Influence of Observer’s Gender on Induction of Fear in Virtual Reality : A Quantum Cognitive Effect?

(1)

Influence of Observer’s

Gender on Induction of Fear

in Virtual Reality

A Quantum Cognitive Effect?

Linda Hes, 10627545 || Olga Pietras, 5965667 || Boudewijn Poot, 10686355

GER POST & MACHIEL KEESTRA

DOMAIN: COGNITION Number of words: 6518 words

(2)

ABSTRACT

It is important to know whether the observer’s gender influences the experience of fear, especially as this is an overlooked factor in fear-research, while experiments on the effects of the gender of the observer in other fields indicate its effects can be major. Therefore, this research investigated the influence of the observer’s gender on subjective and physiological experience of fear. Sixteen subjects, male and female, were immersed in a fearful environment in Virtual Reality, while having either a male or female observer present. Heart rate variability and skin conductance level were measured and questionnaires were used to score the experience of fear. Differences in the subjective fear, but not in the physiological fear measurements are significantly correlated to the gender effect. These findings are approached in a Quantum Cognition framework, suggesting that contextuality has an influence on how people consciously experience their feelings, but not on the physiological manifestation thereof. Future research should investigate this effect with larger samples and for other emotions.

(3)

INTRODUCTION

Fear is a functional emotion that has survival value with a deep evolutionary origin (Ohman, 2008). However, fear is subjected to contextual factors, that influence fear outcomes in research. A common determining factor in intersubjectivity is the gender of the people involved. Nonetheless, this is an often overlooked effect in most studies, while it can have a significant impact on results. Indeed, the effect of the gender of the observing researcher on fear outcomes appears significant (Larkin, Ciano-Federoff & Hammell, 1998). Knowledge about these contextual effects on cognition could explain side effects that were previously unexplained.

A new branch in cognition research called Quantum Cognition (QC) allows an elegant explanation of effects of a contextual factor like gender on a complex concept like fear (Busemeyer, Pothos, Franco, & Trueblood, 2011). Another new trend in cognition research is the use of immersive virtual environments, like Virtual Reality (VR), to simulate more realistic situations than can be achieved in a traditional experimental setting. VR has the capacity to trigger realistic behaviours and emotions in humans (Slater, 2009). For example, unjustified violence in a virtual reality game produced guilt in the players (Hartmann et al., 2010). Another example is a study that simulated Milgram’s electric shock experiment. Even though the participants were fully aware that neither the shocks nor the strangers in the virtual environment were real, they responded at subjective, behavioural, and physiological levels as if they were real (Slater et al., 2006).

The current research acts as a proof of principle and looks into how gender influences the experience of emotions. In case observer’s gender indeed influences induction of fear, future fear research should be more attentive to this context effect. The main question of this research is: “Does observer’s gender influence the physiological and psychological induction of fear in a Virtual Reality setting and can this be explained by Quantum Cognition?”. To find an answer to this research question, we have multiple sub-questions, namely:

1. What is fear and how is it induced? 2. How can VR be used for fear induction? 3. What is Quantum Cognition?

4. How can QC account for contextual factors like observer’s gender? 5. Does observer’s gender influence fear induction?

(5)

Research like this needs an interdisciplinary approach. The three disciplines used to answer the research question are Biomedical Science (BS), Artificial Intelligence (AI) and Physics. Both theoretical and methodological integrations of the three disciplines are needed. Theoretically, Physics explains Quantum Cognition and applies it on the current research. BS defines the psychological and biological processes underlying emotion in general and fear in particular. Moreover, it discusses the psychophysiological effects of contextual factors as gender. AI provides a deeper understanding of emotions and cognition of humans in connection to an immersive Virtual Environment. All theoretical insights each answer part of the research hypotheses and thus add to an integration. Integration of methods occurs by combining an experimental design -including physiological and psychological measurements- from BS, with the medium of Virtual Reality, which is common in AI and Computer Science research. Physics accounts for the experiment manipulation, based on inspirations from QC. Integration of results is formed by explaining the psychobiological data from the experiment within a Quantum Cognition framework.

The following section discusses relevant theories from all three disciplines. The third section elaborates on the integrated techniques, followed by the results in section four. The last section provides a summary and interpretation of the findings, discusses relevant contributions and limitations of the study, and makes suggestions for further research.

THEORETICAL FRAMEWORK

This section discusses relevant theories. First, an elaboration of the complexity of the concept of fear is presented. Then, fear induction using a VR environment is explained. A short discussion of gender effects follows. After a description of Quantum Cognition, the current study is placed into the QC framework.

Fear

Fear is often defined as a basic emotion that is experienced as unpleasant (i.e. negative valence) and is associated with high arousal. According to LeDoux (2013), the scientific use of ‘fear’ refers to two distinct phenomena. One meaning refers to conscious feelings of being afraid (LeDoux, 2014). The other meaning refers to (nonconscious) behavioural and physiological responses, and is more appropriately called threat conditioning, where

(6)

threat is defined as a stimulus that elicits defense responses (LeDoux, 2014). Fear as feeling has many different manifestations, like animal-phobia fear or social fear, which have in common that the brain is aware of a current of potential danger (LeDoux, 2013). In literature, fear is often used as an interchangeable synonym of anxiety. Although these words define somewhat similar emotional states, there are some significant differences between the two. Fear differs from anxiety primary in that it follows an identifiable stimulus: there is an obvious danger that is clearly located in space and time (Ohman, 2008). Anxiety can be regarded as a more complex state of being, where there is not a present threat and there are more emotions involved, such as fear, distress and nervousness (McNeil, Vargovich, Ries & Turk, 2012). When coping behaviours like escape and avoidance fail, fear can turn into anxiety (Ohman, 2008).

Virtual Reality

Virtual Reality (VR) is a digital environment containing a number of displays connected to a computer, combined with stereo sound and sometimes tactile feedback. A common characteristic between VR systems is a high immersiveness. The degree of immersion is a property of a system, that is independent of the human experience that it produces, and includes factors like the extent of the field of view, degree and quality of multisensory stimulation, and the realism of proprioception (e.g. quick and accurate update of visual input based on head movements) (Sanchez-Vives & Slater, 2005). A Head-Mounted Display (HMD) displays images very close to the eyes and includes head tracking. It provides slightly different images for the left and right eye, that are updated based on head movements, simulating the effect of moving in a 3D environment (Slater, 2009). A HMD allows for experiences in a first-person perspective, i.e. a simulation of how humans normally experience perception as located in the self (Seth, 2014).

Immersive virtual environments are capable of inducing emotions because they can simulate a sense of presence (Slater, 2009). Presence is a state of consciousness whereby the sensation of being absorbed in an external world is felt, as opposed to absorption in the internal world of thoughts (Waterworth et al., 2010). Mediated presence occurs when this feeling is evoked by computer-mediated environments (Waterworth et al., 2010). This presence is more operationally defined as the extent to which people respond realistically within a virtual environment, both on physiological and emotional/behavioural level (Slater, 2009). Presence can be decomposed into two parts: Place Illusion (Pi) and

(7)

Plausibility Illusion (Psi). Pi is the feeling of being in a certain (virtual) place, despite the knowledge of physically being in a different place (Slater, 2009). Psi is the feeling that perceived events are actually occurring, while knowing that this is not the case (Slater, 2009). In other words, Pi refers to how the environment is perceived, while Psi refers to

what is perceived. Pi is caused by the level of immersion of a virtual environment (Gorini

et al., 2011). Psi is caused by the realism of situations: correlations between actions and reactions, events directly referring to the person while being outside the person's direct control, and events that conform to what would be expected in similar situations in reality (Slater, 2009). Psi can also be invoked by narrative (Gorini et al., 2011). Although the dynamics between presence and emotion are not yet well understood (Diemer et al., 2015), it seems that presence has a direct influence on the emotion of fear (Peperkorn et al., 2015).

The feeling of presence induced by a virtual environment can be explained by situated cognition and simulation theory. Situated cognition states that information from previous situations that is stored in memory, is used to interpret and infer what is happening in the current situation (Wilson-Mendenhall et al., 2013). Simulation theory states that thinking about certain events involves the same processes in the brain as actually experiencing these events (Hesslow, 2012). Indeed, neural overlap between sensorimotor perception/action and sensorimotor imagery has been found in previous research (Kosslyn et al., 2001). According to the situated simulation theory of Barsalou (2009), perceptions, actions, and introspections from the past, are (partially) re-enacted, i.e. simulated, in order to be used for current representations. Thus, current perception activates knowledge to generate simulations in the relevant modalities, representing a prediction of what is going to happen (Barsalou, 2009). For example, perceiving an object predicts settings in which the object is likely to occur (Barsalou, 2009). Similarly, in the VR experiment, the visual input that simulates a first person perspective, activates knowledge that such input implies that the observed events are experienced by oneself. This simulates the prediction of a sense of presence, thus activating the illusion of non-mediation.

Influence of gender / observer

The research of Larkin et al. (1998) suggests that there is an effect on the responses of the autonomous nervous system when the gender of observer differs. They found that male subjects with high social fear displayed higher blood pressure reactions when observed by

(8)

a female. Also, Miner & Eischeid (2012) found that emotions induced by situations are dependent on the gender. Same-gender targets gave a higher level of discomfort compared to other-gender targets. Another study investigated a similar effect: stress-induction during public speaking in front of a panel of judges composed of either male or female panelists (Duchesne et al., 2012). They found that stress response as measured by cortisol increase was only found when subjects were exposed to panelists of the other gender. Interestingly, this is an opposite effect of what Miner & Eischeid (2012) found. It is clear from this short analysis that a gender effect on emotion can be expected, although little is known about the dynamics of such effects.

Quantum cognition

Quantum Cognition (QC) is a framework in which cognitive effects can be placed, under circumstances of uncertainty, and order- and context-dependency. QC is one of many applications of the abstract concepts from the physics theory of quantum mechanics (Pothos & Busemeyer, 2013). It should be noted, however, that this does not suggest the application of quantum physics to brain physiology, for this remains a controversial issue (Pothos & Busemeyer, 2013). Considering the brain as a quantum system is unsupported by physics, as the overall thermal energy does not allow for any quantum mechanical action to take place on a scale that would influence emotions, according to Ben Freivogel (pers. comm., November 3, 2015).

Quantum cognitive effects can be explained with the concepts of superposition and entanglement. Superposition is the uncertainty of knowledge about the state of a system (Pothos & Busemeyer, 2013), which in QC means that knowledge of the state of a thought or event, implies uncertainty about the state of another thought or event (Bruza, Wang & Busemeyer, 2015). Entanglement occurs when the probability distribution of a system cannot be formed by the probability distribution of its constituent parts. This is due to interdependencies among the parts, as a change in one part can lead to a change in another (Pothos & Busemeyer, 2013).

The principle of complementarity states that a distinction has to be made between two types of events: compatible and incompatible events (Bruza et al., 2015). Event A and event B are compatible when they can be considered simultaneously (Bruza et al., 2015). This could give equal results as classical cognition, unless the events are entangled, and thus interdependent (Pothos & Busemeyer, 2013). Event A and event B are incompatible

(9)

when they have to be considered sequentially (Bruza et al., 2015), thus accounting for relative order and contextual effects (Pothos & Busemeyer, 2013). Contextuality is the effect where measuring one variable influences another correlated variable (Grudka et al., 2014). The outcome of event B is thus in superposition with the outcome of event A, as the outcome of A might influence the outcome of B (Pothos & Busemeyer, 2013).

Busemeyer et al. (2011) state that earlier research shows that human judgement has strong order effects. Bruza et al. (2015) give another example where quantum probability dominates over classical probability. They discuss the prisoner’s dilemma where defecting is the dominant choice. It has been shown that the percentage where the player defects when the choice of the other player is unknown is a violation of classical probability. To explain this, Shafir & Tversky (1992) reasoned it to be a failure of non-sequential reasoning. However, quantum probability allows for a simple explanation using incompatibility of choices (Bruza et al., 2015).

Conte et al. (nd) analyse the optical illusion of ambiguous figures (such as the Necker cube) in a quantum like way. One can see both options (superposition) but neither at the same time. Conte et al. (nd) found evidence that mental states like this follow a quantum-like probability theory. These results are considered anomalous in a classical cognition framework, however are simple and elegant in Quantum Cognition (Busemeyer et al., 2011).

Based on the QC framework, a prediction of the gender effects can be made. In this research, event A is the exposure of the observer's gender effect on the subject, and event B entails the fear reactions of the subject on the horror scene. In case of incompatibility of these events, there are two uncertainties: the classical uncertainty (Pothos & Busemeyer, 2013) of the various outcomes of the fear induction, and the quantum uncertainty about how the state of event B will collapse from superposition (Pothos & Busemeyer, 2013). The value of event B is constructed during this collapse, based on the contextual information (i.e. the outcome of event A). Thus, a gender effect implies incompatibility between event A and event B. As the literature on gender and observer effects suggests context-dependency, incompatible events are expected.

To conclude, two hypotheses are predicted:

1. Subjective and physiological fear is significantly higher during horror compared to control condition (i.e. fear induction effect)

(10)

2. Subjective and physiological fear differs significantly between groups with different observer’s gender (i.e. gender effect)

METHODS

To research the observer’s gender effect on fear, an empirical experimental set-up in an immersive Virtual Reality system is used. The subjects are college students. Limited time and budget allow for a sample size of 16 subjects. The subjects are divided into two groups: a female observer condition and a male observer condition. Gender of the subjects is equally spread among the two groups.

Virtual Reality setting to induce fear

Fear induction is tested in a Virtual Reality (VR) setting. The Oculus Rift Development Kit 2 (DK2) provides the virtual environment. The Rift is a binocular Head-Mounted Display (HMD), which gives the possibility of a full view immersion in the virtual environment, combined with changing visual input as the head turns.

The VR horror movie ‘The Chosen’ (Supergravity Pictures, 2015) is used to induce fear. It includes elements that can potentially increase Place Illusion (Slater, 2009), namely a first-person perspective, a virtual body that is seen when looked down, and first-person audio (terrified sounds). It also includes elements that potentially increase Plausibility Illusion (Slater, 2009), namely the demonic girl that addresses the participant directly. As typical for VR applications, the movie is short: 1:41 minute. This scene is comparable to scenes from ‘The Ring’, a horror movie that is used in other studies to successfully induce fear (Feinstein et al., 2011; Verhoef et al., 2009).

Distractions from outside the VR are limited as much as possible, since they can break Place Illusion (Slater, 2009). As passing light from outside can occur through small gaps, the lights in the room are switched off. To minimize perceived interference from other observers, curtains of the research room are closed and the smaller area of the VR system is surrounded with screens. A surround sound setting with five speakers on high volume both minimize distracting sounds from outside, and contribute to the immersiveness of the system.

(11)

Manipulation: observer’s gender

Based on the principle of contextuality from the Quantum Cognition framework, subjects are exposed to the observer in two ways: the physical presence of a male or female researcher, and a recorded voice fragment of the same researcher. The audio fragment is played before the fear induction video, and contains an introducing narrative to the horror scene (see appendix 1). The gender effect is thus reinforced right before the fear induction. An additional benefit of this design is that the narrative is likely to increase Plausibility Illusion and thus presence, as explained in the literature (Giorini et al., 2011).

Physiological measurements

The activity of the autonomous nervous system (ANS) has the potential to give an indication of a person’s emotional status, as it is a representation of the neural activity related to the brain and body regulation. It has been suggested that measuring components of the ANS can give insight in emotional processing. Activity of the ANS can be measured by cardiovascular, respiratory and electrodermal measures (Kreibig, 2010). The most used psychophysiological measures for fear and anxiety are cardiovascular (Kreibig, 2010). The assessment of electrodermal activity, i.e. sweat gland activity in the skin, is also a common used measure (McNeil et al., 2012). The two most common measuring variables are Heart Rate (HR), a cardiovascular measure, and Skin Conductance Level (SCL), an electrodermal measure (Kreibig, 2010). In line with previous research, this study measures the ANS activity by the HR and SCL. HR is measured with electrodes on the chest, while SCL is measured with two electrodermal sensors for the fingers. As HR is not a generalizable measurement, it is transformed to Heart Rate Variability (HRV) to compare the test results. More specifically, a normalized version of HRV (i.e. normalized fluctuation with respect to the mean) should be used, as this overcomes mathematical bias associated with HRV (Sacha, 2013). The induction of fear has been shown by several studies to give an acceleration in HR, a decrease in HRV, and an increase in SCL (Kreibig, 2010). It has also been shown that an increasing intensity in negative emotional stimuli in general correlates with increased electrodermal activity (Fusar-Poli, Landi & O’Connor, 2009).

Subjective measurements

As measurements of the ANS are not enough to differentiate between emotions (Kreibig, 2010), fear is additionally measured using a questionnaire. To compare the neutral state

(12)

(baseline) and the fear induction state, a selection from the State-Trait Anxiety Inventory - State (STAI-S) questionnaire is used (see appendix 2). This questionnaire is a commonly used measure of anxiety and it has been proven to be a sufficient measure to study anxiety in research and clinical settings (Vitasari, Wahab, Herawan, Othman & Sinnadurai, 2011). As mentioned in the theoretical framework, anxiety and fear are related emotional states but do not entirely describe the same phenomenon. However, Anxiety self-reports can be used to measure anxiety but also fear (Julian, 2011) and because more self-reports are available for anxiety, as this is also a state that is associated with several medical disorders, this study will use a situational anxiety questionnaire for the subjective measurement of fear.

To collect general data about the subjects, some demographics, general anxiety in the last four weeks, and experience and sensitivity to horror scenes are asked (see appendix 2). To assess general anxiety two common questionnaires are used: Beck Anxiety Inventory (BAI) and Hospital Anxiety and Depression Scale - Anxiety (HADS-A). These self-reports measure the generalized symptoms of anxiety and fear and are both extensively tested on reliability and validity (Julian, 2011).

Design of experiment

First, the experiment is explained to the subject. The subject knows a horror scene will be shown, but does not know the main topic of the research, namely the effect of observer’s gender. After the introduction, the subject is asked to fill in the general questionnaire (including BAI and HADS-A) and the baseline state questionnaire (STAI-S) about the current emotional state. Only after these questionnaires the measuring equipment is attached to the subject, as to allow a baseline measurement without potential anxiety caused by the equipment. Next, a configuration of the VR is followed by two neutral VR movies as baseline measurement of physiology (4 minutes). After the baseline measurement the subject is exposed to the fear induction movie of 1:41 minutes, introduced by a spoken scenario of 15 seconds (by either a male of female voice). When the movie has ended, the subject fills in the fear induction state questionnaire, which is the same selection of items from the STAI-S as the baseline, but this time referring to the last movie. No interference of the observing researcher occurs between the fear induction movie and the last questionnaire, in order to allow the subjects to stay attached to the induced emotions. After the experiment the subject is offered ‘aftercare’, which involves

(13)

discussing the VR and fear experience with one of the researchers. This discussion also functions as qualitative input for the study.

Techniques for data analysis

For normality, the Shapiro-Wilk test will be used as Razali & Wah (2011) found that it is the strongest test for normality compared to three other often used normality tests. Next, the heart rate and the skin conductance level are metric data. Normally, Likert-Scales are ordinal, but since all questions test for the same trait in a person, and have numerical value, they can also be analysed as metric data.

The predictor variable is categorical in all cases (before versus after; female observer versus male observer). Depending on the normality assumptions, the number of predictor variables and depending on if they are the same or different participants in the research this gives way to the independent t-test, the dependent t-test, the Mann-Whitney test, the Wilcoxon Matched-Pairs test, the one-way independent ANOVA and the Kruskal-Wallis test. The results discuss which tests are used on a case-by-case basis.

Figure 1: Experimental design. 2: the general questionnaires consist of questions about demographics, fear of horror movies, the BAI and HADS-A. 3 and 9: the situational questionnaire consists of a selection of items from the STAI.

(14)

RESULTS

During the experiments, 16 subjects were tested. For each of these subjects, there are several sets of data to process. First, there is the questionnaire, which has four parts. It measures a subset of the questions of the STAI twice, before and after the induction phase. These results are analyzed if there are any significant changes in the situational fear of the subject. Next, the questionnaire also includes the HADS-A and the BAI. These results will be used to try and explain possible outliers or effects in our research we cannot predict. The next datasets to analyze are the Skin Conductance Level and the ECG. An alpha-level of 0.05 will be used.

Questionnaires

STAI

The subset of the STAI has ten questions and has a ten-level Likert scale in an attempt to minimize skewness of distributions (Wittink & Bayer, 2003). The questionnaire had some reverse worded questions, the scores on these questions were reversed. The ten-level Likert Scale is treated as metric data.

Within-group

The first thing to analyze is if there is a significant difference before and after the induction phase for each group (male / female observer). The sum of the Likert-items is tested for normality using the Shapiro-Wilk test. Normality was rejected for the female observer but accepted for the male observer. Using the Wilcoxon Matched-Pairs test between the before and after results when observed by a female shows that there was no significant change. Using the dependent t-test for the male observer did give a significant change.

(15)

Figure 2: An histogram displaying the STAI Scores of the male observer group. The blue bars show the before scores while the red bars show the after scores. The bin size of this histogram is 5. Please note that there are overlaying bars.

Figure 3: An histogram displaying the STAI Scores of the female observer group. The blue bars show the before scores while the red bars show the after scores. The bin size of this histogram is 5. Please note that there are overlaying bars.

(16)

Between-group

The second to analyze is if there are significant differences between the groups during the same phases (baseline & induction). Both ‘before’-datasets are normally distributed according to the Shapiro-Wilk test. Using the independent t-test, it was shown that the baseline datasets are not from identical populations (significantly). Normality for the ‘after’-datasets was rejected. Using the Mann-Whitney test, it was significantly shown that the induction datasets are from identical populations.

HADS-A

To analyze if both groups (female & male observer) come from identical populations, normality was accepted using Shapiro-Wilk. Using the dependent t-test, it was shown that the scores from the HADS-A test were significantly identical.

Figure 4: An histogram displaying the HADS-A Scores of both groups. The blue bars show the scores for the male observer group while the red bars show the scores for the female observer group. The bin size of this histogram is 1. Please note that there are overlaying bars.

(17)

BAI

Using Shapiro-Wilk, we accepted normality. Using the dependent t-test, it was shown that the scores from the BAI test were significantly identical enough.

Skin Conductance Level

Within subject

Since SCL continually rises, one cannot simply compare datasets. To give a good measure, the gradient of both datasets were calculated. This avoids problems of dependence between groups, since the height and a part of the absolute change in the induction phase is coming from baseline skin conductance levels. The gradients are calculated between five second means to reduce variability due to active skin responses.

These gradients were tested for normality using Shapiro-Wilk. Depending on this result, either the Wilcoxon Matched-Pairs test, or the dependent t-test were performed. From 16 subjects, one subject from the male observer group had false data. From the other 15 subjects 13 showed no significant change in gradient of the skin conductance level between the baseline and the induction phase. The two subjects that were significantly different Figure 5: An histogram displaying the BAI Scores of both groups. The blue bars show the scores for the male observer group while the red bars show the scores for the female observer group. The bin size of this histogram is 2. Please note that there are overlapping bars.

(18)

enough to be from different populations were from both the male and the female observer group.

Figure 6: The Skin Conductance Level means (of five-second segments) in μS plotted against the time unit. Please note that this shows both the datasets of the baseline and the induction phase. Also note that this is a dataset taken from one random subject to give an indication of how the data looks like.

Figure 7: The gradient of the SCL means in μS plotted against the time unit. Please note that this shows both the datasets of the baseline and the induction phase. Also note that this is a dataset taken from one random subject (the same for both plots) to give an indication of how the data looks like.

(19)

Within group

To see if the gradient of the baselines of each group were the same, they were tested for normality using Shapiro-Wilk. Normality was rejected. Thus, to see if they were from identical populations, the Kruskall-Wallis test was applied. Both groups had gradients from significantly different populations.

ECG

Within subject

The ECG data with the 50Hz notch filter was used to make sure that no electric fields from power outlets were influencing the results. To determine the heart rate, the distance between R-tops was calculated. Then, the heart rate variability was calculated by taking the square root of the mean of the squares of the successive differences between R-tops (RMSSD). Subsequently, the datasets were tested if they were significantly different from the baseline during the induction phase. To do this, they were first all tested for normality using the Shapiro-Wilk test. All normality for baselines was rejected. Normality for induction phases was rejected approximately fifty per cent of the time. To test for significant change, the Wilcoxon Matched-Pairs test was used. For three subjects in both groups (six in total), this gave a significant change. The others (5 from female observed group and 4 from male observed group) were not changed significantly.

(20)

Within group

To see if the results between subjects but within group were comparable, the Kruskall-Wallis test was used. Both groups had significantly different baselines and significantly different induction phases from each other. To see if the group as a whole had significant changes before and after the induction phase, normality over the entire male observer group and over the entire female observer group was tested and rejected. Thus, the Wilcoxon Matched-Pairs test was applied and this concluded that the groups as a whole did have significantly different HRV’s during the induction phase.

Figure 8: The Heart Rate Variability means (of five-second segments) in bpm plotted against the time unit. Please note that this shows both the datasets of the baseline and the induction phase. Also note that this is a dataset taken from one random subject to give an indication of how the data looks like. Several data points show a much higher HRV compared to the other data points. These could be taken as outliers, but since they represent mostly true values with only the uncertainty of the measuring device to take into account, they were not scratched from the dataset.

(21)

DISCUSSION AND CONCLUSION

Summary and interpretation of results

The results of the HADS-A and BAI show that the samples under the influence of male and female observers were randomized enough to be considered from identical populations. The results of the STAI show that the group with the male observer displayed significant change of fear level after the fear induction while the female observer group showed no significant change. Analyzing the skin conductance levels of the subjects showed that only two of the subjects had a significant change in gradient of skin conductance levels after the induction. These two subjects were from both groups and thus, gave no sign of a gender of observer effect. Looking at the heart rates of the participants, and specifically, the variability of the heart rates, 6 participants showed significant change after the fear induction. However, these participants were evenly distributed across both groups so no gender of observer effect was found.

Summarizing, a fear induction effect is found based on self-reported fear and HRV measurements, but not on SCL measurements. Therefore, the first hypothesis is mostly confirmed. As for gender effect, the physiological measurements show no significant differences between the two conditions, however self-reported fear shows a significant increase in the male observer condition, but not in the female observer condition. Therefore, the second hypothesis is partly confirmed. These findings suggest that there is a gender effect on how people think they feel, while the body does not show any significant differences in the measure of fear.

As the increase in subjective fear can be attributed to the gender effect, while the decrease in HRV cannot, this suggests different underlying mechanisms. Indeed, there is no one-to-one relationship between felt emotions and the autonomic system (Kreibig, 2010). Fear can be activated in (at least) two pathways: the perceptual pathway triggered by e.g. visual and auditive cues, and the conceptual pathway, triggered by fear-related information (Diemer et al., 2015). Perceptual cues are assumed to lead to physiological and behavioural responses, while information is likely to lead to subjective fear and less so to physiological activation (Diemer et al., 2015). This is in line with the view that the automatic system is involuntary: it does not interfere with attention, is not sensitive to distractions and is hard to access consciously (Ohman, 2008). Immersiveness in a VR seems therefore enough for the automatic system to react on the perceived threat, without interference from contexts outside the horror scene.

(22)

These findings can be explained by the Quantum Cognition framework, although in a different way than expected. The original hypothesis did not differ between psychology and physiology, and thus expected a significant effect on both. As self-reported fear showed a context effect, while physiology did not, it seems that the observer’s gender is an incompatible observable with subjectively felt fear, while a compatible observable with ANS activity. This could be evidence that contextuality does have an influence on how people consciously experience their feelings, but that it has no subconscious influence on how the emotion is actually induced. This interpretation is in line with the recently stressed distinction between fear as conscious feeling and threat conditioning as behavioural and physiological response (LeDoux, 2013). The self-reports refer to the feeling, while the ANS measurements refer to threat conditioning.

The results provide a refinement on an earlier overlooked aspect: the questionnaire measuring the state of fear, more specifically measures the subject’s own judgement of felt fear. Quantum Cognition allows for the possibility that the actual asking for this judgement breaks down the superposition of the fear induction outcome, thereby explaining why a gender effect occurs in the questionnaire scores, but not in the physiological measurements made during the induction phase. As discussed before, previous studies suggest that human judgement, like several other cognitive modalities, is subjected to context-dependency and order-effects (Busemeyer et al., 2011). However, no studies so far seem to confirm that automatic responses like those of the ANS are subjected to equal effects.

Extending the idea of measuring a human judgement as opposed to an automatic response, an effect called social norm effect could provide a further explanation of the different psychological and physiological results of the influence of observer’s gender. As found in previous research (Fisher, 2007), the gender of the experimenter influenced questionnaire answers. More specifically, men scored higher on number of sexual partners, when the researcher was female in case that the cover sheet of the questionnaire stated research findings of women being more sexually permissive than men (Fisher, 2007). Thus, like in this research, contextual factors influenced the results of the questionnaire. The higher subjective fear effect of the male as compared to the female observer, could be a function of complying to social expectations rather than of actual increase in frightful feelings.

(23)

Contributions and Interdisciplinarity

The main contribution of this study is the combination of insights and techniques from various disciplines to do a few things that have not been done before: testing the effect of manipulating a contextual variable in physical reality on the experience in virtual reality, applying the QC framework on an emotional state, and testing for an observer’s gender effect in relation to fear induction as well as to VR. The results contribute to the interdisciplinary literature body. On the intersection of Physics and Cognition, the findings show support to quantum cognitive effects in human judgement. Assuming a correlation between the measurement of the judgement of an emotional state and the actual emotional state, the findings also suggest relevance of QC in emotion research. An important finding is the lack of a physiological response to contextuality, suggesting compatibility between the events, and thus no important order effects between observer’s gender and ANS reactions.

The interdisciplinary insights of this study also contribute to disciplinary insights of each discipline. A contribution to Psychobiology is that the novel induction method of Virtual Reality offers new possibilities, like simulating real situations and overcoming ethical constraints. However, the technology is currently limited and should be given time to mature. A contribution to AI is that contextual factors seem to be relevant for experiences induced in a Virtual Reality system. Furthermore, an important contribution to emotion and VR research, is the influence of gender effects on outcomes.

Limitations and Suggestions for Further Research

A few problems arose with the physiological measuring of fear. As of yet, there is no clear consensus about the exact relation between the activity of the ANS and the emotional status of a person (Kreibig, 2010), which makes conclusions about emotions difficult. In the interpretation of results this problem actually contributed to a view of the automatic responses as part of threat-conditioning, and less so of the experience or feeling of fear. In this study, SCL was used as a measure for skin conductance, but the results of this measure did not turn out to be helpful in the recognition of a fearful emotion. It was expected that changes in the gradient would be more recognizable and useful for the discrimination. Another measure for skin conductance, Skin Conductance Response (SCR), was probably more useful for this research than SCL, as SCR data shows the small

(24)

changes in skin conductance (Braithwaite, Watson, Jones & Rowe, 2015). However, for this study, SCR measurement devices were not available.

The video that is used to induce fear can also induce other effects that influence the ANS response. There might be a contribution of stress, shock and surprise, rather than only fear. Moreover, the movie is quite short, which makes measurements less reliable.

As part of the VR system, the volume of sounds, position of speakers and distracting sound and light from the environment were taken into account. In any replication of this research these factors have to be included in the experimental set-up to increase immersion. One of the most important limitations, is the limited sample size (n=16), due to time and budget constraints. Nonetheless, this research serves as an initial investigation in a new and interesting area. Further research could include experiments with larger sample sizes and applications of other contextual effects in other contexts.

(25)

BIBLIOGRAPHY

1) Barsalou, L. W. (2009). Simulation, situated conceptualization, and prediction.

Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1521),

1281-1289.

2) Braithwaite, J., Watson, D., Jones, R., & Rowe, M. (2015). A Guide for Analysing

Electrodermal Activity (EDA) & Skin Conductance Responses (SCRs) for Psychological Experiments (2nd ed., Rep.). University of Birmingham: Behavioural

Brain Sciences Centre.

3) Bruza, P., Wang, Z., & Busemeyer, J. (2015). Quantum cognition: A new theoretical approach to psychology. Trends in Cognitive Sciences, 383-393.

4) Busemeyer, J., Pothos, E., Franco, R., & Trueblood, J. (2011). A quantum theoretical explanation for probability judgment errors. Psychological Review, 118(2), 193-218.

5) Conte, E., Khrennikov, A., Todarello, O., Federici, A., Mendolicchio, L., & Zbilut, J. (n.d.). Mental States Follow Quantum Mechanics During Perception and Cognition of Ambiguous Figures. Open Systems & Information Dynamics, 85-100.

6) Diemer, J., Alpers, G. W., Peperkorn, H. M., Shiban, Y., & Mühlberger, A. (2015). The impact of perception and presence on emotional reactions: a review of research in virtual reality. Frontiers in psychology, 6.

7) Duchesne, A., Tessera, E., Dedovic, K., Engert, V., & Pruessner, J. C. (2012). Effects of panel sex composition on the physiological stress responses to psychosocial stress in healthy young men and women. Biological psychology, 89(1), 99-106.

8) Feinstein, J. S., Adolphs, R., Damasio, A., & Tranel, D. (2011). The human amygdala and the induction and experience of fear. Current Biology, 21(1), 34-38. 9) Fisher, T. D. (2007). Sex of experimenter and social norm effects on reports of

sexual behavior in young men and women. Archives of Sexual Behavior, 36(1), 89-100.

10) Fusar-Poli, P., Landi, P., & O’Connor, C. (2009). Neurophysiological response to emotional faces with increasing intensity of fear: A skin conductance response study. Journal of Clinical Neuroscience, 16(7), 981-982.

11) Gorini, A., Capideville, C. S., De Leo, G., Mantovani, F., & Riva, G. (2011). The role of immersion and narrative in mediated presence: The virtual hospital experience.

(26)

12) Grudka, A., Horodecki, K., Horodecki, M., Horodecki, P., Horodecki, R., Joshi, P., Wójcik, A. (2014). Quantifying Contextuality. Phys. Rev. Lett. Physical Review Letters, 112(12).

13) Hartmann, T., Toz, E., & Brandon, M. (2010). Just a game? Unjustified virtual violence produces guilt in empathetic players. Media Psychology, 13(4), 339-363. 14) Hesslow, G. (2012). The current status of the simulation theory of cognition. Brain

research, 1428, 71-79.

15) Julian, L. J. (2011). Measures of Anxiety. Arthritis Care & Research, 63(0 11). 16) Kosslyn, S. M., Ganis, G., & Thompson, W. L. (2001). Neural foundations of

imagery. Nature Reviews Neuroscience, 2(9), 635-642.

17) Kreibig, S. D. (2010). Autonomic nervous system activity in emotion: A review.

Biological Psychology, 84(3), 394-421.

18) Larkin, K., Ciano - Federoff, L., & Hammell, D. (1998). Effects of gender of observer and fear of negative evaluation on cardiovascular reactivity to mental stress in college men. International Journal of Psychophysiology, 29(3), 311-318.

19) LeDoux, J. E. (2013). The slippery slope of fear. Trends in cognitive sciences, 17(4), 155-156.

20) LeDoux, J. E. (2014). Coming to terms with fear. Proceedings of the National

Academy of Sciences, 111(8), 2871-2878.

21) McNeil, D. W., Vargovich, A. M., Ries, B. J., & Turk, C. L. (2012). Anxiety and fear. In V. S. Ramachandran (Ed.), Encyclopedia of human behavior (second edition) (pp. 161-168). San Diego: Academic Press.

22) Miner, K. N., & Eischeid, A. (2012). Observing Incivility toward Coworkers and Negative Emotions: Do Gender of the Target and Observer Matter? Sex Roles, 66(7-8), 492-505.

23) Ohman, A. (2008). Fear and anxiety. In: Lewis, M., Haviland-Jones, J. M., & Barrett, L. F. (Eds.), Handbook of emotions (pp. 709-729). Guilford Press.

24) Peperkorn, H. M., Diemer, J., & Mühlberger, A. (2015). Temporal dynamics in the relation between presence and fear in virtual reality. Computers in Human

Behavior, 48, 542-547.

25) Pothos, E. M., & Busemeyer, J. R. (2013). Can quantum probability provide a new direction for cognitive modeling?. Behavioral and Brain Sciences, 36(03), 255-274. 26) Razali, N., & Wah, Y. (2011). Power comparisons of Shapiro-Wilk,

(27)

27) Sacha, J. (2013). Why should one normalize heart rate variability with respect to average heart rate. Frontiers in physiology, 4.

28) Sanchez-Vives, M. V., & Slater, M. (2005). From presence to consciousness through virtual reality. Nature Reviews Neuroscience, 6(4), 332-339.

29) Seth, A. K. (2014). A predictive processing theory of sensorimotor contingencies: explaining the puzzle of perceptual presence and its absence in synesthesia.

Cognitive neuroscience, 5(2), 97-118.

30) Shafir, E., & Tversky, A. (1992). Thinking through uncertainty: Nonconsequential reasoning and choice. Cognitive Psychology, 24(4), 449-474.

31) Slater, M., Antley, A., Davison, A., Swapp, D., Guger, C., Barker, C., & Sanchez-Vives, M. V. (2006). A virtual reprise of the Stanley Milgram obedience experiments. PloS one, 1(1), e39.

32) Slater, M. (2009). Place illusion and plausibility can lead to realistic behaviour in immersive virtual environments. Philosophical Transactions of the Royal Society

B: Biological Sciences, 364(1535), 3549-3557.

33) Supergravity Pictures (2015, September 3). THE CHOSEN - 360 VR Experience (Horror Movie in VR!) [Video file]. Retrieved from https://youtu.be/oPiX5-_w5vw 34) Target Heart Rates. (n.d.). Retrieved January 8, 2016, from

http://www.heart.org/HEARTORG/GettingHealthy/PhysicalActivity/FitnessBasics /Target-Heart-Rates_UCM_434341_Article.jsp#.VpAlf3UrJkU

35) Verhoef, T., Lisetti, C., Barreto, A., Ortega, F., Van der Zant, T., & Cnossen, F. (2009). Bio-sensing for emotional characterization without word labels.

Human-Computer Interaction. Ambient, Ubiquitous and Intelligent Interaction, V, 693-702.

36) Vitasari, P., Wahab, M. N. A., Herawan, T., Othman, A., & Sinnadurai, S. K. (2011). Re-test of state trait anxiety inventory (STAI) among engineering students in malaysia: Reliability and validity tests. Procedia - Social and Behavioral Sciences,

15, 3843-3848.

37) Waterworth, J. A., Waterworth, E. L., Mantovani, F., & Riva, G. (2010). On feeling (the) present. Journal of Consciousness Studies, 17(1-2), 167-188.

38) Wilson-Mendenhall, C. D., Barrett, L. F., & Barsalou, L. W. (2013). Situating emotional experience. Frontiers in human neuroscience, 7.

39) Wittink, D., & Bayer, L. (2003). The Measurement Imperative. Marketing

(28)

APPENDICES

Appendix 1: Narrative introduction to horror scene

The following narrative was played after the baseline to introduce participants to the horror scene in the induction phase.

“You travelled to Science Park at sunset to participate in a research. After filling in some questionnaires, you lost consciousness. You are about to wake up and find out that you are chained to a chair in some unknown bedroom.”

(29)

Appendix 2: questionnaire

(30)

(31)

Influence of Observer’s Gender on Induction of Fear in Virtual Reality : A Quantum Cognitive Effect?