Graduate School of Psychology
Pharmacological manipulations of fear memory and
fMRI: not thé golden couple (yet)
Internship report; Final Version
Name: Bianca Westhoff
Student Number: 10000662
Email address: BiancaWesthoff@live.nl Specialisation: Brain & Cognition Daily Supervisor: R.M. Visser, MSc ResMas Supervisor: dr. H.S. Scholte
Research location: Spinoza Centre for Neuroimaging
Date: 8 October 2014
Page 1 of 48
Pharmacological manipulations of fear memory and
fMRI: not thé golden couple (yet)
ABSTRACT
Only a fraction of the people who experience a traumatic event develops posttraumatic stress disorder (PTSD). Studies suggest that people who are more vulnerable to develop PTSD, have increased noradrenergic activity during the traumatic event which leads to strengthened fear memories. We aimed to investigate whether there are markers at the time of fear learning that indicate the formation of strengthened fear memories.
Healthy participants received either Yohimbine HCl (n = 21) to increase the noradrenergic activity, or a Placebo (n = 20). Participants were subjected to a differential fear conditioning paradigm in the MRI scanner that assessed fear acquisition (Session 1) and fear extinction, reinstatement and generalization (Session 2, 48 hours later). Simultaneously to functional scanning, behavioural fear expressions (pupil dilation responses) were collected. We applied Multivoxel Pattern Analysis (MVPA) to the fMRI data in a trial-by-trial manner to quantify the development of fear during the different stages of the experiment.
Results show that there was no effect of Yohimbine HCl on neural and behavioural measures of fear learning and memory, which is inconsistent with lab findings. The increased noradrenergic activity just before scanning in the Placebo group suggests that the procedure in itself induced stress, causing a ceiling effect. Yet there was strong fear learning and memory indicating that neural pattern similarity still seems a promising marker for the study of enhanced fear. Thus, studying pharmacological manipulations of fear with fMRI holds perspective, but is subject to constraints.
1. INTRODUCTION
Fear is a strong, uncontrollable, aversive emotion caused by actual or perceived danger or threat
(Gluck, Mercado, & Myers, 2007). When an organism experiences fear, its defensive fear system is activated to prepare to adjust to the threat, which makes fear an important emotional state to
survive. However, in cases of long-term unremitting fear, the associated stress can lead to
Page 2 of 48
cause psychological problems such as anxiety disorders - including phobias and posttraumatic stress
disorder (PTSD) - which may specifically involve a dysfunction of the normal fear response. As a result
of this dysfunctional fear response, PTSD patients persistently re-experience the traumatic event
because their ability to unlearn fear-related behaviour is impaired and they generalize the fear to
intrinsically safe stimuli (American Psychiatric Association, 2013). The fact that only a fraction of the people that experience a traumatic event will eventually develop PTSD, indicates that some people
are more vulnerable to develop PTSD than others (Gluck et al., 2007).It is suggested people are more vulnerable when they form strengthened fear memories (Vanelzakker, Kathryn Dahlgren, Caroline Davis, Dubois, & Shin, 2013). Considering the high impact consequences of PTSD for patients’ daily life but also for society (e.g., the enormous societal costs) (Brunello et al., 2001), it would be of great value to have a marker that indicates which people form strengthened fear memories and are
therefore more prone to develop PTSD. Such markers could, for example, be used in a screening of
war veterans before they go on a mission, as they are commonly at high risk for PTSD (Stein, Ipser, & Seedat, 2009).
In experimental settings fear is typically studied in a classical fear conditioning paradigm
(Pavlov, 1927). In this paradigm an intrinsically aversive stimulus (unconditioned stimulus (US); e.g., an electric shock) is repeatedly preceded by an initially neutral stimulus (conditioned stimulus (CS⁺);
e.g., a picture of a house), while another stimulus (CS⁻; e.g., a picture of another house) never
precedes the US. After sufficient US-CS⁺ pairing, the CS⁺ acquires the same aversive qualities as the
US and accordingly the CS⁺ will elicit a conditioned fear response even when the US is absent. If the
CS⁺ is then repeatedly presented in the absence of the US, this conditioned response to the CS⁺ will
reduce or even erase (i.e., fear extinction). Fear can be reinstated when the US is presented again,
indicating that fear is recovered (i.e., fear reinstatement). When stimuli have overlapping features
with the CS⁺ and evoke a fear response, even though they have never been paired with the US, this
designates fear generalization (Lissek et al., 2008). If fear is generalized, the initial fear association is allegedly stronger compared to when fear is not generalized (Dunsmoor, Prince, Murty, Kragel, &
Page 3 of 48
LaBar, 2011). Experimental studies investigating fear often assess the behavioural expression of fear using measures such as the skin conductance response (SCR) (Tabbert, Stark, Kirsch, & Vaitl, 2006), the eye blink startle reflex (Asli & Flaten, 2008) or the pupil dilation response (Reinhard, Lachnit, & . An increasing number of studies investigating fear also use functional magnetic König, 2006)
resonance imaging (fMRI), either to examine the average activation level (i.e., involvement) of a brain
region, or to examine patterns of neural activation (i.e., representational content) in a brain region
using MultiVoxel Pattern Analysis (MVPA) (Haxby et al., 2001; Kriegeskorte & Bandettini, 2007; Kriegeskorte, Mur, & Bandettini, 2008). This multivariate technique evaluates blood-oxygen-level dependent (BOLD) response of multiple voxels at the same time, without averaging over these
voxels. MVPA allows trial-by-trial analysis, which may reveal changes in patterns over time, e.g.,
learning dependent changes of neural representations and categorization of stimuli (Li, Mayhew, & Kourtzi, 2009; Visser, Scholte, & Kindt, 2011; Xue et al., 2010; Zhang, Meeson, Welchman, & Kourtzi, 2010). These changes in patterns may elucidate how relevant information is encoded in the brain and how semantic networks are formed, which may be relevant for examining the difference between
normal and dysfunctional (i.e., in anxiety disorders) fear learning. The relevance of the neural
patterns is shown by a recent study using MVPA (Visser, Scholte, Beemsterboer, & Kindt, 2013): it revealed that the neural activation patterns – but not the behavioural fear expression – at the time
of fear learning are a unique marker for the formation of long-term fear memory, as those patterns
could predict whether someone will show a long-term behavioural expression of the fear memory, or
not.
Several animal and human studies have shown that for fear learning and fear memory
formationthe neurotransmitter noradrenaline is of great importance (e.g., Dębiec, Bush, & LeDoux, 2011; Gazarini, Stern, Carobrez, & Bertoglio, 2013; Soeter & Kindt, 2011). For example, knock-out mice that were unable to synthesize noradrenaline showed impaired fear learning during a cued fear
conditioning paradigm (Toth, Ziegler, Sun, Gresack, & Risbrough, 2013). Therefore, it is suggested that the strengthened memories in PTSD patients are a result of increased noradrenergic activity
Page 4 of 48
during the traumatic event (Southwick et al., 1999). Experimental studies often induce an increase of noradrenergic activity with Yohimbine hydrochloride (HCl) (e.g., Gazarini et al., 2013; Soeter & Kindt, 2011), an antagonist of the pre-synaptic α2-adrenergic autoreceptor. This is a G protein-coupled
receptor that inhibits the release of noradrenaline through negative feedback. Yohimbine HCl blocks
this negative feedback which thereby results in an increased release of noradrenaline.
In the current study we aimed to investigate whether increased noradrenergic activity during
fear learning indeed results in strengthened fear memory, and whether there are markers at the time
of fear learning that indicate the formation of this strengthened fear memory. To investigate this,
healthy participants received either 20 mg Yohimbine HCl, to increase noradrenergic activity during
fear learning, or a Placebo. Participants were subjected to a differential fear conditioning paradigm in
the MRI scanner. Fear acquisition was assessed in Session 1 and fear extinction, reinstatement and
generalization in a second session 48 hours later. We used the pupil dilation response to assess the
behavioural expression of fear and MVPA was applied to the fMRI data to assess the development of
fear during the different stages of the experiment.
We tested whether Yohimbine HCl effects during learning indeed strengthens long-term fear
memory as would be indicated by an increased differential (CS⁺ versus CS⁻) long-term behavioural
fear expression (i.e., increased pupil dilation response during Session 2), denoting impaired
extinction of the fear memory. We hypothesized that, in line with previous findings (Soeter & Kindt, 2011, 2012), the increased noradrenergic activity during acquisition would not affect the fear expression at that time. As there is more information about future fear memory in neural response
patterns than in the expression of fear at the time of learning (Visser et al., 2013), the question is whether the noradrenergic enhancement of fear can also be observed in neural response patterns at
Page 5 of 48
2. METHODS
2.1 Participants
Fifty-two students participated in the first session of the experiment. Data from 11 participants were
discarded from analysis because they did not return for the second session (n = 3), fMRI data was
confounded by substantial head motion (>2 mm in any direction, n = 2) or because of excessive
sleepiness, as judged on the basis of eye-tracker data combined with self-report (n = 6). The final
sample included 41 participants (11 male, 32 right-handed) between 18 and 25 years of age (20.5 ±
1.9, mean ± SD age). Participants were randomly assigned to the Placebo (n = 20) or Yohimbine (n =
21) condition. Gender was equally distributed across the groups (χ2(1) = 0.20. p = 0.73, φ = 0.70).
All participants were assessed to be free from any current or previous medical or psychiatric
condition that would contraindicate taking a single 20 mg oral dose of Yohimbine hydrochloride (i.e.,
pregnancy, seizure disorder, respiratory disorder, cardiovascular disease, blood pressure ≥ 140/90.
diabetes, liver-/kidney disorder, use of medication that could involve potentially dangerous
interactions with Yohimbine HCl, depression, mania or psychosis). Participants had normal or
corrected-to-normal vision and were naïve to the purpose of the experiment.
Participants received either partial course credits or a small amount of money for their
participation. Witten informed consent was obtained from all participants before the start of the
experiment. Procedures were executed according relevant laws and institutional guidelines, and
were approved by the ethics committee of the University of Amsterdam.
2.2 Experimental design
The experimental design (Figure 1), procedure and statistical analyses are adopted from and similar to previous studies (Visser et al., 2013, 2011) and are explained in detail in the next sections.
Page 6 of 48
2.2.1 Experimental task
In the two sessions, participants were subjected to a differential fear conditioning paradigm during
functional scanning. The four stimuli that were used during the task were greyscale pictures of two
neutral faces and two houses. The faces were selected from the Karolinska Directed Emotional Faces
dataset. The pictures were presented one by one in the middle of the screen, on a grey background
to minimize afterimages. In the centre of each picture a white fixation cross indicated where the
participants had to fixate as long as the stimulus was presented. This prevented varying image
representations in retinotopically organized areas in the visual cortex. All stimuli were presented for
4500 ms (Figure 2). In between the stimuli, a grey screen was shown, followed by a fixation cross which enabled participants to focus in time for the next stimulus. These inter-trial intervals (ITI) were
fixed (19.5 s) and long enough to reduce intrinsic noise correlations. The onset of each trial was
triggered by the start of the acquisition of a BOLD-MRI volume.
Figure 1. Experimental design. The experiment consisted of an acquisition phase (Session 1) and a memory phase (Session 2, 48 hours later). For each time point in the experiment, the dependent measures are indicated. Pill intake (either Placebo or 20 mg Yohimbine HCl) took place at T1. During functional scanning in both sessions, the pupil dilation responses were measured.
Abbreviations: sAA = salivary alpha amylase, BP = blood pressure, HR = heart rate, ASI = Anxiety Sensitivity Index, STAI = State and Trait Anxiety Inventory, SAM = Self-Assessment Manikin, Exit = Exit questionnaire.
Page 7 of 48
In the acquisition phase (Session 1) these four stimuli served as the to-be-conditioned stimuli in a
differential fear conditioning paradigm. One picture (CS⁺) of each category was followed by an US in
46% of the trials (CS⁺), while the other face and house were never followed by an US (CS⁻), creating
the four stimulus categories CS⁺ Face, CS⁻ Face, CS⁺ House, and CS⁻ House. Assignment of the stimuli
as CS⁺ and CS⁻ was counterbalanced across participants.
The unconditioned stimulus was a mild electric shock (duration: 2 ms), delivered 4300 ms
after onset of the CS⁺. The electrical stimulation was applied by a Digitimer DS7A through
MRI-compatible carbon electrodes attached to the right shin bone. Prior to the start of the task the shock
intensity was determined by gradually increasing shock intensity (starting at 1 mA) until participants
indicated the shock to be ‘uncomfortable, but not painful’, with a maximum of 60 mA.
In the memory phase (Session 2), the same stimuli were presented but were not followed by
the US. This phase functioned as the extinction phase and was used to assess whether an (increased)
fear memory had been formed during Session 1. After the extinction phase, a single unsignalled US
was applied to reinstate fear for the CS⁺. After this unexpected shock, the four stimuli were
presented once to test for fear reinstatement. Thereafter, the stimuli were presented in colour to
assess fear generalization.
In both sessions, a fixed sequence of the four stimuli (counterbalanced over participants) was
repeatedly presented. These target trials were interspersed with filler trials containing the same
stimuli but were – in contrast to the target trials – not used for analyses. The interval between two
consecutive target trials (e.g., trial 3 and 4) of one stimulus category (e.g., CS⁺ Face) was the same for
all four stimuli. It was necessary to include filler trials for three reasons. Firstly, the interval between
two consecutive presentations of a stimulus had to be equal for all stimulus categories in order to
prevent that temporal proximity or other factors (not related to the experimental manipulation)
would influence the trial-to-trial correlation strengths. Secondly, the filler trials assured that the
participants perceived the order of stimuli as random. Thirdly, in the reinforced trials in Session 1 the
Page 8 of 48
not be analysed. The acquisition phase (Session 1) consisted of 28 target trials (7 per stimulus
category) and 24 filler trials (6 per stimulus category). The memory phase (Session 2) consisted of 36
target trials (9 per stimulus category) and 24 filler trials (6 per stimulus category). Note that the
memory phase consists of 2 more target trials of each stimulus category because this phase also
contains the trials that assessed reinstatement and generalization of fear.
Figure 2. Stimuli presentation. During functional scanning, a differential fear conditioning paradigm was used. Four stimuli were repeatedly presented for 4500 ms. In the acquisition phase, one face and one house were followed by the US (electrical shock) in 46% of the trials. The US was presented 4300 ms after stimulus onset. Inter-trial intervals had a fixed duration of 19500 ms.
2.2.2 Pupil dilation response
The conditioned fear response was measured as the pupil dilation response to the conditioned
stimuli. Pupil dilation responses and eye movements were recorded continuously throughout MRI
scanning using a remote non-ferromagnetic infrared Eyelink-1000 Long Range Mount eye tracker (SR
Research). Before task onset, a nine-point calibration procedure was performed. Data 100 ms before
and after an eye blink were removed. If data samples were obscured by one or more eye blinks, they
Page 9 of 48
A baseline pupil diameter was calculated for each trial as the average diameter during the
500 ms before each conditioned stimulus onset. The pupil dilation response in each trial was
calculated as the maximal peak change from that baseline in the 4000 ms after stimulus onset. If
substantial signal loss affected either the baseline or more than 50% of the 4000 ms time window
after stimulus onset, the pupil dilation response in that trial could not reliably be calculated. Those
trials were eliminated (M = 1.27% per participant) and values were replaced using the linear trend at
point. Three participants were excluded from the analysis because dilation responses were not
recorded due to technical failures.
For each of the four stimulus categories separately, the maximal peak change in the first two
trials (i.e., when no US was presented yet) was averaged. This was set as baseline peak change to
that conditioned stimulus. The percentage change from this baseline was calculated for each trial,
expressing the change in fear response to that stimulus over time in the acquisition phase. The
percentages were averaged over faces and houses, resulting in a percentage pupil dilation response
change to CS⁺ and to CS⁻ for each trial.
2.2.3 Retrospective US-expectancy ratings
To assess declarative knowledge of the fear association, after both sessions the US-expectancy was
measured retrospectively. In the first session participants had to indicate to what extent they
expected the shock at the beginning (i.e., before the first shock), halfway, and at the end of the
experiment. In the second session they had to rate US-expectancy at the beginning, just before the
unexpected shock, right after the unexpected shock (reinstatement) and when the stimuli were
presented in colour (generalization). This was done on a 9-point scale ranging from “certainly no
Page 10 of 48
2.2.4 Pharmacological treatment & manipulation check
To investigate the effect of increased noradrenergic activity on fear memory formation, participants
received double-blind an oral dose of either Placebo or 20 mg Yohimbine HCl prior to fear learning.
Yohimbine is α2-adrenergic receptor antagonist which stimulates the central noradrenergic activity
by blocking the α2-adrenergic autoreceptor (Charney, Woods, Goodman, & Heninger, 1987; Peskind
et al., 1995). Administration was planned so that peak plasma levels (60 minutes after administration) are reached halfway through the acquisition phase.
To check whether Yohimbine HCl exerted its intended physiological effect, salivary alpha
amylase (sAA), blood pressure levels and heart rate were determined at several time points during
the experiment. sAA is a reliable indicator of noradrenergic activation (Noto, Sato, Kudo, Kurata, & Hirota, 2005), and the levels were assessed out of unstimulated saliva samples obtained using regular cotton Salivette sampling devices without chemical stimulants. Participants were instructed to place
the swab in their mouths until it absorbed much saliva and the swab was soaked. After removal, the
Salivettes were stored at -25°. To facilitate salivary sampling, participants were instructed to refrain
from exercise, caffeine and alcohol during the 12 hours before each session. Also, they were
instructed to abstain from brushing their teeth for 1 hour and to avoid food intake, drinking any
beverages except for water, and smoking for 2 hours before each session. Upon completion of the
study, the samples were sent to Dresden, Germany for biochemical analysis (Technische Universität
Dresden). In addition to sAA, also cortisol levels were determined from the saliva samples. Cortisol is
a steroid hormone produced by the adrenal cortex and is released in response to stress (Hellhammer, Wüst, & Kudielka, 2009). Cortisol levels were determined to check whether stress levels changed during the first session. Blood pressure and heart rate were measured using an electronic
sphygmomanometer (OMRON M2, Healthcare Europe BV, Hoofddorp, The Netherlands) with a cuff
Page 11 of 48
2.2.5 Subjective assessment
Trait anxiety was assessed once and state anxiety multiple times throughout the experiment with the
State-Trait Anxiety Inventory (STAI-T and STAI-S; Spielberger, C D; Gorsuch, R L; Lushene, 1970). The Anxiety Sensitivity Index (ASI) was used to assess one’s tendency to respond fearfully to
anxiety-related symptoms (Peterson & Reiss, 1992). In addition, after each session the evaluation of the US was assessed on a 9-point scale ranging from “very unpleasant” (1) to “not unpleasant” (9). Finally,
participants rated to US by means of arousal and valence dimensions of the Self-Assessment Manikin
(SAM; Bradley & Lang, 1994) on a 9-point scale. Higher ratings indicate higher arousal and valence.
2.2.6 Neuroimaging
Neuroimaging was conducted using a 3T Philips Achieva MRI scanner with a 32-channel head-coil.
Whole-brain blood-oxygenation-level-dependent MRI images were acquired using gradient-echo,
echo planar imaging (GE-EPI) (TR = 2000 ms; TE = 27.63 ms; FA = 76.1°; 37 axial slices with ascending
acquisition; 3 x 3 x 3.3 mm voxel size; 80 x 80 matrix; 240 x 133.98 x 240 FoV). For the acquisition
phase 636 volumes were acquired and 740 volumes for the memory phase. Foam pads minimized
head motion. For anatomical localization of functional activation, a high-resolution T1-weighted
image was collected for each participant (TR = 8.2 ms, TE = 3.8 ms, FA = 8°, voxel size = 1 x 1 x 1 mm,
FOV = 240 x 240 x 188 mm). Stimuli were backward-projected onto a screen that was viewed through
a mirror attached to the head-coil.
FEAT (fMRI Expert Analysis Tool) version 6.0.0. part of FSL (Oxford Centre for Functional MRI
of the Brain (FMRIB) Software Library: http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/) was used to analyse the
fMRI data. First, the functional images were motion corrected (MCFLIRT), slice-time corrected
(Regular up), spatial smoothed using a 5 mm full-width-at-half-maximum Gaussian Kernel, high-pass
filtered in the temporal domain (σ = 100 s) and prewhitened.
The structural images were brain extracted (BET) to which the functional images were
Page 12 of 48
(FMRIB’s Linear Image Registration Tool, FSL). The resulting normalization parameters were applied
to the functional images. All trials were modelled as separate events, resulting in a single-trial design.
The data were further analysed in Matlab (version 8.0. MathWorks) as described later on.
The analyses were restricted to a set of a-priori defined Regions of Interest (ROIs) that were
selected on the basis of their role in fear learning and (extinction) memory. The ROIs included the
anterior cingulate cortex (ACC), amygdala, caudate nucleus, fusiform gyrus (FUS), hippocampus,
inferior frontal gyrus (IFG), insula, medial temporal gyrus (MTG), occipital cortex (OC), orbitofrontal
cortex (OFC), superior parietal lobe (SPL), and ventromedial prefrontal cortex (vmPFC). The superior
frontal gyrus (SFG) is also included to illustrate that robust learning-dependent changes, as revealed
by similarity matrices, can also be observed outside the salience network (Visser et al., 2011). ROIs were obtained from the Harvard-Oxford cortical and subcortical structural atlases (Harvard Center
for Morphometric Analysis).
For each participant separately multiple vectors were created for each ROI. Each vector
contained the spatial neural activation pattern in that particular ROI elicited by one event in the
experiment, and each value in a vector represents the Z-value of one voxel in that ROI during that
event. Subsequently, Pearson correlations were calculated between all vectors of a ROI, resulting in a
similarity matrix for each participant. The similarity matrices thus contain correlations of neural
activation patterns between all pairs of events. Since the strength of the correlations indicate to
what degree the neural response pattern of two stimuli is similar (i.e., higher correlations denote
higher similarity), this matrix reveals to what extent the activation patterns elicited by one event is
similar to the activation pattern elicited by another event. All correlations in the similarity matrices
(i.e., for each ROI in each participant) were Fisher-transformed and then averaged across
participants. In total there were 26 similarity matrices: for each ROI the neural activation pattern
similarity in the acquisition phase (Session 1) and the memory phase (Session 2) (see Figure 4 for an
Page 13 of 48
From the obtained average similarity matrices, three different kinds of correlations were
selected (Visser et al., 2011). Firstly, within-stimulus correlations between consecutive trials (trial 1 and 2, trial 2 and 3, etc.) were selected for each stimulus category separately (e.g., correlation
between trial 1 CS⁺ Face and trial 2 CS⁺ Face; Figure 3). This type of correlation reveals whether the neural representation of a stimulus refines, as indicated by in increasing within-stimulus similarity
over time. The stimulus correlations were averaged over faces and houses, resulting in
within-stimulus correlations for CS⁺ and for CS⁻. Secondly, between-within-stimulus correlations between adjacent
trials for original categories (i.e., faces and houses) were selected (e.g., correlation between trial 1
CS⁺ Face and trial 1 CS⁻ Face, etc.). These correlations were averaged, resulting in 1 correlation
indicating whether the neural activation patterns similarity of stimuli within their original category
changes over time, regardless different associated reinforcement-outcomes (Figure 3). Thirdly, between-stimulus correlations between adjacent trials for the new established categories, based on
reinforcement (i.e., CS⁺ and CS⁻) were selected (e.g., correlation between trial 1 CS⁺ Face and trial 1
CS⁺ House; Figure 3). If this correlation increases over time during the acquisition phase, this would denote that new categories have been formed and that this new representation is present in the
brain. Differential neural pattern similarity (CS⁺ vs. CS⁻) would denote a higher order of fear learning
since a neural representation is formed based on affective significance (threat/no threat) rather than
a pre-existing semantic category (faces/houses).
The first and second mentioned correlations were averaged over faces and houses in order to
constrain the number of comparisons, and because it was not of interest whether faces and houses
would differ in terms of stimulus refinement, learning stimulus-shock associations or changing as
original stimulus (regardless reinforcement) over the course of conditioning.
To correct for multiple testing (13 ROIs) a sequential False Discovery Rate (FDR ≤ 5%)
Page 14 of 48
Figure 3. Calculation of the within-stimulus correlation, between-stimulus correlation of the original association and between-stimulus correlation of the new association. The within-stimulus correlations and between-stimulus correlations of the original associations are averaged over faces and houses. See Figure 4 for the similarity matrices with their selected correlations.
Figure 4. Similarity matrices of the ACC during the acquisition phase for the Placebo groups and Yohimbine group separately. 28x28 correlation matrices were created for the Placebo group and the Yohimbine group for each ROI (shown is half of the correlation matrix for the ACC). Indicated by the rectangles are the correlations obtained for analyses. For the within-stimulus correlations, the off-diagonals represent correlations between
Page 15 of 48
2.3 Experimental procedure
The experiment consisted of an acquisition phase in Session 1 and a memory phase in a second
session 48 hours later (Figure 1). At the start of the first session (T0), participants gave their written informed consent and were again medically screened to assure their participation was without any
risk. Subsequently, saliva samples were obtained, blood pressure and heart rate were measured, and
participants completed the ASI, STAI-T, STAI-S and SAM. On T1, they took an oral dose of either
Placebo or 20 mg Yohimbine HCl. Administration was planned so that peak plasma levels (60 minutes
after administration) are reached upon completion of the acquisition phase. Next, the participants
were placed on the scanner table, shock electrodes were attached, and US intensity was determined
as described before. We explicitly stated that in the task four pictures would be presented, of which
two pictures could be followed by a shock while the other two would never be followed by a shock,
and that they had to learn and remember these contingencies. Foam pads were placed around
participants’ heads to minimize head motion during scanning. Once inside the scanner, just before
task onset, salivary samples were collected (T2). Thereafter, the nine-point calibration procedure was
performed to calculate the angle between participants’ eyes and the screen which enables
measuring the pupil dilation responses. Then a structural scan was made. Subsequently, the
differential fear conditioning started, while functional scanning took place and the pupil dilation
responses to the stimuli were measured. After this first scan (T3), again salivary samples were
collected and blood pressure and heart rate were measured. Participants completed the STAI-S and
SAM, as well as an exit-questionnaire which assessed correct/incorrect identification of the CS⁺ and
CS⁻, the US evaluation, the US-expectancy ratings and the alertness throughout the experiment.
Participants returned for the second session approximately 48 hours after pill intake. This break was
inserted in order to substantiate consolidation of the fear memory. The session started with
measuring the blood pressure and heart rate, and completion of the STAI-S (T4). After that, the
participants entered the scanner room and shock electrodes were attached. The US intensity was set
Page 16 of 48
pictures would be presented as in Session 1, and they had to think about what they had learned
about them. It was repeated that the same two pictures could be followed by a shock, while the
other two never would be followed by a shock. In fact, no shocks were given, except for the
unexpected shock near the end of the second session that was applied to reinstate fear. During
functional scanning, fear extinction, fear reinstatement, and fear generalization were assessed. After
the second scan (T5), again blood pressure and heart rate were measured, and the STAI-S and SAM
were completed. Finally, an exit-questionnaire assessed the US-expectancy and alertness throughout
the experiment.
2.4 Statistical analysis
Pupil dilation responses and neural pattern similarity were subjected to repeated-measures ANOVA
with Trial and Stimulus Category (CS⁺ and CS⁻) as within-subjects factors and Group (Placebo and
Yohimbine) as between-subjects factor. For the acquisition phase (Session 1) it was tested whether
differential fear learning (CS⁺ vs. CS⁻) took place, as expressed by a significant Trial x Stimulus
Category interaction. This was only tested in case of a significant main effect of Stimulus Category.
Group differences in speed of fear acquisition were expressed by significant Trial x Stimulus Category
x Group interactions. Exploratory paired t-tests were performed to determine from which trial the
difference between CS⁺ and CS⁻ became significant, thus when fear was acquired.
Fear memory was expressed by a significant main effect of Category in Session 2. Extinction
learning (Session 2, trial 1 – 13) was expressed by a significant Trial x Stimulus Category interaction
indicating a decrease of stimulus differentiation over time. Differences in speed of fear extinction
learning were expressed by significant Trial x Stimulus Category x Group interactions. Exploratory
analyses using paired-samples t-test determined when extinction learning was completed by testing
from which trial on there was no differential effect of stimulus category anymore. Reinstatement of
fear was analysed by testing with paired t-test whether a differential fear response reappeared after
Page 17 of 48
Reinstatement was only tested when extinction was completed at least in the last trial before the
reinstatement trial (Trial 14). Generalization of fear was expressed by a significant difference
between CS⁺ and CS⁻ in the generalization trial (Trial 15) as tested with a paired-samples t-test.
Exploratory, STAI-T score and ASI score were added as covariates to the repeated measures ANOVA
to test whether these measures influence fear acquisition or extinction.
The analyses were tested two-tailed and significance level was set at p < 0.05. If the data did
not meet the assumptions for the parametric statistical tests, non-parametric alternatives were used.
Analyses were performed using Statistical Package for the Social Sciences (SPSS, version 20; SPSS),
and figures were made using R (i386, version 3.1.1).
3. RESULTS
All 41 participants had learned the contingencies (i.e., which pictures were followed by the US)
immediately after differential fear conditioning in Session 1. The Placebo group and the Yohimbine
group did not differ in terms of trait anxiety (STAI-T), baseline state anxiety (STAI-S) or anxiety
sensitivity (ASI score) (Table 1). The shock intensity was rather low in the Yohimbine group as opposed to the Placebo group, probably due to slight individual differences in shock sensitivity, but
no significant group differences were observed (t(39) = 1.31. p = 0.20). Both groups evaluated the US
equally unpleasant at the start of the first session (t(39) = 1.03. p = 0.31), and over the course of the
first session the US unpleasantness equally increased for both groups (t(39) = 1.01, p = 0.32). State
anxiety did not differ between groups at baseline (t(39) = 0.42, p = 0.68) and was not affected by
Yohimbine HCl administration since STAI-S scores remained stable in both groups throughout the
Page 18 of 48
Table 1. Mean values (SD) of the shock intensity (US), US evaluation (average begin and end Session 1), trait anxiety (STAI-T), state anxiety (STAI-S) on T0 and Anxiety Sensitivity (ASI score) for the Placebo group and Yohimbine group.
Placebo Yohimbine Test statistics Shock (US) intensity (mA) 29.30 (15.07) 23.90 (11.09) (t(39) = 1.31. p = 0.20) US evaluation 3.18 (0.77) 4.07 (1.85) (t(39) = 1.03. p = 0.31) Trait anxiety 33.75 (7.97) 35.52 (9.70) (t(39) = 0.64, p = 0.53) State anxiety (T0) 31.75 (7.31) 32.86 (9.32) (t(39) = 0.42, p = 0.68) ASI 10.10 (4.95) 11.05 (5.60) (t(39) = 0.57, p = 0.57)
Manipulation check Yohimbine HCl
To check whether the Yohimbine HCl manipulation exerted its intended physiological effects, we
analysed changes over the course of the experiment of sAA levels, systolic and diastolic blood
pressure and heart rate.
The Placebo and Yohimbine group did not differ in terms of baseline sAA levels (U = 103.0. p = 0.77) and both showed a significant increase in sAA levels over the course of Session 1
(Timepoint(3); Placebo: F(1.25, 14.96) = 4.42, p = 0.046, ηp2 = 0.27; Yohimbine: F(1.28, 20.54) = 6.93, p = 0.011,
ηp2 = 0.30)(Figure 5). However, contrary to expectations, the sAA increase over the course of the first
session was not greater for the Yohimbine group compared to the Placebo group (Time point (3) x Group (2), F(1.28, 35.70) = 0.56, p = 0.50). The sAA levels already significantly increased with from T0 to
T2 (F(1, 28) = 14,70. p = 0.001, ηp2 = 0.34). Since peak levels should have been reached just before T3
this early sAA increase was probably due to stress related to the scanning procedure. As a check, also
levels of the stress hormone cortisol throughout Session 1 were analysed. Also the salivary cortisol
levels at baseline did not differ between groups (U = 102.00. p = 0.15) and the stress hormone levels – in contrast to the sAA levels – remained stable over the course of the first session in both groups
(Timepoint (3), F(1.01, 28.16) = 0.37, p = 0.69; Time point (3) x Group (2), F(1.01, 28.16) = 1.14, p = 0.30). This
may suggest that the sAA increase in both groups just before scanning was not due to increasing
Page 19 of 48
response (Hellhammer et al., 2009; van Stegeren, Wolf, & Kindt, 2008a), a possible brief stress increase at T2 may not be reflected in salivary cortisol levels.
Figure 5. Neuroendocrine responses to drug intake (Placebo or Yohimbine HCl) on different time points during Session 1. (A) Percentage salivary alpha amylase (sAA) level change from baseline on different time points during the first session. The Yohimbine group and Placebo group had similar sAA level increases throughout the experiment during Session 1, indicating that the drug manipulation by Yohimbine HCl was not successful. (B) Percentage salivary cortisol level change from baseline on different time points during the first session. Stress levels - as indicated by cortisol - remained stable over the course of the first session in both groups. Error bars represent standard errors of the mean.
The Yohimbine administration did not affect heart rate (Time point (4) x Group (2), F(2.20. 85.81)
= 0.56, p = 0.59), nor diastolic blood pressure (Time point (4) x Group (2), F(3, 117) = 1.20. p = 0.32).
Thus, Yohimbine administration did not affect sAA or cortisol levels, nor heart rate or diastolic blood
pressure throughout the experiment. However, it did affect the systolic blood pressure (Time point (4) x Group (2), F(3, 117) = 8.01, p < 0.001, ηp2 = 0.17) as it remained stable in the Placebo group
throughout the entire experiment (F(3, 57) = 0.74, p = 0.53), whereas in the Yohimbine group the
systolic blood pressure increased increase from 114.38 mm Hg (SD = 1.80) on T0 to 123.91 mm Hg
(SD = 2.58) on T3 (F(1, 20) = 27.07, p < 0.001, ηp2 = 0.58) but remained stable during the second session
Page 20 of 48
Together these results suggest that Yohimbine HCl administration only affected systolic blood
pressure, but did not have the other intended physiological effects. We therefore conclude that the
pharmacological manipulation was not in increasing noradrenergic activity. This also implies that any
significant differences in neural activation patterns, as discussed in the next sections, are likely the
result of other factors than the experimental manipulation.
Retrospective US-expectancy ratings
US-expectancy ratings were obtained as a measure for declarative knowledge of the fear association.
Figure 6 shows the US-expectancy ratings over the course of the entire experiment for CS⁺ and CS⁻ separately.
Analyses revealed a differential increase (CS⁺ vs. CS⁻) of US-expectancy ratings during
acquisition (Session 1) (Category (2) x Time point (3), groups together; F(1.29, 24.55) = 171.97, p < 0.001,
ηp2 = 0.90). The degree to which participants’ US-expectancy increased for CS⁺ and decreased for the
CS⁻ over the course of acquisition, did not differ between groups (Category (2) x Time point (3) x Group (2): F(1.56, 60.77) = 0.22, p = 0.74). In the memory phase the Yohimbine group expected CS⁺ to a
higher extent than the Placebo group (Category (2) x Group (2), F(1, 39) = 10.84, p = 0.002). During the
memory phase (Session 2), both groups reported a similar change of differential US-expectancy over
time (Category (2) x Time point (4), groups together: F(1.76, 33.41) = 26.33, p < 0.001, ηp2 = 0.58; Category
(2) x Time point (4) x Group (2): F(1.69, 65.79) = 1.00. p = 0.36. First, participants showed extinction
learning by a significant decrease of differential US-expectancy from the beginning to halfway the
session (i.e., decrease in US-expectancy for CS⁺, but US-expectancy remained low for CS⁻) (Category (2) x Time point (2), groups together: F(1, 39) = 107,57, p < 0.001, ηp2 = 0.73). Then, participants showed
fear reinstatement induced by the reminder shock as indicated by an increase of the differential
US-expectancy from just before the shock to the trial right after the shock (F(1, 39) = 27.26, p < 0.001, ηp2 =
0.0.41). Finally, for the stimuli that appeared in colour, participants reported a significantly decreased US-expectancy for the CS⁺ but it remained low for the CS⁻ (F(1, 39) = 12.24, p < 0.001, ηp2 = 0.24). Thus,
Page 21 of 48
the differential effect was less than in the reinstatement trial but there was still a significant
difference between CS⁺ and CS⁻ expectancy in the generalization trial (t(40) = 9.88, p < 0.001).
Together, the results show that participants had declarative knowledge of the fear
associations during the acquisition phase, extinction phase, the fear reinstatement trial and the fear
generalization trial.
Figure 6. Retrospective US-expectancy ratings. Shown are the US-expectancy ratings of the Placebo group and Yohimbine group over the course of the entire experiment for CS⁺ and CS⁻ separately. After both sessions participants retrospectively rated to what extent participants expected the shock on several time points during the task (i.e., Session 1: beginning (i.e., before the first shock), halfway, and end of Session 1; Session 2: beginning of the experiment, just before the unexpected shock, right after the unexpected shock (reinstatement) and when the stimuli were presented in colour (generalization)). US-expectancy reflects the declarative knowledge of the fear associations, indicating that on the cognitive level there was fear acquisition, extinction, reinstatement and generalization. Both groups showed a similar pattern of changes in expectancy, although in Session 2 the Yohimbine group expected the CS⁺ more than the Placebo group. Error bars represent standard error of the mean.
Stimulus valence and arousal (SAM)
The stimuli were rated by means of arousal and valence dimensions of the Self-Assessment Manikin
(SAM) (Figure 7).
Analyses revealed increasing differential valence (CS⁺ vs. CS⁻) in both groups throughout the
experiment (Category (2) x Time point (4); Placebo: F(3, 57) = 11.24, p < 0.001, ηp2 = 0.37; Yohimbine:
Page 22 of 48
throughout the experiment (Category (2) x Time point (2) x Group (2), F(3, 117) = 0.47, p = 0.87): both
groups reported an equally increasing negative feeling of the CS⁺ but a stable valence of the CS⁻
during Session 1 (T0 vs. T3, F(1, 39) = 67.19, p < 0.001, ηp2 = 0.63), then a differential valence decrease
in between the sessions (T3 vs. T4, F(1, 39) = 9.92, p = 0.003, ηp2 = 0.20), and again a differential
increase during Session 2 (T4 vs. T5, F(1, 39) = 12.90. p = 0.001, ηp2 = 0.25).
Figure 7. Stimulus valence and arousal. Using the Self-Assessment Manikin (SAM), stimulus valence and arousal for the CS⁺ and CS⁻ were assessed on different time points throughout the experiment. It should be noted that the stimulus arousal scores were recalculated so that higher values indicate higher arousal. For Error bars represent standard error of the mean.
Regarding stimulus arousal, both groups showed an increasing differential effect of stimulus
category (CS⁺ vs. CS⁻) to that category throughout the experiment (Category (2) x Time point (4); Placebo: F(3, 57) = 23.80. p < 0.001, ηp2 = 0.56; Yohimbine: F(3, 60) = 25.52, p < 0.001, ηp2 = 0.56) and
changes in differential stimulus arousal over the course of the experiment was similar for both
groups (Category (2) x Time point (2) x Group (2), F(3, 117) = 0.47, p = 0.71, ηp2 = 0.01). First the
differential arousal increased during Session 1 (T0 vs. T3, F(1, 39) = 117.47, p < 0.001, ηp2 = 0.75), then
decreased in between the sessions (T3 vs. T4, F(1, 39) = 11.69, p = 0.001, ηp2 = 0.23), and then increased
Page 23 of 48
Together these results show that participants rated the arousal provoked by, and the valence
of the different stimuli in a different extent, dependent on the associated fear on that moment.
Pupil dilation response
The pupil dilation response to the conditioned stimuli was measured during the task in both sessions
to assess the behavioural expression of fear.
Successful fear acquisition was evident from an increased differential pupil dilation response
(CS⁺ vs. CS⁻) over the course of conditioning in Session 1 (Category (2) x Trial (13), F(1.94,75,70) = 7.11, p
= 0.002, ηp2 = 0.15).However, both groups showed a similar increase of the differential pupil dilation
response (Category (2) x Trial (13) x Condition (2), F(1.94, 75.70) = 0.68, p = 0.51). Thus, the Yohimbine
manipulation did not affect the degree of fear acquisition.
Within-subjects contrasts revealed that the differential fear response decreased from the last
trial of Session 1 to the first trial of Session 2 (Category x Trial, F(1, 39) = 5.08, p = 0.03), but was still
present as indicated by a main effect of category in Session 2 (Placebo: F(1, 19) = 5.97, p = 0.025, ηp2 =
0.24; Yohimbine: F(1, 20) = 27.28, p < 0.001, ηp2 = 0.58). As there was a significant difference between
CS⁺ and CS⁻ in Session 2, this indicates that fear memory was present in both groups. Although it
seems from Figure 8 that the Placebo group had greater pupil responses to the CS⁺ than the Yohimbine group at the start of the extinction phase, there was no group effect on Category
(Category (2) x Group (2), F(1, 39) = 0.66 p = 0.42).
Over the course of the extinction phase (Trial 1 – 13), it is evident that in both groups the
pupil response to CS⁺ decreased (Category (2) x Trial (13), F(2.15, 83.71) = 7.82, p = 0.001, ηp2 = 0.17), and
this was similar in both groups (Category (2) x Trial (15) x Group (2), F(2.33, 91.01) = 0.76, p = 0.72).
However, exploratory analyses revealed that extinction was never completed as in all trials there was
a differential fear response. Compared to e.g., Visser et al. (2013) where extinction learning started after trial 3, extinction learning in the current sample seems to be impaired.
Page 24 of 48
Near the end of the second session, an unexpected shock was applied to reinstate fear
responses (i.e., differential pupil responses for CS⁺ and CS⁻). However, since fear was not vanished
(i.e., extinction not completed), fear could not be reinstated and within-subjects contrasts revealed
that the unexpected shock did not induce a significant stronger differentiation (Trial 13 vs. Trial 14; Category (2) x Trial (2), F(1, 39) = 3.98, p = 0.053). The participants had significant differential fear
responses to the stimuli that appeared in colour, thus fear was generalized (Z = 2.42, p = 0.016). Trait anxiety (STAI-T) and the anxiety sensitivity (ASI) were not related to the pupil dilation
response in the acquisition phase (STAI-T: F(1, 38) = 0.74, p = 0.40; ASI: F(1, 38) = 0.58, p = 0.45), nor in
the extinction phase (STAI-T: F(1, 38) = 1.77, p = 0.19; ASI: F(1, 38) = 0.45, p = 0.50).
Together these results confirmed that the Yohimbine manipulation was unsuccessful, since
no significant group differences were observed regarding the pupil dilation response during
acquisition, extinction, reinstatement or generalization of fear. Both groups showed impaired
extinction which may suggest that a strengthened memory had been formed in both groups.
Figure 8. Pupil dilation responses in the Placebo group and Yohimbine group to CS⁺ and CS- during the acquisition phase (Session 1) and memory phase (Session 2). Clear fear learning (trial 1-13 in Session 1), fear extinction (trial 1-13 in Session 2), reinstatement of fear (trial 14 in Session 2), and generalization of fear (trial 15 in Session 2) can be observed in both the Yohimbine group and the Placebo group. Error bars represent standard error of the mean.
Page 25 of 48
Neural pattern similarity
On the behavioural level, fear acquisition and impaired extinction were observed. Analyses were
performed to test whether this finding is also present on the neural level. For this, we analyses
within-stimulus and between-stimulus neural pattern similarity during fear acquisition, extinction,
reinstatement and generalization in 13 ROIs (Table 2-5; Figure 9).
Anterior Cingulate Cortex (ACC)
In the ACC both groups showed a similar differential increase (CS⁺ vs. CS⁻) of both within-stimulus
and between-stimulus neural pattern similarity over the course of acquisition (Within: Category (2) x Trial (6) x Group (2): F(5, 195) = 1.50, p = 0.19; Category (2) x Trial (6), groups together: F(5, 195) = 13.88, p
< 0.001, ηp2 = 0.26; Between: Category (3) x Trial (7) x Group (2): F(8.36, 362.09) = 0.93, p = 0.49; Category
(3) x Trial (7), groups together: F(8.36, 362.09) = 6.77, p < 0.001, ηp2 = 0.15). A paired t-test revealed that
the within-stimulus neural patterns of CS⁺ and CS⁻ became significantly different from trial 3 to 4 (t(40)
= 7.52, p < 0.001). For the between stimulus, that was since trial 3 (t(40) = 2.92, p = 0.01).
Fear memory was present, as indicated by a main effect of Stimulus Category (Within: F(1, 39) =
46.48, p < 0.001, ηp2 = 0.54; Between: F(1.68, 65.50) = 49.59, p < 0.001, ηp2 = 0.56). During the extinction
phase (Session 2, trial 1 to 13) in both groups the stimulus differentiation equally decreased over
time, indicating that extinction learning occurred and the speed of extinction learning was the same
for both groups. (Within: Category (2) x Trial (6) x Group (2): F(5, 195) = 1.02, p = 0.41; Category (2) x
Trial (6), groups together: F(5, 195) = 13.38, p < 0.001, ηp2 = 0.26; Between: Category (3) x Trial (7) x
Group (2): F(8.17, 318.58) = 0.52, p = 0.85; Category (3) x Trial (7), groups together: F(8.19, 327.43) = 4.67, p <
0.001, ηp2 = 0.10). Extinction learning was completed when the neural patterns were not significantly
different for CS⁺ and CS⁻ anymore; for within stimulus this was in trial 4 to 5 (t(40) = 1.07, p = 0.29)
whereas for between stimulus this was in trial 6 (t(40) = 1.41, p = 0.17). Fear was not reinstated by the
Page 26 of 48
to the stimuli in colour (Within: Z = 1.76, p = 0.08; Between: Z = 0.99, p = 0.32), as there was no differential neural pattern similarity for CS⁺ and CS⁻ in those trials.
In sum, in the ACC clear fear learning and extinction were observed in within-stimulus and
between-stimulus neural patterns, but fear was neither reinstated nor generalized.
Amygdala
In the Amygdala there was no fear acquisition, neither for the within-stimulus nor the
between-stimulus neural pattern similarity (Within: Category (2) x Trial (6), groups together, F(5, 195) = 2.08, p =
0.07; Between: Category (3) x Trial (7), groups together, F(8.07, 314.87) = 1.24, p = 0.27). Noteworthy, fear
memory was observed in the second session, as indicated by a significant main effect of Stimulus
Category (Within: Category (2); F(1, 39) = 15.27, p > 0.001, ηp2 = 0.28; Between: Category (3); F(1.59, 62.00)
= 23.70, p < 0.001, ηp2 = 0.38). For the within-stimulus even extinction learning occurred (Category (2)
x Trial (6), groups together: F(5, 195) = 5.73, p < 0.001, ηp2 = 0.13) but the speed of extinction learning
was the same for both groups (Category (2) x Trial (6) x Group (2): F(5, 195) = 1.82, p = 0.11). Extinction
learning was completed in trial 4 to 5 (t(40) = 1.55, p = 0.13). For the between-stimulus however,
extinction learning did not occur (Category (3) x Trial (7), groups together: F(12, 480) = 1.62, p = 0.08).
Fear was not reinstated by the unexpected shock (Within: Z = 0.46, p = 0.65; Between: Z = 0.12, p = 0.90), nor was fear generalized to the stimuli in colour (Within: Z = 0.53, p = 0.60; Between: Z = 0.10,
p = 0.92), as there was no differential neural pattern similarity for CS⁺ and CS⁻ in those trials.
Thus, in the amygdala for both the within-stimulus and between-stimulus we observed no
fear acquisition, reinstatement or generalization, but fear memory formed. Extinction occurred for
the within-stimulus but not for the between stimulus.
It is unlikely that fear memory has been formed, when no differential neural pattern
similarity was observed during the acquisition phase. However, an explanation is that the amygdala is
only involved in the fear expression after the consolidation of the fear memory. Since there is only
Page 27 of 48
fear memory (between-stimulus) remains, but the low level fear memory (within-stimulus)
diminished in the absence of fearful outcomes (US).
Caudate nucleus
In the caudate, there was no fear acquisition (Within: Category (2) x Trial (6), groups together, F(5, 195)
= 1.67, p = 0.14; Between: Category (3) x Trial (7), groups together, F(7.38, 287.98) = 1.68, p = 0.11).
However, fear memory was observed in the second session, as indicated by a significant main effect
of Stimulus Category (Within: Category (2), F(1, 39) = 5.78, p = 0.02, ηp2 = 0.13; Between: Category (3),
F(1.66, 64.62) = 16.38, p < 0.001, ηp2 = 0.30). For the within-stimulus, extinction learning occurred
(Category (2) x Trial (6), groups together: F(5, 195) = 5.70, p < 0.001, ηp2 = 0.13) but the speed of
extinction learning was the same for both groups (Category (2) x Trial (6) x Group (2): F(5, 195) = 1.69, p
= 0.16). Extinction learning was completed in trial 3 to 4 (t(40) = 1.16, p = 0.25). ). For the
between-stimulus, extinction learning also occurred (Category (3) x Trial (7), groups together: F(12, 480) = 1.87, p
= 0.035) but it did not reach the FDR-corrected significance level. The unexpected shock did not reinstate fear (Within: Z = 0.94, p = 0.35; Between: Z = 1.72, p = 0.09). For the between-stimulus, fear was not generalized (Z = 0.63, p = 0.53), but for the within-stimulus, fear was generalized to the stimuli in colour (Z = 2.02, p = 0.04), as the neural patterns for CS⁺ were more similar than for CS⁻.
Altogether, the caudate showed - just like the amygdala - no fear acquisition but there was
fear memory in Session 2. And again, only the within-stimulus showed fear extinction, and fear was
not reinstated by the unexpected shock. However, in the caudate there was fear generalization, but
only for the within-stimulus.
Fusiform Gyrus
For the Fusiform Gyrus (FUS) very different results were observed: both groups showed a similar
differential decrease of the within-stimulus and between-stimulus over the course of the acquisition
Page 28 of 48
groups together: F(3.76, 146.67) = 2.77, p = 0.03, ηp2 = 0.07; Between: Category (3) x Trial (7) x Group (2),
F(7.94, 309.57) =0.85, p = 0.56; Category (3) x Trial (7), groups together: F(7.94, 309.57) = 3.13, p < 0.001, ηp2 =
0.07).During the extinction phase a main effect of stimulus category was observed (Within: Category (2), F(1, 39) = 10.19, p = 0.003, ηp2 = 0.21; Between: Category (3), F(1.69, 65.76) = 20.62, p < 0.001, ηp2 =
0.35), but there was no differential change over time (Within: Category (2) x Trial (6) x Group (2), F(5, 195) = 1.51, p = 0.19; Category (2) x Trial (6), groups together: F(5, 195) = 1.51, p = 0.55; Between:
Category (3) x Trial (7) x Group (2), F(8.061, 314.37) = 1.22, p = 0.29; Category (2) x Trial (6), groups
together: F(7.85, 313.91) = 0.32, p = 0.96). Fear was not generalized to the stimuli in colour (Within: Z =
0.16, p = 0.87; Between: Z = 0.16, p = 0.87).
Taken together, the observed neural pattern similarities in the FUS suggest that this brain
area is not involved in fear memory formation or extinction.
Hippocampus
In the Hippocampus, both groups showed similar fear acquisition for the within-stimulus (Within: Category (2) x Trial (6) x Group (2), F(5, 195) = 1.09, p = 0.37; Category (2) x Trial (6), groups together:
F(5, 195) = 4.14, p = 0.001, ηp2 = 0.10). The difference between CS⁺ and CS⁻ was significant since trial 4-5
(t(40) = 3.87, p < 0.001). However, no fear acquisition for the between-stimulus was observed
(Between: Category (3) x Trial (7) x Group (2), F(12.468) = 1.64, p = 0.08; Category (3) x Trial (7), groups
together: F(12, 468) = 1.55, p = 0.10). Fear memory was present as indicated by a significant main effect
of stimulus category, not only for within-stimulus, but also for the between-stimulus (Within: Category (2), F(1, 39) = 26.40, p < 0.001, ηp2 = 0.40; Between: Category (3), F(2, 78) = 20.87, p < 0.001, ηp2
= 0.35). For the within-stimulus this difference between CS⁺ and CS⁻ diminished equally in both groups during the extinction phase (Within: Category (2) x Trial (6) x Group (2), F(5, 195) = 1.21, p =
0.30; Category (2) x Trial (6), groups together: F(5, 195) = 3.98, p = 0.002, ηp2 = 0.09). Extinction was
completed in trial 3-4 (t(40) = 1.76, p = 0.09). Also for the between-stimulus extinction occurred
Page 29 of 48
together: F(12, 480) = 1.85, p = 0.038, ηp2 = 0.04) but it did not reach the FDR-corrected significance
level. Fear was not reinstated (Within: Z = 0.68, p = 0.50; Between: Z = 0.05, p = 0.96), nor generalized (Within: Z = 0.580.16, p = 0.56; Between: Z = 0.06, p = 0.95).
Thus, in the hippocampus the within-stimulus showed fear acquisition, memory and
extinction but not fear reinstatement or generalization. The between-stimulus, however, did not
show fear acquisition but fear memory was observed in Session 2. This fear memory was not
vanished during the extinction phase, nor was it generalized.
Inferior Frontal Gyrus (IFG)
In the IFG, both groups showed a similar differential increase (CS⁺ vs. CS⁻) over the course of
acquisition for both within-stimulus and between-stimulus neural pattern similarity (Within: Category (2) x Trial (6) x Group (2), F(5, 195) = 0.97, p = 0.44; Category (2) x Trial (6), groups together, F(5. 195) = 13.54, p < 0.001, ηp2 = 0.26; Between: Category (3) x Trial (7) x Group (2), F(8.19, 319.42) = 0.98, p =
0.46; Category (3) x Trial (7), groups together, F(8.19, 319.42) = 7.17, p < 0.001, ηp2 = 0.16). A paired t-test
revealed that the within-stimulus neural patterns of CS⁺ and CS⁻ became significantly different from
trial 2 to 3 (t(40) = 3.80, p < 0.001); for the between-stimulus this was in trial 3 (t(40) = 4.18, p < 0.001).
Fear memory was present as indicated by a significant main effect of stimulus Category (Within: Category (2), F(1, 39) = 63.19, p < 0.001, ηp2 = 0.62; Between: Category (3), F(2, 78) = 60.82, p < 0.001, ηp2
= 0.61) but this difference between CS⁺ and CS⁻ diminished equally in both groups during the extinction phase (Within: Category (2) x Trial (6) x Group (2), F(5, 195) = 1.17, p = 0.33; Category (2) x
Trial (6), groups together, F(5, 195) = 7.50, p < 0.001, ηp2 = 0.16; Between: Category (3) x Trial (7) x
Group (2), F(8.17, 318.43) = 0.44, p = 0.90; Category (3) x Trial (7), groups together, F(8.18, 327.39) = 4.57, p <
0.001, ηp2 = 0.10). Extinction was completed only in trial 6-7 for the within-stimulus (t(40) = 1.31, p =
0.20) and in trial 6 for the between-stimulus (t(40) = 1.68, p = 0.10)
For the within-stimulus, fear was not reinstated (Z = 0.53, p = 0.60), nor generalized (Z = 1.55,
p = 0.12). For the between-stimulus, fear was reinstated (Z = 2.13, p = 0.03), but this did not survive FDR correction. Fear was not generalized (Z = 0.51, p = 0.61).
Page 30 of 48
Thus, in the IFG very early fear acquisition was observed for both within- and
between-stimulus as already after one US presentation a differential neural pattern similarity was present.
Also the extinction was impaired since it took 6 trials for the fear memory to vanish. However, fear
was not reinstated nor generalized.
Insula
In the Insula, both groups showed a similar differential increase (CS⁺ vs. CS⁻) over the course of
acquisition (Within: Category (2) x Trial (6) x Group (2), F(5, 195) = 0.97, p = 0.44; Category (2) x Trial (6),
groups together, F(4.09, 159.38) = 19.82, p < 0.001, ηp2 = 0.34; Between: Category (3) x Trial (7) x Group
(2), F(7.84, 305.61) = 1.27, p = 0.26; Category (3) x Trial (7), groups together, F(7.84, 305.61) = 5.08, p < 0.001,
ηp2 = 0.12). A paired t-test revealed that the neural patterns of CS⁺ and CS⁻ became significantly
different already from trial 2 to 3 for the within-stimulus (t(40) = 4.29, p < 0.001) and in trial 3 for the
between-stimulus (t(40) = 2.97, p = 0.01).
Fear memory was present as indicated by a significant main effect of stimulus Category
(Within: Category (2), F(1, 39) = 58.87, p < 0.001, ηp2 = 0.60; Between: Category (3), F(1.68, 65.42) = 60.16, p
< 0.001, ηp2 = 0.61), but this difference between CS⁺ and CS⁻ diminished equally in both groups
during the extinction phase (Within: Category (2) x Trial (6) x Group (2), F(5, 195) = 1.14, p = 0.34;
Category (2) x Trial (6), groups together, F(5, 195) = 9.93, p < 0.001, ηp2 = 0.20; Between: Category (3) x
Trial (7) x Group (2), F(7.67, 299.04) = 0.84, p = 0.56; Category (3) x Trial (7), groups together, F(7.81, 312.24) =
5.48, p < 0.001, ηp2 = 0.12). Extinction was completed in trial 5 – 6 for the within-stimulus (t(40) = 1.93,
p = 0.60) and in trial 6 for the between-stimulus. Fear was not reinstated for the within-stimulus (Z = 0.15, p = 0.88) but for the between-stimulus fear reinstatement occurred (Z = 3.78, p < 0.001) as the CS⁺ showed a significant greater neural pattern similarity than the CS⁻. However, fear was
generalized for the within-stimulus (Z = 2.35, p = 0.02) since there was a difference between CS⁺ and CS⁻ in the trials with the stimuli in colour, but not with the between-stimulus (Z = 1.43, p = 0.15).
Page 31 of 48
So, the insula showed a similar pattern as the IFG as an early acquisition was observed for
both between- and within-stimulus. Extinction seemed to be impaired since it took 6 trials to vanish
the difference for CS⁺ versus CS⁻ in degree neural pattern similarity. Only between-stimulus showed
fear reinstatement, indicating that the higher order fear memory was recovered after one US
presentation. Also, fear was generalized to the stimuli in colour. In sum, in the insula a strong higher
order fear memory had been formed early and after extinction it could easily be reinstated.
Medial Temporal Gyrus (MTG)
For both the within-stimulus and between-stimulus neural pattern similarity in the MTG, both groups
showed a similar differential increase (CS⁺ vs. CS⁻) over the course of acquisition (Within: Category (2) x Trial (6) x Group (2), F(3.76, 146.67) = 0.65, p = 0.62; Category (2) x Trial (6), groups together, F(5. 195) =
11.32, p < 0.001, ηp2 = 0.23; Between: Category (3) x Trial (7) x Group (2): F(8.16, 318.09) = 0.90, p = 0.52;
Category (3) x Trial (7), groups together: F(8.16, 318.09) = 6.14, p < 0.001, ηp2 = 0.14). A paired t-test
revealed that the neural patterns of CS⁺ and CS⁻ became significantly different from trial 3 to 4 for
the within-stimulus (t(40) = 5.60, p < 0.001) and from trial 3 for the between-stimulus (t(40) = 4.47, p <
0.001).
Fear memory was present as indicated by a significant main effect of Stimulus Category
(Within: Category (2), F(1, 39) = 27.76, p < 0.001, ηp2 = 0.42; Between: Category (3), F(2, 78) = 36.90, p <
0.001, ηp2 = 0.49), and this difference between CS⁺ and CS⁻ diminished equally in both groups during
the extinction phase (Within: Category (2) x Trial (6) x Group (2), F(5, 195) = 1.12, p = 0.35; Category (2)
x Trial (6), groups together, F(5, 195) = 5.39, p < 0.001, ηp2 = 0.12; Between: Category (3) x Trial (7) x
Group (2), F(8.01, 312.38) = 0.39, p = 0.93; Category (3) x Trial (7), groups together, F(8.01, 312.38) = 3.13, p
0.002, ηp2 = 0.07). Extinction was completed in trial 4-5 for the within-stimulus (t(40) = 1.45, p = 0.15),
and in trial 5 for the between-stimulus (t(40) = 1.98, p = 0.054). Fear was not reinstated (Within: Z =
0.33, p = 0.74; Between: Z = 0.75, p = 0.46), nor generalized (Within: Z = 0.28, p = 0.78; Between: Z = 1.04, p = 0.30).
