Presence is often used as a quality measure for virtual reality experiences. It refers to the sensation of \being there" that users feel while wearing a head-mounted display. In contrast, simulator sickness refers to the feeling of unease of some users while experiencing virtual motion. Nowadays, many virtual reality games do not allow the player to walk, trying to minimize the generation of unpleasant symptoms. This study explores how presence is a ected by the ability to walk in VR games, as well as how simulator sickness actually grows when the player takes a virtual stroll. For this purpose, two prototypes of a small puzzle were built. In the rst one, the player is able to walk, whereas the second one does not allow the user to move in any way. Presence and simulator sickness were measured using standard questionnaires while real players faced our puzzle. The results point to a strong correlation between the action of walking and an increment of the level of presence achieved by the subjects. However, there is no clear correspondence between walking and simulator sickness in our experiment. This last observation opens the way for further research and questioning of early studies about simulator sickness, as technical di erences between current virtual reality devices and older ones may in uence how uncomfortable users feel while wearing them.
In spite of the big success virtual reality (VR) is experiencing these days, most games designed for VR headsets do not include one fundamental feature: the ability to walk in a rst-person (FP) perspective through the virtual environment. The reason for making this decision is that both academics and developers seem to agree that such an immersive virtual movement produces a disagreeable e ect known as simulator sickness (SS). However, this presumption is based on dated papers [1{3] which were not written taking into consideration current advances in VR technologies and experiences. We are interested in knowing how relevant walking inside a VR environment is for achieving higher presence levels, as well as checking whether the mere act of walking actually increases SS.
Thus, the aim of this research is to measure how presence is related to the action of walking in a FP VR environment, and to check how this relation is connected to SS when using a speci c VR device: the Oculus Rift Development Kit 2. The alternative hypothesis (H1) is that presence increases when the user is able to walk in a VR experience, while the null hypothesis (H0) is that presence does not change at all.
Usoh, Arthur, Whitton et al. [
Presence [5{7] is a widespread way of measuring the capacity of an interactive
experience to make its users feel \they really are" in a virtual environment.
Experts in this eld, however, refer to this concept in a variety of ways [
Besides, now that virtual reality is becoming a commonly available
technology, and considering how high is the game industry's contribution to its
popularization, it would be useful to gather presence measures and relate them to SS
uctuations. Simulator sickness is a byproduct of high quality simulation
technologies [
The concept of SS refers to the feeling of unease generated by a simulation in which the user experiences virtual motion that is not synchronized with real movements of his or her body. Devices like the HTC Vive let users move to a certain extent, and re ect real-world movement inside the experience with very low latency (with a target of around 11 ms from moving the head to perceiving the e ects of that movement). However, their tracking capacities are not enough to allow a player to walk long distances in VR | just up to an area of 5 x 5 meters.
Previous studies [
Earlier studies [4, 14{16] suggest that virtual walking representation techniques have in uence over presence in virtual environments. However, they do not go deep into the relation between staying still or walking and presence. That is why we designed the following experiment, aiming at discovering how an independent boolean variable (walking or not walking) is related to two numerical variables: presence and SS. 2
During this investigation, we used a series of principles to design two virtual
reality experiences which constitute the main core of our experiment. Sherman &
Craig [
In our simulation, walking was similar to what Usoh, Arthur, Witton et al.
[
So as to measure the e ect of walking on presence and SS, we built two slightly di erent VR game prototypes. Both of them were based on a small puzzle that had to be resolved to trigger the end of the experience; however, they presented di erences in the distribution of in-game elements, as we will explain later.
These prototypes were made in Unreal Engine 4, and used some free assets from the video game In nity Blade (made available for use with this engine by its developer, Epic Games), together with some self-made assets.
The above-mentioned categories de ned by Sherman & Craig [
A neutral subject was asked to identify all actions and elements present in both experiences while wearing a HMD, so as to ensure the noticeable core elements were the same. This subject is a 25 years old male, with a background in Computer Science, and no knowledge about this experiment beforehand. When he was told to separate actions and game elements present in the two prototypes, he identi ed the following ones: { Game actions: move, look at, interact and listen. { Game elements: walls, oor, rocks, ceramic and metal objects, windows, ceiling, pedestals, orbs of three di erent colours, statues, re, fog e ects, sounds and ambiance.
The rst scenario (W ) (see gures 1 & 2) consisted on a 3D environment that the user could explore by using the speci ed HMD and game controller. This environment contained a puzzle which could be resolved by following a series of steps: the player had rst to walk towards one of the three statues (S), located behind a pillar (P ) on top of which there was a lit orb. This orb could be red, green or blue, depending on its position. Once in front of one of the statues, the player had to extinguish a re that the statue held in her hands by pressing the 'A' button in the controller (an in-game tooltip indicated which button to press). The correspondent orb would also turn o as a result, leaving only the other two lit. The player had to repeat this process until no orbs were lit. Then, a gate would open in the place where the player rst spawned (I ). If the player stepped through it, the experience ended, and the puzzle was considered resolved.
S P WPSI P S P S NPSWI S P Fig. 1. Schematic representation of the layout of the two prototypes.
Fig. 3. View of a section of the 3D environment designed for not walking.
The second scenario (N W ) (see gures 1 & 3) was designed in a similar fashion, and contained the same puzzle, but in-game elements were distributed in a way that made all of them accessible from the player's spawn location. The only movement allowed for the player in this case was to rotate their head | and so, the camera view. The solution to the puzzle was, however, the same: the player had to turn o three res so that the gate appeared in the spawn location. This gate acted as a portal in this case, activating and ending the experience as soon as the lights turned o . 2.1
The measures collected during the experiment were the following:
{ Presence questionnaires: The Slater-Usoh-Steed Questionnaire (SUS) [
Only the engagement, spatial presence and perceptual realism subscales of the TPI were used, due to the nature of our experiment. Social presence, social richness and social realism were not interesting in this case, as there are not social interactions throughout the experiences.
All users were asked simply to play each game, and then to answer a series of questions about it. The order in which questionnaires were o ered to the subjects was determined by the need of passing the SSQ rst, while the potential e ects of SS were still active. SUS and TPI came after it, in that order. 2.2
Our experimental group was a set of 12 subjects that went through both experiences, so as to be able to compare their reactions to them. Half of them did the two experiments in one order; while the other half did them inversely. Besides, none of the subjects who took part in the experiment knew about the contents of the second experience before nishing the rst one.
To participate in this experiment, all subjects had to sign-up by answering an email which retrieved the following pieces of information: { Name. { Age. { Gender. { Education. { VR device familiarity. { VR devices present at home. { Video game familiarity.
Among the subjects, 11 (91.67 %) were men, and 1 (8.33 %) was a woman. Their ages ranged from 18 to 26 years old. About their background, 8 (66.67 %) had an education in Computer Science, while 3 (25 %) studied a degree related to Video Game Development and 1 (8.33 %) had studied Graphic Design. About VR, 4 (33.33 %) of them did not have any familiarity with VR devices, and had never tested a HMD, while the rest had used at least one VR system; 8 (66.67 %) did not own any VR device, and 4 (33.33 %) had a Google Cardboard at home. All of them said to be familiar with video games, and declared themselves frequent players. 3
The obtained results point to a strong correlation between the action of walking and an increment in presence. Both SUS and TPI tests presented a p-value much lower than 0.05, and both o ered very similar normalized results, as can be seen in tables 1 and 2. This is relevant if we consider that in-game movement did not imply real-world motion to any extent. The lack of correspondence between in-game and real-life motion does not seem to a ect how presence rises when the user is able to walk (or y, as we could call it) in VR.
Figures 4 and 5 show subject numbers in the abscissa axis, and normalized levels of presence achieved in the ordinate axis. Figure 4 presents a more irregular shape than gure 5, but this seems to be caused by the di erent number of questions in both tests. The rst one had only 6, whereas the version of the TPI that we used had 18, which allowed for a less polarized mean. The correlation is, in both cases, high enough to allow us to think the results of these tests are consistent.
Yes 12
The SSQ, however, o ered results which, in spite of varying between both prototypes, did not present a signi cant correlation (with a p-value of approximately 0.12) when processed with a one-way ANOVA, as can be seen in table 3. Walking did not seem to notoriously a ect SS. Only subject number 7 (see gure 6) presented a clear increment in SS while walking. However, this subject had de ned himself, before taking part in our experiment, as especially prone to nausea and motion sickness.
Yes 12 2.416667 0.201389 0.045718 Notwithstanding the strong relation between the SUS and TPI tests results, the reduced group of subjects we had to work with leads us to think the obtained data can be considered reliable just to a certain extent, and should be broadened in the future. Besides, most voluntaries who took part in the experiment were men, which means there could exist a genre bias in the data.
Though the results of both SUS and TPI questionnaires were consistent, it would also have helped to count on a strongest numeric measure, such as the data from an electrocardiogram (ECG). This was not possible due to limited time and resources, but would be desirable in future research.
SS was not measured over time, as we wanted to check if only the mere act of walking increased it. However, we do not know the e ects a higher level of SS granted by a longer experience may have when comparing the results of both prototypes. Besides, the order in which both experiences were taken could have in uenced the level of SS achieved.
Moreover, all of our subjects said to have a notable experience in playing video games, and some of them had already tested VR devices, which means they could be more used to virtual motion, and therefore more resistant to simulator sickness in short and focused experiences as those of our experiment.
In conclusion, the results of this experiment negate the null hypothesis (H0) and enforce our alternative (H1): there is a signi cant di erence between walkable and non-walkable rst person VR games when comparing the independent variable to presence. Being able to move longer distances than what the tracking capacities of the HMD allow, even if there are not real body movements associated with the act of walking, generates a clear increment in the level of presence achieved by the player.
On the other hand, judging solely by the results of this experiment, there does not seem to be a clear correlation between the act of walking and SS. Though this may have been caused by the lack of a broader set of subjects, the stable nature of the results obtained from SUS and TPI presence tests with the same, narrow group, leads us to think there is actually a lack of SS generation in our game prototypes. This opens a new line of investigation, as it would be useful to know if it is related to the technical properties of current VR devices, the design principles assumed by the authors, or to something else.
Current game developers could bene t from this combination of results. Most VR experiences lack the possibility to walk the environment due to a fear of SS; however, brief and simple experiences like the ones we designed for this experiment have shown no clear correlation between walking and SS. This paves the way to more walkable experiences in the future, as we think player movement is a fundamental asset in modern video games. It allows for complex interaction while increasing presence, and di erentiates video games from static experiences which also use a HMD, like VR video.
Acknowledgements We would like to thank all volunteers who took part in our experiment for giving us their time and e ort.
This research was supported through EC funding for the project WHIM 611560 by FP7, the ICT theme, and the Future Emerging Technologies (FET) programme.