Unravelling the Role of Confidence and Negative Outcome Anticipation in
Obsessive-Compulsive Disorder
Student: Maura Fraikin Student number: 11642807 Supervisor: Monja Hoven
First Assessor: mw. M. (Monja) Hoven
Second Assessor: dhr. dr. Z.L. (Zainal) Haberham Submission Date: 19-06-2020
2
M. Fraikin1,2
1 University of Amsterdam, Amsterdam, The Netherlands
2 Department of Psychiatry, Amsterdam UMC, Amsterdam, The Netherlands Correspondence: Maura Fraikin (maurafraikin@gmail.com)
Abstract
Obsessive-compulsive disorder (OCD) is a psychiatric disorder characterised by time-consuming obsessions, compulsions or both. Previous studies have repeatedly found metacognitive confidence distortions in multiple cognitive domains in OCD patients. Additionally, from studies in healthy controls (HCs) the ventromedial prefrontal cortex (VMPFC) emerges as an area involved in the computation of confidence. However, no studies have looked into neurobiological basis of confidence abnormalities in OCD. Therefore, we performed a functional magnetic resonance imaging using a perceptual decision-making task to look into the neurobiological basis of confidence abnormalities in OCD.
Also, earlier research has shown that negative outcome expectancy (NOE) increases ineffective compulsive checking in HCs, indicating that an increased sensitivity to negative outcome anticipation might be related to the development of compulsive behaviours. However, the role of negative outcome anticipation in compulsive behaviour in OCD patients has not been investigated yet. Therefore, this study used a NOE task to investigate if anticipating a negative outcome is related to compulsive checking and whether negative outcome anticipation affects compulsive behaviour differently in HCs and OCD patients. Additionally, we examined the relationship between compulsive checking and confidence.
We showed that OCD patients are underconfident compared to HCs. Furthermore, we found that the VMPFC encodes subjective confidence in HCs, but no group differences were found in confidence coding in this area. Also, symptom severity was not correlated to confidence. Additionally, we showed that individuals perform more compulsive checking when a possible negative outcome is expected, but we did not find increased overall compulsive checking in OCD patients or increased checking specifically when anticipating negative outcomes. Lastly, compulsive checking negatively related to confidence, but no relationship between compulsive checking and symptom severity was found.
3
1. Introduction
Obsessive-compulsive disorder (OCD) is a devastating psychiatric disorder that affects approximately 2 percent of the population and is characterised by obsessions, compulsions or both (Kessler et al., 2005; Sasson et al., 1997; Goodman, 1999). These obsessions and compulsions are time-consuming and interfere with daily life. Obsessions are persistent and intrusive thoughts, impulses or images that often cause anxiety and distress. Compulsions are often the repetitive actions that are performed to reduce the distress and anxiety associated with the obsessions (American Psychiatric Association, 2013). OCD is ranked among the ten disorders with the most disabling effects on a human life by the World Health Organisation (Murray and Lopez, 1996 cited in Roberts, 2019), highlighting the relevance to expand research on treating the symptoms patients suffer from.
Many studies have been investigating behavioural patterns typical for OCD. One possible underlying mechanism for the compulsive behaviour seen in OCD, is the anticipation of a possible negative outcome. Outcome anticipation entails mentally preparing for upcoming consequences and generates arousal, which in turn may provoke a strong drive to act. It is hypothesised that the expectation of a negative outcome may lead to the urge to perform a behaviour and the feeling of
having to act, a key aspect of compulsivity (Luigjes et al., 2015; Szechtman & Woody, 2004; Luigjes et
al., 2019). Both a high outcome magnitude (i.e. potentially high gains or losses) and an uncertain outcome influence outcome anticipation, since both would generate more arousal and may
therefore result in a stronger urge to perform a certain behaviour (Knutson & Greer, 2008; Luigjes et al., 2015). Luigjes et al. (2016) investigated this hypothesis and found that negative outcome
expectancy (NOE) (i.e. both a high outcome magnitude and an uncertain outcome) increases compulsive checking in healthy controls (HCs), even though it is ineffective and deviates from the overall aim of the task. Therefore, an increased sensitivity to negative outcome anticipation might be related to the development of compulsive behaviours in disorders such as OCD. Another aspect that is getting increasing attention in psychiatric disorders in general, are distortions in someone’s ability to estimate his or her confidence (Hoven et al., 2019; Rouault et al., 2017). Confidence can be described as a subjective feeling on the correctness of a choice, decision or idea given the evidence (Luttrell et al., 2013) and is an aspect of metacognition (Pouget, Drugowitsch & Kepecs, 2016). Metacognition refers to “cognition about cognition” or “thinking about thinking” and is a higher order thinking skill that can be defined as the knowledge and understanding of one’s own thoughts and cognition (Cambridge Dictionary, 2020; Fleming et al., 2012; Grimaldi, Lau & Basso, 2015). These confidence estimations are important in optimising one’s decisions (Ott et al., 2018). Importantly, distortions in confidence estimations can lead to pathological behaviour. For example, having too
4
little confidence in your actions can lead to compulsive checking behaviour, a feature typically seen in OCD (Fisher & Wells, 2008; Hoven et al., 2019). The first clinical study to indicate that confidence in memory plays a role in OCD was performed by McNally & Kohlbeck (1993). In their study, no
differences were found in memory recall between HCs and OCD patients, but patients with OCD were shown to have less confidence in their memories than HCs. Since then, several researchers have replicated these findings in multiple domains. Reductions in confidence in OCD patients compared to HCs have been found in memory (Tolin et al., 2001; Moritz & Jaeger, 2018; MacDonald et al., 1997), attention, perception (Hermans et al., 2008), general knowledge (Dar et al., 2000) and sensory modalities (Taylor & Purdon, 2016), whereas no group differences in performance were found. The reductions in memory confidence were also found to be associated with compulsive checking (MacDonald et al., 1997; Moritz & Jaeger, 2018; Cuttler et al., 2013). Importantly, Dar et al. (2000) reported a negative correlation between confidence and severity of symptoms as measured by the Yale-Brown Obsessive-Compulsive Scale (Y-BOCS). It is important to further investigate the
relationship between confidence and OCD, since it might reveal new targets for intervention. Metacognitive confidence is a relatively new research field and understanding the underlying neurobiological mechanisms is essential for a better understanding of individual differences in this ability and the aetiology of confidence distortions. Several studies have been investigating the neural signatures of confidence, both in humans and animals. Kepecs et al. (2008) investigated the
neurobiological basis of confidence in rats and found that the activity of the orbitofrontal cortex corresponded to differences in confidence. Later, Lak et al. (2014) reported similar results; they found that inactivation of the orbitofrontal cortex in rats led to distortions in confidence related behaviour, whilst decisions remained unaffected. Furthermore, in a human functional magnetic resonance imaging (fMRI) study, Rolls, Grabenhorst and Deco (2010) reported a linear relationship between ventromedial prefrontal cortex (VMPFC) activation and decision confidence. Additionally, De Martino et al. (2013) showed that the Blood-Oxygen-Level-Dependent (BOLD) signal in the VMPFC represented both decision confidence and value of a choice. Similarly, Lebreton et al. (2015) reported that next to subjective value, the VMPFC activity also encoded confidence whilst making a choice. Strikingly, even though the participants were not instructed to rate their confidence, this encoding of confidence in the VMPFC was still seen, indicating that the VMPFC automatically integrates
confidence. Similar to the proposed automatic encoding of confidence in the VMPFC, Gherman & Philiastides (2018) reported that early confidence signals are visible in the VMPFC in a perceptual decision making task before explicit confidence report. Further support for the role of the VMPFC in confidence estimations comes from lesion studies. Hebscher et al. (2016) showed that patients with lesions in the VMPFC were impaired in their confidence monitoring, indicating a causal relationship
5
between VMPFC activity and confidence. Interestingly, multiple neuroimaging studies have reported abnormalities in frontal brain regions in OCD patients that are also found to be involved in
confidence coding, including the VMPFC, anterior cingulate cortex (ACC) (Nakao et al., 2014; Stern et al., 2013; Yücel et al., 2007).
The aforementioned studies suggest that confidence distortions are present in multiple cognitive domains in OCD patients. Furthermore, from studies in HCs, the VMPFC and other medial areas of the prefrontal cortex emerge as areas involved in the computation of confidence. Finding links between the symptoms OCD patients experience and the underlying behavioural and
neurobiological basis is crucial and provides more insight into the clinical relevance of confidence in OCD (Ouellet-Courtois et al., 2018). Therefore, an fMRI study will be performed using a perceptual decision-making task, looking into the neurobiological basis of confidence abnormalities in OCD. Moreover, earlier research showed that anticipating a negative outcome increases ineffective compulsive checking in HCs, indicating that an increased sensitivity to negative outcome anticipation might be related to the development of compulsive behaviour (Luigjes et al., 2015). However, the role of negative outcome anticipation in compulsive behaviour in OCD patients has not been
investigated yet. Therefore, this study will investigate if anticipating a negative outcome is related to compulsive checking behaviour in both HCs and OCD patients and whether negative outcome
anticipation affects compulsive behaviour differently in both groups. Lastly, we will examine whether there is an association between compulsive checking behaviour and confidence.
We hypothesise that OCD patients are underconfident compared to HCs. Furthermore, we expect to find a positive relationship between BOLD activity in the VMPFC and confidence reports. Moreover, we hypothesise that confidence encoding in the VMPFC is distorted in OCD patients. With regards to checking behaviour, we hypothesise that both OCD patients and HCs show increased compulsive checking behaviour during NOE, where OCD patients show more checks overall, as well as specifically during NOE. Furthermore, since previous research has indicated a relationship between checking behaviour and confidence in OCD patients (MacDonald et al., 1997; Moritz & Jaeger, 2018; Hermans et al., 2008; Ouellet-Courtois et al., 2018), it is expected that compulsive checking is negatively correlated with confidence: subjects that check more will show a decrease in confidence level. Generally, we also expect that OCD patients with more severe symptoms will express a lower confidence level and higher rates of checking behaviour. To test these hypotheses, two tasks were performed by both OCD patients and HCs: a perceptual confidence task, performed in an fMRI-scanner, in which participants rated their confidence about their perceptual decisions and a Negative Outcome Expectancy (NOE) task to investigate compulsive checking behaviour.
6
2. Materials & Methods
2.1 Participants
A total of 29 OCD patients (11 males) and 49 HCs (28 males) participated in this study, matched on age, sex and educational level. All participants had normal or corrected to normal vision. OCD patients were recruited from the psychiatry department at the Academic Medical Centre (AMC) in Amsterdam and HCs were recruited through advertisement via posters and online advertisements. The ages of all participants ranged from 18-65 years old (OCD: 34.3 ± 8.1; HCs: 39.5 ± 14.0). The participants were compensated with €40,- and additional gains up to €16,- depending on task performance. Prior to participating in this study all participants were screened. OCD patients were included based on DSM-V diagnostic criteria and a score above 12 on the Y-BOCS (Goodman et al., 1989). The mini international neuropsychiatric interview (MINI) was also administered to exclude OCD patients and HCs with (other) psychiatric diagnoses (Sheehan et al., 1998).
All study procedures were approved by the Medical Ethics Committee of the Academic Medical Centre, University of Amsterdam and prior to participating in this study all participants provided written informed consent. Exclusion criteria for participating in this study were an IQ below 80, insufficient command of the Dutch language, MRI contraindications or current treatment with tricyclic antidepressants or antipsychotic medication. Furthermore, participants were instructed not to use any psychotropic medication or recreational drugs over a period of 72 hours and not to use alcohol over a period of 24 hours prior to participating. Moreover, data were excluded from both behavioural and fMRI analyses when the average accuracy in a session was lower than 50%, when deviation from the initial position of the cursor on the confidence scale was lower than 5% or when, due to a bug in the task, in 100% of the trials the correct answer was left or right. In total, one session was excluded for 13 participants (10 HCs) and another 4 subjects (4 HCs) were excluded for all analyses. In addition, fMRI data were excluded when head movement was more than 3.5 mm. This led to the full exclusion of one session for 4 participants (1 HCs) and the exclusion of one participant (1 HCs) for all fMRI analyses.
2.2 Perceptual confidence task
The perceptual confidence task was adapted from the study of Lebreton et al. (2018), making it usable for fMRI. The task was implemented using MATLAB ® (The MathWorks, Natick, MA, USA) and the COGENT toolbox (www.vislab.ucl.ac.uk/cogent.php). Each trial began with a fixation cross (750ms) followed by two Gabor patches on both sides of the screen (150ms). Participants then had to decide which of the two Gabor patches had the highest contrast (self-paced) using their left or right index finger. After a random jitter period (4500-6000ms), a cue was presented indicating
7
whether participants could win 100 points (green frame), lose 100 points (red frame) or neither win nor lose any points (grey frame). After this cue, the participants had to indicate how confident they were that their decision was correct on a scale from 50% to 100%. This was done by moving the cursor in steps of 5% with the left and right index fingers and selecting the answer with the right middle finger. The starting point of the cursor on the confidence scale randomly varied from 65% to 85% to avoid anchoring of confidence ratings on 75%. At the end of the trial, participants received feedback on whether they won or lost points (900ms). The inter-trial interval varied randomly between 4500 and 6000ms (Figure 1).
Two sessions of the task were performed with 72 trials each. Prior to the task, there were two practice sessions with 10 trials each, one before entering the fMRI scanner and one inside the scanner. Before the task started and after the first session, a calibration session consisting of 144 trials was performed to adjust the difficulty of the task stimuli per subject. The calibration session only consisted of the Gabor contrast discrimination part without confidence rating and incentive. In this session, every 12 trials the contrast differences of the Gabor patches (i.e. difficulty) were adapted in such a way that a performance of approximately 70% correct was reached (Lebreton et al., 2018). We chose to set a fixed performance accuracy for every subject to investigate confidence independently of performance.
Figure 1. Perceptual confidence task
Subjects first had to decide which of the two Gabor patches had the highest contrast. After a random jitter period (4500-6000ms), a cue was presented indicating whether participants could win 100 points (green frame), lose 100 points (red frame) or neither win nor lose any points (grey frame). Thereafter, participants had to indicate how confident they were that their preceding decision was correct on a scale from 50% to 100%. The calibration session only consisted of the Gabor contrast discrimination part without confidence rating and incentive. At the end of the trial participants received feedback on whether they won or lost points. The intertrial interval varied randomly between 4500 and 6000ms.
2.2.1 Task measures
The subjective confidence rating, ranging from 50-100%, and accuracy, either correct or incorrect, were assessed per trial. Furthermore, evidence per trial was calculated by the following formula:
8
|
𝐶𝑜𝑛𝑡𝑟𝑎𝑠𝑡 𝑉𝑎𝑙𝑢𝑒 𝐺𝑎𝑏𝑜𝑟 𝑅𝑖𝑔ℎ𝑡−𝐶𝑜𝑛𝑡𝑟𝑎𝑠𝑡 𝑉𝑎𝑙𝑢𝑒 𝐺𝑎𝑏𝑜𝑟 𝐿𝑒𝑓𝑡𝐶𝑜𝑛𝑡𝑟𝑎𝑠𝑡 𝑉𝑎𝑙𝑢𝑒 𝐺𝑎𝑏𝑜𝑟 𝐿𝑒𝑓𝑡+𝐶𝑜𝑛𝑡𝑟𝑎𝑠𝑡 𝑉𝑎𝑙𝑢𝑒 𝐺𝑎𝑏𝑜𝑟 𝑅𝑖𝑔ℎ𝑡
|
The intensity of the contrast of the Gabor displayed on the left or right side of the screen were matched to every subject’s performance during the calibration session. Lastly, the confidence calibration was calculated by the following formula:
(𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝐶𝑜𝑛𝑓𝑖𝑑𝑒𝑛𝑐𝑒 − 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒) ∗ 100
A confidence calibration, or confidence bias, of zero indicates that subjective confidence rating perfectly matches to performance level, whereas a nonzero value indicates miscalibration. This value could be either negative, indicating underconfidence, or positive, indicating
overconfidence.
2.3 Negative Outcome Expectancy task
The Negative Outcome Expectancy (NOE) task was adapted from Luigjes et al. (2016) and was implemented using MATLAB ® (The MathWorks, Natick, MA, USA) and the Psychtoolbox
(http://psychtoolbox.org/). The task consisted of 3 sessions, each with a duration of 8 minutes. The participants started each session with 75 points and the aim of the task was to collect as many points as possible in one session. Prior to every trial a fixation cross shortly appeared on the screen
(1000ms). Thereafter, a cue indicating whether participants could lose points with an incorrect answer was presented (700ms). The green border indicated that participants could win 10 points with a correct answer or lose nothing with an incorrect answer (no loss trial) and the orange border indicated that participants could win 10 points for a correct answer or lose 75 points when the answer was incorrect (loss trial). Then, the stimulus matrix with red and blue boxes was presented (1000ms). The stimulus was either difficult (51/49) or easy (80/20). After the stimulus matrix, a short mask (100ms) appeared consisting of a flash of coloured boxes. After this, participants had to decide whether there were more red or blue boxes in the matrix or the participants could choose to see the same matrix again (check) by pressing the return button (self-paced; max 5000ms). They could go back to see the field of boxes again for a maximum of four times, but checking costs time and would therefore decrease the number of opportunities to win points. Importantly, the stimulus was displayed so briefly that checking is not functional in the difficult condition. The inter-trial interval had a duration of 1000ms. At the end of every session, the participants received feedback about how many points they earned in total in that session (Figure 2).
9
Participants performed a practice session of 3 minutes prior to the main task sessions. During the practice trial the participants received feedback on how many points they won or lost after each choice and how much time they had left.
Figure 2. The Negative Outcome Expectancy task
At the beginning of each trial a cue was presented indicating whether participants could lose points with an incorrect answer (green border: win 10 points for correct answer, lose 0 points for incorrect answer; orange border: win 10 points for correct answer, lose 75 points for incorrect answer). Thereafter, the stimulus matrix with blue and red boxes was presented, which was either difficult (51/49) or easy (80/20). After a short post-stimulus mask consisting of a flash of coloured boxes, participants had to decide which colour had more boxes in the matrix or they could choose to see the matrix again. Participants could check the same matrix up to 4 times. The intertrial interval had a duration of 1000ms.
2.3.1 Behavioural measures
The number of checks (ranging from 0 to 4), accuracy (correct vs. incorrect), matrix difficulty (difficult vs. easy) (i.e. outcome uncertainty) and the possibility of losing points (loss vs. no loss) (i.e. outcome magnitude) were assessed per trial. The two-by-two factorial design (loss vs. no loss and easy vs. difficult) resulted in four possible conditions. The loss/difficult condition (both high outcome magnitude and outcome uncertainty) resulted in a Negative Outcome Expectancy (NOE). The
proportion of this NOE condition was 1 in 8 trials. This proportion was chosen in order to make sure checking was ineffective.
2.4 Procedure
All participants first received study information and gave written informed consent before participating in the study. After that, the NOE task was performed on a computer for a period of approximately 25-30 minutes. Instructions for the task were given on the screen and participants first performed a practice session with feedback. Participants were explicitly instructed that going back takes time and decreases the opportunities to win points. Then, both the NLV (Schmand, Lindeboom & Van Harskamp, 1992), a short Dutch reading comprehension test, and the WAIS-IV Digit Span (Wechsler, 2008) were administered to evaluate general intelligence. Next, the MINI (Sheehan et al.,
10
1998) and questions about demographics were assessed. Furthermore, the Y-BOCS was conducted to investigate OCD symptoms (Goodman et al., 1989).
After the questionnaires, participants performed two ten-trial practice sessions of the perceptual confidence task: one before entering the fMRI scanner and one inside the fMRI scanner. Then, both the calibration session and the main task, consisting of 2 sessions of 72 trials, were performed in the fMRI scanner. The duration of the perceptual confidence task was approximately 20 minutes per session and after session 1 participants had a short break. After scanning, questions regarding the quality of the compulsions of OCD patients were assessed.
2.5 Behavioural analyses
Behavioural data were collected and analysed using MATLAB (The MathWorks, Natick, MA, USA) and R (R Core Team).
2.5.1 Perceptual confidence task
To investigate the effect of group, accuracy and evidence on confidence, a linear mixed-effects model was performed, using the lmer function from the lme4 package in R (Bates et al., 2015). The final model was selected by model comparison, assessing model fit by using chi-square tests on the log-likelihood values and by comparison of the AIC model values. The final model contained fixed effects of group, accuracy and evidence and an interaction effect of group, accuracy and evidence, as well as a random intercept for each subject to account for individual differences. The model can be summarised as follows:
Confidence ~ Group * Accuracy * Evidence + (1|Subject)
Visual inspection of residual plots did not reveal any obvious deviations from homoscedasticity or normality. P-values were obtained by using the lmerTest package (Kuznetsova, Brockhoff &
Christensen, 2017). A linear mixed-effects model was used instead of a two sample t-test, since it includes both fixed and random effects and thereby gives a better understanding of what contributes to confidence. Additionally, it takes trial-by-trial data, without losing information due to averaging over trials.
Due to violation of the linearity and homoscedasticity assumptions of the Pearson’s correlation, the Spearman’s rank correlation coefficient was computed to investigate whether there was a relationship between Y-BOCS score and confidence level or confidence calibration in OCD patients.
11
2.5.2 Negative Outcome Expectancy task
For the main analysis of the NOE task, a Zero-Inflated Poisson model was used to investigate the interaction effect of group, possibility of losing points and difficulty on number of checks, using the glmmTMB function from the glmmTMB package in R (Brooks et al., 2017). The final model was selected by model comparison, assessing model fit by using chi-square tests on the log-likelihood values and by comparison of the AIC model values. The final model with a single zero inflation parameter contained fixed effects of group, loss possibility and difficulty and the interaction term of group, loss possibility and difficulty, as well as a random slope of session within subject. A Zero-Inflated Poisson model was used to account for excessive zero counts in the dependent variable (i.e. number of checks). The model can be summarised as follows:
Number of Checks ~ Group * Loss Possibility * Difficulty + (Session|Subject), ziformula= ~1
Due to non-normality of the data, the Spearman’s rank correlation coefficient was used for the correlational analyses of the NOE task. The Spearman’s rank correlation coefficient was
computed to investigate whether: 1) the average checks over all conditions in the NOE task related to confidence or confidence calibration in the perceptual confidence task; 2) the average checks in the NOE condition (loss x difficult trials) related to confidence or confidence calibration; 3) average checks over all conditions correlated with Y-BOCS score in OCD patients and 4) average checks in the NOE condition related to Y-BOCS score in OCD patients.
2.6 fMRI acquisition and pre-processing
The perceptual confidence task was presented on an Iiyama monitor (resolution of 1920x1080; 120 Hz) in the fMRI environment. Subjects viewed the stimuli through a 45-degree angle mirror on the head coil. A T1-weighted structural anatomical image followed by T2*-weighted echo-planar images (EPI) sensitive to BOLD contrast were acquired on a 3.0 Tesla Philips MRI scanner (Philips Medical Systems, Best, The Netherlands). A multi-echo combined interleaved scan sequence was used to increase BOLD sensitivity in brain regions (Poser et al., 2006). The following parameters were used: TR=2.375, TEs (3 echoes)=9ms, 24.0ms and 43.8ms, 37 slices, 3mm voxel size, 0.3mm interslice gap. Two experimental sessions, consisting of 570 volumes each, were executed.
fMRI pre-processing and all further analyses were performed in MATLAB R2018a (The
MathWorks, Natick, MA, USA) using the SPM12 software (https://www.fil.ion.ucl.ac.uk/spm/). First, all three echoes were weighted and averaged to combine them into one functional image. All functional images were realigned to the first volume with linear interpolation during the combining. The first 30 dummy scans were discarded. Moreover, since the number of task-based scans varied among participants, all non-task scans were discarded on a subject-by-subject basis. The functional
12
images were co-registered to the T1-weighted structural image and segmented for normalisation. Next, the functional images were spatially normalised to MNI space. Subsequently, spatial smoothing was performed using a full-width at half-maximum 6mm Gaussian kernel to improve signal-to-noise ratio. Some volumes had movement artefacts due to the interleaved scanning method in
combination with fast motion. The ArtRepair toolbox (Mazaika et al., 2009) was used to remove these artefacts. This toolbox searches for outliers using the mean movement and interpolates the data values in the outliers based on the 12 adjacent volumes. Volumes with a deviation of more than 1.5% from the mean intensity of the BOLD signal were repaired.
2.7 fMRI analyses
A General Linear Model (GLM) was used to analyse the fMRI data. Three moments of interest were used, namely the perceptual choice moment (onset of the Gabor patches), the moment of confidence rating (onset of the display of the rating scale) and the moment of feedback. The choice regressor was parametrically modulated by the confidence rating participants gave each trial. This parametrically modulated regressor was chosen based on earlier findings of automatic confidence encoding in the VMPFC and early confidence signals prior to explicit report (Lebreton et al., 2015; Gherman & Philiastides, 2018). Furthermore, the feedback regressor was parametrically modulated by accuracy (correct vs. incorrect). As this study mainly focuses on confidence coding, we do not provide analyses for this parametric modulator (pmod). Furthermore, the motion parameters for each subject were included as regressors of no interest to control for head motion. A high-pass filter (128s cut-off) was included to filter low frequency noise. A canonical haemodynamic response function (HRF) (double-gamma) was used to convolve each regressor.
All contrasts used in our analysis were calculated per subject. Contrast images containing parameter estimates for each comparison of interest were then fed into second level analyses to assess within- and between-group differences. First, to find areas related to confidence encoding, a contrast was set to compare the perceptual choice moment modulated by the pmod confidence rating to baseline. This was analysed on whole-brain level and tested using a one-sample t-test in the HC group and OCD group separately.
Next, to test for group differences in confidence encoding in the VMPFC, this contrast was used in a region of interest (ROI) analysis. A ROI analysis was used because of our a-priori hypothesis on the role of the VMPFC in confidence encoding. An 8 mm sphere around the peak MNI coordinates reported in the study of Lebreton et al. (2015) was used in our study to create the ROI (-2, 52, -2). A small volume correction was applied with this mask for the ROI analysis. The mask was created using the WFU Pickatlas 2.5 toolbox in SPM12. Group differences were tested with an independent two
13
sample t-test. A cluster defining threshold of p<0.001 was used and statistical tests were family-wise error corrected for multiple comparison at cluster level (pFWE_clu<0.05).
3. Results
3.1 Behavioural results
3.1.1 Perceptual confidence task
A linear mixed-effects model was constructed to investigate the effect of accuracy, evidence and group on confidence. The model showed a significant main effect of group on confidence (β=6.18 ±1.92, t=3.21, p=0.0018), revealing that HCs (M=76.13, SD=15.07) have increased confidence compared to OCD patients (M=72.59, SD=16.61). Furthermore, a significant interaction between evidence and accuracy on confidence was found (β=5.88 ±2.45, t=2.41, p=0.02), indicating decreased confidence for incorrect trials with high evidence and increased confidence for correct trials with high evidence. Moreover, a significant interaction between group, accuracy and evidence on
confidence was found (β=22.57 ±4.19, t=5.39, p<0.001), revealing a disturbed integration of evidence and accuracy in subjective confidence in OCD patients: confidence does not increase with high evidence for correct answers in OCD patients (Figure 3). Lastly, a main effect of accuracy (β=4.79 ±0.60, t=8.03, p<0.001) and evidence (β=9.45 ± 2.54, t=3.73, p<0.001) on confidence was found, as well as an interaction effect of group and evidence on confidence (β=-23.41 ± 4.13, t=-5.67, p<0.001).
Figure 3. The interaction effect of accuracy, evidence and group
The mean confidence level as a function of evidence for the different accuracy levels is displayed separately for the two groups. OCD patients are shown in the left panel and healthy controls in the right panel. The interaction effect of accuracy, evidence and group (β=22.57 ±4.19, t=5.39, p<0.001) reveals a disturbed integration of evidence and
14
accuracy in OCD patients: confidence does not increase with high evidence for correct answers in OCD patients. Shaded areas represent standard errors of the regression lines.
The Spearman’s rank-order correlation revealed no significant association between Y-BOCS score (M=20.52, SD=6.09) and confidence level (M=72.59, SD=16.61) (S=4879, rs=-0.20, p=0.29) or confidence calibration (M=-0.63, SD=9.24) (S=4520, rs=-0.11, p=0.56).
3.1.2 Negative Outcome Expectancy task
A Zero-Inflated Poison model was constructed to investigate the interaction effect of group, difficulty and possibility of losing points on number of checks. The descriptive statistics for each condition are shown in table 1. A significant interaction effect of difficulty and possibility of losing points was found (β=0.42 ±0.20, z=2.16, p=0.03), indicating that checking increased in the
difficult/loss trials (Figure 4). Furthermore, a significant main effect of difficulty was found (β=1.91 ±0.13, z=14.77, p<0.001), revealing an increase in checks with difficulty. However, no significant main effect of group (β=-0.57 ±0.78, z=-0.73, p=0.47) or a main effect of possibility of losing points on checking was found (β=0.11 ±0.18, z=0.63, p=0.53). Also, no significant interaction between group, difficulty and possibility of losing points on number of checks was found (β=0.13 ±0.43, z=0.30, p=0.76). Lastly, a significant interaction of group and difficulty was found (β=0.66 ±0.26, z=2.53, p=0.01), with HCs showing a slight increase in checks in the difficult trials compared to OCD patients.
Healthy controls OCD
M SD M SD
Difficult x Loss 0.144 0.430 0.111 0.340 Difficult x No Loss 0.084 0.328 0.083 0.303 Easy x Loss 0.014 0.137 0.005 0.069 Easy x No Loss 0.012 0.122 0.007 0.081
Table 1. Descriptive statistics for each condition in the negative outcome expectancy task
15
Figure 4. The interaction effect of difficulty and possibility of losing points
The mean number of checks as a function of possibility of losing points for the different difficulty levels is displayed for both groups. The interaction effect of difficulty and possibility of losing points (β=0.42 ±0.20, z=2.16, p=0.03) reveals an increase in checking in the difficult/loss condition.
One outlier was detected in our correlational data. However, since the Spearman’s rank correlation is found to be robust to outliers (de Winter et al., 2016; Kim et al., 2015), we chose to keep it in the data. For exploratory purposes we conducted the correlational analyses without this outlier as well, these results can be found in the Appendix. The average number of checks in both groups together (M=0.04, SD=0.24) negatively correlated with both confidence level (S=74237, rs=-0.24, p=0.04) and confidence calibration (S=80403, rs=-0.35, p=0.003) (Figure 5A and 5B).
Furthermore, confidence calibration negatively correlated with number of checks in the NOE condition (S=76438, rs=-0.28, p=0.02) (Figure 5C). However, no significant correlation was found between average checks in the NOE condition (M=0.13, SD=0.40) and confidence level (S=4803, rs=-0.17, p=0.15) (Figure 5D). Lastly, no significant association between Y-BOCS score and average checks over all conditions (M=0.04, SD=0.21) (S=2327, rs=0.29, p=0.14) or in the NOE condition (M=0.11, SD=0.34) (S=2465, rs=0.25, p=0.21) was found in OCD patients.
16
Figure 5. Scatterplots of the correlational analyses
A) The significant negative correlation between average number checks and confidence level (S=74237, rs=-0.24, p=0.04). B) The significant negative correlation between average number of checks and confidence calibration (S=80403, rs=-0.35, p=0.003). C) The significant negative correlation between average number checks in the NOE condition and confidence calibration (S=76438, rs=-0.28, p=0.02). D) No significant correlation between average number checks in NOE condition and confidence (S=4803, rs=-0.17, p=0.15).
3.2 fMRI results
3.2.1 Confidence encoding
To investigate which areas are involved in confidence coding, a one sample t-test was conducted separately for both HCs and OCD patients. In HCs, during choice, subjective
confidence linearly related to BOLD activation in the VMPFC, stretching into the ventral anterior cingulate cortex (vACC) (peak MNI coordinates [0 44 -4] mm, k=163, Z=4.63, pFWE_clu <0.001) (Figure 6). Also, activity in the bilateral striatum (peak MNI coordinates [-27 5 -7] mm, k=264, Z=5.43, pFWE_clu <0.001 and peak MNI coordinates [24 2 -10] mm, k=184, Z=5.10, pFWE_clu <0.001 for left and right striatum respectively) positively related to confidence ratings in HCs (Figure 7). Looking at the results in OCD patients, we found activity in the bilateral striatum to be linearly related to confidence (peak MNI coordinates [9 20 -4] mm, k=247, Z=5.29, pFWE_clu <0.001), but no relationship within the VMPFC.
Concerning areas that related negatively to confidence (i.e. increased in activity with lower confidence), we found significant effects for the bilateral insula in HCs (peak MNI coordinates
[-17
33 20 5] mm, k=67, Z=4.45, pFWE_clu =0.016 and peak MNI coordinates [33 23 5] mm, k=103, Z=4.50, pFWE_clu =0.002 for left and right insula respectively). Additionally, BOLD activity in the dorsal anterior cingulate cortex (dACC) negatively related to subjective confidence in HCs (peak MNI coordinates [-9 20 38] mm, k=259, Z=4.54, pFWE_clu <0.001). Regarding OCD patients, we found that BOLD activation in the right insula negatively related to confidence ratings (peak MNI coordinates [48 8 2] mm, k=56, Z=4.51, pFWE_clu =0.018), as well as dACC activity (peak MNI coordinates [0 17 44] mm, k=50, Z=3.64, pFWE_clu =0.028). A comprehensive list of cluster activation from these analyses can be found in table 2.
Contrast Group Brain region k T Z Peak voxel
MNI
coordinates
pFWE_clu
Confidence + HC Left Striatum 264 6.65 5.43 [-27 5 -7] <0.001
HC Right Striatum 184 6.10 5.10 [24 2 -10] <0.001
HC Ventromedial Prefrontal Cortex Ventral Anterior Cingulate Cortex
163 5.36 4.63 [0 44 -4] <0.001
HC Left Primary Motor Cortex 103 4.94 4.34 [-33 -25 53] 0.002
OCD Bilateral Striatum 247 7.03 5.29 [9 20 -4] <0.001
OCD Left Primary Motor Cortex 176 6.55 5.06 [-33 -19 53] <0.001
OCD Cerebellum 368 6.45 5.01 [15 -64 -25] <0.001
OCD Right Visual Cortex 72 6.09 4.82 [21 82 23] 0.005
OCD Left Visual Cortex 117 5.23 4.33 [-9 -91 20] <0.001
Confidence - HC Right Primary Motor Cortex 423 6.84 5.54 [36 -16 56] <0.001 HC Dorsal Anterior Cingulate Cortex 259 5.23 4.54 [-9 20 38] <0.001
HC Right Insula 103 5.18 4.50 [33 23 5] 0.002
HC Left Insula 67 5.11 4.45 [-33 20 5] 0.016
OCD Right Insula 56 5.53 4.51 [48 8 2] 0.018
OCD Right Primary Motor Cortex 126 5.26 4.35 [33 -28 53] <0.001 OCD Dorsal Anterior Cingulate Cortex 50 4.16 3.64 [0 17 44] 0.028 Table 2. Significant clusters in whole brain analysis in healthy controls and OCD patients
Peak coordinates for the second level GLM analysis are shown. For each significant cluster, the number of voxels, t-statistic, z-t-statistic, anatomical region and coordinates (mm) in Montreal Neurological Institute (MNI) space are presented (pFWE_clu<0.05; cluster-defining threshold p<0.001).
18
Figure 6. Activity positively related to confidence in healthy controls
During the moment of choice, subjective confidence linearly related to both BOLD activation in the VMPFC (peak MNI coordinates [0 44 -4] mm, k=163, T=5.36, Z=4.63, pFWE_clu <0.001) and in the bilateral striatum in healthy controls
(peak MNI coordinates [-27 5 -7] mm, k=264, T=6.65, Z=5.43, pFWE_clu <0.001 and peak MNI coordinates [24 2 -10] mm,
k=184, T=6.10, Z=5.10, pFWE_clu <0.001 for left and right putamen respectively).
3.2.2 Group differences in confidence coding
To investigate if confidence coding differs between HCs and OCD patients, a two sample t-test was conducted. No significant differences were found in confidence encoding in the VMPFC between HCs and OCD patients using the ROI analysis. For exploratory reasons, we also conducted a whole-brain analysis on the difference in confidence coding between groups. However, we did not find any differences in confidence encoding between the groups on whole-brain level.
4. Discussion
The first aim of our study was to investigate if OCD patients are less confident compared to HCs and whether this is represented in the brain as distorted confidence encoding in the VMPFC. In addition, we investigated the relationship between confidence and symptom severity. First off, we found that OCD patients showed decreased confidence levels compared to HCs, but this effect did not result in altered brain activity during confidence encoding. Whereas we did show that subjective confidence is encoded in the VMPFC, no group differences were found in confidence coding in this area. Additionally, no significant associations between symptom severity and confidence level or confidence calibration were found. The second aim of our study was to investigate the role of negative outcome anticipation in compulsive checking and to explore whether this affects
compulsive checking differently in OCD patients. We found a significant interaction effect of difficulty (i.e. outcome uncertainty) and possibility of losing points (i.e. outcome magnitude), indicating that individuals perform more compulsive checking when a possible negative outcome is expected.
19
However, we did not find evidence for an overall increase in compulsive checking in OCD patients or for increased checking specifically when anticipating negative outcomes. Lastly, we examined the link between compulsive checking and both confidence and symptom severity. A significant negative correlation between compulsive checking and both confidence level and confidence calibration was found, whereas no significant association between compulsive checking and symptom severity was found.
4.1 Confidence distortions in OCD
In line with our hypothesis and previous findings, we found that OCD patients show decreased confidence compared to HCs. However, contrary to our expectations, we did not find evidence for a relationship between symptom severity and confidence level or confidence calibration. Previous studies have reported mixed results on the relationship between confidence and symptom severity in OCD patients (Ouellet-Courtois et al., 2018). One explanation for this finding might be that OCD patients usually show a range of different symptoms. This heterogeneity in OCD patients may lead to decreased power (Mataix-Cols, do Rosario-Campos & Leckman, 2005) and it is plausible that confidence is related only to specific symptom dimensions in OCD. Additionally, distortions in different cognitive domains may be responsible for different subsets of OCD symptoms (Ouellet-Courtois et al., 2018). For example, low perceptual confidence may lead to compulsive cleaning, whereas low memory confidence may be related to repetitive compulsive behaviour and checking (Ouellet-Courtois et al., 2018). Future studies would benefit from using questionnaires that do take different symptom dimensions into account, instead of looking at overall OCD symptoms, such as the dimensional Y-BOCS (do Rosario-Campos et al., 2006).
4.2 Neural correlates of confidence
In this study we replicated previous findings by showing that subjective confidence is encoded in the VMPFC in HCs (Lebreton et al., 2015; Gherman & Philiastides, 2018). The linear relationship between confidence and VMPFC activity was observed during the moment of choice, hence providing further support for the idea of automatic confidence encoding in the VMPFC. The VMPFC is known to be part of the brain valuation system (Lebreton et al., 2009; Lopez, 2016) and our finding of confidence encoding in the VMPFC adds to this idea that next to subjective value, the VMPFC also automatically incorporates confidence (Lebreton et al., 2015; De Martino et al., 2013).
Furthermore, BOLD activation in the vACC also encoded confidence in HCs. The linear relationship of ACC activation and confidence is consistent with earlier findings of Bang & Fleming (2018), who showed that the perigenual ACC tracked determinants of perceptual decision
20
OCD patients. The striatum has been implicated in confidence encoding in previous research as well (Vaccaro & Fleming, 2018; Hebart et al., 2016). As such, Hebart et al. (2016) showed that activity in the ventral striatum correlated to confidence in a perceptual decision making task. Also, Daniel & Pollmann (2012) found prediction error signals of confidence in the putamen, with increased activity when confidence was higher than expected. The finding of the striatum being involved in confidence coding might be attributable to the promise of reward one gets when being really confident about a decision.
Moreover, BOLD activity in the bilateral insula in HCs and right insula in OCD patients was found to be negatively related to confidence level. The negative association between confidence ratings and insula activity is in accordance with previous studies (Gherman & Philiastides, 2018; Heereman et al., 2015; Vaccaro & Fleming, 2018). Paul et al. (2015) showed that the insula
represents perceptual uncertainty. The insula is an area that is thought to be involved in conscious error perception, showing increased activity when one is aware of an error (Ullsperger et al., 2010). This might explain our findings of increased insula activity with lower confidence ratings. This awareness of an incorrect answer subsequently activates the insula. Furthermore, dACC activity negatively related to subjective confidence in both HCs and OCD patients. Multiple studies reported similar findings of an increasing BOLD response in the dACC with decreasing reported confidence (Fleming & Dolan, 2012; Shapiro & Grafton, 2020). The dACC has been proposed to play a role in conflict detection and the adjustment of erroneous responses (Carter & Van Veen, 2007; Hochman et al., 2014), which might be the reason for the inverse relationship with confidence we found, since lowered confidence might lead to activation of the dACC in order to adjust future behaviour.
We did not find evidence for distorted confidence coding in the VMPFC in OCD patients when comparing OCD patients with HCs. This finding is not in line with our hypothesis and the behavioural results, in which OCD patients showed decreased confidence compared to HCs. There are several possible explanations for this result. This null finding might be explained by the low sample size in our study, causing the analysis to be underpowered. At sub-threshold level, we did observe some exploratory group differences in VMPFC activation linearly related to confidence, with HCs showing increased VMPFC activity compared to OCD patients. In addition, BOLD activity in the VMPFC was not found to be linearly related to confidence ratings in OCD patients. Both of these findings indicate that there might be a case of a lack of power. Therefore, more subjects are needed to ensure sufficient statistical power. However, we do not rule outthe possibility that there actually are no group differences in confidence coding in the VMPFC.
21
Another possibility is that the abnormalities in confidence judgments do not arise through deviations in VMPFC activity, but through defective connective pathways. Bang et al. (2020)
discovered that medial areas of the prefrontal cortex coded an internal sense of confidence, whereas regions more lateral in the prefrontal cortex tracked the confidence we explicitly report. Additionally, De Martino et al. (2013) showed that the VMPFC encodes decision confidence, whereas the
rostrolateral prefrontal cortex (RLPFC) was found to encode self-reported confidence. Moreover, functional connectivity between VMPFC and RLPFC was modulated by confidence level and was predictive for the relationship between confidence and performance across subjects. The authors therefore hypothesised that the RLPFC reads out the confidence signals, computed in the VMPFC, and makes them available for metacognitive report. Considering these findings, it might be that the transfer of the confidence signal computed in the VMPFC to the RLPFC is distorted. Future studies could examine if functional connectivity between VMPFC and RLPFC is distorted in OCD patients.
Lastly, we found an interaction effect of group and evidence on confidence, revealing a distorted relationship between evidence and confidence in OCD patients. This is supported by previous studies by Hauser et al. (2017) who observed that the effect of evidence on confidence was reduced in high obsessive-compulsive individuals: high obsessive-compulsives needed more evidence in order to make a perceptual decision (i.e. higher decision threshold). Similarly, Seow & Gillan (2020) reported that the transdiagnostic symptom dimension compulsivity was linked to distorted evidence accumulation in order to update behaviour. Future analyses could investigate this evidence
integration distortion by adding the perceptual evidence in each trial as a parametric modulator to the choice regressor and compare the brain activity in HCs with OCD patients.
4.3 The effect of negative outcome anticipation on compulsivity
We found an interaction effect of difficulty and possibility of losing points on number of checks, replicating previous findings of Luigjes et al. (2016). Both HCs and OCD patients showed elevated checking in the difficult/loss condition (i.e. high outcome magnitude and uncertain
outcome), indicating that negative outcome anticipation might indeed be an underlying mechanism of compulsive checking. However, incongruent with our hypothesis, we found neither an effect of group nor an interaction effect of difficulty, possibility of losing points and group on number of checks: thus OCD patients did not show increased checking while anticipating negative outcomes compared to HCs. Our results therefore do not confirm the proposed idea that an increased sensitivity to negative outcome anticipation would be related to the development of compulsive behaviours in disorders such as OCD. An important finding in our study is that there is low individual variation in number of checks: participants do not check much at all, even in the negative outcome condition. Previous studies have reported mixed results on checking behaviour in OCD patients in
22
tasks measuring checking behaviour in the lab, probably due to differences in task type and how checking is measured (Strauss et al., 2020). Strauss et al. (2020) mentioned an important
consideration for studies investigating checking behaviour in OCD, namely that tasks in which
participants are punished for checking (e.g. tasks in which the points one could possibly win decrease with checks) may not represent checking solely, but rather other processes like sensitivity to
penalisation. Since participants in our study were explicitly instructed that checking interferes with the overall goal of the task to win points and OCD patients are thought to be more prone to
punishments (Fullana et al., 2004), this might explain our findings of no group differences in checking behaviour. It is therefore possible that our task is confounded by the fact that OCD patients are more sensitive to punishment. To overcome this problem, in future studies instructions could be adjusted in not stressing the penalisation.
Furthermore, in accordance with our hypotheses, the overall average number checks was negatively related to both confidence level and confidence calibration and the average checks in the NOE condition negatively related to confidence calibration. Nevertheless, contrary to expectations, we did not find a significant association between number of checks in the NOE condition and confidence level. The finding of a negative relationship between checks and confidence is in agreement with past research in which reductions in confidence were found to be associated with compulsive checking (MacDonald et al., 1997; Moritz & Jaeger, 2018; Cuttler et al., 2013). However, these studies measured both subjective confidence and checking behaviour in one task, whereas we compared confidence and checking behaviour in two tasks separately. Also, we chose to keep an outlier in the data due to robustness of the Spearman’s rank correlation, but our exploratory analysis without the outlier revealed no significant association between confidence level and number of checks (Appendix). Thus, these findings must be interpreted with caution due to the limitations, skewed distribution to zero and the outlier. Furthermore, since correlational analyses do not allow for causal inferences, it is unclear whether confidence reductions are a consequence or a cause of increased checking. Cuttler et al. (2013) tried to investigate the causal effect of confidence distortions on doubt and urges to check in a prospective memory task and found that, in a non-clinical sample, reduced memory confidence leads to increased doubt and urges to check, whereas Van Den Hout and Kindt (2003) paradoxically showed that checking in itself leads to decreases in memory confidence.
Incongruent to our hypothesis, no significant associations were found between checking and symptom severity. This could be explained by the low variability in checks and the skewed
23
the harming/checking dimension (Mataix-Cols, do Rosario-Campos & Leckman, 2005; Van Den Heuvel et al., 2009).
Interestingly, the VMPFC and striatum are thought to be involved in outcome anticipation (Spirou et al., 2018). These regions have been implicated in confidence encoding in our study and previous studies. Moreover, multiple neuroimaging studies have reported abnormalities in these regions in OCD patients (Nakao et al., 2014). This might indicate a common underlying mechanism to both confidence distortions and negative outcome anticipation in OCD. Future research could perform the NOE task in an fMRI environment to investigate the neural correlates of negative outcome anticipation in OCD patients and see whether VMPFC and striatum show aberrant activity.
4.4 Limitations
First, in this study, we did not use any bias-free measures of confidence precision. Although we did use a fixed performance accuracy for every subject, measures such as metacognitive
sensitivity (meta-d’) or metacognitive efficiency (meta-d’/d’), based on the framework of the signal detection theory, may be more reliable and more informative since these measures provide
information about the precision of confidence (Fleming & Lau, 2014; Hoven et al., 2019; Maniscalco & Lau, 2012). In addition, the linear relationship between VMPFC activity and confidence might be attributable to aspects that themselves vary with confidence instead of confidence on its own, such as reaction time, however, we did not control for this. This can be done by adding the reaction time parameters for each subject as regressors to the GLM. Furthermore, the effects of feedback on confidence have not been taken into account in this study. Participants received feedback after every trial in the perceptual confidence task, which might have a direct effect on confidence. Importantly, decision feedback is usually absent in real-life situations, thus this might not be a valid representation of OCD-like behaviour. It would be interesting to investigate to what extent feedback influences confidence ratings and see whether confidence distortions in OCD patients get worse when they do not receive feedback (i.e. uncertain environment). Moreover, Elliott et al. (2020) recently revealed an important limitation to fMRI tasks in general. They showed that the test-retest reliability of fMRI tasks is low, which may further exacerbate the deficiency in power commonly seen in fMRI studies, since even more participants are needed to ensure sufficient power. Another limitation is the confidence scale, ranging from 50-100%, used in our study. Individuals might differ in how they interpret this scale. Although, Tekin and Roediger (2017) showed that the relationship between confidence and accuracy does not differ much using different scales, it could be the case that the neural signature of confidence is influenced by this within-group difference.
24
4.5 Recommendations for Future Research
The perceptual decision-making task used in our study to measure confidence does not represent real-life scenarios, making it difficult to assess how cognitive confidence in the context of OCD looks like (Ouellet-Courtois et al., 2018). Gaining more knowledge about confidence distortions considering various OCD-contexts is important, since it could help with specifically targeting the confidence distortions and therefore improving treatment outcomes. Future research should therefore investigate confidence abnormalities in OCD in symptom-specific contexts. Furthermore, some studies have shed some light on the possibility of training confidence. As such, Carpenter et al. (2019) showed that adaptive training enhances metacognitive calibration, which generalised over domains. Furthermore, multiple studies have shown that mindfulness-based meditation improves metacognitive abilities (Fox et al., 2012; Baird et al., 2014; Heyes et al., 2020). An important future research endeavour is to investigate whether these metacognitive training strategies could be used in OCD patients to target specific confidence distortions and to examine whether the improved metacognitive abilities also result in reduced symptoms.
To conclude, we found that OCD patients are underconfident compared to HCs, but this effect did not translate to altered brain activity. Moreover, we showed that -normally- subjective confidence is encoded in the VMPFC in HCs. Furthermore, we demonstrated that expectancy of a negative outcome increases ineffective checking behaviour in both HCs and OCD patients. Finally, compulsive checking was found to be negatively related to confidence.
25
References
American Psychiatric Association. (2013) Diagnostic and statistical manual of mental disorders, Fifth Edition. Washington, DC American Psychiatric Association.
https://doi.org/10.1176/appi.books.9780890425596
Baird, B., Mrazek, M. D., Phillips, D. T., Schooler, J. W. (2014). Domain-specific enhancement of metacognitive ability following meditation training. Journal of Experimental Psychology.
General, 143(5), 1972–1979. https://doi.org/10.1037/a0036882
Bang, D., Ershadmanesh, S., Nili, H. & Fleming, S. (2020). Private-public mappings in human prefrontal cortex. https://doi.org/10.1101/2020.02.21.954305
Bang, D., & Fleming, S. M. (2018). Distinct encoding of decision confidence in human medial prefrontal cortex. Proceedings of the National Academy of Sciences, 115 (23) 6082-6087. https://doi.org/10.1073/pnas.1800795115
Brooks, M. E., Kristensen, K., van Benthem, K. J., Magnusson, A., Berg, C. W., Nielsen, A., Skaug, H. J., Mächler, M. & Bolker, B. (2017). glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. The R Journal, 9(2), 378-400.
Cambridge Dictionary. (2020). Metacognition definition. Achieved from: https://dictionary.cambridge.org/dictionary/english/metacognition
Carpenter, J., Sherman, M. T., Kievit, R. A., Seth, A. K., Lau, H., & Fleming, S. M. (2019). Domain-general enhancements of metacognitive ability through adaptive training. Journal of
Experimental Psychology: General, 148(1), 51–64. https://doi.org/10.1037/xge0000505
Cuttler, C., Sirois-Delisle, V., Alcolado, G. M., Radomsky, A. S., & Taylor, S. (2013). Diminished confidence in prospective memory causes doubts and urges to check. Journal of Behavior
Therapy and Experimental Psychiatry, 44(3), 329–334.
https://doi.org/10.1016/j.jbtep.2013.01.001
Daniel, R., & Pollmann, S. (2012). Striatal activations signal prediction errors on confidence in the absence of external feedback. NeuroImage, 59(4), 3457–3467.
https://doi.org/10.1016/j.neuroimage.2011.11.058
Dar, R., Rish, S., Hermesh, H., Taub, M., & Fux, M. (2000). Realism of confidence in obsessive-compulsive checkers. Journal of Abnormal Psychology, 109(4), 673–678.
26
De Martino, B., Fleming, S. M., Garrett, N., & Dolan, R. J. (2013). Confidence in value-based choice. Nature neuroscience, 16(1), 105–110. https://doi.org/10.1038/nn.3279 de Winter, J. C. F., Gosling, S. D., & Potter, J. (2016). Comparing the pearson and spearman
correlation coefficients across distributions and sample sizes: A tutorial using simulations and empirical data. Psychological Methods, 21(3), 273–290. https://doi.org/10.1037/met0000079 Elliott, M., Knodt, A., Ireland, D., Morris, M., Poulton, R., Ramrakha, S., Sison, M., Moffitt, T., Caspi,
A., & Hariri, A. (2020). What is the Test-Retest Reliability of Common Task-fMRI Measures? New Empirical Evidence and a Meta-Analysis. Biological Psychiatry, 87(9), S132–S133.
https://doi.org/10.1016/j.biopsych.2020.02.356
Fisher, P. L., & Wells, A. (2008). Metacognitive therapy for obsessive-compulsive disorder: A case series. Journal of Behavior Therapy and Experimental Psychiatry, 39(2), 117–132.
https://doi.org/10.1016/j.jbtep.2006.12.001
Fleming, S. M., & Dolan, R. J. (2012). The neural basis of metacognitive ability. Philosophical
Transactions of the Royal Society B: Biological Sciences, 367(1594), 1338–1349.
https://doi.org/10.1098/rstb.2011.0417
Fleming, S. M., Dolan, R. J., & Frith, C. D. (2012). Metacognition: Computation, biology and function.
Philosophical Transactions of the Royal Society B: Biological Sciences, 367(1594), 1280–1286.
https://doi.org/10.1098/rstb.2012.0021
Fleming, S. M., & Lau, H. C. (2014). How to measure metacognition. Frontiers in Human Neuroscience,
8, 1–9. https://doi.org/10.3389/fnhum.2014.00443
Fox, K. C. R., Zakarauskas, P., Dixon, M., Ellamil, M., Thompson, E., & Christoff, K. (2012). Meditation Experience Predicts Introspective Accuracy. PLoS ONE, 7(9).
https://doi.org/10.1371/journal.pone.0045370
Fullana, M. A., Mataix-Cols, D., Caseras, X., Alonso, P., Manuel Menchón, J., Vallejo, J., & Torrubia, R. (2004). High sensitivity to punishment and low impulsivity in obsessive-compulsive patients with hoarding symptoms. Psychiatry Research, 129(1), 21–27.
https://doi.org/10.1016/j.psychres.2004.02.017
Goodman, W., Price, L., Rasmussen, S., Mazure, C., Delgado, P., Heninger, G., & Charney, D. (1989). The Yale-Brown Obsessive Compulsive Scale. II. Validity. Archives of General Psychiatry., 46(11), 1012–1016. https://doi.org/10.1001/archpsyc.1989.01810110054008
27
clinical psychiatry. 60 (suppl 18):27-32
Gherman, S., & Philiastides, M. G. (2018). Human VMPFC encodes early signatures of confidence in perceptual decisions. eLife, 7, e38293. https://doi.org/10.7554/eLife.38293
Grimaldi, P., Lau, H., & Basso, M. A. (2015). There are things that we know that we know, and there are things that we do not know we do not know: Confidence in decision-making. Neuroscience
& Biobehavioral Reviews, 55, 88–97. https://doi.org/10.1016/j.neubiorev.2015.04.006
Hauser, T. U., Allen, M., Rees, G., Dolan, R. J., Bullmore, E. T., Goodyer, I., Fonagy, P., Jones, P., Fearon, P., Prabhu, G., Moutoussis, M., St Clair, M., Cleridou, K., Dadabhoy, H., Granville, S., Harding, E., Hopkins, A., Isaacs, D., King, J., … Pantaleone, S. (2017). Metacognitive impairments extend perceptual decision making weaknesses in compulsivity. Scientific Reports, 7(1), 1–10. https://doi.org/10.1038/s41598-017-06116-z
Hebart, M. N., Schriever, Y., Donner, T. H., & Haynes, J. D. (2016). The Relationship between
Perceptual Decision Variables and Confidence in the Human Brain. Cerebral Cortex, 26(1), 118– 130. https://doi.org/10.1093/cercor/bhu181
Hebscher, M., Barkan-abramski, M., Goldsmith, M., Aharon-peretz, J., & Gilboa, A. (2016). Memory , Decision-Making , and the Ventromedial Prefrontal Cortex ( vmPFC ): The Roles of Subcallosal and Posterior Orbitofrontal Cortices in Monitoring and Control Processes. Cerebral cortex (New
York, N.Y. : 1991), 26(12), 4590–4601. https://doi.org/10.1093/cercor/bhw220
Heereman, J., Walter, H., & Heekeren, H. R. (2015). A task-independent neural representation of subjective certainty in visual perception. Frontiers in Human Neuroscience, 9, 1–12.
https://doi.org/10.3389/fnhum.2015.00551
Hermans, D., Engelen, U., Grouwels, L., Joos, E., Lemmens, J., & Pieters, G. (2008). Cognitive confidence in obsessive-compulsive disorder: Distrusting perception, attention and memory.
Behaviour Research and Therapy, 46(1), 98–113. https://doi.org/10.1016/j.brat.2007.11.001
Heyes, C., Bang, D., Shea, N., Frith, C. D., & Fleming, S. M. (2020). Knowing Ourselves Together: The Cultural Origins of Metacognition. Trends in Cognitive Sciences, 24(5), 349–362.
https://doi.org/10.1016/j.tics.2020.02.007
Hochman, E. Y., Vaidya, A. R., & Fellows, L. K. (2014). Evidence for a role for the dorsal anterior cingulate cortex in disengaging from an incorrect action. PLoS ONE, 9(6).
https://doi.org/10.1371/journal.pone.0101126
28
Abnormalities of confidence in psychiatry: an overview and future perspectives. Translational
Psychiatry, 9(1). https://doi.org/10.1038/s41398-019-0602-7
Kepecs, A., Uchida, N., Zariwala, H. A., & Mainen, Z. F. (2008). Neural correlates , computation and
behavioural impact of decision confidence. 455. https://doi.org/10.1038/nature07200
Kessler, R. C., Berglund, P., Demler, O., Jin, R., Merikangas, K. R., & Walters, E. E. (2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the national comorbidity survey replication. Archives of General Psychiatry, 62(6), 593–602.
https://doi.org/10.1001/archpsyc.62.6.593
Kim, Y., Kim, T. H., & Ergün, T. (2015). The instability of the Pearson correlation coefficient in the presence of coincidental outliers. Finance Research Letters, 13, 243–257.
https://doi.org/10.1016/j.frl.2014.12.005
Knutson, B., & Greer, S. M. (2008). Anticipatory affect: Neural correlates and consequences for choice. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1511), 3771– 3786. https://doi.org/10.1098/rstb.2008.0155
Lak, A., Costa, G. M., Romberg, E., Koulakov, A. A., Mainen, Z. F., & Kepecs, A. (2014). Orbitofrontal Cortex Is Required for Optimal Waiting Based on Decision Confidence. Neuron, 1–12.
https://doi.org/10.1016/j.neuron.2014.08.039
Lebreton, M., Abitbol, R., Daunizeau, J., & Pessiglione, M. (2015). Automatic integration of confidence in the brain valuation signal. Nature Neuroscience, 18(8), 1159–1167. https://doi.org/10.1038/nn.4064
Lebreton, M., Jorge, S., Michel, V., Thirion, B., & Pessiglione, M. (2009). An Automatic Valuation System in the Human Brain: Evidence from Functional Neuroimaging. Neuron, 64(3), 431–439. https://doi.org/10.1016/j.neuron.2009.09.040
Lebreton, M., Langdon, S., Slieker, M. J., Nooitgedacht, J. S., Goudriaan, A. E., Denys, D., van Holst, R. & Luigjes, J. (2018). Two sides of the same coin: Monetary incentives concurrently improve and bias confidence judgments. Science advances, 4(5), eaaq0668. DOI: 10.1126/sciadv.aaq0668 Luigjes, J. (2015). Out of control: Losing oneself in compulsivity. 's-Hertogenbosch: Boxpress. Luigjes, J., Fontenelle, L. F., Youssef, G., Greenwood, C., Guo, K., Denys, D., and Yucel, M. (2016).
Negative outcome expectancy promotes compulsive behaviour.
29
outcome expectancy promotes compulsive behaviour. European Neuropsychopharmacology, 26, s629-s630. https://doi.org/10.1016/S0924-977X(16)31722-9
Luigjes, J., Lorenzetti, V., de Haan, S., Youssef, G. J., Murawski, C., Sjoerds, Z., Van den Brink, W., Denys, D., Fontenelle, L. F. & Yücel, M. (2019) Defining Compulsive Behavior. Neuropsychol
Rev 29, 4–13 https://doi.org/10.1007/s11065-019-09404-9
Luttrell, A., Briñol, P., Petty, R. E., Cunningham, W., & Díaz, D. (2013). Metacognitive confidence: A neuroscience approach. Revista de Psicología Social, 28(3), 317–332.
https://doi.org/10.1174/021347413807719148
MacDonald, P. A., Anthony, M. M., Macleod, C. M. & Richter, M. A. (1997). Memory and confidence in memory judgments among individuals with obsessive compulsive disorder and non-clinical controls. Behaviour research and therapy 35(6), 497–505.
https://doi.org/https://doi.org/10.1016/S0005-7967(97)00013-2
Maniscalco, B., & Lau, H. (2012). A signal detection theoretic approach for estimating metacognitive sensitivity from confidence ratings. Consciousness and Cognition, 21(1), 422–430.
https://doi.org/10.1016/j.concog.2011.09.021
Mataix-Cols, D., Do Rosario-Campos, M. C., & Leckman, J. F. (2005). A multidimensional model of obsessive-compulsive disorder. American Journal of Psychiatry, 162(2), 228–238.
https://doi.org/10.1176/appi.ajp.162.2.228
Mazaika, P., Hoeft, F., Glover, G., & Reiss, A. (2009). Methods and Software for fMRI Analysis for Clinical Subjects. Neuroimage.
McNally, R. J., & Kohlbeck, P. A. (1993). Reality monitoring in obsessive-compulsive disorder.
Behaviour Research and Therapy, 31(3), 249–253.
https://doi.org/10.1016/0005-7967(93)90023-N
Moritz, S., Jaeger, A., & Moritz, S. (2018). Decreased memory confidence in obsessive – compulsive disorder for scenarios high and low on responsibility : is low still too high ? European Archives of
Psychiatry and Clinical Neuroscience, 268(3), 291–299.
https://doi.org/10.1007/s00406-017-0783-0
Nakao, T., Okada, K., & Kanba, S. (2014). Neurobiological model of obsessive-compulsive disorder: Evidence from recent neuropsychological and neuroimaging findings. Psychiatry and Clinical
Neurosciences, 68(8), 587–605. https://doi.org/10.1111/pcn.12195
