogy 215, 71—79.

Tollenaar, M.S., Elzinga, B.M., Spinhoven, P., Everaerd,., 2009. Immediate and prolonged effects of cortisol, butot propranolol, on memory retrieval in healthy young men.eurobiol. Learn. Mem. 91, 23—31.

Tops, M., Van Peer, J.M., Wijers, A.A., Korf, J., 2006.Unpublished data referenced in the published paper,‘Acute cortisol administration reduces subjective fatiguen healthy women’’]. Unpublished raw data.

Vasa, R.A., Pine, D.S., Masten, C.L., Vythilingam, M.,ollin, C., Charney, D.S., . . . Monk, C.S., 2009. Effects ofohimbine and hydrocortisone on panic symptoms, auto-omic responses, and attention to threat in healthy adults.sychopharmacology 204, 445—455.

Vaz, L.J., Pradella-Hallinan, M., Bueno, O.F.A., Pom-éia, S., 2011. Acute glucocorticoid effects on theulticomponent model of working memory. Hum. Psy-

hopharmacol. 26, 477—487.Wingenfeld, K., Wolf, S., Krieg, J.C., Lautenbacher,

., 2011. Working memory performance and cognitiveexibility after dexamethasone or hydrocortisone admin-

stration in healthy volunteers. Psychopharmacology 217,23—329.

M

G.S. Shields et al.

Wolf, O.T., Convit, A., McHugh, P.F., Kandil, E., Thorn,.L., De Santi, S., . . . De Leon, M.J., 2001. Cortisol differ-ntially affects memory in young and elderly men. Behav.eurosci. 115, 1002—1011.

Yehuda, R., Harvey, P.D., Buchsbaum, M., Tischler, L.,chmeidler, J., 2007. Enhanced effects of cortisol adminis-ration on episodic and working memory in aging veteransith PTSD. Neuropsychopharmacology 32, 2581—2591.

ppendix A. Supplementary data

upplementary data associated with this article can beound, in the online version, at http://dx.doi.org/10.1016/.psyneuen.2015.04.017.

eferences

lexander, J.K., Hillier, A., Smith, R.M., Tivarus, M.E., Beversdorf,D.Q., 2007. Beta-adrenergic modulation of cognitive flexibilityduring stress. J. Cogn. Neurosci. 19, 468—478.

llen, A.P., Kennedy, P.J., Cryan, J.F., Dinan, T.G., Clarke, G.,2014. Biological and psychological markers of stress in humans:focus on the trier social stress test. Neurosci. Biobehav. Rev. 38,94—124.

utts, K.A., Weinberg, J., Young, A.H., Phillips, A.G., 2011.Glucocorticoid receptors in the prefrontal cortex regulatestress-evoked dopamine efflux and aspects of executive func-tion. Proc. Natl. Acad. Sci. U. S. A. 108, 18459—18464.

iamond, A., 2013. Executive functions. Annu. Rev. Psychol. 64,135—168.

ickerson, S.S., Kemeny, M.E., 2004. Acute stressors and cortisolresponses: a theoretical integration and synthesis of laboratoryresearch. Psychol. Bull. 130, 355—391.

gger, M., Smith, G.D., Schneider, M., Minder, C., 1997. Bias in meta-analysis detected by a simple, graphical test. Br. Med. J. 315,629—634.

ranz, C.E., O’Brien, R.C., Hauger, R.L., Mendoza, S.P., Panizzon,M.S., Prom-Wormley, E., Eaves, L.J., Jacobson, K., Lyons, M.J.,Lupien, S., Hellhammer, D., Xian, H., Kremen, W.S., 2011. Cross-sectional and 35-year longitudinal assessment of salivary cortisoland cognitive functioning: the Vietnam Era twin study of aging.Psychoneuroendocrinology 36, 1040—1052.

edges, L.V., Tipton, E., Johnson, M.C., 2010. Robust variance esti-mation in meta-regression with dependent effect size estimates.Res. Synth. Methods 1, 39—65.

et, S., Ramlow, G., Wolf, O.T., 2005. A meta-analytic review ofthe effects of acute cortisol administration on human memory.Psychoneuroendocrinology 30, 771—784.

oëls, M., Fernandez, G., Roozendaal, B., 2011. Stress and emo-tional memory: a matter of timing. Trends Cogn. Sci. 15,280—288.

i, G., Cherrier, M.M., Tsuang, D.W., Petrie, E.C., Colasurdo, E.A.,Craft, S., Schellenberg, S., Peskind, G.D., Raskind, M.A., Wilkin-son, C.W., 2006. Salivary cortisol and memory function in humanaging. Neurobiol. Aging 27, 1705—1714.

upien, S.J., Maheu, F., Tu, M., Fiocco, A., Schramek, T.E., 2007.The effects of stress and stress hormones on human cognition:implications for the field of brain and cognition. Brain Cognit.65, 209—237.

cEwen, B.S., 2004. Protection and damage from acute and chronicstress: allostasis and allostatic overload and relevance to the

pathophysiology of psychiatric disorders. Ann. N. Y. Acad. Sci.1032, 1—7.

iyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., Howerter,A., Wager, T.D., 2000. The unity and diversity of executive

S

S

S

T

T

T

T

T

U

v

V

Young, A.H., Sahakian, B.J., Robbins, T.W., Cowen, P.J., 1999. The

Cortisol and core executive functions

functions and their contributions to complex ‘‘frontal lobe’’tasks: a latent variable analysis. Cogn. Psychol. 41, 49—100.

Mizoguchi, K., Ishige, A., Takeda, S., Aburada, M., Tabira, T., 2004.Endogenous glucocorticoids are essential for maintaining pre-frontal cortical cognitive function. J. Neurosci. 24, 5492—5499.

Morris, S.B., DeShon, R.P., 2002. Combining effect size estimates inmeta-analysis with repeated measures and independent-groupsdesigns. Psychol. Methods 7, 105—125.

Newcomer, J.W., Selke, G., Melson, A.K., Hershey, T., Craft, S.,Richards, K., Alderson, A.L., 1999. Decreased memory per-formance in healthy humans induced by stress-level cortisoltreatment. Arch. Gen. Psychiatry 56, 527—533.

Oei, N.Y.L., Everaerd, W.T.A.M., Elzinga, B.M., Van Well, S.,Bermond, B., 2006. Psychosocial stress impairs working memoryat high loads: an association with cortisol levels and memoryretrieval. Stress 9, 133—141.

Plessow, F., Fischer, R., Kirschbaum, C., Goschke, T., 2011. Inflex-ibly focused under stress: acute psychosocial stress increasesshielding of action goals at the expense of reduced cognitiveflexibility with increasing time lag to the stressor. J. Cogn. Neu-rosci. 23, 3218—3227.

Reul, J.M.H.M., de Kloet, E.R., 1985. Two receptor systems forcorticosterone in rat brain: microdistribution and differentialoccupation. Endocrinology 117, 2505—2511.

Robbins, T.W., Arnsten, A.F., 2009. The neuropsychopharmacologyof fronto-executive function: monoaminergic modulation. Annu.Rev. Neurosci. 32, 267—287.

Roozendaal, B., Okuda, S., De Quervain, D.F., McGaugh, J.L., 2006.Glucocorticoids interact with emotion-induced noradrenergicactivation in influencing different memory functions. Neuro-science 138, 901—910.

Scammacca, N., Roberts, G., Stuebing, K.K., 2014. Meta-analysiswith complex research designs: dealing with dependence frommultiple measures and multiple group comparisons. Rev. Educ.Res. 84, 328—364.

Scholz, U., La Marca, R., Nater, U.M., Aberle, I., Ehlert, U.,Hornung, R., Kliegel, M., 2009. Go no-go performance underpsychosocial stress: beneficial effects of implementation inten-tions. Neurobiol. Learn. Mem. 91, 89—92.

Schoofs, D., Preuß, D., Wolf, O.T., 2008. Psychosocial stress inducesworking memory impairments in an n-back paradigm. Psy-choneuroendocrinology 33, 643—653.

Schoofs, D., Wolf, O.T., Smeets, T., 2009. Cold pressor stress impairsperformance on working memory tasks requiring executive func-tions in healthy young men. Behav. Neurosci. 123, 1066—1075.

Schwabe, L., Höffken, O., Tegenthoff, M., Wolf, O.T., 2013.Stress-induced enhancement of response inhibition depends on

mineralocorticoid receptor activation. Psychoneuroendocrinol-ogy 38, 2319—2326.

Schwabe, L., Tegenthoff, M., Höffken, O., Wolf, O.T., 2012. Simul-taneous glucocorticoid and noradrenergic activity disrupts the

103

neural basis of goal-directed action in the human brain. J. Neu-rosci. 32, 10146—10155.

ilverman, M.N., Sternberg, E.M., 2012. Glucocorticoid regulationof inflammation and its functional correlates: from HPA axis toglucocorticoid receptor dysfunction. Ann. N. Y. Acad. Sci. 1261,55—63.

tauble, M.R., Thompson, L.A., Morgan, G., 2013. Increases incortisol are positively associated with gains in encoding andmaintenance working memory performance in young men. Stress16, 402—410.

teptoe, A., Hamer, M., Chida, Y., 2007. The effects of acutepsychological stress on circulating inflammatory factors inhumans: a review and meta-analysis. Brain Behav. Immun. 21,901—912.

anner-Smith, E.E., Tipton, E., 2014. Robust variance estimationwith dependent effect sizes: practical considerations includinga software tutorial in Stata and SPSS. Res. Synth. Methods 5,13—30.

anner-Smith, E.E., Wilson, S.J., Lipsey, M.W., 2013. The com-parative effectiveness of outpatient treatment for adolescentsubstance abuse: a meta-analysis. J. Subst. Abuse Treat. 44,145—158.

averniers, J., Van Ruysseveldt, J., Smeets, T., von Grumbkow,J., 2010. High-intensity stress elicits robust cortisol increases,and impairs working memory and visuo-spatial declarative mem-ory in Special Forces candidates: a field experiment. Stress 13,324—334.

aylor, V.A., Ellenbogen, M.A., Washburn, D., Joober, R., 2011.The effects of glucocorticoids on the inhibition of emo-tional information: a dose—response study. Biol. Psychol. 86,17—25.

ipton, E., 2014. Small sample adjustments for robust vari-ance estimation with meta-regression. Psychol. Methods,http://dx.doi.org/10.1037/met0000011 (Epub ahead of print).

lrich-Lai, Y.M., Herman, J.P., 2009. Neural regulation of endocrineand autonomic stress responses. Nat. Rev. Neurosci. 10,397—409.

an Stegeren, A.H., Roozendaal, B., Kindt, M., Wolf, O.T., Joëls,M., 2010. Interacting noradrenergic and corticosteroid systemsshift human brain activation patterns during encoding. Neuro-biol. Learn. Mem. 93, 56—65.

asa, R.A., Pine, D.S., Masten, C.L., Vythilingam, M., Collin, C.,Charney, D.S., Monk, C.S., 2009. Effects of yohimbine andhydrocortisone on panic symptoms, autonomic responses, andattention to threat in healthy adults. Psychopharmacology 204,445—455.

effects of chronic administration of hydrocortisone on cognitivefunction in normal male volunteers. Psychopharmacology 145,260—266.