Decision Making, Physical Fitness, &
Heart Rate Variability in Virtual Reality
Author: Maarten R. Struijk Supervisor: Dr. Jasper Winkel
Date: 17-‐07-‐2015 Student ID: 5618002
Abstract
Effort-‐Based Decision Making (EBDM) has been researched in humans using standard psychological tools. These tools might suffer from low ecological validity in order for them have high internal validity. Virtual Reality (VR) using a Head Mounted Display (HMD) might be a valuable tool for psychological research, allowing for both high internal validity and increased ecological validity. Research shows that EBDM is closely linked to the Anterior Cingulate Cortex (ACC), which in turn is affected by aerobic physical exercise. The ACC is also one of the main brain regions controlling heart rate and Heart Rate Variability (HRV) through it’s mediating effects on the parasympathetic nervous system. HRV is inYluenced by aerobic physical exercise. A novel EBDM
paradigm is presented using a standard psychological paradigm and a VR paradigm. Results indicate that the presentation method has no effects on the EBDM task. Results indicate that the HRV is able to differentiate between the presentation, allowing for optimism regarding the future of VR in psychological research.
Introduction
People make decisions many times a day. Most of these decisions are based on obtaining some kind of reward (e.g. getting food), while simultaneously requiring some kind of effort (walking to the shop). When these decisions include a choice between two kinds of reward (very tasty food versus fastfood) and differing levels of effort which at least somewhat reYlects these differing rewards (walking further for the tasty food versus getting fastfood from around the corner), these scenarios can be studied using various decision making frameworks.
One of these frameworks is Effort-‐Based Decision Making (EBDM). EBDM is about decision making based on the amount of effort needed to expend, to receive a certain reward associated with that effort (Kurniawan, Guitart-‐Masip & Dolan, 2011). Another framework is Delay-‐Based Decision Making (DBDM) where participants need to wait a set amount of time before the reward is given. Longer waits result in higher rewards (Floresco et al., 2008-‐1). Other decision making theories like Probabilistic Decision Making theory apply when the reward is not completely certain, such as in optimal-‐ foraging theory (Floresco, Maric & Ghods-‐SharifYi 2008; Niv, Daw, Joel & Dayan, 2007), but not all researchers apply this strict a separation between above mentioned
(Wardle, Treadway, Mayo, Zald & de Wit, 2011).
EBDM is traditionally measured in lab-‐rats using a T-‐maze paradigm (Salamone, Cousin & Bucher, 1994). The animal is placed in the bottom arm, and food pellets have been placed in the ends of the remaining arms. A barrier is placed in one of the arms to obstruct easy access. This is the high effort (HE) arm. The other arm is left open, this is the low effort (LE) arm. The animal can choose only one of the arms before the trial ends. In most experiments the amount of food in the HE arm is higher than in the LE arm, creating a high-‐effort/high-‐reward (HE/HR) arm and a low-‐effort/low-‐reward (LE/LR) arm. Healthy rats will choose the HE/HR arm most of the time (Walton, Bannerman & Rushworth, 2002).
A confound in EBDM research occurs when the HE/HR condition takes longer to complete than the LE/LR condition, like in most T-‐maze setups. This paradigm
measures aspects of DBDM as well as EBDM (Floresco et al., 2008-‐1). It potentially allows the animal to quickly complete many LE/LR trials to obtain a higher net reward than spending the same amount of time performing the slower HE/HR trials. For this reason it is important in EBDM research to make sure both conditions take the same amount of time (Floresco et al., 2008-‐1).
With people EBDM can be measured using different paradigms. In one
experiment participants were shown one of eight shapes which corresponded to a set amount of effort and reward (Croxson, Walton, O’Reilly, Behrens & Rushworth, 2009). Participants had to use a trackball to click on square shapes shown on the monitor. In this experiment the participants could not choose whether they wanted to do the LE/ LR option or the HE/HR option, which has implications for the neurobiological basis of these results (Walton, Croxson, Behrens, Kennerley & Rushworth, 2007). The
experiment of Westbrook, Kester and Braver (2013) consisted of a cognitive EBDM task by means of the N-‐back paradigm. Participants were repeatedly made to choose
between a low-‐effort 1-‐back task (LE), or the high-‐effort 2-‐back task (HE). The
participant received $2.00 for each HE task completed. The reward for the LE task was initially set at $1.00. After each iteration, the amount of the LE task was varied
depending on the choice made. When for example the HE task was selected in the Yirst trial, the LE task was then set to $1.50 on the second trial. When in the second round the LE task was chosen, the reward for the LE task dropped to $1.25 in the third trial, and so on. This way the balance between desired reward per expended effort was approached for each participant.
Extensive studies have been done on how the brain processes EBDM. Dopamine is associated with overcoming the effort needed in the EBDM paradigms (Phillips, Walton & Jhou, 2007; Walton, Kennerley, Bannerman, Phillips & Rushworth, 2006; Bardgett et al., 2009), and with the mental calculation needed to assess the relationship between effort and reward (Walton et al., 2006). In rats and in people, dopamine antagonists reduce percentage of times the HE/HR option is chosen (Bardget et al., 2009; Floresco et al., 2008; Schweimer & Hauber, 2006). In contrast, a small to medium dose of the dopamine agonist amphetamine increases the percentage the HE/HR option is chosen in rats and people (Bardget et al., 2009; Floresco et al., 2008; Wardle et al., 2011), while a high dose of amphetamine seems to have a deleterious effect in rats (Floresco et al., 2008).
Decision making is for a large part mediated in the anterior cingulate cortex (ACC) (Kurniawan, Guitart-‐Masip & Dolan, 2011; Schweimer & Hauber, 2006) and the striatum, which are connected with each other through glutamatergic projections (Balleine et al., 2007; Winkel, 2014). Damage to the ACC reduces the amount of effort rats are willing to spend on a particular amount of reward. This is not due to inability to calculate the relationship between effort and reward, because this effect is reversed by raising the reward in the HE/HR arm or placing a barrier in the LE/LR arm (Walton et al., 2002). Rudebeck et al. (2006) surgically damaged the ACC in some rats, and the orbito-‐frontal cortex (OFC) in other rats. In rats with ACC damage a similar pattern was observed as in the experiments of Walton et al. (2002), as well as when dopamine antagonists are used (Bardget et al., 2009; Floresco et al., 2008). This was not the case in rats with OFC damage. These rats showed a pattern of reduced willingness to wait for a higher reward in the DBRM paradigm, and instead chose the low-‐delay/low-‐ reward option most of the time (Rudebeck, Walton, Smyth, Bannerman & Rushworth, 2006). Croxson et al. (2009) showed that fMRI BOLD activity in the ventral striatum and the dACC corresponds to the expected reward and the needed effort. The ACC is active when adaptation is needed in the choice-‐and-‐control process. According to this view, activation in the ACC means a that adaptations in performance might be needed, for instance when the situation is error-‐prone or when tasks require a lot of attention (Colcombe et al., 2004).
Increasing the function of this brain area could possibly increase the functions associated with it. This increase in function can be mediated through physical exercise (Hillman, Erickson & Kamer, 2008). Six months of aerobic training does increase the brain volume of the dACC (Colcombe et al., 2006), but there is some debate about how this relates to the effectiveness of the area. Most research suggests that the ACC is less activated in adults who exercise regularly (Colcombe et al., 2004). In the experiment of Colcombe et al. (2004) conYlict-‐monitoring has been measured using an Eriksen Ylanker task. After six months of aerobic training a reduction in activation in the ACC was shown, in contrast to the prefrontal cortex (PFC), which had increased activation (Colcombe et al., 2004). According to the interpretation of Voss, Nagamatsu, Liu-‐ Ambrose and Kramer (2011), this result means that the ACC was more efYicient in processing the information provided by the PFC. An experiment by Chaddock et al. (2011) shows similar results with children. Brain activation was measured with fMRI while the participants performed an Eriksen Ylanker task. Based on their physical Yitness, children were divided into two groups; lower and higher Yitness. To study the effects of mental fatigue, the results of the task were divided into early and late trials. No difference was observed between the two groups in accuracy or brain activation in the congruent trials; children in both groups showed increased brain activation in early trials, and decreased brain activation combined with decreased accuracy in late trials. This decline in accuracy in both groups was attributed to mental fatigue. In the
incongruent trials however, only the higher Yit children showed increased brain activation in the ACC in the early incongruent trials, followed by decreased ACC activation but no decline in accuracy in the late incongruent trials. The lower Yit
children did not show an increase in brain activation in the early incongruent trials, but did show a decrease in accuracy in the late incongruent trials. This shows that ACC activation is scaled according to the level of action-‐monitoring is required in higher Yit people, whereas their performance remains on a steady level. ACC activity does not scale along with the level of action-‐monitoring required in less Yit people, which results in lower levels of performance.
Other research shows that other executive control processes like planning, inhibition and cognitive control are most strongly affected by physical exercise
(Colcombe & Kramer, 2003), while other functions are less affected. Seventeen weeks of running exercise of three times per week results in increased executive control processes, while working memory is not affected (Stroth, Reinhardt, Thöne, Hille, Schneider, Härtel & Spitzer, 2010). Physically trained older adults have less strong ACC activation than their unYit counterparts, which is evident by a lessening in error-‐related negativity (ERN) ERP amplitude in the dorsal ACC (Hillman, Erickson & Kramer, 2008). The ERN is a measure of error processing, and a lower ERN is one indication of more efYicient executive control processes (Stroth, Kubesh, Dieterle, Ruchsow, Heim & Kiefer, 2009). Fitter people show better task preparation processes, evident by an increase in contingent negative variation (CNV) ERP amplitude and a decrease in the N2 ERP amplitude, which is another marker for more efYicient executive control processes. This signal is associated with response monitoring and inhibition (Stroth et al., 2009). The prevailing theory is that physical Yitness is associated with better top-‐down control, which is a result of a more efYicient ACC, which results in better performance on tasks which are related to executive functions (Hillman et al., 2008).
One way to measure physical exercise is by using self-‐report questionnaires. However, because self-‐report questionnaires can be skewed to reYlect recent exercise practices instead of accurately measuring long-‐term exercise habits, physical measures can be useful in conjunction.
Heart Rate Variability (HRV) has been shown to be a reliable indicator of physical Yitness (Achten & Jeukendorp, 2003; Algra, Tijssen, Roelandt, Pool & Lubsen, 1993; Levy, Cerqueira, Harp, Johannessen, Abrass, Schwartz & Stratton, 1998; Tsuji, Venditti, Manders, Evans, Larson, Feldman & Levy, 1994). HRV is the variance in time between heartbeats. This difference between R-‐R peaks can vary substantially, even when heart rate is stable (Achten & Jeukendorp, 2003; Levy et al., 1998). HRV is inYluenced by the sympathetic and parasympathetic nervous system, where the parasympathetic
inYluence is responsible for fast deceleration of the heart, which is indicated by a higher HRV (Thayer & Lane, 2005). HRV is measured with indexes such as SDNN and rMSSD. SDNN is the standard deviation of all normal R-‐R intervals; this measure can show long term and short term differences between R-‐R intervals. The root mean square of all successive differences is called rMSSD and is used as an index parasympathetic
inYluence. Higher HRV is associated with better health, lower mortality risk and higher quality of life (Algra, Tijssen, Roelandt, Pool & Lubsen, 1993; Tsuji, Venditti, Manders, Evans, Larson, Feldman & Levy, 1994).
Most HRV measures are inYluenced by aerobic exercise, where higher trained individuals have higher SDNN and rMSSD during rest and activity than their less trained counterparts (Achten & Jeukendorp, 2003; Levy et al., 1998). Research shows that the effect of aerobic exercise on HRV measures is only evident after several weeks of training. SigniYicant differences in HRV are evident after 12 or 16 weeks of training, but not after Yive weeks (Achten & Jeukendorp, 2003; Amano, Kanda, Ue & Moritani, 2001). HRV decreases when physical exercise is required (Achten & Jeukendorp, 2003). Participants had lower rMSSD and SDNN shortly after performing Yifteen minutes of incremental physical exercise than they had before the intervention (Luft et al., 2009).
HRV is also a measure of cognitive events, such as increased mental load, divided attention and stress (Hjortskov, Rissén, Blangsted, Fallentin, Lundberg & Søgaard, 2004; Thayer, Hansen, Saus-‐Rose & Johnsen, 2009). HRV can be an indication of mental functioning. When participants were separated into two groups based on their resting rMSSD, the group with higher rMSSD performed better on mental tasks than the group with lower rMSSD (Hansen, Johnsen & Thayer, 2003). This difference is evident for tasks which tax executive functions (Luft, Takase & Darby, 2009: Thayer et al., 2009), but not nonexecutive tasks (Luft, Takase & Darby, 2009).
HRV seems to be linked to the ACC. Some patients with lesions in the ACC have unchanged rMSSD when performing cognitive tasks such as the N-‐back task, or physical tasks such as applying manual grip pressure. People without such lesions show sharply decreased rMSSD when performing these tasks (Critchley, Mathias, Josephs, O’Doherty, Zanini, Dewar, et al, 2003).
Taken together, this means that the ACC might very well be responsible for HRV through mediation of the parasympathetic nervous systems, while also being
responsible for an important part of decision making and executive functions in general, which can be positively affected by physical exercise.
A difYiculty when studying associations like these is that some psychological lab experiments have low ecological validity. The scenarios studied do not always
correspond well to real-‐world situations. Tasks which are often used in psychological research such as the N-‐back task (Kirchner, 1958), random-‐dot task (Julesz, 1971) and the Eriksen Ylanker task (Eriksen, 1974) have high internal validity, and allow for a high degree of experimental control, but they do not allow for the kind of generalisation that Yield research provides. However, Yield research does not always allow for the
experimental control needed to accurately measure the desired constructs. Within the Yield of psychological research, a need has emerged for research methods that combine high ecological validity with high internal validity (Bohil, Alicea & Biocca, 2011). An interesting option has arisen in recent developments in Virtual Reality (VR) combined with Head-‐Mounted Displays (HDMs). These devices could possibly form a bridge between more traditional lab research and Yield research. This is possible due to the fact that virtual worlds can be created which resemble our own world, but with the advantage of having high experimental control. In VR it is possible to create a world which provides multiple sensory stimulations, such as visual,
auditory and kinetic. Some degree of kinetic stimulation can be achieved by using sensitive head-‐tracking systems, where the user’s head movements are translated to movements in the software, and using input devices which resemble the devices used in the VR world. Adding sensory stimulations increases the level of immersion of the user, which causes the user to feel more present in the simulated world (Bohil et al., 2011). However, is not necessary to have perfect sensory integration to create a high level of presence; reading a good book achieves high levels of presence by calling upon the fantasy of the reader to create this effect (Bohil et al., 2011). To achieve this feeling of presence it is important to have an absence of breaks within the stimulus such as interruptions, errors or glitches.
Increasing the level of immersion has effects on physiological measures, which is an indication that it is experienced as more realistic than methods with lower levels of immersion. Adding tactile markers or increasing the video frame-‐rate in VR can
signiYicantly increase stress measures such as heart rate and skin conductance (Meehan, Insko, Whitton & Brooks, 2002). It has been shown that experimental techniques can have more realistic results when they are more immersive when compared to less immersive measures. Participants are better able to assess value (Bateman, Day, Jones & Jude, 2009) and risk (Fiore, Harrison, Hughes & Rutström, 2009) when the scenarios are presented with more immersive techniques.
Since HRM based VR technology as used in present research is a recent
development, previous research does not use the same technology as available today, so generalisations of those results to current research should be viewed with some skepticism. Regardless of those limitations, a trend is visible where more immersive methods provide better and more realistic research results than less immersive methods.
A setup of a HMD and VR software has some advantages. It allows for a cheap and mobile experimental setup. Participants can be virtually transported to realistic
locations and situations which would otherwise not be possible due to budgetary, safety or ethical concerns. Within these worlds the variables of interest can be manipulated freely by the researcher, creating high levels of control.
The experiments described in this thesis investigates an EBDM paradigm at multiple levels of immersion. The way the reward is presented will be held constant over all levels of immersion, while the way the effort is displayed varies over the conditions. It is expected that the effort is perceived as more realistic when presented in a more immersive environment. It is expected that in the more immersive conditions the participants will perform less effort for the same amount of reward than in the less immersive conditions due to the differences in perceived effort between conditions. Because physically Yit participants have better executive functions it is expected that they require less reward per effort than less physically Yit participants. Because HRV correlates both to aerobic Yitness, physical effort and mental effort, it is expected that this measure differentiates between higher Yit and lower Yit people, and between the different levels of immersion.
In experiment one three conditions have been used to measure the effect of different levels of immersiveness. In this experiment a questionnaire has been used to determine physical Yitness. In experiment two the effect of different levels of
immersiveness have been measured using two conditions. These conditions have been performed with and without the addition of wrist-‐weights, which were used to
increase the effort, creating four conditions. Physical Yitness has been measured using a questionnaire and a physical HRV measurement.
Methods Experiment One Participants
A total of thirty-‐eight people (12 male) participated in experiment one. The average age of was 23 years old (min: 19, max: 28). Research funding of €20,-‐ or two research credits was available for the Yirst segment of the participants. Due to
budgetary constraints the remainder of the participants received either a box of Merci chocolates or two research credits. The participants were promised an incentive, where they could earn more money depending on the amount of effort they chose to do. The maximum of this incentive was €2.50. The participants were recruited using the university required lab hours system DPMS, through Ylyers around the university campus, through street recruitment and through friends and family.
Materials
An Oculus Rift DK2 HMD was used in the VR condition. A standard Ylat screen monitor was used for all questionnaires, the listening span task (not used in this thesis), and both 2D and 3D conditions of the task. An X-‐Box One controller was used for the task. It was taped to the table so it could not be picked up, and had to be operated using the balled Yists.
A non-‐operational webcam was used to incite a sense of being monitored, to insure task-‐adherence. Questionnaires were presented using Google Forms.
The Short QUestionnaire to ASses Health enhancing physical activity (SQUASH) by Wendel-‐Vos, Schuit, Saris & Kromhout (2003) is used in this thesis. It measures how much physical movement and exercise is performed in a typical week over the last few months. It has 43 items (over 15 categories). An intensity-‐factor between one and nine is given to each activity depending on the amount of physical strain this activity costs. For example, an intense cycling exercise has an intensity-‐factor of six, whereas an average session of DIY around the house has an intensity-‐factor of two. This intensity-‐ factor is multiplied by the amount of time spent on each activity per week.
From these data a number of measures can be calculated. The Activities Hours per Week index states how many hours per week the participant is performing general activities such as walking or cycling to and from work and performing general
housework duties. The Activities Activity score multiplies the Activities Hours per Week index by the intensity factor of each activity. The Sports Hours per Week index is similar to the Activities Hours per Week index but reYlects sports related vocations. The Sports Activity score multiplies the Sports Hours per Week index with each sports’ activity index, similar to the Activities Activity score. The Total Hours per Week index is a summation of the Activities Hours per Week and the Sports Hours per Week indexes. The Total Activity score equals the Activities Activity score and the Sports Activity score combined. The SQUASH has a good reproducibility (Spearman’s correlate: 0.58).
The Igroup Presence Questionnaire (IPQ) was used to measure sense of presence. This questionnaire has 14 items for each of the conditions (42 total) about the amount of immersion experienced during each of the tasks. Scores can range from 14 to 70 on each of the conditions. Higher scores mean more immersion experienced. Sample question: “I was no longer aware of my actual surroundings during the task”. Possible responses were: completely disagree, partly disagree, neither disagree nor agree, partly agree, completely agree.
Other questionnaires were obtained but not used in this thesis (Locus of Control questionnaire, Temporal Experience of Pleasure questionnaire, Social Economic Status Questionnaire, Immersive Tendencies Questionnaire, Game Experience Questionnaire, Listening Span task).
The task consisted of three times the exact same EBDM task, with various amounts of immersion created by increasing amounts of visual and kinetic stimuli.
In the 2D task the two options were presented, one on the left and one on the right, see Image 1. The required effort was presented in coloured blocks (green = low effort, orange = medium effort, red = high effort). Above each option the amount of reward was presented numerically. After choosing which option the participant wanted to do, a powerbar was presented, indicating if the amount of effort expended was high enough, see Image 2. If the effort expended was too low the powerbar dropped down and turned from green to orange. During the medium effort of the trial the powerbar dropped faster than in the green parts, whereas it dropped fastest in the high effort parts of the trial. The words “Choose track”, “Waiting for track”, “Get ready” or “Go!” provided additional information on which stage of the task the participant was at the moment. The background of the 2D task was black, and no other visual stimulation was given. There was no auditive stimulation. Both tracks took the same amount of time to complete, regardless of the amount of effort needed.
Image 1. Visual representation of 2D condition during effort/reward selection.
Image 2. Visual representation of 2D condition during trail performance.
The 3D version of the task was similar to the 2D version, but instead of a black background a representation of a manually operated train-‐cart was presented. At the Yirst stage of each trial the effort was still displayed by the coloured blocks, see Image 3. However, while operating the train cart, the effort was displayed as shrubbery
view was directly above the seat in the train cart, with the two handlebars which
operated the cart directly in view.
Image 3. Visual representation of 3D condition during effort/reward selection.
Image 4. Visual representation of 3D condition during trail performance.
The VR version was identical to the 3D version, only presented through the Oculus Rift DK2, see Image 5 and Image 6. This way the participant was able to look around freely.
Image 5. Visual representation of VR condition during effort/reward selection.
Image 6. Visual representation of VR condition during trail performance.
The Point of Indifference (POI) was measured in each of the conditions. It is a measure where the participant chooses the high-‐effort/high reward (HE/HR) option 50% of the times. It means that the reward in the HE/HR option is exactly that much higher than in the low-‐effort/low-‐reward (LE/LR) option, that the difference in the required amounts of effort is exactly covered by the amount of reward (Bardgett, Depenbrock, Downs, Points & Green, 2009). A high POI means that a lot of extra reward needs to be given to the participant for the extra effort required in the HE option. A low POI means that the extra effort is not considered that hard, and less reward is needed to perform the HE option. A POI of 0 means that the effort is not taken into account at all, and that simply the highest rewarding option is chosen. The POI is a reliable measure for the relationship between effort and reward (Westbrook et al., 2013). The last Yive trials of each condition were used for another measurement not used in this thesis. The four trials previous were considered to be where the participants reached their POI. The average reward scores of these four trials were taken and used as POI.
Procedure
Participants were tested in separate cubicles. The participants read the information brochure and the informed consent form, after which the latter was signed. Depending on which counterbalancing order the participant was assigned to, the questionnaires and listening-‐span task were performed Yirst, or the task was started directly. If the participant had to start with the questionnaires and listening-‐ span task, they were escorted to the cubicle where the questionnaires were obtained using Google Forms. This took roughly thirty minutes. After completing the
questionnaires, the research assistant explained the listening-‐span task, after which the participant completed this task. It took roughly twenty minutes. If the participant had to start with the task directly, they were escorted to the cubicle where the task was obtained, where the research assistant explained the task and provided a live
demonstration. Depending on the counterbalancing the participant started in the 2D, 3D or VR condition.
In the task itself the participant could choose between the track on the left and the track on the right by moving the joysticks of the controller in that direction. The joysticks were operated using the proximal and intermediate phalanges of the pinky Yingers of the balled Yists, which rested on top of the small joysticks of the X-‐Box controller. After choosing the desired track the participant had to move their Yists in alternating fashion backwards and forwards. They had to move just fast enough so that the powerbar neared full, but not faster.
Each condition took roughly seventeen minutes, with a total of the entire task of a little over Yifty minutes. After completing the task the IPQ was obtained using Google Forms. Each participant was debriefed, after which the research credits or cash reward were handed over. The €2.50 added bonus was presented to each participant,
regardless their level of effort.
Changes
The X-‐Box One controller was considered to be irritating by some participants, due to the unconventional way of operating it. It was replaced by two Logitech joysticks for the second experiment. The joysticks were more similar to the presentation on screen than the X-‐Box controller was. It was also thought to increase the effort needed. In experiment one HRV data was collected originally. Unfortunately this data was lost due to theft. Seventeen out of 39 participants were considered to be non-‐responders due to a POI of zero. It was thought that the amount of effort was not sufYicient for these participants to take this into consideration when choosing between the HE or the LE track. Conditions were added where weights were added to the participants’ wrist to increase the effort needed and to decrease the amount of non-‐responders in
experiment two. Due to time constraints and the addition of the wrist weights, the 3D condition was removed from the second experiment. The reasoning was that if the amount of immersion would inYluence the POI, this difference should be visible between the 2D and the VR conditions.
Methods Experiment Two Participants
A total of thirty-‐six participants (21 men) completed experiment two. The average age was 25 years (min = 18, max = 44). Participants received a box of Merci chocolates, or two research credits as compensation. The participants were recruited via the university research credits website DPMS and through friends and family.
Materials
An Oculus Rift DK2 HMD was used in the VR task. The questionnaires and the 2D version of the task were presented on a standard Ylatscreen monitor. Two Logitech Attack 3 joysticks were used as input devices. Two 0.5KG wrist-‐weights were used in the weighted condition. A noise cancelling headset was used to block extraneous noise. A Polar H7 bluetooth heart beat monitor (chest-‐strap style) was used for HRV
measurements. HRV data were logged on an iPhone 4S using the application HRV Logger (Marco Altini, 2015) which connected to the Polar H7 HRM via bluetooth. Google Forms was used for questionnaires. SQUASH and IPQ were used in this
experiment. ITQ and Game Experiences Questionnaire were also taken, but not used in this thesis.
The amount of effort required was changed slightly in this version of the task compared to the task in experiment one. Green blocks considered no effort, and the powerbar reYlected this change. Orange and red blocks were still medium and high effort parts respectively. Some visual aspects were changed in the 2D version to better show which part of the track the participant was at any moment. This was done by moving the powerbar to the middle of the screen while displaying the current effort requirements.
The last four trials were considered to be the POI of the participant. The average of the reward scores of these trials were used as value for POI.
HRV were measured during the time when the participant reached their POI. The HRV Logger application has built in formulas for HRV calculations, and these data were matched to the POI data using timestamps imbedded in the POI data Yiles.
Procedure
The participants read the information brochure and the informed consent form, after which the latter was signed. The participants were told they could earn more chocolates depending on the amount of effort they were exerting. These chocolates were displayed in the area where the participants were Yirst received. Next the
participants Yilled out the SQUASH, ITQ and Game Experiences questionnaire, this took roughly 20 minutes. Once this part was completed the participants were Yitted with the Polar H7 heart beat monitor, and the connection to the iPhone application was tested by the research assistant. The participant was asked to sit still for two minutes to gain
HRV baseline. Once baseline was completed the research assistant explained the task requirements and provided a live demo. Depending on the counterbalancing the participant was Yitted with the wrist-‐weights or not.
Depending on the counterbalancing the participant was Yitted with the Oculus Rift DK2 HMD, or started with the 2D version of the task. Once the participant was clear on what was expected the experiment began. After each trial block the research assistant assisted the participant in taking off or putting on the wrist weights and/or the Oculus Rift, depending on the counterbalancing. The task took approximately 40 minutes to complete. After completion of the task the participant Yilled out the IPQ, which took approximately 10 minutes. Finally a debrieYing was done by the research assistant and the compensation was provided. If the participant had no more
questions the experiment was completed.
Results Experiment One
Of the 38 original subjects in the Yirst experiment, 16 had a POI of three or lower, indicating that the required effort had no effect, the experimental manipulation had failed, and their performance was based on reward only. The mean age of these participants was 23 years (min 19, max 28). These non-‐responding subjects were omitted from all calculations involving POI, but remained in the rest of the calculations. Of one participant the questionnaires were not properly saved due to technical
difYiculties. Three subjects made errors in the SQUASH questionnaire. The subjects reported riding their bikes for excessive durations (22, 25 and 30 hours per day). Their SQUASH scores were omitted from calculations.
The average POI scores are presented in Table 1 and Figure 1:
A one-‐way within-‐subjects ANOVA was conducted on the mean POI scores. The
main effect of condition was not statistically signiYicant: F(2,42) = 0.031, p = 0.969. There was no difference in POI between the 2D, the 3D and the VR condition. This goes against the main hypothesis of this thesis, stating that the POI of the VR condition should be higher than the 3D condition, which should be higher than the 2D condition.
The average IPQ scores are presented in Table 2 and Figure 2:
A one-‐way within-‐subjects ANOVA was conducted on the mean IPQ scores. The main effect of condition was signiYicant: F(2,42) = 157.557, p < 0.01. There was a signiYicant lineair trend: F(1,21) = 352.238, p < 0.01. The subjects reported feeling much more present in the VR condition than in the 3D condition, and much more present in the 3D condition than in the 2D condition. No signiYicant correlations between the POI measures and 2D IPQ (r = -‐0.24, N = 22, p = ns), 3D IPQ (r = 0.16, N = 22, p = ns) or VR IPQ (r = 0.182, N = 22, p = ns) were found. This means that the level of perceived presence in a certain condition did not correspond to the POI in that
condition.
The average Exertion scores are presented in Table 3 and Figure 3:
A one-‐way within-‐subjects ANOVA was conducted on the mean Exertion scores. The Mauchly’s test of sphericity was signiYicant (χ2(2) = 10.901, p = 0.004), so the
Greenhouse-‐Geisser Epsilon was used. The main effect of condition was signiYicant: F(2,42) = 5.720, p = 0.015. There was a signiYicant quadratic effect: F(1,21) = 7.789, p = 0.011. This means the perceived effort was higher in the VR condition than in the 2D or 3D condition, which were similar.
Pearson’s correlations were calculated between perceived exertion and POI. There was a signiYicant positive correlation between exertion and POI in the 2D condition: r = 0.456, N = 22, p = 0.016. There was neither a signiYicant correlation between exertion and POI in the 3D condition (r = -‐0.271, N = 22, p = ns) nor the VR condition (r = -‐0.06, N = 22, p = ns). This means that the perceived effort in the 3D and VR conditions had no relationship with the POI, but that such a relationship between perceived effort and POI did exist in the 2D condition.
Correlations were calculated between the POIs of the different conditions and the various SQUASH measures. The Activities Activity index correlated signiYicantly with the 3D POI (r = 0.502, N = 21, p = 0.01) and the VR POI (r = 0.485, N = 21, p = 0.013). The Activities Hours per Week index correlated with the VR POI (r = 0.377, N = 20, p = 0.046). The Total Activities index correlated positively with the 3D POI (r = 0.405, N = 20, p = 0.038).
If however a Bonferroni correction for multiple comparisons was used, and thus the desired p value should be below 0.0028 (0.05/18 comparisons) (Field, 2013), none of the above mentioned correlations between POI and the SQUASH measures were signiYicant.
If the conservative Bonferroni correction is disregarded, these results mean that people who reported higher levels of daily activities and higher levels of total activities had higher POIs in the 3D and the VR conditions than people who had reported lower levels of activity. This is not in line with the predictions made in this thesis, because higher levels of reported activity were expected to correlate with lower levels of POI. If the Bonferroni correction is upheld, the levels of POI and reported activity do not relate to each other, which is also against the hypothesis of this thesis. The use of the
Bonferroni correction will be discussed further in the discussion paragraph.
No correlations were found between either of the Sports indexes and any of the POI measures. This result is not in line with the hypothesis of this thesis regarding the level of sports and POI. It was expected that people reporting higher levels of sports would have lower POIs than people who report lower levels of sports.
Correlations between perceived exertion and various SQUASH measures were calculated. SigniYicant negative correlations were found between perceived exertion in 2D and the Activities Hours per Week index (r = -‐0.428, N = 36, p = 0.005), the
Activities Activity index (r = -‐0.396, N = 35, p = 0.009), the Total Hours per Week index (r = -‐0.418, N =36, p = 0.006) and the Total Activities index (r = -‐0.331, N = 36, p = 0.032). This means that higher daily activity levels and higher total activity levels are related to lower perceived exertion scores in the 2D condition, which is in line with the predictions.
The perceived exertion in VR and the Sports Activity index show an unexpected positive correlation (r = 0.316, N = 33, p = 0.036). This is not in line with the
predictions and goes against the expected hypothesis. It could be the case that people who play more sports also work harder in VR than people who play less sports. If this result however is combined with the absence of a signiYicant correlation between POI and perceived exertion this interpretation seems unlikely. It seems like people who play more sports perceive higher levels of exertion, however they do not actually work harder in the corresponding condition.
No other correlations were signiYicant.
If a Bonferroni correction for multiple comparisons was used, and thus the desired p value should be below 0.0028 (0.05/18 comparisons) (Field, 2013), none of the above mentioned correlations between POI and exertion were signiYicant.
Results Experiment Two
Twelve out of 36 participants did not reach a POI of three or higher, and were excluded from all calculations involving POI. Since these participants did not reach a POI of three or higher they were considered non-‐responders as far as the data concerning POI, for the same reason as the non-‐responders in experiment one. The mean age of these participants was 27 (min = 22, max = 38). The participants whose data were omitted did not differ as a group from the rest of the participants. The scores from the group whose POI data were omitted were still used in all other calculations. One participant reported feeling nauseated in the VR condition and only completed the 2D conditions. None of the participants were considered outliers in the SQUASH
Table 1
POI Means and Standard Errors in 2D, 3D & VR conditions
Mean Std. Error
POI 2D 18.318 3.593
POI 3D 17.000 3.809
POI VR 17.500 4.528
Table 2
IPQ Means and Standard Errors in 2D, 3D & VR conditions
Mean Std. Error
IPQ 2D 21.909 1.059
IPQ 3D 34.682 1.664
IPQ VR 49.500 1.133
Table 3
Perceived Exertion Means and Standard Errors in 2D, 3D & VR conditions Mean Std. Error Exertion 2D 1.523 0.304 Exertion 3D 1.455 0.210 Exertion VR 2.523 0.363 Me a n PO I 12 14 16 18 20 22 Condition 2D 3D VR
Figure 1. POI averages in the three presentation conditions. Error bars represent Standard Error. n = 22 participants.
POI = Point of Indifference, 2D = 2D condition, 3D = 3D condition, VR = Virtual Reality condition.
Me a n I PQ 18 23 28 33 38 43 48 53 Condition 2D 3D VR
Figure 2. IPQ averages in the three presentation conditions. Error bars represent Standard Error. n = 22 participants.
IPQ = Igroup Presence Questionnaire, 2D = 2D condition, 3D = 3D condition, VR = Virtual Reality condition.
Me a n Ex e rti o n 0.5 1 1.5 2 2.5 3 Condition 2D 3D VR
Figure 3. Perceived Exertion averages in the three presentation conditions. Error bars represent Standard Error. n = 22 participants. 2D = 2D condition, 3D = 3D condition, VR = Virtual Reality condition.