Changes in Navigation over Time: A Comparison of Teleportation and Joystick-Based Locomotion

Little research has studied how people use Virtual Reality (VR) changes as they experience VR. This article reports the results of an experiment investigating how users’ behavior with two locomotion methods changed over 4 weeks: teleportation and joystick-based locomotion. Twenty novice VR users (with no more than 1 hour prior experience with any form of walking in VR) were recruited. They loaned an Oculus Quest for 4 weeks on their own time, including an activity we provided them with. Results showed that the time required to complete the navigation task decreased faster for joystick-based locomotion. Spatial memory improved with time, particularly when using teleportation (which starts disadvantaged to joystick-based locomotion). In addition, overall cybersickness decreased slightly over time; however, two dimensions of cybersickness (nausea and disorientation) increased notably over time using joystick-based navigation.


INTRODUCTION
Navigation is among the important tasks in most Virtual Reality (VR) applications.It is the process by which people identify where they are, where everything else is, and how to reach specific objects or zones [18].Navigation includes both wayfinding (the mental decision-making procedure by which a motion is intended) and traveling (the actual motion from the existing spot to the new spot) [6].
Navigation is a crucial aspect of VR that allows users to move around and interact with the virtual environment.Several studies have highlighted the importance of navigation in VR, as it can enhance the user's sense of presence, immersion, and overall experience [19,27].A study by Cummings and Bailenson [12] found that navigation in VR, specifically through walking, improved the user's sense of immersion, presence, and overall experience compared to using a joystick to move.The authors suggest that walking in VR can provide a more natural and intuitive means of navigation, leading to a more engaging and realistic experience.
Locomotion (how people move through space) is a component of navigation in VR applications.Many different locomotion techniques have been developed to allow users to travel inside Immersive Virtual Environments 16:2 • M. Nasiri et al.
(IVEs), including using game controllers or joysticks [11,32], walking in place [37], teleportation [8], redirected walking [29], and arm swinging [10].Some locomotion methods enable users to walk through IVEs physically; however, most instead require users to move through an IVE using some level of abstraction (e.g., a handheld device or via partial gait).
Due to the recent availability of consumer Head-Mounted Displays (HMDs), people are using HMDs in all sorts of different locations.This underscores the need for locomotion methods that allow users to move through large IVEs when occupying a small physical space or even seated.Joystick-based motion and teleportation are commonly used locomotion methods that support motion through an IVE, even while seated or standing in a small area.When using joystick-based locomotion, users employ some sort of controller (e.g., a joystick, a touchpad, or a trackball) to direct their motion in the IVE.For teleportation, users first indicate where they want to move to, activate the teleporter, and are instantaneously moved to that position.They can indicate where they want to move using a controller or pointing gestures [4,7,8].
Although many aspects of locomotion in VR have received extensive research, very little work has considered how locomotive behaviors and effects might change over time as users become more accustomed to navigating IVEs.As HMDs were rarely encountered outside of a lab before 2016, most locomotion research before this was likely conducted with VR novices who had no prior experience with the technology.However, as this is no longer the case, it is important to consider whether locomotive behaviors and effects may evolve over time with user experience.This article specifically studies locomotive behaviors and effects that may evolve over time when using teleportation or joystick-based locomotion.These techniques were selected because they are commonly used in consumer applications.Joystick-based locomotion is perceived as an easy-to-use method due to the users' familiarity with controllers [5], and teleportation is a mainstream VR locomotion technique, fully supported by commercial HMDs, such as the HTC Vive and the Oculus Rift [8].
In this article, we report the results of a longitudinal experiment where participants completed a navigation activity four times during a 4-week period.Before beginning the experiment, participants were randomly assigned to one of two between-subject conditions: (1) using teleportation for locomotion or (2) using joystick-based locomotion.Participants were assigned 10 navigation trials each time they performed the activity, during which the time required to complete the task and their accuracy with a spatial updating task was measured.We hypothesize that the change in participants' locomotive performance would be influenced by the locomotion technique they use in VR.In addition, we expected to see an effect of time/experience on participants' performance on locomotion and navigation in an IVE.
Our research questions were as follows: RQ1: Will participants' locomotive behaviors change between sessions as they become more familiar with VR? RQ2: Will a different pattern of change be observed between locomotive methods?RQ3: Will the effects of the locomotion method on simulator sickness change with time?RQ4: Will the effects of the locomotion method on presence change with time?

RELATED WORKS 2.1 Locomotion and Navigation
VR locomotion is an essential interaction element that facilitates navigation in VR environments [8,16].Since the early days of VR development, different locomotion techniques have been designed and analyzed to make efficient and user-friendly navigation in virtual environments [4].The literature review established the diverse character of the different VR locomotion techniques under comparative settings.Joystick locomotion employing artificial interaction allows for a less physically intense experience, with the user being stationary and simply using a controller; however, it can be cognitively intense, and it can lead more easily to VR sickness [14].Teleportation effortlessly takes the users from place to place, but the visual jumps may ruin the immersion and spatial orientation [5].
Langbehn et al. [22] compared joystick, teleportation, and redirected walking.Their results showed that travel time was the shortest for teleportation, and the joystick had the highest VR sickness.In addition, no difference has been found in presence scores between the three locomotion methods, but teleportation and redirected walking were most preferred [22].As well, comparing teleportation to three virtual locomotion techniques, including the joystick method, joystick with tunneling (with a restricted field of view), and body tilt, showed no difference in presence.However, the quality of the experience was significantly higher for teleportation [40].
Coomer et al. [11] examined four locomotion methods: arm cycling, joystick, teleportation, and point-tugging (users select a point in space and pull themselves toward it by pushing on a button on a controller).They found that teleportation and arm-cycling had lower simulator sickness than joystick and point-tugging.In addition, users walked farther by teleportation than the other three methods.Moreover, teleportation caused more spatial disorientation than others, so they looked around more [11].
Navigation includes wayfinding and traveling, which are closely affiliated and used.Thus, the travel technique may affect the ability to perform wayfinding tasks and the user's spatial orientation [6].One study examined the effect of three locomotion techniques (joystick, pointing-and-teleporting, and walking-in-place) on object location learning and recall.Participants were asked to memorize the location of a virtual object in a virtual environment.Unexpectedly, results indicated that the average placement error-the distance between the original and recalled object location-was approximately the same for all locomotion techniques [43].

Prior Experience
Unfortunately, most experimental analyses on navigation assets do not differentiate between experienced and inexperienced users.Navigation assets suitable for experienced users may not provide a proper level of support for inexperienced users.Thus, resolutions enhancing inexperienced users' navigation performance may not benefit experienced users [9].
Lin et al. [23] analyzed the effect of repeated exposures to indoor environments on people's indoor wayfinding performance under normal conditions and during a fire emergency, which could cause mental stress.Indoor wayfinding experiments were performed in an immersive virtual museum.Collected data included participants' wayfinding performance, sense of direction, wayfinding anxiety, and simulator sickness.The results indicated a significant positive effect of repeated exposure on participants' wayfinding performance, which decreased the time needed to complete the required task [23].
Burigut and Chittaro [9] ran an experimental study whose purpose was (1) to compare three navigation assets that allow users to perform wayfinding tasks in desktop virtual environments by pointing out the location of objects or places and (2) to evaluate the impacts of user experience with 3D desktop virtual environments on the navigation assets efficacy.They compared the navigation performance (the whole time to conduct a search task) of 48 subjects separated into two groups of experienced and inexperienced VR users.Based on their results, differences are strongly influenced by the virtual environment where navigation takes place, like abstract vs. geographic environments [9].
Nasiri [26] compared novice and experienced VR users' gait differences in IVEs and real environments by analyzing subjects' gait parameters (walking speed, step length, and trunk angle) to assess the effect of prior VR experience on their motions in the IVE.Based on the study mentioned, the trial found a significant variable affecting the gait parameters of novice and expert users.Although the main effect of expertise was not observed, an interaction effect between expertise and the trial number was shown.Bailenson and Yee [1] conducted longitudinal research, tracking users over 45-minute sessions when interacting with each other in VR; evaluating results showed that the feeling of presence did not change over time.
Since the modern 3D computer game is one of the most common virtual environments or 3D interfaces, using computer games as the basis for virtual environment evolution would be helpful [38].The present era of computer games gives the experience of virtual worlds featuring user-friendly interaction and the simulation of the real world.Frey et al. [15] studied the effects of game experience on psychological experimenting in IVEs, considering if training could lower the performance distinctions between users who play games and users who do not.Differences in navigation performance between different levels of experience were reduced by training, implying that inner validity can be obtained.
Richardson et al. [30] studied the relationship between prior video game experience and spatial performance in virtual and real environments.Across two experiments, the gaming experience was associated with performance in desktop virtual environments; those with more video game experience pointed more accurately to non-visible targets [30].The results suggest that the gaming experience is related to the ability to make precise spatial representations while moving in an IVE.
The results of the study by Murias et al. [25] confirmed that individuals who have played video games for a longer time perform better on a virtual navigation task.However, this effect was most pronounced in players of video games that involve navigation, and it cannot be solely attributed to their mastery of game controls.Additionally, participants who frequently play video games involving navigation reported employing more efficient navigation strategies, such as utilizing cognitive maps or relying on learned routes.These findings support the idea that improved navigation and orientation skills in video game players are likely a result of regular practice of these skills for entertainment purposes [25].
Moreover, another research project has explained the impacts of gaming experience on virtual environment evaluations involving navigation tasks [36].Results revealed that perceived gaming skill and progress in a firstperson-shooter (FPS) game were the most compatible metrics demonstrating significant correlations with performance in time-based navigation tasks.Based on the study by Martelli et al. [24], participants adapt to the virtual environment and the perturbations over time.Results indicate increasing stride length and reducing stride width over time.However, there is yet a significant lack of research on the expected time users need to become proficient at traveling in the VR world.

Participants
Twenty people participated in this study (five females and one non-binary).Ages ranged from 18 to 33 years, with a median age of 20.All participants had a normal or corrected-to-normal vision.Participants were recruited via an e-mail sent to undergraduate and graduate students enrolled at Clemson University.
The participants were recruited based on the following prerequisites: (1) having less than 1 hour of prior experience using VR, (2) volunteering to commit to using the HMD regularly for 4 weeks, (3) being ready to complete detailed activities as part of the experiment every week, (4) approving to not let anyone else use the device, and (5) confirmation of having an open floor space at home that could be used for the Quest.Participants received a $50 Visa gift card when enrolling in the study and a $75 Amazon gift card after finishing the study.

Procedure
Each participant was loaned an Oculus Quest for the experiment duration (4 weeks).Upon picking up the HMD, participants were provided with straightforward instructions about (1) operating the headset, (2) using the Side-Quest application1 to load the custom activities onto the Quest, and (3) accessing data saved to the headset about their activities.Each week, participants were requested to upload log files about their activity to a specific Google Drive folder created for each participant.This allowed us to keep track of their progress and remind participants who fell behind.
Participants were asked to complete three activities each week and any personal use they were interested in (games, entertainment, etc.).Participants were instructed to complete no more than one of our activities on any given day so as to avoid any immediate effects of one activity on another activity.Participants accessed our custom activities via a single application.Participants completed three custom activities in counterbalance order: (1) an activity assessing sensitivity to rotational gains, (2) an activity assessing sensitivity to proprioceptive offsets, and (3) a navigation activity.Activities 1 and 2 were completed while stationary; no locomotion or navigation was required from participants.A brief description of each of these activities is provided in the following.This article only reports results of the navigation activity that is focused on locomotion and navigation in VR over time.

Rotational Gain Activity.
In this activity, participants were asked to stand in a garage environment and look at targets to the right and left in multiple trials.One of these rotations had a gain applied in each trial.Participants were then asked to identify which rotation had a gain applied to it.Considering that rotational gains can sometimes lead to cybersickness, the average scores reported by participants were low [31].
Proprioceptive Offsets Activity.In this activity, participants were asked sit down in front of a virtual table and to stack blocks at a target location.The Oculus Quest's hand tracking was enabled for this activity, and participants could pick up a block using a pinching gesture.An offset to the hand's position was applied during some stacking tasks, and participants were asked to identify when an offset was applied [21].
Navigation Activity.In this activity, participants were supposed to complete 10 trials each time they performed the activity (each session).During each trial, participants were asked to navigate through a maze-like environment (see Section 3.3 for details).As part of the trial, they had to find a key to open a locked door leading to the exit.Upon reaching the exit, participants were asked to point in the direction of where they found the key and where they had begun the trial.Participants were randomly assigned to one of two between-subject conditions in the navigation activity: the teleportation or the joystick-based locomotion condition.They were asked to complete this activity four times, once each week.They did not receive feedback on their performance during experiment implementation to minimize learning impacts.Participants could quit the session at any time due to the risk of cybersickness associated with this activity.In addition, participants completed the Simulator Sickness Questionnaire (SSQ) [20] and the iGroup Presence Questionnaire (IPQ) [34] for each session.

Apparatus
Twenty different floorplans were automatically generated using the Dungeon Architect plugin [39] for Unity.All floorplans were designed with the following constraints: (1) the main path from start to finish was eight rooms long, (2) four side paths containing three rooms led to a dead end, (3) a keycard was placed at the final room of one of these side paths, and (4) a locked door that could be opened with the keycard was placed on the main path after the four side paths.All rooms were directly connected to other rooms; no corridors were present in the floorplans.These rules ensured that the different floorplans contained the same number of rooms, the same connectivity, and the shortest path of the same length (crossing through 10 unique doorways, 3 of which were traversed twice).An example floorplan can be seen in Figure 1, and an example of a typical room can be seen in Figure 2. Locomotion was implemented using VRTK 4 2 and VRTK Prefabs v1.1.8[13].Joystick locomotion was calibrated to move participants at a constant speed of 2.25 m per second, equivalent to a fast walk.Pushing forward on the joystick would translate participants in the direction they were facing; participants could move side-toside or backward by pushing the joystick in the appropriate direction.A dead zone of 10% was used to prevent very slight adjustments of the joystick from moving participants.Teleportation was implemented where participants pressed down on the trigger to activate a parabolic raycast that could be used to select where they wanted to move to.Once they had indicated where they wanted to move using the raycast, participants released the trigger to teleport to that location.Teleportation occurred instantaneously without any fading of the scene view.The maximum distance participants could teleport at once was 10 m.

RESULTS
Linear mixed models analyzed the experiment results and tested the hypotheses.Models were created in R [28] using the 'buildmer' [41] and 'lme4' [2] R packages.Buildmer automatically tests different possible models based on a set of independent variables and uses the model's likelihood-ratio test and the minimum Bayesian information criterion to select the model that best matches the observed data [35].Fixed effects that are not included in the final model can be assumed to have had little effect on the modeled variable.Once the final model was specified, it was fitted to the data using the lmer command provided by lme4.The lmerTest package was used to estimate p-values using the Satterthwaite degrees of freedom method [33] for the models generated by lmer.Figures were generated using ggplot2 [42].Unless stated otherwise, condition, session, and trial number were input into Buildmer as potential fixed effects when modeling a given independent variable.
Sixteen of our 20 participants completed all four sessions.Of these participants, 6 were in the joystick condition and 10 were in the teleportation condition.Although we asked participants to complete 10 trials per session, they completed an average of 5.9 trials per session in joystick mode and 8.2 trials per session in teleportation mode.The higher drop rate for the joystick condition can likely be attributed to the increased cybersickness associated

Locomotion Time
We fitted a linear mixed model to predict locomotion time with Session, Condition, and Session by Condition.The model included the Participant Identification Number (PID) as a random effect.The model's total explanatory power is substantial (conditional R 2 = 0.39), and the part related to the fixed effects alone (marginal R 2 ) is 0.25.
The coefficients for the fixed effects are reported in Table 1.
The time required to complete a trial decreased by 22.95 seconds for each additional completed session (Figure 3).Additionally, the time taken to complete the trial decreased by 72.51 seconds compared to the joystick when using teleportation.However, the two-way interaction effect between Session and Condition [Teleportation] indicates that the effect of Session was less pronounced for participants who used teleportation such that the effect of Session decreased by 12.10 seconds when using teleportation.

Spatial Memory
After reaching the end of a trial, participants were asked to point in the direction of where they started the level and where they found the key card.The angular error between their direction and the true direction was then calculated (Figure 4).This analysis used an absolute angular error in the horizontal XZ plane, as we do not anticipate any right/left directional effects since the entire level was on a single XZ plane.
We fitted a linear mixed model to predict absolute angular error with Session, Condition, Trial number, and all interaction effects as fixed effects.The model included PID as a random effect.The model's total explanatory power is moderate (conditional R 2 = 0.15), and the part related to the fixed effects alone (marginal R 2 ) is 0.03.The coefficients for the fixed effects are reported in Table 2.
The baseline error predicted by the model was 33.79°.This error decreased by 3.85°for each additional Session completed and decreased by 3.85°for each additional trial completed.The effect of Trial on error was moderated by Session such that the effect of Trial diminished by 1.15°for each additional session completed, indicating that the effect of Trial on error diminished as participants completed more sessions.Similarly, the effect of Trial on Fig. 3.The time taken to complete the navigation trials decreased across sessions in both conditions; however, it decreased at a faster rate in the joystick condition.
error was moderated by Condition such that the effect of Trial diminished by 3.47°when in the teleportation condition, indicating that Trial had little effect on error when participants moved via teleportation.However, the three-way interaction indicates that the effect of Trial on error became meaningful as Session increased.The effect of Trial in the teleportation condition decreased error by an additional 1.28°for each additional session completed.

Simulator Sickness
Before starting and after finishing the experiment, participants were asked to answer the SSQ.We report separate analyses for each of the four factors of the SSQ in this section (Figure 5).The questionnaire asks participants to score 16 symptoms on a 4-point scale (0-3).A factor analysis revealed that these symptoms could be placed into three general categories: Oculomotor, Disorientation, and Nausea [20].The Total score represents the overall severity of motion sickness experienced by the users of VR systems.

Nausea.
We fitted a linear mixed model to predict Nausea with Session and Condition.The model included PID as a random effect.The model's total explanatory power is weak (conditional R 2 = 0.07), and the part related to the fixed effects alone (marginal R 2 ) is 2.76e-03.The coefficients for the fixed effects are reported in Table 3.Neither Session nor Condition had a significant effect on Nausea scores.

Disorientation.
We fitted a linear mixed model to predict Disorientation with Session, Condition, and Session by Condition.The model included PID as a random effect.The model's total explanatory power is mod-   4. Disorientation increased by 1.53 for each additional session that was completed.However, the two-way interaction between Session and Condition [Teleportation] indicates that this increase in disorientation was primarily seen in the joystick condition, as the effect of Session in the teleportation condition decreased by 1.98.

Oculomotor.
We fitted a linear mixed model to predict Oculomotor with Session and Condition.The model included PID as a random effect.The model's total explanatory power is substantial (conditional R 2 = 0.43), and the part related to the fixed effects alone (marginal R 2 ) is 0.17.The coefficients for the fixed effects are reported in Table 5.Although Session did not significantly affect Oculomotor discomfort, it did decrease by 10.09 points when using teleportation compared to joystick locomotion.

Total Sickness
Score.We fitted a linear mixed model to predict the Total with Condition.The model included PID as a random effect.The model's total explanatory power is moderate (conditional R 2 = 0.25), and the part related to the fixed effects alone (marginal R 2 ) is 0.10.The coefficients for the fixed effects are reported in Table 6.Although Session did not significantly affect Total Sickness, it did decrease by 5.46 points when using teleportation compared to joystick locomotion.

Presence by Session
Presence as the subjective psychological response in a VR system varies for different users.Participants were asked to answer questions regarding their sense of presence.We used the IPQ [34] as a scale for measuring the sense of presence experienced in a virtual environment (Figure 6).The IPQ is developed in three main categories, which make up 16 questions with slightly different themes altogether.The 16-item scale evaluated presence in terms of spatial presence, involvement, and judgment of realness.

Spatial.
We fitted a linear mixed model to predict Spatial with Session.The model included PID as a random effect.The model's total explanatory power is substantial (conditional R 2 = 0.58), and the part related to the fixed effects alone (marginal R 2 ) is 0.03.The coefficients for the fixed effects are reported in Table 7. Spatial presence decreased by 0.17 for each additional session completed.We fitted a linear mixed model to predict Involvement with Session.The model included PID as a random effect.The model's total explanatory power is substantial (conditional R 2 = 0.33), and the part related to the fixed effects alone (marginal R 2 ) is 0.02.The coefficients for the fixed effects are reported in Table 8.The session did not have a significant effect on Involvement.

Realism.
We fitted a linear mixed model to predict Realism with Session.The model included PID as a random effect.The model's total explanatory power is substantial (conditional R 2 = 0.41), and the part related to the fixed effects alone (marginal R 2 ) is 3.77e-04.The coefficients for the fixed effects are reported in Table 9.The session did not have a significant effect on Realism.

Total Presence
Score.We fitted a linear mixed model to predict the Total score with Session.The model included PID as a random effect.The model's total explanatory power is substantial (conditional R 2 = 0.53), and the part related to the fixed effects alone (marginal R 2 ) is 0.01.The coefficients for the fixed effects are reported in Table 10.The session did not have a significant effect on Total Presence.

DISCUSSION
Regarding RQ1, Session was observed to have several effects on participant behavior: participants (1) completed trials an average of 22.63 seconds more quickly for each session they completed (p < 0.001), (2) improved their performance on the spatial memory task by an average of 3.85°for each session completed (p = 0.038), (3) reported experiencing slightly more disorientation with each session completed (p = 0.043), and (4) reported experiencing slightly less spatial presence with each session completed (p = 0.029).Although simple learning effects likely played a large role in the improvement in trial completion time between sessions, participants' improvement on the spatial memory task is not as easily attributable to a learning effect because participants were not provided with any feedback about their performance on this task.These results show how at least some aspects of navigation and locomotion can change over time as users become more proficient with a locomotion technique.
Regarding RQ2, we observed differences between both locomotion methods, as expected from prior work: participants (1) completed the navigation trials an average of 72 seconds faster when using teleportation and (2) were more accurate on the spatial recall task when using joystick locomotion.However, we also observed interaction effects between the locomotion method and session, indicating that the effect of experience impacted performance on these tasks differently depending on the locomotion method used.Participants improved their completion time more rapidly in the joystick condition than in the teleportation condition.Participants' performance on the spatial memory task improved more rapidly in the joystick condition.Although teleportation initially allowed participants to complete the navigation tasks substantially faster, this advantage diminished substantially by the study's conclusion.This highlights how some differences in locomotion methods may diminish over time as users become more familiar with their use.However, a contrasting effect was seen for the locomotion method's effect on performance in the spatial memory task.The gap between locomotion methods widened substantially with time as participants in the joystick condition saw marked improvements, whereas participants in the teleportation condition showed little improvement.This shows how differences between locomotion methods can grow more pronounced with time and experience.
A more complex observation can be made regarding sessions' effect on spatial memory as expressed in the three-way interaction effect involving the locomotion method, trial number, and session.The main effect of trial (μ = −3.45,p = 0.009) suggests that participants improved their spatial memory within a given session across the 10 trials.However, this effect was moderated by two-way interaction effects with both the locomotion method (μ = 3.47, p = 0.022) and session (μ = 1.15, p = 0.012): participants in the teleportation condition showed little improvement across trials within a given session, and the effect of the trial diminished as the session increased (likely due to an improvement in baseline performance leaving less room for improvement).Finally, a three-way interaction effect was observed for the session on the locomotion method by trial (μ = −1.28,p = 0.015): as the session increased, participants who used teleportation began to improve their performance across trials within a given session.In sum, participants who moved via joystick locomotion were immediately able to improve their performance on the spatial memory task within a given session; in contrast, participants who moved via teleportation were initially unable to improve their performance on the spatial memory task within a given session but learned to do so as session increased.
The concepts of calibration and attunement may help to explain why trials' effect on accuracy was mediated differently between conditions [17] (accuracy was directly mediated by trial in the joystick condition, but this effect was further mediated by the session in the teleportation condition).Calibration and attunement are concepts about sensory information's role in the perception-action system.Calibration occurs when an organism adapts its behavior in response to salient information acquired through its sensory system.Attunement is the process by which an organism identifies what information is salient to a given activity, which can then be used for calibration.Optic flow is an important source of information regarding self-motion in the real world; its presence when using joystick locomotion and absence when using teleportation is an often cited reason explaining why spatial awareness suffers when using teleportation compared to other continuous forms of locomotion [3].As users are already familiar with the information provided by optic flow, the steady improvement in spatial awareness observed in participants who used joystick locomotion may manifest their calibrating to the varying properties of optic flow in VR compared to the real world.In contrast, as no optic flow is present when teleporting, this source of information was not available to participants in the teleportation condition to using when calibrating their spatial awareness.Instead, we see a three-way interaction effect whereby participants initially failed to improve their performance across trials in a given session but later gradually improved their performance across trials.This may be a sign that participants in this condition were attuning to other sources of information that could be used for calibration in the absence of optic flow.
Regarding RQ3 and RQ4, the present findings shed light on the anticipated effects of locomotion on sickness in the context of VR experiences.Our study revealed that teleportation exhibited a notable association with reduced feelings of overall sickness (μ = −5.46,p = 0.041) and oculomotor discomfort (μ = −10.09,p = 0.018).These results align with previous expectations and provide empirical evidence supporting the potential benefits of teleportation as a preferred locomotion technique to mitigate sickness symptoms.
The effects of the session were less pronounced on sickness and presence.Feelings of disorientation were reported to increase across sessions (μ = 1.53, p = 0.043), but no effects of the session were observed for the other measured dimensions of simulator sickness.An interaction effect between session and locomotion method on disorientation was also observed where the effect of session on disorientation was diminished for participants who moved via teleportation (μ = −1.98,p = 0.048).
The relation between locomotion methods, sessions, and simulator sickness suggests additional research to explain the underlying factors contributing to these relationships.By advancing our understanding of locomotion effects on sickness, these findings contribute to optimizing VR experiences, enhancing user comfort, and facilitating the development of immersive applications across various domains.

Limitations
When interpreting these results, it is important to note that participants in the joystick condition completed fewer activities overall and fewer trials within those activities, most likely due to the higher simulator sickness associated with joystick locomotion.As such, the results regarding sickness should be interpreted cautiously, as they may under-report the typical amounts of sickness associated with joystick locomotion.It should also be noted that sickness and presence scores were only collected once for each activity, meaning that fewer data points existed for our analysis than the data for completion for time and spatial memory, which were collected during each trial.
It should also be noted that as this experiment was conducted in the wild, we did not control for what applications participants used and when they engaged in them.We chose to allow for more naturalistic conditions akin to those real consumers would engage in after acquiring a VR HMD to increase the ecological validity of this experiment.We believed this to be important because real users are likely to encounter multiple different forms of locomotion simultaneously across different applications.

CONCLUSION
The presented results demonstrate that a user's familiarity with a given locomotion technique can influence locomotive behaviors and effects.While not conclusive, it is particularly interesting how the effect of session on completion time was more pronounced in the joystick condition, which performed worse overall than the teleportation condition; similarly, the effect of session on spatial memory, as indicated by the three-way interaction effect, was more pronounced for the teleportation condition, which also performed worse overall than the joystick condition.This may suggest that some of the tradeoffs between locomotion methods may become less meaningful over time as users become more familiar with the technique.Cybersickness was a notable exception to this pattern, as cybersickness generally increased in the joystick condition.More research is needed to understand better how specific behaviors and effects associated with different locomotion techniques are affected when users become more familiar with the technique.

Fig. 1 .
Fig.1.Three examples of generated levels.In these images, the main path from start to end is shown in white, and the side path to the key is colored orange.The other colored paths are the three remaining dead-end paths.

Fig. 2 .
Fig. 2.An example of a typical room in the different levels.

Fig. 4 .
Fig. 4. Angular error decreased markedly across sessions in the joystick condition but only decreased slightly in the teleportation condition.

Fig. 5 .
Fig.5.Although some effects were observed for simulator sickness, reported sickness scores were relatively stable overall.

Fig. 6 .
Fig.6.Spatial presence was observed to decrease slightly across sessions.No significant effects were observed for other presence factors.

Table 1 .
Parameters of the Locomotion Time Model with joystick-based locomotion.In total, 141 trials were completed in the joystick condition and 321 trials were completed in the teleportation condition.

Table 3 .
Nausea Model Parameters

Table 4 .
Disorientation Model Parameters

Table 6 .
Total Sickness Score Model Parameters

Table 7 .
Spatial Model Parameters

Table 8 .
Involvement Model Parameters

Table 9 .
Realism Model Parameters

Table 10 .
Total Presence Score Model Parameters