Virtual reality (VR) is currently seen as the next digital Frontier. In recent years, various commercial VR head-mounted displays (HMD) have become available at a low cost and VR is already widely used for training, rehabilitation, education, work, and entertainment. Unfortunately, VR sickness is still considered a major barrier to mass consumer adoption (Nelson, 2015). VR sickness is a type of motion sickness unique to VR and can include symptoms such as nausea, pallor, sweating, stomach awareness, increased heart rate, drowsiness, disorientation, and general discomfort (Kennedy et al., 1993). Not all individuals who are exposed to VR will experience VR sickness, however, those who do may find it debilitating (Regan, 1995; Stanney et al., 2003).

Several VR sickness mitigation strategies have already been developed, including the use of dynamic or foveated field-of-view (FOV) restriction (Fernandes and Feiner, 2016; Adhanom et al., 2020; Yamamura et al., 2020) and vection blurring (Budhiraja et al., 2017). In practice though, the bulk of locomotor VR experiences that are currently available rely primarily on the use of teleportation (Al Zayer et al., 2020; Monteiro et al., 2021) which has been shown in a recent meta review (Prithul et al., 2021) to result in significantly lower VR sickness incidence but at the cost of lower presence and potential spatial disorientation. The reason for the success of teleportation may be in part due to the lack of optic flow generated during use (Prithul et al., 2021). Optic flow can contribute to both vection (Palmisano et al., 2011; Seno et al., 2018) and sensory conflict (Bubka et al., 2007; Palmisano et al., 2011), which cause sickness in certain contexts (Bubka et al., 2007).

A current challenge with VR sickness is that available mitigation strategies aren’t 100% effective since the effectiveness of any particular mitigation strategy varies across individuals (Reason, 1978). This makes it difficult to create a one-size-fits-all strategy for mitigation. A better strategy may be to supplement existing mitigation techniques by investigating how the user might best prepare themselves for continued VR use.

Research which focuses on using adaptation to the virtual environment has shown promising results for VR sickness mitigation (Reason and Brand, 1975; Cobb et al., 1999; Hill and Howarth, 2000; Bailenson and Yee, 2006; Domeyer et al., 2013; Beadle et al., 2021). The general assumption is that this repeated exposure facilitates adaptation to sensory conflict in the virtual environment, reducing sickness. However, much of this success is dependent on the researchers exposing users to the same virtual environment. Development of a training paradigm which facilitates the transfer of adaptation effects between multiple environments has prove more difficult to achieve (Mouloua et al., 2005; Smither et al., 2008; Duzmanska et al., 2018). These issues may be in part a result of the dissimilarities between the environments (Reason and Brand, 1975; Reason, 1978); adaptation to sensory conflict is thought to be largely dependent on the user becoming accustomed to the conflict which arises in a particular environment through developing the appropriate expectations for sensory input in that environment (Reason and Brand, 1975; Reason, 1978; Keshavarz, 2016). Therefore, facilitating more general adaptation may require training through exposure to factors that are consistent between all VR environments, for example, optic flow stimuli.

Therefore, we aimed to develop a training technique that facilitates the transfer of adaptation effects across environments. Our study borrows concepts from motion sickness mitigation work by emphasizing gradual exposure to sickness-inducing stimuli (Rine et al., 1999; Takeda et al., 2001) such as expanding optic flow (Ebenholtz, 1992; Bubka et al., 2007) and graphic realism (Golding et al., 2012; Davis et al., 2015) to facilitate adaptation. To this end, we use a simplified training environment consisting only of the walls and floor of a labyrinth that users navigate. Over successive days, we gradually adapt users to increasing optic flow by incrementally increasing luminance contrast of the visual stimulus. We show that the adaptation that is achieved using this simplified environment with gradual increases in optic flow strength successfully transfers to a richer, more naturalistic environment. Results of this study suggest that it should be possible to develop standardized training paradigms to reduce and possibly even eliminate VR sickness in susceptible individuals.

There are three main theories aimed at explaining VR sickness: postural instability theory, eye movement theory, and sensory conflict theory (Reason and Brand, 1975; Riccio and Stoffregen, 1991; Ebenholtz, 1992). Postural instability theory (Riccio and Stoffregen, 1991) attributes VR sickness to a disruption of postural stability caused by unnatural or unexpected motion in the virtual environment. Eye movement theory (Ebenholtz, 1992) postulates that optic flow drives rapid reflexive eye movement to stabilize the image (optokinetic nystagmus) and that the associated sensory and motor signals innervate the vagal nerve, leading to VR sickness. Sensory conflict theory (Reason and Brand, 1975) seems to be the most widely accepted (Kolasinski, 1995; Keshavarz et al., 2014) and attributes VR sickness to the conflicting signals between the visual, vestibular, and somatosensory senses causing discomfort and nausea among other symptoms. These theories are not mutually exclusive, and this list does not encompass all efforts to explain VR sickness (Keshavarz et al., 2015).

One element these theories have in common is that they all seek to explain how sickness may be caused by exposure to full-field visual motion, or optic flow. Optic flow is thought to cause VR sickness (Davis et al., 2015; Tian et al., 2022) particularly when the visually-simulated self-motion does not match natural self-motion of the user, i.e. when there is sensory conflict (Bubka et al., 2007). Many mitigation strategies, including FOV restriction, aim to reduce optic flow during movement (Fernandes and Feiner, 2016; Adhanom et al., 2020; Shi et al., 2021). Other strategies instead try to reduce conflict by providing some form of vestibular or efferent feedback (Peng et al., 2020), for example, by walking-in-place (Tregillus and Folmer, 2016). It is important to note that optic flow is not the same as vection; optic flow refers to the visual motion stimulus itself while vection refers to the visually-induced perception of self-motion (Fujii and Seno, 2020). Optic flow can cause vection (Palmisano et al., 2015; Seno et al., 2018), so their separate effects can be easily confounded during discussions of VR sickness (Chang et al., 2020). Like optic flow, vection has been linked to feelings of VR sickness (Keshavarz et al., 2015; Palmisano et al., 2017), however the relationship is still not perfectly understood (Keshavarz et al., 2015; Palmisano et al., 2017) nor should it be oversimplified. As vection may be experienced without inducing VR sickness, vection alone is not sufficient to induce sickness (Keshavarz et al., 2015). Instead, it must be paired with other factors such as sensory conflict, postural instability, etc. This makes it difficult to disentangle the distinct effects of the optic flow versus vection when investigating sickness.

Other work related to the current study has investigated adaptation to VR sickness. Prior studies have argued that repeated exposure to an aversive stimulus allows developing an altered expectation of sensory input (Reason and Brand, 1975; Keshavarz, 2016), with the required length of exposure increasing as the stimuli become more complex (Reason, 1978). It has been suggested that this makes it unlikely for adaptation effects in one virtual environment to transfer to another. The degree of transfer may depend on the degree of dissimilarity between two environments (Reason and Brand, 1975; Reason, 1978). Previous research in similar environments has been rather consistent in demonstrating effective adaptation, with reports of symptoms decreasing over the course of exposure (Hill and Howarth, 2000; Bailenson and Yee, 2006; Howarth and Hodder, 2008; Domeyer et al., 2013; Beadle et al., 2021), even on timescales as low as 45 min (Sinitski et al., 2018). In one study by Lampton et al. (2000), the researchers saw VR sickness decrease following repeated exposure to a training environment, however, these effects disappeared when the participants entered the mission environment (Lampton et al., 2000). Another study by Hill and Howarth (2000) had participants view a game being played through an HMD on five consecutive days. Results also demonstrated significantly less nausea on the fifth day, compared to the first. More recently, a study by Beadle et al. (2021) had participants perform a shooting task in an HMD over 3 sessions. SSQ scores significantly decreased following each session. This is not an exhaustive list of VR adaptation studies (for review, we recommend Duzmanska et al. (2018)), but they do demonstrate a common theme; adaptation to a single VR environment is possible for mitigating VR sickness. In this experiment, we hope to expand upon this research by examining whether adaptation can transfer between VR environments.

Other relevant prior work has examined how various methods of training against sickness may offer a potential solution to VR sickness more generally, rather than to a single environment (Mouloua et al., 2005; Smither et al., 2008). One such study by Smither et al. (2008) showed that when participants were pre-exposed to simulated rotary stimulation they experienced less symptoms of VR sickness than individuals who had no prior exposure (Smither et al., 2008). Similarly, an experiment by Mouloua et al. (2005) demonstrated cross-platform training was possible, but that it was only successful when using optokinetic nystagmus (OKN) training, and not a VR HMD (Mouloua et al., 2005). Another study by Pöhlmann and colleagues (2021) found that individuals that spend extensive amounts of time playing action video games (not on HMDs) experienced less VR sickness when exposed to a simple virtual environment compared to individuals that do not play any video games (Pöhlmann et al., 2021). All of these experiments were examining transfer in a way that used a non-VR device as their training mechanism. It is possible that these experiments were training against motion sickness more broadly, rather than adaptation to VR specifically. While this may be another effective methodology for sickness mitigation, it stands to reason that this would not be readily available for the average user. Here, we seek to facilitate adaptation solely through the use of a VR HMD.

Specifically, we explicitly aim to facilitate adaptation between distinct VR environments. This means that experiencing one environment will lead to a reduction in VR sickness in a dissimilar environment. Additionally, we hypothesize that this adaptation can occur without exposing participants to the full optic flow and luminance contrast normally experienced in the VR environment. In other words, training can be tailored to act as a “ramp up” period for the participants. Studies which center on training against motion sickness have suggested that adaptation should be facilitated through a gradual increase of the provocative stimulus (Rine et al., 1999; Takeda et al., 2001). Similarly, if we want to train participants against VR sickness, best practice may be to increase intensity over time such that adaptation has time to take place while avoiding extreme illness.

Our study aims to answer two research questions:

• RQ 1. Does ramped exposure to increased optic flow strength in an abstract environment allow gradual adaptation?

To align our results with our research questions, we report the results in two parts: the first subsection reports results for the test sessions (days 1 and 5) and the second subsection reports the results for the training sessions (days 2, 3, and 4). The results for the test sessions aim to address RQ2 - whether the adaptation training was transferred from the abstract training environment to the complex testing environment.

For all statistical tests, we used Shapiro–Wilk tests to test the normality of our data, and we used the box plot method to identify outliers outside the Q1 - 1.5 × IQR to Q3 + 1.5 × IQR range. Q1 and Q3 are the first and third quartile, respectively. IQR is the interquartile range (IQR = Q3 - Q1). There were outliers in some of our results due to participants showing high levels of discomfort, especially during the early days. We retained these outliers because they are a natural part of the data, rather than errors. A significance level of 0.05 was set for all analysis. All our statistical tests were performed in R version 4.1.0 using the rstatix package, and results were duplicated SPSS.

This study had two aims, to investigate VR sickness adaptation methods that allow gradual exposure to increasing optic flow and the transferability of adaptation from one environment to another. To avoid floor effects, we explicitly recruited individuals that reported susceptibility to VR sickness and limited VR use. Subjects were tested over 5 days, consisting of a baseline on day 1, three consecutive training days, and a final test session on day 5. The test environment consisted of a cityscape, while the training environment was a less realistic labyrinth with black and white striped walls. Contrast and thus optic flow strength was gradually increased over consecutive training days. Results indicate that, despite the increase in optic flow strength, subjective sickness reports and objective physiological measures did not increase, but instead tended to remain constant or even decrease over consecutive training days, though this decrease was generally not significant. Comparison between measures obtained on days 1 and 5, on the other hand, tended to show significant decreases in sickness. These results suggest that exposure to ramped optic flow strength supports gradual adaptation of VR sickness and its symptoms. Additionally, adaptation that is developed in more abstract and impoverished environments can transfer to richer and more naturalistic environments. These results suggest that carefully controlled adaptation procedures can be a successful mitigation strategy for VR sickness.

The conclusions of the study would be strengthened by collection of additional data to control for both ramping optic flow as well as the transfer of adaptation between abstract and rich natural environments. For example, the drop in sickness and EDA observed between days 1 and 5 could be compared against the drop in sickness observed in a comparison group exposed to the test stimulus on consecutive days without being exposed to the training environment. Similarly, the levels of sickness observed on training days 2 through 4 could be compared against sickness observed in a comparison group for which contrast and optic flow were held constant over the 3 days. These avenues were not explored for two reasons; first, these measurements were not possible because they would have required tripling the number of subjects enrolled in our study. Due to our restrictive inclusion criteria (see methods) and COVID-19 concerns it was already a challenge to recruit the current pool of subjects. Second, these methodologies were already largely exploratory, with previous literature already investigating much of the aforementioned control results. As mentioned in the related works section, it is well-known that consistent exposure to the same VR environment results in adaptation (Reason and Brand, 1975; Cobb et al., 1999; Hill and Howarth, 2000; Bailenson and Yee, 2006; Domeyer et al., 2013; Beadle et al., 2021), but these effects have also been known to dissipate in as little the same day (Lampton et al., 2000). This makes it unlikely that the results from day 5 are solely due to exposure to the city environment on the first day. Similarly, investigations into adaptation transfer have been largely unsuccessful or conflicting, so it is noteworthy that scores between days 4 and 5 were so similar. We suggest that this methodology was successful because it borrows concepts from motion sickness training such as gradually increasing exposure (Rine et al., 1999; Takeda et al., 2001) to a root cause of VR sickness (i.e., optic flow or the associated vection).

From our results, we concluded that subjects experienced lower levels of sickness over time, and these effects persisted between environments. The mechanisms behind this change, whether they be through the actions of the subjects or via some neurological mechanism, are unfortunately beyond the scope of this study. However, future work exploring changes in behavior in subjects in VR over time would be a worthy follow-up to this study.

Subjective feelings of vection were not directly measured during this experiment. However, previous research has shown increasing luminance contrast increases feelings of vection (Sauvan and Bonnet, 1993; Patterson and York, 2009; Guo et al., 2021), and so it is logical to assume the same effect was experienced in our experiment. Independent of this matter, further research has shown that increases in optic flow lead to greater feelings of VR sickness (Davis et al., 2015). Therefore, regardless of vection, we can assume that the increase in luminance contrast and optic flow did lead to greater stimulus intensity as intended.

Regarding physiological data, we used the Empatica E4 watch, which is limited to a sampling rate of 4 Hz for EDA data. It is important to note that EDA signals are particularly sensitive to motion artifacts. Because our subjects were using their hands in order to manipulate the controller for locomotion, these artifacts were likely present in our data.

In this study, we tested a novel method for managing VR sickness through training. We used an abstract environment and reduced stimulus contrast to limit sickness during training. We showed that sickness remained constant or decreased on consecutive training days, even though contrast and optic flow strength was increasing. We also argue that adaptation developed in an abstract training environment transferred to a richer test environment. These results demonstrate that adaptation to VR sickness via careful and gradual exposure to successively more provocative VR stimuli shows promise as a VR sickness mitigation strategy.