Are Embodied Avatars Harmful to our Self-Experience? The Impact of Virtual Embodiment on Body Awareness

Our body constantly gathers, processes, and filters information about our environment. It is sensitive to the outside temperature, the intensity of touch, or the noise of our surroundings. In addition to external signals from our environment, signals from inside the body maintain our self-awareness [63]. The processing of these internal body signals, especially the interoceptive and proprioceptive signals, is called body awareness. It is defined as the “subjective, phenomenological aspect of proprioception and interoception that enters conscious awareness and is modifiable by mental processes such as attention, interpretation, evaluation, beliefs, memories, conditioning, attitudes, and affect” [45, p. 4]. Therefore, body awareness is a central part of perceiving the body’s sensations and includes the perception of various internal body signals, such as hunger and heart activity or other more complex perceptive syndromes. It is often captured via self-reports or operationalized as interoceptive accuracy (IAC) and assessed via heartbeat-counting tasks.

Body awareness is closely linked to mental health and subjective well-being [28] and is negatively related to symptoms of depression [49], eating disorders [35], or migraine [55]. On the other hand, body awareness dysfunctions are associated with increased suicidal thoughts and actions [30]. Following these findings, Gibson [25] proposed in a recent discussion that a strengthened IAC or body awareness accounts for the benefits of mindfulness practice in different research. The processing of the body’s internal signals has become the focus of several therapeutic approaches, so-called mind-body therapies, aiming to integrate mind and body awareness into daily life via breathing, meditation, or movement exercises [46]. Although the practical application of body awareness in therapy varies widely, in a qualitative study on the definition of body awareness in therapy, Mehling et al. [46] found a great deal of commonality in understanding body awareness among practitioners. Therapists have defined body awareness in two ways, as a core element of integrity and an essential part of self-awareness and as an individual’s capacity and ability for embodiment.

Therapeutic approaches aiming to increase or adapt body awareness mostly rely on modifying body awareness via attention regulation. Directing attention to external body signals can facilitate the processing of interoception [3, 42]. Especially in the field of mindfulness, some developments and design ideas have been proposed to integrate VR into mind-body therapy approaches. In this context, VR allows arbitrarily adapting the visible environment or augmenting feedback to body movements or physiological measures using virtual stimuli. While research in this area has predominantly relied on mindfulness, the influence of virtual bodies on body awareness could provide new insights into the mechanisms of body awareness and embodiment and how virtual stimuli could help maintain or manipulate body awareness in a virtual therapy scenario [4, 14].

VR experiences rely on supplementing, modifying, or replacing a particular part of body signals with virtual stimuli. Typically, this is done by displaying visual stimuli while excluding visual information from the real environment. Adapting visual movements to the user’s actions establishes a state of congruence between the digital (visual) and non-digital (proprioceptive, vestibular, and kinesthetic) stimuli [39]. Upon meeting this state, the virtual experience is perceived as plausible and thus elicits a sense of presence. When embodying a virtual body, the congruence of a virtual body’s behavior and look can lead to plausibility [43] and a perceptual shift towards the virtual body. Kilteni et al. [36, p. 375] define this state as the Sense of Embodiment (SoE), “the sense that emerges when [the body’s] properties are processed as if they were the properties of one’s own biological body”. In the context of our work, the question arises whether one’s body perception is influenced when the visual body signals do not come from the own body. Through bottom-up processing of congruent visuotactile or visuomotor stimulation, the perception of a virtual body is integrated into one’s physical body perception causing the virtual body to be perceived as a part, extension, or substitute of the physical body. A typical method to enhance the SoE towards a virtual body is the mirror metaphor [32]. By adding a mirror to the virtual environment and consciously juxtaposing the user with their virtual mirror image, the effect of visuomotor or visuotactile congruence is intended to be reinforced [57].

In VR, external and internal body signals may be overridden or suppressed by the external signals presented through the embodiment of virtual bodies. For example, in a study on temperature sensitivity in the palm, Llobera et al. [41] showed that external temperature stimuli are processed less dominant during the embodiment of a virtual body. In their study, half of the participants were presented with a visuomotor congruent virtual body whose movements and posture corresponded to their own. In contrast, the other half of the participants were presented with an incongruent representation. It turned out that participants in the congruent condition were less sensitive to temperature differences. The authors stated a distraction by the visual stimuli could not explain this effect but an integration of the congruent virtual body into the own body perception. Concerning the processing of internal body stimuli, Kasahara et al. [33] showed in a study on visuomotor congruence that delays in the body movement of a virtual body produced a feeling of heaviness in one’s physical body. In contrast, faster virtual body movements produced a feeling of physical lightness. In addition to the visuomotor congruence between the physical and the virtual body, it has been investigated to what extent an inconsistency between dimensions of body parts impacts body perception, for example, proprioception and the perception of one’s body position and dimensions. Van der Veer et al. [68] demonstrated that the positioning of virtual body parts relative to the physical body might lead to a proprioceptive shift when estimating the position of physical body parts. Kilteni et al. [37] showed that the length of virtual arms influences the perception of one’s own reach and body space. With remark to an embodiment scenario with virtual bodies, various works demonstrated that embodying virtual bodies of different sizes impacts body weight perception and the estimation of one’s body size [48, 50, 76].

However, it remains unclear whether these influences on the different aspects of body perception are equivalent to an impact on body awareness. When considering the goal of mind-body interventions, strengthening the connection between body and mind, the question arises of whether VR can be a suitable tool for mental health interventions. Suppose body perception is affected by those external stimuli. Does the embodiment of virtual bodies and the associated distraction from the real body towards a virtual body have a disruptive effect on body awareness?

The processing of exteroceptive signals from the RHI or embodiment of virtual bodies might partially compete with the processing of internal body signals and thus limit body awareness [54, 63]. The presented research highlights the importance of visuotactile or visuomotor congruence in the embodiment of artificial bodies or body parts to maintain or even strengthen body awareness. However, especially when embodying artificial faces, visuotactile congruence does not rule out a negative influence on body awareness [19]. During enfacement illusions, congruence might even have an adverse effect. In summary, prior work suggests a relationship between IAC and SoE and between self-reported body awareness and SoE. However, research is still pending on how VR affects body awareness and IAC compared to reality. It further needs to be investigated to what extent the presented perspective on a personalized virtual body affects the perceived body awareness and IAC.

To address these research gaps, we present a study investigating the effects of having a mirror image in a body awareness movement task in VR. Additionally, we investigate to what extent the embodiment of a highly personalized, photorealistic virtual body affects body awareness and IAC. In a 2 × 2 design, we evaluated the effects of virtuality and perspective on body awareness. Our participants performed movement exercises from Basic Body Awareness Therapy [27] either in a laboratory of the University of Würzburg or in a virtual model of that laboratory in VR in counterbalanced order (virtuality). When in VR, they embodied a virtual replica of themselves. Half of our participants performed the exercises in front of a mirror, and the second half performed them without a mirror (perspective). As dependent variables, we recorded their self-reported body awareness and SoE, measured in experience directly following the performed exercises. Additionally, we assessed their self-reported body awareness, SoE, and IAC measured after leaving the virtual or real laboratory environment. The results of our study intend to provide new insights into the effects of VR on body awareness and, thus, new insights into the relationship between one’s virtual and physical body. Based on the work presented above, we hypothesize the following:

We conducted our study according to the Declaration of Helsinki and received approval from the ethics committee of the Institute Human-Computer-Media (MCM) of the University of Würzburg ¹. Given the prolonged exposure to the mirror image, we referred participants during acquisition and after the study to the freely available support services from the Anorexia Nervosa and Associated Disorders organization (ANAD) ², which they could contact in case they felt uncomfortable about their body shape. Participants were informed in advance about the risks of VR regarding simulation sickness and epilepsy symptoms according to the local VR-usage guidelines. Before entering VR, participants were instructed to report any discomfort they felt during the VR experience immediately. In addition, we set up an area where participants could sit down in silence, hydrate, or lie down if needed.

A total of 45 students and employees of the University of Würzburg participated in our study and received either course credit or 30 EUR in return. Ahead of the evaluation, we defined four exclusion criteria queried by self-disclosure. Participants were not eligible when they (1) had visual impairments not compensated by contact lenses, (2) currently suffered from a diagnosed eating or body image disorder, (3) had less than three years of experience with the German language, or (4) reported simulation sickness symptoms during the experiment. We excluded one participant due to their visual impairment and four participants due to technical issues during the VR session (n = 3) or heart rate tracking (n = 1). Thus, we included 40 participants (25 female, 15 male) in our analysis. The participants were between 19 and 53 years (M = 22.00, SD = 1.48). Twenty-nine participants had spent less than 5 hours, seven participants had spent 5-10 hours, and 4 participants had spent 10-20 hours in VR. Six participants had never used a VR system before their participation.

Our study was designed in a 2 × 2 mixed design with the two independent variables virtuality and perspective. The first independent variable, virtuality, included two experimental conditions performed by each participant: reality and VR. In reality, the tasks were performed in the local laboratory, while in VR, they were performed in a virtual replica of the local laboratory. The order of the two conditions was counterbalanced. The second independent variable, perspective, varied between participants. Participants performed the tasks described in Section 3.6.1 either in front of a (virtual) mirror or without a mirror. Thus, participants only received additional external cues about their bodies in the mirror condition. As dependent variables, we assessed the participants’ self-reported body awareness and their IAC. As a possible mediator between the independent and dependent variables, we assessed their SoE towards their visible body. As control variables, we captured the participants’ body awareness, body consciousness, and IAC prior to our experimental tasks and the two VR-related measures of simulator sickness and avatar uncanniness.

The whole procedure of the study is illustrated in Figure 2. It was split into four phases: Preparation, reality condition, VR condition, and Closure. Both experimental conditions were presented in counterbalanced order and were executed with or without a mirror.

Based on our main effects, we calculated a mediation analyses on virtuality as the independent variable, the three dimensions of SoE as mediator, and the body awareness ratings noticing internal, attention regulation, and visual attention as dependent variable.

As expected (H1.1), we found a significant effect of virtuality on the SoE. Both in-experience and post-experience, participants reported feeling more body ownership and agency towards their physical body in the real environment than towards their virtual body in VR. Additionally, they stated that they experienced more change in their bodily experience in VR than in reality. The ratings in the reality condition were generally very high for body ownership and agency. When assuming that all perception and cognition are body-based [72], the feeling of owning and controlling our physical body should be at a maximum at all times. However, some participants still rated their body ownership and agency in the reality condition lower than the maximum score and stated a feeling of change, although they did not have to split their body ownership between two competing bodies in this condition. There can be various reasons for this. There are certain states in which people do not feel embodied in their physical body or able to control it, such as depersonalization or derealization. The mere question of body ownership or agency may elicit a questioning of one’s bodily state. Similar to the sense of presence in VR compared to reality [66], it seems to be possible that people generally do not report full SoE in reality. To what extent this should impact the interpretation of SoE ratings in virtual reality remains open for future work.

Contrary to our expectations (H1.2), our participants did not report different SoE when confronted with their mirror image than without it. Similar to our results, two recent studies investigated the effects of mirrors on the SoE. Wolf et al. [73] stepwise increased the distance to the third-person perspective provided by a virtual mirror from two to eight meters and could not find any sign of a declining SoE. Bartl et al. [6] investigated the effects of virtual bodies in VR-based physical exercises and did not find an effect of placing a virtual mirror in front of the participants. Past research shows that confrontation with (virtual) mirrors – while being used and proposed as a tool to reinforce SoE [11, 38, 57, 75] – has not yet been investigated extensively until recently. Studies on the impact of a third-person perspective on SoE often rely on a perspective where the participants see their virtual body only from behind, compared to an egocentric first-person perspective [11, 12, 26]. Apparently, both via first-person or third-person perspective, a certain amount of SoE can be achieved via visuomotor or visuotactile congruence. However, depending on the tracking accuracy, the first-person perspective, or alternating between both, lead to higher body ownership and agency than the third-person perspective [11].

Regarding our study, there are three possible explanations for the lack of perspective effects. First, our sample size was relatively small, and minor effects such as those visible in the descriptive data could have been detected with a larger sample. Second, our participants had potentially high expectations concerning the appearance and movements of their virtual bodies. To our knowledge, no research exists on the expectations of VR users toward their personalized virtual bodies. Thus, while Waltemate et al. [70] found a positive effect of personalization on SoE, minor deviations in facial features could have impacted the SoE in our study, especially as we did not contrast the personalized virtual bodies with generic virtual bodies. Future work should investigate whether the embodiment of generic or less realistic virtual bodies leads to similar results concerning the existence of a virtual mirror. Third, despite considerable technological progress, the embodiment of virtual bodies still does not work flawlessly. Contrary to the beneficial effect of mirrors in a virtual embodiment lab proposed by Spanlang et al. [57], in a study on the effect of mirrors on SoE, Rey et al. [51] found higher ratings in SoE in conditions without a mirror than in conditions with a mirror. They explained this effect based on the properties of the mirror they used. Inoue and Kitazaki [32] propose that SoE decreases during exposure to a virtual mirror image when the virtual body does not move synchronously. In our study, we used a low-threshold embodiment system with six-point tracking where the pose between points was calculated approximately. Thus, minor deviations in the posture of arms and legs and missing facial animations could have gradually reduced SoE over time. Consequently, more accurate tracking could be necessary to hold up SoE for such tasks. For future work, we recommend showing a mirror image only for a short introduction to the virtual body, if at all, to avoid possible disturbances caused by minor tracking deviations.

Using a realistic scenario and photorealistically personalized virtual bodies, we found some effects of virtuality on body awareness (H2.1) that did not last over the experience. During the experience, our participants found it significantly more challenging to focus on their bodies, reported noticing fewer signals from within their bodies, and relied more on what they saw than what they felt in VR than in reality. Filippetti and Tsakiris [19] reported a positive effect of the RHI on body awareness, operationalized as IAC. We could not extend this result to virtual bodies in our study, as we did not find an effect of virtuality on IAC. Our effects on self-reported body awareness indicate a negative impact of virtuality.

Since we did not work with generic body parts in our setup but with personalized virtual bodies, our results are more comparable to the second experiment of Filippetti and Tsakiris [19]. They showed that prolonged confrontation with images of one’s face in an enfacement illusion could harm IAC. While IAC and self-reported body awareness are discussed as independent concepts [18], our results on self-reported body awareness indicate a similar effect of the confrontation with photorealistically personalized virtual bodies on body awareness. Still, we did not find an effect of virtuality on IAC. In our study, the use of a mirror without additional haptic stimulation or the inclusion of facial animations had close to no effect. This result contradicts the hypothesis that the confrontation with one’s face would be a causal factor in differences in body awareness (H2.2). While participants reported that they paid more attention to their visuals than to their other bodily sensations, they did not report reduced body awareness in the other measures. Future work could investigate how the personalization of virtual bodies contributes to the found effects. In previous work, personalization has affected SoE positively [70] and IAC negatively [19]. However, the extent to which it affects body awareness when embodying a virtual body has not yet been investigated. In addition, future work should address to what extent not only latency but posture accuracy and tracking performance [23] affect body awareness. Previous studies mainly focused on the effects of visuotactile congruence, while no transfer to virtual bodies and visuomotor congruence has been performed yet. It may be concluded that virtuality, at least for our realistic scenario, had neither a lasting supportive nor a disruptive effect on body awareness. Further, providing a mirror to supposedly strengthen the SoE did not affect body awareness negatively. To ensure that the focus during a virtual mind-body exercise remains on the body’s sensations, and as the positive effect of prolonged mirror exposure on SoE is questionable, we would still argue against using a mirror during the whole length of virtual exercises.

Future research will bring further insights into how virtual body design can support users in maintaining body awareness. Although we found only a partial impact of VR on body awareness, caution should be exercised when using virtual bodies in VR-based mind-body exercises. When creating such scenarios, designers should consider how the VR environment, the performed task, and the virtual body itself affect body awareness. For example, if a mirror is task-immanent, designers should identify solutions to draw attention back to internal body signals. When an avatar is used to guide the user, its appearance and behavior should aim to draw attention to the body while avoiding visual distractions. Depending on the intended outcome, designers should carefully consider to what extent a distraction from internal body signals is likely to happen, problematic, or even desirable.

Based on the work of Filippetti and Tsakiris [19] and Döllinger et al. [15], we expected that a manipulation of the SoE would mediate the perceived body awareness in our tasks (H3). Our results partly confirmed this assumption as we found significant mediating effects of SoE on each of the variables that were affected by virtuality. We found a significant partial mediating effect of perceived agency on attention regulation and a significant partial mediating effect of perceived change on noticing internal and visual attention. While a higher agency was associated with a higher attention regulation, higher change ratings were associated with less noticing internal and more attention to visual signals. However, we did not find mediating effects for the effect of perspective on body awareness. The result is consistent with Döllinger et al. [15], who found a positive correlation between body ownership and agency and body awareness (assessed via SMS). However, it extends the findings as we could show that not only was SoE related to body awareness ratings but also explained part of the effects of virtuality on body awareness. The negative correlation between change and noticing internal is particularly interesting. When embodying virtual bodies, we are confronted with potentially contradicting signals about our bodies. If these lead us to perceive our body as changed, the visual signals seem to have more influence than the internal signals. This result thus fits well with the assumptions of research on individual differences in SoE towards a rubber hand or virtual body [13, 54, 63, 64]. It supports the hypothesis that external and internal stimuli compete in such scenarios. While prior work focused on individual capacity to process external signals, we showed that, at least in the short term, increased processing of external stimuli appears to be associated with reduced processing of internal stimuli.

In addition to the limitations already mentioned above, such as the sample size or possible tracking imprecision, we would like to mention a few limitations of our study design. Our results are limited to virtual experiences where the virtual environment and the virtual body of the participants strongly resemble reality. In developing the virtual environment, we replicated the local laboratory as closely as possible and created personalized photorealistic virtual replicas of the participants. This level of realism and personalization is not feasible in most cases. Work on virtually supported mind-body interventions presents very heterogeneous virtual environments and virtual bodies that are adapted to the goal of the task rather than to the user or do not include virtual bodies at all [14]. To generalize our results, it is necessary to replicate them in diverse virtual spaces, with less personalized or generic virtual bodies, or even without an anthropomorphic self-representation. We can only conclude that even in a scenario like ours, a negative influence of the embodiment of virtual bodies on body awareness cannot be excluded completely. In addition to the degree of realism, our choice of tasks also limits our results. The subjects in our study performed tasks designed to increase body awareness specifically. Since we focused on the application context of mind-body therapies, we initially limited our task selection. However, it remains to be investigated whether a more substantial effect on body awareness has to be expected in other tasks that are less movement- or body-focused. For further application, it would be vital to conduct investigations on body awareness in different virtual scenarios. Finally, our design is limited because a mirror exposure was performed at the end of each condition to highlight the difference between virtuality and reality more clearly. However, it limits the results on the influence of perspective to the extent that the post-experience surveys could not be investigated concerning perspective.