Acute Psychological Stress Reduces Working Memory-
Related Activity in the Dorsolateral Prefrontal Cortex
Shaozheng Qin, Erno J. Hermans, Hein J.F. van Marle, Jing Luo, and Guillén Fernández
Background: Acute psychological stress impairs higher-order cognitive function such as working memory (WM). Similar impairments are
seen in various psychiatric disorders that are associated with higher susceptibility to stress and with prefrontal cortical dysfunctions,
suggesting that acute stress may play a potential role in such dysfunctions. However, it remains unknown whether acute stress has
immediate effects on WM-related prefrontal activity.
Methods: Using functional magnetic resonance imaging (fMRI), we investigated neural activity of 27 healthy female participants during a
blocked WM task (numerical N-back) while moderate psychological stress was induced by viewing strongly aversive (vs. neutral) movie
material together with a self-referencing instruction. To assess stress manipulation, autonomic and endocrine, as well as subjective,
measurements were acquired throughout the experiment.
Results: Successfully induced acute stress resulted in significantly reduced WM-related activity in the dorsolateral prefrontal cortex
(DLPFC), and was accompanied by less deactivation in brain regions that are jointly referred to as the default mode network.
factor may play an important role in higher-order cognitive deficits and hypofrontality observed in various psychiatric disorders.
Key Words: Dorsolateral prefrontal cortex, fMRI, psychological
stress, working memory
various psychiatric disorders that are associated with higher
susceptibility to stress and prefrontal dysfunction (5–9). This
implicates acute stress as a potential mediating factor in symp-
toms of higher-order cognitive dysfunction. However, little is
known about the immediate effects of acute stress on WM-
related prefrontal function in humans.
WM refers to a system maintaining relevant information in a
temporary buffer that is constantly updated to guide behavior
and is well-known to be supported by a frontoparietal network
(10). Exposure to acute stress leads to rapid activation of the
sympathetic nervous system (SNS), accompanied by the release
of norepinephrine (NE) from a widely distributed brain network
of synapses including abundant projections to the prefrontal
cortex (PFC; 1–4,7). Acute stress also results in rapid activation of
the prefrontal dopamine (DA) system (11). Hence, detrimental
effects of acute stress on WM are thought to result from supraop-
timal levels of catecholamines in the PFC. Empirical evidence
from multiple pharmacological studies in nonhuman primates
has revealed that catecholamines exert an inverted U-shaped
xposure to acute stress impairs higher-order cognitive
function as exemplified by impairment of working mem-
ory (WM) (1–4). Similar impairments are observed in
influence on prefrontal cognitive function in which sub- or
supraoptimal levels weaken WM processing (7,8), and high
doses of catecholamines are indeed associated with decreased
neuronal firing in the dorsolateral PFC (DLPFC) (2,12,13). On a
slightly longer time scale, acute stress results in activation of the
hypothalamic-pituitary-adrenal (HPA) axis, which regulates the
release of glucocorticoids (1,14,15). Animal studies have demon-
strated detrimental effects of glucocorticoids on WM, but only in
the presence of concomitant arousal-related noradrenergic acti-
vation (16), which in turn appears to be dependent on the
amygdala (17). In humans, similar detrimental effects of glu-
cocorticoids on WM have been shown to be limited to a time
window during which the SNS and the HPA axis are synergisti-
cally activated and do not persist after SNS recovery while
glucocorticoid levels are still elevated (3,4). In sum, the HPA axis
appears to exacerbate detrimental effects of supraoptimal levels
of catecholamines on PFC functioning. From this neurobiological
account, we therefore predicted that acute stress would lead to
attenuated WM-related DLPFC activity.
In addition, activation in WM-related frontoparietal executive
function networks is consistently accompanied by deactivation in
a set of brain regions referred to as the default mode network
(DMN) (18,19). Performing a WM task while coping with an
acutely stressful situation can be considered a form of continuous
dual processing: acute stress may result in more difficulty inhib-
iting stress-related task-irrelevant internal thoughts (4,20–22) and
therefore lead to alterations in, and reallocation of, attentional
resources. Because the DMN and the frontoparietal executive
network are known to exhibit reciprocal activity (23,24), we
conjectured that acute stress induction could lead to redistribu-
tion of neural resources away from executive functioning net-
works and toward the DMN.
To address these issues, we used blocked-design functional
magnetic resonance imaging (fMRI) to investigate how experi-
mentally induced stress modulates neural activity during a nu-
merical N-back task. Moderate psychological stress was induced
using strongly aversive (vs. neutral) movie material with a
Nijmegen Medical Center, Nijmegen, The Netherlands; and Key Labora-
tory of Mental Health (JL), Institute of Psychology, Chinese Academy of
Sciences, Beijing, China.
Address reprint requests to Shaozheng Qin, M.S., Centre for Cognitive Neu-
roimaging, Donders Institute for Brain, Cognition and Behaviour, Rad-
erlands; E-mail: firstname.lastname@example.org.
Received November 5, 2008; revised February 27, 2009; accepted March 2,
BIOL PSYCHIATRY 2009;66:25–32
© 2009 Society of Biological Psychiatry
self-referencing instruction. Participants were trained extensively
on the WM task before scanning to minimize interindividual
variability and reduce practice effects. To assess the effects of
stress induction on the SNS and HPA axis activation, heart rate
(HR) was continuously recorded throughout scanning, and sali-
vary cortisol samples were collected at baseline and at various
time delays. We predicted that acute stress would reduce WM-
related DLPFC activity, potentially in combination with less
deactivation of the DMN.
Methods and Materials
Twenty-nine young, healthy, right-handed female university
students (aged 18–25 years) with normal or corrected-to-normal
vision participated in this study. Participants reported no history
of neurological, psychiatric, or endocrine disease; no current use
of any psychoactive drugs or corticosteroids; and no habit of
watching violent movies or playing violent video games. None of
them had experienced severe physical or emotional trauma.
Avoiding confounds related to gender differences and menstrual
cycle–dependent variance in stress responsiveness (25,26), only
women taking standard single-phase oral contraceptives were
included. They were tested in the final 2 weeks of their cycle to
ensure stable hormone levels. Data from two participants were ex-
cluded because of technical failure and failure to complete the
experiment. Written informed consent was obtained before the
experiment in accordance with local ethical board requirements.
Participants were tested in a mixed-factorial design with stress
induction as between-subject factor and WM-load (0- vs. 2-back)
as within-subject factor. They were randomly assigned to either
the stress induction (n ? 14; aged 21 ? 2.1 years) or the control
group (n ? 13; aged 20 ? 1.8 years).
The experiment was carried out between 2 and 7 PM to ensure
relatively stable and low levels of endogenous cortisol. After
arrival, 1.5 hours before scanning, participants trained on the
WM task extensively and completed various questionnaires.
Baseline measurements of cortisol and subjective affect (positive
and negative affect scales [PANAS]; 27) were obtained. After this,
participants were told to which of the two experimental groups
they were randomly assigned. The actual fMRI experiment
consisted of four short movie clips to ensure that tasks of interest
were fully embedded in a continuously stressful (or neutral
control) context; it ended with a structural scan. Between the
second and third movie clips, participants performed the nu-
meric N-back task (Figure 1).
In the stress-induction group, acute psychological stress was
induced by showing short movie clips in the MRI scanner
containing scenes with extremely aversive content (extreme
male to male and female violence), selected from a commercial
movie (Irreversible, 2002, by Gaspar Noé). In the control group,
Salivary cortisol (nmol/l)
Negative affect score
4 2 1 3 2 9 6 5 6
Figure 1. Experimental design and subjective, endocrine, and autonomic
measurements of stress. Experimental design: the experiment started with
facial expressions (12 min), a second movie clip (M2: 1.30 min), the N-back
task (13.60 min), a third movie clip (M3: 1.30 min), and other tasks (30 min);
subjective (positive and negative affective scale [PANAS]), endocrine (corti-
sol), and autonomic (heart rate [HR], HR variability [HRV]) measurements of
stress were acquired throughout the experiment. (A) The digit sequence in
the rectangular box is an example of the 0- and 2-back conditions in the
N-back task (see Methods and Materials for more details). (B and C) Aver-
aged and baseline-corrected negative affect ratings and free salivary corti-
sol at different time points for the two groups: four PANAS measurements
coinciding with five salivary samples were acquired (i.e., two baseline sali-
baseline-corrected HR and HRV during the N-back task and its surrounding
movie clips (M2 and M3) for the stress and the control groups. Control,
control group; Stress, stress group. *p ? .05; **p ? .01; ***p ? 0. 001.
26 BIOL PSYCHIATRY 2009;66:25–32
S. Qin et al.
participants watched equally long movie clips from another
movie (Comment j’ai tué mon père, 2001, by Anne Fontaine),
which was equalized in luminance and similar in language and
human presence to the stress-induction film but contained only
nonarousing scenes. After short introductory texts, participants
were asked to watch the movies attentively and imagine them-
selves in the scene from an eyewitness perspective, thereby
attempting to involve them maximally in the movie.
The present stress induction method closely corresponds with
the determinants of human stress response described by Mason
(28)—that is, unpredictability, novelty, and uncontrollability. It
also meets the criteria for stress impaired WM to occur—that is,
close proximity of stressor and task to ensure concurrent (nor)
adrenergic activity (3).
Using a blocked-design, participants completed 10 cycles of
alternating 0- and 2-back conditions interleaved by a jittered
resting-fixation baseline ranging from 8 to 12 sec (average 10
sec). Within each block, a random digit sequence consisting of 15
single digits was shown to participants (see Figure 1A). Each digit
was presented for 400 msec, followed by an inter-stimulus-
interval of 1400 msec. Each block lasted 27 sec, and started with
a 2-sec cue presentation indicating the 0- or 2-back condition.
During the 0-back condition, participants were asked to detect
whether the current item on the screen was a “1” or not. During
the 2-back condition, participants were asked to detect whether
the current item had appeared two positions back in the se-
quence. Participants were instructed to make a button press with
their index finger when detecting a target. Before fMRI scanning,
they were extensively trained in performing the task (i.e., 10
cycles of alternating 0- and 2-back conditions) to minimize
interindividual variability and reduce practice effects. Data from
the last four training cycles served as prestress induction baseline
Subjective and Physiological Measurements of Stress
Subjective mood was assessed using the PANAS at baseline
and three additional time points coinciding with collection of
salivary samples (see Figure 1B and 1C). To monitor the HPA axis
response, saliva samples were collected using salivette collection
devices (Sarstedt, Rommelsdorf, Germany). Participants were
requested to abstain from eating, drinking, or smoking for 1 hour
before arrival. Salivary sampling consisted of two baseline mea-
surements (before MRI scanning) and three additional ones (right
before the N-back task, right after the last movie clip, and 20 min
after leaving scanner). All samples were stored at ?20°C until
analysis. Samples were prepared for biochemical analysis by
centrifuging at 3000 rpm for 5 min, which resulted in a clear
supernatant of low viscosity. Salivary-free cortisol concentrations
were determined employing a chemiluminescence assay (CLIA)
with high sensitivity of .16 ng/mL (IBL, Hamburg, Germany).
To assess the autonomic nervous system response, HR was
recorded continuously throughout MRI scanning using an MR-
compatible pulse oximeter attached to the left index finger.
Offline analysis included calculation of both HR frequency and
HR variability (HRV; calculated as the root mean square of
successive differences [rMSSD], an index of respiratory sinus
arrhythmia) (29). Data from two participants were excluded from
this analysis because of excessive artifacts (one in the stress
group). Additionally, eye tracking was performed using an
MR-compatible eye-tracking device (MEye Track-LR camera unit,
SensoMotoric Instruments, Teltow, Germany) to confirm atten-
tive viewing of the movie clips.
fMRI Data Acquisition
During MRI scanning, whole brain T2*-weighted echo planar
imaging based on blood oxygenation level-dependent contrast
(EPI-BOLD) fMRI data were acquired with a Siemens Trio 3.0-T
MR-scanner (Erlangen, Germany) using an ascending slice acqui-
sition sequence (37 axial slices, volume repetition time [TR] ?
2.18 sec, echo time [TE] ? 25 msec, 80° flip angle, slice matrix
size ? 64 ? 64, slice-thickness ? 3.0 mm, slice gap ? .3 mm,
field of view [FOV] 212 ? 212 mm). Three hundred seventy-six
volumes were acquired during the N-back task. High-resolution
structural images (1 ? 1 ? 1 mm) were acquired using a
T1-weighted three dimensional magnetization-prepared rapid
gradient-echo (MP-RAGE) sequence (TR 2.3 sec, TE 2.96 msec, 8°
flip-angle, 192 contiguous sagittal slices, slice matrix size 256 ?
256, FOV 256 ? 256 mm), and Siemens’ integrated parallel
acquisition technique (iPAT) in conjunction with generalized
autocalibrating partially parallel acquisitions (GRAPPA) recon-
struction (factor two accelerated) (30).
fMRI Data Analysis
Image preprocessing and statistical analysis was performed
using SPM5 (http://www.fil.ion.ucl.ac.uk/spm). The first five EPI
volumes were discarded to allow for T1 equilibration. Remaining
functional images were rigid-body motion corrected and the
mean image was coregistered to each participant’s T1-weighted
MR-image. Subsequently, images were transformed into a com-
mon stereotactic space (MNI152 T1-template), and resampled
into 2 mm isotropic voxels. Finally, images were spatially
smoothed by convolving with an isotropic 3D-Gaussian kernel
(8-mm full width at half maximum). The data were statistically
analyzed using general linear models and statistical parametric
To assess neural activity associated with 0- and 2-back condi-
tions, the two conditions were modeled separately as boxcar
regressors and convolved with the canonical hemodynamic re-
sponse function in SPM5. Additionally, realignment parameters
were included to account for movement-related variability. The
analysis furthermore included high-pass filtering using a cutoff of
1/128 Hz, global intensity normalization, and serial correlations
correction using a first-order autoregressive (or AR) model.
The contrast parameter images for both conditions relative to
baseline, which were generated at the single-subject level, were
submitted to a second-level analysis within a 2 (group) by 2
(WM-load) mixed factorial analysis of variance (ANOVA). We
used an alpha of .05 corrected for multiple comparisons based
suprathreshold cluster size statistics (32). The initial threshold for
this analysis was set at p ? .001, uncorrected, which was also
used for visualization of activations. Given our clear hypotheses
regarding the DLPFC, this region was additionally investigated
with a reduced search region consisting of a sphere (radius 20
mm) at coordinates reported in previous studies with similar
N-back tasks (33,34), using a small volume correction procedure
(SVC). The SVC procedure was also employed for brain regions
within the DMN at coordinates reported by Greicius and col-
leagues (35). Parameter estimates were extracted from those
regions to characterize the response patterns of 0- and 2-back
conditions of the two groups using MarsBar (36).
S. Qin et al.
BIOL PSYCHIATRY 2009;66:25–32 27
Subjective and Physiological Measurements of Stress
Subjective negative affect scores at different time points are
shown in Figure 1B for the two groups. A 2-by-3 ANOVA with
group as the between-subjects factor and time as the within-
subjects factor (three post-baseline time points) revealed signif-
icant main effects of group [F(1,25) ? 18.56, p ? .001] and time
[F(2,24) ? 12.31, p ? .001], and a significant interaction effect
[F(2,24) ? 7.56, p ? .003], indicating that stress induction
resulted in significantly increased negative affect.
Baseline-corrected salivary cortisol measures are shown in
Figure 1C. A 2 (group) by 3 (time: three post-baseline time
points) ANOVA revealed a significant downward pattern in
cortisol for both groups over time [F(1,25) ? 9.59, p ? .005], most
likely due to diurnal rhythm and stress anticipation, and a
significant interaction effect of group and time [F(2,24) ? 3.43, p
? .049]. Further testing revealed significantly higher cortisol
levels for the stress group than the control group at the time point
directly preceding the N-back task and surrounding movie clips
[t(15.8) ? 1.91, p ? .037 one-tailed].
Baseline-corrected HR and HRV were averaged separately for
the N-back task and surrounding movie clips (see Figure 1D and
1E). A 2 (group) by 3 (time: pre-, during-, and post-N-back task)
ANOVA was conducted separately for HR and HRV data. A
significant main effect of group was found for HR [F(1,23) ?
10.77, p ? .003] as well as HRV [F(1,23) ? 6.69, p ? .016], with
significantly increased HR, and decreased HRV, in the stress
group compared with the control group. The two groups did not
differ in either HR [t(24) ? .85, ns] or HRV [t(24) ? ?1.15, ns] at
Taken together, the results from subjective and physiological
measurements of stress consistently confirm that the N-back task
was indeed embedded in a stressful context for the stress group.
Two separate ANOVAs for accuracy and reaction times (RTs)
were conducted with session (prestress baseline vs. scanning)
and WM-load as within-subject factors and group as between-
subject factor. There were robust main effects of WM-load for
both accuracy and RTs [F(1,25) ? 25.197 and F(1,25) ? 36.672,
respectively, both p values ? .001]. We found no interaction
between WM-load and session (both F values ? 1), indicating
no significant change in the WM-load effect from prestress
baseline to scanning. Also, we found no three-way interaction
effect involving group, indicating no significant stress effect on
WM performance change. Additionally, two separate 2 (group)
by 2 (WM-load) ANOVAs were conducted for accuracy and RTs
specifically on the data acquired during scanning. Again, robust
main effects of WM-load on accuracy and RTs were found
[F(1,25) ? 31.572 and F(1,25) ? 40.097, respectively, both
p values ? .001]. Neither a main effect of group nor an interaction
effect was found (all F values ? 1; see Figure 2). Thus,
performance data show robust WM-load effects but no changes
in WM-load effects from prestress baseline to scanning and no
effects of stress induction on WM-load effects.
To investigate further whether stress-induced performance
decreases may have occurred in participants with a stronger
physiological stress response exclusively, we calculated correla-
tions between physiological stress measurements and changes in
performance from prestress baseline to scanning within the stress
group. Cortisol levels just before the N-back task (r ? .546, p ?
.043) and HR during the N-back task (r ? .649, p ? .016)
correlated positively with RT change (see Figure 1 in Supplement
1), showing that participants with the strongest stress response
slowed down most.
First, by contrasting 2- with 0-back conditions (collapsing
across groups), we replicated robust activations of a WM-
related network including the bilateral DLPFC (local maxima
at [36,48,18] and [?34,52,14], p ? .05, whole-brain family-wise
error [FWE] corrected), bilateral intraparietal cortex (local
maxima at [?44,?42,50] and [40,?44,46], p ? .05, whole-brain
FWE corrected), cerebellum (local maxima at [30,?58,?32]
and [?30,?58,?34], p ? .05, whole-brain FWE corrected), and
other related regions (see Table 1).
Reaction time (ms)
Figure 2. Behavioral performance in the N-back task. Mean accuracy
(? SEM) and mean RTs (? SEM) of 0- and 2-back conditions for the stress
induction and the control groups. Stress, stress group; Control, control
Table 1. Brain Activations Related to WM Load and Modulations of
Brain RegionsBAT Valuex y z
Main Effect of WM Load (2- vs. 0-Back, Collapsing Across Two Groups)
Superior/middle PFCR 6
28 4 58
?30 2 58
32 24 0
?28 24 4
36 48 18
?34 52 14
18 ?4 0
?16 ?4 0
40 –44 46
?44 ?42 50
18 ?66 60
?14 ?68 58
30 ?58 ?32
?30 ?58 ?34
Inferior parietal cortex
Superior parietal cortex
Interaction Effect Between WM Load and Group (2- vs. 0-Back ? Control
30 46 20
?36 48 8
Only clusters significant at p ? .05, corrected at cluster level, are re-
tal cortex; R, right; stress, stress group; WM, working memory.
ap ? .05, whole-brain corrected.
bp ? .05, small volume correction procedure.
28 BIOL PSYCHIATRY 2009;66:25–32
S. Qin et al.
More important for the question at issue, we found significant
clusters in the bilateral DLPFC (local maxima at [30,46,20] and
[?36,48,8], cluster p ? .05, SVC) when contrasting neural activity
related to 2- versus 0-back in the control group with that of the
stress group (Figure 3). In other words, there was an interaction
between WM-load (2- vs. 0-back) and group (control vs. stress),
indicating that WM-related DLPFC activation was significantly
reduced in participants exposed to stress induction compared
with participants in the control group. Subsequent whole-brain
regression analyses within the stress group, with physiological
measurements of stress as separate covariates, revealed no
significant clusters in WM-related structures. However, a more
specific region of interest analysis on averaged parameter esti-
mates from the cluster of voxels exhibiting a stress-induced
reduction effect in the left DLPFC did reveal a significant positive
correlation between HRV and left DLPFC activation (see Figure 2
in Supplement 1).
In addition, by contrasting the active task-demanding condi-
tions with fixation baseline (collapsing across groups), we rep-
licated earlier findings showing deactivation in the DMN includ-
ing the posterior cingulate cortex (local maxima at [6,?52,16] and
[?6,?54,16], p ? .05, whole-brain FWE corrected), the ventral
medial PFC extending into the orbitofrontal cortex (local maxima
at [0,58,2] and [10,52,?6], p ? .05, whole-brain FWE corrected;
see Table 2). Moreover, we found that stress induction led to
significantly less deactivation in regions within the DMN, more
specifically in the posterior cingulate cortex and the medial
aspect of orbitofrontal cortex (local maxima at [?4,?40,28] and
[12,46,?12] respectively; cluster p ? .05, SVC; see Figure 4).
Within the stress group, whole-brain regression analysis revealed
that cortisol levels correlated with activity in the medial PFC
extending into the anterior cingulate cortex (ACC; local maxima
at [?8,48,?2], cluster p ? .05, SVC; see Figure 3 in Supplement
1), indicating that participants with larger cortisol responses
exhibited less deactivation of this DMN subregion.
We aimed to investigate stress-induced modulations in WM-
related prefrontal activity. Results confirmed our hypothesis of
reduced WM-related activation in the DLPFC. This reduction was
accompanied by less deactivation of brain structures within the
DMN. As indicated by increased HR and decreased HRV, our
stress induction procedure resulted in a shift toward more sym-
pathetic, and less parasympathetic, autonomic nervous system
activity. Moreover, stress induction increased HPA axis activity as
measured from salivary cortisol. We therefore discuss elevations
of stress-sensitive catecholamines, which are associated with
increased sympathetic tonus, and cortisol as potential, but not
mutually exclusive, factors that may account for our observed
alterations in neural activity.
The PFC is sensitive to its neurochemical environment, and
small changes in catecholamine modulation of this region can
have substantial impact on higher-order cognitive function such
as WM (2,7,8). Exposure to acute stress is thought to result in
activation of the locus coeruleus (LC), which rapidly increases
NE projections to a widely distributed brain network (1,14). In
this way, the LC plays a critical role in promoting behavioral
adaptation to stressful situations (37–39). According to an inte-
grative theory of the LC-NE system in neuromodulation of
cognitive function (40,41), LC-NE activity exhibits an inverted
Table 2. Brain (De)Activations Related to the Default Mode Network and
Modulations of Stress Induction
Brain Regions BAT Valuex y z
Deactivations During Active Conditions (0- and 2-Back vs. Fixation
Baseline, Collapsing Across Groups)
Posterior cingulate cortexR 30/23
6 ?52 16
?6 ?54 16
10 52 ?6
0 58 2
26 ?18 ?18
?26 ?26 ?14
26 ?42 ?10
?28 ?40 ?10
42 ?16 0
?44 ?6 ?4
Ventral medial PFC
Main Effect of Group (Stress vs. Control, Collapsing Across 0- and 2-Back
Posterior cingulate cortex
?4 ?40 28
12 46 ?12
Only clusters significant at p ? .05, corrected at cluster level, are re-
tal cortex; R, right; stress, stress group.
ap ? .05, whole-brain corrected.
bp ? .05, small volume correction procedure.
Para. Esti. (a.u.)
Figure 3. Brain regions involved in working memory in general (shown in
blue, thresholded at p ? .05, whole-brain family-wise error corrected) and
(vs. control) group (coded in red, thresholded at p ? .001, uncorrected, for
spatially normalized and averaged (n ? 27) high-resolution T1-weighted
transversal view of activation in the right DLPFC (right panel; marked by
white circle). (B and C) Bar graphs representing parameter estimates of 0-
group; T, corresponding t values.
S. Qin et al.
BIOL PSYCHIATRY 2009;66:25–32 29
U-shaped relationship with outcome performance of goal-di-
rected behavior. Optimal performance is associated with an
intermediate level of LC activity and a strong phasic LC firing
pattern in response to a focused task. In contrast, both LC
hypoactivity and tonic hyperactivity lead to an impairment of
performance and a reduced phasic LC firing pattern. Moreover,
high tonic LC activity has been associated with a state of
hypervigilance and increased exploration of adaptive options. In
this study, stress induction may thus have led to high tonic LC
activity resulting in a right-sided shift on the inverted U-shaped
curve. Therefore, the reduction in WM-related DLPFC activity
may be explained by a shift from phasic to tonic activation of
ascending noradrenergic projections to the PFC.
Evidence from numerous pharmacological studies supports
the notion that NE, but also other stress-sensitive catecholamines
such as DA, exhibit inverted U-shaped dose–response relation-
ships with cognitive performance (7,8). On the cellular level, a
recent pharmacological study implementing intracellular record-
ings suggests that catecholamines indeed have such dose–
response relationships with neural firing patterns of the DLPFC
underlying WM (42). The existence of such an inverted U-shaped
pattern is further substantiated by dissociations of detrimental
and beneficial effects through distinct cellular mechanisms. At
optimal levels of NE, prefrontal function is strengthened through
actions of ?-2A-adrenoceptors and increasing neural firing via
inhibition of cAMP-HCN (cyclic adenosine monophosphate–
hyperpolarization-activated cyclic nucleotide-gated cation chan-
nel) signaling, whereas optimal levels of DA D1 receptor de-
crease task-irrelevant neural firing by increasing cAMP-HCN
signaling (8,13,42). In contrast, stress-induced excessive levels of
catecholamines impair WM-related prefrontal function by high
levels of cAMP-HCN signaling and high levels of NE engaging the
low-affinity ?-1-adrenoceptor, which suppresses the neural firing
pattern (7,8,12). In this study, the stress-induced shift in auto-
nomic nervous activity toward more sympathetic tonus impli-
cates strong engagement of the LC-NE system and therefore has
likely resulted in the observed reduction of WM-related DLPFC
In addition to this catecholaminergic mechanism, stress-
induced glucocorticoids are also known to target the PFC, where
corticosteroid receptors are abundantly expressed. On the be-
havioral level, studies in humans have shown that cortisol and
NE activation have additive effects in WM impairment and that
NE activation is a necessary condition for glucocorticoid effects
to occur (3,4). In contrast to the control group, which only
showed elevated SNS activity during the N-back task (see Figure
1D), most likely because of arousal related to performing the
task, our physiological stress measurements show that stress
induction indeed resulted in significant elevations of both HPA
axis and SNS activity. Moreover, some measures of stress corre-
lated with brain activation and WM performance changes within
the stress group. Therefore, it is plausible that elevated levels of
cortisol, in conjunction with high levels of catecholamines, play
a role in stress-related DLPFC hypoactivation. In animal studies,
such interactions between corticosteroids and NE have been
found in the basolateral amygdala, where glucocorticoids poten-
tiate noradrenergic actions (16,43). It is likely that similar inter-
actions also occur in the PFC, because cortisol blocks extraneu-
ronal catecholamine transporters that remove catecholamines
from the synaptic cleft (44). Future studies are required to address
such potential interactions in the PFC.
It is well known that the DMN and the frontal-parietal
executive network activate reciprocally: frontoparietal activation
has often been found to be accompanied by DMN suppression
(18,19,23,24). Our data robustly replicate earlier findings of DMN
deactivation during WM processing but also show stress-induced
reduction of DLPFC activation accompanied by less deactivation
in the DMN. Interestingly, recent studies suggest that the DMN is
involved in processing information unrelated to a current goal-
directed task, or “mind wandering” (45,46). Performance of a
WM task in the context of acute stress, which can be taken as a
form of dual processing, may result in difficulties inhibiting
task-irrelevant internal thoughts (4,20–22) such as intrusive
recollection of aversive content of the movie. These notions are
supported by our findings of reduced DMN suppression posi-
tively correlating with cortisol response in the stress group. Our
data converge with other findings showing stress-induced in-
creases in cerebral blood flow in a similar region (47). Thus,
stress-related psychological factors may lead to a reallocation of
neural resources away from a WM-related network and toward
On a broad functional level, such a stress-induced realloca-
tion of resources away from executive function networks may
represent a mechanism that is essential for survival. Alongside
rapid activation of autonomic and endocrine systems, excessive
catecholamines released during acute stress may take prefrontal
function “offline” to facilitate more adaptive and habitual re-
sponses like the “fight-or-flight” response (2,48), trading the
accuracy of slow, higher-order cognitive processing for speed.
Para. Esti. (a.u.)
Figure 4. Brain regions showing deactivation during general active task
conditions (coded in blue; thresholded at p ? .05, whole-brain family-wise
error corrected) and reduced deactivation in the stress (vs. control) group
(coded in red, thresholded at p ? .001, uncorrected, for visualization pur-
ized and averaged (n ? 27) high-resolution T1-weighted images. (A) Coro-
left panel; marked by white circle) and coronal view of deactivation in the
graphs representing parameter estimates of 0- and 2-back conditions for
the stress induction and the control groups in orbital PFC and posterior
brain regions coded in red. Control, control group; L, left; p, posterior; R,
right; stress, stress group; T, corresponding t values.
30 BIOL PSYCHIATRY 2009;66:25–32
S. Qin et al.
Despite its utility in life-threatening situations, such a mechanism
may exacerbate symptoms of prefrontal dysfunction in various
psychiatric disorders characterized by higher susceptibility to
stress, such as depression or schizophrenia. By showing similar
prefrontal cortical dysfunctions in healthy individuals under
acute stress, our data support the notion that there may be a
direct association between symptoms of hypofrontality and
heightened stress in these disorders (5,7–9). Future studies in
patients with these psychiatric disorders should take this factor
In conclusion, the stress-induced reduction of WM-related
activity in the DLPFC and the concomitant reduction of DMN
deactivation are most likely caused by supraoptimal elevations of
catecholamines (such as NE and DA), potentially in concert with
elevated levels of cortisol. Such a reallocation of neural resources
away from executive function networks may play an important
role in higher-order cognitive dysfunctions observed in many
psychiatric disorders, which lead to far-reaching problems for
patients and their social environment.
This work was supported by Grant No. 918.66.613 from the
Netherlands Organization for Scientific Research (NWO). We
thank the anonymous reviewers for insightful suggestions.
The authors reported no biomedical financial interests or
potential conflicts of interest.
Supplementary material cited in this article is available
1. de Kloet ER, Joëls M, Holsboer F (2005): Stress and the brain: From
adaptation to disease. Nat Rev Neurosci 6:463–475.
2. Arnsten AFT, Goldman-Rakic PS (1998): Noise stress impairs prefrontal
cortical cognitive function in monkeys: Evidence for a hyperdopamin-
3. Elzinga BM, Roelofs K (2005): Cortisol-induced impairments of working
memory require acute sympathetic activation. Behav Neurosci 119:98–
4. Schoofs D, Preuss D, Wolf OT (2008): Psychosocial stress induces work-
ing memory impairments in an n-back paradigm. Psychoneuroendocri-
5. Mazure CM (1995): Does stress cause psychiatric illness? In: Spiegel D,
editor. Progress in Psychiatry, vol 46. Washington, DC: American Psychi-
atric Publishing, 270.
6. Hammen C, Gitlin M (1997): Stress reactivity in bipolar patients and its
relation to prior history of disorder. Am J Psychiatry 154:856–857.
7. Arnsten AFT, Li BM (2005): Neurobiology of executive functions: Cate-
cholamine influences on prefrontal cortical functions. Biol Psychiatry
8. Arnsten AFT (2007): Catecholamine and second messenger influences
on prefrontal cortical networks of “representational knowledge”: A ra-
tional bridge between genetics and the symptoms of mental illness.
Cereb Cortex 17(suppl 1):i6–15.
9. Driesen NR, Leung HC, Calhoun VD, Constable RT, Gueorguieva R, Hoff-
man R, et al. (2008): Impairment of working memory and response in
schizophrenia: Functional magnetic resonance imaging evidence. Biol
10. Baddeley A (2003): Working memory: Looking back and looking for-
11. Deutch AY, Roth RH (1990): The determinants of stress-induced activa-
tion of the prefrontal cortical dopamine system. Prog Brain Res 85:367–
al. (2004): Protein kinase C overactivity impairs prefrontal cortical regu-
lation of working memory. Science 306:882–884.
Alpha2A-adrenoceptor stimulation strengthens working memory net-
works by inhibiting cAMP-HCN channel signaling in prefrontal cortex.
14. Joëls M, Pu Z, Wiegert O, Oitzl MS, Krugers HJ (2006): Learning under
stress: How does it work? Trends Cogn Sci 10:152–158.
of corticosteroids on short-term and long-term memory. Neurology 64:
16. Roozendaal B, Okuda S, De Quervain DJ, McGaugh JL (2006): Glucocor-
encing different memory functions. Neuroscience 138:901–910.
17. Goldstein LE, Rasmusson AM, Bunney BS, Rober HR (1996): Role of the
amygdala in the coordination of behavioral, neuroendocrine, and pre-
18. Hampson M, Driesen NR, Skudlarski P, Gore JC, Constable RT (2006):
19. Esposito F, Berolino A, Scarabino T, Latorre V, Blasti G, Popolizio T, et al.
tion: Assessing the impact of active thinking. Brain Res Bull 70:263–269.
21. Seibert PS, Ellis HC (1991): Irrelevant thoughts, emotional mood states,
and cognitive task performance. Mem Cogn 19:507–513.
ference predicts cognitive function in old age. Psychol Aging 21:535–
23. Selemon LD, Goldman-Rakic PS (1998): Common cortical and subcorti-
cal targets of the dorsolateral prefrontal and posterior parietal cortices
in the rhesus monkey: Evidence for a distributed neural network sub-
serving spatially guided behavior. J Neurol Sci 8:4049–4068.
24. Pomarol-Clotet E, Salvador R, Sarró S, Gomar J, Vila F, Martínez Á, et al.
(2008): Failure to deactivate in the prefrontal cortex in schizophrenia:
Dysfunction of the default mode network? Psychol Med 38:1185–1193.
25. Kirschbaum C, Kudielka BM, Gaab J, Schommer NC, Hellhammer DH
(1999): Impact of gender, menstrual cycle phase, and oral contracep-
26. Kudielka BM, Kirschbaum C (2005): Sex differences in HPA axis re-
sponses to stress: A review. Biol Psychol 69:113–132.
27. Watson D, Clark LA, Tellegen A (1988): Development and validation of
brief measures of positive and negative affect: The PANAS scales. J Pers
28. Mason JW (1968): A review of psychoendocrine research on the pitu-
itary-adrenal cortical system. Psychosom Med 30:576–607.
29. van Aken MO, Romijn JA, Miltenburg JA, Lentjes EG (2003): Automated
measurement of salivary cortisol. Clin Chem 49:1408–1409.
(2002): Parallel magnetic resonance imaging using the GRAPPA opera-
unified statistical approach for determining significant signals in im-
ages of cerebral activation. Hum Brain Mapp 4:58–73.
33. Callicott JH, Mattay VS, Bertolino A, Finn K, Coppola R, Frank JA, et al.
(1999): Physiological characteristics of capacity constraints in working
memory as revealed by functional MRI. Cereb Cortex 9:20–26.
34. Callicott JH, Bertolino A, Mattay VS, Langheim FJ, Duyn J, Coppola R, et
al. (2000): Physiological dysfunction of the dorsolateral prefrontal cor-
tex in schizophrenia revisited. Cereb Cortex 10:1078–1092.
35. Greicius MD, Krasnow B, Reiss AL, Menon V (2003): Functional connec-
tivity in the resting brain: A network analysis of the default mode hy-
36. Brett M, Anton JL, Valabregue R, Poline JB (2002): Region of interest
analysis using an SPM toolbox. Neuroimage 16:Abstract:497.
37. Sved AF, Cano G, Passerin AM, Rabin BS (2002): The locus coeruleus,
38. Berridge CW, Waterhouse BD (2003): The locus coeruleus-noradrener-
S. Qin et al.
BIOL PSYCHIATRY 2009;66:25–32 31
(2005): Role of brain norepinephrine in the behavioral response to
40. Aston-Jones G, Cohen JD (2005): An integrative theory of locus coer-
uleus-norepinephrine function: Adaptive gain and optimal perfor-
41. Aston-Jones G, Rajkowski J, Cohen J (1999): Role of locus coeruleus in
attention and behavioral flexibility. Biol Psychiatry 46:1309–1320.
Inverted-u dopamine D1 receptor actions on prefrontal neurons en-
gaged in working memory. Nat Neurosci 10:376–384.
43. Roozendaal B, Quirarte GL, McGaugh JL (2002): Glucocorticoids in-
teract with the basolateral amygdala ?-adrenoceptor–cAMP/cAMP/
PKA system in influencing memory consolidation. Eur J Neurosci 15:
identification of the corticosterone-sensitive extraneuronal catechol-
amine transporter. Nat Neurosci 1:349–351.
(2007): Wandering minds: The default network and stimulus-indepen-
dent thought. Science 315:393–395.
47. Wang J, Rao H, Wetmore GS, Furlan PM, Korczykowski M, Dinges DF, et
al. (2006): Perfusion functional MRI reveals cerebral blood flow pattern
48. Diamond DM, Campbell AM, Park CR, Halonen J, Zoladz PR (2007): The
temporal dynamics model of emotional memory processing: A synthesis
32 BIOL PSYCHIATRY 2009;66:25–32
S. Qin et al.