Discrepancies between Autonomic and Self-Report Measures of Shame Reactivity and their Neurobiological Underpinnings: Evidence for Repression?

1

Discrepancies between Autonomic and Self-Report Measures of

Shame Reactivity and their Neurobiological Underpinnings:

Evidence for Repression?

Sanne Riemsma

10670890

Master Brain and Cognitive Sciences, University of Amsterdam

42 ECs

September 2015 – May 2016

Supervisor: Rick Wassing, Msc

Research Group: Sleep and Cognition, Netherlands Institute of Neuroscience

Co-assessor & Uva-Representative: Lucia Talamini, PhD

2

Shame is a painful self-evaluative emotion and maladaptive coping with shame can greatly hinder therapeutic progress. Previous research indicates a relationship between the adoption of different emotional coping styles, and the discrepancy between self-report and autonomic responses to shame. Consequently, investigating the BOLD-signal responses,

emotional reactivity and autonomic response to shame and relating these responses to different types of emotional coping strategies can greatly improve our knowledge of the processes behind shame and coping with shame. By examining the different BOLD-signal responses in a shame-eliciting task, the main objective of the current study is to investigate which brain activation patterns are indicative of repressive coping behaviour. The results of the present study could not confirm brain areas typically related to shame found in previous studies. The cerebellum showed significant associations between our measure of repression (the emotional discrepancy score) and activations in response to shameful stimuli relative to neutral stimuli, indicating a role for the cerebellum in the regulation of shameful emotions. Many different brain areas related to shame, emotion processing and emotional memory retrieval were significantly positively or negatively associated with shame proneness, adaptive coping, and the coping styles denial, aggression and avoidance. The present study was the first to investigate and report neurobiological underpinnings of both the repression of shame and different types of coping behavior in response to shame

Key words

3

Introduction 4

Methodology 7

Ethical Approval and Informed Consent 7

Participants 7

Experimental Stimuli and Task Procedure 8 Compass of Shame Scale Questionnaire 9 Skin Conductance – Preprocessing 9

MRI Acquisition 9

MRI Preprocessing 10

Emotional Discrepancy Factor 10

First-level Analysis 11

Second-level Analysis 12

Third-level Analysis 12

Results 14

Emotional Reactivity – Subjective Responses 14 Assessing negative stimulus specific activations for all participants 14 Assessing associations between negative stimulus specific activations and the EDF 14 Assessing associations between negative stimulus specific activations, the EDF and Coping Styles 16

Discussion 20

Emotional Discrepancy Factor (EDF): Measure of Repression 21

The Role of the Cerebellum 24

Coping Styles 24 Shame Proneness 25 Adaptation 25 Denial 26 Aggression 26 Avoidance 28

Most Striking Findings and Implications 29

Methodological Evaluation 30

Future Research 31

References 31

Appendix I: Exclusion Criteria 37

4

Complex self-conscious emotions such as shame, guilt, pride and humiliation, play an important role in modern psychoanalytical practices. Of these complex emotions, shame is a painful and self-evaluative emotion, which can make the patient want to hide from his or her feelings and thoughts (Harper, 2011). Moreover, shame is an emotion that can overwhelm a person, leading to a characteristic hyper-aroused state (Dickerson et al., 2004). Individuals that are prone to shame can cope with or defend against their feelings in a variety of unhealthy ways, examples of which are externalizing coping styles (such as denial or aggression) and internalizing coping styles (such as avoidance and self-devaluation) (Nathanson, 1992; Harper, 2011).

Previous research indicates a relationship between the adoption of different emotional coping styles, and the difference between the self-report and physiological responses to an emotion. For instance, individuals that employ a repressive coping technique, often referred to as ‘repressors’, generally tend not to recognize and label negative emotions (Lane et al., 2000). In addition, repressors have dissociated their somatic reactions from their perceptions of distress, reporting low levels of emotional distress and anxiety but exhibiting high levels of autonomic nervous system activity as seen in studies using physiological and behavioural measures (Weinberger et al., 1979; Asendorf & Scherer, 1983; Barger, Kircher, & Croyle, 1997; Lambie & Baker, 2003; Pauls & Stemmler, 2003; Myers, 2010). This discrepant pattern of response to emotions calls into question the validity of the self-report measures used, but several studies provide evidence that repressors deceive themselves rather than others (Millham & Kellogg, 1980; Derakshan & Eysenck, 1999), validating the self-report measures of repressors.

As of yet, the theory and practice of emotion repression is mostly examined within the psychoanalytical practice. Investigating the neurobiological underpinnings of repression through neuroscientific methodologies can provide a much needed scientific background for our current understanding of repression. Partly responsible for this lack of research is the challenge of designing a study that uses neurobiological measures to validate a psychoanalytical construct. In the current interdisciplinary study we aim to examine repression, as identified in psychoanalytical practice, by investigating different coping techniques following shame through the use of multiple neuroscientific methodologies. One method is the use of blood oxygen-level-dependent (BOLD) signals, which are in vivo changes in blood oxygenation that can be detected with magnetic resonance imaging. Consequently, in this functional magnetic resonance imaging (fMRI) study, we investigate the blood-oxygen-level dependent contrast imaging to the emotion shame in a shame-eliciting autobiographical memory task. During the task, skin conductance response (electrodermal activity) is measured, reflecting sympathetic tone and frequently used as a solid index of autonomic arousal (Critchley, 2000). Secondly, subjective self-report measures of shame

5

are taken. Previous studies identified an affective-autonomic response discrepancy as a factor of repression (Bonnano et al., 1995; Coifman et al., 2007); this discrepancy is used as an index of repression in the current study.

In Panksepp’s theory of affective consciousness (1998) a distinction is made between three processes that negate affective consciousness. In this theory, affective consciousness is considered a comparatively intrinsic function of all mammalian species of animals. So although much of the evidence for the model is based on animal studies in rats, its implications are generalized to humans. First, Panksepp identifies a primary form of affective consciousness, which is an intensely affective state with autonomic nervous system activity and arises directly from medial subcortical networks, and runs from midbrain periaqueductal gray (PAG) regions to various basal ganglia nuclei that interact with paleocortical brain functions (cingulate, insula, medial- and orbitofrontal cortices) (Panksepp, 1998a,b; Liotti & Panksepp, 2004; Watt & Pincus, 2004; Panksepp, 2010 and in humans: Damasio et al., 2000). Secondary emotional processes arise in the upper limbic regions (the cingulate gyrus, anterior thalamus and fornix) that facilitate simple emotional learning. An example are classical and operant fear conditioning models (LeDoux, 2014). Lastly, the third level of affective processing is higher-order affective cognitive awareness over consciousness, or meta-cognition, elaborated by medial-frontal regions that can only be studied well in humans (Panksepp, 2010). When applying Panksepp’s theory of affect to people utilizing a repressive coping style, it becomes clear that repressors lack the abstract meta-cognition that integrates the perception of emotion that is typical of the tertiary level of affective consciousness. Derakshan and Eysenck (2005) come to a similar conclusion when their findings indicate that repressors are typically self-deceivers when it comes to acknowledging emotion, because they lack meta-consciousness of their internal states.

Very few studies exist thus far that have investigated brain activations related to repressive coping behaviour. Findings from electroencephalographic (EEG) research suggest that frontal brain areas may play an important role in repressive coping (Tomarken and Davidson, 1994; Kline et al., 1998). An fMRI study by Sander et al. (2003) examined cerebral activation in repression though identification tasks with acoustically presented sad and happy speech material, presented to healthy women that were classified as either high or low repressive. In four different types of tasks, they found significant orbitofrontal cortical activation for repressors compared to nonrepressors. Repressors also showed larger left than right hemisphere activation in temporo-parietal regions during the identification of sad intonations. Rauch et al. (2007) studied the neural correlates of coping modes during the perception of threatening and nonthreatening socially relevant information, using pictures of faces with different emotions that were presented masked or unmasked. Two groups of subjects were used that were defined either as consistent repressors (high cognitive avoidance and low vigilance) or sensitizers (high vigilance and low cognitive avoidance) using the Mainz Coping Inventory (MCI) to assess vigilant or avoidant coping strategies in threatening situations (Krohne and Egloff, 1999; Krohne et al., 2000). Their results show that

6

repressors exhibit more top-down cortical regulation of the amygdala (as seen by increased ventromedial prefrontal cortex activation) than sensitizers during the processing of fearful faces.

Furthermore, the neurobiological substrates of shame are still relatively unknown, since neuroscientific research has often focussed on less complex emotions such as anger, fear and happiness. Generally, brain-imaging studies indicate that the fronto-temporo-limbic network is involved in the generation of emotions. A recent pilot fMRI study on the neurobiological underpinnings of shame and guilt showed activations for both emotions in the temporal lobe (anterior cingulate cortex, parahippocampal cortex), and specific activations for shame in the frontal lobe (medial and inferior frontal gyrus) (Michl et al., 2014). The visceral sensation of shame – including both affective and cognitive components – activates both limbic and paralimbic structures: the insular, anterior cingulate and prefrontal cortices (Aziz et al., 2000). The amygdala-hippocampus network is shown to be involved in the arousal-dimension modulation of emotional memories (Dolcos et al., 2004; Kensinger and Corkin, 2004). Moreover, since maladaptive coping with shame has been recognized as a process that can dramatically hinder therapeutic progress (Lewis, 1971; Seidler and Jenkins, 2000; Shearn et al., 1990), a better understanding of shame and responses to shame are important for the psychoanalytical clinical practice. Consequently, investigating the BOLD-signal responses and the ANS response to shame and relating these responses to different types of self-reported emotional coping strategies can greatly improve our knowledge of the fundamental processes behind shame and its coping styles.

By examining the different BOLD-signal responses in a shame-eliciting task, the main objective of this study is to investigate which brain activation patterns are indicative of repressive coping behaviour. We expect to see a significant BOLD-signal response in brain areas involved in the experience of shame following the negative (i.e. shameful) stimuli compared to the neutral stimuli. Thereafter, we expect to find brain activations indicative of repression in response to shame by evaluating the discrepancies between an individual’s subjective rating of emotional reactivity and skin conductance response. We expect associations between brain areas that are found in response to our measure of repression to be further clarified by different coping styles in response to shame.

7

Ethical Approval and Informed Consent

The present study received ethical approval (as part of the larger PhD project) by the ethics committee of the faculty of Social and Behavioural Sciences at the University of Amsterdam. After explaining all procedures of the experiment in detail, participants gave their written informed consent. All experiments were performed according to the ethical guidelines of the latest revision of the Declaration of Helsinki.

Participants

Of all participants that voluntarily took part in the larger study, a total of 36 participants were included in this study (mean age = 45.9; SD = 15.8; 17 men and 19 women). Participants were recruited via the website of the Sleep Registry (Benjamins et al., 2013), social media, adverts in popular science magazines and electronic mailing. Participants subsequently signed up for the study via the website of the research group Sleep & Cognition from the Netherlands Institute for Neuroscience. Since this experiment is part of a larger study about sleep, there were clinical insomniacs amongst the sample (17 insomniacs and 10 controls). Insomniacs were diagnosed based on the Diagnostic and Statistical Manual of Mental Disorders (DSM-V, 5th _{edition) criteria that included sleep complaints such as difficulty to initiate sleep, difficulty}

maintaining sleep and early-morning awakenings with inability to return to sleep. Complaints had to cause significant distress of impairment in the participant’s functioning, be present for at least 3 months, at least 3 nights a week. Importantly, these sleep complaints could not be explained by another sleep-wake disorder, substance use or coexisting mental disorders (American Psychiatric Association, 2013). To test the severity of sleep complaints, a questionnaire consisting of 7 questions about sleep habits called the Insomnia Severity Index (ISI) was assessed (Morin & Barlow, 1993). The participant’s total score on the ISI was used as an indication of severity of insomnia. The mean ISI-score of insomniacs was 17.4 (SD = 4.9) and 4.7 (SD = 3.8) for controls. No distinction was made between healthy participants and insomnia patients in this study because only data is used from before the night of sleep in the sleep laboratory, thereby discarding the influence of the sleep manipulation from the larger study. Exclusion criteria were the presence of other sleep disorders different from Primary Insomnia, depression, psychological trauma, neurological and endocrinal disorders, claustrophobia and other factors excluding a person from MRI studies (see appendix I for a full list). Participants were instructed to refrain from consuming any caffeine (8 hours) or alcohol (24 hours) before and during the experiment. Before the start of the experiment, the participants filled in a questionnaire about their sleep habits and emotional characteristics. Participants received financial compensation including travel expenses for their participation in the study.

8

The fMRI task consists of an emotion-induction task based on shameful autobiographical memories. One week prior to the experiments, participants were instructed to think of at least 5 shameful and 5 neutral autobiographical memories, matched for time since the event, and describe the memories in four key words. The cue words are shown during the experiment to reactivate these shameful memories with the help of the 4 cue-words, so participants can imagine they are the protagonist in a movie about their own memory. Only 4 words were used to describe the particular memory, allowing the participants to remember and replay the memory during the task, while at the same time giving some degree of privacy to the participant (i.e.: the shameful memory remains unknown to the investigators).

Each trial started with a fixation cross in the center of the screen for 7-9 seconds (figure 1). Then, four words of one condition (neutral or negative) were shown simultaneously on the screen (16 s). After this, the participants were asked to rate the intensity of their emotional response on a scale from 1 to 4 (none, a bit, somewhat, strong; 3.5 s). Subsequently, the participants performed a 1-back task (15 s) during which 9 simultaneously written and spoken letters were presented. After each letter participants had to respond whether the letter was the same or different from the previous letter. The 1-back task was introduced to ease their focus from the negative or neutral stimuli and concentrate on the 1-1-back attention task. The participants were then asked to rate the effort that was needed to perform the 1-back task on a scale from 1 to 4 (none, a bit, somewhat, strong; 3.5 s). The complete procedure consisted of 5 neutral and 5 negative trials, with the different conditions being presented in a counterbalanced manner. After a short break, the same task was repeated with the same words in a different presentation order of the neutral and negative conditions. During the fMRI task, the skin conductance signal was collected using the BrainVision physiology module. The skin conductance signal was measured by placing two electrodes on the inside of the left hand on the second digit of the index and middle fingers.

Figure 1. Timeline of the fMRI task. The task starts with a 7-9 second (jittered) fixation cross, after which the

experimental stimulus (4 negative or 4 neutral words describing a memory) were presented for 16 seconds and emotional reactivity was rated on a scale from 1-4 during 3.5 seconds. The 1-back task followed for 15 seconds, with visuoauditory letter presentation, after which effort of the 1-back task was rated on a 1-4 scale during 3.5 seconds.

9

Before coming to the lab to do the fMRI task, each participant filled out the Compass of Shame Scale questionnaire (Nathanson, 1992). In this questionnaire, participants answer questions about their responses to certain (shameful) situations on a scale from 1 to 5 (never, almost never, sometimes, often and almost always). One example is the following situation: ‘You make a mistake in the presence of others. What is your response?’, followed by 10 possible responses (e.g. ‘I pretend that I don’t care’, or ‘I am ashamed’) that are rated on the scale. After rating the 40 response items on the scale, the responses are divided into six subscales: Shame Proneness (an example response to the previous example situation: ‘I’m ashamed’), Adaptation (e.g. ‘this did not go very well, but next time I will do better’), and the four types of coping styles: Denial (e.g. ‘I pretend nothing is going on’), Aggression (e.g. ‘I blame others’), Devaluation (e.g. ‘I think I am stupid’) and Avoidance (e.g. ‘I would prefer to run away’). Of these four coping styles, two types are externalizing coping styles (Denial and Aggression), and two are internalizing coping styles (Devaluation, Avoidance).

Skin Conductance – Preprocessing

Preprocessing of the skin conductance response was done using Matlab (Mathworks, 2010). The skin conductance signal was first processed using the FMRIB FASTR (fMRI Artefact Slice Template Removal; Niazy et al., 2004) plug-in for EEGLAB which removes MRI gradient artefacts in EEG and physiology data. The skin conductance signal was recorded at a sampling frequency of 5000 Hz (to ensure optimal FASTR processing), with a low-pass filter of 250 Hz and a DC high-pass filter. For the purpose of this study, the data was down sampled to 250 Hz and a notch filter of 50 Hz was applied to filter out power line noise. The skin conductance response filter settings included a low-pass filter of 0.5 Hz and a high-pass filter of 3 Hz. Lastly, the skin conductance responses were epoched according to stimulus onset and duration times and loaded into a Matlab cell array structure to be used in further analysis stages.

MRI Acquisition

The structural (T1-weighted) and functional (T2*-weighted echo planar imaging (EPI)) Magnetic Resonance Imaging (MRI) data was acquired in the Spinoza Centre for Neuroimaging in Amsterdam, using a Philips 3 Tesla Achieva MRI scanner. The structural scan was collected for each participant in an earlier recording session about a week before the fMRI session (TR = 8.194 ms, TE = 3.77 ms, flip angle = 8 degrees, Field Of View (FOV) = 240 x 220 x 188 mm (ap, fh, rl), slice thickness = 1.0 mm, inter-slice gap = 0.0 mm). Each experimental run consisted of 212 EPI volumes (2.5 × 2.5 mm voxels in plane; 2.5 mm slice thickness with a 10% inter-slice gap; 240 × 240 mm in-plane FOV; 42 slices; TE = 27.63 ms; SENSE factor = 2; transverse slices with phase encoding in the R-L direction, bottom-top sequential slice acquisition).

10

Preprocessing was performed using FEAT (FMRIB’s Expert Analysis Tool, version 6.00), a tool from the FSL toolbox (FMRIB Software Library, v5.0 of the Analysis Group, FMRIB, Oxford, UK; Jenkinson et al., 2012; Smith et al., 2004; Woolrich et al., 2009). Brain extraction of the T1 image was optimized using Matlab and Fieldtrip functions (Oosterveld et al., 2011) in the SPM Toolbox (Penny et al., 2011), by improving segmentation of grey and white matter (GM, WM) (probability threshold of 0.5) and adding the two segmented GM and WM brain mask images together. A Gaussian kernel of 4.7096 mm was used to smooth the data. A second brain extraction was done using the FSL Brain Extraction Tool (BET; Smith, 2002). Unwarping of B0 (direction: -x, or R-L) was done by first creating a fieldmap image and fieldmap magnitude image by rescaling and unwrapping the phase image to radians per second, and by creating and dilating a brain mask from the magnitude image. The signal loss threshold for unwarping was 10%. Brain extraction of the EPI images was also optimized in the same manner as the T1 image. No slice-time correction was applied since the experimental design is a robust block design and temporal derivatives were added to the general linear model. Motion correction was performed using MCFLIRT (based on optimization and registration techniques used in FLIRT (Jenkinson et al., 2002). Spatial smoothing was carried out on each EPI volume separately, to reduce noise without reducing valid activation (a 5 mm full-width-at-half-maximum (FWHM) Gaussian kernel was used). A high-pass temporal filter cut-off of 100 s was applied to filter low-frequency artefacts such as background noise and temperature fluctuations.

Using FLIRT (FMRIB’s Linear Image Registration Tool; Jenkinson et al., 2002; Jenkinson & Smith, 2001), the EPI images were linearly registered to the participant’s brain extracted structural image (DOF = 6). The T1 image was then non-linearly registered to standard space MNI template image (MNI152_T1_1mm) using FNIRT (FMRIB’s Non-Linear Image Registration Tool). Boundary-Based Registration (BBR) was included to map white-matter boundaries from T1 to the EPI image and optimise the linear registration by the use of the grey-white matter boundaries. Both registrations were combined in a single transformation to transform the EPI image into standard space, by applying the two concatenated transformation matrices at one.

Emotional Discrepancy Factor – Analysis

In order to compare the emotional reactivity rating data to the skin conductance response, and calculate an ‘Emotional Discrepancy Factor’ (EDF) between the two variables, a couple of processing steps were required. A dependent samples t-test was done on the emotional reactivity scores and reported. Next, the rating of emotional reactivity following the neutral and negative trials were z-scored. Since it was possible for the participant to respond to the question too late, some data points were missing. In one instance 4 observations were missing. The mean and standard deviations of the remaining

11

elements were used for the calculation of the z-score. Next, the subjective responses were normalized by subtracting the emotional reactivity responses from the neutral condition from the negative condition, and dividing the difference by the emotional reactivity responses from the neutral condition.

The skin conductance response (SCR) was calculated as the standard deviation of the electrodermal signal recording within the stimulus presentation and baseline. To investigate the amplitude and length of the variations in the SCR, the standard deviation of the raw skin conductance response for the emotional stimulus and the baseline (the 7-9 second fixation cross before the beginning of the stimulus presentation) were calculated. Next, the skin conductance response is calculated relative to the baseline preceding the emotional stimulus in that trial (the fixation cross presentation, see figure 1). Hence, the standard deviations for the skin conductance response during the baseline were subtracted from the standard deviations during the stimulus presentation, and divided by the standard deviation of the baseline. Because the median better approached the central tendency of the data than the mean (due a skewed distribution of the data, i.e. non-centrality), the median of the data was taken for the z-scoring of the data.

Finally, the EDF was calculated by subtracting the standardized mean emotional reactivity score from the standardized median SCR. Subtracting the two different indicators of emotional responses (one subjective, and one more objective measure) from each other in this way enables us to extract from the emotional discrepancy factor the degree of discrepancy between the two measures of emotional response (i.e. if the emotional discrepancy factor is high, this means the skin conductance response is relatively stronger than the subjective response).

First-level Analysis

In the first-level analysis, the whole brain BOLD responses were modelled for the presentation of the negative and neutral autobiographical stimuli using FSL FEAT (Jenkinson et al., 2012). A general linear model (GLM) was used to model the BOLD activations to the stimulus conditions and estimate the effects for each voxel. The predicted responses to the experimental stimuli were modelled through the use stimulus onset, event duration and magnitude of response information for each regressor. Combined, this information created the task-related regressors used in the regression. The regressors were subsequently convolved with a double-gamma hemodynamic response function. To assess the relative relevance of each regressor, contrasts between regressors are created with which parameter estimates (PE) could be compared (i.e. contrasts of parameter estimates; COPEs). The specified contrast evaluated the voxels where the negative stimuli elicited more activation than the neutral stimuli: the Neg>Neu contrast.

The analysis resulted in unthresholded z-statistic maps for each of the two runs, which were subsequently used in the second-level analysis.

12

For the second-level analysis, the z-statistic maps of the two runs of each session (i.e. the outputs of the two separate runs from the first level analysis on the Neg>Neu contrast) were averaged by transferring the resulting z-statistic maps to the FSL FEAT higher-level analysis, and averaging the mean for both runs into one mean z-statistic map. Since the experiment consisted of two runs in each session, the run-to-run variance could not be estimated well, and Fixed Effects (FE) higher-level modelling was chosen in which this variance is ignored. The resulting mean z-statistic map for the negative stimulus specific activations was used as input for the following higher-level analyses.

Third-level Analysis

In the third-level analysis, three separate general linear models were run, which are described per model in the following paragraphs.

Model 1: Assessing group mean negative stimulus specific activation

To investigate if there was a significant (z > 2.3; p < 0.5) activation in response to the negative stimuli relative to the neutral stimuli for all participants, a one sample t-test was performed on the second-level output for the Neg>Neu contrast from the first-level analysis. Analysis was carried out using a mixed effects model: FLAME (FMRIB’s Local Analysis of Mixed Effects) stage 1 with automatic outlier detection (Woolrich, 2008). The mixed effects analysis, contrary to the fixed effects analysis from the second-level analysis, models subject-to-subject variability and therefore allows inferences to be made about the wider population from which the subjects were drawn. FLAME estimates higher-level parameter estimates and ME variance using sophisticated estimation techniques (the higher-level model is fit using a fast approximation to the final estimation). FLAME also detects outlier data points (each subject’s data is considered in relation to the other subjects regarding whether each voxel appears to be an outlier) and then automatically de-weights the outliers in the multi-subject statistics. The same mixed effects model is used for all third-level analyses.

Model 2: Assessing associations between mean negative stimulus specific activations and EDF

In this model, the emotional discrepancy factor (EDF) is added to the model next to the stimulus specific activation to negative stimuli (the Neg>Neu contrast from the first level output). Correlations between stimulus specific activation to negative stimuli and the EDF are investigated by the use of a two-tailed t-test (z > 2.3; p = 0.05) in a mixed effects model. In order to investigate both positive and negative associations for the two regressors with the BOLD signal responses (negative stimulus specific activation), four different contrasts were specified (table 1).

13

Contrasts Group Mean EDF Group_mean_pos 1 0 Group_mean_neg -1 0

EDF_pos 0 1

EDF_neg 0 -1

Table 1. The exploratory variables used in the second third-level regression analysis and their relative contrasts in the

model. Pos = positive associations, neg = negative associations. EDF = emotional discrepancy factor. Model 3: Assessing associations between negative stimulus specific activation, EDF and coping styles

In the final third-level model, the six outcomes of the Compass of Shame Scale questionnaire are added to the model. These subscales are ‘Shame Proneness’, ‘Adaptation’, ‘Denial’, ‘Aggression’, ‘Devaluation’, and ‘Avoidance’. As with the previous third-level models, the mixed effects model was used to investigate if there are significant positive or negative associations (two-tailed t-test) between the group mean negative stimulus specific activation (Neg > Neu contrast), the EDF and the six outcomes from the Compass of Shame Scale questionnaire. Z-statistic images were thresholded using threshold-free cluster enhancement determined by z > 2.3 (Smith & Nichols, 2009). In order to investigate both positive and negative associations for each of the eight regressors with the BOLD signal responses, 16 different contrasts were specified (a positive and negative contrast for each of the 8 regressors).

To localize the coordinates of significant peak voxels in each cluster, the statistical z-maps from all third-level analyses were overlaid onto the MNI152 anatomical template and Harvard-Oxford probabilistic atlases (both the cortical and subcortical structures atlas; Desikan et al., 2006) were used. The Harvard-Oxford atlases were used to report the coordinates of significant peak voxels, because they take inter-individual variability into account and therefore provide reliable labels (Poldrack & Nichols, 2011). Since the peak voxel can occasionally be found on the edge of a significant cluster, labeling of that peak voxel alone does not always give an accurate location of a significant cluster. Therefore, in addition to the Harvard-Oxford atlases, anatomy literature is used to accurately label brain areas (Sobotta, 2006). For each contrast, cluster label(s), cluster significance, maximum z-statistic value in that cluster and peak voxel coordinates are mentioned from the largest cluster to the smallest cluster in the results. Additional cluster information is reported in appendix II.

14

Emotional Reactivity – Subjective Responses

There was a significant difference (t(35) = -8.92, p < 0.001) between the emotional reactivity responses between the negative trials and the neutral trials for all participants.

Assessing negative stimulus specific activations for all participants

There were significant activations for all participants on average for the negative stimuli compared to the neutral stimuli in multiple areas: the right Lateral Occipital Cortex (p < 10-5_{; Z}

-Max = 5.45; Xz-max = 40, Yz-max = -57, Zz-max = 56), the left

Lateral Occipital Cortex (superior division) (p < 0.001; Z-Max = 4.63; Xz-max = -29, Yz-max = -65, Zz-max = 49), the left Middle

and Inferior Temporal Gyrus (temporooccipital part) (p < 0.01; Z-Max = 4.92; Xz-max = -59, Yz-max = -46, Zz-max = -11) (see

figure 2), the right Middle Frontal Gyrus (identified as a significant cluster twice, a more inferior cluster: p < 0.01; Z-Max =

4.68; Xz-max = 52, Yz-max = 32, Zz-max = 25, and a superior cluster: p < 0.05; Z-Max = 4.67; Xz-max = 40, Yz-max = 15, Zz-max =

55). More information for the significant clusters and cluster sizes for all results are included in the appendix (Appendix II, table 1).

Figure 2. Significant increase in the Middle Frontal Gyrus for negative stimulus specific activation for all participants. A horizontal section of the thresholded z-map for the Neg>Neu contrast overlaid onto the MNI152

anatomical template, showing the peak significant voxel (x = 40, y = 15, z = 55) in the (superior division) Middle Frontal Gyrus (crosshair) and the cluster it lies in (green circle; also Inferior Frontal Gyrus). The other two significant activations in the image are the left and right Lateral Occipital Cortices. A = Anterior, L = Left, P = Posterior, R = Right. Labelling according to the (sub)cortical Harvard-Oxford probability atlases.

Assessing associations between negative stimulus specific activation and EDF

For the second hypothesis, in which associations between the mean negative stimulus specific activation and the emotional discrepancy factor (EDF) are tested, there are several significant positive and negative associations between the group mean and the EDF. Significant clusters of associations were found for three out of four contrasts.

15

Significant positive associations were found for the group mean and the negative stimulus specific activations (the Neg>Neu contrast), very similar to the results found for the first hypothesis: the right Lateral Occipital Cortex (superior division) (p < 10-5_{; Z}

-Max = 5.07; Xz-max = 41, Yz-max = -56, Zz-max = 57), the left Lateral Occipital Cortex (superior division)

(p < 0.01; Z-Max = 4.28, Xz-max = -29, Yz-max = -65, Zz-max = 49), the left Middle Temporal Gyrus (temporooccipital part) (p

< 0.01; Z-Max = 4.76; Xz-max = -59, Yz-max = -45, Zz-max = -12) and the right Middle Frontal Gyrus (two clusters: one closer

to the frontal lobe with p < 0.05; Z-Max = 4.78; Xz-max = 52, Yz-max = 32, Zz-max = 25, and a second cluster with p < 0.05; Z -Max = 4.45; Xz-max = 40, Yz-max = 14, Zz-max = 55).

Contrary to the previous model, the second regression model included negative contrasts, including the negative associations between the group mean and the negative stimulus specific activations. For this contrast, significant negative associations were found for the group mean: a very large cluster (see figure 3) in the right Juxtapositional Lobule Cortex that also spans across the right Superior Frontal Gyrus, the Cingulate Gyrus (anterior division), the Paracingulate Gyrus, the right Caudate and the left and right Thalamus (p < 10-5_{; Z}

-Max = 4.3; Xz-max = 2, Yz-max = -4, Zz-max = 62), the left

Supramarginal Gyrus (anterior division) and the Postcentral Gyrus (p < 10-5_{; Z}

-Max = 4.83; Xz-max = -67, Yz-max = -31, Zz-max

= 23), the Superior and Middle Temporal Gyrus (posterior division) and the Insular Cortex (p < 10-5_{; Z}

-Max = 4.66; Xz-max

= 50, Yz-max = -18, Zz-max = -7), the Cerebellum (posterior) (p < 10-5; Z-Max = 5.16; Xz-max = 26, Yz-max = -78, Zz-max = -34),

the right superior Frontal Gyrus and Pole (p < 0.01; Z-Max = 4.11; Xz-max = -25, Yz-max = 38, Zz-max = 24) and the right

Supramarginal Gyrus (anterior division) and Parietal Operculum Cortex (p < 0.05; Z-Max = 4.05; Xz-max = 65, Yz-max = -33,

Zz-max = 29).

Figure 3. Significant negative associations (blue) between the group mean and negative stimulus specific activations.

A mid-sagittal section of the thresholded z-map overlaid onto the MNI152 anatomical template, showing the peak significant voxel (x = 2, y = -4, z = 62) in the right Juxtapositional Lobule Cortex (crosshair) and the cluster it lies in (green circle). The image also shows the cluster ranging to the Superior Frontal Gyrus, the anterior Cingulate Gyrus, and the Thalamus, but the peak significant voxel of this cluster is found in the Juxtapositional Lobule Cortex. The other clusters in

16

the image are (from left to right): the cerebellum, the right thalamus (part of a larger cluster) and the right superior frontal gyrus and pole. Labelling according to the (sub)cortical Harvard-Oxford probability atlases.

Significant positive associations were found between several brain areas and the EDF on the negative stimulus specific activations, of which peak voxels in the Cerebellum (posterior) (p < 10-5_{; Z}

-Max = 4; Xz-max = 44, Yz-max = -68, Z z-max = -34), the right Middle Temporal Gyrus (temporooccipital part) (p < 0.05; Z-Max = 4.09; Xz-max = 56, Yz-max = -49, Z z-max = -4) and the left Occipital Pole, the Lateral Occipital Cortex, the right Cuneal Cortex and the Precuneus Cortex (p <

0.05; Z-Max = 3.89; Xz-max = -13, Yz-max = -95, Zz-max = 23).

No significant negative associations were found between negative stimulus specific activations and the regressor EDF.

Assessing associations between negative stimulus specific activations, EDF and coping styles

In the final regression model, the positive and negative associations between the group mean, the EDF and the 6 outcomes of the Compass of Shame Scale questionnaire (shame proneness, adaptation, and the coping styles denial, aggression, devaluation and avoidance) for all participants on the mean negative stimulus specific activation (contrast Neg>Neu) are assessed.

The first four contrasts of this model show very similar clusters of activation (see table 2) as seen in the previous model (positive and negative group mean associations, positive and negative EDF associations), with some differences. First of all, a negative contrast was added for the group mean, investigating significant deactivations in the shameful condition. Secondly, the very large cluster seen in the negative associations between the mean negative stimulus specific activations and the EDF (contrast 2) around the Juxtapositional Lobule Cortex and the Cingulate Cortex is split in two smaller clusters. Only the Cerebellum (posterior) (p < 0.01; Z-max = 3.74) remains significant for the positive associations

between EDF and the negative stimulus specific activations.

Peak Voxel Co-ordinates (in mm)

Contrast Brain Area p-value Z-max X Y Z

Group_mean pos

Lateral Occipital Cortex (Superior Division)

< 10-5 _{4.53 42 -59 59}

Lateral Occipital Cortex (Superior Division)

< 10-5 _{4.64 -29 -75 34}

Middle Frontal Gyrus 0.0016 4.5 40 15 56 Middle and Inferior

Temporal Gyrus (Temporooccipital part)

0.00278 4.47 -59 -45 -11 Middle Frontal Gyrus 0.00337 4.57 55 22 33 Group_mean

neg

17

Caudate, Cingulate Gyrus (Anterior Division) < 10-5 _{4.43 22 25 9} Supramarginal Gyrus (Anterior Division), Postcentral Gyrus < 10-5 _{4.6 -68 -22 26} Juxtapositional Lobule Cortex, Superior Frontal Gyrus, Cingulate Gyrus (Anterior Division)

< 10-5 _{4.38 2 -4 60}

Superior and Middle Temporal Gyrus, Insular Cortex

< 10-5 _{4.39 50 -18 -7}

Supramarginal Gyrus and Parietal Opercular Cortex

0.0361 3.82 66 -33 27 Temporal Pole 0.0438 4.16 -54 3 -24 EDF_pos Cerebellum 0.00356 3.74 43 -68 -31

Table 2. Results associations between mean negative stimulus specific activation and EDF in regression model 3.

Contrast, Brain area label, cluster significance, maximum z-value in cluster and peak voxel coordinates per significant cluster for the first three contrasts in the third regression model (GLM3). Labelling done according to the (sub)cortical Harvard-Oxford probability atlases and Sobotta (2006).

There were significant positive associations for all participants on average between the regressor Shame Proneness and the mean negative stimuli specific activations in the right Precentral Gyrus and the right Superior Frontal Gyrus (p < 0.0001; Z-max = 4.14; Xz-max = 19, Yz-max = -12, Zz-max = 65). In addition, significant negative associations between Shame

Proneness and the mean negative stimulus specific activations were found: the Lateral Occipital Cortex (superior division) (p < 0.01; Z-max = 4.08; Xz-max = -37, Yz-max = -87, Zz-max = 39) and the right Frontal Pole (p < 0.05; Z-max = 3.76; Xz-max =

26, Yz-max = 49, Zz-max = -12) (see figure 4).

Figure 4. Significant positive (red) and negative (blue) associations between shame proneness and the mean negative stimuli specific activations. A mid-sagittal section of the thresholded z-map overlaid onto the MNI152 anatomical

template, showing the significant cluster for the positive associations: the Superior Frontal Gyrus and the Precentral Gyrus (p < 0.000; Z-max = 4.14; in the green circle) and the surrounding cluster (red). The cluster shown for the significant negative

18

There were no significant positive associations between Adaptation and the negative stimulus specific activations. However, significant negative associations were found for the right Middle Frontal Gyrus (p < 0.05; Z-max = 3.71; Xz-max =

42, Yz-max = 11, Zz-max = 45).

Significant positive associations were found in multiple locations between the coping style Denial and the mean negative stimulus specific activations (see figure 5): the largest cluster includes the left Juxtapositional Lobule Cortex, the Cingulate Gyrus (anterior division), the Paracingulate Gyrus and the Superior Frontal Gyrus (p < 10-5_{; Z}

-max = 4.08; Xz-max

= -2, Yz-max = 2, Zz-max = 64), the left Lateral Occipital Cortex (superior division) (p < 0.001; Z-max = 4.45; Xz-max = -44, Y z-max = -63, Zz-max = 50), the right Superior Temporal Gyrus (posterior division) (p < 0.001; Z-max = 4.22; Xz-max = 67, Yz-max

= -34, Zz-max = 16), the left Superior Frontal Gyrus (p < 0.05; Z-max = 3.98; Xz-max = -26, Yz-max = 18, Zz-max = 59) and the

right Inferior Frontal Gyrus (p < 0.05; Z-max = 3.55; Xz-max = 58, Yz-max = 18, Zz-max = 4). Significant negative associations

between Denial and the mean negative stimulus specific activations were found in the Cingulate Gyrus (posterior division) (p < 0.01; Z-max = 3.61; Xz-max = -8, Yz-max = -25, Zz-max = 22).

Figure 5. Significant positive (red) and negative (blue) associations between the coping style denial and the mean negative stimuli specific activations. A mid-sagittal section of the thresholded z-map overlaid onto the MNI152

anatomical template, showing the largest significant cluster for the positive associations (from top to bottom: the Superior Frontal Gyrus, the Paracingulate Gyrus and the anterior Cingulate Gyrus (p < 10-5_{; Z}

-max = 4.08)) and the significant cluster

for the negative associations between Denial and the mean negative stimuli specific activations: the posterior Cingulate Gyrus (p < 0.01; Z-max = 3.61). S = Superior, A = Anterior, I = Inferior, P = Posterior.

On the sixth regressor, Aggression, no significant positive associations were found between the coping style Aggression and the mean negative stimuli specific activation. However, significant negative associations were found in multiple areas: the right Precentral Gyrus (p < 10-5_{; Z}

-max = 4.07; Xz-max = 58; Yz-max = 9; Zz-max = 8), the right Superior

Frontal Gyrus, the left and right Juxtapositional Lobule Cortex, the anterior Cingulate Gyrus and the right Paracingulate Gyrus (p < 10-5_{; Z}

-max = 3.85; Xz-max = 1; Yz-max = 37; Zz-max = 54), the left Angular Gyrus (p < 10-5; Z-max = 4.57; Xz-max =

19

left Insular Cortex (p < 0.01; Z-max = 3.87; Xz-max = -33; Yz-max = -14; Zz-max = 9) and the left Precuneus Cortex (p < 0.01; Z -max = 3.92; Xz-max = -4; Yz-max = -61; Zz-max = 30) (see figure 6).

Figure 6. Significant negative associations in the precuneus cortex between the coping style aggression and the mean negative stimuli specific activations. A mid-sagittal section of the thresholded z-map overlaid onto the MNI152

anatomical template, showing two largest significant clusters: the left Precuneus Cortex (p < 0.01; Z-max = 3.9; crosshair)

and the second cluster here showing the Juxtapositional Lobule Cortex, the Superior Frontal Gyrus and the Paracingulate gyrus (p < 10-5_{; Z}

-max = 3.85). S = Superior, A = Anterior, I = Inferior, P = Posterior.

No significant positive or negative associations were found for the coping style Devaluation between Devaluation and the negative stimuli specific activation.

On the eighth and final regressor, the coping style Avoidance, no significant positive associations were found between avoidance and the mean negative stimuli specific activation. There were significant negative associations between several locations and Avoidance and the mean negative stimuli specific activation. The cluster included the left Lingual Gyrus and the left and right Precuneus Cortex and the Cuneal Cortex (p < 10-5_{; Z}

-max = 4.84; Xz-max = -13, Yz-max = -79, Z z-max = -13), the right Middle Frontal Gyrus (p < 10-5; Z-max = 4.74; Xz-max = 34, Yz-max = 1, Zz-max = 65) and the left Lateral

Ventricle (p < 10-5_{; Z}

20

The present study investigated the brain activation patterns indicative of repressive coping behaviour. The BOLD-signal response, skin conductance response during, and emotional reactivity response following a shame-eliciting task have been studied to examine if discrepancies between emotional reactivity responses and skin conductance responses (i.e. the emotional discrepancy factor, EDF: measure of repression) can provide new insights into the neurobiological underpinnings of the repression of shame, and can be further clarified by subsequent coping behaviour. To answer this main hypothesis, we first discuss whether there are significant BOLD-signal responses in brain areas involved in the experience of shame following the negative compared to the neutral stimuli. In addition, the whole-brain analysis of the associations between the EDF and the mean negative stimulus specific activations are discussed. Next, the brain areas associated with the different coping styles are discussed per coping style. Concluding the discussion is a summary of the main findings, some potential methodological improvements are discussed and suggestions for future research are made. According to the first hypothesis, a significant increase in activation in response to the shameful stimuli compared to the neutral stimuli in the brain areas involved in the experience of shame is expected. The main findings are shameful stimuli-specific activations in the lateral occipital cortices, the left middle and inferior temporal gyrus and the right middle frontal gyrus. The lateral occipital cortex, besides playing a role in object recognition (Malach et al., 1995; Grill-Spector et al., 2001), is also involved in the processing of information about objects and emotions (Gläscher et al., 2007). The autobiographical memory task was presented to the participant through the use of four key words describing the shameful experience; no pictures were present. A possible explanation for this unexpected finding is then that due to the role of strong shameful emotions in the negative condition, stronger emotion-related visual processing occurred. The left middle and inferior temporal gyri play a role in semantic memory and language processing, mostly by accessing word meaning while reading (Acheson & Hagoort, 2013). Finding activations in the middle and inferior temporal gyri is unexpected and might also be explained by stronger semantic interpretation of the stimuli for emotional autobiographical memories. The left middle frontal gyrus has been shown to play a role in the reorienting of attention (Japee et al., 2015), which could indicate greater attention for emotionally-loaded stimuli. Interestingly, there is an increase in activation in the middle frontal gyrus, which can be caused by a regulatory process involving the suppression of emotional memories. Depeu et al. (2007) found evidence that emotional memories are suppressed via two distinct neural mechanisms; an initial suppression by the right inferior frontal gyrus over regions involved in sensory components of the memory (i.e. visual cortex, thalamus), followed by the right medial frontal gyrus control over emotional components of the memory representation (i.e. hippocampus, amygdala). Both mechanisms are influenced by fronto-polar regions, indicating that emotional memory suppression is under control of prefrontal regions. Although finding significant activations in the middle temporal gyrus

21

can be considered unexpected because the participants are explicitly instructed to focus on retrieving the emotional memory, they are not entirely surprising. Finding a brain area that suggest that a significant degree of suppression of emotional memories takes place for the shameful condition across all participants is entirely in line with the other hypotheses in this study.

Following previous research about the neurobiological underpinnings of shame, it is hypothesized that the anterior cingulate cortex, the parahippocampal cortex and the medial and frontal inferior gyrus show activations (Michl et al., 2014). Moreover, the visceral sensation of shame is known to activate the insular cortex, the anterior cingulate cortex and the prefrontal cortices (Aziz et al., 2000; Singer et al., 2009) and the arousal-dimension modulation of emotional memories activates amygdala-hippocampal networks (Dolcos et al., 2004; Kensinger and Corkin, 2004). The failure to find either of these areas associated with shame in our results does not agree with previous research into self-conscious emotions and emotional memory processing. One explanation for the failure to find any of the before mentioned activations can be that there is a trend towards these areas but this trend does not reach significance with the current z-statistic threshold of > 2.3. Unthresholded z-statistic images do show activations in the right posterior and anterior cingulate cortex, the left and right parahippocampal cortices and the left and right frontal pole but these do not reach significance. Another possibility is that the manipulation of shame in this study (the retrieval of autobiographical shameful memories) failed. However, subjective emotional reactivity responses do indicate a significant difference between the negative and neutral stimuli.

Emotional Discrepancy Factor (EDF): Measure of Repression

In the second regression model the positive and negative associations between the BOLD-signal and the group mean negative stimulus specific activations were tested, with the addition of positive and negative associations for the EDF. Although it is important to keep in mind that this is a separate regression model, it is interesting to shortly discuss some of the negative associations between certain brain areas and the mean negative stimulus specific activations. The second regression model does not directly answer what brain areas deactivate during the experience of shame, because the EDF is also included in this model as a regressor and therefore influences the results for the negative group mean contrast. To limit the discussion of this contrast, only brain areas related to the experience of shame are discussed. A variety of significant deactivations were found in a collection of clusters, including brain areas of interest like the anterior cingulate gyrus, the supramarginal gyrus, the insular cortex, the superior frontal gyrus and the cerebellum. Although increases in the anterior cingulate gyrus and the superior frontal gyrus are expected according to previous research (Michl et al., 2014), our results show deactivations in these areas. Similarly, insula activation was expected but the results showed a deactivation in this area. The insula is involved in a variety of tasks, but most relevant is its function as the area where visceral sensations of

22

emotions occur (Singer et al., 2009). These findings are unexpected, and might place some question marks on whether the shameful manipulation was successful. The supramarginal gyrus plays a crucial role in overcoming an emotional egocentricity bias in social judgements (Silani et al., 2013). A decrease in the supramarginal gyrus might suggest that the autobiographical memories contain social context judgements remembered from a very egocentric, subjective experience point of view. The deactivation found in the cerebellum for this contrast is discussed in a separate paragraph.

For our second hypothesis, we expect an association between the mean negative stimulus specific activations and the EDF. The EDF is hypothesized as a valid measure of emotional response repression, and we expect participants that have a higher EDF to show more associations with brain areas related to the repression of emotions. To our knowledge, only two fMRI studies previously investigated repression, and little is known about the brain areas active during repression of shameful or negative emotions. Previous research points to the role of frontal brain areas (the orbitofrontal cortex, ventromedial prefrontal cortex) in the top-down regulation of emotional responses (Tomarken and Davidson, 1994; Kline et al., 1998; Sander et al., 2003 and Rauch et al., 2007). Significant positive associations between the EDF and the mean negative stimulus specific activations were found in the cerebellum, the middle temporal gyrus, the occipital pole, the cuneal cortex and the precuneus cortex. The activation found in the cerebellum will be discussed in a separate paragraph further in the discussion. The middle temporal gyrus plays a role in accessing word meaning while reading (Acheson & Hagoort, 2013) and will not be discussed further. Rauch et al. (2007) also found enhanced responsivity in temporo-occipital visual systems, which might indicate greater reliance on posterior visual processing for stimulus analysis by repressors. This finding is supported by the negative associations found between the EDF and negative stimulus specific activations in the occipital pole. The cuneal cortex is known for its role in basic visual processing. In addition, gray matter volume in the cuneus is associated with better inhibitory control in bipolar depression patients (Haldane et al., 2008), but pathological gamblers show higher activity in the cuneus relative to controls (Crockford et al., 2005). The relation between the cuneal cortex and inhibitory control of behaviour is not entirely clear, and more research about the role of the cuneal cortex in emotion processing will benefit the interpretation of this finding. The precuneus plays a central role in a wide spectrum of tasks, including visuo-spatial imagery, episodic memory retrieval and self-processing operations (i.e. first-person perspective taking) (Cavanna and Trimble, 2006). In addition, the review by Cavanna and Trimble also suggests that the precuneus is involved in the network of the neural correlates of self-consciousness, and is engaged in self-related mental representations during rest. The positive association of the precuneus with the EDF is unexpected, because it might suggest increased episodic memory retrieval, first-person perspective taking or even increased self-consciousness for repressors. But because of the variety of functions of the precuneus it is not currently clear what the relation is between the precuneus and repression, and a future region of interest analysis for the precuneus might further resolve this unexpected result.

23

The cerebellum has in the recent years become more and more known for its role in emotional cognition (Bastian, 2011; Schmahmann, 2010; Stoodley and Schmahmann, 2009; Baumann & Mattingley, 2012), and shows activation more often found in experiments with negative-emotion conditions than positive-emotion conditions (Strick et al., 2009). An fMRI study by Bermpohl et al. (2006) found that emotional picture perception activated the cerebellum. Similarly, another fMRI study investigated aversion-related responses in the cerebellum to noxious heat and unpleasant images, and found activations in overlapping areas in the posterior cerebellum. Additionally, cerebellar areas that showed functional overlap with both heat pain and unpleasant picture viewing were significantly inversely correlated with activations found in limbic structures, including the anterior cingulate cortex and the parahippocampal gyrus (Moulton et al., 2011). One theory following these studies is that the cerebellum is involved in the perceptive components of emotional stimulus processing, but this would not entirely explain the activations that are found following heat pain. A more likely explanation is in line with our results, namely that the cerebellum plays a key role in shameful emotion processing. A significant negative association between the cerebellum and the mean activations for the shameful condition indicate less cerebellar activity during the experience of shame. This finding indicates that over all participants, the cerebellum decreases in activation during the shameful condition. Contrary to this finding, there is a significant positive association between the cerebellum and the EDF, indicating a regulatory role of the cerebellum in the suppression of negative emotion processing. More evidence for a role of the cerebellum in negative emotion processing can be found in neuroanatomy and in clinical studies. The neuroanatomy of the cerebellum interconnects with the limbic system, along with associated frontal and parietal areas (Middleton and Strick, 1994; Schmahmann and Pandyat, 1997; Kelly and Strick, 2003). Moreover, lesions of the cerebellum lead to high prevalence of negative affect states (Schmahmann and Sherman, 1998; Wolf et al., 2009), and patients with mood and anxiety disorders consistently display prominent dysfunctions in the cerebellum (Andreasen et al., 1998; Brambilla et al., 2002; Liu et al., 2010; Nakao et al., 2011). The current findings are in line with previous research and confirms the regulatory role of the cerebellum in emotion processing.

Coping styles

When people often experience shame, they develop particular types of coping behaviour to respond to the negative experience (Harper, 2011). In order to further clarify the relations between the repression of shame, shame proneness, adaptive responses to shameful experiences and different coping styles, the associations between the negative stimulus specific activations, the EDF and coping styles were tested in the final regression model. The current analysis does not allow for a full comparison of the relative influence of each individual Compass of Shame Scale outcome, since there was

24

only one model that included all coping styles. To limit the discussion, only brain regions related to shame or emotion regulation were discussed in detail.

Shame Proneness

The amount of shame proneness of an individual is measures in the Compass of Shame Scale through how strongly a person identifies with the response ‘I’m ashamed’ to shameful situations. This requires a certain amount of self-reflection and insight into emotional experiences, something certainly not common when employing a repressive coping style. In fact, it is likely that repressors score low on shame proneness due to lack this lack of meta-cognition. Positive associations between the mean negative stimulus specific activations and shame proneness are found in the precentral gyrus and the right superior frontal gyrus. The precentral gyrus is generally known as the site of the primary motor cortex, but is also involved in motor imagery (mental rehearsal of motor acts without body movement) (Porro et al., 1996). This finding indicates that shame-prone individuals have stronger activations in motor imagery areas, suggesting more active (motor) imagination. A positive association was also found between shame proneness and the mean negative stimulus specific activations in the superior frontal gyrus (SFG). The SFG is involved in a variety of cognitive and motor control tasks (Li et al., 2013). Moreover, a study investigating the neural correlates of conscious self-regulation of emotion (Beauregard et al., 2001) found activation in the right SFG (and the right anterior cingulate gyrus) when participants attempted to inhibit sexual arousal generated by viewing erotic stimuli. This finding indicates that emotional self-regulation is implemented by prefrontal regions and is stronger for shame prone individuals.

Associations between the mean negative stimulus specific deactivations and shame proneness are found in the lateral occipital cortex and the frontal pole. Previous studies show the role of the inferior frontal pole in automatic emotion regulation (Jackson et al., 2003; Cunningham et al., 2004; Hynes et al., 2006) and the suppression of sensory components of emotional memories (Depeu et al., 2007). Furthermore, the inferior frontal gyrus co-activates with the amygdala and hippocampus during autobiographical memory retrieval (Greenberg et al., 2005). It is possible that shame prone individuals show deactivations in the frontal pole due to less automatic emotion regulation when retrieving shameful autobiographical memories.

Adaptation

Although many coping styles in response to shame can be dysfunctional, individuals can also employ adaptive coping mechanisms. A negative association was found between the negative stimulus specific activations and adaptation in the middle frontal gyrus. As explained in more detail in a previous paragraph, the middle frontal gyrus plays a role in regulatory processes involving the suppression of emotional memories (Depeu et al., 2007). The association between adaptation and

25

the deactivation in this prefrontal area suggests that individuals that employ adaptive coping mechanisms show significantly less emotional memory suppression. Interestingly, this is the first instance of a study providing compelling evidence for the role of the middle frontal gyrus in the suppression of emotional memories through deactivations in this area that are significantly associated with the use of an adaptive coping style for shameful memories.

Denial

One of the coping styles that is closely related to repression is denial. When responding to shame with denial, a person typically does not acknowledge the negative experience of self and does not accept the message of shame as valid. Overall, there is little awareness of shame or shameful actions, and the aim of the coping style is to minimize the conscious experience of shame (Elison et al., 2006). When investigating the coping style denial, many positive and negative associations between denial and negative stimulus specific activations were found. Positive associations were found in the juxtapostional lobule cortex, anterior cingulate gyrus, the paracingulate gyrus, the superior and inferior frontal gyrus, the lateral occipital cortex and the superior temporal gyrus. The juxtapositional lobule cortex (or supplementary motor area) is known for its control of postural stability, coordinating temporal sequences of actions, bimanual coordination and internally generated movement (reviews: Goldberg, 1985; Nachev et al., 2008). This finding is unexpected, and no clear explanation is found for the association between denial and the negative stimulus specific activation of the juxtapositional lobule cortex. The superior temporal gyrus contains the primary auditory cortex and is responsible for processing sounds and the comprehension of language. In addition, studies have found a role for the superior temporal gyrus in social cognition (Jou et al., 2010) and social perception (Pelphrey and Carter, 2008; Redcay, 2008). Although this finding is also unexpected, it is possible that the superior temporal gyrus is associated with denial due to this relation to social cognition. One would expect however that the adaptation of the coping style denial in response to shame does not result in increased social cognition or perception. The varied functional properties of the superior temporal gyrus, combined with the unexpected positive (instead of negative) association of the superior temporal gyrus with denial, makes this result difficult to interpret.

Positive associations between denial and the shameful condition on different brain areas that are related to the experience and sensations of shame (i.e. the anterior cingulate gyrus, the paracingulate gyrus and the frontal gyrus) are found. The anterior cingulate cortex is known for its activation in response to shame (Michl et al., 2014), the cognitive and affective components of the visceral sensation of shame (Aziz et al., 2000) and for its role in generating and coordinating emotional responses through connections to areas that generate behaviour and autonomic responses (Bush et al., 2000; Sturm et al., 2013). Even though one might assume that individuals that deny shameful emotions also do not experience the visceral sensations of shame, both previous research and the current study confirms this is not the case. A denying

26

coping response to shame is most likely out of all coping styles to operate outside of conscious thought, but this does not diminish the autonomic experience of emotion. Research about repressors confirms the continued physiological response to shame while cognition and understanding of shame might be minimal (Lambie and Baker, 2003; Myers, 2010). Our finding further confirms this theory of discrepant emotional responses for repressors. The anterior paracingulate cortex is a key region attending to theory of mind (Gallagher & Frith, 2003) and plays a role in the understanding of intentions in social interaction (Walter et al., 2004). Finding activations in the paracingulate gyrus for the coping style denial is unexpected, and could mean that there is an additional role for the paracingulate gyrus in coping with shame. In addition, finding significant associations on both the superior and inferior frontal gyrus for the coping style denial and the negative stimulus specific activations provides additional evidence of the role of frontal brain regions in the regulation of emotional responses (Tomarken and Davidson, 1994; Kline et al., 1998; Sander et al., 2003 and Rauch et al., 2007). Furthermore, because of the close similarities between the repression of shame and the use of the coping style denial in response to shame, these findings further illuminate the neurobiological underpinnings of repression.

The posterior cingulate gyrus is also significantly negatively associated with the negative stimulus specific activations and the coping style denial (see figure 5). Previous research suggests a role for the posterior cingulate gyrus in autobiographical memory retrieval (Maddock et al., 2001) and the monitoring of sensory events (Vogt et al., 1992). The connections between the posterior cingulate and the Parahippocampal cortices likely contribute to its role in memory. It is possible that individuals that employ the coping style denial show deactivation in this region because of the repression of emotional elements of the autobiographical memory, but more detailed investigation into the coping style denial will be necessary to answer this question.

Aggression

Aggression is an externalizing coping style, in which the individual responds to shame by becoming aggressive towards others and blaming others for a shameful situation. When employing an aggressive coping style (also often referred to as ‘attack other’), an individual typically does not accept shame’s message, and responds to the emotional experience by making someone else feel worse (Elison et al., 2006). No positive associations between aggression and certain brain areas were found for the negative stimulus specific activations. However, a large number of negative associations of certain brain areas between aggression and negative stimulus specific activations were found. To limit this discussion to brain areas related to shame and emotion processing, we will not further discuss the juxtapositional lobule cortex, the precentral gyrus and the precuneus cortex.