Procesamiento dinámico de amenazas
Christian Meyer*, Srikanth Padmala*, and Luiz Pessoa
Abstracto
■ During real-life situations, multiple factors interact dynami-
cally to determine threat level. In the current fMRI study involv-
ing healthy adult human volunteers, we investigated interactions
between proximity, direction (approach vs. retreat), and speed
during a dynamic threat-of-shock paradigm. As a measure of
threat-evoked physiological arousal, skin conductance re-
sponses were recorded during fMRI scanning. Some brain re-
gions tracked individual threat-related factors, and others were
also sensitive to combinations of these variables. En particular,
signals in the anterior insula tracked the interaction between
proximity and direction where approach versus retreat re-
sponses were stronger when threat was closer compared with
farther. A parallel proximity-by-direction interaction was also
observed in physiological skin conductance responses. En el
right amygdala, we observed a proximity by direction inter-
acción, but intriguingly in the opposite direction as the anterior
insula; retreat versus approach responses were stronger when
threat was closer compared with farther. In the right bed nu-
cleus of the stria terminalis, we observed an effect of threat
proximity, whereas in the right periaqueductal gray/midbrain
we observed an effect of threat direction and a proximity by
direction by speed interaction (the latter was detected in ex-
ploratory analyses but not in a voxelwise fashion). Juntos,
our study refines our understanding of the brain mechanisms
involved during aversive anticipation in the human brain.
En tono rimbombante, it emphasizes that threat processing should be
understood in a manner that is both context-sensitive and
dynamic. ■
INTRODUCCIÓN
Anticipation of aversive events leads to a repertoire of
changes in behavioral, physiological, and brain responses
that contribute to the handling of the negative conse-
quences of such events. Al mismo tiempo, abnormalities
in aversive anticipatory processing are thought to un-
derlie many mental disorders, such as anxiety and de-
presion (Dillon et al., 2014; Grupe & Nitschke, 2013).
Por eso, understanding the brain mechanisms of aversive
anticipation is important from both basic and clinical
standpoints.
In humans, aversive anticipation has been investigated
with paradigms in which punctate cues signal an upcom-
ing negative event (Marrón, Seymour, Boyle, El-Deredy, &
jones, 2008; Nitschke, Sarinopoulos, Mackiewicz, Schaefer,
& Davidson, 2006; Simmons, Strigo, Matthews, Paulus, &
piedra, 2006; Jensen et al., 2003; Böcker, Baas, Kenemans,
& Verbaten, 2001) or by blocked manipulations with con-
stant threat level (McMenamin, Langeslag, Sirbu, padmala,
& Persona, 2014; Vytal, Overstreet, Charney, robinson, &
Grillon, 2014). Sin embargo, during most real-world situa-
ciones, aversive anticipation changes dynamically over time.
An important factor in determining threat level is proxim-
idad, as when a prey reacts differently to the presence of a
predator when the latter is proximal compared with
Universidad de Maryland
* These authors contributed equally to this work.
© 2018 Instituto de Tecnología de Massachusetts
distant (Figura 1A; Blanchard, Griebel, Pobbe, &
Blanchard, 2011; Blanchard & Blanchard, 1990). Otro
factors involve direction, namely whether threat is
approaching versus retreating (Figura 1B) and speed,
reflecting how fast or slow the threat is moving
(Fanselow & Lester, 1988). Some studies have taken initial
strides at investigating how some of these factors influ-
ence brain responses during aversive anticipation. Para
instancia, the contrast of proximal versus distal threats
revealed fMRI responses in a host of brain regions, incluir-
ing the anterior insula, midbrain periaqueductal gray
(PAG), and bed nucleus of the stria terminalis (BST;
Mobbs et al., 2010; Somerville, Whalen, & kelly, 2010);
evidence for amygdala involvement linked to threat prox-
imity is mixed (Mobbs et al., 2010; Somerville et al., 2010).
Similarmente, comparison of approaching versus retreating
threats has revealed responses in the anterior insula,
BST, and amygdala (Mobbs et al., 2010).
Hasta ahora, studies have considered the effects of threat
proximity and direction independently. Por eso, it is cur-
rently unknown how such factors potentially interact in
the brain during aversive anticipation (Figura 1C). Este
is an important gap in our knowledge base because behav-
ioral findings have extensively documented interactions
between threat-related factors, which have produced
several influential theoretical accounts (for excellent dis-
cussion, see Mobbs, Hagan, Dalgleish, Silston, & Prévost,
2015). Además, it is not only important to investi-
gate how multiple threat-related factors interact but to
Revista de neurociencia cognitiva 31:4, páginas. 522–542
doi:10.1162/jocn_a_01363
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
understand how the brain tracks them continuously. En
particular, do signal fluctuations in brain regions track
threat-related factors dynamically? En ese caso, to what factor(s)
and factor combinations are they sensitive?
To address these questions, we devised a paradigm in
which threat was dynamically modulated during fMRI
scanning. Two circles moved on the screen, sometimes
moving closer and sometimes moving apart, and at vary-
ing speeds (Cifra 2). Participants were instructed to pay
attention to the circles on the screen and were explicitly
informed that, if they touched, the participants would
receive an unpleasant shock. As a measure of threat-
evoked physiological arousal, skin conductance re-
sponses (SCRs) were recorded during scanning. Nuestro
paradigm allowed us to investigate the role played by
the interaction between proximity (nearer vs. farther
circles), direction (approach vs. retreat), and speed (fas-
ter vs. slower) in determining brain responses during
anticipatory threat processing. En tono rimbombante, the impact
of the factors “proximity” and “speed” were assessed
parametrically (es decir., continuously) as they varied dynami-
cally. Por lo tanto, the paradigm allowed us to test how
Cifra 1. Threat-related factors and their interaction. (A) Closer and
farther threat, where threat is represented by an aversive shock
when circles touched. (B) Direction of threat: approach versus retreat.
(C) Threat level may depend on both proximity (closer and farther) y
direction (left panels indicate approach; right panels indicate retreat).
Cifra 2. Experimental paradigm. Two circles moved randomly on the
screen and a shock was administered to the participant if they touched.
The inset represents threat proximity (the distance between the two
circles), which varied continuously. A central goal of the study was to
determine the extent to which signal fluctuations in brain regions
(such as the anterior insula) followed threat-related factors (incluido
proximity) and their interactions.
multiple threat-related factors “dynamically” influence
signals fluctuations across brain regions. Específicamente, hacer
they provide independent contributions or do they inter-
act in regions important for threat processing, como
the anterior insula, amygdala, PAG, and BST? Intuitivamente,
probing interactions allowed us to evaluate the extent to
which the influence of one factor on threat anticipation
depended on the values of other factor(s). Por ejemplo,
in terms of a two-way interaction, we anticipated that the
influence of direction (es decir., approaching vs. retreating
amenaza) would depend on proximity (es decir., whether the
threat was near versus far; Figura 1C). In terms of
three-way interactions, we sought to evaluate if the inter-
action between the continuously manipulated factors of
proximity and speed depended on direction.
MÉTODOS
Participantes
Eighty-five participants (41 women, ages 18–40 years,
average = 22.62 años, DE = 4.85) with normal or cor-
rected-to-normal vision and no reported neurological or
psychiatric disease were recruited from the University of
Maryland community (of the original sample of 93, datos
from seven participants were discarded because of
technical issues during data transfer [específicamente, campo
maps were lost], and one other participant was removed
because of poor structural–functional alignment). El
Meyer, padmala, and Pessoa
523
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
project was approved by the University of Maryland,
College Park institutional review board, and all partici-
pants provided written informed consent before partici-
pation. The data analyzed here were investigated in an
entirely separate fashion at the level of networks and pub-
lished previously (Najafi, Kinnison, & Persona, 2017). El
sample size was not based on an explicit statistical power
análisis. At the outset, we sought to collect around 90
participants to allow investigation of the data in terms of
separate “exploratory” and “test” sets in the network study
(Najafi et al., 2017). For the investigation of activation
(this article), our intention was to use the available data
in a single type of analysis.
Anxiety Questionnaires
Participants completed the trait portion of the Spielberger
State–Trait Anxiety Inventory (Spielberger, Gorsuch, &
Lushene, 1970) before scanning (average = 17.23 días,
DE = 15.90) and then completed the state portion of
the State–Trait Anxiety Inventory immediately before the
scanning session.
Procedure and Stimuli
Two circles with different colors moved around on the
screen randomly. When they collided with each other,
an unpleasant mild electric shock was delivered.
En general, proximity, direction of movement, and relative
speed of the circles were used to influence perceived
amenaza. The position of each circle (on the plane), xt,
was defined based on its previous position, xt−1, plus a
random displacement, Δxt:
xt ¼ xt−1 þ Δxt
The magnitude and direction of the displacement was
calculated by combining a normal random distribution
with a momentum term to ensure motion smoothness,
while at the same time remaining (relatively) unpredict-
able to the participants. Específicamente, the displacement
was updated every 50 msec as follows:
d
Δxt ¼ 1 − c
Þ
ÞΔxt−1 þ cN 0; 1d
where c = 0.2 y N(0, 1) indicates the normal distribu-
tion with zero mean and standard deviation of 1.
Visual stimuli were presented using PsychoPy (www.
psychopy.org/) and viewed on a projection screen via a
mirror mounted to the scanner’s head coil. Each partici-
pant viewed the same sequence of circle movements.
The total experiment included six runs (457 sec each),
each of which had six blocks (3 de 85 participants had
only five runs). In each block, the circles appeared on
the screen and moved around for 60 segundo; blocks were
separated by a 15-sec off period during which the
screen remained blank. Each run ended with a 7-sec blank
pantalla.
To ensure that the effects of threat proximity and
direction were uncorrelated, half of the blocks in each
run were temporally reversed versions of the other
blocks in that run. Temporally reversing the stimulus
trajectories guarantees that proximity and direction are
uncorrelated because reversing time changes the sign
of the direction (es decir., approach becomes retreat). A
optimize the experimental design, 10,000 candidate
stimuli trajectories and block orders were generated.
We then selected six runs, which minimized collinearity
between all predictors of interest (see below), measured
as the sum of respective variance inflation factors (Neter,
Kutner, Nachtsheim, & Wasserman, 1996).
In each run, the circles collided eight times within
four of six blocks (one to three times in a block); en
the remaining two blocks, there were no collisions.
Each collision resulted in the delivery of an electric
shock. The 500-msec electric shock (composed of a
series of current pulses at 50 Hz) was delivered by an
electric stimulator (Model E13-22 from Coulbourn Instru-
mentos, Whitehall, Pensilvania) to the fourth and fifth fingers
of the nondominant left hand via MRI-compatible elec-
trodes. To calibrate the intensity of the shock, cada
participant was asked to choose his or her own stimu-
lation level immediately before functional imaging,
such that the stimulus would be “highly unpleasant but
not painful.” After each run, participants were asked
about the unpleasantness of the stimulus to recalibrate
shock strength, if needed. SCR data were collected using
the MP-150 system (BIOPAC Systems, Cª, Goleta, California)
at a sampling rate of 250 Hz by using MRI-compatible
electrodes attached to the index and middle fingers of
the nondominant left hand. Because of technical problems
and/or experimenter errors during data collection, SCR
data were not available in two participants, and six par-
ticipants had only five runs of the SCR data; one par-
ticipant who had only three runs of data was excluded
from the analysis of SCR data.
MRI Data Acquisition
Functional and structural MRI data were acquired using a
3-T Siemens TRIO scanner with a 32-channel head coil.
Primero, a high-resolution T2-weighted anatomical scan
using Siemens’s SPACE sequence (0.8 mm isotropic)
was collected. Después, we collected 457 functional
EPI volumes in each run using a multiband scanning
secuencia (Feinberg et al., 2010), with repetition time =
1.0 segundo, echo time = 39 mseg, campo de visión = 210 mm, y
multiband factor = 6. Each volume contained 66 non-
overlapping oblique slices oriented 30° clockwise relative
to the AC–PC axis (2.2 mm isotropic). A high-resolution
T1-weighted MPRAGE anatomical scan (0.8 mm isotropic)
was collected. Además, in each session, double-echo
field maps (TE1 = 4.92 mseg, TE2 = 7.38 mseg) eran
524
Revista de neurociencia cognitiva
Volumen 31, Número 4
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
acquired with acquisition parameters matched to the
functional data.
fMRI Preprocessing
To preprocess the fMRI and anatomical MRI data, a
combination of packages and in-house scripts were
usado. The first three volumes of each functional run
were discarded to account for equilibration effects.
Slice-timing correction—with Analysis of Functional
Neuroimages’ (AFNI; Cox, 1996) 3dTshift—used Fourier
interpolation to align the onset times of every slice in
a volume to the first acquisition slice, and then a six-
parameter rigid body transformation (with AFNI’s
3dvolreg) corrected head motion within and between
runs by spatially registering each volume to the first
volumen.
en este estudio, we strived to improve functional–
anatomical coregistration given the small size of some
of the structures of interest. Skull stripping determines
which voxels are to be considered part of the brain
y, although conceptually simple, plays a very important
role in successful subsequent coregistration and nor-
malization steps. Actualmente, available packages perform
suboptimally in specific cases, and mistakes in the brain-
to-skull segmentation can be easily identified. Respectivamente,
to skull strip the T1 high-resolution anatomical image
(which was rotated to match the oblique plane of the
functional data with AFNI’s 3dWarp), we used six different
packages, ANTs (Avants, Tustison, & Song, 2009; http://
stnava.github.io/ANTs/), AFNI (Cox, 1996; http://afni.nimh.
nih.gov/ ), ROBEX (Iglesias, Liu, Thompson, & Tu, 2011;
https://www.nitrc.org/projects/robex), FSL (Smith et al.,
2004; http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/ ), SPM (www.fil.
ion.ucl.ac.uk/spm/), and BrainSuite (shattuck & Leahy,
2002; http://brainsuite.org/), and used a “voting scheme”
como sigue: Based on T1 data, a voxel was considered to
be part of the brain if four of six packages estimated it to
be a brain voxel; de lo contrario, the voxel was not considered
to be brain tissue (for six participants whose T1 data were
lost due to issues during data transfer, the T2 image was
used instead and only the ANTs package was used for skull
stripping).
Después, FSL was used to process field map im-
ages and create a phase distortion map for each partici-
pant (by using bet and fsl_prepare_fieldmap). FSL’s
epi_reg was then used to apply boundary-based coregis-
tration to align the unwarped mean volume registered
EPI image with the skull-stripped anatomical image (T1
or T2), along with simultaneous EPI distortion correction
(Greve & pescado, 2009).
Próximo, ANTS was used to estimate a nonlinear transfor-
mation that mapped the skull-stripped anatomical image
(T1 or T2) to the skull-stripped MNI152 template (enterrar-
polated to 1-mm isotropic voxels). Finalmente, ANTS com-
bined the nonlinear transformations from coregistration/
unwarping (from mapping mean functional EPI image to
the anatomical T1 or T2) and normalization (from map-
ping T1 or T2 to the MNI template) into a single trans-
formation that was applied to map volume-registered
functional volumes to standard space (interpolated to
2-mm isotropic voxels). In this process, ANTS also uti-
lized the field maps to simultaneously minimize EPI dis-
tortion. The resulting spatially normalized functional data
were blurred using a 4-mm FWHM Gaussian filter. Espacial
smoothing was restricted to gray matter mask voxels.
Finalmente, the intensity of each voxel was normalized to a
significado de 100 (separately for each run).
Voxelwise Analysis
Each participant’s preprocessed fMRI data were analyzed
using multiple linear regression with AFNI (restricted to
gray matter voxels) using the 3dDeconvolve program
(https://afni.nimh.nih.gov/afni/doc/manual/3dDeconvolve.
pdf ). Time series data were analyzed according to the
following model (additional nuisance variables are de-
scribed below):
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Y ¼ β
PP þ β
þ β
PDSPDS
DD þ β
SS þ β
PDPD þ β
PSPS þ β
DSDS
(1)
where P indicates proximity, D represents direction,
and S represents speed. Variables were determined
based on circle positions on the screen. Proximity was
defined as the Euclidean distance between the two cir-
cles; direction indicated approach versus retreat; velocidad
was the discrete temporal difference of proximity. El
products PD, PS, and PDS represent the interactions
terms; the individual terms P, D, and S were mean-
centered before multiplication to reduce potential col-
linearity. The resulting regressors exhibited pairwise
correlations that were relatively small (the largest was
.41), and all variance inflation factors were less than
1.3, indicating that model estimation was unproblem-
atic (Mumford, Poline, & Poldrack, 2015).
In addition to the variables above, we included re-
gressors for visual motion (velocity tangential to the
difference vector of the combined circle-to-circle stimu-
lus), sustained block event (60-sec duration), and block-
onset and block-offset events (1-sec duration) to account
for transient responses at block onset/offset. All regres-
sors were convolved with a standard hemodynamic re-
sponse based on the gamma variate model (cohen,
1997). Note that interaction regressors were multiplied
before convolution; también, as stimulus-related display in-
formation was updated every 50 mseg (20 Hz), convolu-
tion with the hemodynamic response was performed
before decimating the convolved signal to the fMRI
sample rate (1 Hz). To simplify plotting, decimated re-
gressors were scaled by their corresponding root mean
square value (de este modo, multiplicative interactions terms
Meyer, padmala, and Pessoa
525
were on the same scale as simple effects). Other regres-
sors included in the model included six motion param-
eters (three linear displacements and three angular
rotations) and their discrete temporal derivatives. A
further control for head motion-related artifacts in the
datos (Siegel et al., 2014), we excluded volumes (en
promedio 0.4%) with a frame-to-frame displacement of
más que 1 mm. To model baseline and drifts of the
MRI signal, regressors corresponding to polynomial
terms up to the fourth order were included (for each
run separately). Finalmente, to minimize effects due to the
physical shock event, data points in a 15-sec window
after shock delivery were discarded from the analysis. Él
should be pointed out that to partly account for the fact
that the circles were most proximal just before shock
events, the design included time periods when circles
were very close but did not touch eventually.
Group Analysis
Whole-brain voxelwise random-effects analyses were
conducted using response estimates from individual-
level analyses (restricted to gray matter voxels) in AFNI.
To probe the effects of the regressors of interest, we ran
separate one-sample t tests against zero using the AFNI’s
3dttestþþ program.
The alpha-level for voxelwise statistical analysis was
determined by simulations using the 3dClustSim pro-
gram (restricted to gray matter voxels). For these simu-
laciones, the smoothness of the data was estimated using
3dFWHMx program (restricted to gray matter voxels)
based on the residual time series from the individual-
level voxelwise analysis. Taking into account the recent
report of increased false-positive rates linked to the as-
sumption of Gaussian spatial autocorrelation in fMRI data
(Eklund, Nichols, & Knutsson, 2016), we used the -acf
(es decir., autocorrelation function) option recently added to
the 3dFWHMx and 3dClustSim tools, which models
spatial fMRI noise as a mixture of Gaussian plus mono-
exponential distributions. This improvement was shown
to control false-positive rates around the desired alpha
nivel, especially with relatively stringent voxel-level
uncorrected p values such as .001 (Cox, Chen, Glen,
Reynolds, & taylor, 2017). Based on a voxel-level un-
corrected p value of .001, simulations indicated a mini-
mum cluster extent of 13 vóxeles (2.0 × 2.0 × 2.0 mm)
for a cluster-level corrected alpha of .05.
BST ROI Analysis
The BST is a basal forebrain region and has been fre-
quently implicated in threat-related processing (Fox,
Oler, Tromp, Fudge, & Kalin, 2015; davis, Caminante, Miles,
& Grillon, 2010), along with other regions such as the
amygdala and anterior insula (Persona, 2016). Porque
the BST is a small region, analysis based on spatially
smoothed data would be susceptible to signals from
surrounding structures. To reduce this possibility, nosotros
conducted an additional BST ROI analysis using spatially
unsmoothed data. Bilateral BST ROIs were defined
anatomically according to the probabilistic mask of the
BST (en 25% límite) recently reported by Blackford
and colleagues (Theiss, Ridgewell, McHugo, Heckers, &
Blackford, 2017). For this analysis, no spatial smoothing
was applied. In each participant, for each ROI, a repre-
sentative time series was created by averaging the un-
smoothed time series from all the gray matter voxels
within the anatomically defined ROI (izquierda: nine voxels,
bien: eight voxels). Entonces, as in the individual-level voxel-
wise analysis, multiple linear regression analysis was run
using the 3dDeconvolve program to estimate condition-
specific responses. A nivel de grupo, as in the voxelwise
análisis, we ran separate one-sample t tests against zero
using the corresponding regression coefficients from the
individual-level analysis.
SCR Analysis
Each participant’s SCR data were initially smoothed with
a median filter over 50 muestras (200 mseg) to reduce
scanner-induced noise. In each run, la primera 3 sec of
data were discarded (corresponding to first three vol-
umes excluded in the fMRI analysis), and the remaining
data were resampled by decimating the 250-Hz sample
rate to the sample rate of fMRI data (1 Hz) and sub-
sequently Z scored. The preprocessed SCR data were
then analyzed using multiple linear regression using the
3dDeconvolve program in AFNI (for related approaches,
see Bach, Flandin, Friston, & Dolan, 2009; Engelmann,
Meyer, Fehr, & Fallar, 2015). We used the same regres-
sion model as the one used for fMRI data (see Equation
1). Además, we included regressors for visual motion
(velocity tangential to the difference vector of the com-
bined circle-to-circle stimulus), sustained block event
(60-sec duration), and block-onset and block-offset
events (1-sec duration) to account for transient responses
at block onset/offset. All regressors were convolved with
Cifra 3. Skin conductance response model based on the sigmoid
exponential function (Lim y col., 1997). A.U. = arbitrary units.
526
Revista de neurociencia cognitiva
Volumen 31, Número 4
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
responses to physical shock itself. For the anticipatory
actividad, we considered the proximity by direction inter-
action and focused on the right anterior insula and right
amygdala clusters, which exhibited this interaction (ver
Resultados). To estimate responses to physical shocks, nosotros
ran a separate multiple regression analysis with all the
regressors as in the original model along with an addi-
tional regressor that modeled physical shock events
(500 mseg). As noted above, these events were dis-
carded in the main analyses to minimize potential con-
tributions from actual electrical stimulation. Entonces, para
each cluster, we ran a robust correlation (Rousselet &
Pernet, 2012; Wilcox, 2012) across participants. For each
partícipe, we considered the average regression coeffi-
cient corresponding to the proximity by direction inter-
acción (from the original model so as to estimate it with
minimal contamination from shocks) and regression
coefficient corresponding to physical shock events.
Plotting Parametric Effects as a Function
of Proximity
Ecuación 1 allowed us to estimate the contributions of
the seven main regressors to fMRI responses. Porque
of the parametric nature of the design, to illustrate re-
sponses in a more intuitive manner, we estimated re-
sponses separately for approach and retreat for a range
of proximity values (Cifra 8). para hacerlo, the value of
z scored proximity was varied (in the range of [−2, 1.5]
and at the mean speed value), and the estimated regres-
sion coefficients were used to estimate the response at
each value of proximity.
To provide an indication of variability of the fit across
Participantes, we adopted the following approach. En el
case of the proximity by direction interaction (Figures 8
and 11A), at each level of proximity, we calculated the
difference between the estimated response for the ap-
proach and retreat conditions. We then calculated the
standard error of the approach-minus-retreat difference
across participants (at each value of proximity). Nosotros
Mesa 1. SCR Results
Regressor
Proximity
Direction
Speed
Direction × Speed
Proximity × Direction
Proximity × Speed
Proximity × Direction × Speed
t(81)
4.57
9.37
−4.20
−0.92
10.99
−2.43
−2.78
pag
.0000
.0000
.0001
.3602
.0000
.0175
.0067
Bonferroni correction for multiple comparisons: 0.05/7 = 0.0071.
Meyer, padmala, and Pessoa
527
Cifra 4. Skin conductance response proximity by direction
interacción. Estimated responses for a range of proximity values. A
display estimated responses, we varied proximity and estimated the
response based on the linear model for SCR (analogous to the model of
Ecuación 1). The approach versus retreat difference was greater when
circles were near compared with far. The confidence bands were
obtained by considering within-subject differences (approach minus
retreat); see Methods. A.U. = arbitrary units.
a canonical SCR model based on the sigmoid exponen-
tial function (Lim y col., 1997; Cifra 3). Además,
constant and linear terms were included (for each run
separately) to model baseline and drifts of the SCR. A
minimize effects due to the physical shock event, datos
points in a 15-sec window after shock delivery were
discarded from the analysis. A nivel de grupo, to probe
the effects of the regressors of interest, we ran separate
one-sample t tests against zero using the corresponding
regression coefficients from the individual-level analysis.
Relationship between SCR and Brain Activity
To probe the relationship between brain activity and
physiological arousal, we focused on the right anterior
insula and the right amygdala clusters that exhibited a
proximity by direction interaction (see Results). For each
grupo, an interaction index was created by averaging
the corresponding regression coefficients (βPD in Equa-
ción 1) from all the voxels within the cluster (después
cluster-level thresholding). Entonces, for each cluster, nosotros
ran a robust correlation (Rousselet & Pernet, 2012;
Wilcox, 2012) across participants. Para cada participante,
we considered the average fMRI interaction regression
coefficient and the corresponding interaction term in
the SCR data (específicamente, the coefficient βPD obtained
from the SCR regression analysis).
Relationship between Threat Anticipation and
Physical Shock Responses
In an exploratory analysis, we probed the relationship
between activity related to threat anticipation and
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Cifra 5. Brain responses as a
function of threat proximity.
Clusters in red show regions
with stronger responses for
closer versus farther; grupos
in blue show the reverse.
Clusters were thresholded at
a whole-brain corrected alpha
de .05. SMA = supplementary
motor area; FEF = frontal eye
campo; IPS = intraparietal sulcus;
SFG = superior frontal gyrus;
PCC = posterior cingulate
corteza; vmPFC = ventromedial
corteza prefrontal.
display the 95% confidence bands at each proximity value
(note that because the intervals were based on differ-
ences between approach and retreat conditions, el
same band widths are used for approach and retreat).
An analogous procedure was used for the proximity by
direction interaction of SCRs (Cifra 4). The BST exhib-
ited a proximity effect but no interaction. Por lo tanto, en
Cifra 9 we computed error bands separately for ap-
proach and retreat based on the variability of estimated
responses across participants as a function of proximity.
Statistical Approach and p Values
The null hypothesis significance testing framework has
come under increased scrutiny in recent years. In partic-
ular, the hard threshold of .05 has come under attack,
with reasonable researchers calling for both stricter
umbrales (Benjamin et al., 2018) o, conversely, para
p values to be abandoned (McShane, Gal, Gelman,
Roberto, & Tackett, 2017). Sin embargo, like McShane and
colegas, we do not consider a binary threshold to be
satisfactory and believe that p values should be treated
continuously. Respectivamente, in select cases, we show
p values and discuss findings that do not survive correc-
tion for multiple comparisons; in the context of Table 9,
we discuss the general results of the BST given its im-
portant role in threat-related processing.
RESULTADOS
Our paradigm allowed us to investigate the role played
by threat proximity, direction, and speed and their
interactions on SCRs and fMRI responses. Intuitivamente,
interactions evaluated the extent to which factor com-
binations were relevant in explaining the data. Para
instancia, the contrast of approach versus retreat (direction)
was anticipated to depend on proximity (Figura 1C).
Además, as proximity and speed varied continu-
iosamente, their roles and their interactions were assessed
parametrically.
Our design did not include a standard control condi-
ción (p.ej., circles colliding but no shock administered),
as often is the case in fMRI studies. Nota, sin embargo, eso
our main goal was not to investigate the shock event
itself but potential threat. De este modo, approach and retreat
can be viewed as paired conditions insofar as processes
Cifra 6. Brain responses as a function of direction (approach vs. retreat). Clusters in red show regions with stronger responses for approach
versus retreat; clusters in blue show the reverse. Clusters were thresholded at a whole-brain corrected alpha of .05. PAG = periaqueductal gray;
SMA = supplementary motor area; FEF = frontal eye field; IPS = intraparietal sulcus; PreCG = precentral gyrus.
528
Revista de neurociencia cognitiva
Volumen 31, Número 4
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Mesa 2. Clusters that Exhibited the Effect of Proximity in Voxelwise Analysis at Whole-brain Cluster-level Corrected Alpha of .05
k
12470
1690
1489
1453
1188
1088
995
869
796
576
526
336
333
209
138
125
118
117
117
88
83
79
72
70
68
66
65
61
55
42
40
36
34
34
34
31
25
24
20
X
−14
36
−34
28
64
14
−16
−60
−32
−2
8
−22
−12
44
22
−4
−34
−62
18
26
−26
−12
4
58
56
−26
−34
42
−42
32
−12
−58
36
4
−4
20
−22
−4
38
y
−88
22
−92
−90
−38
10
−76
−46
22
46
−18
26
−72
−58
32
−20
48
−6
6
42
6
−54
32
−30
−4
−24
−6
−12
−24
40
−2
−22
−70
12
−68
−14
−8
−24
−12
z
28
8
−6
4
30
64
−34
42
6
−10
6
46
−44
−30
48
30
28
−14
18
22
−10
66
48
−6
−18
54
50
48
18
−10
66
48
−10
−10
50
−24
−22
54
−8
t
−13.47
7.93
8.84
8.93
7.83
6.40
7.84
6.58
7.28
−5.91
7.49
−5.23
5.89
5.88
−4.55
5.27
4.96
−5.84
5.41
4.47
5.18
−5.71
4.88
4.36
−4.98
−4.83
4.57
−4.85
−4.11
−5.02
4.73
−4.09
5.04
−5.17
−4.38
−5.47
−4.36
−4.52
4.36
Cluster
Occipital cortex/cuneus/posterior cingulate cortex
Right anterior/mid-insula
Left inferior/middle occipital gyrus
Right inferior/middle occipital gyrus
Right supramarginal/postcentral gyrus
Right superior frontal gyrus
Left cerebellum
Left supramarginal gyrus
Left anterior insula
Ventromedial prefrontal cortex
Right/left thalamus
Left superior frontal gyrus
Left cerebellum
Right cerebellum
Right superior frontal gyrus
Right/left posterior cingulate cortex
Left middle frontal gyrus
Left middle temporal gyrus
Right dorsolateral caudate
Right middle frontal gyrus
Left putamen
Left precuneus/superior parietal lobule
Medial superior frontal gyrus
Right superior/middle temporal gyrus
Right middle temporal gyrus
Left precentral gyrus
Left middle frontal gyrus
Right precentral gyrus
Left posterior insula
Right lateral orbitofrontal cortex
Left superior frontal gyrus
Left postcentral gyrus
Right inferior temporal gyrus
Subcollosal area
Precuneus
Right hippocampus/amygdala
Left hippocampus/amygdala
Left paracentral lobule
Right planum polare
Meyer, padmala, and Pessoa
529
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Mesa 2. (continued )
k
20
17
17
17
17
17
16
15
14
13
X
54
12
−52
−16
66
14
−26
12
−56
−18
y
−26
12
−28
64
−4
−54
62
−74
−50
−34
z
10
−2
10
14
18
68
−12
−40
−10
72
t
−4.20
5.02
−3.91
−4.22
−4.78
−4.57
4.42
5.49
−4.11
−4.34
Cluster
Right tranverse temporal gyrus
Right ventral caudate
Left tranverse temporal gyrus
Right superior frontopolar gyrus
Right precentral gyrus
Right precuneus/superior parietal lobule
Left frontomarginal gyrus
Right cerebellum
Left superior/middle temporal gyrus
Left postcentral gyrus
Peak MNI coordinates, t(84) valores, and cluster size (k) refer to number of 2.0 × 2.0 × 2.0 mm3 voxels. Peak coordinates are presented for
completeness and potential meta-analysis; with cluster-based thresholding, it is not possible to conclude that all the reported peaks were activated
(see Woo, Krishnan, & Apostar, 2014).
related to tracking the movement of the circles are con-
cerned, Por ejemplo. Además, as stated in the pre-
ceding paragraph, an important focus of the research
was to assess whether or not brain regions were sensitive
to variable interactions, an approach that further helped
reduce the contributions of non-threat related processing
(see also Discussion).
Skin Conductance Responses
Analysis of SCR data revealed that all three main variables
had robust effects on responses (Mesa 1). Además, nosotros
detected an interaction of proximity by direction; en esto
caso, responses to approach versus retreat were sensitive
to threat distance, such that the effect was larger when
near versus far. To visualize this result, Cifra 4 muestra
estimated SCRs for approach and retreat for a range of
proximity values (because the circles moved contin-
uously on the screen, Cifra 4 used an approach similar
to that of Figure 8 for plotting; see Methods). Finalmente, a
three-way interaction between proximity, direction, y
speed also survived correction for multiple comparisons.
fMRI Voxelwise Analysis
Figures 5 y 6 (Tables 2 y 3) show the effects of
proximity and direction (Mesa 4 shows the effect of
velocidad). The main focus of this study was to investigate
interactions between threat-related factors. Cifra 7
(Mesa 5) shows interactions between proximity and
direction; positive voxels (rojo) show effects when the
contrast of approach versus retreat was greater during
closer versus farther circles, and blue voxels indicate the
opposite. Cifra 8 shows estimated responses for
approach and retreat for a range of proximity values, cual
aids in visualizing the parametric effects of proximity on
the signals in the two regions (see Methods). Para el
right anterior insula (Figure 8A), when the circles were
closer to each other, a larger approach versus retreat
differential response was observed compared with when
the circles were farther from each other. Responses
for the right amygdala (Figure 8B) exhibited the opposite
pattern as responses were larger for retreat compared
with approach, and the contrast was enhanced when
circles were closer compared with farther. Tables 6 y
7 show two-way interactions between direction and
speed and between proximity and speed. Mesa 8 muestra
the three-way interaction of proximity, direction, y
velocidad.
BST ROI Analysis
Given that the BST is a rather small region that is in-
volved in threat-related processing, we ran a focused
ROI analysis using anatomically defined left/right BST
masks and unsmoothed data to minimize the influence
of signals from surrounding structures. We observed a
robust effect of threat proximity in the right BST (y
weak evidence in the left BST), with stronger responses
when circles were closer than farther (Figure 9A; Mesa 9).
For the right BST, some evidence for proximity by speed
interaction was seen.
Relationship between SCR Responses and
Brain Activity
We evaluated the linear relationship between SCR and
fMRI by running a robust correlation analysis (across
Participantes). Because multiple aspects of both the SCR
530
Revista de neurociencia cognitiva
Volumen 31, Número 4
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Mesa 3. Clusters that Exhibited the Effect of Direction in Voxelwise Analysis at Whole-brain Cluster-level Corrected Alpha of .05
k
761
633
572
565
546
472
277
269
246
230
205
151
144
134
130
108
106
105
96
90
82
63
60
43
38
37
37
35
26
26
24
23
22
22
22
21
19
18
18
X
−36
34
−46
−24
48
36
26
−28
26
36
14
−26
−56
−34
54
56
−16
−12
32
−30
−12
−28
14
12
6
−8
−42
46
−40
−12
−12
44
10
2
−20
20
26
−18
12
y
−44
−50
−62
−8
−60
−4
−98
−72
−68
28
−86
−98
2
18
6
−44
−86
−94
−72
−28
−76
−58
−80
−70
−28
−74
−74
20
−54
−46
−22
−46
−74
−54
20
−24
8
−72
2
z
56
60
12
52
10
50
−4
26
−4
4
24
−4
38
6
34
20
22
16
32
52
−8
−8
2
−20
−10
−42
−8
26
−18
52
40
−14
−38
−32
48
66
−10
−22
70
t
7.77
6.67
7.29
8.40
6.78
6.62
−6.22
7.02
−6.49
7.45
−6.17
−6.63
6.91
6.73
5.89
4.73
−5.48
−5.63
5.32
−4.79
−5.77
−4.87
−4.54
5.39
4.71
5.30
4.35
4.88
4.46
5.04
4.68
4.93
6.09
4.59
−4.80
−4.74
4.60
4.60
4.81
Cluster
Left inferior parietal cortex
Right inferior parietal cortex
Left superior temporal gyrus
Left frontal eye field
Right superior temporal gyrus
Right frontal eye field
Right superior occipital gyrus
Left parieto-occipitalis
Right lingual gyrus
Right anterior insula
Right parieto-occipitalis (posterior)
Left superior occipital gyrus
Left precentral gyrus
Left anterior insula
Right precentral gyrus
Right parietal operculum
Left parieto-occipitalis (posterior)
Left parieto-occipitalis (posterior)
Right parieto-occipitalis
Left postcentral gyrus
Left lingual gyrus
Left lingual/fusiform gyrus
Right occipital gyrus
Right cerebellum
Right periaqueductal gray
Left cerebellum
Left inferior temporal gyrus
Right inferior/middle frontal gyrus
Left fusiform gyrus
Left paracentral lobule
Left posterior cingulate cortex
Right fusiform gyrus
Right cerebellum
Cerebellum
Left superior frontal gyrus
Right postcentral gyrus
Right putamen
Left cerebellum
Right superior frontal gyrus
Meyer, padmala, and Pessoa
531
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Mesa 3. (continued )
k
16
13
13
X
36
42
8
y
−14
42
−86
z
44
−2
18
t
−4.56
−4.33
−4.80
Cluster
Right precentral gyrus
Right inferior frontal gyrus
Right parieto-occipitalis (posterior)
Peak MNI coordinates, t(84) valores, and cluster size (k) refer to number of 2.0 × 2.0 × 2.0 mm3 voxels. Peak coordinates are presented for
completeness and potential meta-analysis; with cluster-based thresholding, it is not possible to conclude that all the reported peaks were activated
(see Woo et al., 2014).
Mesa 4. Clusters that Exhibited the Effect of Speed in Voxelwise Analysis at Whole-brain Cluster-level Corrected Alpha of .05
k
4402
3332
339
283
235
178
120
114
112
110
102
63
50
45
42
35
33
29
28
25
24
19
17
16
15
14
13
13
13
X
−46
46
−26
−32
26
−32
36
36
−6
−12
−50
10
54
4
50
12
18
20
−34
−12
−14
−14
−36
24
12
−8
−8
36
−50
y
−74
−68
−56
−4
−50
24
−4
24
6
−24
4
20
−44
0
4
−94
4
−74
−46
−74
−30
−46
−12
−70
−20
−72
26
2
−26
z
0
4
52
46
52
6
50
6
50
40
36
36
20
56
34
20
64
40
−20
12
−2
48
−6
10
40
−38
30
34
36
t
9.74
9.38
5.70
5.68
6.23
6.92
5.35
5.42
5.52
6.08
6.63
5.22
4.98
4.50
5.36
4.48
4.66
4.81
5.21
4.56
5.14
4.43
4.36
4.07
4.00
5.52
4.10
4.26
4.50
Cluster
Left inferior/middle temporal gyrus/fusiform gyrus
Right inferior/middle temporal gyrus/fusiform gyrus
Left intraparietal sulcus
Left frontal eye field
Right intraparietal sulcus
Left anterior insula
Right frontal eye field
Right anterior insula
Left mid-cingulate cortex/supplementary motor area
Left posterior cingulate cortex
Left precentral gyrus
Right mid-cingulate cortex
Right parietal operculum
Right supplementary motor area
Right precentral gyrus
Right parieto-occipitalis (posterior)
Right superior frontal gyrus
Right parieto-occipitalis
Left fusiform gyrus
Left precuneus/occipital gyrus
Left ventral thalamus
Left superior parietal lobule
Left postcentral insular cortex
Right precuneus/occipital gyrus
Right posterior cingulate cortex
Left cerebellum
Left mid-cingulate cortex
Right precentral gyrus
Left supramarginal gyrus
Peak MNI coordinates, t(84) valores, and cluster size (k) refer to number of 2.0 × 2.0 × 2.0 mm3 voxels. Peak coordinates are presented for completeness
and potential meta-analysis; with cluster-based thresholding, it is not possible to conclude that all the reported peaks were activated (see Woo et al., 2014).
532
Revista de neurociencia cognitiva
Volumen 31, Número 4
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Cifra 7. Brain responses
exhibiting a proximity by
direction (approach vs. retreat)
interaction in areas of interest.
Clusters in red show regions
with approach versus retreat
responses greater when closer
versus farther; clusters in blue
show the reserve pattern.
Clusters were thresholded at
a whole-brain corrected alpha
de .05. FEF = frontal eye field;
PreCG = precentral gyrus.
Mesa 5. Clusters that Exhibited the Proximity × Direction Interaction in Voxelwise Analysis at Whole-brain Cluster-level Corrected
Alpha of .05
k
528
398
358
263
243
215
159
148
138
109
109
105
99
98
74
73
54
52
35
34
31
31
29
27
X
10
−10
28
−20
16
26
−54
−16
−28
−54
50
16
66
−30
34
40
22
20
8
18
−18
−36
54
−54
y
−74
−96
−96
−12
−86
−2
−32
−86
−98
4
−32
−92
−16
−28
28
−60
−32
−60
−44
−74
−66
−16
4
−22
z
−2
12
2
60
26
58
−2
22
0
38
58
18
20
52
2
50
72
12
62
24
8
16
36
54
t
−9.42
−7.09
−5.98
6.26
−7.16
5.55
−5.16
−5.94
−5.24
6.72
−4.88
−5.51
−4.95
−4.83
5.63
−4.61
−5.81
−4.85
−4.42
−5.00
−5.02
−4.80
5.11
−4.95
Cluster
Right occipital cortex
Left occipital gyrus
Right occipital gyrus
Left frontal eye field
Right occipital gyrus
Right frontal eye field
Left superior temporal gyrus
Left occipital gyrus
Left inferior occipital gyrus
Left precentral gyrus
Right postcentral gyrus
Right occipital gyrus
Right supramarginal gyrus
Left precentral gyrus
Right anterior Insula
Right angular gyrus
Right postcentral gyrus
Right occipital gyrus
Right superior postcentral sulcus
Right posterior angular gyrus
Left occipital gyrus
Left posterior insula
Right precentral gyrus
Left postcentral gyrus
Meyer, padmala, and Pessoa
533
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Mesa 5. (continued )
k
25
24
24
23
22
22
21
20
20
20
19
18
17
15
15
15
14
13
X
−26
−60
38
56
28
−50
−36
−54
−50
−2
32
58
44
54
−8
32
−52
−36
y
−70
−32
−16
−58
−10
22
−18
24
−20
22
−76
−58
−70
20
−76
−36
22
−60
z
−28
16
40
−6
−20
−8
44
18
18
54
−38
28
−20
−4
16
50
24
−42
t
−4.45
−4.51
−5.01
−4.01
−4.16
−4.66
−5.06
−4.34
−4.87
−4.24
−4.32
−4.07
−3.88
−4.59
−4.19
4.21
−3.98
−3.70
Cluster
Left cerebellum
Left supramarginal gyrus
Right postcentral gyrus
Right inferior temporal gyrus
Right amygdala
Left inferior temporal gyrus
Left postcentral gyrus
Left inferior frontal gyrus
Left parietal operculum
Left paracentral lobule
Right cerebellum
Right supramarginal gyrus
Right inferior temporal gyrus
Right inferior temporal gyrus
Left precuneus
Right postcentral gyrus
Left middle frontal gyrus
Left cerebellum
Peak MNI coordinates, t(84) valores, and cluster size (k) refer to number of 2.0 × 2.0 × 2.0 mm3 voxels. Peak coordinates are presented for
completeness and potential meta-analysis; with cluster-based thresholding, it is not possible to conclude that all the reported peaks were activated
(see Woo et al., 2014).
and fMRI data could be probed (simple effects and inter-
comportamiento), we chose to focus the interrogation on the prox-
imity by direction interaction. De este modo, for both SCR and
resonancia magnética funcional, the strength of the two-way interaction was
considered for the analysis (as given by the regression
coefficient in Equation 1). To minimize the problem of
multiple statistical comparisons, for this analysis, we fo-
cused on clusters exhibiting a two-way interaction in the
Cifra 8. Proximity by direction (approach vs. retreat) interacción. Estimated responses for a range of proximity values. (A) For the right anterior
insula, activity increased as a function of proximity for both approach and retreat, but more steeply for the former. (B) For the right amygdala,
activity decreased as a function of proximity during approach, but changed little during retreat. The confidence bands were obtained by considering
within-subject differences (approach minus retreat); see Methods. A.U. = arbitrary units.
534
Revista de neurociencia cognitiva
Volumen 31, Número 4
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Mesa 6. Clusters that Exhibited the Direction × Speed Interaction in Voxelwise Analysis at Whole-brain Cluster-level Corrected
Alpha of .05
k
125
105
95
95
55
46
30
28
22
22
22
21
18
18
17
17
15
15
15
14
14
13
13
13
X
−46
−30
34
−32
34
16
−8
−30
−24
−8
−40
8
−30
16
−26
66
0
40
30
−30
−4
46
20
−4
y
−74
−100
−94
−26
26
−26
−22
−46
−78
−72
−32
−76
26
34
−54
−2
6
−2
−26
26
−24
48
46
−78
z
0
−2
0
58
0
68
62
42
30
34
42
−2
−4
56
−8
16
32
48
60
6
56
−8
36
42
t
4.80
−5.40
−4.82
−4.34
4.44
−4.54
−4.53
4.33
5.11
4.06
4.53
−4.71
4.68
−4.58
−3.94
−4.99
4.34
4.21
−4.11
4.52
−4.55
−4.55
−4.54
3.94
Cluster
Left middle temporal gyrus
Left superior occipital gyrus
Right middle occipital gyrus
Left precentral gyrus
Right anterior insula
Right precentral gyrus
Left paracentral lobule
Left inferior parietal cortex
Left parieto-occipitalis
Left precuenus
Left inferior postcentral sulcus
Right lingual gyrus
Left anterior insula
Right superior frontal gyrus
Left parahippocampal gyrus
Right precentral gyrus
Mid-cingulate cortex
Right precentral gyrus
Right precentral gyrus
Left anterior insula
Left paracentral lobule
Right inferior frontal gyrus (orbital)
Right superior frontal gyrus
Left precuneus
Peak MNI coordinates, t(84) valores, and cluster size (k) refer to number of 2.0 × 2.0 × 2.0 mm3 voxels. Peak coordinates are presented for
completeness and potential meta-analysis; with cluster-based thresholding, it is not possible to conclude that all the reported peaks were activated
(see Woo et al., 2014).
right anterior insula and the right amygdala, regions that
feature in most models of threat processing. We did not
detect a relationship between SCR and fMRI responses
in either the right anterior insula, r(77) = .07, pag = .550,
or the right amygdala, r(75) = -.04, pag = .697.
Relationship between Anticipatory Activity and
Physical Shock Responses
Our interpretation of the proximity by direction inter-
action was that it reflected, at least in part, threat-related
Procesando, especially in brain regions important for this
type of processing, such as the anterior insula. In an ex-
ploratory analysis, we tested if the strength of this inter-
action effect was associated (across participants) con el
strength of responses evoked by physical shock. Para el
right anterior insula cluster that exhibited a proximity by
direction interaction, we detected a positive linear rela-
tionship between the two measures, r(80) = .33, pag =
.002 (Cifra 10). Given the importance of the amygdala
in threat processing, we also tested the relationship in
the right amygdala (also considering the cluster that
exhibited a proximity by direction interaction), but no
effect was detected, r(80) = -.02, pag = .888.
Individual Differences in State and Trait Anxiety
Linear relationships between state/trait anxiety and SCR
o, separately, fMRI interactions of proximity and direc-
tion in the right anterior insula were not detected (todo
Meyer, padmala, and Pessoa
535
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Mesa 7. Clusters that Exhibited the Proximity × Speed Interaction in Voxelwise Analysis at Whole-brain Cluster-level Corrected
Alpha of .05
k
116
50
46
39
37
33
23
22
21
19
15
13
13
13
X
4
−6
48
26
−30
30
40
8
−62
18
46
−24
12
14
y
−2
16
−66
−90
28
−74
−80
22
4
−66
−58
−96
−72
−54
z
56
34
4
6
2
38
20
38
8
52
−6
2
6
64
t
5.58
4.68
−5.23
−5.35
5.25
−4.46
−4.38
4.65
4.19
−4.47
−4.49
−4.09
4.15
−3.85
Cluster
Supplementary motor area
Left mid-cingulate cortex
Right middle temporal gyrus
Right occipital gyrus
Left anterior insula
Right parieto-occipitalis
Right occipital gyrus
Right mid-cingulate cortex
Left precentral gyrus
Right superior parietal lobule
Right inferior temporal gyrus
Left middle occipital gyrus
Right occipital gyrus
Right superior parietal lobule
yo
D
oh
w
norte
oh
a
d
mi
d
F
r
oh
metro
h
t
t
pag
:
/
/
d
i
r
mi
C
t
.
metro
i
t
.
mi
d
tu
/
j
/
oh
C
norte
a
r
t
i
C
mi
–
pag
d
yo
F
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
oh
C
norte
_
a
_
0
1
3
6
3
pag
d
.
F
b
y
gramo
tu
mi
s
t
t
oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3
Peak MNI coordinates, t(84) valores, and cluster size (k) refer to number of 2.0 × 2.0 × 2.0 mm3 voxels. Peak coordinates are presented for
completeness and potential meta-analysis; with cluster-based thresholding, it is not possible to conclude that all the reported peaks were activated
(see Woo et al., 2014).
rs < .1 in absolute value). We detected a modest positive
relationship between state anxiety and fMRI interactions
of proximity and direction in the right amygdala (in the
cluster that exhibited a proximity by direction inter-
action; state: r(77) = .2107, p = .0544; trait: r(79) =
.0769, p = .4870). Given the multiple tests involved here,
we do not believe these findings are noteworthy.
the direction effect (approach vs. retreat) previously re-
ported (Figure 6). Upon plotting, we discerned an effect
of proximity for the approach condition, but not for re-
treat, consistent with a proximity by direction interaction
(which was not detected in the voxelwise analysis). Given
the importance of the PAG in the orchestration of defen-
sive responses in the face of threat (Pessoa, 2016;
Exploratory Analyses: PAG Responses
To visualize the responses of the PAG, we plotted esti-
mated responses (Figure 11A), as done above for the
right anterior insula, right amygdala, and right BST. To
do so, we used the cluster (38 voxels) that exhibited
Table 8. Clusters that Exhibited the Proximity × Direction ×
Speed Interaction in Voxelwise Analysis at Whole-brain Cluster-
level Corrected Alpha of .05
k
x
y
17 −54 −24
z
38
t
Cluster
3.98
Left central sulcus
17
24
30
52 −5.02
Right superior frontal gyrus
Peak MNI coordinates, t(84) values, and cluster size (k) refer to number
of 2.0 × 2.0 × 2.0 mm3 voxels. Peak coordinates are presented for
completeness and potential meta-analysis; with cluster-based threshold-
ing, it is not possible to conclude that all the reported peaks were
activated (see Woo et al., 2014).
Figure 9. Proximity effect in the BST ROI analysis. Estimated responses
for a range of proximity values. Activity increased as a function of
proximity for both approach and retreat. The confidence bands were
obtained by considering variability during approach and retreat,
separately; see Methods. A.U. = arbitrary units.
536
Journal of Cognitive Neuroscience
Volume 31, Number 4
Table 9. BST ROI Analysis Results
Regressor
Proximity
Direction
Speed
Direction × Speed
Proximity × Direction
Proximity × Speed
Left BST
Right BST
t(84)
p
t(84)
p
1.91 .0591
4.17 .0000
1.29 .1997
0.64 .0000
0.83 .4099
1.78 .0001
−0.91 .3664 −0.26 .3602
−1.63 .1066 −0.35 .0000
−0.08 .9353
2.60 .0175
Proximity × Direction × Speed
0.00 .9986 −1.69 .0067
Bonferroni correction for multiple comparisons: 0.05/7 = 0.0071.
Exploratory Analyses: Potential Nonlinear Effects
of Proximity
The regression model we used (Equation 1) makes
the assumption that the effect of proximity is linear.
In additional exploratory analyses, we investigated
potential nonlinear effects of proximity on brain activ-
ity. To do so, we inspected the pattern of the residuals
as a function of proximity in the right anterior insula,
right amygdala, right BST, and right PAG. For example,
Figure 12 shows the residuals when using Equation 1
for the right anterior insula. Based on the pattern of
residuals, the linear modeling approach adopted
here appears to be reasonable in the context of our
experiment.
Bandler & Shipley, 1994), we performed an additional
exploratory analysis in this region. First, we generated a
representative time series for the PAG by averaging the
time series of the voxels within the cluster (based on
the voxelwise effect of direction) and then evaluated
the full model (Equation 1). As shown in Table 10, a
robust proximity by direction interaction was detected
(note that the interaction effects were nearly inde-
pendent from the selection criterion, which was based
on direction; the correlation between the interaction
and direction was −.14). Given this result, we inspected
again the results at the voxelwise level and observed
some voxels that exhibited such an interaction, but too
few to survive cluster thresholding.
Notably, we also observed a robust three-way inter-
action. As the three factors simultaneously affected PAG
responses, the finding can be visualized via a contour
plot (Figure 11B). During approach periods, when prox-
imity increased (circles moved closer to each other),
stronger responses were observed as speed increased
from slower to faster (compare the top right vs. bottom
left quadrants).
DISCUSSION
In this study, we investigated the role of threat-related
factors and their temporally evolving interactions. Our
findings support the view that threat processing is
context sensitive and dynamic (Mobbs et al., 2015;
Kavaliers & Choleris, 2001; Blanchard & Blanchard,
1990; Fanselow & Lester, 1988). In some brain regions,
signal fluctuations were sensitive to continuous manipu-
lations of proximity and speed indicating that threat pro-
cessing is dynamic. Importantly, whereas some brain
regions tracked individual threat-related factors (prox-
imity, direction, or speed), others were also sensitive to
combinations of these variables revealing the context-
sensitive nature of threat processing. In this section, we
will focus the discussion on a few of the brain regions
that have been most heavily implicated in threat-related
processing in the literature, specifically the anterior
insula, amygdala, BST, and PAG.
To investigate how threat-related factors influence
physiological arousal during dynamic threat anticipation,
we recorded SCR during scanning. We observed robust
effects of proximity and direction, with larger responses
Figure 10. Relationship
between anticipatory activity
and physical shock responses
in the right anterior insula.
For the anticipatory activity,
the proximity by direction
interaction was considered
for the analysis. Data points
correspond to participants
(red points indicate outliers
deemed based on the robust
correlation algorithm).
A.U. = arbitrary units.
Meyer, Padmala, and Pessoa
537
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
d
i
r
e
c
t
.
m
i
t
.
e
d
u
/
j
/
o
c
n
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
o
c
n
_
a
_
0
1
3
6
3
p
d
.
f
b
y
g
u
e
s
t
t
o
n
0
8
S
e
p
e
m
b
e
r
2
0
2
3
Table 10. Exploratory Right PAG ROI Analysis Results
Regressor
Proximity
Direction
Speed
Direction × Speed
Proximity × Direction
Proximity × Speed
Proximity × Direction × Speed
t(84)
2.33
6.75
2.69
1.92
3.89
1.17
2.86
p
.0220
.0000
.0087
.0588
.0002
.2455
.0054
Bonferroni correction for multiple comparisons: 0.05/7 = 0.0071.
during near versus far and approach versus retreat, re-
spectively. Of note, we observed a robust proximity by
direction interaction, where responses to threat direction
(approach vs. retreat) were enhanced when the circles
were “near” compared with “far,” suggesting that the in-
fluence of dynamic threat anticipation on physiological
arousal was context dependent.
Responses in the anterior insula were driven by prox-
imity, direction, and speed. Importantly, in the right
hemisphere, anterior insula responses also exhibited
an interaction between proximity and direction, such
that the approach versus retreat contrast was enhanced
when the circles were “near” compared with far. The an-
terior insula supports subjective awareness of bodily
states (Craig, 2002, 2009) and is consistently engaged
during threat-related processing (Nitschke et al., 2006;
Simmons et al., 2006). In particular, the anterior insula
is implicated in tracking threat proximity and direc-
tion during aversive anticipation (Mobbs et al., 2010;
Somerville et al., 2010). Our results replicated these
findings while extending them by showing that the effects
of threat proximity and direction are not independent
but jointly contribute to responses in the anterior insula.
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
d
i
r
e
c
t
.
m
i
t
.
e
d
u
/
j
/
o
c
n
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
o
c
n
_
a
_
0
1
3
6
3
p
d
.
f
b
y
g
u
e
s
t
t
o
n
0
8
S
e
p
e
m
b
e
r
2
0
2
3
Figure 11. Exploratory analysis of the PAG. (A) Estimated responses for a range of proximity values. During approach, activity increased as a
function of proximity; activity changed little during retreat periods. The confidence bands were obtained by considering within-subject differences
(approach minus retreat); see Methods. A.U. = arbitrary units. (B) Contour plots show estimated responses for different combinations of proximity
and speed during approach and retreat periods. Arrows point in the direction of signal increase. During approach, both proximity and speed
simultaneously influenced responses, which increased when the circles were closer and speed was higher. A.U. = arbitrary units.
538
Journal of Cognitive Neuroscience
Volume 31, Number 4
published (Theiss et al., 2017; Torrisi et al., 2015; Avery
et al., 2014), which should enhance the reproducibility of
published findings. We analyzed BST data using an ana-
tomical mask and unsmoothed data, which is important
because nearly all studies have used some voxelwise
spatial smoothing, which blends BST signals with those
of adjacent territories (beyond the inherent point spread
function of imaging itself ), but note that smoothing within
the BST was accomplished by averaging unsmoothed
time series of voxels within the anatomically defined
ROI. In the right BST, we observed an effect of proximity
and a proximity by speed interaction (but note that these
effects were less robust as they would not survive cor-
rection for the seven tests used or 14 if one were to
consider both hemispheres). The observed effect of prox-
imity is consistent with previous findings that the BST re-
sponds to threat proximity (independent of direction;
Somerville et al., 2010; Mobbs et al., 2010; although the
activated region was sufficiently large as to make anatomi-
cal localization challenging in the study by Mobbs and
colleagues).
The PAG of the midbrain has been implicated in aver-
sive and defensive reactions (Bandler & Shipley, 1994;
Bandler, 1988), in line with more recent studies (Tovote
et al., 2016). In humans, the PAG has been suggested to
be involved in negative emotional processing more
generally (Satpute et al., 2013; Lindquist, Wager, Kober,
Bliss-Moreau, & Barrett, 2012). The virtual tarantula
manipulation by Mobbs et al. (2010), where participants
were shown a prerecorded video of a spider moving to-
ward or away from their feet was particularly effective in
engaging the PAG when threat was proximal (although
the activation was very extensive and thus difficult to
localize). Here, in the voxelwise analysis, we only de-
tected an effect of direction in the right midbrain/PAG
where stronger responses were observed when circles
were approaching compared with retreating. However,
exploratory analyses revealed a robust proximity by
direction interaction, as well as a proximity by direction
by speed interaction. These results are potentially im-
portant because they suggest that threat-related re-
sponses in the PAG are sensitive to multiple factors that
jointly determine the PAG’s activity. Interestingly, unlike
in the amygdala and anterior insula where we only ob-
served an interaction between proximity and direction,
speed also played a role in the PAG. However, given the
exploratory nature of our analysis, future converging find-
ings are needed to more precisely delineate the role of
multiple threat-related factors on PAG activity during
aversive anticipation.
A limitation of this study was that it did not include two
types of control condition. First, only aversive events
were encountered and not motivationally positive ones.
Thus, the extent to which signals investigated here were
linked to threat and not “motivational significance” more
generally needs to be further investigated. Second, be-
cause a “no-shock condition” was not included, it is
Meyer, Padmala, and Pessoa
539
Figure 12. Exploratory analysis of potential nonlinear effects of
proximity. The residuals from the model fit are plotted as a function of
proximity. No appreciable lack of fit is evident. To plot residuals for
all participants, they were first studentized ( jitter as a function of
proximity was also used to reduce overlap).
In this study, we observed a proximity by direction
interaction in the right amygdala, but in the opposite
direction to that seen in the anterior insula: When far,
direction had a weak or no effect on responses, but
when near responses were greater for retreat relative to
approach. In fact, the differential response to retreat
versus approach became more pronounced as the circles
approached each other, with approach responses de-
creasing with increased proximity. In paradigms in-
vestigating the independent effects of proximity and
direction on threat anticipation, Somerville et al. (2010)
suggested a limited role of the amygdala in tracking
threat proximity, whereas Mobbs et al. (2010) observed
amygdala responses that responded to the proximity
and direction of threat. In a study involving virtual pred-
ators, Mobbs et al. (2007) reported increased activation
in the dorsal amygdala when threat was near, whereas
responses were “stronger” in the inferior-lateral amygdala
with distant threats. Thus, our results more closely re-
semble the latter amygdala subregion. It should be noted
that in previous studies, similar to the pattern of re-
sponses observed in the current study, we and others
have observed amygdala deactivations during short and
long periods of sustained threat (relative to safe con-
ditions; Grupe, Wielgosz, Davidson, & Nitschke, 2016;
McMenamin et al., 2014; Choi, Padmala, & Pessoa, 2012);
see also (Wager et al., 2009; Pruessner et al., 2008) in case
of social stress/threat.
The role of the BST in threat processing has gained
increased attention in the past two decades (Shackman
& Fox, 2016; Davis & Whalen, 2001), especially during
conditions involving temporally extended and less pre-
dictable threats. Given the small size of the structure
and its anatomical location, studying the BST with fMRI
is particularly challenging. Recently, anatomical masks
for both regular and higher field scanning have been
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
d
i
r
e
c
t
.
m
i
t
.
e
d
u
/
j
/
o
c
n
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
o
c
n
_
a
_
0
1
3
6
3
p
d
.
f
b
y
g
u
e
s
t
t
o
n
0
8
S
e
p
e
m
b
e
r
2
0
2
3
possible that signal fluctuations were due to processes
linked to tracking circle movement, including predicting
future circle positions based on current position and
prior movement statistics. In this context, the anterior
insula is an interesting case because it is a highly func-
tionally diverse region and is sensitive to a very broad
range of influences (Anderson, Kinnison, & Pessoa, 2013).
But because anterior insula signals were sensitive to
interactions between proximity and direction, it is un-
likely that prediction/updating processes explained
responses, as participants presumably engaged in such
processing in a similar fashion when the circles were
closer or father. In addition, we observed a positive
correlation between proximity by direction interaction
responses and responses evoked to physical shock, con-
sistent with the fact that responses were at least in part
related to anticipation of the aversive event. Finally, and
more generally, the three-way interaction in the PAG
(but see Results for the exploratory aspect of this result)
exhibited a degree of specificity (compare left and right
panels in Figure 11B) that are difficult to explain by
visuocognitive processes of circle movement tracking.
Another limitation of the preset study was that par-
ticipants did not have control over the threat. Unlike
active avoidance paradigms where participants could
perform instrumental actions to terminate or completely
avoid the threat (for instance, see Mobbs et al., 2007), the
passive nature of our task likely constrained the types of
“defensive processing” observed. In particular, investi-
gation of a richer set of behaviors and brain responses,
such as described in the threat imminence continuum
framework (Fanselow & Lester, 1988), will require novel
approaches and experimental designs attuned to findings
in ethology and behavioral ecology (see Mobbs,
Trimmer, Blumstein, & Dayan, 2018). Finally, our choice
of using nonpainful aversive stimulation was motivated
by our goal to minimize potential harm to participants,
and using painful stimulation likely would have generated
stronger threat-related responses. Of note, a recent meta-
analysis reported a large number of shared neural sub-
strates during the processing of nonpainful and painful
aversive stimuli (Hayes & Northoff, 2012).
To conclude, we investigated how multiple threat-
related factors (proximity, direction, and speed) interact
when varied continuously. In particular, we asked whether
signal fluctuations in brain regions track threat-related
factors dynamically? If so, to what factor(s) and factor
combinations are they sensitive? We observed a prox-
imity by direction interaction in the anterior insula where
approach versus retreat responses were enhanced when
threat was proximal. In the right amygdala, we also ob-
served a proximity by direction interaction, but in the op-
posite direction as that found for the anterior insula; retreat
responses were stronger than approach responses when
threat was proximal. In the right BST, we observed an effect
of proximity and in the right PAG/midbrain we observed an
effect of direction as well as a proximity by direction by
speed interaction (the latter was detected in exploratory
analyses but not in a voxelwise fashion). Overall, this
study refines our understanding of the mechanisms
involved during aversive anticipation in the typical
human brain. Importantly, it emphasizes that threat
processing should be understood in a manner that is
both context sensitive and dynamic. As aberrations in
aversive anticipation are believed to play a major role in
disorders such as anxiety and depression (Dillon et al.,
2014; Grupe & Nitschke, 2013), our findings of inter-
actions between multiple threat-related factors in regions
such as the amygdala, anterior insula, and PAG may
inform the understanding of brain mechanisms that are
dysregulated in these disorders.
Acknowledgments
We would like to thank Brenton McMenamin for paradigm de-
velopment, Dan Levitas for data collection, Jason Smith for help
with processing scripts, Mahshid Najafi for help with initial pre-
processing of the fMRI data, and Nicole Friedman and Jessica
Berman for help with participant recruitment. The authors also
acknowledge the Behavioral and Social Sciences College,
University of Maryland, high-performance computing resources
(http://bsos.umd.edu/oacs/bsos-high-performance) made avail-
able for conducting the research reported in this article. The
authors acknowledge funding from the National Institute of
Mental Health (R01 MH071589 and R01 MH112517) and a
National Science Foundation Graduate Research Fellowship to
S. P.
Reprint requests should be sent to Luiz Pessoa, Department of
Psychology, University of Maryland, 1147 Biology-Psychology
Building, College Park, MD 20742, or via e-mail: pessoa@umd.
edu.
REFERENCES
Anderson, M. L., Kinnison, J., & Pessoa, L. (2013). Describing
functional diversity of brain regions and brain networks.
Neuroimage, 73, 50–58.
Avants, B. B., Tustison, N. J., & Song, G. (2009). Advanced
normalization tools (ANTs). Insight Journal, 2, 1–35.
Avery, S. N., Clauss, J. A., Winder, D. G., Woodward, N., Heckers,
S., & Blackford, J. U. (2014). BNST neurocircuitry in humans.
Neuroimage, 91, 311–323.
Bach, D. R., Flandin, G., Friston, K. J., & Dolan, R. J. (2009).
Time-series analysis for rapid event-related skin conductance
responses. Journal of Neuroscience Methods, 184, 224–234.
Bandler, R. (1988). Brain mechanisms of aggression as revealed
by electrical and chemical stimulation: Suggestion of a central
role for the midbrain periaqueductal grey region. In A. N.
Epstein & A. R. Morrison (Eds.), Progress in psychobiology
and physiological psychology (Vol. 13, pp. 67–154). San Diego:
Academic Press.
Bandler, R., & Shipley, M. T. (1994). Columnar organization in
the midbrain periaqueductal gray: Modules for emotional
expression? Trends in Neurosciences, 17, 379–389.
Benjamin, D. J., Berger, J. O., Johannesson, M., Nosek, B. A.,
Wagenmakers, E.-J., Berk, R., et al. (2018). Redefine statistical
significance. Nature Human Behaviour, 2, 6–10.
Blanchard, R. J., & Blanchard, D. C. (1990). Anti-predator
defense as models of animal fear and anxiety. In P. F. Brain, S.
540
Journal of Cognitive Neuroscience
Volume 31, Number 4
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
d
i
r
e
c
t
.
m
i
t
.
e
d
u
/
j
/
o
c
n
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
o
c
n
_
a
_
0
1
3
6
3
p
d
.
f
b
y
g
u
e
s
t
t
o
n
0
8
S
e
p
e
m
b
e
r
2
0
2
3
Parmigiani, R. J. Blanchard, & D. Mainardi (Eds.), Ettore
Majorana international life sciences series, Vol. 8. Fear and
defence (pp. 89–108). Amsterdam: Harwood Academic
Publishers.
Blanchard, D. C., Griebel, G., Pobbe, R., & Blanchard, R. J.
(2011). Risk assessment as an evolved threat detection and
analysis process. Neuroscience & Biobehavioral Reviews,
35, 991–998.
Böcker, K. B. E., Baas, J. M. P., Kenemans, J. L., & Verbaten,
M. N. (2001). Stimulus-preceding negativity induced by fear:
A manifestation of affective anticipation. International
Journal of Psychophysiology, 43, 77–90.
Brown, C. A., Seymour, B., Boyle, Y., El-Deredy, W., & Jones,
A. K. P. (2008). Modulation of pain ratings by expectation
and uncertainty: Behavioral characteristics and anticipatory
neural correlates. Pain, 135, 240–250.
Choi, J. M., Padmala, S., & Pessoa, L. (2012). Impact of state
anxiety on the interaction between threat monitoring and
cognition. Neuroimage, 59, 1912–1923.
Cohen, M. S. (1997). Parametric analysis of fMRI data using
linear systems methods. Neuroimage, 6, 93–103.
Cox, R. W. (1996). AFNI: Software for analysis and visualization
of functional magnetic resonance neuroimages. Computers
and Biomedical Research, 29, 162–173.
Cox, R. W., Chen, G., Glen, D. R., Reynolds, R. C., & Taylor, P. A.
(2017). fMRI clustering in AFNI: False-positive rates redux.
Brain Connectivity, 7, 152–171.
Craig, A. D. (2002). How do you feel? Interoception: The sense
of the physiological condition of the body. Nature Reviews
Neuroscience, 3, 655–666.
Craig, A. D. (2009). How do you feel—Now? The anterior insula
and human awareness. Nature Reviews Neuroscience, 10,
59–70.
Davis, M., Walker, D. L., Miles, L., & Grillon, C. (2010). Phasic vs.
sustained fear in rats and humans: Role of the extended
amygdala in fear vs. anxiety. Neuropsychopharmacology, 35,
105–135.
Davis, M., & Whalen, P. J. (2001). The amygdala: Vigilance and
emotion. Molecular Psychiatry, 6, 13–34.
Dillon, D. G., Rosso, I. M., Pechtel, P., Killgore, W. D. S.,
Rauch, S. L., & Pizzagalli, D. A. (2014). Peril and pleasure:
An RDoC-inspired examination of threat responses and
reward processing in anxiety and depression. Depression &
Anxiety, 31, 233–249.
Eklund, A., Nichols, T. E., & Knutsson, H. (2016). Cluster
failure: Why fMRI inferences for spatial extent have inflated
false-positive rates. Proceedings of the National Academy of
Sciences, U.S.A., 113, 7900–7905.
Engelmann, J. B., Meyer, F., Fehr, E., & Ruff, C. C. (2015).
Anticipatory anxiety disrupts neural valuation during risky
choice. Journal of Neuroscience, 35, 3085–3099.
psychological perspective. Nature Reviews Neuroscience,
14, 488–501.
Grupe, D. W., Wielgosz, J., Davidson, R. J., & Nitschke, J. B.
(2016). Neurobiological correlates of distinct post-traumatic
stress disorder symptom profiles during threat anticipation
in combat veterans. Psychological Medicine, 46, 1885–1895.
Hayes, D. J., & Northoff, G. (2012). Common brain activations for
painful and non-painful aversive stimuli. BMC Neuroscience,
13, 60.
Iglesias, J. E., Liu, C.-Y., Thompson, P. M., & Tu, Z. (2011).
Robust brain extraction across datasets and comparison with
publicly available methods. IEEE Transactions on Medical
Imaging, 30, 1617–1634.
Jensen, J., McIntosh, A. R., Crawley, A. P., Mikulis, D. J.,
Remington, G., & Kapur, S. (2003). Direct activation of
the ventral striatum in anticipation of aversive stimuli.
Neuron, 40, 1251–1257.
Kavaliers, M., & Choleris, E. (2001). Antipredator responses and
defensive behavior: Ecological and ethological approaches
for the neurosciences. Neuroscience & Biobehavioral
Reviews, 25, 577–586.
Lim, C. L., Rennie, C., Barry, R. J., Bahramali, H., Lazzaro, I.,
Manor, B., et al. (1997). Decomposing skin conductance into
tonic and phasic components. International Journal of
Psychophysiology, 25, 97–109.
Lindquist, K. A., Wager, T. D., Kober, H., Bliss-Moreau, E., &
Barrett, L. F. (2012). The brain basis of emotion: A meta-analytic
review. Behavioral and Brain Sciences, 35, 121–143.
McMenamin, B. W., Langeslag, S. J. E., Sirbu, M., Padmala, S.,
& Pessoa, L. (2014). Network organization unfolds over
time during periods of anxious anticipation. Journal of
Neuroscience, 34, 11261–11273.
McShane, B. B., Gal, D., Gelman, A., Robert, C., & Tackett,
J. L. (2017). Abandon statistical significance. arXiv preprint
arXiv:170907588.
Mobbs, D., Hagan, C. C., Dalgleish, T., Silston, B., & Prévost, C.
(2015). The ecology of human fear: Survival optimization
and the nervous system. Frontiers in Neuroscience, 9, 55.
Mobbs, D., Petrovic, P., Marchant, J. L., Hassabis, D., Weiskopf,
N., Seymour, B., et al. (2007). When fear is near: Threat
imminence elicits prefrontal-periaqueductal gray shifts in
humans. Science, 317, 1079–1083.
Mobbs, D., Trimmer, P. C., Blumstein, D. T., & Dayan, P.
(2018). Foraging for foundations in decision neuroscience:
Insights from ethology. Nature Reviews Neuroscience, 19,
419–427.
Mobbs, D., Yu, R., Rowe, J. B., Eich, H., FeldmanHall, O., &
Dalgleish, T. (2010). Neural activity associated with
monitoring the oscillating threat value of a tarantula.
Proceedings of the National Academy of Sciences, U.S.A.,
107, 20582–20586.
Fanselow, M. S., & Lester, L. S. (1988). A functional behavioristic
Mumford, J. A., Poline, J.-B., & Poldrack, R. A. (2015).
approach to aversively motivated behavior: Predatory
imminence as a determinant of the topography of defensive
behavior. In R. C. Bolles & M. D. Beecher (Eds.), Evolution
and learning (pp. 185–212). Hillsdale, NJ: Erlbaum.
Feinberg, D. A., Moeller, S., Smith, S. M., Auerbach, E., Ramanna,
S., Glasser, M. F., et al. (2010). Multiplexed echo planar
imaging for sub-second whole brain fMRI and fast diffusion
imaging. PLoS One, 5, e15710.
Fox, A. S., Oler, J. A., Tromp, D. P. M., Fudge, J. L., & Kalin,
N. H. (2015). Extending the amygdala in theories of threat
processing. Trends in Neurosciences, 38, 319–329.
Greve, D. N., & Fischl, B. (2009). Accurate and robust brain
image alignment using boundary-based registration.
Neuroimage, 48, 63–72.
Grupe, D. W., & Nitschke, J. B. (2013). Uncertainty and
anticipation in anxiety: An integrated neurobiological and
Orthogonalization of regressors in fMRI models. PLoS One,
10, e0126255.
Najafi, M., Kinnison, J., & Pessoa, L. (2017). Dynamics of
intersubject brain networks during anxious anticipation.
Frontiers in Human Neuroscience, 11, 552.
Neter, J., Kutner, M. H., Nachtsheim, C. J., & Wasserman, W.
(1996). Applied linear statistical models (4th ed.). Chicago:
Irwin.
Nitschke, J. B., Sarinopoulos, I., Mackiewicz, K. L., Schaefer,
H. S., & Davidson, R. J. (2006). Functional neuroanatomy of
aversion and its anticipation. Neuroimage, 29, 106–116.
Pessoa, L. (2016). The emotional brain. In P. M. Conn (Ed.),
Conn’s translational neuroscience (pp. 635–656).
Amsterdam: Elsevier.
Pruessner, J. C., Dedovic, K., Khalili-Mahani, N., Engert, V.,
Pruessner, M., Buss, C., et al. (2008). Deactivation of the
Meyer, Padmala, and Pessoa
541
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
d
i
r
e
c
t
.
m
i
t
.
e
d
u
/
j
/
o
c
n
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
o
c
n
_
a
_
0
1
3
6
3
p
d
.
f
b
y
g
u
e
s
t
t
o
n
0
8
S
e
p
e
m
b
e
r
2
0
2
3
limbic system during acute psychosocial stress: Evidence
from positron emission tomography and functional magnetic
resonance imaging studies. Biological Psychiatry, 63, 234–240.
Rousselet, G. A., & Pernet, C. R. (2012). Improving standards in
brain-behavior correlation analyses. Frontiers in Human
Neuroscience, 6, 119.
Satpute, A. B., Wager, T. D., Cohen-Adad, J., Bianciardi, M.,
Choi, J.-K., Buhle, J. T., et al. (2013). Identification of discrete
functional subregions of the human periaqueductal gray.
Proceedings of the National Academy of Sciences, U.S.A.,
110, 17101–17106.
Shackman, A. J., & Fox, A. S. (2016). Contributions of the
central extended amygdala to fear and anxiety. Journal of
Neuroscience, 36, 8050–8063.
Shattuck, D. W., & Leahy, R. M. (2002). BrainSuite: An automated
cortical surface identification tool. Medical Image Analysis,
6, 129–142.
Siegel, J. S., Power, J. D., Dubis, J. W., Vogel, A. C., Church, J. A.,
Schlaggar, B. L., et al. (2014). Statistical improvements in
functional magnetic resonance imaging analyses produced
by censoring high-motion data points. Human Brain
Mapping, 35, 1981–1996.
Simmons, A., Strigo, I., Matthews, S. C., Paulus, M. P., &
Stein, M. B. (2006). Anticipation of aversive visual stimuli is
associated with increased insula activation in anxiety-prone
subjects. Biological Psychiatry, 60, 402–409.
Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann,
C. F., Behrens, T. E. J., Johansen-Berg, H., et al. (2004).
Advances in functional and structural MR image analysis
and implementation as FSL. Neuroimage, 23(Suppl. 1),
S208–S219.
Somerville, L. H., Whalen, P. J., & Kelley, W. M. (2010).
Human bed nucleus of the stria terminalis indexes
hypervigilant threat monitoring. Biological Psychiatry, 68,
416–424.
Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970).
Manual for the State–Trait Anxiety Inventory. Palo Alto, CA:
Consulting Psychologists Press.
Theiss, J. D., Ridgewell, C., McHugo, M., Heckers, S., &
Blackford, J. U. (2017). Manual segmentation of the human
bed nucleus of the stria terminalis using 3 T MRI. Neuroimage,
146, 288–292.
Torrisi, S., O’Connell, K., Davis, A., Reynolds, R., Balderston, N.,
Fudge, J. L., et al. (2015). Resting state connectivity of the
bed nucleus of the stria terminalis at ultra-high field. Human
Brain Mapping, 36, 4076–4088.
Tovote, P., Esposito, M. S., Botta, P., Chaudun, F., Fadok, J. P.,
Markovic, M., et al. (2016). Midbrain circuits for defensive
behaviour. Nature, 534, 206–212.
Vytal, K. E., Overstreet, C., Charney, D. R., Robinson, O. J., &
Grillon, C. (2014). Sustained anxiety increases amygdala–
dorsomedial prefrontal coupling: A mechanism for
maintaining an anxious state in healthy adults. Journal of
Psychiatry & Neuroscience, 39, 321–329.
Wager, T. D., Waugh, C. E., Lindquist, M., Noll, D. C.,
Fredrickson, B. L., & Taylor, S. F. (2009). Brain mediators
of cardiovascular responses to social threat: Part I.
Reciprocal dorsal and ventral sub-regions of the medial
prefrontal cortex and heart-rate reactivity. Neuroimage,
47, 821–835.
Wilcox, R. R. (2012). Introduction to robust estimation and
hypothesis testing (3rd ed.). Cambridge, MA: Academic
Press.
Woo, C.-W., Krishnan, A., & Wager, T. D. (2014). Cluster-
extent based thresholding in fMRI analyses: Pitfalls and
recommendations. Neuroimage, 91, 412–419.
l
D
o
w
n
o
a
d
e
d
f
r
o
m
h
t
t
p
:
/
/
d
i
r
e
c
t
.
m
i
t
.
e
d
u
/
j
/
o
c
n
a
r
t
i
c
e
-
p
d
l
f
/
/
/
/
3
1
4
5
2
2
1
7
8
8
3
3
4
/
j
o
c
n
_
a
_
0
1
3
6
3
p
d
.
f
b
y
g
u
e
s
t
t
o
n
0
8
S
e
p
e
m
b
e
r
2
0
2
3
542
Journal of Cognitive Neuroscience
Volume 31, Number 4
