Supporting Information
Mobbs et al. 10.1073/pnas.1009076107
SI Experimental Procedures
Participants.
Twenty-
fi
ve healthy nonphobic volunteers took part in
the study. Subjects were rejected if they had any history of neu-
rological damage or psychiatric disorder. After the MRI scan,
fi
ve
participants were excluded as they expressed reservations about
whetherthestudywasgenuinelyinrealtimeduringdebrie
fi
ng.This
left 20 participants (10 female; mean age, 25.8
±
3.7 y). Trait
anxiety was measured using the Spielberger questionnaire, with
a mean score of 40
±
9.9. These scores are comparable to the
published norms for this age group mean (36
±
10) (1). Subjects
were remunerated £30 for time, travel, and inconvenience. All
subjects gave informed consent, and the study was approved by
the Essex Research Ethics Committee (United Kingdom).
FSQ Measures.
During recruitment, subjects were asked if they had
any signi
fi
cant fear of spiders, including tarantulas, and given
details on the experiment. Those who expressed reservations were
not followed up and not included in the study. We also collected
FSQ scores on each subject (2). These mean FSQ scores were in
medium to low ranges (mean
±
SD, 30.7
±
18.4; range, 14
–
86.1).
Overall, this mean score is signi
fi
cantly lower than that of di-
agnosed arachnophobic subjects (89.1
±
19.6) (3).
Spider Stimuli Creation and Validation.
We recorded footage of
a Brazilian salmon pink tarantula (
L.parahybana
; body size, 22 cm
length
×
15 cm width;
Fig. S1
A
) being placed in each compartment of
the imminence box. Video clips were edited to 4 s and a border was
added to the footage to decontextu
alize the environment. Sixty
black-and-white 4-s
fi
lm clips of the tarantula were then presented in
the study (12 in each compartment of the imminence box). Each
fi
lm
was prerated for movement of the tarantula on a four-point scale: 1,
no movement; 2, very little movement (
<
25% of the time); 3,
movement 25% to 50% of the time; and 4, movement 50% to 100%
of the time. Ratings were made by 10 independent observers. No
signi
fi
cant differences were found between compartments (
P
>
0.05).
There was also no signi
fi
cant relationship between participants
’
experienced fear ratings and movement (
r
= 211;
P
= 0.105). In
addition, we tested to see if the direction the spider was facing
(toward vs. away from the foot) in
fl
uenced fear ratings. We
found no signi
fi
cant correlation between the tarantula
’
s angle of
orientation to the foot, from 0° to 180°, and fear ratings (
P
>
0.05). Finally, mindful of likely habituation effects, we pseu-
dorandomized the presentation of the tarantula in each box so as
to decorrelate distance from time (Pearson correlation,
r
= 0.09;
P
<
0.495).
Experimental Setup and Paradigm.
Immediately before the scan-
ning session, participants were shown a tarantula housed in
a distant glass tank, but were not told whether it was real (in fact, it
was a robotic model). This robot tarantula was viewed in a box
containing soil, bark, and plants, and emitted subtle lifelike
movements. After this, subjects were placed supine on the MRI
scanner bed (
Fig. S1
B
) and placed their foot, with the shoe re-
moved, into the custom-built open-topped imminence box, which
contained six compartments separated by sliding partitions (
Fig.
S1
C
). A curtain was placed at the end of the scanner bore so
subjects could not see their foot or the box.
Although they were not explicitly told so, participants believed
they could view the imminence box compartments in real time via
a live camera feed from above this open-topped box. Via this
video feed, participants watched the tarantula placed pseudor-
andomly in the compartments at distances of approximately 1 to
18 cm, 18
–
to 36 cm, 36 to 54 cm, 54 to 72 cm, and 72 to 90 cm from
their foot (
Fig. S1
C
). Although participants believed contact with
the spider was possible, the compartment containing the partic-
ipant
’
s foot was not used. The top of the compartment was open,
and there was potential for contact (e.g., there was no physical
barrier between the top of the subject
’
s foot and the tarantula).
This possibility was drawn to the subjects
’
attention before scan-
ning. Subjects were
fi
rst showed the imminence box and told that
their foot would be placed in the tall chamber. The experimenter
then stated:
“
The barrier [between the foot chamber and
fi
rst
spider chamber] only covers half of the chamber where your foot
will be, so the top of your foot will be exposed to box 5. However,
the experimenter will be in the room at all times, and it is unlikely
that the spider will have any direct contact with you.
”
In reality, participants were viewing prerecorded
fi
lms of the
spider in the different compartments as described earlier. To
complete the illusion that activity involving the real and unseen
imminence box was occurring in real time, the experimenter
remained in the scanner room and moved the robot tarantula from
compartment to compartment in synchrony with the video footage
of the real Tarantula that the participants were actually viewing.
Postscan debrie
fi
ng revealed that participants believed that they
were viewing a genuine real-time video feed of a proximate live
tarantula.
Each experimental trial commenced with viewing a schematic
representation of the imminence box indicating which compart-
ment the tarantula would enter next. Participants then used a VAS
to predict how afraid they would feel when subsequently viewing
thetarantulaintherelevantcompartment(i.e.,expectedfear).The
participants next saw the Tarantula in the compartment and di-
rectly rated how afraid they felt (i.e. experienced fear;
Fig. S2
).
Image Acquisition.
MRI scanning was conducted at the Medical
Research Council Cognition and Brain Sciences Unit on a 3-T Tim
Trio MRI scanner (Siemens) by using a head coil gradient set.
Whole-brain data were acquired with echoplanar T2*-weighted
imaging (i.e., EPI), sensitive to BOLD signal contrast (48 sagittal
slices, 3-mm thickness; repetition time, 2,400 ms; echo time, 30
ms;
fl
ip angle, 78°;
fi
eld of view, 192 mm; voxel size, 3
×
3
×
3 mm).
To provide for equilibration effects, the
fi
rst
fi
ve volumes were
discarded. T1-weighted structural images were acquired at a res-
olution of 1
×
1
×
1 mm.
Image Preprocessing.
SPM5 software (
www.
fi
l.ion.ucl.ac.uk/spm/
)
was used for data analysis. The EPI images were sinc interpolated
in time for correction of slice timing differences and realignment
to the
fi
rst scan by rigid body transformations to correct for head
movements. Field maps were estimated from the phase difference
between the images acquired at the short and long TE and un-
wrapped by using the FieldMap toolbox. Field map and EPI im-
aging parameters were used to establish voxel displacements in
the EPI image. Application of the inverse displacement to the EPI
images served the correction of distortions. For each participant
the mean EPI was calculated and examined to guarantee that
none exhibited excessive signal dropout in insula and ventral
striatum. Using linear and nonlinear transformations, and
smoothing with a Gaussian kernel of full-width half-maximum 8
mm, EPI and structural images were coregistered and normalized
to the T1 standard template in MNI space (International Con-
sortium for Brain Mapping). Moreover, global changes were re-
moved by proportional scaling, and high-pass temporal
fi
ltering
Mobbs et al.
www.pnas.org/cgi/content/short/1009076107
1of5
with a cutoff of 128 s was used to remove low-frequency drifts in
signal.
Statistical Analysis.
After preprocessing, statistical analysis was
performed using the general linear model. Our regression matrix
included the cue periods (1 s), both expectancy (8 s) and outcome
VAS (8s)time periods, and a4-speriod when the spider was shown
in the relevant box. Analysis was carried out to establish each
participant
’
s voxel-wise activation during the 4-s presentation of
the spider. Activated voxels in each experimental context were
identi
fi
ed using an event-related statistical model representing
each of the experimental contexts, convolved with a canonical
hemodynamic response function and mean-corrected.
Allsubjectshad headmotion oflessthanonevoxel(3mm) across
the whole session. To correct for possible confounding effects of
head motion under different conditions, we included the head
motion parameters as regressors of no interest in our subject-
speci
fi
c (i.e.
fi
rst level) models. Multiple linear regression was then
run to generate parameter estimates for each regressor at every
voxel. For group analysis, a random-effects model was used with
a small volume correction for FWE within a priori areas of interest,
including the amygdala (left,
−
18,
−
4,
−
30; right, 10,
−
6,
−
18),
BNST (left,
−
9, 0,
−
12; right, 12,
−
1,
−
12), medial orbital frontal
cortex (
−
3, 48,
−
18), and midbrain PAG (8,
–
32,
–
21). These co-
ordinates were taken from previous studies (4
–
7), and when
needed, converted into MNI space. A spherical ROI 6 to 12 mm in
diameter was centered on these coordinates, and statistical in-
ferences were corrected for multiple comparisons within this ROI
(FWE,
P
<
0.05). This produces statistical inferences that are
FWE-corrected centered on the aforementioned independent
ROI coordinates. Outside of these areas of interest, we also
present results at
P
<
0.05 FWE-corrected for whole-brain multi-
ple spatial comparisons. When false-negative results at these cor-
rected thresholds would be of particular relevance, we also provide
results at the exploratory uncorrected threshold of
P
<
0.001 un-
corrected. For areas outside our a priori ROIs, we use FWE cor-
rection for the whole brain to give strong control of type I error.
There are situations in which exploration at more liberal thresh-
olds is likely to be of interest to the reader, especially if type II error
is relevant. We therefore do present selected results at this liberal
uncorrected threshold. Our principal conclusions, and tests of our
hypotheses of interest, do not rely on these uncorrected tests.
Questionnaires.
Following the MRI scan, participants were asked
to complete the FSQ and to rate how much scarier than expected
the spider was. (For example, they were asked,
“
Was the spider as
scary as you thought?
”
) We also asked subjects to complete
a memory for tarantula size test to rate how large they thought
the tarantula was by using
fi
ve different sizes of the spider
printed on an A3 sheet of paper (
Fig. S4
). This was administered
between 45 and 60 min after the experiment.
1. Spielberger CD (1983)
Manual for the State Trait Anxiety Inventor
(Consulting
Psychologists, Palo Alto, CA).
2. Szymanski J, O
’
Donohue W (1995) The fear of spiders questionnaire.
J Behav Ther Exp
Psychiatry
26:31
–
34.
3. Muris P, Merckelbach H (1996) A comparison of two spider fear questionnaires.
J Behav
Ther Exp Psychiatry
27:241
–
244.
4. Somerville LH, Whalen PJ, Kelley WM (2010) Human bed nucleus of the stria terminalis
indexes hypervigilant threat monitoring.
Biol Psychiatry
68:416
–
424.
5. Mobbs D, et al. (2007) When fear is near: threat imminence elicits prefrontal-
periaqueductal gray shifts in humans.
Science
317:1079
–
1083.
6. Hermann A, et al. (2009) Emotion regulation in spider phobia: Role of the medial
prefrontal cortex.
Soc Cogn Affect Neurosci
4:257
–
267.
7. Mobbs D, et al. (2009) From threat to fear: The neural organization of defensive fear
systems in the human brain.
J Neurosci
29:12236
–
12243.
Fig. S1.
(
A
) Photograph of the Brazilian pink salmon tarantula (
L. parahybana
) used in the video footage. (
B
) Picture of the experimental setup showing
a subject supine in the MRI bore and with his/her foot in the (
C
) custom built imminence box.
Fig. S2.
A schematic representation of the experiment order and timings.
Mobbs et al.
www.pnas.org/cgi/content/short/1009076107
2of5
Fig. S3.
Changes in fear and expectancy (i.e., underestimation) errors and fear of spiders over time. (
A
) Expectancy errors over time. Red dots signal mean
value. The correlations between (
B
) increasing underestimation errors and sustained amygdala activation (
−
10,
−
6,
−
18;
P
<
0.039, svc), and (
C
) FSQ scores and
sustained midbrain PAG activation (8,
−
30,
−
4;
P
<
0.038, svc) over the course of the experiment.
Fig. S4.
The memory for tarantula size test (not to scale).
Table S1. Comparison between compartments 5/4 and 1/2 [(compartment 5 + compartment 4)
−
(compartment 1 + compartment 2)]
Region
MNI coordinates, x/y/z
z value
P
value
Compartment (5 + 4)
–
(1 + 2)
Midbrain
4/
−
30/
−
24
3.10
0.028*
Dorsal ACC
−
4/28/36
4.13
<
0.0005
†
Left amygdala
−
20/2/
−
16
2.74
0.030*
Right amygdala
16/8/
−
14
3.49
0.006*
Compartment (1 + 2)
–
(5 + 4)
omPFC
2/52/
−
6
2.94
0.032*
*
P
value small volume corrected using independent coordinates.
†
P
value whole-brain corrected.
Table S2. Parametric modulation weighted by tarantula compartment
Region
MNI coordinates, x/y/z
z value
P
value
Closer to the foot
Midbrain
4/
−
30/
−
24
2.86
0.028*
Dorsal ACC
28/36/24
4.13
<
0.0005
†
dlPFC
30/48/30
4.25
<
0.0005
†
Right insula
44/14/
−
4
3.47
<
0.0005
‡
Left insula
−
44/2/10
3.52
<
0.0005
‡
Visual cortex (V1)
2/
−
66/0
4.83
<
0.0005
†
Intraparietal cortex
12/
−
68/64
4.38
<
0.0005
‡
BNST
12/0/
−
4
3.14
<
0.014*
Further from the foot
omPFC
−
6/54/
−
16
3.09
0.016*
PCC
−
10/
−
54/20
3.29
<
0.001
‡
Visual cortex (V1)
−
2/
−
94/12
4.91
<
0.0005
‡
PCC, posterior cingulate cortex.
*
P
value small volume corrected using independent coordinates.
†
P
value whole-brain corrected.
‡
P
value uncorrected.
Mobbs et al.
www.pnas.org/cgi/content/short/1009076107
3of5
Table S3. Parametric modulation by tarantula distance correlated with FSQ scores
Region
MNI coordinates, x/y/z
z value
P
value
Increased FSQ scores
Midbrain
14/
−
26/
−
4
3.30
<
0.009*
Right insula
54/
−
4/8
3.53
<
0.0005
†
Left insula
−
46/
−
4/4
3.16
<
0.001
‡
Decreased FSQ scores
Perigenual ACC
20/44/6
2.99
<
0.036*
*
P
value small volume corrected using independent coordinates.
†
P
value whole-brain corrected.
‡
P
value uncorrected.
Table S4. Parametric modulation by postexperimental ratings of how much scarier than
expected the subjects found the spider
Region with Increased fear of
approach scores
MNI coordinates, x/y/z
z value
P
value*
Midbrain
10/
−
24/
−
6
2.94
<
0.023
Amygdala
36/
−
2/
−
22
2.69
<
0.042
*Small volume corrected using independent coordinates.
Table S5. Approach minus retreat independent of distance
Region
MNI coordinates, x/y/z
z value
P
value
Amygdala
14,/
−
2/
−
16
2.90
<
0.024*
Right BNST
12/0/
−
4
2.85
<
0.027*
Left BNST
−
12/4/
−
4
3.10
<
0.014*
Left insula
−
40/16/0
2.94
0.037*
Right insula
40/20/
−
2
3.68
0.002*
Dorsal ACC
6/34/22
3.37
<
0.0005
†
Striatum
12/4/4
3.64
<
0.003*
All regions were still signi
fi
cant after covarying out FSQ scores.
*
P
value small volume corrected using independent coordinates.
†
P
value uncorrected.
Table S6. Parametric modulation by tarantula distance correlated with expectancy error (fear
underestimation) scores
Region
MNI coordinates, x/y/z
z value
P
value*
Left amygdala
−
20/0/
−
18
—
<
0.043
Right insula
−
30/26/2
3.39
0.013
Underestimate (covarying out FSQ scores)
Left amygdala
−
18/2/
−
22
2.80
<
0.034
Underestimate (covarying out mean fear scores)
Left amygdala
−
26/
−
6/18
3.88
<
0.004
*Small volume corrected using independent coordinates.
Table S7. Habituation for four time bins (T1
–
T4) over the course of the experiment independent
of distance
Region
MNI coordinates, x/y/z
z value
P
value
Decreased activity
Midbrain PAG
12/
−
32/
−
14
2.81
<
0.029*
Hypothalamus
−
4/2/
−
12
3.46
<
0.005*
sgACC
8/26/
−
12
2.84
<
0.046*
Amygdala
−
24/
−
2/
−
14
2.77
<
0.030*
Insula
52/26/0
3.22
<
0.001
†
Increased activity
MFD
−
22/46/0
3.07
<
0.001
†
*
P
value small volume corrected using independent coordinates.
†
P
value uncorrected.
Mobbs et al.
www.pnas.org/cgi/content/short/1009076107
4of5
Table S8. Habituation for four time bins over the course of the experiment independent of
distance examining the correlation with expectancy errors.
Region increased with errors
MNI coordinates, x/y/z
z value
P
value
Left amygdala
−
10/
−
/6/
−
20
3.29
<
0.001*
Right amygdala
34/
−
2/
−
18
2.73
<
0.046
†
MFD
16/56/6
3.52
<
0.0005*
Mediodorsal thalamus
0/6/
−
2
3.46
<
0.0005*
No signi
fi
cant voxels
———
*
P
value uncorrected.
†
P
value small volume corrected using independent coordinates.
Table S9. Habituation for four time bins over the course of the experiment independent of
distance examining the correlation with FSQ scores
Region
MNI coordinates, x/y/z
z value
P
value
Increased with FSQ
Midbrain PAG
8/
−
30/
−
4
2.71
<
0.038*
Parahippocampal gyrus
−
38/
−
20/
−
20
3.24
<
0.001
†
Ventral motor cortex
48/16/22
3.18
<
0.001
†
Decreased with FSQ
Dorsal caudate
−
14/2/24
3.31
<
0.0005
†
dlPFC
−
36/50/18
3.05
<
0.001
†
*
P
value small volume corrected using independent coordinates.
†
P
value uncorrected.
Mobbs et al.
www.pnas.org/cgi/content/short/1009076107
5of5