may have contributed to the decreased peripersonal space boundary since it has been shown that larger perceptions of body size decrease personal space [Phillips 1979; Sanders 1976]. But our results do not show a clear correlation since the monster has a body size intermediate between the intimidating male and the rest of the avatars.
Figure 13 also shows us that once the peripersonal space boundary is crossed, reaction times trend slower for both the average and intimidating agents. The reaction times for both of these agent groupings appear to be non-monotonic. The implication here is that depending on the social characteristics that are conveyed, there may be an initial acceptance of an agent once the peripersonal space boundary has been crossed. This may be tied to the idea that when a person breaches another’s peripersonal space, behavior is adjusted to respond appropriately to this new situation fast (i.e., a fight or flight situation). Under such circumstances, one’s capacity is tied up to cope with the emergency situation of having another body directly in the space around his or her own body. Distress could lead to an increase in reaction time and thus a decrease in attention and concentration in cognitive tasks Kato et al. [2009]. This trend did not happen with the monster, as the reaction times to this agent monotonically increase. The latter two findings, to our knowledge, have not been reported previously.
which seems at odds with the results from the likeability survey about these objects. Users responded significantly more slowly to the average agents than they did both the intimidating agents and the monster. Additionally, the monster, who was rated as the most “unlikeable” by participants of the likeability survey, was the only character that had reaction times that monotonically increased (meaning that the further away it was perceived, the slower the reaction time got). For both the average and intimidating characters, there was an increase in reaction time when the initial peripersonal space boundary was breached before it decreased again.
The type variations and reaction time results can give guidance to designers of virtual spaces who are interested in specific types of interactions. Direct interaction methods within virtual reality is difficult [Mine et al. 1997; Poupyrev et al. 2000; Kjeldskov 2001]. Collaborative interactions of avatars have been the subject of significant study but remain a difficult topic because the fidelity of the interaction is still low [Hrimech and Merienne 2010; Zibrek et al. 2017]. Our results can give some indication for the proper spatial requirements of virtual environments that lie within comfortable interaction distances for social and object interaction. Reaction times may serve as proxies for attention, with shorter reaction times indicating stronger reactions, and give guidance about the nature of what users will be tempted to attend to within such spaces.
4.4.1 Limitations and Future Work
This study has limitations due to the nature of commodity-level virtual reality equipment and avatars typically employed in virtual reality. Mutual gaze is a known moderator of peripersonal space [Argyle and Dean 1965;
Bailenson et al. 2001, 2003; Yee et al. 2007], and eye tracking in virtual reality is only on the cusp of accessibility.
The agents used in our study did not exhibit naturalistic gaze behavior, and this can change people’s perception of them [Bailenson et al. 2003; Garau et al. 2003]. In particular, facial expressions are also a moderator of peripersonal space [Ruggiero et al. 2017], as well as preconceived notions of an interaction partner’s personality traits [Iachini et al. 2015a; Pellencin et al. 2018] and the general nonverbal cues that body language conveys [Yee et al. 2007]. While it is possible to generate and give agents facial expressions and to convey some body language cues in immersive virtual environments, the process is arduous and inverse kinematics for commodity level equipment do not yet convey naturalistic movements. In addition, there are other social factors involving things such as the personality of the agent that we did not account for [Hayduk 1983]. Achieving lifelike virtual agents
in both appearance and action is an open problem, of course.
Giving participants surveys based on their attitudes toward a broad range of factors could give us a more refined model of how users treat and form peripersonal space. Such factors might include the mental health of the user (the level of anxiety they possess [Iachini et al. 2015b] or if they have any behavioral medical diagnosis [Alcorn et al. 2011; Delevoye-Turrell et al. 2011]) and the personality type of the user [Iachini et al. 2015b].
Thus, they might be fairly invasive, but understanding the complexities of human interaction through a psychological lens could lead to models that would help developers refine interaction distances in applications with more
specificity and nuance than is currently possible. Such applications may include the assistive and rehabilitative ones mentioned in the Introduction.
It would be interesting to see how the dimensions of a self-avatar affect the modulation of peripersonal space and thus interactions, since the dimensions of self-avatars affect action possibilities in virtual environments [Lin et al. 2012; Kilteni et al. 2012b]. Determining how self-localization affects perception of action capabilities and the modulation of peripersonal space in multi-user environments considering seems fruitful given recent work [van der Veer et al. 2019] on errors in self-localization. It would also be interesting to see if and how peripersonal space boundaries change with embodiment and the factors influencing it [Gonzalez-Franco and Peck 2018]. And finally, it would be interesting to determine the required level of copresence to evoke naturalistic peripersonal space modulation considering the findings of Podkosova and Kaufmann [2018a] that show collocated users to modulate some level of peripersonal space while distributed users do not.
4.4.2 Conclusions
Again, this Chapter represents the second contribution of this dissertation. We have shown that peripersonal space boundaries in immersive virtual environments can be measured easily with commodity level virtual reality equipment. These boundaries can be modulated and reactions to these boundaries can be varied based on salient features of objects and agents in those environments. These results are promising and have implications for the design of immersive virtual environments using current technology to improve user experience and
facilitate successful social interaction. As mentioned in the Introduction, these findings can have a positive impact on applications such as those related to assistive therapy for children and adults [Alcorn et al. 2011;
Boyd et al. 2018], training and simulation applications for medical, military and educational purposes [Fox et al. 2009]. While these findings are novel and have promise, future work should focus on the factors affecting
peripersonal space boundaries in virtual reality. The future work that we intend to conduct to expand our understanding of peripersonal and interpersonal space is discussed in the following chapter.
This chapter was reprinted with permission ©IEEE 2020 (see Buck et al. [2020b]).
5 Measuring Personal Space with Relation to Bodily Manipulation
This chapter takes a closer look at how personal space is maintained in immersive virtual environments.
We now know that peripersonal space can be measured and that its boundaries depend on the type of interaction occurring in virtual reality. In this chapter we want to understand how different characteristics of self-avatars affect the boundaries of interpersonal and peripersonal space. This chapter uses the same methodology as in the previous chapter, and presents a novel understanding of how both embodiment and manipulated arm dimensions affect both interpersonal and peripersonal space in an immersive virtual environment. Two experiments using visual and tactile stimuli to determine interpersonal and peripersonal space boundaries were employed.
In the first experiment, users experience differing levels of embodiment, and both interpersonal and peripersonal spaces are affected. In the second experiment, users experience self-avatars with differing arm dimensions, and we found that interpersonal space was affected by the change in dimension while peripersonal space was not. These findings have important design implications for virtual reality developers and researchers to consider, and tease out interesting results that lead toward new research questions.
5.1 Research Motivation
We consider personal space here in two categories: interpersonal and peripersonal space. Interpersonal space is the distance that an individual maintains between themselves and another individual [Hall, 1963; Hayduk, 1983]. Peripersonal space is an individual’s perceived reaching and grasping distance [Berti and Frassinetti, 2000; Delevoye et al., 2010; Farne et al., 2005; Rizzolatti et al., 1997]. We measure interpersonal space by
asking virtual reality users to report their comfort distance, or the distance at which the user is no longer comfortable with an object or person being in their personal space. We measure peripersonal space through reaction times extrapolated by a multisensory task that requires users to react to objects and people within peripersonal space.
Interpersonal and peripersonal space are intermingled, and the expansion and contraction of both spaces is contingent upon interaction context in a similar way [Iachini et al., 2014].
Embodiment is a sensation that can easily be induced in immersive virtual environments [Kilteni et al.,
2012a; Slater et al., 2010b], and the level of embodiment can affect the way users feel about self-avatars [Waltemate et al., 2018]. The way space is mediated around the body can be directly affected by the sense of embodiment [Kilteni
et al., 2012a; Lenggenhager et al., 2009]. Virtual reality affords developers the option to design environments that offer users different senses of embodiment. Therefore we seek to determine what degree of embodiment is required to evoke naturalistic peripersonal space boundary modulation in immersive virtual environments.
Physical characteristics of humanoid models are readily manipulated in immersive virtual environments;
users can be given self-avatars that do not match their physical appearance whatsoever. Prior literature suggests users of immersive virtual environments adapt to the size of their given self-avatar and behave accordingly [Creem-Regehr et al., 2015; Jun et al., 2015; Lin et al., 2012; Stefanucci and Geuss, 2009, 2010]. Additionally, there is work showing that perception of immediate reaching and grasping space extends when tools are used that increase arm length [Berti and Frassinetti, 2000; Farne and Ladavas, 2000; Holmes and Spence, 2006;
Iriki et al., 1996; Ladavas and Serino, 2008; Witt et al., 2005] and can vary when virtual reality users are given self-avatars of different volume [Normand et al., 2011].
Thus, these are the research questions this work seeks to answer:
• Q1: Does the level of embodiment a user experiences affect the interpersonal and peripersonal space of users in an immersive virtual environment?
• Q2: Do an avatar’s physical characteristics, such as arm length, affect the interpersonal and peripersonal space of users in an immersive virtual environment?