The use of virtual environments (VE) is nowadays widespread. Their potential in academia has been discussed extensively, and numerous research applications have been presented. Since VEs offer the possibility to create settings of high ecological validity that can be fully controlled, they have been suggested for and employed in fundamental research (Blascovich et al., 2002; Mapala et al., 2017) and for therapeutic and training purposes (e.g., Bossard et al., 2007; Potkonjak et al., 2016). Fundamental research uses virtual environments to study and understand fundamental mechanisms, for example regarding deceptive behavior (Mapala et al., 2017) or proxemics behavior (Yee et al., 2007; Iachini et al., 2016). Moreover, VEs can be employed to examine and reduce stereotype bias in terms of racial or age stereotypes (Banakou et al., 2016; Oh et al., 2016; Hasler et al., 2017). In the applied area of therapeutic interventions, virtual scenarios are being tested for the treatment of paranoia, post-traumatic stress disorders and other anxiety disorders (Gerardi et al., 2010) such as flight anxiety (Cardoş et al., 2017) and speech anxiety (Pertaub et al., 2002). Increasingly, they are also being used for training purposes, mainly in the area of training motor skills, for example regarding surgery (Seymour et al., 2002), motor rehabilitation training (Holden, 2005; Pedreira da Fonseca et al., 2017) or to perfect skills in sports (Miles et al., 2012).

The development and evaluation of virtual environments for training resilience and future behavior in stressful situations has not been extensively addressed. Most notable among the exceptions is the stress resilience training conducted with military service members prior to their initial deployment (Rizzo et al., 2013). Here, users are immersed in a challenging context and train a range of psychoeducational and cognitive-behavioral emotional coping strategies believed to enhance stress resilience. More recently, a virtual environment application has been presented that enables resilience training for bullying situations (Feng et al., 2017; Jeong et al., 2017). Given its societal importance, bullying represents a critical, and yet widely neglected, field of application for virtual environment resilience training. Bullying can occur in various forms: as physical (e.g., slapping), verbal (e.g., offensive utterances), and relational (e.g., betrayal, social exclusion, spreading harmful gossip) violence (Berger, 2007). Due to its high prevalence rates in the population, bullying can be seen as a pressing societal problem. Indeed, in a meta-analytic review, Berger (2007) estimated that approximately 9–25% of school children worldwide have already been victims of bullying.

Bullying can have a powerful impact on the victims, in terms of negative affect (e.g., feeling nervous) and physiological reactions (e.g., stress, headaches, pain, sleep problems) (Hansen et al., 2006). It can also cause long-term consequences such as depression (Agervold and Mikkelsen, 2004; Sapouna and Wolke, 2013).

In the current paper, we test the recently developed virtual environment application (Feng et al., 2017) with a special focus on the influence of gender (of the bully as well as the victim) and personality factors. The aim is twofold: We (a) employ virtual environment technology as an empirical testbed in order to learn more about the mechanisms and resilience factors influencing the effects of bullying (specifically regarding the impact of the bully’s gender and the participant’s personality variables) and (b) evaluate the effects of the environment on different groups of participants with regard to their stress levels, emotional states and behavioral intentions. The results of these analyses should form the basis for an effective training intervention which could be applied to train victims or enhance prevention workshops in schools or universities.

To examine H1–H3, we conducted a MANOVA with the independent factors bully’s gender and participants’ gender and the dependent variables physiological reactions (electrodermal activity, heart rate), self-reports (bullying perception, mental state) and behavioral intentions to report the misbehavior of the instructor (informal, formal).

Regarding H1, which stated that female participants would experience more adverse reactions, the analysis showed a difference in heart rate between female and male participants, F(1,77) = 5.01, p = 0.028, ηp2 = 0.061: Males showed higher heart rate changes (M = 30.29, SE = 2.75, CI [24.82, 35.76]) than females (M = 22.05, SE = 2.45, CI [17.16, 26.93]). There were no significant effects on SCL [F(1,77) = 1.64, p = 0.205, ηp2 = 0.021]. The self-reports on mental state, F(1,77) = 0.38, p = 0.542, ηp2 = 0.005, and the perception of the bullying situation, F(1,77) = 0.05, p = 0.831, ηp2 = 0.001, did not differ significantly. Moreover, there was no difference between men and women in the intentions to report the bullying situation in a formal, F(1,77) = 0.15, p = 0.698, ηp2 = 0.002, or informal way, F(1,77) = 0.00, p = 0.956, ηp2 = 0.000.

Concerning H2, which stated that a male bully would elicit more adverse effects than a female bully, the analysis revealed no significant difference in the physiological reactions depending on the bully’s gender [SCL F(1,77) = 2.34, p = 0.130, ηp2 = 0.030; HR F(1,77) = 1.93, p = 0.169, ηp2 = 0.024]. The self-report did not reveal a difference for the variable “mental state ratings,” F(1,77) = 0.15, p = 0.705, ηp2 = 0.002, but a difference was found for the variable “perception of the bullying situation” depending on the bully’s gender, F(1,77) = 5.08, p = 0.027, ηp2 = 0.062. The male bully elicited a greater threat perception (M = 5.52, SE = 0.25, CI [5.03, 6.02]) than did the female bully (M = 4.72, SE = 0.25, CI [4.22, 5.23]). Regarding the behavioral intentions to report the bullying, the analysis did not reveal a difference depending on the bully’s gender [informal F(1,77) = 0.14, p = 0.707, ηp2 = 0.002; formal F(1,77) = 0.63, p = 0.432, ηp2 = 0.008].

The interaction of bully’s gender and participants’ gender (H3) did not show significant differences for physiological reactions [SCL F(1,77) = 0.01, p = 0.908, ηp2 = 0.000; HR F(1,77) = 0.00, p = 0.974, ηp2 = 0.000], self-reports [mental state F(1,77) = 0.37, p = 0.546, ηp2 = 0.005; perception of bullying situation F(1,77) = 0.45, p = 0.505, ηp2 = 0.006] and behavioral intentions to report the mistreatment by the instructor [informal F(1,77) = 2.56, p = 0.114, ηp2 = 0.032; formal F(1,77) = 0.70, p = 0.406, ηp2 = 0.009].

To examine whether the personality variables self-esteem, need to belong, neuroticism, and social gender moderate the results of H1–H3, we conducted three-way moderations (model 3) using the PROCESS macro by Hayes (2013). To this end, we consecutively conducted moderations, with each personality trait (self-esteem, need to belong, neuroticism, and social gender) as a moderator and the physiological reactions [electrodermal activity (EDA) and heart rate], self-reports on experiences (perception of the bullying, mental state), and behavioral intentions to report the bullying behavior on a formal and informal level as dependent reactions.

The aim of the current study was twofold: We (a) employed virtual environment technology as a testbed in order to learn more about the influence of a bully’s gender and of participants’ resilience factors on the effects of bullying, and (b) evaluated the effects of the gender of the bully (male vs. female) in one bullying situation (acting rehearsal) on the stress level, self-reported mental state, and behavioral intentions of two groups of participants (men and women). This should form the basis for an effective training intervention which might be applied as a training environment in prevention workshops in schools or universities. Therefore, we conducted an experimental between-subjects study in which we varied the bully’s and participants’ gender. A virtual environment was used to create a mild bullying situation by a figure of authority and captured participants’ (adverse) reactions by means of physiological data, self-reports and behavioral intentions to report the mistreatment. Contrary to expectation (H1), we did not find that female victims experience more adverse reactions during a bullying situation than male victims. On the contrary, males experienced a stronger increase in heart rates than did females. In line with Fowles (1980), this can be interpreted as an increased action readiness, and might be an indication that men tend to react more physically to threat. However, this needs to be further investigated in future studies. Despite this one difference, it seems that taken individually, the reactions of female and male victims are very similar. Moreover, moderators hardly changed these results. For potential applications of the virtual environment, this means that it might be useful to provide resilience training to both women and men.

We further assumed (H2) that a male bully would elicit more adverse reactions in participants, due to stereotypical beliefs about men and their different physical appearance. While we did not find any main effects for the physiological measures, the analyses revealed that the evaluation of the bullying situation is indeed more negative when the bully is male. Although participants did not report feeling worse, they described the situation as more threatening. While this might seem unsurprising at first glance, it is nevertheless remarkable that the same behavior leads to different effects if only the gender of the bully is changed. The fact that the same behavior displayed by women and men does not necessarily lead to the same effects or attributions has already been demonstrated in other realms (Deutsch et al., 1987). Although the male and the female bully’s behavior were experienced differently, behavioral intentions to report the mistreatment were not affected by the bully’s gender. For potential future application in resilience training, the more menacing effect of the male bully nevertheless suggests that to increase the effectiveness of such training, it may be more beneficial to include a male rather than a female bully. In addition, the gender could be customized to the “victim’s” preferred degree of experienced threat.

According to H3, we expected that the male bully would trigger the strongest adverse reactions in female victims, due to the above-mentioned reasons. In contrast, we supposed that the female bully would elicit less adverse reactions especially when interacting with male victims. The analyses did not reveal any significant interaction effect of the bully’s gender and participants’ gender, which is in line with the results regarding the main effects of bully’s and victims’ gender, suggesting that overall, gender is rather unimportant concerning the effects of bullying.

However, the consideration of further moderators changes the influence of the bully’s gender on different adverse reactions and the interaction of bully’s gender and participants’ gender. We considered neuroticism, need to belong, self-esteem and social gender as potential moderators. Surprisingly, the only moderator that did not influence the results was self-esteem. This was particularly unexpected given that previous research (Sapouna and Wolke, 2013) indicated that high self-esteem would enable victims to cope better with such a situation. Our results indicate that high self-esteem did not lead participants to evaluate the situation as less threatening or to feel better. However, as we focused on the immediate reaction in the situation, this does not preclude that long-term coping might be more successful in participants with high self-esteem.

The impact of the bully’s gender on the adverse reactions was affected by neuroticism, the need to belong and social gender. Starting with neuroticism, the results indicate that the higher the participants’ neuroticism scores, the more they perceived the female bully as threatening. The opposite pattern was observed for the male bully. While participants with low neuroticism found female bullies less threatening than males, people with high neuroticism evaluated both genders to be equally threatening. Moreover, participants with high neuroticism scores were more likely to formally report mistreatment from a female bully than from a male bully. Male bullies elicit more threat, while female bullies are perceived as less threatening, which may lead to less fear of complaining about mistreatment by females. Moreover, as female bullies are acting against their perceived female role of being warm and sincere, this violation may elicit a desire to penalize them (Bosson and Michniewicz, 2013), in this context through a formal report.

Moreover, the need to belong (Baumeister and Leary, 1995) moderated the relation between the bully’s gender and the perception of the bullying situation. The higher the participants’ need to belong scores, the more threatened they felt by the female bully, while the male bully induced less threat. We assume that these findings are also attributable to gender stereotypes. Participants with a high need to belong, who have a strong wish for attachments and acceptance, might believe that it is easier to befriend females, as females are expected to be friendlier and more communicative and approachable (Eagly and Karau, 2002). However, in the case of the present scenario, the female bully violated her gender role, which in turn might be perceived as particularly threatening. In contrast to this, males are perceived as less approachable, more dominant and less communicative (Prentice and Carranza, 2002); thus, the male bully was acting more in accordance with his perceived role as a male. Another indication that the female bully was perceived as a role-violating person is provided by the interaction effect of the bully’s gender and self-reported femininity on increase in heart rate. The more feminine attributes participants hold, the higher the heart rate increases when encountering the female bully, while the heart rate in response to the male bully was unaffected. One might assume that participants who indicate being more sensitive might more easily notice such a role violation, resulting in a higher heart rate, although this assumption is highly speculative at this point. The fact that only a physiological measure was affected, which is hard to control, might indicate that stereotypical beliefs are embedded on an implicit level, but controlled on an explicit level (i.e., self-report on perceived bullying). However, it needs to be acknowledged that the results on the other psychophysiological variable, skin conductance level, did not manifest themselves in exactly the same way. Although it might be seen as troubling that the two physiological measures did not yield the same results, such findings were also demonstrated in recent studies employing first-person shooter games (Drachen et al., 2010). A potential explanation for the differing impact on different psychophysiological measures might lie in the distinction between the behavioral activation system (BAS) and the behavioral inhibition system (BIS) (Fowles, 1980): While the BAS initiates behavior (approach) and is strongly associated with heart rate, the BIS is an anxiety system, which inhibits behavior and is associated with electrodermal activity. Against this background, a uniform reaction of heart rate and electrodermal activity would not be expected.

Besides the aforementioned effects, neuroticism and the need to belong also affected the interaction between bully’s gender and participants’ gender regarding physiological responses and behavioral intentions. All of these moderations show the same pattern, which indicates specific cross-gender effects. High neuroticism values affected the relation between bully’s gender and participants’ gender with respect to heart rate and informal report. Highly neurotic female participants showed larger heart rate changes in response to a male bully than to a female bully. The opposite was the case for highly neurotic male participants (note, however, that the highest increase in heart rate was observed in men with low neuroticism scores being bullied by a male, which is not in line with the pattern described). When only the cross-gender effects are addressed, it seems that in line with the construct of neuroticism (Eysenck, 1947), highly neurotic persons experience high stress levels especially when they are bullied by a person of the opposite sex. While research has shown that males more often bully males and that both females and males mistreat females (Narayanan and Betts, 2014), it might be that the unusual situation of male participants being oppressed by a female bully led the heart rate to increase. In contrast, although females might have experience of being bullied by both genders, the physical appearance of the male bully might have been more intimidating, leading to the increased heart rate. Although this claim cannot be corroborated by previous research, it seems generally plausible to assume that male bodies are perceived as more threatening. However, would this matter in a VR environment, in which no physical harm can be done? This also needs to be addressed in future research, but for the moment, in line with media equation assumptions (Reeves and Nass, 1996; Krämer et al., 2015), we suggest that people automatically react to virtual characters in the way they would toward real humans.

The same pattern of results was revealed for the intention to informally report the mistreatment: Highly neurotic males would be more likely to report bullying by a female to their friends, and might be disturbed by the role-incoherence of the female bully. In turn, highly neurotic females would be more likely to report bullying by a male to their friends.

The analyses showed a further interaction effect of participants’ gender and bully’s gender on skin conductance for participants with a very low and a very high need to belong. Participants with a low need to belong had very similar skin conductance levels in response to the male and female bully, with the notable exception that male participants reacted more strongly to the female oppressor. However, those participants with a high need to belong seem to react in a special way to a same-sex bully: Female participants showed an increase in skin conductance level in the presence of the female bully, while male participants showed such an increase in response to the male bully. Female participants were rather unaffected by their need to belong level in the presence of a male bully; indeed, those with a high need to belong even showed a slight decrease in their skin conductance level. Most notably, the combination of a male participant being oppressed by a female bully was strongly affected by the participants’ need to belong: While male participants with a low need to belong had a very strong skin conductance increase, the conductance was rather low in those with a high need to belong. This pattern is – as is customary with three-way interactions – very difficult to interpret, but seems to indicate that especially for male participants, the need to belong influences their reactions, rendering male participants with a low need to belong especially susceptible to the female bully. Moreover, the results indicate again that people with a high need to belong rather strive for same-sex connections and are more affected when they are bullied by their own sex. With regard to psychophysiological reactions, however, this pattern only emerged for skin conductance. This might indicate that in this regard, reactions are less connected to energizing activity, and are rather associated with inhibition and anxiety (Fowles, 1980). Given that we cannot exhaustively explain the patterns (e.g., why people even feel threatened when they have a low need to belong, which appears to suggest that the need to belong is not a prerequisite for reactions), future research needs to incorporate the need to belong.

It is also very important to take a closer look at what the results might mean for the identification of resilience factors: In line with results by Friborg et al. (2005), neuroticism affected the outcomes and especially influenced the impact of the bully’s gender. However, low neuroticism or emotional stability did not ease reactions in general, but only depending on the bully’s gender. Therefore, it cannot be seen as a general resilience factor. Findings regarding the role of need to belong were also mixed. While this trait affected perceived bullying and skin conductance, the results did not reveal a clear pattern. At the very least, this construct is worthy of inclusion and testing in future studies. Concerning the question of whether social gender can serve as a resilience factor, we found one single effect of self-reported masculinity on the willingness to formally report the bullying, indicating that with increasingly masculine attributes, the likelihood of reporting the mistreatment in a formal way increased. Given that masculine attributes comprise self-confidence and the ability to deal with pressure (Stein et al., 1992), it is logical that these attributes would support the participant to defend her/himself by reporting the mistreatment in formal situations. This is in line with findings that self-esteem fosters resilience (Sapouna and Wolke, 2013), although this effect did not directly emerge in our study. Moreover, social gender influenced the effect of the bully’s gender on heart rate changes. In conclusion, we recommend the inclusion of social gender in further studies. However, as the reliability of the scales used in our study proved to be rather low, we would suggest the use of different instruments. Furthermore, an explicit consideration of androgyny would be beneficial, as this could turn out to be a more adequate resilience factor than masculinity, the effects of which were rather ambiguous.

With regard to the overarching question of whether the system could be used to train resilience and appropriate behavior after a bullying event, we can conclude that future developments in this direction would be worthwhile. As the results show that people react to bullies in virtual environments, and feel especially threatened by a male bully, the environment could conceivably be used to train resilience against bullying. The experience might, for example, be integrated into a workshop, in which participants learn to withstand the bullying, regulate their emotions, and are taught appropriate responses – to the bully as well as regarding the reporting of the behavior.

The study is, of course, not without limitations. Most importantly, our results are limited to showing the effects of bullies of different gender on participants of different gender, additionally considering the role of several person variables. Our design cannot provide insights into the general question of whether the effects in the virtual environment differ from those in the real world and/or whether the effects are due to the specific bullying rather than the acting rehearsal situation. Although questions such as these have been targeted in previous research, it might be useful to address them again in further studies that include the appropriate control groups.

Although the sample is of a reasonable size for a laboratory study, the number of participants is nevertheless rather low when considering three-way interactions. Furthermore, the sample was quite homogenous and included mostly students. Therefore, the results are only generalizable to students – although students constitute one of the most important target groups for future training interventions. Given the specific setting we used (bullying by a figure of authority in an institutional setting), the results might also not be generalizable to other, more informal bullying by peers. However, as a first step, we aimed to gain insights into people’s willingness to report misbehavior of a superior, as this might even be more difficult and worthy of training compared to reporting bullying by peers. The specific setting also entailed a situation that might not have been appealing to all participants, although this would likely have been true for any type of task. However, it should be noted here that the situation might not have been sufficiently threatening, as the mean values show only mild perceptions of threat. Moreover, the setting only included a small amount of ridiculing, although this is a frequent element of bullying. Another potentially problematic aspect of the virtual environment is the fact that the second virtual character, the fellow student, was always male. Although this was kept constant in all conditions, it might have influenced especially female participants in specific ways.

Some limitations regarding the dependent variables also need to be noted. For example, the mental state scale was developed as a scale for clinical samples, which always bears the risk that there is only limited variance in a sample with non-clinical participants. However, variance appeared to be in a normal range in the present study. The psychophysiological measures have to be treated with caution, as the Empatica E4 has not been validated in previous studies, meaning that it is unclear whether the data might be influenced by artifacts. Another methodological problem is that during the baseline measurement, participants already knew whether they would be interacting with a female or male instructor, since they had seen a picture of the virtual character in the instructions. This might have attenuated the effects. For future studies including gender (especially gender of the bully), it would be advisable to collect data on stereotypical beliefs. An awareness of the participants’ gender stereotypes might facilitate the interpretation of some of the results. With regard to the person variables and potential moderators, we included those which have already been described in the literature (such as neuroticism, gender and self-esteem), but other variables, for instance prior experience with verbal and physical bullying, might, of course, also have influenced the results.

The present study demonstrates that a virtual bullying situation can have distinct effects. With regard to basic research, we conclude that the use of such an environment enables researchers to deepen their understanding of processes in bullying situations and to identify factors that influence victims’ reactions and resilience. Specifically, the results demonstrate that a male bully is perceived to be more threatening than a female bully, and that men in particular react to bullying with an increased heart rate, which might indicate their readiness to act. Moreover, the personality factors neuroticism, need to belong and social gender moderate the results. With a view to future applications, the environment could indeed be used in order to prepare people for potential future bullying situations – especially when a male bully is used. The experience might be used as part of an education program that builds on the emotional reactions by reflecting on appropriate reactions and training self-regulation of one’s own emotions, as well as learning about appropriate further actions such as formal reporting.

This study was carried out in accordance with the recommendations of the APA with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the local ethics committee of the Department of Computer Science and Applied Cognitive Science of the University of Duisburg-Essen.

Conceptualization of the study: NK, SM, and DF. Development of the virtual environment: SM and DF. Collection of the data: ET. Technical support for data collection: DF. Data analyses: ET and SS. Writing of the manuscript: NK and SS. Editing of the manuscript: SM.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.