The effects of social stress and cortisol responses on the
preconscious selective attention to social threat
Karin Roelofsa,*, Patricia Bakvisa,b, Erno J. Hermansc, Johannes van Peltd, Jack van Honke
aSection Clinical and Health Psychology, University of Leiden, P.O. Box 9555, 2300 RB Leiden, The Netherlands
bStichting Epilepsie Instellingen Nederland (SEIN), Heemstede, The Netherlands
cF.C. Donders Centre for Cognitive Neuroimaging, Nijmegen, The Netherlands
dDepartment of Clinical Chemistry, Leiden University Medical Center (LUMC), Leiden, The Netherlands
eHelmholtz Institute, University Utrecht, The Netherlands
Received 7 March 2006; accepted 1 September 2006
Available online 6 October 2006
The purpose ofthe presentstudy was to investigatethe effects of social stress and stress-induced cortisol on the preconscious selectiveattention
to social threat. Twenty healthy participants were administered a masked emotional Stroop task (comparing color-naming latencies for angry,
neutral and happy faces) in conditions of rest and social stress. Stress was induced by means of the Trier social stress test. Based on the stress-
induced increase in cortisol levels, participants were allocated post hoc (median-split) to a high and low responders group. In contrast to low
responders, high responders showed a negative or avoidant attentional bias to threat (i.e. shorter latencies for angry than neutral faces) in the rest
condition. Most importantly, although low responders became avoidant, the high responders became vigilant to the angry faces after stress
glucocorticoid stress-responsiveness with diminished avoidance and prolonged freezing reactions during stress.
# 2006 Elsevier B.V. All rights reserved.
Keywords: Cortisol; Avoidance; Emotional Stroop task; Masked angry faces; Trier social stress test; Selective attention
Social submissiveness and avoidance in social situations is
associated with hypercortisolism and increased activity of the
hypothalamic pituitary adrenal (HPA)-axis in primates (e.g.
the relation between social avoidance tendencies and HPA-axis
unequivocal. Some recent studies found significant relations
between self-reports of avoidance motivation, such as low self-
esteem, negative affect and social submissiveness on the one
hand and elevated cortisol responses to social stress on the other
hand (e.g. Gruenewald et al., 2004; Pruessner et al., 1997). Also
Hessl et al. (2006) found a significant relation between gaze
avoidance and cortisol responses during a social stress test in
healthy children. In contrast, a study among army recruits
responses to social stress. Finally, Buss et al. (2003) studied
avoidance reactions (observations of fear, sadness and with-
of these studies social avoidance tendencies were indirectly
measured using self-report questionnaires of withdrawal
motivation (Gruenewald et al., 2004; Pruessner et al., 1997) or
(Buss et al., 2003; Hellhammer et al., 1997; Hessl et al., 2006).
activation and social avoidance tendencies directly by means of
controlled reaction time paradigms. Using such paradigm, Van
Honk et al. (1998) found that the tendency to avoid social threat
stimuli was associated with high basal cortisol levels. The latter
authors measured the preconscious emotional response to social
neutral facial expressions were presented backwardly masked
and subjects were instructed to color name the masks. In
emotional Stroop tasks, attentional bias scores are calculated by
subtracting the reaction times (RT) for neutral faces from those
Biological Psychology 75 (2007) 1–7
* Corresponding author. Tel.: +31 71 5273955; fax: +31 71 5274678.
E-mail address: Roelofs@fsw.leidenuniv.nl (K. Roelofs).
0301-0511/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
indicatevigilance,whilenegative attentionalbias scores (i.e. RT
forangry faces are shorter than RT forneutral faces) are thought
to indicate avoidance (e.g. Putman et al., 2004; Van Honk et al.,
1998, 2000; Mathews and MacLeod, 1994). The notions of
to both masked and unmasked angry faces not only find support
in light of correlational studies using personality questionnaires
and hormone levels but is also causally supported by human
studies applying hormone administration and repetitive tran-
Haan, 2001; Van Honk and Schutter, 2005). This task, thus,
seems to provide a research paradigm for the study of human
avoidance reactions and the way they are influenced by
individual differences, such as differences in basal cortisol
levels. However, the masked emotional Stroop task has not yet
been applied in stress challenge studies that allow studying the
the masked emotional Stroop test before and after a social
stressor and investigated the effects of social stress and
glucocorticoid stress-responsiveness on preconscious attention
processing of social threat stimuli.
Although glucocorticoid stress-responsiveness has never
been studied in relation to the attentional processing of masked
social threat stimuli, it has been studied in relation to the
processing of unmasked social threat stimuli in two recent
investigations. Roelofs et al. (2005) used a manual approach–
the emotional valence of pictures presenting happy or angry
facial expressions, by making either approaching (arm-flexion)
or avoidant (arm extension) arm-movements. The manual
responses were made in affect-congruent (i.e. happy face-
avoid; angry face-approach) instruction conditions. Subjects
were tested in a rest and social stress condition and stress was
induced by the Trier social stress test (TSST: Kirschbaum et al.,
1993) that is known to elicit significant cortisol responses in the
majority of the subjects (Dickerson and Kemeny, 2004).
Comparison of high and low cortisol-responders revealed that
high responders showed larger congruency effects, involving
faster avoidance reactions to angry faces in the rest condition.
Most significantly, in the social stress condition the initial
avoidance reactions of the high responders disappeared. Thus,
subjects characterized by a high stress-responsiveness of the
HPA-axis displayed relatively high avoidance tendencies in
tendencies during stress (Roelofs et al., 2005). Although this
manual avoidance task offers a direct and controlled operatio-
nalization of avoidance tendencies (Rotteveel and Phaf, 2004) it
remains difficult to determine whether stress and cortisol
reactivity primarily affected processes involved in the response
initiation, such as movement planning and preparation or also
earlier processes such as attentional processing of the threat
stimuli. In an attempt to shed light on this question we added an
unmasked Stroop color-word task, presenting social threat,
general threat and neutral words in the same experimental setup
(Roelofs and Elzinga, 2005). The results again showed that the
initially increased avoidance reactions of high cortisol-respon-
ders disappeared during stress and, whereas high responders
became vigilant, the low responders turned avoidant during
the subjects’ motivated attention to social threat stimuli. These
et al., 1999; Okuda et al., 2004) and human (Abercrombie et al.,
changes during stress depend on an interplay between cortisol
findings from unmasked Stroop test have been criticized for the
fact that subjects may be able to override emotional Stroop
den Hout et al., 1995; Van Honk et al., 1998; Williams et al.,
1996). In contrast, subliminal Stroop tests using masked threat
stimuli seem to provide for a more reliable index of motivated
attention (Putman et al., 2004) that is less vulnerable to
uncontrollable subject and task factors.
The purpose of the present study was to investigate whether
the previously found interaction between social stress and
glucocorticoid stress-responsiveness on avoidance behavior
(Roelofs and Elzinga, 2005; Roelofs et al., 2005) would hold
for the preconscious selective attention to social threat. To this
end, healthy subjects were administered the masked emotional
Strooptask mentionedearlier,applyingangry andneutralfacial
expressions for stimuli. During the task, backwardly masked
pictures of angry, neutral and happy faces were briefly
presented. The happy faces were added to check for a possible
attentional bias for emotional stimuli per se. Stress was again
induced using the TSST and high and low cortisol-responders
were compared with respect totheir avoidance reactionsinboth
a resting and a social stress condition.
Based on the previous findings (Roelofs and Elzinga, 2005;
Roelofsetal.,2005) we expected tofinda significantinteraction
between the test condition (rest versus stress) and the subject
initially increased avoidance tendencies of high cortisol-
responders would diminish under conditions of social stress.
We tested these hypotheses in a sample of twenty volunteers (18 females, 2
males) with a mean age of 22.1 years (S.D. = 4.2) whowere originally recruited
as a control group for a larger patient study addressing a different research
question. Nine subjects used oral contraceptives and all except one females had
registered the first day of the last menstruation allowing to calculate the current
week of the menstrual cycle (week 1 (n = 7); week 2 (n = 2); week 3 (n = 4);
week 4 (n = 4)).1The participants were recruited via advertisements and
participated in the experiment for financial credit reasons. Exclusion criteria
were: any psychiatric disorder on AXIS-I (DSM-IV, APA, 1994), any clinical
significant medical disease, use of medication, and age <18 or >40. Partici-
K. Roelofs et al./Biological Psychology 75 (2007) 1–72
1On average, the first two weeks of the menstrual cycle include the follicular
phase followed by the ovulation and the last twoweeks involvethe luteal phase.
pants were instructed to minimize physical exercise during the hour preceding
the experiment and not to take large meals, coffee, drinks with low pH or
normal or correct-to-normal vision. The study was approved by the local ethics
committee and all participants provided written informed consent.
2.2.1. Emotional Stroop task
Selective attention to happy and angry faces was assessed using a masked
pictorial emotional Stroop task. Facial stimuli of 10 different individuals (5
males, 5 females) were taken from Ekman and Friesen’s Pictures of Facial
Affect (Ekman and Friesen, 1976), each displaying a neutral, a happy and an
angry expression. The facial stimuli were presented for 14 ms (below the
threshold for explicit visual awareness). Immediately after the stimulus pre-
sentation the pictures were replaced by a masking stimulus. The masking
stimuliconsisted of randomlycut, reassembledand re-photographed pictures of
response initiation, the presentation of the masking stimulus was terminated.
After a random inter-trial interval (ITI; 2–4 s) new trials started with a 750 ms
lasting fixation point. A total of 30 happy, 30 angry and 30 neutral faces were
presented in a random order with the restriction that the same color was never
repeated more than twice consecutively. The dependent measures in the
emotional Stroop task are attentional bias scores for emotional facial expres-
sions (i.e. the mean individual color-naming latencies of angry (or happy) faces
minus the individual mean color-naming latencies on neutral faces). A positive
attentional bias score, indicating slower color-naming responses to emotional
stimuli as compared to neutral stimuli, is interpreted as a vigilant response,
whereas a negative attentional bias score, indicating faster color-naming
responses to emotional stimuli as compared to neutral stimuli, is interpreted
as an avoidant response (e.g. Putman et al., 2004; Van Honk et al., 1998, 2000;
Mathews and MacLeod, 1994).
To maximize the quality of the voice key registration, the subjects were
instructedto speak loud and clearly, to keep their mouthopen duringthe task,to
avoid smackingtheir lips or coughing beforerespondingand to not correct their
answer in case they had already started vocalizing an erroneous response. All
instructions were practiced in a practice phase of nine stimulus presentations in
which only masks were used (i.e. without facial stimuli).
To ascertain that subjects remained unaware of thevariable ofinterest in the
Stroop task, subliminal thresholds for the facial expressions were controlled for
by an awareness check after the experiment. During this three-alternative,
forced choice, emotional-neutral recognition procedure, a random set of 30
maskedfaces wasshownto the subjects. In advance ofthe test the subjectswere
explicitly told that the set contained 10 happy, 10 neutral and 10 angry faces,
and they were instructed to indicate (or guess), whether the presented picture
contained a neutral, happy or angry expression by pushing the corresponding
2.2.2. The Trier social stress test
This psychological challenge test, which mainly consists of a free speech
and a mental arithmetic task of 15 min duration, has been found repeatedly to
induce significant endocrine and cardiovascular responses in the majority of the
the test is reported under the subheading ‘procedure’.
2.2.3. Physiological and subjective measures
All physiological and subjective stress-measures were obtained at nine
assessment points overa 200-minperiod, at respectively ?60, 0, +20, +40, +60,
+80, +100, +120 and +140 min with reference to the start of the stressor. All
assessments were performed between 13.15 and 17.00 p.m.
Cortisol: Saliva samples were obtained using Salivette collection devices
(Sarstedt, Rommelsdorf, Germany). Saliva samples were stored at ?20 8C
before assaying. Biochemical analysis of free cortisol in saliva was performed
using a competitive electrochemiluminescence immunoassay (ECLIA, Elecsys
2010, Roche Diagnostics), as described elsewhere (Van Aken et al., 2003).
Heart rate (HR) was recorded continuously by an Ambulatory Monitoring
System of the Vrije Universiteit Amsterdam (VU-AMS) version 3.6, a small
battery powered device for ambulatory recording. It was measured via three
left side of the chest, and at the bottom right side of the chest (cleaned with
alcohol). For each participant, HR was post hoc averaged for 2 min starting
from a marker given at each of the nine assessment points.
non-dominant arm using an automatic blood pressure monitor (Omron R5-I)
that could be initiated manually.
Finally, a subjective measure of anxiety, rated on a visual analogue scale,
ranging from 0 to 10, was administered at each assessment point.
2.2.4. Psychological measures
Symptoms of anxiety and agoraphobia were measured using the corre-
sponding sub-scales of the Dutch version (Arrindell and Ettema, 1986) of the
Symptom Checklist (SCL-90, Derogatis, 1983).
Participants arrived at the experimental room at 13.15 p.m., after which the
VU-AMS devicewas connected and checked by the experimenter. After the first
During this 8 min lasting task, participants were instructed to name as quickly as
possible the color of the masking stimulus, which presentation was terminated
after vocal response initiation. Task administration took place just before the
and was preceded by a battery of additional cognitive tests, of which the results
will be reported elsewhere. Subsequently, the experimenter introduced the TSST
by telling the participants that they would be taking on the role of a job applicant
(the job description was selected a priori, defining a position that would be
challenging and relevant to the current situation of the participant). Participants
were given 5 min to prepare a 5-min long, free speech to an audience of three
individuals whowere in another room waiting to interview them. They were told
would be performed, and that the speech would be criticized on content and
accuracy. After this introduction, the experimenter left the room. Following the
preparation time, the audience (three persons) entered the room and prominently
switched on the camera and microphone. Participants were instructed to stand in
seated in the center asked the participant to describe his/her qualifications for the
were expected to utilize the entire 5 min for the speech as described by
Kirschbaum et al. (1993). For the mental arithmetic task (5 min duration)
K. Roelofs et al./Biological Psychology 75 (2007) 1–73
Fig. 1. Free salivary cortisol in nmol/l (mean ? S.E.M.) before, during and
participants were instructed to serially subtract 13 from 1587. The audience
responded to any mistakes by saying: ‘‘Incorrect, start from the top; subtract 13
from 1587’’. Immediately after the TSST, the audience called the experimenter
back into the room and requested him to administer the cognitive tests including
the emotional Stroop task again in their presence. In this way, the social stress
context remained present during the second administration of the Stroop task
and returned for a short debriefing after the last physiological assessment had
2.4. Statistical analyses
Reaction time outliers were filtered using a <150 and >1500 ms cut-off,
and subsequent removal of all RTs exceeding 2.5 S.D. from the mean. For each
participant, the remaining latencies (99.5%) for the correct responses were
averaged over the facial expression types presented in each condition. The
influence of stress induction on physiological and subjective stress measures, as
well as the influence of stress and cortisol on the task performance, were tested
using repeated measures analyses of variance (ANOVA rm). The relation
between cortisol responses on the one hand and anxiety scores on the other
hand were calculated using Pearson’s correlations. All statistical analyses
described employed a two-tailed alpha of 0.05 and effect sizes of significant
results are reported using the partial eta squared (h2).
3.1. Physiological and subjective stress responses
Separate one-way ANOVA’s rm for the physiological and
subjective stress measures (each measured at nine time points)
showed significant increases on all stress measures over time:
cortisol (F(1, 19) = 19.13, p < 0.0001, h2= .50); HR (F(1,
19) = 10.44, p < 0.005, h2= .36), SBP (F(1, 19) = 43.84,
p < 0.0001, h2= .69); DBP (F(1, 19) = 13.49, p < 0.005,
19) = 58.41, p < 0.0001, h2= .49). For all measures the mean
rates assessed before stress induction (assessments 1–2) were
significantly lower than the mean rates assessed after stress
induction (assessments 3–9) (all p < 0.001), indicating that
stress induction was successful.
3.1.2. Cortisol responses
The stress-related cortisol-response (CR) of each individual
was computed by calculating the percentage increase from the
individual minimum cortisol level before stress induction to the
individual maximum after stress induction. The mean CR ofthe
(S.E.M.) = 35.04%). Subjects were post hoc (median split)
allocated to a high (n = 10) and low (n = 10) responders group.
(S.E.M. = 63.96) and differed significantly from the low
responders (mean CR = 7.27%; S.E.M. = 6.43) in their cortisol
responses (F(1, 18) = 9.49, p < 0.01, h2= .35) (see Fig. 1). The
high and low responders did not differ with respect to HR (F(1,
18) = 0.84, p = 0.52), DBP (F(1, 18) = 0.04, p = 0.84) and
subjectively experienced anxiety (F(1, 18) = 2.08, p = 0.13).
However, the high responders showed a relative increase with
respect to the SBP (F(1, 18) = 4.47, p < 0.05, h2= .20),
compared to the low responders.
3.2. Behavioral results
3.2.1. Forced choice emotional-neutral recognition check
Chance performance in a three-alternative forced choice
recognition check using 30 stimuli is 10 correct identifications
per subject (33.33%). Of the total number of 600 trials, 199
were correct (=33.17%). Non-parametric tests showed that
therewas no significant deviation from the expected value (cut-
point = 13) and that masking was successful (n = 20, p = 0.70).
3.2.2. Error rates
The error rates of the Stroop performance in the rest and
stress conditions were 2.3% and 1.7%, respectively. Therewere
no differential effects of facial expression, stress or cortisol
response with respect to the error rates.
3.2.3. The effects of stress and CR on the Stroop
To investigate the effects of the social stress context and the
cortisol responses (CR) on the selective attention to angry and
neutral faces we conducted a two-way ANOVA rm for the
attentional bias scores for angry faces (RT angry faces minus
RT neutral faces) with condition (rest, stress) as within-subject
factor and CR (high, low responders) as between-subject factor.
There were no main-effects for condition (F(1, 18) = 1.07,
p = 0.31) and CR (F(1, 18) = 0.18, p = 0.68) but there was a
significantcondition ? CR
p < 0.05, h2= .24), showing opposite response patterns in
high and low responders in rest and stress conditions (see
Fig. 2). Compared to low responders, high responders showed a
negative or avoidant attentional bias to threat (i.e. shorter
latencies for angry than neutral faces) in the rest condition.
Most importantly, although low responders became avoidant,
the high responders became vigilant to social threat stimuli
after stress induction. To test whether these effects were
affected by gender, we excluded the two males from the
analyses and found that the condition ? CR interaction
remained significant (F(1, 16) = 5.32, p < 0.05, h2= .25).
Finally, we controlled for the use of Oral contraceptives and for
the menstrual cycle (ranging from weeks 1 to 4 with reference
to the onset of the menstruation). Therewas no significant main
interaction(F(1,18) = 5.55,
K. Roelofs et al./Biological Psychology 75 (2007) 1–74
Fig. 2. Mean attentional bias scores for angry faces for (cortisol) high respon-
ders and low responders, tested in rest and stress conditions.
effect for menstrual cycle (F(1, 13) = 0.25, p = 0.62), but a
significant main effect for Oral contraceptives (F(1, 13) = 5.0,
p < 0.05, h2= .27) indicated a general RT slowing for non-
users compared to users. Most importantly the condition ? CR
interaction remained significant after controlling for the effects
of oral contraceptives and menstrual cycle (F(1, 13) = 6.24,
p < 0.05, h2= .32). Univariate post hoc F-tests (also while
controlling for oral contraceptives and menstrual cycle) for the
attentional bias scores for angry faces showed that there were
trends towards significant main effects for CR in both the rest
(F(1, 13) = 3.60, p = 0.08, h2= .22) and the stress condition
(F(1, 13) = 4.18, p = 0.06, h2= .24). There were no main and
interaction effects of condition and CR for the attentional bias
scores for happy faces (all p > 0.11).
3.2.4. Correlational analyses
The stress induced cortisol responses (maximum increase)
were significantly correlated to the SCL90 agoraphobia
subscale (r = .76, p < 0.0001) and showed a trend into the
same direction for the SCL90 anxiety scale (r = .46,
p = 0.054). In all reported correlations we controlled for the
effects of oral contraceptives andmenstrual cycle usingpartial
of social stress andglucocorticoid stress responsiveness onthe
preconscious selective attention to social threat stimuli. The
results showed that subjects with high and low glucocorticoid
stress-responsiveness showed opposite reaction patterns to the
highresponders showeda negative or avoidant attentional bias
to threat (i.e. shorter latencies for angry than neutral faces) in
the rest condition. Most importantly, although low responders
became avoidant after stress induction, the high responders
became vigilant to the social threat stimuli during stress. In
other words, in the context of social stress the high responders
showed a relative increase in selective attention for angry
faces. Another important result was that these differential
effects of stress and cortisol counted only for the attentional
bias scores for angry faces and not for happy faces. These
the cortisol stress-responsiveness were specific for context
relevant (social threat) cues and not for unspecific emotional
The diminished avoidance to threat in the high responders
group during stress is in line with the results of two previous
threat cues were measured using a manual avoidance task
(Roelofs et al., 2005) and an emotional Stroop color-word task
(Roelofs and Elzinga, 2005), respectively. In the latter studies a
significant interaction between the test condition (rest versus
social stress) and group (high versus low cortisol-responders)
also showed that high cortisol-responders displayed a decrease
in avoidance tendencies towards the threat cues during stress. A
new finding that emerges from the present study is that these
interacting effects are already manifested on a preconscious
level of attentional processing. The results from the awareness
check in the present study confirmed that the subjects did not
consciously perceive the stimuli. This makes it unlikely that
they have exerted strategic effort to control attentional bias
effects (e.g. MacLeod and Hagan, 1992; Putman et al., 2004;
Van den Hout et al., 1995; Van Honk et al., 1998; Williams
et al., 1996) and makes the findings less vulnerable to
uncontrollable subject factors. This additional value of using a
masked Stroop task is supported by the fact that masked
compared to unmasked versions of the emotional Stroop task
have yielded more consistent results (Putman et al., 2004) and
by the finding that masked, as compared to unmasked
emotional Stroop performance is more predictive of actual
coping with stressful life events (MacLeod and Hagan, 1992).
The avoidant reactions of high responders in the rest
condition correspond with the findings of a recent investigation
applying the same masked Stroop task in patients with social
anxiety disorder (Hermans et al., 2006). The patients were also
tested in a non-stress condition and displayed avoidant
reactions to the social threat cues as well. Interestingly, social
anxiety disorder is also associated with elevated cortisol
responses to social stress (e.g. Condren et al., 2002), which
raises the hypothesis that glucocorticoid stress-responsiveness
plays a role in the altered attentional processing in socially
Previous studies have indicated that subjects who show
relatively high cortisol responses in social stress tests tend to
report lower self-esteem, more negativeaffect and higher social
submissiveness than their low responding counterparts (Grue-
newald et al., 2004; Pruessner et al., 1997). Also, we found
increased glucocorticoid stress responsiveness to be associated
with increased symptoms of anxiety and agoraphobia. In the
light of these findings it is likely that our high responders group
may have perceived the ‘non-stress’ test condition as more
threatening than our low responders group. The fact that the
high responders showed no increases on physiological and
subjective stress measures in this rest condition may be due to
the fact that these subjects tended to display more avoidance
behavior. This was evidently shown by their Stroop perfor-
mance in the rest condition but may also have been manifested
in their general testing attitude in the rest condition. A previous
study by Van Honk et al. (2000), for example, showed that
individuals who pre-consciously directed their attention away
from angry faces in the masked Stroop task, exhibited
decreased post-task cortisol levels. These findings were
interpreted as indicative of an adaptive response to social
stress stimuli and it was speculated that these individuals may
be best served by inhibited behavioral and physiological
responses to avoid injury and energy loss (Flinn et al., 1998;
Sapolsky, 1990). Whereas our high responders were able to
show such an adaptive response in the rest condition, their
initial avoidance reactions seemed to fail in the social stress
condition. These diminished avoidance reactions during stress
may be explained in several ways. In the first place the high
responders may have had difficulty inhibiting the threatening
K. Roelofs et al./Biological Psychology 75 (2007) 1–75
attentional bias towards the stress context and away from the
emotional Stroop task. However, such general drop in selective
attention for the task would likely have been accompanied by a
were not the case. Also, the fact that the high responders
directed relatively more attention to angry faces compared to
neutral ones indicates that the facial stimuli presented in the
Stroop test were still processed in the stress condition. A more
plausible explanation for the failing avoidance and increased
vigilance for social threat stimuli in the social stress context
may be found in a behavioral inhibition that is comparable to
so-called freezing responses observed in animals with elevated
glucocorticoid stress-reactivity when theyare exposedtostress.
Research in animals has, for example, shown that high basal
and reactive cortisol levels are associated with increased
freezing and diminished active avoidance reactions during
stress in primates (Kalin et al., 1998) and rats (Nunez et al.,
1996). Freezing is regarded as an extreme form of behavioral
inhibition and in primates it is suggested to be analogous to
fearful responses frequently observed in intensely inhibited
children (Kalin et al., 1991). Studies in these children also
demonstrated extreme inhibition to be associated with
increased levels of salivary cortisol (Kagan et al., 1988).
be considered is the possibility that the relative increase in the
high responders’ attention to angry faces reflects an approach
reaction towards threat evoked by the social stress context.
However, this explanation is unlikely because such reaction is
typical for subjects featured by high testosterone as opposed to
are associated with quite opposite motivations in situations of
social threat. Generally, high levels of cortisol are related to
socially avoidant, submissive behavior (Kagan et al., 1988;
Sapolsky, 1990; Schulkin et al., 1998) whereas high levels of
testosterone are related to socially dominant behavior (Mazur
and Booth, 1998; Zuckerman, 1991). Moreover, the stress
induced cortisol levels in the present study were positively
related to self-report measures of anxiety and agoraphobia,
which is also incompatible with an approach motivation.
To our knowledge the present study is the first to show the
effects of stress and stress-induced cortisol on the preconscious
selective attention to social threat stimuli, and suggests that the
diminished avoidance reactions during stress are related to the
glucocorticoid stress-responsiveness. However, some limita-
tions should be considered when evaluating the findings. In the
first place, although we did find a significant group (high, low
responders) ? condition (rest, social stress) interaction, the
groupsizes(N = 10)wererelativelysmall.Astrongpointisthat
our results replicated the findings from two previous studies
that applied the samestudydesign, butused different avoidance
tasks (Roelofs et al., 2005; Roelofs and Elzinga, 2005).
Nevertheless, a replication of our findings with the masked
emotional Stroop task in larger sample sizes would strengthen
the results. Second, our sample primarily consisted of female
subjects. Future studies are clearly needed to systematically
examine gender differences. Another possible limitation is that
the assessment of the menstrual cycle was not based on
biochemical testing and depended on the accuracy with which
subjects had registered the first day of the last menstruation.
Despite these limitations the present findings may have
several important implications. First, the data reveal that the
relation between high cortisol responses and diminished
avoidance can already be observed at a preconscious level of
attention processing. Second, a more general implication
involves the importance of taking the glucocorticoid stress-
That cortisol high responders and low responders may show
quite opposite approach–avoidance reactions to social threat
has been demonstrated several times now (see Roelofs and
Elzinga, 2005; Roelofs et al., 2005). Finally, the data provide
implications for future research. Testing participants featured
by social anxiety such as patients with social phobia or post-
traumatic stress is an important step to reveal the mechanisms
by which stress-induced cortisol may affect avoidance
In sum, subjects with a high glucocorticoid stress-respon-
siveness (high responders) showed opposite reaction patterns to
the social threat stimuli compared to subjects with low
glucocorticoid stress responsiveness (low responders). In
contrast to low responders, high responders showed a negative
or avoidant attentional bias to threat in the rest condition. Most
importantly, although low responders became avoidant during
stress, high responders’ initial avoidant reactions to the social
threat stimuli disappeared in the context of social stress. These
findings are in line with previous studies in both animals and
humans, showing high glucocorticoid stress-responsiveness to
be associated with diminished avoidance and prolonged
freezing reactions in stressful situations. A new finding is that
the relation between cortisol responses and failing avoidance
already counts for the preconscious processing of threat. These
findings may not only provide insight into fundamental
processes mediating human avoidance reactions. They, more-
over, offer a fruitful experimental model for the study of
avoidance reactions in patients with anxiety disorders, such as
social phobia and post-traumatic stress disorder.
This study was supported by VENI Grant (#451-02-115)
from The Netherlands Organization for Scientific Research
(NWO) awarded to K. Roelofs. The authors thank Nathalie van
der Krogt for assistance during data collection and Peter
Edelbroek (Stichting Epilepsie Instellingen Nederland, Heem-
stede) and Bart Ballieux (Department of Clinical Chemistry,
Leiden University Medical Centre: LUMC) for conducting the
Abercrombie, H.C., Speck, N.S., Monticelli, R.M., 2006. Endogenous cortisol
elevations are related to memory facilitation only in individuals who are
emotionally aroused. Psychoneuroendocrinology 31, 187–196.
American Psychiatric Association, 1994. Diagnostic and Statistical Manual
of Mental Disorders, 4th ed. American Psychiatric Association, Washing-
K. Roelofs et al./Biological Psychology 75 (2007) 1–76
Arrindell, W.A., Ettema, J.H.M., 1986. SCL-90: Dutch translation and adapta- Download full-text
tion. Lisse, Swets and Zeitlinger.
Buss, K.A., Schumacher, J.R., Dolski, I., Kalin, N.H., Goldsmith, H.H.,
Davidson, R.J., 2003. Right frontal brain activity, cortisol, and withdrawal
behavior in 6-month-old infants. Behavioral Neuroscience 117, 11–20.
Condren, R.M., O’Neill, A., Ryan, M.C.M., Barrett, P., Thakore, J.H., 2002.
HPA axis response to a psychological stressor in generalized social phobia.
Psychoneuroendocrinology 27 (6), 693–704.
De Kloet, E.R., Oitzl, M.S., Joe ¨ls, M., 1999. Stress and cognition: are
corticosteroids good or bad guys? Trends in Neurosciences 22, 422–426.
Derogatis, L.R., 1983. Symptom Checklist-90-R: Administration, Scoring, and
Procedures Manual. Clinical Psychometric Research, Baltimore, MD.
Dickerson, S.S., Kemeny, M.E., 2004. Acute stressors and cortisol responses: a
theoretical integration and synthesis of laboratory research. Psychological
Bulletin 130, 355–391.
Ekman, P., Friesen, W.V., 1976. Pictures of Facial Affect. Consulting Psychol-
ogist Press, Palo Alto, CA.
Elzinga, B.M., Roelofs, K., 2005. Cortisol-induced impairments of working
memory require acute sympathetic activation. Behavioral Neuroscience
Flinn, M., Baerwald, C., Decker, S., England, B., 1998. Evolutionary functions
of neuroendocrine responses to the social environment. The Behavioral and
Brain Sciences 21, 372–374.
Gruenewald, T.L., Kemeny, M.E., Aziz, N., Fahey, J.L., 2004. Acute threat to
the social self: shame, social self-esteem, and cortisol activity. Psychoso-
matic Medicine 66, 915–924.
Hellhammer, D.H., Buchtal, J., Gutberlett, I., Kirschbaum, C., 1997. Social
hierarchy and adrenocortical stress reactivity in men. Psychoneuroendocri-
nology 22, 643–650.
Hermans, E.J., Montagne, B., Schutters, S., Putman, P., Westenberg, H.G., Van
Honk, J., 2006. Reduced attentional processing of facial threat in social
anxiety disorder, submitted for publication.
Hessl, D., Glaser, B., Dyer-Friedman, J., Reiss, A.L., 2006. Social behavior and
cortisol reactivity in children with fragile ? syndrome. Journal of Child
Psychology and Psychiatry 47, 602–610.
Kagan, J., Reznick, J.S., Sidman, N., 1988. Biological bases of childhood
shyness. Science 240, 167–171.
Kalin, N.H., Shelton, S.E., Rickman, M., Davidson, R.J., 1998. Individual
differences in freezing and cortisol in infant and mother rhesus monkeys.
Behavioral Neuroscience 112, 251–254.
rhesus monkeys: ontogeny and context-dependent selective expression.
Child Development 62, 1175–1183.
Kirschbaum, C., Pirke, K.M., Hellhammer, D.H., 1993. The Trier social stress
test: a tool for investigating psychobiological stress response in a laboratory
setting. Neuropsychobiology 28, 76–81.
MacLeod, C., Hagan, R., 1992. Individual differences in the selective proces-
sing of threatening information, and emotional responses to a stressful life
event. Behavior Research and Therapy 30, 151–161.
Mathews, A., MacLeod, C., 1994. Cognitive approaches to emotion and
emotional disorders. Annual Review of Psychology 45, 25–50.
Mazur, A., Booth, A., 1998. Testosterone and dominance in men. The Beha-
vioral and Brain Sciences 21, 253–297.
Nunez, J.F., Ferre, P., Escorihuela, R.M., Tobena, A., Fernandez-Teruel, A.,
1996. Effects of postnatal handling of rats on emotional, HPA-axis, and
prolactin reactivity to novelty and conflict. Physiology & Behavior 60 (5),
Okuda, S., Roozendaal, B., McGaugh, J.L., 2004. Glucocorticoid effects on
object recognition memory require training-associated emotional arousal.
Proceedings of the National Academy of Sciences of the USA 101,
Pruessner, J.C., Gaab, J., Hellhammer, D.H., Lintz, D., Schommer, N., Kirsch-
baum, C., 1997. Increasing correlations between personality traits and
cortisol stress responses obtained by data aggregation. Psychoneuroendo-
crinology 22, 615–625.
Putman, P., Hermans, E.J., Van Honk, J., 2004. Emotional Stroop performance
for masked angry faces: it’s BAS, not BIS. Emotion 4, 305–311.
Roelofs, K., Elzinga, B.M., Rotteveel, M., 2005. The effects of stress-induced
cortisol responses on approach-avoidance behavior. Psychoneuroendocri-
nology 30, 665–677.
Roelofs, K., Elzinga, B.M., 2005. The effects of social stress and stress-induced
cortisol on the selective attention to social threat cues. Poster Presented at
the 36th Annual Conference of the International Society of Psychoneur-
oendocrinology, Montreal, Quebec, Canada, September 24-27.
Rotteveel, M., Phaf, R.H., 2004. Automatic affective evaluation does not
automatically predsipose for arm flexion and extension. Emotion 4 (2),
Sapolsky, R.M., 1990. Adrenocortical function, social rank, and personality
among wild baboons. Biological Psychiatry 28, 862–878.
Sassenrath, E., 1970.Increased adrenalresponsiveness related to social stress in
rhesus monkeys. Hormones and Behavior 1, 283–298.
Schulkin, J., Gold, P.W., McEwen, B.S., 1998. Induction of corticotropin-
releasing hormone gene expression by glucocorticoids: implications for
understanding the states of fear and anxiety and allostatic load. Psycho-
neuroendocrinology 23, 219–243.
Van Aken, M.O., Romijn, J.A., Miltenburg, J.A., Lentjes, E.G.W.M., 2003.
Automated measurement of cortisol. Clinical Chemistry 49 (8),
Van den Hout, M., Tenney, N., Huygens, K., Merckelbach, H., Kindt, M., 1995.
Responding to subliminal threat cues is related to trait anxiety and emo-
tionalvulnerability: a successful replication ofMacLeod and Hagan(1992).
Behavior Research and Therapy 33, 451–454.
Van Honk, J., De Haan, E., 2001. Cortical and subcortical routes for conscious
and unconscious processing of emotional faces. In: De Gelder, B., Hey-
wood, C.A., De Haan, E. (Eds.), Varieties of Unconscious Processing.
Oxford University Press, Oxford.
Van Honk, J., Schutter, D.J.L.G., 2005. In: Harmon-Jones, E., Winkielman, P.
(Eds.), Vigilant and Avoidant Responses to Angry Facial Expressions:
Dominance andSubmissionMotives. Fundamentalsof SocialNeuroscience
Guilford Press, New York.
Van Honk, J., Tuiten, A., van den Hout, M., Koppeschaar, H., Thijsen, J., de
Haan, E., Verbaten, R., 1998. Baseline salivary cortisol levels and precon-
scious selective attention for threat: a pilot study. Psychoneuroendocrinol-
ogy 23, 741–747.
Van Honk, J., Tuiten, A., van den Hout, M., Koppeschaar, H., Thijssen, J., de
Haan, E., Verbaten, R., 2000. Conscious and preconscious selective atten-
tion to social threat: different neuroendocrine response patterns. Psycho-
neuroendocrinology 25, 577–591.
Van Honk, J., Tuiten, A., Verbaten, R., van den Hout, M., Koppeschaar, H.,
Thijssen, J., De Haan, E., 1999. Correlations among salivary testosterone,
mood, and selective attention to threat in humans. Hormones and Behavior
Williams, J.M., Mathews, A., MacLeod, C., 1996. The emotional Stroop task
and psychopathology. Psychological Bulletin 120, 3–24.
Zuckerman, M., 1991. The Physiology of Personality. Cambridge University
Press, New York.
K. Roelofs et al./Biological Psychology 75 (2007) 1–77