Body ownership and response to threat

The rubber hand illusion is the experience of an artificial body part as becoming a real body part. This illusion was first reported by Botvinick and Cohen (1998), who placed a rubber hand in front of participants whose corresponding real hand was hidden from sight. When the real hand and the visible rubber hand were stroked in a synchronous fashion, participants reported to experience the rubber hand as being a part of their body. This method is widely used, with various minor and major variations to induce illusions of body ownership (Ide, 2013; Lloyd, 2007; Tsakiris & Haggard, 2005; Zopf, Savage, & Williams, 2010). Among other things, the illusion can also be produced by replacing the rubber hand by a virtual hand that moves synchronously with one’s own hand (Ma & Hommel, 2013; Padilla et al., 2010; Sanchez-Vives, Spanlang, Frisoli, Bergamasco, & Slater, 2010; Slater, Perez-Marcos, Ehrsson, & Sanchez-Vives, 2008).

Perceiving an object as being part of one’s own body has been shown to go along with increased affective reactions to threat directed at this object. Armel & Ramachandran, (2003) repeatedly tapped and stroked participants’ real hidden hand and a rubber hand synchronously (which according to Botvinick and Cohen would induce a sense of ownership for the rubber hand). If the rubber hand was then “injured”, participants displayed a strong skin conductance response (SCR), which is a widely accepted indicator of autonomic arousal (Armel & Ramachandran, 2003; Guterstam & Ehrsson, 2012; Petkova & Ehrsson, 2009). Brain imaging studies also showed that threat to an “owned” rubber hand can induce brain-activity patterns that are commonly associated with anxiety and introspective awareness (in insular and anterior cingulate cortex) and that are also obtained if the participant’s real hand is threatened (Ehrsson, Wiech, Weiskopf, Dolan, & Passingham, 2007).

Yuan & Steed, (2010) measured SCR responses to what they considered threats to a virtual hand. Participants were to play games in a virtual environment by operating a virtual hand or an arrow. During the game, a virtual lamp would fall on the operated virtual effector, which induced a reliable increase in SCR for the virtual hand but not for the virtual arrow. Ma & Hommel, (2013) pointed out that the falling lamp, which only contacted but did not damage the effector, might be taken to represent more of an impact (i.e., a contact-inducing event) than a threat (i.e., a potentially damaging event). To test whether contacting and potentially damaging events trigger different affective states, they combined a standard synchronization technique with the exposure of a virtual hand to either a contact with a ball (which was considered an impact with little damaging potential) or a contact with a cutting knife (which was considered a threat with considerable damaging potential). Their findings show that SCR increased with synchrony (i.e., perceived ownership) in the face of impact but not in the face of threat, which, however, produced elevated SCR levels independently of synchrony/ownership.

The available evidence can thus be taken to suggest that ownership is related to affective reactivity, in the sense that perceived ownership for artificial effectors is associated with stronger affective responses if these effectors are under threat. However, previous studies have used SCR to assess affective reactivity, and employed this measure merely as a convergent measure to assess ownership, while the kind and quality of the affective processes were less relevant. This has several disadvantages. While it is generally accepted that SCR is related to affective reactivity, it is a particularly non-selective, undifferentiated measure that assesses the general level of arousal (Ehrsson et al., 2007; Guterstam, Petkova, & Ehrsson, 2011; Ma & Hommel, 2013) rather than a specific emotion. As a consequence, it is difficult to exclude that SCR effects reflect general motivational attitudes (e.g., preparedness to react) or mere surprise rather than specific emotions. And, even if emotions are involved, it remains unclear which emotions that might be. Several studies have used SCR in the context of conditions that were designed to remind the participant of painful situations (e.g., Armel & Ramachandran, 2003; Yuan & Steed, 2010) and interpreted the thereby induced SCR effects as affective responses. One obvious affective response that such situations are likely to evoke is anxiety, which is why we focused on this emotion in the present study. Indeed, it makes sense to assume that people become particularly anxious if some part of their body is targeted by a threatening event, which suggests that anxiety should be more pronounced for (virtual) effectors that are perceived as part of one’s body. Given that synchrony between one’s own movement and the movement of a virtual effector increases perceived body ownership (Ma & Hommel, 2013; Yuan & Steed, 2010), we thus expected higher anxiety levels under threat to synchronized as compared to unsynchronized virtual effectors.

A second independent variable we considered was the modality of the virtual effector. Similar to the classical rubber hand setup, studies using virtual reality commonly use virtual representations of human hands as candidate body parts. Given that some authors have argued that ownership illusions require a close similarity between the candidate effector and the internal representation of one’s body (e.g., Tsakiris, 2010), this seems to be an obvious choice. However, recent studies have revealed that people can experience body ownership for body-dissimilar effectors as well: Ma & Hommel, (2015a, b) found synchrony-induced increases in ownership perception for virtual balloons and rectangles if participants could control their size, orientation, or color by moving their own hand. They concluded that people may be able to perceive ownership for any event that they can intentionally control. And yet, the findings of Yuan & Steed, (2010) and Ma & Hommel, (2013) suggest that perceived ownership for body-dissimilar effectors may not necessarily translate into the same degree of affective responsivity. With respect to our present study, this suggests that synchrony-induced anxiety under threat may be less pronounced for body-dissimilar than for body-similar effectors. To test that, we manipulated the modality of the virtual effector, which in one condition was a human hand (as in previous virtual-hand experiments) and in another condition was a cat’s claw. The claw was presented in the same orientation as the human hand (see Fig. 1) but clearly different in terms of skin and other details. We were interested to see whether the two effectors would differ in terms of ownership or agency, which we assessed in a pilot study. We were also interested to see whether and how such possible differences would translate into differences in anxiety under threat, tested in the following experiments.

We used an anxiety questionnaire to assess the subjective level of the emotion. The advantage of this method is that it provides direct insight into a specific emotion and the degree to which the participant is experiencing it. However, the disadvantage of this method is that it does not provide a continuous measure, as SCR does, and that the assessment itself takes time and attention. Among other things, this makes it difficult to provide an unbiased assessment of the ownership illusion: filling in an ownership questionnaire first is not unlikely to systematically affect the anxiety level and filling in an anxiety questionnaire first might affect perceived ownership. We, therefore, decided to manipulate perceived ownership by means of the standard synchrony manipulation but to restrict the post-induction assessment to anxiety measurements. As this raises the question whether manipulations were indeed able to induce significant ownership illusions, we first carried out a manipulation check that focused on ownership rather than anxiety. In the following, we first report the outcome of this manipulation check in a pilot study before we turn to the outcomes of two experiments that used the same experimental setup but employed anxiety measures only.

The purpose of Experiment 1 was to assess ownership-related changes in anxiety level. We used synchrony manipulations to induce (if the virtual effector moved synchronously with people’s own hand movements) or not induce (if the movements of the virtual effector were delayed with respect to people’s own movements) perceived ownership (as verified in the pilot study). The general expectation was that people would show higher levels of anxiety if the virtual effector is under threat, especially for conditions that lead to perceived ownership (i.e., with synchrony). We compared the effects for two virtual effectors: one resembling a human hand, as in many previous studies, and another resembling a cat’s claw. As the pilot study showed, less ownership is perceived for a claw than for a human hand (consistent with observations of Guterstam, Gentile, & Ehrsson, 2013; Guterstam et al., 2011; Haans, IJsselsteijn, & de Kort, 2008; Tsakiris, Carpenter, James, & Fotopoulou, 2010), which is why we expected a reduced impact of synchrony on anxiety for claws than for hands.

In Experiment 1, we induced anxiety-relevant threats by having participants engage in a game that required them to use the virtual effector to collect virtual coins and to avoid virtual cutting knives. We hoped that this manipulation would be effective in inducing certain levels of anxiety, especially for “owned” virtual effectors. Experiment 2 replicated these conditions but had each participant play only one of the two games, which allowed us to assess the impact of collecting coins and of avoiding knives on anxiety levels separately.

In Experiments 1 and 2, the levels of anxiety were assessed by means of post-experimental measures only. It is true that additional pre-measures of anxiety would have provided more information and helped reducing the statistical noise resulting from individual differences. However, having participants to report about anxiety at the beginning already would have attracted attention to anxiety being an important dimension for the study. This would have been likely to artificially boost the anxiety level, which in turn would have rendered ceiling effects more likely. Consequently, we assessed anxiety only once per participant, which means that we used post-experimental measures only and that we manipulated all independent variables between participants in all three parts of the study.

As explained above, our two actual experiments assessed perceived anxiety. Given that we wanted to avoid influences from the anxiety assessment on perceived ownership and from the ownership assessment on perceived anxiety, we assessed anxiety only. However, with the exception of the second part, we used the exact same experimental setup as in the pilot study, and thus assumed that the manipulations of synchrony and modality would have the same ownership effects than obtained in the pilot study.

The experiment replicated Experiment 1 except that we designed two versions of the game the participants were exposed to in the second part. One version contained descending coins only and participants assigned to this version were to catch them. The other contained descending knives only and participants assigned to this version were to avoid them. This modification of the experimental design created a third between-participants factor (event: catching coins vs. avoiding knives) that was fully crossed with the other two.

The main aim of the present study was to investigate the impact of ownership-relevant conditions on perceived anxiety. Taken altogether, our findings suggest three conclusions.

First, people experience more anxiety for threats of effectors that they perceive as part of their own body (cf., Guterstam et al., 2011). While our experimental approach does not speak to the underlying direct causal connections, we systematically find increased anxiety scores in conditions with synchronous relationships between virtual effectors and real hands. As these conditions also increase the perception of ownership, it makes sense to assume that the perception of ownership and the experience of anxiety are based on at least partly overlapping information.

Second, people experience more anxiety for threats of effectors that look more similar to their own hand. The lack of such similarity does not prevent them from experiencing agency, as our pilot study has shown, but it does lead to reduced anxiety even in synchronous conditions. This suggests that pre-existing internal representations of one’s body mediate the experience of anxiety, which fits with claims that self-perception integrates exogenous intersensory information with a more stable internal body image (Synofzik, Vosgerau, & Newen, 2008; Tsakiris, Longo, & Haggard, 2010). However, given that even the cat’s claw produced significant ownership effects, our findings do not support claims that the perception of body ownership is restricted to effectors or objects that resemble one’s own real body parts (Tsakiris et al., 2010). Rather than seeing top-down factors as censoring bottom-up information, our findings can be taken to suggest that bottom-up information (provided through synchrony and related factors) and top-down information (such as general expectations, perceived possibility, and biases) are integrated into a coherent percept.

Third, pronounced, systematic effects on anxiety are restricted to plausible threats. As compared to the knife condition, the coin condition produced rather low anxiety levels overall, which moreover were only mildly affected by synchrony and not sensitive at all to modality. Given that collecting coins is likely to induce some affect and arousal, it is possible that SCR measures would have been more sensitive to pick up (general, arousal-related) affective processes in the coin task. In any case, inducing substantial increases of subjective anxiety seem to require a “true” (even if virtual) threat to an effector that is perceived as part of one’s body. Along these lines, a comparison of the outcomes of Experiment 1 and 2 suggests that the anxiety pattern that we obtained in Experiment 1 was driven by the knife-avoiding part of the task but not by the coin-catching part.

Taken altogether, our findings suggest that the degree of subjective anxiety that people experience when a virtual effector is under threat is perfectly predicted by the degree to which the particular circumstances evoke ownership illusions. Minimally, this suggests that perceptions/judgments of ownership and anxiety are based on overlapping information. Exciting situations that, however, do not threaten the “physical integrity” of the virtual effector have a mild and rather nonspecific impact on anxiety, which confirms that the major part of the anxiety effect is threat specific. This observation is consistent with previous findings that mere impacts and actual threats affect SCR levels in different ways (Ma & Hommel, 2013). It is also consistent with previous demonstrations of a positive relationship between perceived ownership and SCR responses under threat (Armel & Ramachandran, 2003; Yuan & Steed, 2010) but goes beyond these demonstrations by showing a specific effect on anxiety. This does not rule out effects on other factors that are known to impact SCR, such as surprise or motivation—the demonstration of which would require the employment of more specific measures than SCR. Separating these effects seems useful and important on the way to a better understanding of the functional and neural mechanisms underlying the perception of body ownership.