The sense of embodiment (SoE) is the feeling of one’s own body. When we move our bodies in our everyday lives, we feel the SoE unconsciously. The SoE is intimately related to the sense of self and is considered as the starting point of having self-identity [1,2]. The term embodiment has been defined differently in various contexts. From a philosophical perspective, embodiment is considered to be how a person defines and experiences himself/herself [3,4]. In the field of cognitive neuroscience, embodiment is related to how the brain expresses the body [5,6].

Because the SoE is an unconscious process, its alteration or manipulation is difficult. However, Botvinick and Cohen manipulated the SoE using an experiment [7]. With the rubber hand illusion paradigm, the authors demonstrated that participants experienced the SoE on the rubber hand, generating an ownership illusion [7,8]. The rubber hand illusion experiment showed that concurrence of visual and tactile stimulation could generate an embodied experience for a part of one’s body. Recently, researchers have begun to apply virtual reality (VR) techniques to expand the SoE to the whole body. A life-sized virtual avatar for a full-body ownership illusion has been developed [9]. Slater et al showed that participants perceived the ownership illusion from virtual arms [10], and Jun et al demonstrated that a full-body ownership illusion and changes in an avatar’s facial expression meaningfully affected users’ emotions [11].

The effects of the SoE on perceptual and behavioral alterations can be extended to several medical areas. Pioneers have made efforts to apply them in psychological counseling programs [12], and others have used them to enhance cognitive-behavioral therapy for eating disorders [13]. In addition to these applications, systematic desensitized experiences with a virtual avatar could be extended to pain distraction among burn patients, phobias, and posttraumatic stress disorder [14-16]. However, for these purposes, we may need to control the SoE, which has not yet been rigorously examined.

Previous theories have suggested that the SoE can generally be divided into the sense of agency, ownership, and self-location [2]. First, the sense of agency is the subjective feeling that “I am the one who is causing or generating an action” [17]. It includes the subjective experience of action, control, intention, and motor selection, as well as the conscious experience of will [3,18]. Second, the sense of ownership is the feeling that “I am the one who is undergoing the current experience” [17]. The sense of ownership is about the self-attribution of a body that distinguishes it from the sense of agency because it can occur even without a behavior being conducted. Finally, the sense of self-location is the determinate volume in space in which one feels to be located [18]. It is determined by visuospatial perspectives, which are usually egocentric [3], and vestibular signals are also considered to have an important role in self-localization [19].

In this study, we proposed three virtual avatar methods (Figure 1) based on the definition of the three subcomponents of the SoE, as approaches to understand and control the SoE. The first method is human avatar (HA), which is expected to maintain agency, ownership, and self-location at high levels. The HA method uses a human-shaped avatar that moves congruently with the movement of the user and matches the gender and body size of the user [11]. The second method is point light avatar (PLA), which is expected to maintain high levels of agency and self-location but a low level of ownership. The PLA method is based on the point light method proposed by Johansson in his biological motion perception study [20]. The point light is placed at the joint position of the HA to sufficiently reflect the movement of the user but does not have human visual characteristics. Because it does not have human visual characteristics, the PLA is expected to produce a low sense of ownership compared with the HA. However, biological motion is the movement of a living creature and contains information such as behavior, intention, emotion, and personality [21]. Thus, we expect PLA users to be aware of their movement, and their agency will not be lowered. The third method is out-of-body point light avatar (OBPLA), which is expected to maintain a high level of agency and low levels of ownership and self-location. The OBPLA method places the user’s viewpoint behind the PLA so that the user can see the PLA from a third person perspective. Therefore, we expect that the OBPLA will show lower self-location compared with the PLA while maintaining a high level of agency. We defined the HA, PLA, and OBPLA and compared them in three user studies. In study 1, we manipulated the SoE by controlling the presence of the virtual avatar and point of view. In studies 2 and 3, we manipulated the SoE by controlling motion synchrony. The differences among the three studies are illustrated in Figure 2.

In this research, we proposed virtual avatar methodologies to understand and control the SoE and compared the HA, PLA, and OBPLA conditions using full-body motion capture and VR technologies. Study 1 suggested that it is possible to control the level of SoE from the HA to PLA conditions and from the PLA to OBPLA conditions. In studies 2 and 3, we manipulated motion synchrony and avatar type to determine whether such manipulations affect the SoE, emotional response, presence, and simulator sickness. The results suggested that motion synchrony affects all subtypes of SoE. Regarding avatar type, differences in both ownership and self-location were found between the HA and PLA conditions, indicating positive effects of the HA compared with the PLA for those two SoE variables. Additionally, we found that the PLA increased agency, ownership, and self-location compared with the OBPLA. Presence, emotion, and simulator sickness were affected by motion synchrony and avatar conditions. Our findings suggest that the SoE and subjective experiences can be affected differently with different avatar types in the full-body illusion. We believe that these results can be applied to various fields. For example, the results can be used for systematic desensitization of phobias or pain reduction by controlling the level of SoE. They can also be used to control the effects of emotion and presence associated with using virtual avatars.

In study 1, no relevant differences were found for agency among the three conditions. Despite the absence of human visual characteristics, if full-body motion-capture is maintained, the level of agency is not lowered. Recently, Zopf et al [29] compared virtual fingers and a virtual sphere moving correspondingly to a participant’s hand and found that the moving sphere did not lower the level of agency. This study suggests that the degree of visual stimulation does not affect the sense of agency for not only the hand but also the whole body. Therefore, based on existing research and this study, we suggest that movements induced by one’s intent and predictions determine the sense of agency. Moreover, in the comparison among the avatar conditions, we found that the details of the avatar’s visual characteristics did not affect the sense of agency.

In studies 2 and 3, we investigated the effects of motion synchrony on sense of agency, ownership, and self-location, and the results suggested that motion synchrony greatly affects all SoE subcomponents. Interestingly, the effects of motion synchrony overwhelmed all other effects. Therefore, motion synchrony may need to be controlled if we wish to control the SoE in general. However, the results of study 1 also suggested that the sense of body ownership and self-location can be controlled once motion synchrony is satisfied. Merging the results from studies 1, 2, and 3, we expect that we can use motion synchrony for the general control of the SoE, and we can use the presence of the virtual body and point of view for more detailed control over the SoE. The sense of ownership under the HA condition was higher than that under the PLA and OBPLA conditions. This result is an extension of previous findings that the sense of ownership is induced when the virtual body and physical body are morphologically coincident [7,30,31]. This also suggests that the HA and PLA are appropriate tools for controlling the level of ownership. The point light (PL) method is known to have sufficient information to not only express human behavior but also determine gender [32,33] and recognize emotion [34]. On the other hand, human visual features, such as important interjoint expressions, finger expressions, rotation of some joints, and skin texture are excluded in the PL method. These characteristics of the PL method seem to be suitable for controlling ownership when the PLA is used as a virtual body. Participants experienced a higher sense of self-location under the HA and PLA conditions than under the OBPLA condition. This result supports previous findings that self-location can be controlled through viewpoint changes in low agency situations [35,36]. This study suggested that the control of self-location through changes of perspective is also possible in a high agency situation, such as in a full-body motion-capture system. In study 1, there was no relevant difference in the sense of self-location between the HA and PLA conditions, but in study 2, there was a difference in the sense of self-location according to the avatar’s visual characteristics. Moreover, in study 3, we found that changes in the viewpoint (first vs third person perspective) affected the sense of agency. This suggests the possibility of mutual dependencies among the three subcomponents or the possibility of the context effect of synchrony and asynchrony. However, additional studies are required on this issue.

This study also found that there were significant differences in emotional response, presence, and simulator sickness according to avatar type and motion synchrony. Although the design of this study did not include an emotional evocation context, it was shown that the participants felt different emotional levels depending on the condition. Participants felt higher arousal under the HA condition than under the PLA condition. The HA expresses a human face, but the PLA expresses the face as point light, so the HA contains more facial information than the PLA. Although the avatar under the HA condition had a neutral face, it is possible that the human appearance influenced arousal. The present results also showed an interaction effect between motion synchrony and the avatar’s visual characteristics and a main effect in the avatar’s perspective. It would be beneficial to control presence more precisely by using these combinations of SoE subcomponents.

This study has some limitations. First, the participants in our study were all healthy young college students. It is possible that the characteristics of the participants affected the results of embodiment and subjective responses. Therefore, in future studies, it is necessary to collect data from participants of various ages and health statuses. Second, this study asked participants to move freely and measured arousal and valence as emotions, but pain, anxiety, and fear were not accurately represented. These emotions should be considered for future medical applications. Therefore, in future studies, it is necessary to give specific tasks to participants and check how these methods of controlling the embodiment affect participants in emotional tasks. Third, this study proposed the HA, PLA, and OBPLA to control the external appearance and perspective of the avatar. However, an extension of these methods can be used to control the SoE. For example, we can use lines to express the body or low-polygon human avatars to control the level of visual characteristics. We can also control the point of view, allowing the human avatar to be seen from another third person perspective, or we can see our avatar from the front. Future studies must use additional methods to regulate the SoE and identify the impacts.

In this study, we proposed and compared the HA, PLA, and OBPLA conditions to understand and control the SoE. We also studied how emotional response, presence, and simulator sickness can be affected by controlling the subcomponents of the SoE. The results suggested that the three avatar generation methodologies and two synchrony levels can control the sense of agency, ownership, and self-location, and emotional response, presence, and simulator sickness were differently affected by each methodology. This study may have implications for boosting the effects of virtual avatar applications in several fields of study, such as psychotherapy, entertainment, and education, by controlling the SoE with a full-body illusion.