TECHNIQUES AND METHODS
Experience-Dependent Plasticity for Attention to
Threat: Behavioral and Neurophysiological Evidence in
Christopher S. Monk, Eric E. Nelson, Girma Woldehawariat, Lee Anne Montgomery, Eric Zarahn,
Erin B. McClure, Amanda E. Guyer, Ellen Leibenluft, Dennis S. Charney, Monique Ernst, and
Daniel S. Pine
Biased attention to threat represents a key feature of anxiety disorders. This bias is altered by therapeutic or stressful experiences,
suggesting that the bias is plastic. Charting on-line behavioral and neurophysiological changes in attention bias may generate insights
on the nature of such plasticity. We used an attention-orientation task with threat cues to examine how healthy individuals alter their
response over time to such cues. In Experiments 1 through 3, we established that healthy individuals demonstrate an increased
attention bias away from threat over time. For Experiment 3, we used functional magnetic resonance imaging to determine the neural
bases for this phenomenon. Gradually increasing attention bias away from threat is associated with increased activation in the
occipitotemporal cortex. Examination of plasticity of attention bias with individuals at risk for anxiety disorders may reveal how
threatening stimuli come to be categorized differently in this population over time.
Key Words: Attention, emotion, threat, fMRI, neuroimaging
threat (Bradley et al 1999; Fox et al 2001; Mogg and Bradley 2002).
Effective treatments for anxiety reduce the bias (Foa and McNally
1986; Mathews et al 1995), and training regimens that produce the
bias increase stress reactivity (MacLeod et al 2002), suggesting such
attention biases are plastic in nature. Changes in brain function are
thought to account for this plasticity. Whereas subcortical structures,
such as the amygdala, are implicated in rapid, invariant response to
threat (Hitchcock and Davis 1986; Pessoa et al 2002), posterior
cortical regions are crucial for more detailed processing that allows
identification and categorization of objects as safe or threatening
(Pine 2003). Categorization involves the occipitotemporal (OT)
pathway, a pathway that exhibits experience-based plasticity man-
changes in activation during priming (Henson et al 2000).
Neuroimaging studies emphasize the utility of paradigms that
relate psychiatric symptoms to both observable behaviors and
objective measures of brain function. Hence, for studies of
plasticity in attention bias, it is important to implement imaging
paradigms sensitive to plasticity in both neural and behavioral
domains. The current study shows that it is possible to capture
attention-related plasticity in both domains. The study uses a
variant of the Posner attention-orientation task (Posner et al
1980), wherein visual probes replace face emotion cues (Bradley
et al 1999; Fox et al 2001; MacLeod et al 1986; Mogg and Bradley
2002). Previous studies with variants of the task demonstrate
attention bias toward threat faces in anxious subjects and away
from threat faces in healthy subjects (Fox et al 2001). In the
ecent research has generated interest in the role of neural
plasticity in mediating associations between anxiety and
attention. Anxiety is associated with attention bias toward
present investigation, the opportunity to perceive the face was
diminished through a brief, masked presentation approach. With
repeated presentations, we anticipated that processing would
become facilitated; with facilitation, healthy subjects would come
to exhibit the expected attention bias away from threat, as
previously reported when stimuli presentations were not dimin-
ished. Thus, plasticity in healthy subjects was hypothesized to
produce facilitated processing, manifest as time-related increases
in attention bias away from threat and enhanced brain activation.
These changes are thought to reflect time-related shifts, whereby
cortical processes associated with categorization become in-
creasingly engaged over time in threat processing.
Methods and Materials
Twenty-one adults participated in Experiment 1 (9 females;
mean age ? 29.57 years; SD ? 3.09). Twenty-three adoles-
cents participated in Experiment 2 (14 females; mean age ?
12.87 years; SD ? 2.14). This study included adolescents to
evaluate the generalizability of the findings. Another sample
of 12 adults participated in Experiment 3 (6 females; mean age
? 30.58 years; SD ? 3.53). All subjects provided written
informed consent/assent for participation after being in-
formed of the risks, as well as being given the opportunity to
ask questions. The National Institute of Mental Health (NIMH)
Institutional Review Board (IRB) approved all procedures.
Subjects were healthy as determined by a physical, psychiatric
interview (Kaufman et al 1997; Spitzer et al 1992); intelligence
quotients (IQs), as measured with the Wechsler Abbreviated
Scale of Intelligence (Wechsler 1999), were above 70.
The same task was used across the three experiments (Figure
1). Trials began with a 500-millisecond fixation point in the
middle of the screen. Two pictures of the same model then
appeared. The emotional face (happy or angry) was presented
either on the right or left side of the screen. A neutral face was
displayed on the other hemifield. Stimuli were drawn from the
Karolinska Directed Emotional Faces set (Lundqvist et al 1998)
(Stockholm, Sweden). Subjects were instructed to attend to the
midline fixation and to press the appropriate button as quickly as
possible. Following the instructions, subjects performed a prac-
(EZ), Columbia University, New York, New York.
Received March 2, 2004; revised June 17, 2004; accepted July 21, 2004.
BIOL PSYCHIATRY 2004;56:607–610
© 2004 Society of Biological Psychiatry
Incorrect trials and trials in which the reaction time for the
button response was ? 200 milliseconds or ? 1000 milliseconds
were discarded. Following this, mean reaction times were de-
rived and trials in which the reaction time exceeded two standard
deviations were removed. Mixed-model repeated measure statis-
tical analysis (Wolfinger 1997) was conducted using SAS mixed
model procedure (SAS Institute, Cary, North Carolina) for reac-
tion time. Previous work using this task found attentional biases
solely for left visual field presentation of angry faces (Mogg and
Bradley 2002). Therefore, we treated the side of the angry face
presentation as a separate factor in the behavioral analysis.
Methods for the functional magnetic resonance imaging
(fMRI) and anatomical data acquisition followed our established
procedures (Monk et al 2003). We analyzed the functional data
with Statistical Parametric Mapping 99 (SPM99, Wellcome De-
partment of Imaging Neuroscience, University College London,
London, United Kingdom) using an event-related model. The
response to each trial was modeled with a cosine basis function
for the 2100-millisecond event, which included the facial dis-
plays, probe, and fixation.
For group-level analyses, the contrast images from each
subject were fit to a second-level random effects analysis. For the
small volume correction (SVC) analysis, we followed established
procedures (Worsley et al 1996). Since the main purpose for
conducting Experiment 3 was to evaluate fMRI signal changes
over time to masked angry faces, we analyzed the effect of time
for trials in which masked angry faces cued subjects to the
location of the probe compared with trials in which the masked
angry faces cued subjects away from the probe. Specifically, the
serial position of each of these event types across the task was
entered as a modulator variable in SPM99. This approach delin-
eates voxels where activation increases linearly with time.
For behavioral data, we performed a mixed-model repeated-
measures analysis for reaction time. Time and the four levels of
trial type (the angry face on the left or right and the probe on the
left or right) served as factors. For Experiment 1, this analysis
revealed a time-by-trial-type interaction, F(483, 357) ? 1.47, p ?
.001. Figure 2 illustrates the attention bias over the first three runs
and the last run of each of the three experiments. For Experiment
2, adolescents also showed a significant time-by-trial-type inter-
action, F(475, 358) ? 1.58, p ? .0001. Finally, for Experiment 3,
healthy adults in the fMRI again displayed a significant time-by-
trial-type interaction, F(275, 135) ? 1.92, p ? .0001. In all
experiments (Figure 2), the hypothesis was confirmed: bias away
from threat increased over time.
For the first step of the fMRI analysis, we contrasted overall
activation (as opposed to time-related changes in activation) to
trials in which masked angry faces were presented on the
opposite versus the same side of the probe. Such a contrast
directly compares activity during events where attention is cued
to the opposite or same location as the motor response. Prior
studies implicate three brain regions in attention to emotional
displays: amygdala, cingulate, and orbitofrontal cortex (OFC)
(Monk et al 2003). Thus, statistical parametric maps (SPMs) used
thresholds derived from SVCs; this revealed significant activation
in each region (Table 1). Statistical parametric maps corrected for
the entire brain revealed no activation.
In our next analysis, to capture neurophysiological changes
associated with behavioral changes, we contrasted temporal
changes in brain activation related to trials in which masked angry
faces and the probe were presented on the opposite sides, com-
pared with such changes during trials in which the angry faces and
probe were on the same side. Trials with threat presented on the left
as well as trials with threat on the right were included in the fMRI
analysis to maximize statistical power. We used trial number as a
adopted a whole-brain approach, since virtually no previous study
examined temporal changes in brain activation during a task
Figure 1. Task procedures. Thirty-five actors displayed the different emo-
tions across 435 trials. The combination of actors and combination of ex-
pressions was random across trials and subjects. Two face presentation
durations were used. In 128 trials, the emotional and neutral faces were
for 467 milliseconds. For another 128 trials, both the emotion and neutral
faces were presented for 500 milliseconds. The nonmasked trials were in-
cluded as a control to evaluate whether documented effects from happy
and angry faces were specific to the masked trials or if they occurred with a
particular emotion or both emotions. The remaining trials comprised vari-
ous control conditions: same stimulus and expression was presented on
both sides trials with only fixations and probes as well as just fixation were
also presented to facilitate imaging analysis for Experiment 3. Following all
trials except the fixation only trials, a probe was presented in either the
location of the face on the left or the face on the right; participants were
instructed to press a button as quickly as possible to the location of the
probe. Thus, motor responses reflected attention bias to or away from the
threat location. The order of the trial types was random for each subject in
experiments. Statistical analysis treated time as a continuous variable in a
tion bias was calculated as reaction time to trials in which the threat was
presented on the left and the probe was on the right minus trials in which
the threat was on the left and the probe was on the left. For the adults in
Experiments 1 and 3, these changes in attention bias reflect decreased
reaction time to targets that are on the opposite side from the threat. For
adolescents in Experiment 2, the bias reflects increased reaction time to
targets presented on the same side as the threat.
608 BIOL PSYCHIATRY 2004;56:607–610
C.S. Monk et al
demonstrating behavioral changes in attention allocation. This
analysis demonstrated selective activation in the right OT cortex,
t(11) ? 7.87, p ? .05 corrected for whole brain (Figure 3). In
addition, we examined the effect of time on the regions of interest
(ROIs) described above and found no significant activation using
the small volume correction (Worsley et al 1996).
All three experiments demonstrate behavioral plasticity in the
magnitude of bias away from masked angry faces. While fMRI
data in Experiment 3 documented engagement of the amygdala,
cingulate, and OFC, these activations did not change over time.
Rather, behavioral plasticity was associated with a pattern of
increasing right OT activation. There are multiple explanations to
account for this neurophysiological-behavioral association. Re-
gardless, the data show that learning is accompanied by in-
creased activation in the posterior cortical region, a crucial area
for categorizing objects as safe or threatening (Pine 2003).
Interactions between the OT pathway and amygdala are thought
to modulate cognitive aspects of anxiety disorders including
threat bias (Pine 2003).
The variables associated with increasing as opposed to de-
creasing brain engagement across various forms of learning
remain incompletely specified. Some fMRI studies of priming,
one category of learning, document suppressed neural response;
others report increased response (Henson et al 2000; Thiel et al
2001). Similarly, learning to enhance working memory perfor-
mance relates to increased activation (Olesen et al 2004). The
current results show that increased neurophysiological response
occurs in OT cortex when humans behaviorally adapt to threat
cues. Moreover, the study uses an attention bias task sensitive to
differences between healthy subjects and those with anxiety
disorders. Therefore, such behavioral differences may reflect
disorder-related perturbations in these same pathways.
Individual differences in adapting to adverse events account
for differential susceptibility to anxiety-related psychopathology
(Halligan et al 2003). The effects of such events are thought to
result at least partially from plastic changes in attention-oriented
processes examined in the current study (MacLeod et al 2002).
Therefore, charting brain activation patterns associated with
plastic as opposed to invariant threat responses may more fully
capture the relevant events that lead to clinically significant
anxiety. The use of such a task to compare healthy individuals
and those with anxiety disorders will bring us one step closer to
documenting the unfolding of maladaptive neurocognitive pro-
cesses associated with adversity.
We thank K.E. Towbin and A.J. Zametkin for medical oversight;
H. Iwamoto for computer programming; and A. Merikangas and A.
Hoberman for technical support.
Draganski B, Gaser C, Busch V, Schuierer G, Bogdahn U, May A (2004):
Neuroplasticity: Changes in grey matter induced by training. Nature
sives: A dichotic listening analysis. Cognit Ther Res 10:477–486.
Fox E, Russo R, Bowles R, Dutton K (2001): Do threatening stimuli draw or
hold visual attention in subclinical anxiety? J Exp Psychol Gen 130:681–
forms of repetition priming. Science 287:1269–1272.
tiated startle paradigm. Behav Neurosci 100:11–22.
Kaufman J, Birmaher B, Brent D, Rao V, Flynn C, Moreci P, et al (1997):
Schedule for affective disorders and schizophrenia for school-age chil-
Lundqvist D, Flykt A, Vhman A (1998): The Karolinska Directed Emotional
Faces. [Pictorial face set available from Department of Neurosciences,
Karolinska Hospital, Stockholm, Sweden].
MacLeod C, Mathews A, Tata P (1986): Attentional bias in emotional disor-
MacLeod C, Rutherford E, Campbell L, Ebsworthy G, Holker L (2002): Selec-
tive attention and emotional vulnerability: Assessing the causal basis of
their association through the experimental manipulation of attentional
Mathews A, Mogg K, Kentish J, Eysenck M (1995): Effect of psychological
treatment on cognitive bias in generalized anxiety disorder. Behav Res
faces in social anxiety. Behav Res Ther 40:1403–1414.
Monk CS, McClure EB, Nelson EE, Zarahn E, Bilder RM, Liebenluft E, et al
emotional facial expressions. Neuroimage 20:420–428.
Olesen PJ, Westerberg H, Klingberg T (2004): Increased prefrontal and pari-
etal activity after training of working memory. Nat Neurosci 7:75–79.
Table 1. Peak Activations for the Contrast of Threat with the Probe on the
Opposite Relative to Threat with the Probe on the Same Side in
Regiontp value x, y, z
Left Orbital Frontal
26, ?8, ?8
?22, 32, ?12
4, 6, 28
?10, 38, 20
6, 40, 18
8, 28, 20
8, 20, 30
0, 8, 28
p values are significant at .05 using the small volume correction method
(Worsley et al 1996).
Figure 3. The contrast of threat with the probe on the opposite relative to
modulator variable to document changes in brain function over time in
Institute [MNI] coordinates).
C.S. Monk et al
BIOL PSYCHIATRY 2004;56:607–610 609
Pine DS (2003): Developmental psychobiology and response to threats: Rele-
Posner MI, Snyder CR, Davidson BJ (1980): Attention and the detection of
Spitzer RL, Williams, JB, Gibbon M, First MG (1992): The Structured Clinical
Interview for DSM-III-R (SCID). I: History, rationale, and description. Arch
The Psychological Corporation.
Worsley KJ, Marrett S, Neelin P, Vandal AC, Friston KJ, Evans AC (1996): A
of cerebral activation. Hum Brain Mapp 4:58–73.
610 BIOL PSYCHIATRY 2004;56:607–610
C.S. Monk et al