Cortisol Variation in Humans Affects Memory for Emotionally Laden and Neutral Information

Abstract

In a test of the effects of cortisol on emotional memory, 90 men were orally administered placebo or 20 or 40 mg cortisol and presented with emotionally arousing and neutral stimuli. On memory tests administered within 1 hr of stimulus presentation, cortisol elevations caused a reduction in the number of errors committed on free-recall tasks. Two evenings later, when cortisol levels were no longer manipulated, inverted-U quadratic trends were found for recognition memory tasks, reflecting memory facilitation in the 20-mg group for both negative and neutral information. Results suggest that the effects of cortisol on memory do not differ substantially for emotional and neutral information. The study provides evidence of beneficial effects of acute cortisol elevations on explicit memory in humans.

Full text

Cortisol Variation in Humans Affects Memory for Emotionally Laden and
Neutral Information
Heather C. Abercrombie, Ned H. Kalin, Marchell E. Thurow, Melissa A. Rosenkranz, and
Richard J. Davidson
University of Wisconsin—Madison
In a test of the effects of cortisol on emotional memory, 90 men were orally administered placebo or 20
or 40 mg cortisol and presented with emotionally arousing and neutral stimuli. On memory tests
administered within 1 hr of stimulus presentation, cortisol elevations caused a reduction in the number
of errors committed on free-recall tasks. Two evenings later, when cortisol levels were no longer
manipulated, inverted-U quadratic trends were found for recognition memory tasks, reflecting memory
facilitation in the 20-mg group for both negative and neutral information. Results suggest that the effects
of cortisol on memory do not differ substantially for emotional and neutral information. The study
provides evidence of beneficial effects of acute cortisol elevations on explicit memory in humans.
Cortisol elevations are one mechanism through which stress
affects learning and memory. Research, primarily in rats, has
shown that mild glucocorticoid
1
elevations enhance memory and
extreme deficiencies or elevations disrupt memory (see McEwen
& Sapolsky, 1995 for review). Circulating glucocorticoids readily
cross the blood–brain barrier, and extensive research has revealed
that glucocorticoids alter functioning of hippocampal neurons (for
reviews, see Lupien & McEwen, 1997; McEwen & Sapolsky,
1995). For instance, an inverted U-shaped function characterizes
the relation between glucocorticoids and long-term potentiation
in hippocampal neurons (Filipini, Gijsbers, Birmingham, &
Dubrovsky, 1991). This effect is due to differential activation of
the two types of corticosteroid receptors, that is, mineralocorticoid
receptors (MRs, with high affinity for cortisol) and glucocorticoid
receptors (GRs, with substantially lower affinity for cortisol; for
review, see McEwen & Sapolsky, 1995). The effects of glucocor-
ticoids on memory depend on both MR and GR activation, with
memory facilitation occurring when MRs are fully occupied and
GRs are only partially activated. It is only when GRs become
highly saturated (e.g., during stress) that deficits in memory related
to elevated glucocorticoids are observed (de Kloet, Oitzl, & Joels,
1999; Oitzl & de Kloet, 1992; Roozendaal, Bohus, & McGaugh,
1996).
Consistent with the animal research, studies in humans have also
shown various effects of cortisol on memory. Although most
human studies have shown deficiencies in memory performance
associated with acute glucocorticoid elevations (see Lupien &
McEwen, 1997 for review), a few studies have shown facilitation
with mild cortisol elevations (Beckwith, Petros, Scaglione, &
Nelson, 1986; Buchanan & Lovallo, 2001; Lupien et al., 2002).
Human data demonstrate that glucocorticoids affect hippo-
campally mediated learning but have few effects on non-
hippocampally mediated cognitive tasks, such as implicit memory
testing or vigilance (e.g., Kirschbaum, Wolf, May, Wippich, &
Hellhammer, 1996; Newcomer et al., 1999; Wolkowitz et al.,
1990). Thus, GR activation in the hippocampus has been suggested
to underlie the effects of cortisol on memory in humans (for
review, see Lupien & Lepage, 2001). However, recent data suggest
that significant species differences exist in the concentrations of
GR receptors in the hippocampus. For instance, using in situ
hybridization, Sanchez, Young, Plotsky, and Insel (2000) recently
found very low concentrations of GR mRNA in the hippocampus
of the rhesus monkey, but Patel and colleagues (2000) found high
concentrations of GR mRNA in the squirrel monkey hippocampus.
The density of GRs in the human hippocampus is currently un-
known (for review, see Lupien & Lepage, 2001). These data call
into question whether the hippocampal model accounts for all of
the observed findings and highlight the importance of expanding
the neural model of glucocorticoids’ effects on memory (Lupien &
Lepage, 2001).
The amygdala is an additional brain region known to mediate
the effects of glucocorticoids on memory (Roozendaal, 2000). In
some memory tasks, the basolateral nucleus of the amygdala is
1
Glucocorticoids are corticosterone in most rodents, and cortisol in
primates.
Heather C. Abercrombie, Marchell E. Thurow, and Melissa A. Rosen-
kranz, Department of Psychology, University of Wisconsin—Madison;
Ned H. Kalin, Department of Psychiatry, University of Wisconsin—Mad-
ison; Richard J. Davidson, Department of Psychology and Department of
Psychiatry, University of Wisconsin—Madison.
The study reported herein was conducted as part of Heather C. Aber-
crombie’s doctoral dissertation. We thank her dissertation committee mem-
bers, Craig Berridge, Morton Ann Gernsbacher, and Joseph Newman, for
their guidance. We also thank Susan Johnston, Sonia Lupien, Clemens
Kirschbaum, Holly McCreary, Dani McKinney, Keith Nuechterlein, Ad-
rian Pederson, Colleen Urben, Karen VandenBrook, and Stephen Weiler
for their consultation or assistance with various aspects of the study.
Correspondence concerning this article should be addressed either to
Heather C. Abercrombie, who is now at the Department of Psychiatry,
Wisconsin Psychiatric Institute and Clinics, 6001 Research Park Boule-
vard, Madison, Wisconsin 53719 or to Richard J. Davidson, Department of
Psychology, University of Wisconsin, 1202 West Johnson Street, Madison,
Wisconsin 53706. E-mail: abercrombie@psyphw.psych.wisc.edu or
rjdavids@facstaff.wisc.edu
Behavioral Neuroscience Copyright 2003 by the American Psychological Association, Inc.
2003, Vol. 117, No. 3, 505–516 0735-7044/03/$12.00 DOI: 10.1037/0735-7044.117.3.505
505

important in mediating the effects of glucocorticoids or GR ago-
nists infused directly into the hippocampus (for review, see
Roozendaal, 2000). For instance, in rats, bilateral lesions of the
basolateral nucleus of the amygdala block the memory-enhancing
effects of the specific GR agonist 28362 administered directly into
the hippocampus after training on an inhibitory avoidance task
(Roozendaal & McGaugh, 1997). Furthermore, lesions of the
basolateral nucleus of the amygdala that, alone, do not impair
retention block the enhancing effects of systemic injections of
glucocorticoids immediately after inhibitory avoidance training in
rats (Roozendaal & McGaugh, 1996). These and other studies have
shown that the effects of glucocorticoids on learning and memory
are importantly mediated by activity not only in the hippocampus,
but also in the basolateral nucleus of the amygdala (Roozendaal,
2000).
The Amygdala and Emotionally Based Memories
It is well established that emotional information tends to be
remembered better than neutral information (Cahill & McGaugh,
1995; Heuer & Reisberg, 1990). For instance, when study partic-
ipants are presented with both emotionally laden and neutral
stimuli (e.g., words like pain vs. cabinet, or pictures of scenes such
as a car accident vs. a boat on a quiet lake), memory tests generally
show superior performance for emotionally arousing, compared
with neutral, stimuli (Bradley, Greenwald, Petry, & Lang, 1992;
Phelps, LaBar, & Spencer, 1997). Animal and human studies have
suggested that the amygdala mediates the effects of emotion on
learning and memory. For instance, activity in the amygdala un-
derlies aversive classical conditioning in humans and animals (e.g.,
LaBar, LeDoux, Spencer, & Phelps, 1995; LeDoux, Cicchetti,
Xagoraris, & Romanski, 1990). Human neuroimaging and brain
lesion data have confirmed a role for the amygdala in the superi-
ority of memory for emotional information (Cahill et al., 1996;
Hamann, Ely, Grafton, & Kilts, 1999). For instance, using func-
tional magnetic resonance imaging, Canli, Zhao, Brewer, Gabrieli,
and Cahill (2000) found that event-related activation in the left
amygdala during encoding predicted memory performance only
for highly emotionally evocative scenes, suggesting a relatively
specific role for the amygdala in memory of emotional, but not
neutral, information. Furthermore, two patients with selective bi-
lateral damage to the amygdala failed to show normal enhance-
ment of memory for emotionally arousing information (Adolphs,
Cahill, Schul, & Babinsky, 1997).
Thus, the amygdala is involved in the superiority of memory for
emotional information, and glucocorticoids are one of several
molecular mechanisms contributing to the amygdala’s role in
memory (which also importantly includes
␤
-adrenergic activation
of basolateral amygdala neurons; for review, see Roozendaal,
2000). As a consequence of the differential roles of the amygdala
and hippocampus in emotional versus nonemotional learning (Be-
chara et al., 1995; Cahill et al., 1996; Canli et al., 2000), the
dose–response curves for the effects of glucocorticoids on memory
may vary for emotionally arousing and neutral information.
One study examining these issues (Buchanan & Lovallo, 2001)
showed memory facilitation associated with acute cortisol eleva-
tions for emotionally arousing, but not for neutral, stimuli. They
found that 20 mg cortisol caused memory facilitation for emotion-
ally arousing stimuli and no effect for neutral stimuli, and thus
concluded that cortisol facilitates memory only for emotional
information. However, other investigators have found cortisol-
related memory facilitation for neutral information (e.g., Beckwith
et al., 1986; Lupien et al., 2002; Lupien, Gillin, & Hauger, 1999).
Thus, the conclusion that cortisol facilitates memory only for
emotionally arousing material but has no effect or causes impair-
ments for neutral information may be oversimplified. A primary
goal of the current study was to further characterize the dose–
response curves for cortisol’s effects on emotional versus neutral
information in humans.
Memory and Glucocorticoids: Dose–Response
Relationship
On the basis of studies in animals, it is known that an inverted
U-shaped function characterizes the relation between memory and
glucocorticoids. However, human studies have yet to replicate
these effects for explicit memory. Human studies have shown
either facilitation or impairment in explicit memory associated
with cortisol elevations. Thus, for the current study, two doses of
exogenously administered cortisol were chosen, one hypothesized
to cause memory facilitation and another hypothesized to cause
impairment.
With few exceptions (e.g., Buchanan & Lovallo, 2001; de
Quervain, Roozendaal, Nitsch, McGaugh, & Hock, 2000), most
investigators who have studied the effects of pharmacologically
manipulated cortisol levels on memory have tested memory re-
trieval while glucocorticoid levels were still pharmacologically
elevated. It has therefore been difficult to distinguish cortisol’s
effects on memory formation from its effects on retrieval in
humans. Thus, in the current study, memory testing was performed
both during the same session as encoding and two evenings later,
when cortisol levels were no longer manipulated. Data from Ses-
sion 2 allow examination of cortisol’s effects on memory forma-
tion independent from its effects on retrieval.
In the current study, participants were given either placebo or a
single administration of one of two doses of hydrocortisone. The
doses of hydrocortisone were chosen to elevate cortisol to levels
observed during mild-to-moderate (20 mg) or extreme (40 mg)
acute stress. Drug administration occurred in the evening, when
endogenous cortisol levels are minimal. Participants were pre-
sented with words and pictures (i.e., photographs) that varied with
respect to their emotional content. Free-recall and recognition
memory for these stimuli were tested on the same evening and two
evenings later. Recognition is considered a relatively pure measure
of memory storage because it is not affected by processes that alter
generation or retrieval of items stored in memory, which are
involved in free-recall. Certain variables affect recall and recog-
nition differently (sometimes even in opposite directions, such as
word frequency; for review, see Brown, 1976). Furthermore, for
certain tests of memory, it is known that verbal encoding of
material can be deleterious for memory performance (e.g.,
Schooler & Engstler-Schooler, 1990). It is currently unknown
exactly what aspects of explicit memory are affected by glucocor-
ticoids. Thus, word and picture recall and recognition were as-
sessed with the goal of fully examining the effects of cortisol on
explicit memory processes.
506
ABERCROMBIE ET AL.

Method
Participants
Ninety paid healthy male volunteers (aged 18–33), weighing between
140–200 lbs (63.5–90.7 kg), were recruited. Individuals who met any of
the following exclusionary criteria were excluded from participation: pre-
vious exposure to the slides used in the study (i.e., International Affective
Picture System; Lang, Bradley, & Cuthbert, 1998); non-native English
speaking; medical illness within the prior 3 weeks; asthma; endocrine
disorders; history of psychopathology; current alcohol or substance abuse;
daily tobacco use; cardiac disorders; hypertension; neurological disorders;
history of head trauma; night-shift work; allergies or sensitivities that
would preclude administration of the study drug; or treatment with psych-
otropic medications, narcotics, beta-blockers, or steroids. Participants were
additionally screened upon arrival to the laboratory for vision problems
(i.e., worse than 20/40 vision) and hypertension (i.e., blood pressure ⬎
160/95). Written informed consent was obtained in accordance with the
University of Wisconsin Health Sciences Human Subjects Committee
guidelines.
Every participant who completed Session 1 also completed Session 2.
Data from two placebo participants and one 20-mg participant were ex-
cluded from data analyses because of extremely high salivary cortisol
values.
2
Procedure
Eligible participants were invited into the lab for two sessions: an initial
session, which always began at 7 p.m. (Session 1), followed two evenings
later by Session 2, which began any time between 5 and 8 p.m. Participants
were instructed to eat a light dinner at least 1 hr prior to the initial session
and to refrain from eating, exercising, and drinking anything but water for
the hour prior to both sessions. Participants were also instructed to refrain
from drinking alcohol for the 24 hr prior to both sessions. Individuals who
were occasional smokers (i.e., ⬍ 1 pack per month) were instructed not to
smoke for the week prior to the sessions. Participants were tested individ-
ually, and tasks were administered on a computer, with the exception of
self-report questionnaires and the free-recall tasks.
Session 1
Drug administration. Participants were orally administered placebo
or 20 or 40 mg hydrocortisone (which is identical to the hormone cortisol).
Drug administration was randomized
3
and double-blind, using identical
opaque capsules. Hydrocortisone tablets (Hydrocortone; Merck & Co,
Whitehouse Station, NJ) were encapsulated along with lactose, and placebo
capsules contained only lactose. After drug administration, participants sat
quietly for 40–43 min while the drug was absorbed. During the rest period,
participants watched an informative video about American cities, Rand
McNally Celebrated Cities of America (International Video Network; San
Ramon, California). This neutral to slightly positively valenced video was
shown to control, to whatever extent possible, the participant’s activities
during the drug-absorption rest period.
Saliva sampling. Salivary cortisol samples were collected with the
Salivette sampling device (Sarstedt, Rommelsdorf, Germany). During Ses-
sion 1, saliva samples were obtained at the following time points: 10 min
after arrival at the laboratory (i.e., immediately prior to drug administra-
tion), approximately every 20 min thereafter for the first 2 hr of the session,
and approximately every 25 min during the last hour of the session (Figure
1). Samples were stored frozen at –80 °C until processing.
Measurement of emotional state. To examine the relation between
cortisol levels and subjective emotional experience, current emotional state
was measured three times during Session 1: approximately 50, 100, and
160 min after drug administration. Ratings were obtained on the 20
adjectives from the Positive Affect and Negative Affect Schedule
(PANAS—State Version; Watson, Clark, & Tellegen, 1988) and on 11
additional emotional adjectives (Feldman Barrett & Russell, 1998), which
allow separation of the valence and arousal dimensions of affective space.
Encoding of negative and neutral stimuli: Rating tasks. Approxi-
mately 43 min after drug administration, all participants performed a
word-rating task followed by a picture-rating task, which exposed them to
negative and neutral stimuli.
4
Words were chosen from the Affective
Norms for English Words (ANEW; Bradley & Lang, 1999). Pictures (i.e.,
photographs) were chosen from the International Affective Picture System
(IAPS; Lang et al., 1998). For both the picture- and word-rating tasks, two
sets of stimuli (deemed Sets A and B) were developed to allow for
counterbalancing of targets and distracters in later tests of recognition
memory. Because counterbalancing adds another factor to the study design
and introduces variability, we chose to limit counterbalancing of variables
2
The 2 excluded placebo participants had salivary cortisol levels of 0.93
and 1.17
␮
g/dl, compared with the placebo group range of 0.03–0.29
␮
g/dl. On a test day information sheet, these 2 participants indicated that
they sleep into the very late morning on a regular basis, suggesting that
these participants may have had altered circadian rhythmicity of cortisol.
Furthermore, 1 participant in the 20-mg group had chewed the capsule
containing the hydrocortisone tablet. His extremely high observed salivary
cortisol level of 22.7
␮
g/dl was assumed to have resulted from hydrocor-
tisone residue left in his mouth and/or from an increased absorption rate.
3
For the first 30 participants, the doses used for the current study were 5
and 20 mg cortisol, and drug administration was truly randomized (i.e.,
1:1:1 ⫽ placebo: 5 mg: 20 mg). However, the 5-mg dose was found to
produce insufficient cortisol elevations and was thus discarded, and a
40-mg group was added. After this point, drug administration was pseudo-
randomized (i.e., 2:2:3 ⫽ placebo: 20 mg: 40 mg).
4
Rating tasks were administered a few minutes following the third
(Minute 41) and before the fifth (Minute 73) saliva samples.
Figure 1. Mean salivary cortisol elevations for each group prior to and
following drug administration (at approximately 7:13 p.m.). Salivary cor-
tisol levels among the groups did not differ at the time of the baseline
sample (3 min before drug administration). Error bars represent SEM. The
apparent lack of error bars for the placebo group is due to the minimal
variation in cortisol concentration in this group. To convert to nanomoles
per liter, multiply values in micrograms per deciliter by 27.6.
507
CORTISOL AND MEMORY FOR EMOTIONAL INFORMATION

to the recognition memory task. Recognition memory targets and distract-
ers were considered the most important items to counterbalance in order to
ensure that memory effects were not related merely to a particular set of
stimuli. Thus, other variables, such as order of presentation of the rating
tasks, were kept constant for all participants.
Both picture sets included 56 pictures (28 negative and 28 neutral) that
were matched on average normative ratings of pleasantness and arousal
(Lang et al., 1998). See Table 1 for average normative ratings. To facilitate
free-recall testing, content overlap among pictures was minimized within
each set. Both word sets included 44 words (22 negative and 22 neutral)
that were matched on average normative ratings of pleasantness and
arousal (Bradley & Lang, 1999; Table 1). Word sets were also matched on
frequency of usage (ps ⬍ .71; Carroll, Davies, & Richman, 1971), and
length (ps ⬍ .28).
During the rating tasks, participants were instructed to rate pictures or
words on the basis of how they felt while viewing each stimulus, and were
not told that they would later be asked to recall the stimuli. Participants
rated pictures (5-s stimulus presentation duration) and words (4-s presen-
tation duration) on two 9-point numeric scales assessing pleasantness and
arousal.
Continuous performance test. After the rating tasks were completed,
the degraded stimulus continuous performance test (DS-CPT) developed
by Nuechterlein and Asarnow (1999) was administered as a control task to
test for potential differences in vigilance as a result of hydrocortisone
administration.
Memory assessment. Explicit memory was assessed for the stimuli in
the word- and picture-rating tasks. Participants were not given feedback on
their performance for any of the memory tests.
Free-recall. Participants completed separate free-recall tasks for words
and pictures, in which they were instructed to list all the words or briefly
describe all the pictures they could remember from the rating tasks.
Participants were given 7 min to complete the word free-recall and 11 min
to complete the picture free-recall tasks. In addition to number of correct
responses, free-recall tasks were also scored for intrusive errors, that is,
errors of commission, which were responses that were not presented in the
word rating task or descriptions of pictures that were not presented in the
picture rating task. Because scoring the free-recall lists for pictures entailed
a degree of subjectivity, interrater reliability was computed for the first 30
participants’ data. All remaining picture free-recall lists were scored by one
person (Heather C. Abercrombie), as interrater reliability was found to be
extremely high (intraclass correlation coefficients ⬎ .98).
Recognition memory. After the free-recall tasks were completed, sep-
arate recognition memory tests for words and pictures were administered.
The tasks involved use of a two-button “yes” or “no” response pad to
indicate whether or not test stimuli were presented during the rating tasks.
Half of the test stimuli were targets (previously viewed), and half were
distracters (new stimuli). Instructions emphasized both speed and accuracy.
Distracters were pictures or words from the alternate set of stimuli not
presented during encoding (i.e., A or B, accordingly). Only half of the
stimuli from each Set A and B were used for the Session 1 recognition
memory tasks. Testing only a subset of the total pool of stimuli during
Session 1 made available a set of distracters for Session 2 that were
completely new, and targets that had been viewed only during the encoding
tasks, allowing for a Session 2 recognition memory not contaminated by
practice effects from the Session 1 recognition tests. The subsets of stimuli
chosen for the recognition tests from Sets A and B were psychometrically
matched on normative ratings. Comparisons between the Set A and B
subsets (within valence) revealed no differences (for pleasantness, fre-
quency, and word length, ps ⬎ .50 for pictures and ps ⬎ .45 for words; for
arousal, ps ⬎ .18).
The participant’s ability to discriminate between previously presented
and new items, (i.e., “sensitivity”) served as the dependent variable for
recognition memory. The sensitivity index Pr was used (Snodgrass &
Corwin, 1988). Pr is the proportion of old items (targets) endorsed minus
the proportion of new items (distracters) endorsed, that is, hits minus false
alarms, with positive scores reflecting more hits than false alarms. This
metric does not require that the data be normally distributed, and it
provides a measure of sensitivity that is independent from bias (Snodgrass
& Corwin, 1988). Cortisol dose was not hypothesized to be related to bias
or reaction time, and these data are therefore excluded.
Session 2
At Session 2, no drug was administered. Three saliva samples (approx-
imately 25 min apart) were collected according to methods identical to
those of Session 1. Participants’ memory for stimuli viewed during the
Session 1 rating tasks was again assessed, and order of presentation of the
memory tests was identical to Session 1. In Session 2, recognition memory
measures were derived from the sets of test stimuli not used during
Session 1 (see explanation above).
Table 1
Stimuli Sets A and B: Average Normative Ratings
Stimuli
Negative stimuli Neutral stimuli
Set A Set B Set A Set B
Picture rating task
IAPS
Pleasantness 2.38 ⫾ 0.42 2.42 ⫾ 0.41 4.89 ⫾ 0.37 4.91 ⫾ 0.33
Arousal 6.00 ⫾ 0.60 5.99 ⫾ 0.58 2.57 ⫾ 0.32 2.50 ⫾ 0.36
Word rating task
ANEW
Pleasantness 2.59 ⫾ 0.53 2.58 ⫾ 0.57 5.12 ⫾ 0.56 5.09 ⫾ 0.57
Arousal 5.59 ⫾ 0.88 5.58 ⫾ 0.72 3.90 ⫾ 0.30 3.94 ⫾ 0.39
Note. Values are reported as means (⫾ SD). Pleasantness and arousal values are average normative ratings for
stimuli used in each set (pleasantness: 1 ⫽ highly unpleasant,5⫽ neutral,9⫽ highly pleasant; arousal: 1 ⫽
low,9⫽ high). Comparisons between Sets A and B (within valence) revealed no differences: for the words, ps ⬎
.71; for the pictures, ps ⬎ .47. IAPS ⫽ International Affective Picture System; ANEW ⫽ Affective Norms for
English Words.
508
ABERCROMBIE ET AL.

Processing of Saliva Samples
Prior to the cortisol assay, samples were centrifuged at 4500 rpm for 10
min, and the supernatant was transferred to 2-ml tubes for storage (⫺70 °C)
until assayed. Cortisol was assayed with the
125
I Cortisol RIA kit (Pantex,
Santa Monica, CA) modified for saliva. Individuals performing cortisol
assays were unaware of group assignment (i.e., dose). The detection limit
of the assay (ED
80
) was 0.03
␮
g/dl. The mean interassay and intra-assay
variation was 7.4% and 3.8%, respectively (for additional details, see
Smider et al., 2002).
Primary Analyses
Effects of dose and stimulus valence on memory performance. To
test for the effects of dose and stimulus valence (i.e., negative or
neutral) on free-recall correct responses and recognition memory sen-
sitivity (and to ensure that these effects did not vary for the alternate
stimulus sets
5
) mixed three-way analyses of variance (ANOVAs) were
computed with dose (placebo, 20 mg, and 40 mg), stimulus valence
(negative or neutral), and set (A or B) as variables. Significant effects
of dose were followed up with comparisons among individual means
and trend analyses to test for the predicted quadratic trend (Keppel,
1991).
For errors of commission (i.e., intrusive errors) in the recall tasks, mixed
two-way ANOVAs were computed with dose and set as variables. Com-
mission error scores were not separated by stimulus valence because errors
were not always easily scored as negative or neutral, and because errors, by
definition, did not correspond to the negative or neutral stimuli obtained
from the ANEW or IAPS normative sets. Thus, stimulus valence was not
included as a variable in the analyses of errors.
Correlational analyses between cortisol levels and memory perfor-
mance. In order to take advantage of the within-group variation in cor-
tisol levels, we tested correlations between observed salivary cortisol levels
and memory scores within each group separately, allowing further exam-
ination of the relation between cortisol and memory. Correlational analyses
were conducted using the average of cortisol samples taken after drug
uptake was regressed on the cortisol concentrations for Sample 1 to remove
variance related to baseline cortisol levels. The residualized scores are
hereafter referred to as the “postdrug cortisol levels.”
Additional Analyses
Effects of dose on vigilance levels. The DS-CPT task provides sig-
nal detection metrics for overall performance and for performance
broken down into three blocks. A one-way, between-group ANOVA
was performed to test for effects of dose (placebo, 20 mg, and 40 mg)
on overall DS-CPT scores (i.e., DS-CPT scores collapsed across block).
To test the effects of dose on vigilance decrements over time, a
two-way mixed ANOVA was performed with dose and block (DS-CPT
Blocks 1, 2, and 3) as variables. In addition, one-way ANOVAs were
computed for each block separately to adequately test for any effects of
dose.
Effects of dose on emotional ratings. The effects of dose on emotional
ratings of stimuli viewed during encoding tasks, and the effects of dose on
current emotional state were examined with one-way ANOVAs.
Results
Salivary Cortisol Levels
Baseline salivary cortisol levels did not differ among the
three groups: 3 min before drug administration, F(2,
84) ⫽ 0.51, ns. See Figure 1 for time course and magnitude of
salivary cortisol elevations following drug administration. Cor-
tisol levels in the 20-mg group were commensurate with en-
dogenous elevations occurring during moderate behavioral
stressors (e.g., final exam) or moderate exercise stress (e.g., 30
min on a stationary bicycle). The cortisol levels observed within
the 40-mg group remained within the physiological range of
cortisol, but such levels would be seen only during extreme
stress, such as a marathon run or surgery (Kirschbaum &
Hellhammer, 1989, 1994).
Analyses were performed within the placebo group to test for
the presence of endogenous cortisol elevations associated with
viewing emotional stimuli. Within-group t tests for the placebo
group, examining the differences between the average of the three
samples prior to the encoding tasks and each sample thereafter,
revealed only declining cortisol values (ts ⬎ 3.0, ps ⬍ .01). Thus,
viewing negative pictures and words did not cause endogenous
cortisol elevations.
No significant differences between groups in cortisol levels
occurred at any Session 2 time point (Fs ⬍ 1.02).
Vigilance
No effects of dose were found for DS-CPT performance ana-
lyzed across block or separately by block (highest F ⫽ 0.92).
Rating Tasks: Pleasantness and Arousal Ratings of Words
and Pictures During Encoding
Pleasantness
There were no effects of dose on pleasantness ratings of nega-
tive or neutral words or pictures (Fs ⬍ 1.33).
Arousal
There were no effects of dose on arousal ratings (for negative
stimuli: Fs ⬍ 1.32). However, marginally significant effects of
dose were found for arousal ratings of neutral stimuli for both
pictures, F(2, 81) ⫽ 2.76, p ⫽ .07, and words, F(2, 81) ⫽ 3.08,
p ⬍ .06. Comparisons of individual means revealed that the 40-mg
group rated neutral stimuli as more arousing than both the placebo
group: pictures, t(56) ⫽ 1.90, p ⫽ .06; words, t(56) ⫽ 2.23, p ⬍
.05, and the 20-mg group: pictures, t(57) ⫽ 2.28, p ⬍ .03; words,
t(57) ⫽ 2.17, p ⬍ .05.
Current Emotional State
No main effects or interactions of dose were found for any of the
current emotional state indices (Fs ⬍ 2.08). Thus, emotional state
was not significantly altered by cortisol elevations.
Session 1 Memory Results
Analyses of free-recall correct responses for both picture and
word free-recall at Session 1 revealed no main effects of dose or
Dose ⫻ Stimulus Valence interactions (Fs ⬍ 1.86). See Table 2
for comparison of free-recall correct responses versus errors of
commission.
5
Because the effects of set rarely interacted with the effects of dose,
significant effects of set are reported as footnotes.
509
CORTISOL AND MEMORY FOR EMOTIONAL INFORMATION

Analyses of errors of commission for picture recall at Session 1
revealed a main effect of dose, F(2, 79) ⫽ 3.07, p ⫽ .05 (see
Figure 2 and Table 2). The 40-mg group made fewer errors than
the placebo group, t(56) ⫽ 2.22, p ⬍ .05. The 20-mg group
showed a trend toward fewer errors than the placebo group,
t(55) ⫽ 1.80, p ⫽ .08. The effect size for the difference between
the 40-mg and placebo groups was 0.58, and between the 20-mg
and placebo groups, the effect size was 0.48.
Similarly, a main effect of dose was found for Session 1 word
recall errors, F(2, 78) ⫽ 3.28, p ⬍ .05 (see Figure 2 and Table 2).
Both the 20-mg group, t(55) ⫽ 2.58, p ⬍ .02, and the 40-mg
group, t(55) ⫽ 2.57, p ⬍ .02, made fewer errors than the placebo
group. Effect sizes for the differences between the 20-mg and
placebo groups and between the 40-mg and placebo groups were
both 0.68. For sensitivity in the picture and word recognition tests
at Session 1, no main effects or interactions of dose were found
(Fs ⬍ 1.7).
Session 2 Memory Results
Free-Recall
No effects of dose on errors of commission were found for
Session 2 (Fs ⬍ 1.47).
6
Analyses of free-recall correct responses
for both picture and word free-recall at Session 2 revealed no main
effects of dose or Dose ⫻ Stimulus Valence interactions
(Fs ⬍ 1.82).
7
Picture Recognition
For picture recognition memory at Session 2, a main effect of
dose on sensitivity was found, F(2, 78) ⫽ 6.37, p ⬍ .01, but a
Dose ⫻ Stimulus Valence interaction was not found, F(2,
78) ⫽ 0.08, ns (see Figure 3). The predicted quadratic trend was
found across stimulus valence, F
quadratic
⫽ 6.58, p ⬍ .02. How
-
ever, both the 20-mg group, t(54) ⫽ 3.51, p ⬍ .01, and the 40-mg
group, t(53) ⫽ 2.28, p ⬍ .05, performed better than the placebo
group. The effect size for the difference between the 20-mg and
placebo group was 0.94, and between the 40-mg and placebo
group, the effect size was 0.61. As expected, negative pictures
were better recognized than neutral pictures across dose levels,
F(1, 78) ⫽ 19.36, p ⬍ .01.
Word Recognition
For word recognition at Session 2, a main effect of dose was
found, F(2, 78) ⫽ 3.21, p ⬍ .05. Again, no Dose ⫻ Stimulus
Valence interaction emerged, F(2, 78) ⫽ 1.06, ns. The predicted
quadratic trend was found across Stimulus Valence, F
quadratic
⫽ 5.8, p ⬍ .05 (see Figure 3). Across negative and neutral words,
the 20-mg group performed better than the placebo group,
t(54) ⫽ 2.15, p ⬍ .05 (effect size: d ⫽ 0.57), and marginally
significantly better than the 40-mg group, t(55) ⫽ 1.91, p ⫽ .06.
There was no difference between the 40-mg and placebo groups,
t(53) ⫽ .30, ns. There was a trend toward a main effect of stimulus
valence, reflecting unexpectedly better performance for recogni-
tion of neutral words than negative words, F(1, 78) ⫽ 3.6, p ⬍
.07.
8
Misses Versus False Alarms
Misses and false alarms were analyzed separately to determine
whether errors of omission or errors of commission solely caused
the effects of dose on recognition memory. Of the four recognition
memory tests (i.e., Sessions 1 and 2, pictures and words), only the
Session 2 pictures revealed a main effect of dose for false alarms,
6
Likely due to the fact that negative words were more semantically
related than were neutral words. See the Discussion.
7
For errors of commission in the word free-recall task during Session 2,
a Dose ⫻ Set interaction was found, F(2, 78) ⫽ 5.08, p ⬍ .01, such that
an effect of dose on errors emerged for Set B, F(2, 39) ⫽ 4.03, p ⬍ .05,
but not for Set A, F(2, 40) ⫽ 1.96.
8
For word free-recall correct responses at Session 2, a Dose ⫻ Set
interaction was found such that only Set B showed the predicted quadratic
relationship between dose and number of words recalled: for Set B, F(2,
39) ⫽ 6.32, p ⬍ .005, but for Set A, F(2, 40) ⫽ 0.75.
Table 2
Comparison of Free-Recall Correct Responses and Errors of Commission
Measures
Session 1 Session 2
Placebo 20 mg 40 mg Placebo 20 mg 40 mg
Picture free-recall
Correct responses
Negative pictures 16.52 ⫾ 3.86 17.38 ⫾ 3.76 18.03 ⫾ 4.22 16.15 ⫾ 3.72 17.00 ⫾ 3.76 16.55 ⫾ 3.84
Neutral pictures 10.11 ⫾ 3.11 11.24 ⫾ 2.64 10.97 ⫾ 2.53 10.15 ⫾ 3.29 11.17 ⫾ 3.50 11.48 ⫾ 3.42
Pictures: Errors of commission 1.11 ⫾ 1.03 0.66 ⫾ 0.86 0.53 ⫾ 0.94* 0.63 ⫾ 0.88 0.66 ⫾ 0.86 0.79 ⫾ 1.11
Word free-recall
Correct responses
Negative words 5.50 ⫾ 2.62 5.93 ⫾ 2.49 5.62 ⫾ 2.24 4.33 ⫾ 2.18 5.17 ⫾ 2.67 4.66 ⫾ 2.24
Neutral words 2.39 ⫾ 1.59 3.38 ⫾ 1.47 3.20 ⫾ 2.06 2.81 ⫾ 2.00 3.69 ⫾ 2.02 2.79 ⫾ 1.70
Words: Errors of commission 3.36 ⫾ 2.70 1.83 ⫾ 1.67* 1.89 ⫾ 1.42* 3.04 ⫾ 2.07 2.76 ⫾ 2.28 2.37 ⫾ 2.00
Note. Values are reported as means (⫾ SD).
*p ⱕ .05, Session 1 reduction in errors related to cortisol elevations.
510
ABERCROMBIE ET AL.

F(2, 78) ⫽ 3.80, p ⬍ .05, as well as a marginally significant main
effect of dose on misses, F(2, 78) ⫽ 2.92, p ⫽ .06. No other effects
of dose were found for analyses conducted separately on misses
and false alarms. The ability to correctly accept targets and cor-
rectly reject distracters must both be taken into account for the
effects of dose to emerge. Thus, for recognition memory, neither
misses (errors of omission) nor false alarms (errors of commission)
alone accounted for the effects of dose.
Correlations Between Observed Salivary Cortisol Levels
and Memory Performance
Within the placebo and 40-mg groups, postdrug salivary cortisol
levels were not correlated with memory performance (ps ⬎ .21).
However, within the 20-mg group, postdrug salivary cortisol levels
were negatively correlated with performance on free-recall tests
(i.e., correct free-recall responses; see Table 3). Scatter plots
confirmed that correlations were not due to outliers. Furthermore,
for word recall, scores for negative stimuli alone were significantly
related to cortisol levels, after variance related to memory for
neutral stimuli had been removed (Session 1 increment in R
2
⫽
.22, p ⬍ .05; Session 2 increment in R
2
⫽ .16, p ⬍ .05). Thus,
within the 20-mg group, higher cortisol levels predicted worse
recall performance, especially for negative stimuli.
Discussion
The current study replicates and extends previous research
showing that acute glucocorticoid elevations affect explicit mem-
ory performance, in the absence of effects on other types of
cognitive measures, such as vigilance (Kirschbaum et al., 1996;
Newcomer et al., 1999; Wolkowitz et al., 1990). In the present
study, memory tests were administered at two time points, once
during the same session as encoding of stimuli, while cortisol
levels were concurrently elevated (Session 1), and two evenings
after the encoding session, when cortisol levels were no longer
manipulated (Session 2). Memory effects differed for the two
sessions. Compared with placebo, single doses of either 20 or 40
mg cortisol administered 40–45 min prior to encoding caused
fewer errors of commission (i.e., intrusive errors) on a free-recall
test during Session 1. However, during Session 2, effects emerged
for recognition memory tests. An inverted-U quadratic trend was
found across negative and neutral stimuli, with memory facilitation
observed most predominantly in the 20-mg group, and less facil-
itation or none at all in the 40-mg group.
Figure 2. Graphs representing group means for Session 1 free-recall
errors of commission. For words, participants in both the 20- and 40-mg
groups made fewer errors than placebo group subjects (ts ⬎ 2.55). For
pictures, participants in the 40-mg group made fewer errors than placebo
group subjects (t ⫽ 2.20, p ⬍ .05), and participants in the 20-mg group
made marginally fewer errors than the placebo group (t ⫽ 1.80, p ⫽ .08).
Error bars represent SEM.
Figure 3. Graphs representing group means for Session 2 recognition
memory tasks. For both words and pictures, main effects of dose were
found (Fs ⬎ 3.21). Error bars represent SEM.
511
CORTISOL AND MEMORY FOR EMOTIONAL INFORMATION

Glucocorticoids and Emotional Memory
In the current study, no interaction was found between stimulus
valence and dose, which suggests that the effects of cortisol on
memory do not differ substantially for negative and neutral infor-
mation. However, within the 20-mg group, higher cortisol levels
predicted poorer free-recall performance, primarily for negative
stimuli. For the words, poorer recall for negative stimuli was
related to higher cortisol levels, even after performance for neutral
stimuli was accounted for. Thus, across memory tests, the 20-mg
group showed memory facilitation, but within this group, poorer
memory (especially for negative stimuli) was associated with
higher cortisol levels. For negative information, the peak of the
hypothetical dose–response curve between explicit memory for-
mation and cortisol levels may have resided in the lower cortisol
elevations in the 20-mg group. These findings suggest the specu-
lative conclusion that there may exist a more lawful relation
between decline in memory performance and increasing cortisol
levels for negative, compared with neutral, information. The drop-
off in memory performance that occurs with extreme cortisol
elevations may occur more reliably for negative information than
for neutral information, and the signal-to-noise ratio between glu-
cocorticoids and memory may be enhanced for negative, compared
with neutral, information. Together, the results from the current
study suggest that an inverted-U-shaped function characterizes the
relation between cortisol elevations and delayed recognition mem-
ory for both negative and neutral information, and that marked
differences in the effects of cortisol on memory for emotional and
neutral information do not exist. However, the findings are possi-
bly suggestive of subtle valence-related differences in dose–
response curves, namely, that the decline in memory associated
with elevated cortisol may be more reliable for negative, compared
with neutral, information. Future studies should specifically ad-
dress this hypothesis.
The results obtained by Buchanan and Lovallo (2001), of
cortisol-related memory facilitation only for emotionally arousing
information, might be explained by a shift in the hypothetical
dose–response curve for negative information. In animals, dose–
response curves for exogenous administration of glucocorticoids
vary on the basis of the inherent stressfulness of the task
(Roozendaal, 2000). A leftward dose–response shift for stressful
tasks is at least partially caused by differences in endogenous
cortisol elevations. For tasks that produce very small glucocorti-
coid responses, moderate doses cause facilitation, but for tasks that
elicit a strong endogenous glucocorticoid response, moderate glu-
cocorticoid doses cause memory impairment (Roozendaal, 2000).
However, passive viewing of aversive stimuli is typically not
sufficient to cause endogenous cortisol elevations (Hubert & de
Jong-Meyer, 1991; Kirschbaum & Hellhammer, 1994), and in the
current study, no cortisol elevations associated with viewing stim-
uli were found in the placebo group. It is possible that other
mechanisms in addition to differences in endogenous cortisol
responses might cause a subtle dose–response shift, such as the
combined effects of cortisol and the differential roles of the amyg-
dala and hippocampus in emotional and nonemotional memory
(e.g., Bechara et al., 1995).
Inverted-U-Shaped Dose–Response Function in Humans
The current study demonstrates in humans the hypothesized
inverted U-shaped relation between cortisol and explicit memory,
with facilitation occurring primarily at the 20-mg dose (which
produced cortisol levels commensurate with endogenous eleva-
tions during moderate stress). However, the 40-mg dose (commen-
surate with extreme stress) was hypothesized to be sufficient to
cause acute impairment in memory, but it did not cause perfor-
mance deficits. Other studies have shown memory impairments at
doses even lower than the moderate 20-mg dose in the current
study (e.g., Kirschbaum et al., 1996). Across different studies,
identical cortisol elevations have caused opposite effects on mem-
ory performance. These differences in findings may be explained
by the circadian rhythmicity of cortisol, given that the timing of
studies has differed (e.g., the Kirschbaum et al., 1996 study oc-
curred in the late morning/early afternoon, whereas both the cur-
rent study and the Buchanan & Lovallo, 2001 study occurred in the
late afternoon and evening). Early in the day, GRs are already
moderately saturated because of high endogenous morning cortisol
levels, and it is likely that a small cortisol dose (which would cause
no deficits or facilitation in the evening) would oversaturate GRs,
causing memory impairment. This view is supported by an abun-
dance of animal and human research showing the importance
of time of day in relation to centrally mediated glucocorticoid
effects (e.g., Bradbury, Akana, & Dallman, 1994; Fehm-
Wolfsdorf, Reutter, Zenz, Born, & Lorenz, 1993; Lupien et al.,
2002).
Additional factors also may account for the differences among
studies. For instance, doses that do not affect memory after 1 day
of exposure have been found to cause memory impairments after
multiple days of treatment (e.g., Newcomer et al., 1999). The rate
of rise of cortisol levels may also be an important factor. In the
current study, the rate of rise of cortisol was relatively slow
compared with that seen in some other studies. For instance, in the
Kirschbaum et al. (1996) study, subjects peaked at a mean salivary
cortisol concentration of 2.3
␮
g/dl within 90 min after adminis-
tration of a 10-mg hydrocortisone tablet, whereas 90-min cortisol
elevations of only 1.6
␮
g/dl (which were continuing to increase)
were found in the current study after a 20-mg encapsulated oral
dose. The relatively slow increase in the current study was likely
due to a time-release effect caused by encapsulation of hydrocor-
tisone tablets. It is known that the negative feedback effects of
glucocorticoids vary depending on the rate of rise of plasma
concentrations (Keller-Wood & Dallman, 1984). It is possible that
Table 3
Correlations Between Cortisol Levels and Free-Recall
Performance Within the 20-mg Group
Session and task All Negative Neutral
Session 1
Word recall ⫺.47* ⫺.48** ⫺.15
Picture recall ⫺.31 ⫺.21 ⫺.22
Session 2
Word recall ⫺.42* ⫺.47* ⫺.23
Picture recall ⫺.45* ⫺.39* ⫺.33
Note. Data represent r values for correlations between postdrug cortisol
levels and free recall correct responses (for all stimuli, collapsed across
negative and neutral, and separately for negative and neutral stimuli).
* p ⬍ .05. ** p ⬍ .01.
512
ABERCROMBIE ET AL.

study procedures that led to slower rate of rise of cortisol may have
contributed to the observations of beneficial effects on explicit
memory processes in humans (Beckwith et al., 1986; Buchanan &
Lovallo, 2001). Examination of this issue would require memory
testing after systematic variation of rate of rise of cortisol by
means of intravenous infusion. It is clear that there does not exist
a simple dose–response relation between cortisol levels and mem-
ory performance. Various factors (such as time of day, rate of rise
of cortisol, or type of task) moderate cortisol’s role in memory.
Differential Effects of Cortisol on Memory Formation
Versus Retrieval
The different findings from Session 1 and 2 in the current study
provide further clarification of glucocorticoids’ differential roles in
memory formation and retrieval. Because memory effects were
found at Session 2 when cortisol levels were not manipulated, the
study provides definitive evidence in humans that cortisol has
effects on memory formation that are independent from its effects
on retrieval (de Quervain et al., 2000). Not only did results from
the current study differ for the types of memory tasks in Session 1
and 2 (free-recall errors vs. recognition memory), but the shapes of
the dose–response curves also differed for Session 1 and 2. Unlike
the Session 2 recognition memory tasks, no quadratic function was
apparent for the group differences in Session 1 errors of commis-
sion. This suggests that different neuropsychological processes
account for the Session 1 and Session 2 results.
One could speculate that effects of glucocorticoids primarily on
retrieval or generative processes might account for the Session 1
results. McEwen (1982) argued that glucocorticoids affect the
capacity of the hippocampus to filter out task-irrelevant stimuli,
and it is known that glucocorticoids modulate the strength of a fear
response to a context, but not to a specific cue (Pugh, Fleshner, &
Rudy, 1997). Thus, glucocorticoids may affect selection of con-
textually appropriate responses and filtering of internally gener-
ated responses. For Session 1 free-recall tasks, hydrocortisone may
have facilitated inhibition of incorrect answers, filtering out inap-
propriate information during retrieval. In addition to the current
study, previous studies have also found effects of glucocorticoids
on errors of commission in the absence of effects on errors of
omission (Wolkowitz et al., 1990). Furthermore, a recent study (de
Quervain et al., 2000) provided evidence that cortisol has effects
on retrieval that are independent of its effects on memory forma-
tion. de Quervain and colleagues found deficits in memory per-
formance when cortisol was given before testing only, rather than
before encoding (although these findings were for correct re-
sponses rather than errors). Thus, it is known that glucocorticoids
affect retrieval processes, and it is plausible that effects isolated to
generative processes during retrieval primarily accounted for
group differences in Session 1 errors of commission in the current
study.
Furthermore, the observation of no differences for recognition
memory during Session 1 in the current study suggests that con-
sequential differences among the dose groups in encoding pro-
cesses did not occur. If group differences in encoding solely
accounted for recognition memory results at Session 2, then rec-
ognition memory findings as strong as those seen during Session 2
should have been apparent in Session 1. Furthermore, direct effects
of dose on retrieval cannot account for the Session 2 findings
(because cortisol levels were not manipulated at Session 2, and
cortisol levels among the groups did not differ at Session 2). Thus,
effects on encoding and retrieval most likely do not account for the
effects of dose on Session 2 recognition. The current study there-
fore suggests that hydrocortisone affects consolidation processes,
which is highly consistent with the animal literature. In rats,
systemic injections of glucocorticoids have their strongest effects
when given in a narrow time window following training, with low
posttraining doses facilitating, and high posttraining doses impair-
ing, performance on various delayed tests of memory occurring
when glucocorticoid levels are no longer manipulated (Lupien &
McEwen, 1997; Roozendaal, 2000). Thus, the Session 1 and 2
results together suggest that cortisol has separable effects on
consolidation and retrieval processes.
Cortisol and Current Emotional State
In the current study, dose was not related to current emotional
state or to pleasantness ratings of stimuli. These data suggest that
at these doses, cortisol is not strongly related to variation in
positive or negative affect. However, the current study showed a
trend toward higher arousal ratings of neutral stimuli for partici-
pants administered 40 mg hydrocortisone, which suggests that
extreme elevations in cortisol may be associated with feeling
aroused in response to stimuli that are objectively nonarousing.
However, these findings should be interpreted with caution. Not
only are the effects marginally significant, but the group difference
occurred only in the 40-mg group, whose salivary cortisol levels
on average were ⬎ 4
␮
g/dl while rating the stimuli. Endogenous
cortisol increases of this magnitude do occur, for example, after a
marathon, during surgery, or after extreme trauma (Kirschbaum &
Hellhammer, 1989, 1994; Resnick, Yehuda, Pitman, & Foy, 1995).
However, they do not constitute normal variation in cortisol levels.
Prior studies have shown that pharmacologically administered
glucocorticoids can increase negative or positive mood and acti-
vation (Plihal, Krug, Pietrowsky, Fehm, & Born, 1996; Schmidt,
Fox, Goldberg, Smith, & Schulkin, 1999). However, doses used
are typically outside the physiological range. For instance, the dose
of cortisol used in the Plihal study (10 mg/hr for 9 hr overnight)
would likely produce cortisol concentrations outside of the phys-
iological range. Thus, evidence exists that cortisol by itself can
affect subjective state, but endogenously occurring variation likely
accounts for a relatively small portion of the variance in mood-
related variables.
Limitations and Future Directions
A limitation of both the current study and the Buchanan and
Lovallo (2001) study is that the same study participants were
presented both neutral and emotional stimuli. It may be that
viewing neutral stimuli in the context of negative stimuli moder-
ates cortisol’s effects on memory. Furthermore, although the study
was adequately powered to detect the basic treatment effect of
cortisol on memory, it may not have been sufficiently powered to
detect valence-related differences in the effects of cortisol on
memory. Important follow-up research includes a placebo-
controlled dose–response study with stimulus valence as a
between-subjects variable (in which participants view only neutral
or only emotionally arousing stimuli). To examine the role of
513
CORTISOL AND MEMORY FOR EMOTIONAL INFORMATION

endogenous cortisol elevations in human emotional memory, in-
vestigators must study tasks that elicit cortisol elevations, such as
social stress tasks (e.g., Kirschbaum, Pirke, & Hellhammer, 1993),
and test memory for emotionally laden and neutral stimuli learned
during endogenous cortisol elevations. Decreasing cortisol or its
ability to interact with GR (e.g., with metyrapone or RU486), as
well as examining interactions with noradrenergic effects on emo-
tional memory (e.g., O’Carroll, Drysdale, Cahill, Shajahan, &
Ebmeier, 1999), will each provide further clarification of the role
of stress hormones in emotional memory. Furthermore, the role of
glucocorticoids in declarative memory for emotional information
may be quite different from their role in aversive associative
learning. It is also known that glucocorticoids are differentially
involved in cue conditioning versus hippocampally based context
conditioning (Pugh et al., 1997). The varying effects of stress-
related elevations in hormones on different types of emotional
learning must be further studied in humans.
Furthermore, if inverse correlations between cortisol levels and
recall of negative stimuli were apparent in both the 20- and 40-mg
groups (i.e., not just in the 20-mg group) in the current study, then
the speculative conclusion that the drop-off in memory that occurs
with increasing cortisol levels is more reliable for negative than for
neutral information could be stated with more confidence. Possibly
studies with intravenous infusion and detection of cortisol, which
keep tighter control on circulating cortisol levels than oral admin-
istration, will allow the opportunity to more precisely examine the
very subtle differences that appear to characterize the dose–
response relations for emotional and neutral information.
This line of research will provide important data regarding
processes involved in traumatic emotional memories. It must be
determined whether cortisol increases during traumatic events play
a role in the formation of traumatic memories. It may be the case
that extreme elevations in cortisol are protective against too strong
a memory trace. This interpretation is consistent with data showing
that inadequate cortisol elevations following trauma is predictive
of posttraumatic stress disorder (PTSD; McFarlane, Atchison, &
Yehuda, 1997; Resnick et al., 1995), which necessarily includes
tenacious recollections or reexperiencing of the traumatic event.
Noradrenergic mechanisms are also implicated in PTSD (Yehuda,
Southwick, Giller, Ma, & Mason, 1992). Future research is needed
to elucidate the role of both cortisol and noradrenergic mecha-
nisms in traumatic memory.
Psychological processes indirectly related to memory may ac-
count for some of the current findings. For instance, reductions in
errors of commission in the free recall tasks during Session 1 may
have been due to effects of cortisol on impulsivity, reducing the
amount of guessing during the tasks. Evidence exists which sug-
gests that endogenous cortisol levels under various conditions may
be negatively correlated with trait impulsivity (Bruce, Davis, &
Gunnar, 2002; King, Jones, Scheuer, Curtis, & Zarcone, 1990;
Moss, Vanyukov, & Martin, 1995), and the possibility exists that
cortisol plays a causal role in reducing impulsive behavior. The
relations among cortisol, impulsivity, and generation of responses
on cognitive tasks must therefore be further studied. In addition, it
cannot be stated definitively that effects of cortisol on consolida-
tion processes accounted for the recognition memory effects at
Session 2. To precisely determine the role of cortisol in memory
consolidation in humans, studies must vary the timing of cortisol
administration with respect to encoding of stimuli, including a
study condition of cortisol administration soon after encoding.
Likewise, studies should be designed to examine the relation
between cortisol and the rate of forgetting (e.g., using “savings
scores”) to more fully investigate cortisol’s role in learning and
memory consolidation.
Another limitation of the current study is that the inference can
only be made tentatively that results depend on glucocorticoid
activity in the brain. For instance, glucocorticoids affect biosyn-
thesis of catecholamines in the adrenal medulla (Axelrod & Rei-
sine, 1984). Thus, the results of the current study may be a result
of glucocorticoid interactions with noradrenergic mechanisms, and
only secondarily related to cortisol levels. Future studies of the
relation between cortisol and memory should study interactions
with noradrenergic mechanisms. Furthermore, it remains to be
validated whether emotional stimuli such as the ones used in this
study can be used to test hypotheses about the effects of glucocor-
ticoids in the amygdala’s role in memory. Neuroimaging studies
should be conducted to test whether activation in the hippocampus,
amygdala, and other brain areas is related to glucocorticoid-
induced variation in memory. Furthermore, glucocorticoids inhibit
transport of glucose into hippocampal neurons and glia, which is a
mechanism implicated in certain glucocorticoid actions (Horner,
Packan, & Sapolsky, 1990). However, this effect is dose- and
time-dependent, with sustained (4 hr) elevations needed for inhi-
bition, and is limited by abundant circulating glucose (Horner et
al., 1990; Virgin et al., 1991). Thus, it is unlikely that the single
administration of cortisol in the current study significantly affected
glucose transport. In addition, participants were instructed to eat a
light dinner 1 hr before Session 1, and were therefore likely in a
variety of postprandial states, which would have limited any
glucose-mediated effects of cortisol administration.
Another limitation of the current study was that several effects
of stimulus set (Sets A and B) were apparent, that is, the two sets
did not produce exactly the same pattern of results. Fortunately,
none of the main effects or interactions of hydrocortisone dose
were ever called into question by interactions with set. Word sets
were also limited in that the negative words appeared to have a
higher degree of semantic relatedness than did the neutral words,
causing a spurious reduction in recognition memory performance
for negative, compared with neutral, words. It is clear that the
careful matching of word sets for multiple psychometric properties
in this study was not sufficient. Future studies will need to more
extensively match word sets for semantic relatedness and, possi-
bly, other factors in addition to the ones used in this study.
Furthermore, two classes of stimuli (words and pictures) and
two types of memory (free-recall and recognition memory) were
studied, and corrections for multiple comparisons were not ap-
plied. However, some of the memory effects reported here would
have survived a more conservative alpha level of .01, and the basic
conclusions about emotional memory would not be altered. A
benefit of having multiple memory tests is that the different types
and timing of memory tests revealed diverse results, providing
strong evidence that cortisol differentially affects various aspects
of memory functioning. The general consistency of results for
pictures and words suggests that these results are generalizable and
not due to particulars of a single type of stimuli.
An additional limitation of the current study is that only male
participants were included. Evidence now exists suggesting that
the effects of cortisol on memory differ for men and women (Wolf,
514
ABERCROMBIE ET AL.

Schommer, Hellhammer, McEwen, & Kirschbaum, 2001). Thus,
future studies must systematically determine whether sex interacts
with the various effects of glucocorticoids on memory.
Summary
The current study provides further evidence of specific effects
of cortisol on explicit memory processes in humans. The result of
an inverted-U-shaped function for Session 2 recognition memory
was consistent with the neurobiological model of glucocorticoid
effects on corticosteroid receptors. The current study revealed
similar results for emotional and neutral stimuli, suggesting that
effects of cortisol on memory do not differ substantially for emo-
tional and neutral information. However, findings also suggest that
the decline in memory performance with increasing cortisol levels
may be more reliable for negative, compared with neutral, infor-
mation. Differences in the dose–response curves for emotionally
arousing and neutral information have been implicated by prior
research showing memory facilitation only for emotionally arous-
ing information (Buchanan & Lovallo, 2001). Memory results
across studies may be explained by a slightly increased signal-to-
noise ratio and/or a subtle leftward shift in the dose–response
relation between glucocorticoids and explicit memory for emo-
tional, compared to neutral, information. Future research is needed
to specifically address these hypotheses.
References
Adolphs, R., Cahill, L., Schul, R., & Babinsky, R., (1997). Impaired
declarative memory for emotional material following bilateral amygdala
damage in humans. Learning & Memory, 4, 291–300.
Axelrod, J., & Reisine, T. D. (1984, May 4). Stress hormones: Their
interaction and regulation. Science, 224, 452–459.
Bechara, A., Tranel, D., Damasio, H., Adolphs, R., Rockland, C., &
Damasio, A. (1995, August 25). Double dissociation of conditioning and
declarative knowledge relative to the amygdala and hippocampus in
humans. Science, 269, 1115–1118.
Beckwith, B. E., Petros, T. V., Scaglione, C., & Nelson, J. (1986). Dose-
dependent effects of hydrocortisone on memory in human males. Phys-
iology & Behavior, 36, 283–286.
Bradbury, M. J., Akana, S. F., & Dallman, M. F. (1994). Roles of type I
and II corticosteroid receptors in regulation of basal activity in the
hypothalamo-pituitary-adrenal axis during the diurnal trough and the
peak: Evidence for a nonadditive effect of combined receptor occupa-
tion. Endocrinology, 134, 1286–1296.
Bradley, M. M., Greenwald, M. K., Petry, M. C., & Lang, P. J. (1992).
Remembering pictures: Pleasure and arousal in memory. Journal of Exper-
imental Psychology: Learning, Memory, and Cognition, 18, 379–390.
Bradley, M. M., & Lang, P. J. (1999). Affective norms for English words
(ANEW). Gainesville: The NIMH Center for the Study of Emotion and
Attention, University of Florida.
Brown, J. (1976). An analysis of recognition and recall and of problems in
their comparison. In J. Brown (Ed.), Recall and recognition (pp. 1–34).
New York: Wiley.
Bruce, J., Davis, E. P., & Gunnar, M. R. (2002). Individual differences in
children’s cortisol response to the beginning of a new school year.
Psychoneuroendocrinology, 27(6), 635–650.
Buchanan, T. W., & Lovallo, W. R. (2001). Enhanced memory for emo-
tional material following stress-level cortisol treatment in humans. Psy-
choneuroendocrinology, 26, 307–317.
Cahill, L., Haier, R. J., Fallon, J., Alkire, M., Tang, C., Keator, D., et al.
(1996). Amygdala activity at encoding correlated with long-term, free
recall of emotional information. Proceedings of the National Academy of
Science, USA, 93, 8016–8021.
Cahill, L., & McGaugh, J. L. (1995). A novel demonstration of enhanced
memory associated with emotional arousal. Consciousness and Cogni-
tion, 4, 410–421.
Canli, T., Zhao, Z., Brewer, J., Gabrieli, J. D. E., & Cahill, L. (2000).
Event-related activation in the human amygdala associates with later
memory for individual emotional experience. Journal of Neuroscience,
20, RC99.
Carroll, J. B., Davies, P., & Richman, B. (1971). Word frequency book.
New York: American Heritage Publishing Company.
de Kloet, E. R., Oitzl, M. S., & Joels, M. (1999). Stress and cognition: Are
corticosteroids good guys or bad guys? Trends in Neuroscience, 22,
422–426.
de Quervain, D. J. F., Roozendaal, B., Nitsch, R. M., McGaugh, J. L., &
Hock, C. (2000). Acute cortisone administration impairs retrieval of
long-term declarative memory in humans. Nature Neuroscience, 3, 313–
314.
Fehm-Wolfsdorf, G., Reutter, K., Zenz, H., Born, J., & Lorenz, H. (1993).
Are circadian variations in taste thresholds cortisol-dependent? Journal
of Psychophysiology, 7,65–72.
Feldman Barrett, L., & Russell, J. A. (1998). Independence and bipolarity
in the structure of current affect. Journal of Personality and Social
Psychology, 74, 967–984.
Filipini, D., Gijsbers, K., Birmingham, M. K., & Dubrovsky, B. (1991).
Effects of adrenal steroids and their reduced metabolites on hippocampal
long-term potentiation. Journal of Steroid Biochemistry and Molecular
Biology, 40,87–92.
Hamann, S. B., Ely, T. D., Grafton, S. T., & Kilts, C. D. (1999). Amygdala
activity related to enhanced memory for pleasant and aversive stimuli.
Nature Neuroscience, 2, 289–293.
Heuer, F., & Reisberg, D. (1990). Vivid memories of emotional events:
The accuracy of remembered minutiae. Memory & Cognition, 18, 496–
506.
Horner, H. C., Packan, D. R., & Sapolsky, R. M. (1990). Glucocorticoids
inhibit glucose transport in cultured hippocampal neurons and glia.
Neuroendocrinology, 52, 57–64.
Hubert, W., & de Jong-Meyer, R. (1991). Psychophysiological response
patterns to positive and negative film stimuli. Biological Psychology, 31,
73–93.
Keller-Wood, M. E., & Dallman, M. F. (1984). Corticosteroid inhibition of
ACTH. Endocrine Reviews, 5, 1–24.
Keppel, G. (1991). Design and analysis: A researcher’s handbook (3rd
ed.). Englewood Cliffs, NJ: Prentice-Hall.
King, R. J., Jones, J., Scheuer, J. W., Curtis, D., & Zarcone, V. P. (1990).
Plasma cortisol correlates of impulsivity and substance abuse. Person-
ality and Individual Differences, 11, 287–291.
Kirschbaum, C., & Hellhammer, D. H. (1989). Salivary cortisol in psy-
chobiological research: An overview. Neuropsychobiology, 22, 150–
169.
Kirschbaum, C., & Hellhammer, D. H. (1994). Salivary cortisol in psy-
choneuroendocrine research: Recent developments and applications.
Psychoneuroendocrinology, 19, 313–333.
Kirschbaum, C., Pirke, K., & Hellhammer, D. H. (1993). The ‘Trier Social
Stress Test’—A tool for investigating psychobiological stress responses
in a laboratory setting. Neuropsychobiology, 28, 76–81.
Kirschbaum, C., Wolf, O. T., May, M., Wippich, W., & Hellhammer, D. H.
(1996). Stress- and treatment-induced elevations of cortisol levels asso-
ciated with impaired declarative memory in healthy adults. Life Sci-
ences, 58, 1475–1483.
LaBar, K. S., LeDoux, J. E., Spencer, D. D., & Phelps, E. A. (1995).
Impaired fear conditioning following unilateral temporal lobectomy in
humans. Journal of Neuroscience, 15, 6846–6855.
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1998). International affective
515
CORTISOL AND MEMORY FOR EMOTIONAL INFORMATION

picture system (IAPS): Technical manual and affective ratings. Gainesville:
The Center for Research in Psychophysiology, University of Florida.
LeDoux, J. E., Cicchetti, P., Xagoraris, A., & Romanski, L. M. (1990). The
lateral amygdaloid nucleus: Sensory interface of the amygdala in fear
conditioning. Journal of Neuroscience, 10, 1062–1069.
Lupien, S. J., Gillin, C. J., & Hauger, R. L. (1999). Working memory is
more sensitive than declarative memory to the acute effects of cortico-
steroids: A dose-response study in humans. Behavioral Neuroscience,
113, 420–430.
Lupien, S. J., & Lepage, M. (2001). Stress, memory, and the hippocampus:
Can’t live with it, can’t live without it. Behavioural Brain Research,
127, 137–158.
Lupien, S. J., & McEwen, B. S. (1997). The acute effects of corticosteroids
on cognition: Integration of animal and human model studies. Brain
Research Reviews, 24,1–27.
Lupien, S. J., Wilkinson, C. W., Briere, S., Menard, C., Ng King Kin,
N. M. K., & Nair, N. P. V. (2002). The modulatory effects of cortico-
steroids on cognition: Studies in young human populations. Psychoneu-
roendocrinology, 27, 401–416.
McEwen, B. S. (1982). Glucocorticoids and hippocampus: Receptors in
search of a function. In D. Ganten & D. Pfaff (Eds.), Adrenal action on
brain (pp. 1–22). New York: Springer-Verlag.
McEwen, B. S., & Sapolsky, R. M. (1995). Stress and cognitive function.
Current Opinion in Neurobiology, 5, 205–216.
McFarlane, A. C., Atchison, M., & Yehuda, R. (1997). The acute stress
response following motor vehicle accidents and its relations to PTSD. In
R. Yehuda & A. C. McFarlane (Eds.), Annals of the New York Academy
of Sciences: Vol. 821. Psychobiology of posttraumatic stress disorder
(pp. 437–441). New York: New York Academy of Sciences.
Moss, H. B., Vanyukov, M. M., & Martin, C. S. (1995). Salivary cortisol
responses and the risk for substance abuse in prepubertal boys. Biolog-
ical Psychiatry, 38(8), 547–555.
Newcomer, J. W., Selke, G., Melson, A. K., Hershey, T., Craft, S.,
Richards, K., & Alderson, A. L. (1999). Decreased memory performance
in healthy humans induced by stress-level cortisol treatment. Archives of
General Psychiatry, 56, 527–533.
Nuechterlein, K. H., & Asarnow, R. F. (1999). Degraded stimulus contin-
uous performance test (DS-CPT): Program for IBM-compatible micro-
computers [Computer software]. Los Angeles: Authors.
O’Carroll, R. E., Drysdale, E., Cahill, L., Shajahan, P., & Ebmeier, K. P.
(1999). Stimulation of the noradrenergic system enhances and blockade
reduces memory for emotional material in man. Psychological Medicine,
29, 1083–1088.
Oitzl, M. S., & de Kloet, E. R. (1992). Selective corticosteroid antagonists
modulate specific aspects of spatial orientation learning. Behavioral
Neuroscience, 106,62–71.
Patel, P. D., Lopez, J. F., Lyons, D. M., Burke, S., Wallace, M., &
Schatzberg, A. F. (2000). Glucocorticoid and mineralocorticoid receptor
mRNA expression in squirrel monkey brain. Journal of Psychiatric
Research, 34, 383–392.
Phelps, E. A., LaBar, K. S., & Spencer, D. D. (1997). Memory for
emotional words following unilateral temporal lobectomy. Brain and
Cognition, 35,85–109.
Plihal, W., Krug, R., Pietrowsky, R., Fehm, H. L., & Born, J. (1996).
Corticosteroid receptor mediated effects on mood in humans. Psycho-
neuroendocrinology, 21, 515–523.
Pugh, C. R., Fleshner, M., & Rudy, J. W. (1997). Type II glucocorticoid
receptor antagonists impair contextual but not auditory-cue fear condition-
ing in juvenile rats. Neurobiology of Learning and Memory, 67,75–79.
Resnick, H. S., Yehuda, R., Pitman, R. K., & Foy, D. W. (1995). Effect of
previous trauma on acute plasma cortisol level following rape. American
Journal of Psychiatry, 152, 1675–1677.
Roozendaal, B. (2000). Glucocorticoids and the regulation of memory
consolidation. Psychoneuroendocrinology, 25, 213–238.
Roozendaal, B., Bohus, B., & McGaugh, J. L. (1996). Dose-dependent
suppression of adrenocortical activity with metyrapone: Effects on emo-
tion and memory. Psychoneuroendocrinology, 21, 681–693.
Roozendaal, B., & McGaugh, J. L. (1996). Amygdaloid nuclei lesions differ-
entially affect glucocorticoid-induced memory enhancement in an inhibitory
avoidance task. Neurobiology of Learning and Memory, 65, 1–8.
Roozendaal, B., & McGaugh, J. L. (1997). Glucocorticoid receptor agonist
and antagonist administration into the basolateral but not central amyg-
dala modulates memory storage. Neurobiology of Learning and Mem-
ory, 67, 176–179.
Sanchez, M. M., Young, L. J., Plotsky, P. M., & Insel, T. R. (2000).
Distribution of corticosteroid receptors in the rhesus brain: Relative
absence of glucocorticoid receptors in the hippocampal formation. Jour-
nal of Neuroscience, 20, 4657–4668.
Schmidt, L. A., Fox, N. A., Goldberg, M. C., Smith, C. C., & Schulkin, J.
(1999). Effects of acute prednisone administration on memory, attention
and emotion in healthy human adults. Psychoneuroendocrinology, 24,
461–483.
Schooler, J. W., & Engstler-Schooler, T. Y. (1990). Verbal overshadowing
of visual memories: Some things are better left unsaid. Cognitive Psy-
chology, 22,36–71.
Smider, N. A., Essex, M. J., Kalin, N. H., Buss, K. A., Klein, M. H.,
Davidson, R. J., & Goldsmith, H. H. (2002). Salivary cortisol as a
predictor of socioemotional adjustment during kindergarten: A prospec-
tive study. Child Development, 73(1), 75–92.
Snodgrass, J. G., & Corwin, J. (1988). Pragmatics of measuring recognition
memory: Applications to dementia and amnesia. Journal of Experimen-
tal Psychology: General, 117,34–50.
Virgin, C. E., Jr., Ha, T. P., Packan, D. R., Tombaugh, G. C., Yang, S. H.,
Horner, H. C., & Sapolsky, R. M. (1991). Glucocorticoids inhibit glu-
cose transport and glutamate uptake in hippocampal astrocytes: Impli-
cations for glucocorticoid neurotoxicity. Journal of Neurochemistry, 57,
1422–1428.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and vali-
dation of brief measures of positive and negative affect: The PANAS
Scales. Journal of Personality and Social Psychology, 54, 1063–1070.
Wolf, O. T., Schommer, N. C., Hellhammer, D. H., McEwen, B. S., &
Kirschbaum, C. (2001). The relationship between stress induced cortisol
levels and memory differs between men and women. Psychoneuroen-
docrinology, 26, 711–720.
Wolkowitz, O. M., Reus, V. I., Weingartner, H., Thompson, K., Breier, A.,
Doran, A., et al. (1990). Cognitive effects of corticosteroids. American
Journal of Psychiatry, 147, 1297–1303.
Yehuda, R., Southwick, S. M., Giller, E. L., Ma, X., & Mason, J. W.
(1992). Urinary catecholamine excretion and severity of PTSD symp-
toms in Vietnam combat veterans. Journal of Nervous and Mental
Disease, 180, 321–325.
Received May 13, 2002
Revision received September 16, 2002
Accepted October 21, 2002 䡲
516
ABERCROMBIE ET AL.