Investigating proxemics behaviors towards individuals, pairs, and groups in virtual reality

ORIGINAL ARTICLE Virtual Reality (2025) 29:58 https://doi.org/10.1007/s10055-025-01127-y become. As a result, violating these distinct zones can cause discomfort (Otterbring et al. 2021). For instance, interactions with close friends or family typically occur within the personal zone. In contrast, interactions with acquaintances are generally relegated to the social zone, while interactions with strangers will occur in the public zone. Therefore, investigating when the different spaces are kept and when they are violated allows us to model which factors can shape this aspect of human non-verbal communication. 1.2 Proxemics in virtual environment Interestingly, human-human proxemic behavior in the physical world can be replicated in an immersive virtual environment (VR) populated with human-like avatars (Hecht et al. 2019; Ruggiero et al. 2019). By their human-like appearance, these avatars exert a social presence, attracting interactions and engagements from human users (Kim et al. 2018). However, when an individual approaches a human-like avatar too closely, i.e., invading their personal zone, there is a perceived discomfort due to the invasion of the space occupied by the avatar (Hecht et al. 2019). Similarly, an approaching virtual human can create a sense of 1 Introduction 1.1 The proxemics theory The proxemics theory examines how people use the space surrounding them (Hall 1966), a vital aspect of non-verbal communication. It suggests that individuals keep different distances from one another depending on how familiar and intimate they are (Hall 1966; Sorokowska et al. 2017). Previous research (as seen in Sorokowska et al. 2017) grouped interpersonal distances into four distinct zones: the intimate (up to 0.46 m), personal (0.46–1.22 m), social (1.22–3.65 m), and public zone (beyond 3.65 m). The closer the zone is to a person, the more familiar and intimate the interactions Debora Nolte [email protected] 1 Institute of Cognitive Science, University of Osnabrück, Wachsbleiche 27, 49090 Osnabrueck, Germany 2 Department of Neurophysiology and Pathophysiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany Abstract The proxemics theory explains the consistent social boundaries surrounding individuals as reported (Hall in The Hidden Dimension, Doubleday, Garden City, 1966), yet little is known about the social boundaries surrounding pairs or groups of people. The current study explored interpersonal proxemics behavior in a virtual environment, focusing on distances maintained towards individual pedestrians, pairs, and groups. Using virtual reality to simulate a city center, participants freely navigated it while their movements and gazes were captured. Importantly, the city was populated by pedestrians in different social configurations. Eye movements identified interactions defined by gaze-onsets towards a pedestrian’s head. Our results indicate that participants approached individuals with a median distance of 3.18 unity units aligned with the social space boundary as reported (Hall in The Hidden Dimension, Doubleday, Garden City, 1966). Distances kept from pairs and groups were similarly centered within the social space, revealing no significant difference in approaching behavior across different social configurations. The consistency in approaching distances suggests that personal and social spaces are not substantially altered, irrespective of the social context. Keywords Proxemics · Interpersonal distance · Virtual reality · Nonverbal behavior Received: 29 April 2024 / Accepted: 3 March 2025 / Published online: 25 March 2025 © The Author(s) 2025 Investigating proxemics behaviors towards individuals, pairs, and groups in virtual reality DeboraNolte1· ReemHjoj1· TracySánchez Pacheco1· AnnHuang1· PeterKönig1,2 1 3 Virtual Reality (2025) 29:58 discomfort in participants when their personal space gets invaded (Bailenson et al. 2003) or will lead to participants signaling to stop the approaching virtual human when they feel generally uncomfortable as it became too close (Iachini et al. 2016). Interestingly, being approached by or approaching individuals caused discomfort at similar distances (Ruggiero et al. 2019). Together, empirical work in real and virtual social encounters indicated that personal space in social interaction is shared when individuals’ spaces coincide. When an individual traverses in VR and encounters a virtual human, the individual exhibits similar proxemic behavior observed in the physical world, thereby modulating social dynamics. 1.3 Proxemics behavior of groups While substantial efforts have been invested into better understanding proxemics for individuals, the approaching behavior toward groups of people needs to be investigated and better understood. The main focus of this area of research thus far is the study of the discomfort when being approached by a group compared to an individual. Subjects felt uncomfortable at greater distances when approached by a group than an individual (Bönsch et al. 2018). They displayed a significant increase in negative affect and reactive behavior with increasing crowd density (Dickinson et al. 2019). Similarly, being approached by a group compared to individuals resulted in higher levels of arousal, which can be understood as discomfort (Llobera et al. 2010). Taken together, previous studies indicate a systematic difference in how passive individuals respond to being approached by groups as opposed to individual pedestrians. Conversely, while approaching or being approached by individuals resulted in similar distances kept (Ruggiero et al. 2019), it remains to be investigated whether subjects maintain a more considerable minimum distance when actively approaching a group compared to individual pedestrians. The current study aims to expand on previous findings by investigating the approaching behavior towards individuals and groups. Specifically, we were interested in the proxemics behavior in a naturalistic environment when subjects actively approached pedestrians in different social dynamics: individuals, pairs, and groups. To achieve this, we analyzed the subjects’ behavior recorded during a free-exploration study in a virtual reality city center. Monitoring the subjects’ proxemics behavior allowed us to explore how subjects behaved when naturally and actively approaching individuals and groups of pedestrians in different social dynamics. 1.4 The current study We proposed three hypotheses to understand proxemics behavior across various social dynamics in a virtual reality context. We expect the subjects to respect the personal space when interacting with individual pedestrians (Hall 1966). Specifically, we hypothesize that subjects will keep a social distance of 1.22–3.65 m from the virtual individual pedestrians. To our knowledge, proxemic behavior towards pairs or groups, and specifically individuals within these pairs or groups, has not been systematically investigated. In order to address this gap, we considered two potential ideas to describe space based on set theory (see Fig. 1 for a visual description of these ideas). On the one hand, subjects might respect the personal space of each member of the pair or group (Fig. 1A). On the other hand, they might consider the personal space of the group to be the joint social space of all members in the pair or group (Fig. 1B), treating the personal space of a pair or group as the union of the individual social spaces. Based on these considerations, we derived two competing hypotheses to test empirically. Hypothesis 2A, corresponding to the null hypothesis, assumes that belonging to a pair or group does not influence the distance kept towards this individual. In contrast, hypothesis 2B assumes that being part of a pair or group changes the minimal distance kept towards an individual within that pair or group. To test this hypothesis, we employed Bayes factor analysis because it allows for a direct assessment of the evidence in favor of the null hypothesis, unlike traditional frequentist methods, which primarily focus on rejecting the null. The results affirmed hypothesis 1, demonstrating that participants approached individual pedestrians in VR at distances similar to those reported in previous real-world studies (Bailenson et al. 2003; Hall 1966; Hecht et al. 2019). Furthermore, the lack of a significant difference in approaching distances between individuals, pairs, and groups confirmed our null hypothesis 2A and provided new insights into approaching behavior towards pairs and groups. 2 Methods 2.1 Subjects We recorded a total of 59 subjects, 44 of whom were included in the analysis of this paper (mean age: 23 ± 3.10; 27 females, zero diverse). Subjects who experienced substantial motion sickness, software errors, failed to follow task instructions, and one subject who did not have trials for individuals were excluded from the study (N = 15). All subjects were students at the University of Osnabrück. Participation was voluntary, and all subjects provided written 1 3 58 Page 2 of 11 Virtual Reality (2025) 29:58 informed consent. Recognizing their time, they were offered course credits or monetary compensation. The research protocol received approval from the Ethics Committee of the University of Osnabrück. 2.2 Experimental setup and procedure The data was recorded as part of a more extensive study; the experimental setup is explained in detail by (Nolte et al. 2024). Briefly, the experiment was designed using Unity3D (www.unity.com; version 2019.4.21f1) to represent a city center populated with houses, greenery, objects such as Fig. 1 The two competing ideas of the proxemics behavior towards pairs and groups. a Each group member’s personal space will be unaltered. As a result, the approaching behavior for pairs and groups will be identical to that of individuals. Thus, the minimum distance distributions for individuals, pairs, and groups will be statistically indistinguishable. b The personal space of the group will be the sum of the social spaces of each member of the pair or group; therefore, the distances kept to individual members of a pair or group will be more considerable than and statistically distinguishable from the distances kept to individuals 1 3 Page 3 of 11 58 Virtual Reality (2025) 29:58 throughout the city in a one-minute trial session that doubled as a motion sickness test. After completing this oneminute test, subjects who did not report motion sickness were admitted to the main study. During this familiarization phase, subjects could potentially see virtual pedestrians in the distance but were not able to inspect or approach them closely. During the main study, participants were directed to explore a specified area, demarcated by beige floor tiles (refer to Figs. 2 and 3), as though awaiting a friend, emphasizing mimicking real-life behavior. The entire experimental session, including the EEG preparations, lasted about two and a half hours. For half of the participants, an unrelated study followed the current one, taking an additional five to ten minutes. Participants were allowed 30 min to freely explore the virtual city, with pauses for 5-point validation checks and possible recalibrations or additional breaks as needed. During validation checks every five minutes, participants were teleported to a separate space within the virtual environment. If necessary, an internal calibration and a subsequent 5-point validation were started automatically. After a successful validation, participants were returned to their last position within the virtual city center. 2.3 The pedestrians Overall, 140 pedestrians taken from the Adobe Mixamo collection (Mixamo, 2008, https://www.mixamo.com) populated the virtual city. We selected all realistic-looking pedestrians from this collection that were reasonably benches, and pedestrians (see Sect. 2.3 The Pedestrians for more detail). The dimensions of the virtual city were built to match those of the real world - one unity unit equals one meter. The virtual city was displayed with a constant frame rate of 90 Hz using the HTC ViveProEye HMD (110° field of view, 1440 × 1600 pixels per eye; h t t p s : / / b u s i n e s s . v i v e . c o m / u s / p r o d u c t / v i v e - p r o - e y e - o ffi c e / ), the HTC Vive L i g h t h o u s e 2.0 tracking system ( h t t p s : / / w w w . v i v e . c o m / e u / a c c e s s o r y / b a s e - s t a t i o n 2 /) and the HTC Vive controllers 2.0 ( h t t p s : / / w w w . v i v e . c o m / e u / a c c e s s o r y / c o n t r o l l e r 2 0 1 8 /). Eye m o v e m e n t s were captured with the help of the Eye-Tracking SDK SRanipal (v1.1.0.1; h t t p s : / / d e v e l o p e r - e x p r e s s . v i v e . c o m / r e s o u r c e s / v i v e - s e n s e / e y e - a n d - f a c i a l - t r a c k i n g - s d k / ). The e x p e r i m e n t was played on an Alienware Aurora Ryzen computer using a Windows 10 system (64-bit, build-version 19044; 6553 MB RAM), an Nvidia RTX 3090 GPU (driver version 31.0.15.2698), and an AMD Ryzen 9 3900 × 12-Core CPU. The data was recorded using LabStreamingLayer (LSL; h t t p s : / / g i t h u b . c o m / s c c n / l a b s t r e a m i n g l a y e r) on a Dell Inc. Precision 5820 Tower computer (Windows 10; 64-bit, Build 19044; an Nvidia RTX 2080 Ti GPU; driver version 31.0.15.1694; Intel Xeon W-2133 CPU). We additionally recorded EEG data, but since it is irrelevant to the current paper, it will not be discussed further. The study led participants to explore the virtual environment freely. The participants’ movement was implemented to mimic real-life movement using small displacements controlled via a HTC Vive controller. Each session started with familiarization with the equipment and movement Fig. 2 The movement path of an individual subject plotted on top of a birds-eye view of the walkable city center area 1 3 58 Page 4 of 11 Virtual Reality (2025) 29:58 groups. Ten pairs were accessible from the central square, with varied configurations: five pairs seated, one pair with a standing and a seated pedestrian, and three pairs with both pedestrians standing. Additionally, two groups, each comprising three pedestrians, were present. One group had all seated individuals, while the other had standing ones. All standing pedestrians had an animation attached to them and were moving in place. This setup amounts to a realistic mixture of pedestrians one might encounter walking in a city. 2.4 Defining trials Our study allowed subjects to move freely throughout a virtual city center for 30 min. The movement path of one subject is shown in Fig. 2, and a short video displaying firstperson impressions of the virtual reality city can be seen dressed to appear in a city center. For this study, we categorized the pedestrians as individuals, referring to pedestrians standing or sitting alone, pairs of pedestrians that consisted of two people near each other, and groups of pedestrians, understood as a group of three or more individuals close to one another. An example of each category can be seen in Fig. 3. A subset of the 140 pedestrians was selected for the current analysis. First, only stationary pedestrians were considered to mitigate biases from differing movement speeds. Furthermore, pedestrians were only included in the final data analysis if they were positioned in the walkable area of the experiment. Overall, five static individuals positioned around central fixtures such as tables, benches, or statues were included in the analysis. Next to static individuals, we analyzed the approaching behavior towards pairs of two, from now on simply pairs, and groups of three, in short Fig. 3 Proxemics behavior of an individual participant towards individuals, pairs, and groups. The upper panel displays three examples of pedestrians visible in the VR scene: a an individual pedestrian, b a pair of pedestrians, and c a group of three. The bottom panel displays the different movement paths around the pedestrians in each condition in the example of a single subject. The subject shown in this plot is identical to that shown in Fig. 2. For easy visualization, the location of the pedestrians is set to be the origin. For the condition of pairs b and groups c, the first gazed-at pedestrian of the pair/group is the graph’s origin. The other pedestrians of that pair or group are left out of this display. The dark blue circle in each graph corresponds to the personal space of the approached individual, and the light blue area corresponds to the social space 1 3 Page 5 of 11 58 Virtual Reality (2025) 29:58 to evaluate the effects of avatar group size and participant gender on minimal distance. Specifically, we utilized the generalTestBF function from the BayesFactor package in R (Morey and Rouder 2018), which allowed us to account for the nested nature of the data by treating participants as a repeated measure. This analysis enabled us to compare models and determine whether the data provided evidence for the null hypothesis (H0: no effect of group size and gender) relative to the alternative hypothesis (H1: there is an effect). Importantly, in this step, we included only the distances between the subject and the first gazed-at pedestrian of a pair or group. This approach ensured that the statistical analysis was anchored to a subject-defined event, starting with their first visual engagement with the pair or group. By doing so, we avoided treating the approaching distances to all pedestrians in a pair or group as separate events, as these are not independent measures. The minimum distances towards the remaining pedestrian of a pair and the mean minimum distance towards the remaining two pedestrians of a group were assessed (see Fig. 4) and compared to the distribution of the first gazed-at pedestrian, showing that they remained within the range of the minimum distance kept toward the gaze at pedestrian. 3 Results 3.1 Number of interactions Due to the nature of trial definitions, a diverging number of events were observed across the different conditions. On average, our subjects approached individuals five times (IQR = 4), pairs an average of 13 approaches (IQR = 8), and groups four times (IQR = 2). An example of interactions can be seen in Fig. 3. The upper panel displays pedestrians visible in our virtual reality scene, while the bottom panel shows the different movement paths toward pedestrians of one subject. The different number of gray lines correspond to the different number of approaches. It is visible that for this particular subject, most interactions were had with pairs of pedestrians, while only three different interactions could be found among the group of three. 3.2 Distance kept per condition Before a comprehensive comparison, we analyzed the distances maintained by participants relative to the social constructs: individuals, pairs, and groups. We compared the medians (over trials) of the minimum approaching distance (over time) per condition within each subject. Individually, 16 subjects approached singular pedestrians more closely than pairs or groups. Conversely, 18 subjects kept a bigger in the supplementary material. As the study did not contain predefined trials, we had to define trials offline using the exploration behavior of the subjects. The trials were centered around subjects looking at pedestrians’ heads. As a result, we classified the data into gazes, all eye-stabilization movements, and saccades, utilizing the eye movement classification algorithm described in (Nolte et al. 2024). In short, the eye movement classification employed a velocitybased algorithm with an adaptive threshold to differentiate between gazes and saccades (Nolte, et al. 2024). The classification was based on versions of the REMoDNaV (Dar et al. 2021) and MAD saccade (Voloh et al. 2020) algorithms, adjusted for three-dimensional data (Keshava et al. 2023) while accounting for translational movements of the player within the virtual scene (Nolte et al. 2024). Subsequently, gazes specifically targeting pedestrian heads were isolated. Each trial was demarcated around a single gaze event, encapsulating a period from 50 s prior to 150 s post-gaze initiation. This trial duration was chosen to have an interval longer than the typical duration of an encounter, capturing the complete approach and departure phases. Furthermore, the interval is much shorter than the total experimental duration to distinguish between encounters involving the same avatars, even when other interactions occur. For the pairs and groups, the pedestrian under investigation was defined as the one gazed at first. However, another pedestrian out of that pair or group could be under investigation in a subsequent trial. The trials were categorized into three conditions: approaching individuals, pairs, or groups. It is important to note that the number of trials per condition and subject differed due to the subjects’ varying exploration behaviors. Defining trials based on the explorations allowed us to analyze the proxemics’ behavior in a naturalistic setting. 2.5 Calculating distances We assessed the minimum distance between the participant and the virtual pedestrians during each trial to evaluate proxemics behavior across individuals, pairs, and groups. The minimum distances were determined by calculating the Euclidean distances between the subject and pedestrian positions throughout the trial and then computing the minimum. The Euclidean distances were calculated using two dimensions only, the x-and z-coordinates, as the elevation of the ground in our environment was constant. Then, for each participant, the median (over trials) of the minimum approaching distance (over time) was calculated for each of the three conditions: approaching individuals, pairs, and groups. To compare the three conditions, we computed the median (over subjects) of that median (over trials) of the minimum distance (over time) of each condition across subjects. A series of Bayesian analyses were conducted 1 3 58 Page 6 of 11 Virtual Reality (2025) 29:58 Fig. 4 Distances across subjects for all three conditions. a The plot displays the probability density distribution of all subjects’ median minimal approaching distances for the individual, pair, and group conditions. The shaded areas in the background correspond to the different spaces around the pedestrians. Dark blue represents the personal space, light blue is the social space, and white represents the public space. b The distributions of all subjects’ median minimal approaching distances are additionally displayed using bar plots. The x-axis of this plot is identical to the x-axis of Figure a. We additionally plotted the distributions of all subjects’ median minimal approaching distances towards the other pedestrian of a pair (not the first gazed at) or the mean distance of the other two pedestrians of a group 1 3 Page 7 of 11 58 Virtual Reality (2025) 29:58 and groups identically to individual avatars, corresponding to our null hypothesis 2A. 4 Discussion This study investigated interpersonal distances kept towards pedestrians in different social dynamics, individual pedestrians, pairs of pedestrians, and groups of pedestrians. In detail, we analyzed approaching behavior in a lively virtual reality city center that subjects could explore freely. When evaluating the distances kept toward individuals, we observed a median distance corresponding to the social space of the individuals (Hall 1966). This result substantiates our main hypothesis, suggesting that in casual virtual interactions, participants maintain a distance akin to Hall’s delimitation of social space (Hall 1966). Such a pattern stands in accordance with recent literature (Bailenson et al. 2003; Hecht et al. 2019; Huang et al. 2022), concluding that personal space and non-verbal interactions in virtual reality follow the same implicit rules as in real-world settings. Considering the approaching behavior towards pairs or groups of pedestrians, we found evidence for no difference between the different social conditions. This result supported our null hypothesis 2 A: There is no difference between the approaching behavior towards individuals and that towards pedestrians in pairs or groups. Our observations suggest that proxemic behavior towards individuals remains consistent, irrespective of their engagement in diverse social dynamics. Thus, hypothesis 2B must be discarded as the minimal requirement that belonging to a pair or group results in different minimal distances than being an individual is not met. Overall, we showed that unknown pedestrians in a public setting are being approached up to and including their social space but rarely enter their personal space regardless of the social situation. Interestingly, as opposed to approaching behavior towards individual subjects, the distances kept towards pairs and groups of people were somewhat surprising in light of current research findings. Studying proxemics behavior with a stop-distance-task revealed higher levels of discomfort (Llobera et al. 2010) and the wish to keep a bigger distance (Bönsch et al. 2018) when approached by a group instead of a single person. The present results oppose these findings to some degree, demonstrating that subjects interact with pairs or groups of pedestrians at the same distance as individuals. Several factors might elucidate the observed disparity between our findings and previous studies. Primarily, differences in study designs could be a contributing factor. Contrary to studies where stationary participants were approached by pedestrians (as seen in Bönsch et al. 2018; Llobera et al. 2010), our research emphasized interpersonal distance from individual pedestrians, while ten subjects displayed a mixed pattern; three had the smallest median distance when approaching groups and the most extensive distance when approaching pairs, while seven subjects kept the smallest distance towards pairs and the largest one towards groups. Comparing minimal distance for individual subjects, no overarching pattern emerged. For closer analysis, we compare the median (over trials) of the minimum distances (over time) kept toward individual avatars across all subjects. The data are visualized in Fig. 4. The median (over subjects) of the median (over trials) minimal (over time) distance was 3.18 Unity units (IQR = 2.32) kept from individual pedestrians. Comparing individual subjects, the smallest distance kept was 0.79, while the maximum distance reached 6.94 unity units. Overall, two subjects had a median distance within the personal space of the approached pedestrians, 23 subjects kept a distance aligned with the social space of the pedestrians, and 19 subjects had a median distance greater than the social space. Thus, the distribution of distances kept to individual avatars is fully compatible with previous work, supporting our first hypothesis. We now turn to analyzing approaches towards the gazed-at pedestrians of pairs and groups of avatars. This data is visualized in Fig. 4. The median (over subjects) of the median (over trials) minimal (over time) distance kept towards pairs was 2.7 unity units (IQR = 1.72) and towards groups 2.7 unity units (IQR = 2.03). For pairs, one person had a median within the personal space, 31 subjects kept a median distance aligned with the social space, and 12 had a distance bigger than that. For groups, 32 subjects approached up to and including the social space of the pedestrians, while 12 subjects kept a distance bigger than that. Statistical differences between the conditions were determined using a Bayes factor approach. The analysis indicates moderate evidence supporting the null hypothesis for the effects of avatar group size (BF01 = 9.49) and strong evidence for null in the interaction between avatar group size and participant gender (BF01 = 24.66). This evidence suggests that avatar group size and its interaction with participant gender do not significantly impact minimal distance. There is slight evidence suggesting a potential effect of participant gender (BF01 = 0.41), but it is not strong enough to be conclusive. The inclusion of random intercepts for repeated measures is crucial, as evidenced by the overwhelming Bayes factors favoring models that account for individual differences (e.g., repeated measures alone: BF01 = 4.43e − 13). We also visually examined the minimum distances kept toward the other pedestrians in a pair or group (see Fig. 4B). The distributions were within the same range as those of the first gazedat pedestrians, excluding potential concerns of confounding factors. These results show that subjects approached pairs 1 3 58 Page 8 of 11 Virtual Reality (2025) 29:58 approaching is determined by the participants and, as a result, a dependent variable, the present study is not suitable to examine this effect. Additionally, as the current data were collected as part of a more extensive study, other influencing factors were not controlled, including fundamental cultural differences (Hall 1966; Little 1968; Sorokowska et al. 2017) and environmental factors (Little 1968; Sorokowska et al. 2017). Controlling environmental factors in a VR setup is challenging; for instance, the virtual scene might differ from the actual room temperature experienced by subjects wearing the head-mounted display. Overall, while our results contribute to understanding proxemics behavior towards pairs and groups, additional influential factors warrant further exploration. Several aspects of the current study design could be extended or refined in future research. For instance, the variability in spatial placement, timing, and sequence of different pedestrian configurations may have introduced variability that influenced our results. More consistent control of these variables could enhance the robustness of the findings. Additionally, controlling factors such as age, gender, and general likability of the pedestrians and including responsiveness towards the participants could provide deeper insights into approaching behavior towards groups in future studies. Similarly, future studies could control the approaching direction, i.e., if pedestrians are being approached from the front or back, to examine if this effect influences the approaching behavior towards groups. Moreover, directly comparing active approaches versus passive encounters could help resolve discrepancies with previous studies. While these methodological considerations are not essential for the validity of our current conclusions, they represent important steps toward a more comprehensive understanding of approaching behavior in various social contexts. 5 Conclusion The current study investigated interpersonal distances in a free exploration study in a virtual reality environment. Our findings corroborate the principles of the proxemics theory (Hall 1966), reinforcing the credibility of conducting such explorations in a virtual domain. Additionally, we demonstrated that social dynamics do not modulate this effect, contrary to recent literature findings. This insight illuminates an under-researched facet of approach behavior, emphasizing the influence of group dynamics on individual interpersonal distances. Given the divergence from existing literature, further research is imperative to comprehensively understand this nuanced non-verbal communication dimension. Supplementary Information The online version contains supplementary material available at h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 1 0 0 5 5 - 0 distances within a free exploration framework. Allowing for self-motion could have resulted in less anxiety and led to consistent proximities between subjects and individuals or groups. Additionally, the design and behavior of our virtual pedestrians might have influenced participant behavior. Specifically, the avatars in our virtual environment were passive, lacking responsiveness to participants, and avoided direct eye contact - a factor previously highlighted for its impact on interpersonal distances (Bailenson et al. 2003). However, as individuals were approached at similar distances, as shown in the literature (Bailenson et al. 2003; Hall 1966; Hecht et al. 2019; Huang et al. 2022), this seems a less likely possibility. Taken together, how far being active in the approach towards pairs and groups influences preferred interpersonal distances deserves attention in future work. Virtual reality environments often elicit social behaviors that closely mirror those observed in real-world interactions, with users instinctively adhering to familiar social norms and drawing inferences based on others’ appearances. As a result, factors such as clothing style (Wang and Yeh 2013), perceived attractiveness (Waddell and Ivory 2015; Welsch et al. 2020), and other contextual factors (Otterbring et al. 2022) may influence approach behaviors in VR, similar to their effects on physical spaces, shaping users’ perceptions and affecting their proximity preferences. In addition to these appearance-based influences, several other factors known to affect proxemic behavior could have impacted the present results. Personal factors such as age (Iachini et al. 2016; Webb and Weber 2003; Willis 1966) and gender (Bailenson et al. 2003; Hecht et al. 2019; Huang et al. 2022; Iachini et al. 2016; Webb and Weber 2003) of both the subject and pedestrians have been shown to influence proxemics behavior. While the narrow range of the age distribution of the subjects precluded an assessment of age effects, our results provided limited evidence for a gender effect. The gender effect was rather small to be conclusive on its own but not incompatible with previous literature. Notably, no interaction between gender and social configuration was found. Thus, for the research question in the focus of the present study, gender does not influence the minimal distances kept as a function of individual/pair/group. However, the absence of control or recording of pedestrians’ perceived gender and age limited our exploration of previously reported pedestrian effects (Bailenson et al. 2003; Huang et al. 2022; Iachini et al. 2016). Moreover, the orientation of the person being approached can affect the distance maintained towards them (Bailenson et al. 2003; Hecht et al. 2019), with smaller distances observed when approaching from behind compared to the front (Bailenson et al. 2003; Hecht et al. 2019), a factor not considered in this study. As the pedestrians being approached and their orientation while 1 3 Page 9 of 11 58