Action Opportunities Modulate Attention Allocation Under Social Threat

Abstract

When entering a subway car affording multiple targets for action, how do we decide, very quickly, where to sit, particularly when in the presence of a potential danger? It is unclear, from existing motor and emotion theories, whether our attention would be allocated toward the seat on which we intend to sit on or whether it would be oriented toward an individual that signals the presence of potential danger. To address this question, we explored spontaneous action choices and attention allocation in a realistic context, where a threat-related signal (an angry or fearful individual) and the target for action in that situation could compete for attentional priority. Results showed that participants chose the actions that avoided angry individuals and were more confident when approaching those with a fearful expression. In addition, covert and overt measures of attention showed a stronger avoidance effect for angry, compared to fearful, individuals. Crucially, these effects of anger and fear on attention allocation required the presence of action possibilities in the scene. Taken together, our findings show that in a realistic context offering competing action possibilities, threat-related distractors shape both action selection and attention allocation accordingly to their social function.

Full text

Emotion
Action Opportunities Modulate Attention Allocation
Under Social Threat
Emma Vilarem, Jorge L. Armony, and Julie Grèzes
Online First Publication, April 4, 2019. http://dx.doi.org/10.1037/emo0000598
CITATION
Vilarem, E., Armony, J. L., & Grèzes, J. (2019, April 4). Action Opportunities Modulate Attention
Allocation Under Social Threat. Emotion. Advance online publication.
http://dx.doi.org/10.1037/emo0000598

Action Opportunities Modulate Attention Allocation Under Social Threat
Emma Vilarem
PSL Research University Jorge L. Armony
McGill University and Douglas Mental Health
University Institute
Julie Grèzes
PSL Research University
When entering a subway car affording multiple targets for action, how do we decide, very quickly, where
to sit, particularly when in the presence of a potential danger? It is unclear, from existing motor and
emotion theories, whether our attention would be allocated toward the seat on which we intend to sit on
or whether it would be oriented toward an individual that signals the presence of potential danger. To
address this question, we explored spontaneous action choices and attention allocation in a realistic
context, where a threat-related signal (an angry or fearful individual) and the target for action in that
situation could compete for attentional priority. Results showed that participants chose the actions that
avoided angry individuals and were more confident when approaching those with a fearful expression.
In addition, covert and overt measures of attention showed a stronger avoidance effect for angry,
compared to fearful, individuals. Crucially, these effects of anger and fear on attention allocation required
the presence of action possibilities in the scene. Taken together, our findings show that in a realistic
context offering competing action possibilities, threat-related distractors shape both action selection and
attention allocation accordingly to their social function.
Keywords: facial displays, action, attention, threat, affordances
Supplemental materials: http://dx.doi.org/10.1037/emo0000598.supp
In our natural environment, we are continuously confronted with
opportunities for action. Consider, for example, the everyday sce-
nario of entering a subway car and having to decide, very quickly,
where to sit. It has been suggested that, in such situations that
afford several potential targets for action, our brain prepares mul-
tiple competing actions in parallel, while collecting sensory infor-
mation to select the most appropriate one regarding that specific
situation (Cisek, 2007). More precisely, when entering the subway
car, the specification of potential targets for sitting will be quickly
computed in the dorsal visual stream and premotor cortex. The
selection process, on the other hand, takes longer to compute; it
requires weighting each action’s respective outcomes in the pre-
frontal cortex (Alexander & Brown, 2011), using relevant infor-
mation gained through visual attention, to bias the competition
toward the seat that is expected to yield to the most positive or the
least negative result (Cisek & Pastor-Bernier, 2014). In social
contexts, such as a subway car, the question arises as to what type
of information guides the selection of our imminent actions.
One key piece of information that can influence action decision
making is the emotion expressed by others (Dezecache, Jacob, &
Grèzes, 2015; Grèzes, 2011). Indeed, evolutionary accounts posit
that emotional displays serve a communicative function by con-
veying critical information about the sender’s affective state and
associated behavioral intentions (Fridlund, 1994; Keltner & Haidt,
1999). Moreover, emotional displays are thought to have co-
evolved with the observers’ behavioral responses so that the latter
reflect the social function of the perceived expression (Dezecache,
Mercier, & Scott-Phillips, 2013). Within such a framework, the
emotional displays others produce should influence the selection
of our upcoming actions in social contexts, according to their
social function.
Returning to our everyday scenario of entering a subway car
where different possible actions compete, we would be more likely
Emma Vilarem, Laboratory of Cognitive and Computational Neurosci-
ence - INSERM U960, Ecole Normale Supérieure, PSL Research Univer-
sity; Jorge L. Armony, Department of Psychiatry, McGill University and
Douglas Mental Health University Institute; Julie Grèzes, Laboratory of
Cognitive and Computational Neuroscience - INSERM U960, Départe-
ment d’études cognitives, Ecole Normale Supérieure, PSL Research Uni-
versity.
Emma Vilarem, Jorge L. Armony, and Julie Grèzes designed the study.
Emma Vilarem developed stimuli, conducted the experiments, and performed
data analyses. Emma Vilarem, Jorge L. Armony, and Julie Grèzes prepared the
manuscript. All authors reviewed and approved the final manuscript.
The authors thank M. Chadwick for carefully proofreading this article,
E. Koechlin and V. Wyart for their comments, M. Babo-Rebelo and T.
Griessinger for the stimuli, and A. Petschen and H. Michaux for data
collection. This research was supported by FRM Team DEQ20160334878,
Fondation ROGER DE SPOELBERCH, ANR-11-EMCO-00902, ANR-10-
LABX-0087 IEC, and ANR-10-IDEX-0001-02 PSL and by INSERM.
Correspondence concerning this article should be addressed to Julie
Grèzes, Laboratory of Cognitive and Computational Neuroscience—
INSERM U960, Département d’études cognitives, Ecole Normale Supéri-
eure, PSL Research University, 29, rue d’Ulm - 75005 Paris, France.
E-mail: julie.grezes@ens.fr
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Emotion
© 2019 American Psychological Association 2019, Vol. 1, No. 999, 000
1528-3542/19/$12.00 http://dx.doi.org/10.1037/emo0000598
1

to choose to sit next to a person with a neutral facial expression
than to one displaying an angry expression. However, the choice
may be different for fearful expressions. Indeed, although angry
and fearful are negative expressions signaling the presence of
danger and would thus typically elicit avoidance behaviors (Corr,
2013; Elliot & Covington, 2001), they convey different social
meanings. Facial expressions of anger are thought to have evolved
to enhance cues of strength (Sell, Cosmides, & Tooby, 2014) and
communicate the emitter’s aggressive intent (Reed, DeScioli, &
Pinker, 2014). Being perceived as a direct threat to the observer
(Sander, Grandjean, Kaiser, Wehrle, & Scherer, 2007), they should
lead to avoidance responses. Fearful displays, in contrast, are more
ambiguous in terms of avoidance and approach behaviors. On one
hand, by signaling the presence of an imminent danger in the
observer’s environment, they should prompt avoidance behaviors
(Paulus & Wentura, 2016). Yet, by enhancing facial cues of
vulnerability and affiliation that appease social interactions and
inhibit aggression, fearful displays could also elicit approach be-
haviors (Hammer & Marsh, 2015; Marsh, Ambady, & Kleck,
2005).
To date, the literature regarding actions elicited in response to
anger and fear has yielded mixed results, with both expressions
having been found to elicit approach and avoidance responses. One
possible explanation is that the signaling functions of specific
emotional expressions and their associated impact on observers’
behaviors are susceptible to contextual influences (Van Kleef,
2009), such as the cooperative or competitive nature of the task
employed (Bossuyt, Moors, & De Houwer, 2014; Krieglmeyer &
Deutsch, 2013; Wilkowski & Meier, 2010) and the other emotions
used in the task (Paulus & Wentura, 2016). Moreover, it has been
suggested that interindividual differences in prosocial tendencies
may explain why some studies using fearful expressions observed
approach behaviors while others found avoidance (Kaltwasser,
Moore, Weinreich, & Sommer, 2017). Here, we investigated
whether these two facial expressions (i.e., anger and fear) differ-
entially influence people’s free choices in the presence of compet-
ing targets for action, according to their social function. That is, in
the example of the subway car, would we sit away from an angry
individual but next to a fearful one?
Investigating action processes entails to also consider how po-
tential targets for action are represented within the attentional
system, as visuospatial information is critical for specifying ac-
tions’ parameters (Cisek & Kalaska, 2010). The motor domain
literature has investigated this interplay between attention and
action and has demonstrated that planning an action can automat-
ically guide attention (either covert or overt) toward the end point
of the chosen action, so as to extract action-relevant information
(Fagioli, Ferlazzo, & Hommel, 2007; Kirsch, 2015; Wykowska,
Schubö, & Hommel, 2009). In the context of the premotor theory
of attention, which argues that attention allocation to a specific
location corresponds to preparing a movement toward that location
(Sheliga, Riggio, & Rizzolatti, 1994), such findings predict that,
when entering a subway car, our attention (labeled from now on
action-related attention) should be allocated toward the seat we
intend to sit on. Yet, if among the passengers, one appears as
signaling potential danger, would our attention be diverted from
our destination? Threat-related stimuli, even when neither physi-
cally salient nor task relevant (distractors), have been shown to
capture both covert (e.g., Yiend & Mathews, 2001) and overt
attention (e.g., Nissens, Failing, & Theeuwes, 2017) and alter
performance of ongoing tasks (Ariga & Arihara, 2018; Fenske &
Eastwood, 2003; Vuilleumier, Armony, Driver, & Dolan, 2001).
Threat-related attentional biases have been demonstrated in a large
number of studies, using various paradigms (e.g., visual search
paradigm and spatial cueing paradigm), stimulus material (e.g.,
photographs of spiders, angry faces, and fear-conditioned shapes)
and response modalities (eye or manual movements). Classically,
these effects are reflected by faster and more accurate detection of
threatening stimuli over nonthreatening ones (e.g., Calvo, Avero,
& Lundqvist, 2006; Eastwood, Smilek, & Merikle, 2001; Fox,
Griggs, & Mouchlianitis, 2007; Öhman, Flykt, & Esteves, 2001;
Pinkham, Griffin, Baron, Sasson, & Gur, 2010; Preciado, Mun-
neke, & Theeuwes, 2017; Soares, Esteves, Lundqvist, & Öhman,
2009).
However, the exact nature and underlying mechanisms of atten-
tion capture by threat-related stimuli are still under debate, as
several studies have either failed to find an effect or found an
opposite one (i.e., attention away from threat; see, e.g., Bar-Haim,
Lamy, Pergamin, Bakermans-Kranenburg, & van IJzendoorn,
2007). One possible explanation for this inconsistency in the
literature is that attention allocation in the presence of threat may
be susceptible to procedural influences, such as timing or the
nature of contrasted cues. Several experiments have indeed ob-
served, using longer presentation times and contrasting threat to
nonthreatening stimuli, that following the initial orienting toward
threat, attention can be the directed away from threat (Juth, Lund-
qvist, Karlsson, & Öhman, 2005; Mulckhuyse, Crombez, & Van
der Stigchel, 2013; Schmidt, Belopolsky, & Theeuwes, 2012).
Moreover, recent experiments have tested attention allocation to
threat in the presence of safety cues. This shift in paradigm is
particularly interesting given that trying to reach safety is the most
adaptive reaction in the face of threat. Using spatial cueing para-
digms while recording saccade trajectories or manual reaction
times (RTs), these studies have demonstrated that attention can be
captured by a stimulus signaling safety when the time between
stimulus and cue presentation was relatively long (Schmidt, Be-
lopolsky, & Theeuwes, 2017; Vogt, Koster, & De Houwer, 2017).
Hence, based on these studies, it appears that attention can be
allocated away from threat when sufficient time is available and/or
if action opportunities that permit reaching safety are present in the
environment. Yet, the question remains as to whether such avoid-
ance responses are the result of mere attentional processes to threat
(“vigilance-avoidance hypothesis”; Mogg, Bradley, Miles, &
Dixon, 2004; Pflugshaupt et al., 2005) or are related to motor
processes (Rizzolatti, Riggio, Dascola, & Umiltá, 1987) and as
such would predict behavioral responses. Here, to better under-
stand how threat- and action-related attentional processes unfold
over time as well as their potential relationship with action choices,
we investigated covert and overt attention allocation in a context
where the source of potential threat and the target for action in that
situation could compete for attentional priority.
To address these questions, we created a paradigm to study
participants’ action selection and attention allocation in a realistic
context by manipulating the presence of action opportunities and
emotional displays. The stimuli represented a room with four seats:
The two middle seats were occupied by two individuals, while the
two outer seats remained empty, thus presenting the opportunity
for two possible actions. Critically, the facial expressions of the
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
2VILAREM, ARMONY, AND GRÈZES

seated individuals were manipulated so that one always displayed
a neutral expression while the other displayed a neutral, fearful, or
angry expression of varying intensity. We conducted four exper-
iments. The first experiment addressed whether different facial
expressions (i.e., anger and fear) differentially influence people’s
spontaneous choices between two competing actions (sitting on
either of the two empty outer seats; Experiment 1). Subsequently,
to investigate whether these threat-related facial expressions have
different effects on attention allocation in the presence or absence
of action possibilities, we conducted two additional experiments,
using a spatial attention design. The first spatial attention experi-
ment (Experiment 2) used the exact same stimuli as described
above, whereas the second one (Experiment 3), by removing the
action-related context, addressed the direct effect of action possi-
bilities on attention allocation. Finally, to assess how attention
allocation relates to action selection processes, we conducted a
final experiment (Experiment 4) that measured participants’ overt
attention allocation by online tracking of their eye movements
while they freely chose where they would like to sit. We predicted
that anger and fear displays, signaling aggression (Sell et al., 2014)
and affiliation (Hammer & Marsh, 2015), respectively, would
favor the selection of actions that avoid angry and approach fearful
individuals, as well as trigger attention to the end point of these
threat-prioritized actions. Crucially, if attention allocation is influ-
enced by threat-prioritized actions, we hypothesized that this mod-
ulatory effect should disappear when action opportunities are
removed from the experimental context (Experiment 3).
Experimental Section
Experiment 1
This first experiment investigated people’s spontaneous choices
between two competing actions (choosing one of the two empty
outer seats) in the presence of threat-related facial expressions
(anger and fear). Participants were asked to freely choose where
they would like to sit, thus avoiding any potential confounding
effects of instructions, arbitrary movements, or response labels
(Laham, Kashima, Dix, & Wheeler, 2015). To indicate their
choice, participants were asked to perform a mouse movement
whose kinematic parameters were continuously collected, allowing
to disentangle different stages of the action process. Using such
continuous measure of processing can provide insightful informa-
tion about choice dynamics by capturing an individual’s potential
attraction to competing response options during a choice task
while on route to the chosen destination (Freeman & Ambady,
2010). Hence, by measuring action choices and movement kine-
matics, this first experiment allowed us to better determine
whether and how two action tendencies can compete for selection
in in the face of threat.
Materials and methods.
Participants. Twenty volunteers (nine males, mean age: 23.0
⫾3.6 years) participated in this experiment. All participants were
right-handed, had normal or corrected-to-normal vision, and had
no history of neurological or psychiatric disorders. The experimen-
tal protocol was approved by INSERM and the local research
ethics committee (Comité de protection des personnes Ile de
France III—Project CO7-28, N° Eudract: 207-A01125-48), and it
was carried out in accordance with the Declaration of Helsinki.
The participants provided informed written consent and were
compensated for their participation. The sample size was calcu-
lated at an alpha of 0.05 and a power of 0.80 using G
ⴱ
Power. We
based our calculation on a study by Fagioli et al. (2007) that
investigated the effect of action selection on attention and obtained
a minimum recommended sample size of 9. The same calculation
applies for Experiments 1, 2, and 4.
Stimuli. The stimuli reproduced a realistic social environment
and consisted of photographs depicting a waiting room with four
seats, where the two middle seats were occupied by two individ-
uals and the two outer seats were empty (see Figure 1a). Pairs’
identities were matched for gender as well as perceived trustwor-
thiness and threat traits (see online supplemental material). Faces
varied in emotion (neutral, angry, or fearful expressions) and in
intensity (four levels of morphs for anger and fear, created from
the neutral to the emotional expression using a simple linear
morphing transformation) and were equalized in perceived emo-
tional intensities (El Zein, Wyart, & Grèzes, 2015). In order to
study the influence of anger and fear independently from each
other and exclude emotion contrast effects (Paulus & Wentura,
2016), one actor of the pair always displayed a neutral expression
while the other displayed a neutral, an angry, or a fearful expres-
sion. The identities, as well as the side of the actor expressing
emotions, were fully counterbalanced. This resulted in 480 trials:
10 pairs ⫻(2 emotional expressions ⫻4 levels of morphs ⫹1
neutral expression ⫻4 repetitions) ⫻2 emotional actor’s iden-
tity ⫻2 emotional actor’s side. Further details about the stimuli are
provided in the online supplemental material.
Experimental procedure. Participants were seated at a dis-
tance of 60 cm from eyes to screen so that the eccentricity to the
central fixation cross was of 4.5 degrees for the center of the faces
and of 8 degrees for the center of the seats. The experiment, ran
using Psychtoolbox implemented in MATLAB R2012b, was de-
signed as follows: A gray screen was presented for 1,000 ms, and
then a fixation cross was superimposed upon the gray screen for
500 ms, followed by the appearance of the scene. Participants
were asked to choose which seat they would like to occupy in the
scene as they maintained fixation on the cross displayed between
the faces throughout the trial. They were given a maximum re-
sponse time of 1,400 ms to indicate their choice by clicking on the
mouse, moving the cursor from the bottom center of the scene to
the chosen seat, and releasing the mouse button. An example of a
trial is depicted in Figure 1a. Participants were requested to make
spontaneous free choices and were told that there were no correct
choices in this task. Nevertheless, they were instructed that their
movements needed to be correctly performed for their responses to
be registered. A correct movement was defined by the release of
the button on one of the seats within 1,400 ms after scene onset and
was signaled to the participants by the appearance of their portrait
(taken prior to the experiment), superimposed on the scene at the
release location, for 300 ms (the maximum scene duration for
correct movements was thus 1,400 ms ⫹300 ms). If the button
release was not made on the seat or did not occur within 1,400 ms,
the trial was considered incorrect. The end of the trial was trig-
gered either by the mouse release or 1,400 ms after scene onset in
the absence of a response.
Participants underwent training until their accuracy (proportion
of correct movements) in the task reached 60% and then completed
the experiment. They were informed of their percentage of correct
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
3
ACTION- AND ATTENTION-SELECTION UNDER THREAT

responses at the end of each block and were asked to do their best
to maximize it.
Data preparation and statistical analyses. We rejected in-
valid trials including nonresponses, incorrect responses (trials
where the movement was not correctly performed), and RTs faster
than 250 ms. Our variables of interest being movement direction
and kinematics, we screened for outliers on movement accuracy
and excluded data from two participants (lower than the mean
accuracy minus two standard deviations). Moreover, we screened
for outliers on anxiety, because it is known to impact attention
allocation to threat (for a review, see Grupe & Nitschke, 2013),
and excluded two participants based on the State–Trait Anxiety
Inventory (STAI) administered before the study (score higher than
the mean plus two standard deviations). We therefore analyzed the
data of 17 participants for this experiment (one participant exhib-
ited both poor performance and a high STAI score).
Participants’ responses were first analyzed by running a
repeated-measures analysis of variance (ANOVA)—with Emotion
(Anger and Fear), Side (Away and Toward), and Level (four levels
of morphs) as within-subjects factors—on the proportion of trials
in which they decided to sit next to the neutral or emotional
individual. We also conducted a raw time analysis using a general
linear regression model on peak velocity measures, which were
extracted by generating 60 time bins along the original (nonnor-
malized) movement time, between 0 ms and 1,200 ms, and com-
puting the peak velocity of each trajectory. For comparisons be-
tween conditions, repeated-measures ANOVAs were performed on
the proportion of choice or on the parameter estimates of the
regression, and a Greenhouse-Geisser correction was applied when
the assumption of sphericity was not met. When appropriate, post
hoc analyses with paired Student’s ttests were computed. All data
are expressed as mean (M) and 95% confidence intervals (CIs).
Partial eta squared (␩p
2) is reported as the effect size of the F
statistics and Cohen’s d(d) as the effect size of the tstatistics. A
value of ␩p
2⫽0.1/d⫽0.2 represents a small effect size, ␩p
2⫽
0.06/d⫽0.5 a medium one, and over ␩p
2⫽0.14/d⫽0.8 a large
effect size. Statistical analyses were performed using MATLAB
R2012b and PASW Statistics 18.
Results.
Impact of task-irrelevant emotional displays on action choices
under visual fixation. Overall, participants chose more often the
seat located away from the emotional individual (“away” and
“toward” refer to the neutral and emotional actor side, respec-
tively), F(1, 16) ⫽10.37, p⫽.005, ␩p
2⫽0.39. Moreover, the level
of intensity of the emotional display increased this tendency to
avoid the emotional individual, F(2.04, 32.59) ⫽6.85, p⫽.003,
␩p
2⫽0.30, Greenhouse-Geisser corrected. However, there was an
interaction between Emotion and Side, F(1, 16) ⫽14.90, p⫽.001,
␩p
2⫽0.48: Participants significantly chose more often to sit away
from the angry individuals than next to them but sat equally frequently
Figure 1. Action-related decisions (Experiment 1). (a) Time course of a trial where participants have to indicate
where they would like to sit by moving their cursor from the bottom center to the chosen seat. The face of the
participant appeared for 300 ms after the offset of the movement. Please note that the pair of identities displayed
on the stimuli was not used in this experiment and was selected for illustration purposes only, according to
Radboud Faces Database permission. (b) Emotion ⫻Side interaction on the proportion of choice. c) Emotion ⫻
Side interaction on the peak velocity (a.u.).
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01. ns,p⬎.1. For illustration purposes, error bars
represent within-subject standard errors. See the online article for the color version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
4VILAREM, ARMONY, AND GRÈZES

away from and next to the fearful individuals (see Figure 1b). All
other paired comparisons of the Emotion ⫻Side interaction were
significant (see Table 1). No other significant main effects or inter-
actions were found (all Fs⬍2.10, all ps⬎.10, all ␩ps
2⬍0.12). We
ran a replication of Experiment 1 in a larger sample (n⫽40; see
online supplemental materials) and replicated these results: Partici-
pants displayed a clear preference for away as compared to toward
choices for angry faces and no significant preference for away as
compared to toward choices for fearful faces (Emotion ⫻Side inter-
action F(1, 39) ⫽14.064, p⫽.001, ␩p
2⫽0.265).
Impact of task-irrelevant emotional displays on movements’
peak velocity. To better characterize the distinct influence of
fearful and angry displays on action choices, we further investi-
gated the action-decision process by continuously recording
mouse-tracking data (custom-made MATLAB program). We com-
puted several kinematic measures (initiation times, movement
duration, maximum deviation, and peak velocity) using the Ana-
lyzer tool from the Mouse-Tracker software package (Freeman &
Ambady, 2010). A parametric regression-based approach was ad-
opted, consisting of regressing kinematic measures against the
intensity of the displayed emotion (El Zein et al., 2015; Wyart,
Myers, & Summerfield, 2015). For each analysis, a general linear
regression model was used where emotion intensity was intro-
duced as a trial-per-trial predictor of kinematic measures. The
corresponding parameter estimates of the regression, reported in
arbitrary units, were calculated for each participant and then av-
eraged across participants to produced group-level averages. We
calculated the parameter estimates separately for our conditions of
interest (AA ⫽Anger Away, AT ⫽Anger Toward, FA ⫽Fear
Away, FT ⫽Fear Toward) and tested main effects on the corre-
sponding intercepts. We also tested for a main effect of Emotion
by comparing the neutral condition to anger and fear pooled
together.
We observed a significant Emotion ⫻Side interaction, F⫽
6.81, p⫽.02, ␩p
2⫽0.30, on peak velocity intercepts thought to
reflect choice confidence (Palser, Fotopoulou, & Kilner, 2015),
showing that participants made quicker movements going toward
fearful individuals than going away from them (FA vs. FT:
t(16) ⫽⫺2.61, p⫽.02, d⫽⫺0.08). The difference between AA
and AT did not reach significance, t(16) ⫽1.34, p⫽.20, d⫽0.08
(see Figure 1c). No other paired comparisons (all ts⬍1.66, all
ps⬎.12, d⬍0.09) or main effects were significant (all Fs⬍0.06,
all ps⬎.81, all ␩ps
2⬍0.004).
The other kinematic measures showed that (a) threat-related
displays (both anger and fear), compared to neutral ones, signifi-
cantly reduced the time required to launch a given action plan,
indicating a speeded action selection mechanism when facing
danger, and (b) the selection process took place before movement
initiation, as indicated by an absence of effects on the maximum
deviation and its negative correlation with initiation time. For
detailed analyses and statistics, please refer to the online supple-
mental material.
Discussion 1. This first experiment revealed that angry and
fearful expressions differentially shape the map of potential com-
peting actions provided by the environment, by favoring the se-
lection of actions that avoid angry and approach fearful individu-
als. Compared with existing stimulus-response compatibility
paradigms, which yielded mixed results (i.e., both avoidance and
approach tendencies to angry and fearful expressions depending on
the context; Bossuyt et al., 2014; Carver & Harmon-Jones, 2009;
Krieglmeyer & Deutsch, 2013; Marsh et al., 2005; Paulus &
Wentura, 2016; Wilkowski & Meier, 2010), our paradigm was
ecologically valid (free choice task) and methodologically relevant
(recording of kinematics). Our results revealed that the presence of
task-irrelevant angry faces favored the selection of avoidance
responses. For fearful faces, while neither approach not avoidance
tendency was evidenced by participants’ choices, their level of
confidence in their choice, reflected by peak velocity (Palser et al.,
2015), was higher when approaching fearful individuals. Even
though our data do not allow to uncover the motives associated
with action choices, they reveal that threat-related displays influ-
ence action-related decisions by shaping existing maps of potential
actions.
Experiment 2
Studies on attentional biases have demonstrated that threat
quickly and efficiently captures attention (yet see the beginning of
this article for a discussion of inconsistencies in the literature).
However, consecutively to this initial orienting toward threat, it
has been shown that attention can be allocated away from threat
when sufficient time is available and/or toward safety cues if
action opportunities allowing to reach safety are provided by the
environment (Schmidt et al., 2017; Vogt et al., 2017). These
attentional responses could either result from mere attentional
processes to threat (“vigilance-avoidance hypothesis”; Mogg et al.,
Table 1
Paired Differences and Statistics of the Paired Comparisons of the Emotion-by-Side Interaction
on Choice Proportions (%) in Experiment 1
Proportion of choice (%)
Paired differences
tddl pvalue Cohen’s dMean 95% CI
AA-AT 6.08 [3.14, 9.02] 4.06 16 .001 1.95
FA-FT 0.07 [⫺1.67, 1.81] 0.09 16 .936 0.04
AA-FA 3.26 [1.65, 4.87] 3.97 16 .001 1.25
AT-FT ⫺2.75 [⫺4.32, ⫺1.18] ⫺3.42 16 .004 ⫺1.07
AA-FT 3.33 [1.35, 5.31] 3.30 16 .005 1.27
AT-FA ⫺2.83 [⫺4.72, ⫺0.95] ⫺2.95 16 .009 ⫺1.11
Note. The abbreviations refer to the choice of the subjects with respect to the emotional actor: AA ⫽Anger
Away; AT ⫽Anger Toward; FA ⫽Fear Away; FT ⫽Fear Toward. “Away” refers to the opposite side of the
emotional actor (i.e., side of the neutral actor) and “toward” to the side of the emotional actor.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
5
ACTION- AND ATTENTION-SELECTION UNDER THREAT

2004; Pflugshaupt et al., 2005) or be related to motor processes
(e.g., the preparation of a motor response toward a safety cue). In
order to better understand the nature of these attentional responses,
our second experiment investigated attention allocation in the
presence of both threat-related information and action opportuni-
ties, using the same scenes as in Experiment 1. In this task,
participants had to report the orientation of a target (upright or
inverted “T”) appearing briefly on either outer seat as quickly as
possible (see Figure 2a). Importantly, the cue appeared 200 ms
after the scene onset and was presented for 400 ms. Sufficient time
was therefore available for attention to be reallocated in the scene,
after the initial orienting toward threat. Thus, if late biases away
from threat result from a mere reorientation of attention
(“vigilance-avoidance hypothesis”), participants should be quicker
at detecting the target when located away from threat, irrespective
of the nature of the expression (for both anger and fear). Yet, if
these biases are related to the preparation of a motor response,
participants should be quicker at detecting the target when located
at the end point of the preferred action, that is, away from anger.
In the case of fear, as no clear preference was evidenced between
away and toward choices, an absence of bias for detecting a target
away or toward emotional faces should be observed; however, as
participants’ level of confidence in their choice (peak velocity)
was higher when approaching fearful individuals, participants may
be quicker at detecting the target when located toward fear.
Materials and methods.
Participants. Twenty-five volunteers (13 males, mean age:
22.4 ⫾2.3 years) took part in this experiment. All participants
were right-handed, had normal or corrected-to-normal vision, and
had no history of neurological or psychiatric disorders. The exper-
imental protocol was approved by INSERM and licensed by the
local research ethics committee (Comité de protection des per-
sonnes Ile de France III—Project CO7-28, N° Eudract: 207-
A01125-48) and carried out in accordance with the Declaration of
Helsinki. The participants provided informed written consent and
were compensated for their participation.
Stimuli. The stimuli were identical to Experiment 1.
Experimental procedure. The experimental trials were as fol-
lows: After a 100-ms gray screen, the scene appeared for 200 ms;
then a target (an upright or an inverted “T”) was superimposed on
the scene for 400 ms at a fixed location on either outer seat. The
scene was then replaced by a gray screen for 800 ms. Participants
had a maximum of 1,200 ms (400-ms cue presentation ⫹800-ms
gray screen) from the onset of the target to report whether the “T”
Figure 2. Spatial attention studies. (a, b) Experiment 2. (a) Time course of a trial where participants have to
report the orientation of the “T” appearing on either outer seat by pressing the corresponding key. The mapping
between the responses and the keys was counterbalanced between participants. Please note that the pair of
identities displayed on the stimuli was not used in this experiment and was selected for illustration purposes only,
according to Radboud Faces Database permission. (b) Emotion ⫻Side interaction on the reaction times (RTs).
(c) Experiment 3. On the left, an example stimulus where the scene is covered by a gray mask in order to hide
action-relevant information. On the right, the Emotion ⫻Side interaction of RTs. ⬃p⬍.1.
ⴱ
p⬍.05.
ⴱⴱ
p⬍
.01. ns,p⬎.1. For illustration purposes, error bars represent within-subject standard errors. See the online article
for the color version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
6VILAREM, ARMONY, AND GRÈZES

was upright or inverted by pressing the corresponding key, regard-
less of the side where the “T” appeared. The trial ended after the
keypress or at the end of the 800-ms gray screen in the absence of
a response. Importantly, participants had to answer while focusing
on a fixation cross displayed between the faces throughout the
trial. That is, they were requested to use their peripheral vision to
discriminate the “T” and were instructed not to look at the sur-
rounding scene. The mapping between the responses (upright or
inverted) and the keys was counterbalanced between participants.
Participants underwent training until their accuracy (proportion
of correct detections of an upright or an inverted “T”) in the task
reached 60% and then completed the experiment. They were
informed of their percentage of correct responses at the end of each
block and were asked to maximize it.
Data preparation and analyses. We rejected invalid trials
including nonresponses, incorrect responses (trials where the cue
was not correctly detected), and RTs faster than 250 ms. We found
no outliers on either RTs or anxiety. However, data of one partic-
ipant were excluded due to a misunderstanding of instructions (no
maintenance of gaze on the fixation cross). We therefore analyzed
the data of 24 participants for this study. Spatial attentional biases
were calculated by comparing RTs when the target was displayed
toward or away from the emotional individual using repeated-
measures ANOVAs.
Results.
Impact of task-irrelevant emotional displays on spatial atten-
tion in the presence of action possibilities. Overall, participants
reported the orientation of the target significantly above chance,
M⫽75%, 95% CI [71, 79], t(23) ⫽13.01, p⬍.001, d⫽2.66.
Of interest here, attention biases were tested by performing an
Emotion ⫻Side ⫻Level repeated-measures ANOVA on RTs,
which revealed a significant Emotion ⫻Side interaction, F(1,
23) ⫽9.50, p⫽.005, ␩p
2⫽0.29. Paired comparisons of interest
(AA-AT and FA-FT) revealed that subjects were faster to discrim-
inate the “T” when it appeared next to the fearful individuals rather
than away from them. In the case of anger, although the effect was
in the predicted direction (i.e., trend for a quicker detection when
the “T” appeared away from the angry individual), the comparison
did not reach significance (see Figure 2b and Table 2). No other
main effects or interactions were significant (all Fs⬍2.7, all ps⬎
.11, all ␩ps
2⬍0.11).
Discussion 2. This second experiment demonstrated that an-
gry and fearful displays have a differential impact on attention
allocation. Along with the action choices observed in Experiment
1, these findings suggest that attention is allocated to the space of
the scene corresponding to the end point of the action prioritized
by anger and fear displays, respectively, that is, away from anger
and toward fear. This differential impact of threat-related displays
on attention allocation might depend upon the presence of action
opportunities in the scene and on the concurrent preparation of a
motor response away from anger or toward fear. The present
results would thus support that an action-related attentional bias
emerged consecutively to the initial orienting toward threat. How-
ever, such interpretation would need to be corroborated by findings
showing that the pattern of attention allocation to threat is modi-
fied when action opportunities are removed from the environment.
Experiment 3
To directly test whether the observed effects in Experiment 2
were related to the presence of action opportunities, we conducted
a third experiment in which we removed the action-related context
from the scene. The design and the task were thus identical to
Experiment 2, except that, crucially, we superimposed a mask on
the scene to hide any information susceptible to being associated
with action opportunities. The only visible parts of the original
scene were the faces of the two central individuals and the two
locations where the “T” appeared on the outer seats (see Figure
2c). If the presence of action opportunities indeed impacted on
attention allocation, removing them from the environment should
impact attention orienting to threat and possibly lead to a general
bias toward or away from threat (“vigilance-avoidance hypothe-
sis”; Mogg et al., 2004; Pflugshaupt et al., 2005) for both angry
and fearful expressions.
Materials and methods.
Participants. Twenty-seven volunteers (11 males, mean age:
23.0 ⫾3.1 years) took part in this experiment. All participants
were right-handed, had normal or corrected-to-normal vision, and
had no history of neurological or psychiatric disorders. The exper-
imental protocol was approved by INSERM and licensed by the
local research ethics committee (Comité de protection des per-
sonnes Ile de France III—Project CO7-28, N° Eudract: 207-
A01125-48) and carried out in accordance with the Declaration of
Helsinki. The participants provided informed written consent and
were compensated for their participation. With Experiment 3 being
a modified version of Experiment 2, we based our sample size
calculation on that previous study.
Stimuli. The stimuli were identical to Experiment 1.
Table 2
Paired Differences and Statistics of the Paired Comparisons of the Emotion-by-Side Interaction
on Reaction Times (ms) in Experiment 2
Reaction times (ms)
Paired differences
tddl pvalue Cohen’s dMean 95% CI
AA-AT ⫺9.18 [⫺19.25, 0.89] ⫺1.79 23 .087 ⫺.09
FA-FT 8.36 [1.13, 15.59] 2.26 23 .033 .08
AA-FA ⫺11.65 [⫺20.25, ⫺3.05] ⫺2.66 23 .014 ⫺.12
AT-FT 5.89 [⫺1.05, 12.83] 1.67 23 .109 .06
AA-FT ⫺3.30 [⫺12.61, 6.01] ⫺0.70 23 .494 ⫺.03
AT-FA ⫺2.47 [⫺9.90, 4.96] ⫺0.65 23 .522 ⫺.03
Note. Abbreviations refer to the location of the “T.” For definitions of other abbreviations, see Table 1.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
7
ACTION- AND ATTENTION-SELECTION UNDER THREAT

Experimental procedure. The task and design were identical
to Experiment 2, except that a mask was superimposed on the
scene to hide all the information susceptible to be associated with
action opportunities action, leaving only the faces visible (see
Figure 2c).
Participants underwent training until their accuracy (proportion
of correct detections of an upright or an inverted “T”) in the task
reached 60% and then completed the experiment. They were
informed of their percentage of correct responses at the end of each
block and were asked to maximize it.
Data preparation and analyses. We rejected invalid trials
including nonresponses, incorrect responses (trials where the cue
was not correctly detected), and RTs faster than 250 ms. We found
two outliers based on their mean RTs (lower than the mean
accuracy minus two standard deviations) but no outliers on anxi-
ety. We therefore analyzed the data of 25 participants for this
study. Spatial attentional biases were calculated by comparing RTs
when the cue was displayed toward or away from the emotional
individual, using repeated-measures ANOVAs.
Results.
Impact of task-irrelevant emotional displays on spatial atten-
tion in the absence of action possibilities. Participants reported
the orientation of the target significantly above chance, M⫽73%,
95% CI [69, 77], t(24) ⫽12.09, p⬍.001, d⫽2.42.
To test whether the Emotion ⫻Side effect found in Experiment
2 was replicated in this setting, we performed a repeated-measures
ANOVA on RTs with Emotion and Side as within-subjects factors
and Study as a between-subjects factor. Of importance here,
independent-samples ttests indicated that participants from Ex-
periment 2 and Experiment 3 did not differ on mean RTs (respec-
tively, M⫽608 ms, 95% CI [569, 647]; M⫽625 ms, 95% CI
[589, 662]; t(47) ⫽0.66, p⫽.51, d⫽0.19) or accuracy levels
(respectively, M⫽75%, 95% CI [72, 79]; M⫽73%, 95% CI [70,
77]; t(47) ⫽⫺0.78, p⫽.44, d⫽0.22). Supporting our hypothesis,
the ANOVA revealed a significant Emotion ⫻Side ⫻Study
interaction, F(1, 47) ⫽7.72, p⫽.008, ␩p
2⫽0.14. Contrary to
Experiment 2, the Emotion ⫻Side interaction was no longer
significant in Experiment 3, F(1, 24) ⫽0.97, p⫽.33, ␩p
2⫽0.04,
suggesting that the differential effect of emotional displays on
attention allocation required the presence of action possibilities in
the scene (see Figure 2c). Indeed, in their absence, a significant
main effect of Side, F(1, 24) ⫽4.73, p⫽.04, ␩p
2⫽0.16, replicated
a classical avoidance bias (e.g., Koster, Crombez, Verschuere, Van
Damme, & Wiersema, 2006), such that the RTs were decreased
when the cue was presented away from the emotional individual,
irrespective of the nature of the emotional display. No significant
main effect of Emotion was found, F⫽0.08, p⫽.78, ␩p
2⫽0.003.
Discussion 3. Results from Experiments 2 and 3 revealed a
different pattern of attention allocation depending upon the
presence or absence of action opportunities in the environment:
Attention was allocated away from threat in the absence of
action possibilities to reach safety (threat-related attentional
processes) but oriented accordingly to the preferred action when
such action was possible (action-related attentional processes).
However, the question remained as to when and how threat-
related displays and the best target for action compete for
attentional priority.
Experiment 4
Experiments 1–3 demonstrated that angry and fearful facial
expressions differentially impacted on action selection and covert
attention allocation but only when the environment provided ac-
tion opportunities. In order to understand how attention allocation
unfolds over time in the presence of both threat-related displays
and action opportunities, we ran our action-related decisions par-
adigm while recording a behavioral marker of overt attention
allocation—namely, saccadic activity. We hypothesized that atten-
tion would be first quickly oriented to threat and then reoriented
accordingly to the action that was being planned.
Materials and methods.
Participants. Twenty-two volunteers (eight males, mean age:
22.9 ⫾3.2 years) took part in this experiment. All participants
were right-handed, had normal or corrected-to-normal vision, and
had no history of neurological or psychiatric disorders. The exper-
imental protocol was approved by INSERM and licensed by the
local research ethics committee (Comité de protection des per-
sonnes Ile de France III—Project CO7-28, N° Eudract: 207-
A01125-48) and carried out in accordance with the Declaration of
Helsinki. The participants provided informed written consent and
were compensated for their participation.
Experimental procedure. The procedure was identical to Ex-
periment 1, except that participants were allowed to visually ex-
plore the scene in order to record their saccadic behavior (see
Figure 3a). Before the experiment, participants rested their head on
a chinrest, and their eye movements were calibrated using a
standard 9-point grid. During the experiment, they were asked to
fixate the central cross at the beginning of each trial and then
decide where to sit and make the movement to reach the seat, as
soon as possible, while freely exploring the scene.
Participants underwent training until their accuracy (proportion
of correct movements) in the task reached 60% and then completed
the experiment. They were informed of their percentage of correct
responses at the end of each block and were asked to maximize it.
Data preparation and analyses. We rejected invalid trials
including nonresponses, incorrect responses (trials where the
movement was not correctly performed), and RTs faster than 250
ms. We found two outliers on movement accuracy (lower than the
mean accuracy minus two standard deviations) and one outlier on
anxiety (higher than the mean plus two standard deviations). We
therefore analyzed the data of 19 participants for this study. Par-
ticipants’ action-related decisions were analyzed as in Experiment
1. Saccadic activity was recorded monocularly (left eye for all
subjects) at 1,000 Hz using an Eyelink 1000 (SR Research, Mis-
sissauga, ON, Canada) with a level desktop camera. A saccade was
defined as the first time point at which the velocity exceeded 30°/s
and the acceleration exceeded 8,000°/s
2
. The data related to the
saccadic behavior were cleaned by excluding saccades data on
incorrect trials and microsaccades (inferior to 1 degree of visual
angle; see Martinez-Conde, Macknik, Troncoso, & Hubel, 2009).
Our scene was divided into five regions of interest (ROIs; the cross
area, each actor, and each outer seat). Importantly, because our
previous experiments showed that emotion intensity was not a
critical factor in any of our tasks, we decided to pool all the
intensity levels together in order to maximize the number of
observations per condition of interest. We ran repeated-measures
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
8VILAREM, ARMONY, AND GRÈZES

ANOVAs on the data related to saccade latencies and directions
using Emotion and Side as within-subject factors.
In order to assess whether saccadic behavior (notably the second
saccade) could be a trial-per-trial predictor of action choices, we
used generalized linear mixed models that account for repeated
measures with a logit link function, as implemented in the lme4
package in RStudio (v 1.1–14; Bates, Mächler, Bolker, & Walker,
2015). We included Saccade Direction (away or toward emotion)
and Emotion (anger or fear) as predictors and Choice (away or
toward) as the dependent variable. All models were run with a
random intercept per participant and all main effects and interac-
tions as fixed effects. We compared four models: A null model was
built using intercept and the random factor (mixed logit Model 1:
choice ⬃1⫹(1 | participant)). Model 2 included the main effect
of saccade direction only (choice ⬃Saccade Direction ⫹(1 |
participant)), Model 3 incorporated both main effects of saccade
direction and emotion (choice ⬃Saccade Direction ⫹Emotion ⫹
(1 | participant)), and Model 4 included both main effects and the
interaction between the direction of the saccade and emotion
(choice ⬃Saccade Direction ⫹Emotion ⫹Saccade Direction ⫻
Emotion ⫹(1 | participant)). We extracted the Akaike weights (w)
of each model from the second-order Akaike information criterion
values, using dedicated functions of the package MuMIn (Barto´
n,
2018). Models were then ranked according to their Akaike weights
(w), corresponding to their probability of being the best model
among the candidate models (Wagenmakers & Farrell, 2004).
Results.
Impact of emotional displays on action choices under visual
exploration. Overall, participants chose more often the seat lo-
cated away from the emotional individual (“away” and “toward,”
respectively, refer to the neutral and emotional actor side, respec-
tively), as revealed by a main effect of Side, F(1, 18) ⫽47.33, p⬍
.001, ␩p
2⫽0.72. Importantly, however, the nature of the displayed
emotion influenced participants’ choice (Emotion ⫻Side interac-
tion, F(1, 18) ⫽6.97, p⫽.02, ␩p
2⫽0.28): Angry displays elicited
more away responses than fearful ones. All the paired comparisons
of the Emotion ⫻Side interaction were significant (see Figure 3b
and Table 3). The main effect of emotion was not significant, F(1,
18) ⫽0.17, p⫽.68, ␩p
2⫽0.01.
Impact of emotional displays on saccadic activity. First, we
analyzed the latency of the first saccades as an index of attentional
capture (Jiang, Won, & Swallow, 2014). Importantly, we excluded
saccades whose starting point was not on the fixation cross, as it
was required by the task. Our data revealed that the latency of the
first saccades, which was of 129 ms on average (95% CI [116,
142]), was neither affected by the nature of emotional displays nor
by the side where they were presented (all Fs⬍1.50, ps⬎.24,
␩ps
2⬍0.08).
Figure 3. (a) Time course of a trial where participants have to indicate where they would like to sit by moving
their cursor from the bottom center to the chosen seat. The face of the participant appeared for 300 ms after the
offset of the movement. Please note that the pair of identities displayed on the stimuli was not used in this
experiment and was selected for illustration purposes only, according to Radboud Faces Database permission.
(b) Emotion ⫻Side interaction on the proportion of choice. (c) Emotion ⫻Side interaction on the proportion
of second saccades landing sites.
ⴱ
p⬍.05.
ⴱⴱⴱ
p⬍.001. For illustration purpose, error bars represent
within-subject standard errors. See the online article for the color version of this figure.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
9
ACTION- AND ATTENTION-SELECTION UNDER THREAT

We then explored the distribution of the landing saccades in
order to better understand attention allocation over the time course
of a trial. Using the ROI approach, we observed that the first
saccade was predominantly landing on the faces (89.67%; 95% CI
[86.40, 92.94]), while a small subset of saccades was directed
toward the seat ROIs (1.44%; 95% CI [0.42, 2.46]), and the
remaining ones remained within the cross ROI. Because saccades
directed toward the seats were susceptible to be informative re-
garding action selection and were too limited to be analyzed
separately (data were missing for some subjects who did not make
any first or second saccades toward the seats in some conditions),
we pooled the saccades landing on face or seat for each side and
ran an Emotion ⫻Side ANOVA on these proportion of saccades.
We found that the first saccade was mostly directed toward the
hemifield with the emotional face, M⫽53%, 95% CI [51, 56],
F(1, 18) ⫽7.06, p⫽.02, ␩p
2⫽0.28. No other effects were
significant (all Fs⬍0.07, ps⬎.80, ␩ps
2⬍0.004).
The second saccade, with a mean latency of 492 ms (95% CI
[437, 547]), was also landing mostly on the faces (88.28%; 95% CI
[83.16, 93.39]) and in a lesser proportion on the seats (5.05%; 95%
CI [0.36, 9.73]).
1
We observed that the second saccade was mostly
directed toward the hemifield with the neutral individual, M⫽
56%, 95% CI [54, 58], F(1, 18) ⫽36.35, p⬍.001, ␩p
2⫽0.67.
However, we found a significant Emotion ⫻Side interaction, F(1,
18) ⫽7.47, p⫽.01, ␩p
2⫽0.29, such that there were more saccades
away than toward anger, compared to fear (see Figure 3c). No
main effect of emotion was observed, F(1, 18) ⬍0.001, p⫽.99,
␩p
2⬍0.001.
Impact of saccadic activity on the choice. In order to assess
whether the interaction between Emotion and Side observed on the
direction of the second saccade could significantly predict the
choice in a trial-by-trial fashion, we performed logistic regres-
sions. The best model to account for the data incorporated the
interaction between Saccade Direction and Emotion. This model
was better than the other competing models by at least a factor of
2.9 (Model 4 Akaike weights ⫽0.743 vs. Model 3 ⫽0.257 vs.
Models 2 and 1 ⫽0.007). The interaction term indicated that an
increased number of saccades away from the emotional individual
in the anger versus fear condition were associated with a higher
proportion of away choices, in line with the results observed on the
proportion of choice (see Table 4 and online supplemental Table
S2 for model selection).
Discussion 4. This experiment explored the unfolding of at-
tention allocation to threat while participants completed an action-
related decision task. Overt attention allocated to faces, as indexed
by the first saccade, occurred as early as 130 ms and was prefer-
entially directed toward emotional displays, thereby replicating
superiority effects of emotional compared to neutral displays on
saccadic behaviors (Mulckhuyse et al., 2013; Schmidt et al., 2012;
Schmidt, Belopolsky, & Theeuwes, 2015). Furthermore, and in
agreement with the idea that distractor-related activity might be
strengthened by threatening information, resulting in more, but not
necessarily faster, oculomotor capture (Mulckhuyse & Dalmaijer,
2016), we did not observe faster saccades toward emotional dis-
plays. Finally, no difference between angry and fearful displays on
the latency or the direction of the first saccade was observed,
suggesting that this early pop-out effect was sensitive to the
presence of emotional stimuli but not to the nature of the threat-
related signal.
Yet, while the second saccades were predominantly directed
away from the emotional individual, this effect was significantly
more pronounced for angry displays compared to fearful ones.
Crucially, our data indicated that the landing site of the second
saccade predicted the seat subsequently chosen by the subject,
although, on average, the second saccade average latency preceded
movement onset (respectively 492 and 505 ms).
Thus, our data show that the first saccades were mostly directed
toward the emotional individual while the second saccades were
mostly directed away from the emotional individual. Such results
could support the “vigilance-avoidance hypothesis,” where the
quick vigilance toward threat cues is followed by an avoidance
response at longer stimulus durations (Mogg et al., 2004; Pflug-
shaupt et al., 2005). Nevertheless, given our results showing that
the second saccade distribution pattern is more strongly impacted
by angry as compared to fearful displays, as well as predicts
upcoming action choice, we propose that our results corroborate
the view that attention allocation is also influenced by an action-
related bias reflecting the concurrent selection of an adaptive
1
Note that among the second saccades that landed on a face/seat ROI,
78.1% originated from the other face/seat ROI, 18.9% from the same face,
and 0.03% from the cross ROI.
Table 3
Paired Differences and Statistics of the Paired Comparisons of the Emotion-by-Side Interaction
on Choice Proportions (%) From Experiment 4
Proportion of choice (%)
Paired differences
tddl pvalue Cohen’s dMean 95% CI
AA-AT 18.13 [12.56, 23.70] 6.84 18 ⬍.001 3.12
FA-FT 14.16 [9.40, 18.91] 6.26 18 ⬍.001 2.85
AA-FA 2.10 [0.43, 3.75] 2.64 18 .02 0.39
AT-FT ⫺1.88 [⫺3.56, ⫺0.21] ⫺2.36 18 .03 0.35
AA-FT 16.25 [11.21, 21.28] 6.78 18 ⬍.001 2.97
AT-FA ⫺16.04 [⫺20.92, ⫺11.16] ⫺6.90 18 ⬍.001 3.01
Note. The abbreviations refer to the choice of the subjects with respect to the emotional actor: AA ⫽Anger
Away; AT ⫽Anger Toward; FA ⫽Fear Away; FT ⫽Fear Toward. “Away” refers to the opposite side of the
emotional actor (i.e., side of the neutral actor) and “toward” to the side of the emotional actor.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
10 VILAREM, ARMONY, AND GRÈZES

behavioral response and thus in agreement with the results from
Experiments 1 and 2.
Meta-Analytic Analyses
One limitation of the present study is the relatively small size of
the samples used in our experiments. In order to test the reliability
of our effects of interest, we conducted a meta-analytic analysis on
our results (Emotion ⫻Side interaction on action choices in
Experiment 1, replication of Experiment 1 and Experiment 4, and
RTs in Experiment 3). Following the procedure described in Open
Science Collaboration (2015), we first transformed the Fstatistics
into correlation coefficients in order to take into account both
choice and RT studies. We then conducted a fixed-effect meta-
analysis using the R-package Metafor on Fisher-transformed cor-
relations. We found a significant meta-analytic effect: mean esti-
mate ⫽0.552, 95% CI [0.343, 0.760], p⬍.0001, showing that,
although our sample sizes were relatively small, our effect was
highly reliable. In a second step, we tested the reliability of our
effects of interest for angry and fear faces separately (pairwise test
between toward and away on action choices in Experiment 1,
replication of Experiment 1 and Experiment 4, and RTs in Exper-
iment 3). For both angry and fearful faces, we found a significant
meta-analytic effect (angry: mean estimate ⫽0.626, 95% CI
[0.417, 0.835], p⬍.0001; fearful: mean estimate ⫽0.368, 95% CI
[0.159, 0.577], p⫽.0006), yet the mean estimate for angry faces
was 1.7 times bigger than the mean estimate for fearful faces.
General Discussion
We conducted four experiments to assess the impact of task-
irrelevant social threat displays on action and attentional re-
sponses. In line with our hypothesis that the social function of
threat-related expressions should distinctly influence action selec-
tion, the first experiment revealed that angry and fearful expres-
sions differentially shape the map of potential competing actions
provided by the environment. The presence of task-irrelevant
angry faces favored the selection of avoidance responses, while in
the presence of fearful faces, participants were more confident
when approaching fearful individuals. The second study demon-
strated that, in the presence of action opportunities, attention is
allocated to the space of the scene corresponding to the end point
of the action prioritized by anger and fear displays in the first
experiment. Although findings for the first two experiments do not
allow drawing conclusions regarding the causal relationship be-
tween action selection and attention allocation, our third experi-
ment confirmed this interpretation by revealing that the effects of
anger and fear on attention allocation were absent when action
possibilities were removed from the scene. Finally, our last exper-
iment revealed that a fast attentional allocation toward threat-
related displays was followed by a reorientation of attention to-
ward the end point of the selected action, before movement
initiation. Together, these findings demonstrate that, in a realistic
context offering competing targets for action, action selection
processes and attention allocation are influenced by threatening
stimuli, according to their social function.
Angry and fearful faces influenced both action- and attention-
related decisions to different degrees. Specifically, angry individ-
uals mainly favored the selection of actions that avoided them
when paired with another neutral individual. Although some stud-
ies have shown that anger can be associated with approach behav-
iors, but mostly in the emitter (e.g., Krieglmeyer & Deutsch,
2013), or depending on the contrasted emotion (Paulus & Wentura,
2016), our data indicate that angry faces, by enhancing cues of
strength and communicating the probability to be aggressed (Sell
et al., 2014), primarily signal a direct threat to the observers and
prompt them to increase distance with the source of the threat
(Sander et al., 2007).
In contrast, the results for fearful expressions are less clear.
Indeed, participants were more confident (as indexed by peak
velocity in Experiment 1) when approaching fearful individuals
and allocated their attention to the spatial area next to these
individuals (Experiment 2). However, when participants were able
to visually explore the scene (Experiment 4), we observed a
preference for avoidant responses on both attention allocation and
action selection, although the proportion of away responses was
smaller than in response to anger (Kaltwasser, Moore, et al., 2017).
These mixed findings may be related to the fact that fearful faces
communicate both threat and distress and as such could either
favor avoidant or approach responses. Fearful displays, by signal-
ing the presence of a potential danger in the environment (Ander-
son, Christoff, Panitz, De Rosa, & Gabrieli, 2003; Springer, Rosas,
McGetrick, & Bowers, 2007), should elicit avoidance responses
(as in Paulus & Wentura, 2014). Yet, affiliation has also been
suggested to be a primitive response to danger (Mawson, 2012),
and grouping is an old evolutionary strategy to cope with per-
ceived risk of predation (Isbell, 1994). Therefore, by communicat-
ing distress (e.g., Hammer & Marsh, 2015), fearful displays may
also activate a concern mechanism in the observer (Batson, Fultz,
& Schoenrade, 1987; Nichols, 2001), leading to approach behav-
iors. Interestingly, it has been suggested that interindividual dif-
ferences in prosocial motivations may explain why some studies
observed avoidance (Paulus & Wentura, 2014) while others ob-
served approach (as in the present study). Such individual differ-
ences indeed appear to be a moderating factor of behavioral
responses to fearful displays as approach behaviors were espe-
cially observed in prosocial participants (Kaltwasser, Hildebrandt,
Wilhelm, & Sommer, 2017). However, the present experiments do
not allow assessing the motives underlying behavioral responses to
fear or the prosocial preferences of our participants. Future exper-
iments would be needed to determine whether approaching fearful
individuals is of self-preservative nature (a need for affiliation to
alleviate one’s own fear) or of prosocial nature (a desire to provide
help; Dezecache, Grèzes, & Dahl, 2017; Marsh, Kozak, & Am-
bady, 2007).
The present results provide evidence that, in a realistic context
offering competing targets for action, a fast attentional allocation
Table 4
Summary of Logistic Regression Analysis for Variables
Predicting Action Choices
Predictors Model 4 Odds ratio 95% CI pvalue
Intercept 1.29 [1.01, 1.64] .039
Saccade direction 1.99 [1.68, 2.36] ⬍.001
Emotion 1.06 [0.89, 1.26] .482
Saccade Direction ⫻Emotion 1.29 [1.01, 1.65] .042
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
11
ACTION- AND ATTENTION-SELECTION UNDER THREAT

toward threat-related displays was followed by a reorientation of
attention toward the end point of the selected action, before move-
ment initiation. These results are consistent with the prevailing
view that attentional processes prioritize threatening information,
irrespective of the associated behavioral response (Notebaert,
Crombez, Van Damme, De Houwer, & Theeuwes, 2011; Öhman
& Mineka, 2001; Vuilleumier & Huang, 2009). The wealth of
research demonstrating attentional biases toward threat classically
used stereotypical emotional faces, isolated from their bodies, and
displayed on a plain background (Cisler & Koster, 2010). Here, by
incorporating action opportunities to our design, we further re-
vealed that threat can later bias attention-selection processes to-
ward the most appropriate action possibility. In a similar vein, but
using aversive auditory conditioning during an attentional cueing
paradigm, a recent study demonstrated that when both threat and
safety signals compete for attentional priority, attention is allo-
cated to instrumental safety signals rather than to threat signals
(Vogt et al., 2017).
Threat-related displays reduced the time required to initiate a
given action plan, indicating speeded action selection when facing
danger. We suggest that existing maps of potential actions are
rapidly shaped through a modulation of the value associated with
each competing option. The affordance competition hypothesis
postulates that the brain forms multiple action plans in order to
efficiently respond to environmental challenges, while gathering
information to select the most appropriate according to the observ-
er’s current goal (Cisek, 2007). In the presence of threat, the
emotional information likely modifies the value of each action
plan, biasing the selection toward socially more adaptive behav-
iors. Even when emotional information is task irrelevant, as in the
everyday scenario of entering a subway car, such value modifica-
tion must be very efficient and rapid so as to promote survival.
Here, the action selection process seems to terminate before the
onset of the movement (mean initiation time of 410 ms), consistent
with the idea of a very quick mechanism shaped by evolution
(Thura & Cisek, 2014). Future experiments, using a “go-before-
you-know” design to better follow the deliberation online (Galli-
van, Bowman, Chapman, Wolpert, & Flanagan, 2016), as well as
computational modeling (Lepora & Pezzulo, 2015), are needed to
further specify the mechanisms underlying the prioritization of
threat-specific actions.
Although the present experiments provided important results
regarding the impact of anger and fear on action- and attention-
related decisions, they are not devoid from limitations, notably one
related to our small sample sizes. Therefore, to test the reliability
of our effects of interest, we conducted a meta-analytic analysis on
our results (Open Science Collaboration, 2015) which revealed a
highly significant meta-analytic effect (p⬍.001). Even though
future studies will benefit from larger sample sizes, this meta-
analytic analysis suggests that our effects are reliable. Finally,
besides including realistic contexts offering competing targets for
action, as in the present study, the ecological validity of experi-
mental designs to study the relationship between emotional dis-
plays and attention- and action-related decisions may be increased
using dynamic stimuli (Kaltwasser, Moore, et al., 2017).
To conclude, our findings strongly support that the presence of
threat-related facial displays in the environment rapidly guides the
selection of existing targets for action and determines attention
allocation toward the end point of the selected action. Further, we
found that two negative emotional expressions, anger and fear,
tend to impact action- and attention-related processes, in accor-
dance with their social meaning—namely, aggression and affilia-
tion, respectively. We propose that these processes have been
optimized through evolution to ensure survival.
References
Alexander, W. H., & Brown, J. W. (2011). Medial prefrontal cortex as an
action-outcome predictor. Nature Neuroscience, 14, 1338–1344. http://
dx.doi.org/10.1038/nn.2921
Anderson, A. K., Christoff, K., Panitz, D., De Rosa, E., & Gabrieli, J. D. E.
(2003). Neural correlates of the automatic processing of threat facial
signals. Journal of Neuroscience, 23, 5627–5633. http://dx.doi.org/10
.1523/JNEUROSCI.23-13-05627.2003
Ariga, A., & Arihara, K. (2018). Attentional capture by spatiotemporally
task-irrelevant faces: Supportive evidence for Sato and Kawahara
(2015). Psychological Research, 82, 859–865. http://dx.doi.org/10
.1007/s00426-017-0869-3
Bar-Haim, Y., Lamy, D., Pergamin, L., Bakermans-Kranenburg, M. J., &
van IJzendoorn, M. H. (2007). Threat-related attentional bias in anxious
and nonanxious individuals: A meta-analytic study. Psychological Bul-
letin, 133, 1–24. http://dx.doi.org/10.1037/0033-2909.133.1.1
Barto´
n, K. (2018). MuMIn: Multi-model inference (Version 1.42.1). Re-
trieved from https://CRAN.R-project.org/package⫽MuMIn
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear
mixed-effects models using lme4. Journal of Statistical Software. Ad-
vance online publication. http://dx.doi.org/10.18637/jss.v067.i01
Batson, C. D., Fultz, J., & Schoenrade, P. A. (1987). Distress and empathy:
Two qualitatively distinct vicarious emotions with different motivational
consequences. Journal of Personality, 55, 19–39. http://dx.doi.org/10
.1111/j.1467-6494.1987.tb00426.x
Bossuyt, E., Moors, A., & De Houwer, J. (2014). On angry approach and
fearful avoidance: The goal-dependent nature of emotional approach and
avoidance tendencies. Journal of Experimental Social Psychology, 50,
118–124. http://dx.doi.org/10.1016/j.jesp.2013.09.009
Calvo, M. G., Avero, P., & Lundqvist, D. (2006). Facilitated detection of
angry faces: Initial orienting and processing efficiency. Cognition and
Emotion, 20, 785–811. http://dx.doi.org/10.1080/02699930500465224
Carver, C. S., & Harmon-Jones, E. (2009). Anger is an approach-related
affect: Evidence and implications. Psychological Bulletin, 135, 183–
204. http://dx.doi.org/10.1037/a0013965
Cisek, P. (2007). A parallel framework for interactive behavior. Progress
in Brain Research, 165, 475–492. http://dx.doi.org/10.1016/S0079-
6123(06)65030-9
Cisek, P., & Kalaska, J. F. (2010). Neural mechanisms for interacting with
a world full of action choices. Annual Review of Neuroscience, 33,
269–298. http://dx.doi.org/10.1146/annurev.neuro.051508.135409
Cisek, P., & Pastor-Bernier, A. (2014). On the challenges and mechanisms
of embodied decisions. Philosophical Transactions of the Royal Society
of London, Series B, Biological Sciences, 369, 20130479. http://dx.doi
.org/10.1098/rstb.2013.0479
Cisler, J. M., & Koster, E. H. W. (2010). Mechanisms of attentional biases
towards threat in anxiety disorders: An integrative review. Clinical
Psychology Review, 30, 203–216. http://dx.doi.org/10.1016/j.cpr.2009
.11.003
Corr, P. J. (2013). Approach and avoidance behaviour: Multiple systems
and their interactions. Emotion Review, 5, 285–290. http://dx.doi.org/10
.1177/1754073913477507
Dezecache, G., Grèzes, J., & Dahl, C. D. (2017). The nature and distribu-
tion of affiliative behaviour during exposure to mild threat. Royal
Society open science, 4, 170265.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
12 VILAREM, ARMONY, AND GRÈZES

Dezecache, G., Jacob, P., & Grèzes, J. (2015). Emotional contagion: Its
scope and limits. Trends in Cognitive Sciences, 19, 297–299. http://dx
.doi.org/10.1016/j.tics.2015.03.011
Dezecache, G., Mercier, H., & Scott-Phillips, T. (2013). An evolutionary
approach to emotional communication (SSRN Scholarly Paper No. ID
2345829). Rochester, NY: Social Science Research Network.
Eastwood, J. D., Smilek, D., & Merikle, P. M. (2001). Differential atten-
tional guidance by unattended faces expressing positive and negative
emotion. Perception & Psychophysics, 63, 1004–1013. http://dx.doi.org/
10.3758/BF03194519
Elliot, A. J., & Covington, M. V. (2001). Approach and avoidance moti-
vation. Educational Psychology Review, 13, 73–92. http://dx.doi.org/10
.1023/A:1009009018235
El Zein, M., Wyart, V., & Grèzes, J. (2015). Anxiety dissociates the
adaptive functions of sensory and motor response enhancements to
social threats. eLife, 4, e10274. http://dx.doi.org/10.7554/eLife.10274
Fagioli, S., Ferlazzo, F., & Hommel, B. (2007). Controlling attention
through action: Observing actions primes action-related stimulus dimen-
sions. Neuropsychologia, 45, 3351–3355. http://dx.doi.org/10.1016/j
.neuropsychologia.2007.06.012
Fenske, M. J., & Eastwood, J. D. (2003). Modulation of focused attention
by faces expressing emotion: Evidence from flanker tasks. Emotion, 3,
327–343. http://dx.doi.org/10.1037/1528-3542.3.4.327
Fox, E., Griggs, L., & Mouchlianitis, E. (2007). The detection of fear-
relevant stimuli: Are guns noticed as quickly as snakes? Emotion, 7,
691–696. http://dx.doi.org/10.1037/1528-3542.7.4.691
Freeman, J. B., & Ambady, N. (2010). MouseTracker: Software for study-
ing real-time mental processing using a computer mouse-tracking meth-
od. Behavior Research Methods, 42, 226–241. http://dx.doi.org/10
.3758/BRM.42.1.226
Fridlund, A. J. (1994). Human facial expression: An evolutionary view. San
Diego, CA: Academic Press.
Gallivan, J. P., Bowman, N. A. R., Chapman, C. S., Wolpert, D. M., &
Flanagan, J. R. (2016). The sequential encoding of competing action
goals involves dynamic restructuring of motor plans in working mem-
ory. Journal of Neurophysiology, 115, 3113–3122. http://dx.doi.org/10
.1152/jn.00951.2015
Grèzes, J. (2011). Les emotions, modes d’action et de communication
[Emotions motivate action and communication]. Ms, 27, 683–684.
Grupe, D. W., & Nitschke, J. B. (2013). Uncertainty and anticipation in
anxiety: An integrated neurobiological and psychological perspective.
Nature Reviews Neuroscience, 14, 488–501. http://dx.doi.org/10.1038/
nrn3524
Hammer, J. L., & Marsh, A. A. (2015). Why do fearful facial expressions
elicit behavioral approach? Evidence from a combined approach-
avoidance implicit association test. Emotion, 15, 223–231. http://dx.doi
.org/10.1037/emo0000054
Isbell, L. A. (1994). Predation on primates: Ecological patterns and evo-
lutionary consequences. Evolutionary Anthropology, 3, 61–71. http://dx
.doi.org/10.1002/evan.1360030207
Jiang, Y. V., Won, B.-Y., & Swallow, K. M. (2014). First saccadic eye
movement reveals persistent attentional guidance by implicit learning.
Journal of Experimental Psychology: Human Perception and Perfor-
mance, 40, 1161–1173. http://dx.doi.org/10.1037/a0035961
Juth, P., Lundqvist, D., Karlsson, A., & Öhman, A. (2005). Looking for
foes and friends: Perceptual and emotional factors when finding a face
in the crowd. Emotion, 5, 379–395. http://dx.doi.org/10.1037/1528-3542
.5.4.379
Kaltwasser, L., Hildebrandt, A., Wilhelm, O., & Sommer, W. (2017). On
the relationship of emotional abilities and prosocial behavior. Evolution
and Human Behavior, 38, 298–308. http://dx.doi.org/10.1016/j
.evolhumbehav.2016.10.011
Kaltwasser, L., Moore, K., Weinreich, A., & Sommer, W. (2017). The
influence of emotion type, social value orientation and processing focus
on approach-avoidance tendencies to negative dynamic facial expres-
sions. Motivation and Emotion, 41, 532–544. http://dx.doi.org/10.1007/
s11031-017-9624-8
Keltner, D., & Haidt, J. (1999). Social functions of emotions at four levels
of analysis. Cognition and Emotion, 13, 505–521. http://dx.doi.org/10
.1080/026999399379168
Kirsch, W. (2015). Impact of action planning on spatial perception: Atten-
tion matters. Acta Psychologica, 156, 22–31. http://dx.doi.org/10.1016/
j.actpsy.2015.01.002
Koster, E. H. W., Crombez, G., Verschuere, B., Van Damme, S., &
Wiersema, J. R. (2006). Components of attentional bias to threat in high
trait anxiety: Facilitated engagement, impaired disengagement, and at-
tentional avoidance. Behaviour Research and Therapy, 44, 1757–1771.
http://dx.doi.org/10.1016/j.brat.2005.12.011
Krieglmeyer, R., & Deutsch, R. (2013). Approach does not equal approach:
angry facial expressions evoke approach only when it serves aggression.
Social Psychological and Personality Science, 4, 607–614. http://dx.doi
.org/10.1177/1948550612471060
Laham, S. M., Kashima, Y., Dix, J., & Wheeler, M. (2015). A meta-
analysis of the facilitation of arm flexion and extension movements as a
function of stimulus valence. Cognition and Emotion, 29, 1069–1090.
http://dx.doi.org/10.1080/02699931.2014.968096
Lepora, N. F., & Pezzulo, G. (2015). Embodied choice: How action
influences perceptual decision making. PLoS Computational Biology,
11, e1004110. http://dx.doi.org/10.1371/journal.pcbi.1004110
Marsh, A. A., Ambady, N., & Kleck, R. E. (2005). The effects of fear and
anger facial expressions on approach- and avoidance-related behaviors.
Emotion, 5, 119–124. http://dx.doi.org/10.1037/1528-3542.5.1.119
Marsh, A. A., Kozak, M. N., & Ambady, N. (2007). Accurate identification
of fear facial expressions predicts prosocial behavior. Emotion, 7, 239–
251. http://dx.doi.org/10.1037/1528-3542.7.2.239
Martinez-Conde, S., Macknik, S. L., Troncoso, X. G., & Hubel, D. H.
(2009). Microsaccades: A neurophysiological analysis. Trends in Neu-
rosciences, 32, 463–475. http://dx.doi.org/10.1016/j.tins.2009.05.006
Mawson, A. R. (2012). Mass panic and social attachment: The dynamics
of human behavior. London, UK: Ashgate Publishing, Ltd.
Mogg, K., Bradley, B., Miles, F., & Dixon, R. (2004). Brief report time
course of attentional bias for threat scenes: Testing the vigilance-
avoidance hypothesis. Cognition and Emotion, 18, 689–700. http://dx
.doi.org/10.1080/02699930341000158
Mulckhuyse, M., Crombez, G., & Van der Stigchel, S. (2013). Conditioned
fear modulates visual selection. Emotion, 13, 529–536. http://dx.doi.org/
10.1037/a0031076
Mulckhuyse, M., & Dalmaijer, E. S. (2016). Distracted by danger: Tem-
poral and spatial dynamics of visual selection in the presence of threat.
Cognitive, Affective & Behavioral Neuroscience, 16, 315–324. http://dx
.doi.org/10.3758/s13415-015-0391-2
Nichols, S. (2001). Mindreading and the cognitive architecture underlying
altruistic motivation. Mind & Language, 16, 425–455. http://dx.doi.org/
10.1111/1468-0017.00178
Nissens, T., Failing, M., & Theeuwes, J. (2017). People look at the object
they fear: Oculomotor capture by stimuli that signal threat. Cognition
and Emotion, 31, 1707–1714.
Notebaert, L., Crombez, G., Van Damme, S., De Houwer, J., & Theeuwes,
J. (2011). Signals of threat do not capture, but prioritize, attention: A
conditioning approach. Emotion, 11, 81–89. http://dx.doi.org/10.1037/
a0021286
Öhman, A., Flykt, A., & Esteves, F. (2001). Emotion drives attention:
Detecting the snake in the grass. Journal of Experimental Psychology:
General, 130, 466–478. http://dx.doi.org/10.1037/0096-3445.130.3.466
Öhman, A., & Mineka, S. (2001). Fears, phobias, and preparedness:
Toward an evolved module of fear and fear learning. Psychological
Review, 108, 483–522. http://dx.doi.org/10.1037/0033-295X.108.3.483
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
13
ACTION- AND ATTENTION-SELECTION UNDER THREAT

Open Science Collaboration. (2015). Estimating the reproducibility of
psychological science. Science, 349, aac4716. http://dx.doi.org/10.1126/
science.aac4716
Palser, E., Fotopoulou, A., & Kilner, J. (2015). The role of movement in
metacognition: The relationship between movement speed and confi-
dence (Program No. 804.07, 2015 Neuroscience Meeting Planner).
Washington, DC: Society for Neuroscience.
Paulus, A., & Wentura, D. (2014). Threatening joy: Approach and avoid-
ance reactions to emotions are influenced by the group membership of
the expresser. Cognition and Emotion, 28, 656–677. http://dx.doi.org/
10.1080/02699931.2013.849659
Paulus, A., & Wentura, D. (2016). It depends: Approach and avoidance
reactions to emotional expressions are influenced by the contrast emo-
tions presented in the task. Journal of Experimental Psychology: Human
Perception and Performance, 42, 197–212. http://dx.doi.org/10.1037/
xhp0000130
Pflugshaupt, T., Mosimann, U. P., von Wartburg, R., Schmitt, W.,
Nyffeler, T., & Müri, R. M. (2005). Hypervigilance-avoidance pattern in
spider phobia. Journal of Anxiety Disorders, 19, 105–116. http://dx.doi
.org/10.1016/j.janxdis.2003.12.002
Pinkham, A. E., Griffin, M., Baron, R., Sasson, N. J., & Gur, R. C. (2010).
The face in the crowd effect: Anger superiority when using real faces
and multiple identities. Emotion, 10, 141–146. http://dx.doi.org/10.1037/
a0017387
Preciado, D., Munneke, J., & Theeuwes, J. (2017). Was that a threat?
Attentional biases by signals of threat. Emotion, 17, 478–486. http://dx
.doi.org/10.1037/emo0000246
Reed, L. I., DeScioli, P., & Pinker, S. A. (2014). The commitment function
of angry facial expressions. Psychological Science, 25, 1511–1517.
http://dx.doi.org/10.1177/0956797614531027
Rizzolatti, G., Riggio, L., Dascola, I., & Umiltá, C. (1987). Reorienting
attention across the horizontal and vertical meridians: Evidence in favor
of a premotor theory of attention. Neuropsychologia, 25, 31–40. http://
dx.doi.org/10.1016/0028-3932(87)90041-8
Sander, D., Grandjean, D., Kaiser, S., Wehrle, T., & Scherer, K. R. (2007).
Interaction effects of perceived gaze direction and dynamic facial ex-
pression: Evidence for appraisal theories of emotion. European Journal
of Cognitive Psychology, 19, 470–480. http://dx.doi.org/10.1080/09
541440600757426
Schmidt, L. J., Belopolsky, A. V., & Theeuwes, J. (2012). The presence of
threat affects saccade trajectories. Visual Cognition, 20, 284–299. http://
dx.doi.org/10.1080/13506285.2012.658885
Schmidt, L. J., Belopolsky, A. V., & Theeuwes, J. (2015). Potential threat
attracts attention and interferes with voluntary saccades. Emotion, 15,
329–338. http://dx.doi.org/10.1037/emo0000041
Schmidt, L. J., Belopolsky, A. V., & Theeuwes, J. (2017). The time course
of attentional bias to cues of threat and safety. Cognition and Emotion,
31, 845–857. http://dx.doi.org/10.1080/02699931.2016.1169998
Sell, A., Cosmides, L., & Tooby, J. (2014). The human anger face evolved
to enhance cues of strength. Evolution and Human Behavior, 35, 425–
429. http://dx.doi.org/10.1016/j.evolhumbehav.2014.05.008
Sheliga, B. M., Riggio, L., & Rizzolatti, G. (1994). Orienting of attention
and eye movements. Experimental Brain Research, 98, 507–522. http://
dx.doi.org/10.1007/BF00233988
Soares, S. C., Esteves, F., Lundqvist, D., & Öhman, A. (2009). Some
animal specific fears are more specific than others: Evidence from
attention and emotion measures. Behaviour Research and Therapy, 47,
1032–1042. http://dx.doi.org/10.1016/j.brat.2009.07.022
Springer, U. S., Rosas, A., McGetrick, J., & Bowers, D. (2007). Differ-
ences in startle reactivity during the perception of angry and fearful
faces. Emotion, 7, 516–525. http://dx.doi.org/10.1037/1528-3542.7.3
.516
Thura, D., & Cisek, P. (2014). Deliberation and commitment in the
premotor and primary motor cortex during dynamic decision making.
Neuron, 81, 1401–1416. http://dx.doi.org/10.1016/j.neuron.2014.01.031
Van Kleef, G. A. (2009). How emotions regulate social life: The Emotions
as Social Information (EASI) model. Current Directions in Psycholog-
ical Science, 18, 184–188. http://dx.doi.org/10.1111/j.1467-8721.2009
.01633.x
Vogt, J., Koster, E. H. W., & De Houwer, J. (2017). Safety first: Instru-
mentality for reaching safety determines attention allocation under
threat. Emotion, 17, 528–537. http://dx.doi.org/10.1037/emo0000251
Vuilleumier, P., Armony, J. L., Driver, J., & Dolan, R. J. (2001). Effects of
attention and emotion on face processing in the human brain: An
event-related fMRI study. Neuron, 30, 829–841. http://dx.doi.org/10
.1016/S0896-6273(01)00328-2
Vuilleumier, P., & Huang, Y.-M. (2009). Emotional attention: Uncovering
the mechanisms of affective biases in perception. Current Directions in
Psychological Science, 18, 148–152. http://dx.doi.org/10.1111/j.1467-
8721.2009.01626.x
Wagenmakers, E.-J., & Farrell, S. (2004). AIC model selection using
Akaike weights. Psychonomic Bulletin & Review, 11, 192–196. http://
dx.doi.org/10.3758/BF03206482
Wilkowski, B. M., & Meier, B. P. (2010). Bring it on: Angry facial
expressions potentiate approach-motivated motor behavior. Journal of
Personality and Social Psychology, 98, 201–210. http://dx.doi.org/10
.1037/a0017992
Wyart, V., Myers, N. E., & Summerfield, C. (2015). Neural mechanisms of
human perceptual choice under focused and divided attention. Journal of
Neuroscience, 35, 3485–3498. http://dx.doi.org/10.1523/JNEUROSCI
.3276-14.2015
Wykowska, A., Schubö, A., & Hommel, B. (2009). How you move is what
you see: Action planning biases selection in visual search. Journal of
Experimental Psychology: Human Perception and Performance, 35,
1755–1769. http://dx.doi.org/10.1037/a0016798
Yiend, J., & Mathews, A. (2001). Anxiety and attention to threatening
pictures. Quarterly Journal of Experimental Psychology A: Human
Experimental Psychology, 54, 665–681. http://dx.doi.org/10.1080/
713755991
Received November 29, 2017
Revision received December 12, 2018
Accepted February 5, 2019 䡲
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
14 VILAREM, ARMONY, AND GRÈZES