To see or not to see: An fMRI study about curiosity for valenced stimuli

MSc Brain and Cognitive Sciences

Track Cognitive Neuroscience

Research Report

To see or not to see:

An fMRI study about curiosity for

valenced stimuli

by

October 2017

26 EC

05.01.2017 – 17.07.2017

Supervisor/Examiner:

1

Content

Abstract ... 2

Introduction ... 3

Curiosity vs. morbid curiosity? ... 3

Brain basis of curiosity ... 4

Brain basis of morbid curiosity ... 5

Choice behaviour ... 5

The present study... 6

Materials and methods ... 8

Participants ... 8

Stimulus materials ... 8

Task ... 9

Scan parameters and preprocessing ... 11

Statistical testing... 11 Results ... 12 Behavioural results ... 12 ROI analysis ... 13 Whole-brain analysis ... 14 Conjunction Analysis ... 16 Discussion ... 17 Limitations ... 21 Conclusion ... 21 References ... 22

2

Abstract

Curiosity is a common phenomenon which we experience in our daily life. However, affective neuroscience so far mostly studied curiosity for rather positive stimuli and has not taken into consideration the fact that humans are curious for positive as well as negative events. Hence, little is known about morbid curiosity, i.e. curiosity for events which involve harm, violence or death. We used functional magnetic resonance imaging (fMRI) to examine the underlying brain

mechanisms of curiosity for differently valenced stimuli, especially for negative stimuli.

Curiosity was manipulated by letting participants decide themselves whether they wanted to see a visual stimulus or not. Our results showed that (1) the induction of curiosity for negative stimuli, through presenting descriptions of images, activated bilateral striatum as well as bilateral

orbitofrontal cortex (OFC) and inferior frontal gyrus (IFG), brain regions associated with reward mechanisms; (2) the induction of curiosity for negative and positive stimuli did not significantly differ from each other; (3) the induction of curiosity for both, negative and positive stimuli, activated the insular cortex as well as the paracingulate gyrus, regions associated with salience and (negative) arousal processing. The results provided further evidence that curiosity for negative and positive stimuli share brain activation patterns which involve reward mechanisms.

3

Introduction

Curiosity drives our behaviour in many daily situations. People are curious about the latest sports news or the weather forecast. However, we can also be curious about rather negative events. Imagine you are reading a headline about a natural catastrophe, crime or accident. Eventually, you might become curious and read the whole article to gain more information about the event. Emotional responses to affective stimuli are often studied by presenting pictures to participants. Nevertheless, the underlying neural mechanisms of curiosity for affective stimuli - especially negative valenced stimuli, i.e. morbid curiosity - is a strikingly understudied phenomenon in affective neuroscience. Using fMRI enables us to investigate those underlying neural processes as well as compare and identify possible differences between curiosity for positive and negative stimuli. We present a novel behavioural choice paradigm to investigate the neural mechanisms of curiosity for differently valenced stimuli. The present study will give new insights into the neural mechanisms of curiosity for negative stimuli, a topic mostly ignored in current neuroscience, which has mainly focused on a link between emotions like fear and disgust, and negative stimuli. In addition, the current paradigm provides the opportunity to explicitly compare morbid, i.e. curiosity for negative stimuli, and curiosity for positive stimuli, which has never been done before to our knowledge. Finally, the present study provides new insights into the neural representation of curiosity, suggesting an overlap between curiosity for negative and positive stimuli, as well as the potential involvement of neural reward mechanisms during the experience of curiosity.

Curiosity vs. morbid curiosity?

Curiosity is a common phenomenon which we experience in our daily life. Litman (2005) describes curiosity as a desire to gain new knowledge and the underlying motivation for exploratory behaviour. Although curiosity is a basic biological drive which influences human behaviour, it is a rather unexplored phenomenon in psychological science in general (Jepma, Verdonschot, Van Steenbergen, Rombouts, & Nieuwenhuis, 2012). Especially, the neural mechanisms of curiosity are not well understood, yet (Kidd & Hayden, 2015).

Morbid curiosity refers to seeking information which involves death or violence

4

daily life (Rimé, Delfosse & Corsini, 2005; Garcia-Garcia et al., 2016). However, research does not take into consideration that people can be curious about negative stimuli (Oosterwijk, 2017) and leaves morbid curiosity as a strikingly understudied phenomenon. The current fMRI study will explore this phenomenon for the first time in a neuroimaging context using a validated paradigm (Oosterwijk, 2017).

Even though curiosity and morbid curiosity seem different at first glance, these processes may not be clearly separate. The main reason is that the behavioural tendencies underlying curiosity for positive and negative information seem quite similar. Loewenstein’s (1994)

information gap theory is useful for explaining this similarity. According to Loewenstein (1994), curiosity is an intrinsic motivation aimed at gaining specific information. He specifies that a gap of information is defined by what an individual already knows and what an individual wants to know. In other words, if an individual focuses on a missing piece of information, curiosity for this missing information can evolve. This mechanism eventually leads to the exploration of a certain stimuli. Although Loewenstein does not mention curiosity for negative stimuli

specifically, this mechanism might underlie both, curiosity for negative and positive stimuli. Since both types of curiosity share behavioural patterns, i.e. the exploration of a stimulus, it is important to consider the current state of literature concerning brain processes of curiosity as well as morbid curiosity.

Brain basis of curiosity

Several studies have specifically focused on patterns of neural activity when people are curious. First, Jepma et al. (2012) showed participants blurred and clear pictures of objects. In their design, they distinguished between the induction and the relief of curiosity. Curiosity was induced by letting participants view ambiguous stimuli. On the other hand, the relief of curiosity was achieved by disambiguate the visual stimuli. During the induction of curiosity, the anterior insula as well as the anterior cingulate cortex (ACC) showed increased activation, whereas the striatum was significantly activated during the relief of curiosity (Jepma et al., 2012). In another fMRI study, participants read trivia questions and rated their feelings of curiosity (Kang et al., 2005). The researchers found significant activation of the left caudate, bilateral prefrontal cortex (PFC) and inferior frontal gyrus (IFG), regions which show that the anticipation of reward might be involved during the induction of curiosity (Kang et al., 2005). In addition, Gruber et al. (2014) let participants read trivia questions and rate them how likely they know the answer as well as

5

their curiosity for the answer during a screening phase outside the scanner. During the actual scan phase, questions which were previously rated with low or high curiosity, were presented to the participants followed by a picture of a face about which participants had to made an incidental judgment. Gruber et al. (2014) found increased activation in the dorsal and ventral striatum as well as the IFG during the induction of curiosity. These studies provide evidence that the

striatum, a subcortical region which is strongly associated with reward (Jepma et al., 2012; Kang et al., 2009; Gruber, Geiman & Ranganath, 2014) is activated during the experience of curiosity. Since curiosity for negative and positive stimuli share behavioural patterns, this region might also be involved in curiosity for negative stimuli. It is important to note, however, that none of these studies included valenced stimuli.

Brain basis of morbid curiosity

A neuroimaging study that examined curiosity (or fascination) regarding negative stimuli was performed by Oosterwijk, Lindquist, Adebayo and Barrett (2016). In their study, participants were presented with negative images in combination with a false feedback manipulation with different emotion labels namely fear, disgust and morbid fascination (Oosterwijk et al., 2016). In 70% of the cases, participants agreed with the false feedback given by the researchers. The neuroimaging results demonstrated differential activity when people labelled their response to a negative image as disgusted, fearful or fascinated. The study showed increased activity in the ventrolateral prefrontal cortex (vlPFC), dorsomedial prefrontal cortex (dmPFC) and lateral orbitofrontal cortex (lOFC) when comparing fascination to control feedback, and to the other emotion feedback conditions. The authors interpreted these patterns as evidence for enhanced engagement of meaning making processes, or top-down interpretative processes, when people are fascinated by a negative image. However, participants did not engage in exploratory behaviour in this study, i.e. choosing to explore a stimulus.

Choice behaviour

Choice behaviour and decision making are crucial components of curiosity. However, this component is often missed and instead, self-reports are commonly used to measure curiosity (Kang at al., 2005; Gruber et al., 2014). Recently, studies have started to use behavioural

6

measurements to gain more insight in curiosity for negative stimuli. Hsee and Ruan (2016) studied curiosity for potential aversive stimuli. In their study, participants were told that they could press any of 10 pens provided by the researchers. In the certain condition, pens were marked red or green and the participants were told that they would receive an electrical shock when pressing a red marked pen, whereas green marked pens were not dangerous. In the uncertain condition, all pens were marked yellow and the participants were told that some pens would deliver electrical shocks while others would not. Surprisingly, participants clicked more pens in the uncertain condition as well as pens which would certainly shock them. Hsee and Ruan (2016) also showed that participants would show more curiosity for unpleasant sounds as well as negative images (insects). In another study, Oosterwijk et al. (2017) let participants choose between two images (one positive and one negative) based on two descriptions. They found that participants chose more often for negative social stimuli, compared to neutral social images and equally often as positive social stimuli. However, even though these studies provide some understanding concerning the importance of choice behaviour for curiosity, more research is needed to gain more insights in the underlying neuronal basis of curiosity which leads to exploratory behaviour for negative stimuli.

The present study

In this study, we aimed to examine the underlying brain mechanisms of curiosity for differently valenced stimuli. Participants were assigned to one of two conditions. They either engaged in a passive viewing task in which they read descriptions and saw the corresponding images or a blurred version of the image (passive condition), or they read descriptions and actively chose to view or not to view the corresponding image (active choice condition).

Following previous work by Jepma et al. (2012), we aimed to compare neural activity for curiosity induction (assuming that curiosity drives the choice to view a stimulus based on a description, cf. Oosterwijk, 2017) with a control condition in which people could not follow their curiosity (passive viewing). This comparison is interesting, because Baumeister, Bratslavsky, Finkenauer and Vohs (2001) argue that people engage in more cognitive processing while being confronted with negative events compared to positive events. Thus, participants might have elaborated more while reading negative descriptions in the active choice paradigm, which offered

7

them volitional control concerning the exploration of the stimuli, compared to the passive viewing condition. We examined the following question regarding the comparison between active choice and passive viewing: How does brain activity during normal anticipation of a negative stimuli (passive viewing) differ from brain activity during curiosity about a negative description when people actively choose to view the image?

By using positive as well as negative images we had the opportunity to compare curiosity for different types of valenced stimuli. This allowed us to examine whether the brain response is different when making the decision to view negative stimuli compared to the decision to view positive stimuli.

We compared negative descriptions associated with active choice (choice condition) with negative descriptions without the possibility for choice (passive condition). In the active

condition, people might have anticipated to gain novel information when they process descriptions associated with an active choice to view. Therefore, we expected significant activation in regions associated with anticipated reward (striatum) when people viewed a description that resulted in the choice to view a negative stimulus (as compared to passive viewing) (Kang et al., 2009).Based on this the following a-priori hypotheses were preregistered: We expected more activation in the striatum when participants view a negative description that has been chosen (choice condition) as compared to passively view a negative description. In addition, this increase in activation due to choice (choice vs. passive condition) was expected to be stronger for a negative than for a positive descriptions. Moreover, it was hypothesized that there would be more activation in the IFG when people see a negative description due to choice (active condition) compared to the passive viewing condition. This activation was expected to be stronger for negative than for positive stimuli.

Concerning the whole-brain analysis we predicted and preregistered the following activations: During the induction of curiosity, we expected activation of the ACC and the AIC, regions associated with salience processing (Jepma et al., 2012). In addition, we expected the engagement of the so-called default network, associated with meaning-making processes and self-reflection, in the active choice condition (as compared to passive viewing). This especially involves the mPFC and PCC, previously shown by Oosterwijk et al. (2016) when participants received information that they engage in an internal state of morbid fascination. We expected stronger activation in the active choice condition compared to the passive condition.

8

Materials and methods

Participants

Sixty participants (16 M, 44 F, M age = 22.8, SD age = 2.35) were recruited from the University of Amsterdam through an online recruitment website for participants. Potential participants were excluded if they were using psychoactive medication or were diagnosed with a psychiatric illness at that time. Other exclusion criteria included a disfavor to look at shocking affective material, not having Dutch as first language, a previous participation in the online study “How do you evaluate descriptions of photos” and not having a general practitioner in the Netherlands.

Additionally, we preregistered that participants would be excluded if they would choose to view negative (or positive) images in less than 40 % of the trials (14 out of 35 images). This was necessary to prevent a filling problem, i.e. comparing e.g. the brain activation of chosen negative descriptions, which contains considerably less trials (e.g. 13), with brain activation of chosen positive descriptions, which contains considerably more trials, (e.g. 35). Participants gave informant consent according to the Spinoza Centre for Neuroimaging. One participant was excluded because she felt uncomfortable in the scanner, and five participants were excluded due to their choice patterns (<14 choices to view negative stimuli). Analyses were performed on the remaining fifty-four participants (16 M, 38 F, M age = 22.8, SD age = 2.27). Participants were paid €20 for their participation.

Stimulus materials

The descriptions of the images were written by the researchers. On average, descriptions contained 60 characters. Images were selected from the International Affective Picture System database (IAPS; Lang, Bradley & Cuthbert, 2008) and the Nencki Affective Picture System (NAPS; Marchewka, Zurawski, Jednoróg & Grabowska, 2014) based on their valence scores on a Likert scale from 1 to 9. The negative stimuli had a mean valence rating below 4. Positive images had a mean valence rating above 6. Beforehand, all selected stimuli and descriptions were tested in a pilot study (n=20) to match the valence of the descriptions and images. Participants were asked to rate the descriptions on two scales from 0 (very negative) to 100 (very positive) and from 0 (very relaxed) to 100 (very tense). In addition, the images were presented to the participants and they were asked if they experience the image as more positive, negative or equally valenced as the corresponding description (0 = more negative; 50 = equal; 100 = more

9

positive). Based on the pilot study, some images and descriptions were excluded because images were experienced considerably more negative than the corresponding descriptions.

Task

The experimental task was programmed in the stimulus software Presentation (Neurobehavioral Systems, Inc., Berkeley, CA, www.neurobs.com). Beforehand, participants received information that the study contained images which could be experienced as aversive, shocking or unpleasant. The study was reviewed by the Ethical Review Board of the department of Psychology of the University of Amsterdam. Both the active choice task and the passive viewing task consisted of 70 trials. In order to separately model the processing of the description, the choice moment, and the viewing of the corresponding image, the different phases in each trial were separated by jittered interstimulus- intervals (500-2000ms).

In the active choice task, each trial started with a fixation cross presented for 500 ms, followed by a description of a stimulus presented in the middle of the screen for 3000 ms. After that, there was a jittered inter-stimulus interval of 500-2000ms. Then, participants decided if they want to see the corresponding image or not by pressing one of two buttons on a remote control, corresponding to a 'yes' or 'no' (2000ms), followed by a jittered ISI of 500-2000ms. If the participant had chosen 'yes', he/ she saw the corresponding stimulus on the screen for 3000 ms. If the participants had chosen 'no' he/ she saw a blurred version of the image for 3000 ms followed by an inter-trial interval of 2000-4000ms (Figure 1). Images were blurred in the software Irfan View (version 4.44) using the fast Gaussian blur (filter = 150 pixels).

In the passive viewing task, each trial started with a fixation cross for 500ms followed by the presentation of the description for 3000ms. After that, there was a jittered inter-stimulus interval of 500-2000ms. Then, participants were presented with a screen which indicated if they would see the image or not (‘yes’ or ‘no’ lighted up in green) and were asked to confirm this by pressing ‘yes’ or by pressing ‘no’ (2000ms), followed by a jittered ISI of 500-2000ms. After that, either the corresponding stimulus or a blurred version was presented to the participant for

3000ms. Participants were presented with a description / image sequence of a previous

participant from the active choice condition. Beforehand, they were told that the computer would choose for them if they would see the image or not. After the experiment, participants were enlightened that they saw a description / image sequence of a previous participant. In both

10

conditions, the different trials were separated with an ITI of 2000-4000ms (Figure 1). The setup of the study allowed us to examine the neural activity when people read a description that resulted in an active choice to view a stimulus, with the neural activity when people read a description without the possibility to choose. Furthermore, by letting participants view a description before they viewed the corresponding image or take the decision to view or not to view the image, we had the possibility to investigate the induction of curiosity (i.e. reading the description) as well as the relief of curiosity (viewing the image). However, this report will focus on the induction of curiosity. The only difference between the two conditions was the choice component in the active condition. For the rest, the two conditions were completely identical. Therefore, the paradigm controlled for possible differences of semantic processing between the two groups.

After the scan session, participants were asked to fill in the Interpersonal Reactivity Index (IRI), the Morbid Curiosity in Daily Life questionnaire as well as a questionnaire about

demographics. There were no hypotheses about the questionnaires. In addition, participants were asked how curious they felt about the positive and negative stimuli. These questions served as a manipulation check.

Figure 1 | Overview of task design. In the active condition, participants could choose whether they want to see an image or not.

In the passive condition, participants viewed previously selected sequences of participants from the active choice condition. They were told that the computer would select for them.

11

Scan parameters and preprocessing

Blood oxygenation level-dependent (BOLD) sensitive fMRI scans were acquired with a Philips Achieva (Spinoza Centre for Neuroimaging, University of Amsterdam, The Netherlands) 3.0 Tesla scanner (T2*-weighted echo planar imaging (EPI) scans, repetition time(TR) = 2000 ms, echo time (TE) = 27.63 ms, field of view (FOV) 240mm x 118.5 mm x 240mm, scan resolution = 80mm x 78mm, EPI factor = 39, number of slices = 36, slice thickness = 3mm, interslice gap = 0.3mm, flip angle = 76.1 °). For anatomical reference, high resolution T1-weighted images (TR = 8ms, TE = 3.79, FOV = 240mm x 220mm x 188mm, scan resolution = 240mm x 240mm, slice thickness = 1mm, flip angle = 8°) were acquired.

Functional and structural neuroimaging data were preprocessed with FMRIB’s Software Library (FSL). Registration to standard space was done with FLIRT (2mm MNI 152 standard space) and refined with FNIRT using non-linear registration. The preprocessing pipeline

consisted of a non-brain removal using BET with a threshold of 0.3, motion correction using MC Flirt, slice time correction (Fourier-space time-series phase shifting), spatial smoothing (5mm full-width at half maximum (FWHM) Gaussian kernel) and high-pass temporal filtering (sigma = 64; 1/128 Hz). Grand mean intensity normalization of the entire 4D dataset was done with a single multiplicative factor. Prewhitening was done using FILM with local autocorrelation correction.

Statistical testing

To test our hypotheses, we focused on two a-priori defined regions of interest (ROIs). Based on the ROIs two masks were created using FSL and the Harvard-Oxford Subcortical Atlas (striatum; caudate + putamen + nucleus accumbens) and the Harvard-Oxford Cortical Atlas (IFG; pars opercularis + pars triangularis). These masks are probabilistically liberal and incorporate neighbour regions. A non-parametric group-level analysis was performed on the preprocessed data using FSL’s randomise command in combination with Threshold-Free Cluster Enhancement (TFCE; Smith & Nichols, 2009; FWE-corrected at p < 0.025; Table 1).

Furthermore, a whole brain analysis was performed using FSL’s FEAT higher level analysis (z > 2.6; Table 1). In addition, a conjunction analysis was done to gain insight into a possible overlap in activation between the two contrasts negative active > negative passive and positive active > positive passive (Table 1).

12

Table 1 ROIs and contrasts associated with the hypotheses and contrasts for exploratory analysis

Confirmatory analyses

ROIs Hypotheses (both for induction and relief phase)

Dorsal striatum negative active > negative passive

(negative active > negative passive) > (positive active > positive passive)

Ventral striatum negative active > negative passive

(negative active > negative passive) > (positive active > positive passive)

IFG negative active > negative passive

(negative active > negative passive) > (positive active > positive passive)

Exploratory contrasts

positive active > positive passive

(negative active > negative passive) ∩ (positive active > positive passive) (negative active > negative passive) < (positive active > positive passive)

Notes. Analysis of ROIs is based on a non-parametric TFCE-based test, FWE-corrected at p < 0.025, cluster-wise; and a FEAT

higher-level analysis (z >2.6). Analysis of exploratory contrasts is based on a non-parametric TFCE-based test (cluster-wise p < 0.05, FWE-corrected).

Results

Behavioural results

We found no significant difference on the subscale perspective of the IRI between the active group (M=4.98, SD=0.88) and the passive group (M=5.14, SD=0.88), t (52) =-0.66, p=0.51. There was no significant difference on the subscale fantasy between the active group (subscale fantasy M=4.87, SD=0.98) and the passive group (M=5.21, SD=1.26), t (52) =-1.14, p=0.26. Furthermore, no significant difference between the active group (M=4.95, SD=0.75) and the passive group (M=4.91, SD=0.80) was found on the subscale concern, t (52) =0.2, p=0.82. No significant difference between the active group (M=3.17, SD=0.98) and the passive group (M=3.21, SD=1) was found on the subscale distress, t (52) = -0.12, p=0.91. In addition, the two groups did not differ significantly from each other on the Morbid Curiosity in Daily Life

questionnaire (active M=5.01, SD=0.59; passive M=4.68, SD=0.72), t (52) =1.89, p=0.65. Active (M=4.72, SD=1.82) and passive group (M=4.44, SD=1.60) did not differ significantly from each other on their ratings for curiosity for positive stimuli, t (50) =0.58, p=0.56. There was a

13

significant difference between the active group (M=6.2, SD=0.71) and the passive group (M=5.41, SD=1.28) on their ratings for curiosity for negative stimuli, t (50) =2.74, p=0.009. Participants chose to view the negative stimuli in 80% of the cases (M=28.22), SD=5.91). In 94% of the cases, participants chose to view the positive stimuli (M=33.19, SD=4.88).

ROI analysis

We ran a non-parametric group-level analysis on the first level parameter estimates using FSL’s ‘randomise’ combined with Threshold-Free Cluster Enhancement (TFCE; Smith &

Nichols,2009), FWE-corrected at p < 0.025 on two a-priori determined ROIs.

Our first hypothesis was that there is more activation in the striatum when participants viewed a negative description in the active choice condition (negative active) compared to a negative description viewed in the passive viewing condition (negative passive). Our ROI analysis of the striatum (i.e. caudate + putamen + nucleus accumbens) showed significant activation in the left (caudate and putamen) and right (caudate and nucleus accumbens) striatum (Figure 1; Table 1). This increase in activation was expected to be stronger for negative than for positive descriptions. The ROI analysis did not show any significant clusters for this contrast (negative active > negative passive) > (positive active > positive passive).

Secondly, we hypothesized that there is more activation in the IFG when people viewed a negative description in the active choice condition (negative active) compared to view a

description in the passive viewing condition (negative passive). The ROI analysis of the IFG (pars opercularis + pars triangularis) showed significant activation in the OFC (bilateral) as well as left and right IFG, pars triangularis (Figure 1; Table1). The used masks are probabilistically liberal. Therefore, the IFG mask incorporated neighbour regions such as the OFC. The increase in activation was expected to be stronger for negative than for positive descriptions. The ROI

analysis did not show any significant clusters for this contrast (negative active > negative passive) > (positive active > positive passive).

Table 2 Significant clusters from the ROI analysis

Region Left/Right Cluster Size x y z

14 17% Caudate 92% Caudate 3% Nucleus accumbens Right 679 14 18 0 38% Frontal Orbital Cortex 10% Frontal Pole

1% Inferior Frontal Gyrus (pars triangularis)

Right 1931 38 30 -8

40% Frontal Orbital Cortex

8% Insular

6% Inferior Frontal Gyrus (pars triangularis)

Left 1828 -30 30 0

Notes: Table shows significant clusters (p < 0.025, FWE corrected) within ROI analysis based a non-parametric TFCE-based test

during the induction of curiosity for the contrast negative active > negative passive.

Figure 2 | Striatum / IFG activation associated with induction of curiosity. The colored regions were more active when people

viewed a negative description in the active choice condition (negative active) compared to viewing a negative description in the passive viewing condition (negative passive). R = right; L = left; A = anterior; P = posterior; S = superior; I = inferior. The shown activations are thresholded t-stat images from the non-parametric TFCE-based test (p < 0.025, FWE corrected). The red color shows activation in the striatum. The blue color shows activation in the OFC and IFG.

Whole-brain analysis

In addition to the non-parametric TFCE-based test, we conducted a whole-brain higher level analysis using FSL’s FEAT with a cluster-wise threshold of z > 2.6. For the first contrast, negative active > negative passive, we found significant activation in the OFC. Additionally, significant activation was found in the paracingulate gyrus, superior frontal gyrus, middle

15

temporal gyrus and temporal occipital fusiform gyrus (Table 3). A whole-brain analysis of positive active > positive passive contrast revealed significant activation in the insular cortex (bilateral), paracingulate gyrus and precentral gyrus (Table 3). We predicted that the activation would be stronger for negative than for positive descriptions. The whole-brain analysis did not show any significant clusters for this contrast (negative active > negative passive) > (positive active > positive passive).

Table 3 Significant clusters in whole brain contrast analysis (FEAT)

Region Left/Right Cluster Size x y z

Induction of curiosity: negative active > negative passive

Frontal Orbital Cortex right 10187 34 28 -4

Paracingulate Gyrus right 4967 8 14 46

Superior Frontal Gyrus Left 278 -22 -4 52

Middle Temporal Gyrus Left 254 -64 -58 10

Temporal Occipital Fusiform Cortex Left 224 -44 -48 -30 Temporal Occipital Fusiform Cortex Left 214 -34 -50 -22

Induction of curiosity: positive active > positive passive

Insular cortex right 766 30 24 -4

Paracingulate Gyrus right 619 6 12 48

Insular Cortex Left 288 -40 14 -10

Precentral Gyrus Right 253 42 0 38

Notes: Table shows significant clusters (z > 2.6, corrected) within whole-brain analysis based on FEAT higher level analysis

during the induction of curiosity for the contrast negative active > negative passive and the exploratory contrast positive active > positive passive.

Figure 3 | Whole-brain activation associated with induction of curiosity. The colored regions were more active when people

viewed a negative description in the active choice condition (negative active) compared to view a negative description in the passive viewing condition (negative passive). R = right; L = left; A = anterior; P = posterior; S = superior; I = inferior. The shown activations are thresholded z-stat images from the higher-level FEAT whole-brain analysis (z > 2.6, corrected).

16

Figure 4 | Whole-brain activation associated with induction of curiosity. The colored regions were more active when people

viewed a positive description in the active choice condition (positive active) compared to view a positive description in the passive viewing condition (positive passive). R = right; L = left; A = anterior; P = posterior; S = superior; I = inferior. The shown activations are thresholded z-stat images from the higher-level FEAT whole-brain analysis (z > 2.6, corrected).

Conjunction Analysis

Finally, a conjunction analysis for the two significant contrasts negative active > negative passive and positive active > positive passive with a cluster-wise threshold of z > 2.6 was conducted. The conjunction analysis revealed clusters which were significantly active for both contrasts.

Especially, there was significant activation in the right insular cortex and the right paracingulate gyrus for both contrasts (Table 4; Figure 3).

Table 4 Conjunction analysis negative active > negative passive ∧ positive active > positive passive

Region Left/Right Cluster Size x y z

Insular Cortex Right 992 30 24 -4

Paracingulate Gyrus

Right 685 6 12 48

Notes. Table shows significant clusters (z > 2.6, corrected) of the conjunction analysis of the two contrasts negative active >

17

Figure 5 | Conjunction analysis negative active > negative passive ∧ positive active > positive passive. The colored regions

showed overlap in cluster activation for the two contrasts negative active > negative passive and positive active > positive passive. R = right; L = left; A = anterior; P = posterior; S = superior; I = inferior. The shown activations are thresholded z-stat images from the conjunction analysis in FSL (z > 2.6).

Discussion

The present study used a novel active choice paradigm to investigate the neural representation of curiosity for differently valenced stimuli and whether curiosity for negative and positive stimuli differ from each other. The crucial component of decision making was incorporated in the

research paradigm. This allowed us to study (morbid) curiosity behaviorally instead of only using commonly applied self-report measurements.

Participants showed a solid curiosity for negative stimuli on a behavioral level since they chose to view negative stimuli in 80 % of the cases (94% for positive stimuli). Additionally, the ratings for curiosity for negative stimuli showed that participants in the active choice condition were indeed curious for those stimuli compared to the passive condition. These results indicate that participants experienced curiosity for the negative stimuli in the active condition and that their choice behavior reflected this curiosity.

The first aim of the study was to determine whether brain responses during the anticipation of a negative stimulus when actively choosing to view the image vs. passively anticipating the negative stimulus differ from each other. The results showed that there was a significant difference in brain responses concerning the difference during the anticipation of a chosen negative stimulus vs. passively anticipating a negative stimulus. The ROI analysis showed peak cluster activation in the left and right striatum as well as the left and right OFC and IFG, pars triangularis.

18

The second aim of the present study was to investigate whether there are different brain responses when making the decision to view a negative stimulus vs. making the decision to view a positive stimulus. The results showed that there was no significant difference in brain responses between making the decision to view a negative stimulus vs. making the decision to view a positive stimulus. However, we found significant overlap between those two conditions in the insular cortex and the paracingulate gyrus.

The ROI analysis revealed significant activation of the striatum as well as the IFG, which is in line with previous research by Jepma et al. (2012), Kang et al. (2009) and Gruber et al. (2014). These activation patterns indicate an involvement of reward mechanisms during the induction of curiosity for negative stimuli. Thus, participants might have anticipated reward in the form of new information while reading the descriptions. This is in line with Loewenstein (1994) who describes curiosity as a gap of knowledge as well as Litman (2005) who explains this desire of knowledge as a motivation for explorative behaviour. That the participants in the active choice condition complied with this motivation of actively seeking new information (viewing the negative image), is a possible explanation of the found brain response patterns. This suggestion is in line with the subjective rating for curiosity for negative stimuli in the active choice condition. Participants stated that they were curious for the negative stimuli in the active choice condition. To sum up, the results indicate that participants chose to view the negative stimuli out of their curiosity and anticipated reward in the form of new information about the negative stimuli.

Furthermore, we found significant activation in the IFG, pars triangularis, as well as the OFC. This is in line with Oosterwijk et al. (2016), who also found activation in the OFC and the IFG (labelled as vlPFC in their paper) when people were told that they are morbidly fascinated. The OFC is associated with conceptualization and a part of the so-called default mode network. Compared to Oosterwijk et al. (2016) who gave false feedback to their participants, participants in the present study experienced a real state of morbid curiosity. This was indicated by subjective ratings for curiosity for negative stimuli as well as choice patterns. The activation of the OFC, associated with the default mode network, suggests that participants might have engaged in meaning making processes while deciding to view or not to view a negative image. Additionally, the stronger activation of the IFG in the active choice condition compared to the passive viewing condition suggests the involvement of reward mechanisms (Kang et al., 2005). Especially, participants might have anticipated reward in the form of new details about the negative stimuli

19

while making the decision to view the stimuli, i.e. during the induction of curiosity. To sum up, the subjective ratings and behavioural patterns indicate that participants experienced curiosity when they chose to view a negative stimulus. The state of morbid curiosity was reflected in neural patterns that suggest the involvement of the anticipation of reward as well as meaning making processes when participants engaged in exploratory behaviour, i.e. choose to view a negative stimulus.

In addition to the ROI analysis, a whole-brain analysis was conducted. We found significant activation in the OFC. However, this cluster contained more than 10000 voxels. Therefore, the activation of the OFC possibly extended into other areas such as the insula and might also involve areas such as the IFG, pars triangularis and parts of the medial prefrontal cortex. This activation pattern is another indication that participants might have anticipated reward during the induction of curiosity. One might argue that participants expected reward in the form of new information to comply to their desire for new knowledge (cf. Litman, 2005) invoked by their curiosity.

In addition, we found significant brain activation in the left superior frontal gyrus (SFG), an area which has been shown to play a role in working memory, especially monitoring and executive processing (Boisgueheneuc, 2006). Therefore, participants might have carefully thought about whether to view the negative image or not, i.e. ‘monitoring’ the situation. Nevertheless, associating brain regions with specific functions should be done cautiously.

Regarding curiosity for positive stimuli, significant brain activation was found in the bilateral insular cortex which is consistent with Jepma et al. (2012), who found activation in the anterior insular cortex during the induction of curiosity. Additionally, significant activation was found in the paracingulate gyrus. This area was also found to be significantly active during the induction of curiosity for negative stimuli. Since this area is a part of the ACC, this finding is also in line with Jepma et al. (2012). It provides evidence that salience processing and (negative) arousal might play a role during the induction of curiosity. Moreover, we found activation in the prefrontal gyrus when participants actively chose to view positive stimuli compared to passively view them. Oosterwijk et al. (2016) also found significant clusters in the precentral gyrus when people were told that they are fearful, disgusted or morbidly fascinated compared to control feedback. These activation patterns may indicate that curiosity for positive and negative stimuli do not differ significantly from each other.

20

Especially, we found no significant difference in activation when participants chose to view a negative image compared to a positive image. The ROI analysis already indicated that curiosity and morbid curiosity may share neural patterns. Jepma et al. (2012) and Kang et al. (2009) found similar neural representations of curiosity compared to the brain responses we found for curiosity for negative stimuli. But they did not study curiosity for valenced stimuli and used rather positive/neutral stimuli. When comparing the induction of curiosity for negative and positive stimuli, we found significant overlap in the insular cortex as well as the paracingulate gyrus. First, the paracingulate gyrus is a dorsal part of the anterior cingulate cortex. For example, Jepma et al. (2012) included this area in their brain mask for the ACC. Moreover, Jepma et al. (2012) found activation in the insula (i.e. AIC) and the ACC in their study. This is in line with our finding and suggests that salience, and emotion as well as (negative) arousal processing is crucial during the induction of curiosity. Jepma et al. (2012) explain further that activation in these brain regions is commonly found under aversive conditions and state that curiosity might invoke an unpleasant feeling. Hsee and Ruan (2016) showed that participants in their study showed more curiosity in the uncertain condition, in which it was not obvious to the participants which device would deliver an electrical shock. Not knowing which pen would deliver an electrical shock might have invoked curiosity and the participants showed more exploratory behaviour in this condition because they wanted to solve a gap of knowledge, i.e. which pen would shock them. In the present study, the unpleasant feeling could be as well the result of a gap of information (not knowing what the image looks like). Eventually, participants might have wanted to close this gap by choosing to view the stimuli. This explanation is in line with the activation of the striatum, which we found during the induction of curiosity in the active choice condition. Thus, participants were curious for more information about the stimuli and aimed at gaining more information by choosing to view the stimuli. Therefore, they could have anticipated reward, reflected in the activation of the striatum, in the form of new information about the stimuli. In conclusion, participants might have expected reward in the form of new information about a stimulus when choosing to view the image, regardless of the valence of the image. This would explain the surprising overlap in brain activation between curiosity for negative and curiosity for positive stimuli.

21

Limitations

As one of a few studies which investigates curiosity for negative stimuli, there are some limitations which need to be discussed. First, participants were not asked per trial if they were curious for the following image. Instead, they were asked after the scan session how curious they were for positive and negative stimuli. A reason to do so was to prevent possible confounding brain activity due to ask participants from trial-to-trial for their curiosity. Hence, we could not control for variations of curiosity for stimuli, a limitation also mentioned by Jepma et al. (2012) concerning their study. In addition, participants might have chosen to view stimuli out of boredom in the scanner rather than out of curiosity induced by the descriptions. However, subjective ratings for curiosity for negative stimuli indicated that participants in the active

condition were indeed curious for the negative stimuli. On the other hand, ratings for curiosity for positive stimuli did not differ between the active and the passive condition. Therefore, it might be the case that participants chose to view positive stimuli out of boredom whereas subjective ratings for negative stimuli counteract this assumption. Furthermore, curiosity might have been induced due to the uncertainty in the passive condition whether the participants would see the image or not, compared to the curiosity about the identity of an image in the active choice condition. For example, Kang et al. (2005) showed that uncertainty increases curiosity. Thus, participants in the passive condition might have experienced curiosity in a different way, which in turn might have led to different brain responses. However, this would only reduce the chance of finding a difference between the neural representation of actively choosing to view a negative stimulus versus passively viewing it, since both conditions would show brain processes involved in experiencing curiosity.

Conclusion

To summarize, the present study provides insights into the neural mechanisms which underlie curiosity for negative and positive stimuli. Our findings showed, that the induction of curiosity for differently valenced stimuli may share underlying brain responses, including activation of the striatum and the IFG. Therefore, curiosity and morbid curiosity may not be as different as one might think at first glance, since they not only share behavioral outcomes, i.e. exploratory behavior, but also brain processes. Especially, there is evidence that both might share (aversive) states of arousal, resulting out of a gap of information, and reward mechanisms.

22

References

Aldhafeeri, F. M., Mackenzie, I., Kay, T., Alghamdi, J., & Sluming, V. (2012). Regional brain responses to pleasant and unpleasant IAPS pictures: different networks. Neuroscience letters, 512(2), 94-98.

Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of general psychology, 5(4), 323.

Boisgueheneuc, F. D., Levy, R., Volle, E., Seassau, M., Duffau, H., Kinkingnehun, S. & Dubois, B. (2006). Functions of the left superior frontal gyrus in humans: a lesion study. Brain, 129(12), 3315-3328.

Garcia-Garcia, I., Kube, J., Gaebler, M., Horstmann, A., Villringer, A., & Neumann, J. (2016). Neural processing of negative emotional stimuli and the influence of age, sex and task- related characteristics. Neuroscience & Biobehavioral Reviews, 68, 773-793.

Gruber, M. J., Gelman, B. D., & Ranganath, C. (2014). States of curiosity modulate

hippocampus-dependent learning via the dopaminergic circuit. Neuron, 84, 486-496.

Hsee, C. K., & Ruan, B. (2016). The Pandora effect: The power and peril of curiosity. Psychological science, 27(5), 659-666.

Jepma, M., Verdonschot, R. G., Van Steenbergen, H., Rombouts, S. A., & Nieuwenhuis, S. (2012). Neural mechanisms underlying the induction and relief of perceptual curiosity. Memory and motivational/emotional processes, 100.

Kang, M. J., Hsu, M., Krajbich, I. M., Loewenstein, G., McClure, S. M., Wang, J. T. Y., & Camerer, C. F. (2009). The wick in the candle of learning: Epistemic curiosity activates reward circuitry and enhances memory. Psychological Science, 20, 963- 973.

Kidd, C., & Hayden, B. Y. (2015). The psychology and neuroscience of curiosity. Neuron, 88(3), 449-460.

Lang, P.J., Bradley, M.M., & Cuthbert, B.N. (2008). International affective picture system (JAPS): Affective ratings of pictures and instruction manual. Technical Report A- 8. University of Florida, Gainesville, FL.

23

Litman, J. (2005). Curiosity and the pleasures oflearning: Wanting and liking new information. Cognition & emotion, 19(6), 793-814.

Loewenstein, G. (1994). The psychology of curiosity: A review and reinterpretation. Psychological Bulletin, 116, 75–98.

Marchewka, A., Zurawski, L., Jednoróg, K., & Grabowska, A. (2014). The Nencki Affective Picture System (NAPS): Introduction to a novel, standardized, wide-range, highquality, realistic picture database. Behavior research methods, 46(2), 596-610.

Oosterwijk, S (2017). Choosing the negative: A behavioral demonstration of morbid curiosity. PLOSONE.

Oosterwijk, S., Lindquist, K.A., Adebayo, K., & Barrett, L.F. (2016). The neural representation of typical and a-typical experiences of negative images: Comparing fear, disgust and morbid fascination. Social, Cognitive and Affective Neuroscience, 11, 11-22.

Rimé, B., Delfosse, C., & Corsini, S. (2005). Briefreport. Cognition & Emotion, 19(6), 923-932.

Smith, S. M., & Nichols, T. E. (2009). Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage, 44, 83-98.

Zuckerman, M., & Litle, P. (1986). Personality and curiosity about morbid and sexual events. Personality and Individual Differences, 7(1), 49-56.