An important aim in cognitive neuroscience is to understand the basic principles and functional mechanisms that give rise to our embodied sense of self (Blanke, 2012
; Blanke & Metzinger, 2009
; Ehrsson, 2012
; Tsakiris, 2010
). Studies on body illusions provide an important contribution to our understanding of embodiment by investigating the conditions in which (illusory) embodiment may or may not occur. It has been shown that body illusions may be induced when two or more sensory modalities (e.g. vision and touch) are activated in a synchronous matter. A well-known example is the rubber hand illusion (RHI) in which participants report ownership over a rubber hand when an experimenter synchronously strokes the (visible) rubber hand and the participants’ (invisible) real hand (Botvinick & Cohen, 1998
), but not when stroking is asynchronous and random. Although there is a large body of evidence that multisensory integration between synchronous signals from different modalities is a dominant factor in the construction of the bodily self, contrasting views have emerged with regard to how multisensory integration contributes to embodiment. More specifically, two contrasting ideas have been formulated. Kalckert and Ehrsson (2014
) suggest that (illusory) embodiment operates via an all-or-nothing principle. Once a correlation has been established between two sources of information, be it two signals from different sensory modalities, or between signals in sensory and motor cortices during motor control, the inclusion of information from an additional sensory modality will not enhance the degree of illusory embodiment. Conversely, Samad et al. (2015
) suggest that illusory embodiment varies as a function of the available sensory evidence for a common cause. According to this account, synchronous activity in an additional sensory modality may strengthen the inference that a body or body part belongs to the self and will increase the strength of the body illusion. The current study aims to resolve these opposing views by investigating the question of whether or not the inclusion of additional synchronous sensory information will enhance the degree of illusory embodiment. The outcomes of the study are relevant for advancing our fundamental understanding of the mechanisms that underlie body ownership (Kilteni et al., 2015
), self–location (Blanke, 2012
) and construction of the body schema (Maravita et al., 2003
) in support of action (de Vignemont, 2010
). Insights may hold relevance for the treatment of neuropsychological patients with deficits involving body representations such as phantom limb pain (Moseley et al., 2012
; Ramachandran et al., 1995
), alien hand syndrome (Schaefer et al., 2013
), and xenomelia (Lenggenhager et al., 2015
), or may support interventions that facilitate the acceptance and control of artificial limbs following amputation (Ehrsson et al., 2008
; Rognini et al., 2019
). In addition, the study findings may be relevant for the induction of body illusions in the treatment of chronic pain (Pamment & Aspell, 2017
) and the creation of effective methods for inducing first and third person body illusions in virtual reality (Debarba et al., 2015
; Galvan Debarba et al., 2017
The conclusion by Kalckert and Ehrsson (2014
) that (illusory) embodiment functions as an all-or-nothing phenomenon reflects the outcomes of their experiments in which they compared different methods to induce the RHI. They found body ownership over the rubber hand to be equally strong with the traditional static induction of the RHI through stroking by an experimenter, an active movement condition in which active movements of the participants’ real hand animated the rubber hand, and a passive movement condition in which the real hand and the rubber hand were moved by the experimenter (Kalckert & Ehrsson, 2014
). Comparable effects were found in an earlier study from the same lab that compared active and passive induction of the RHI through movement (Kalckert & Ehrsson, 2012
). Furthermore, other labs have reported comparable levels of ownership with different induction methods (Brugada-Ramentol et al., 2019
; Pyasik et al., 2019
; Riemer et al., 2013
Nevertheless, several inconsistent findings have been reported which cast doubt on the conclusion that the strength of body illusions is independent of the specific method of induction. For instance, Walsh et al. (2011
) found that passively induced synchronous movements led to higher reports of illusory body ownership than movements that were self-induced, whereas Dummer et al. (2009
) found a trend towards stronger illusory embodiment of a rubber hand with active induction than with passive induction. Strikingly, Ma and Hommel (2015
) found that body ownership was much stronger in a condition in which a virtual hand illusion (VHI) was evoked trough self-movement versus a static condition in which the illusion was evoked by synchronous visuotactile stimulation [also see Ma et al. (2017
) for similar results]. These latter findings may be explained by the larger degrees of freedom in virtual reality and the opportunity for participants to acquire a more extensive set of multisensory samples supporting the experience of ownership and agency over the virtual hand.
Altogether, it seems too early to conclude that different methods for inducing illusory embodiment are equally effective and that illusory embodiment operates as an all-or-nothing phenomenon. A complicating factor in this discussion is that comparisons in the strength of body illusions between active, passive and static induction methods may be confounded by differences in the experience of agency during action execution (Gallagher, 2000
; Tsakiris et al., 2006
). Although, research on the (in)dependence of agency and body ownership is still ongoing and unresolved, several studies have found increases in agency to be accompanied by increases in body ownership (Brugada-Ramentol et al., 2019
; Kalckert & Ehrsson, 2017
; Ma & Hommel, 2015
; Ma et al., 2017
; Tsakiris et al., 2006
). Hence, increases in agency in association with active movement might add to the experience of owning an artificial body or body part.
Another problem that potentially invalidates comparisons between illusions evoked by (passive or active) movement and static conditions, is that actions will activate a large number of proprioceptive channels in the skin, muscles, tendons and joints that register changes in limb positions, joint angles, and force applied to the body. Likewise, when movements involve the head or the trunk, vestibular channels that register acceleration and balance (Day & Fitzpatrick, 2005
) will become activated. Considering that interactions between sensory channels may result in suppression or enhancement depending on the task and movement (Walsh et al., 2011
), effects on body illusions and comparisons between static—and movement induced body illusions may be difficult to predict and come to vary across experiments. In the present study, we circumvented these methodological problems by making a direct comparison between two similar movement conditions in which we only vary the inclusion of the tactile modality.
The hypothesis that the strength of body illusions may vary as a function of the available sensory evidence was forwarded by Samad et al. (2015
). The authors adopted a Bayesian causal inference model of multisensory perception to account for the rubber hand illusion (RHI). In their model, a body illusion is characterized by the inference of a common cause for conflicting proprioceptive, visual and tactile sensations. The strength of the overall evidence for a common cause in the end determines the probability that a common cause is inferred and that the illusion arises. Their model predicted that synchronous stroking would strengthen the illusion as compared to a version in which the rubber hand was merely watched and not stroked. This prediction was confirmed in an experiment. The results suggest that the addition of more synchronous visuotactile information strengthened the overall evidence for a common cause, not only tipping the balance to a point where the illusion was experienced but creating a stronger illusion experience as compared to the condition where participants merely watched the rubber hand (Samad et al., 2015
It is the question, however, whether the experiment by Samad et al. (2015
) really proves the idea that an induced body illusion can be strengthened by additional synchronous information. In their study, at least one quarter of all participants scored below the midpoint on the ownership question (Q3) “I feel like the rubber hand is my hand” at the pre-test when presented with the rubber hand. The increase in ownership that was reported after visuotactile stimulation could (at least partly) be explained by this subset of participants who did not report the illusion at the pre-test. Furthermore, it is also possible that participants who scored positively on the ownership question did not experience a full-blown body illusion but simply acknowledged the idea that visual and spatial properties of the hand could pass as their own hand. Without more objective measures of body ownership such as skin conductance in response to threat (Armel & Ramachandran, 2003
) or proprioceptive drift (Botvinick & Cohen, 1998
) it is difficult to determine which participants did experience a RHI and which did not. Consequently, it could be that the average increase in body ownership in the experiment of Samad et al. (2015
) reflects the induction of the RHI in participants who did not experience an illusion at the pre-test.
A study by Choi et al. (2016
) may provide additional support for the hypothesis proposed by Samad et al. (2015
). They enriched the acquisition phase of the VHI by including tactile and auditory feedback on participants’ actions with a virtual xylophone. Action feedback was found to systematically enhance the strength of the VHI, supporting the hypothesis that adding synchronous multisensory evidence may increase the strength of body illusions. It should be noted though that Riemer et al. (2019
) have argued that these effects probably reflect increased emphasis on the goal-directed aspects of the actions. In line with this interpretation, Wen et al. (2016
) have found that proprioceptive drift (an implicit measure of body ownership) and body ownership will increase when participants make goal-directed actions towards a virtual object, as compared to intransitive movements. In addition, including sensory information about the effects of actions will enhance the sense of agency that participants experience over their actions (Haggard, 2017
), as acknowledged by Wen et al. (2016
). Consequently, agency is probably a confounding factor that limits the interpretation of the findings by Choi and Li (2016
). In the current study, the possible confound of action effects is circumvented by using repetitive stroking and waving movements in which the sensory consequences of the action are no goal in itself.
In sum, the question whether the inclusion of additional multisensory information can uplift (illusory) embodiment or whether the illusion should be seen as an all-or-nothing phenomenon can, in our view, not be answered clearly yet on the basis of the current literature. In our study we have chosen to investigate this research question in the context of the full body illusion (FBI) as insights into the nature of illusory embodiment may help to optimize systems of virtual reality and possibly prevent costs for the real body that are considered to accompany online role-playing from a third person perspective (3PP; Ganesh et al., 2012
; Swinkels et al., 2020
). During the experience of a full body illusion (FBI; Lenggenhager et al., 2007
) participants experience touch to originate from a body that they see in front of them and report to feel spatially disjoint from their body. Similar to the RHI, the FBI can be induced by an experimenter who provides synchronous visual and tactile stimulation. The illusion is usually accomplished with the help of a camera positioned behind the participant that provides a live video feed of the stroking by the experimenter that is presented to the participant via a head mounted display (HMD). As is the case with the RHI, the illusion breaks if the visual information and tactile sensations are presented asynchronously and/or or in spatially incompatible locations. Recent studies have found that the FBI may also be self-generated through tactile self-stimulation as mediated by a robotic device (Hara et al., 2014
) or more simply through stroking their own neck (Swinkels et al., 2020
). Importantly, the functional mechanisms that are held responsible for the FBI, e.g. multisensory integration, are considered to play a central role in other bodily illusions such as the RHI as well (Ehrsson, 2012
; Metral et al., 2017
; Olive & Berthoz, 2012
). Hence, the outcomes of the current experiment should be relevant for the majority of bodily illusions that are considered to rely on similar mechanisms.
In the present study, we describe four experiments in which participants self-induced a FBI. Crucially, in one condition the FBI was evoked through movement; that is, participants repeatedly waved their dominant hand back and forth in parallel to the side of their neck, without touching (movement condition); whereas in the other condition, the same movement was made, while touching the neck (stroking condition). The fact that our two conditions only differ in synchronous information in just one sensory modality should enable us to disentangle the two principles in the present study. If body illusions reflect an all-or-nothing phenomenon, we should find that the illusion strength and onset of the FBI should be identical in the two conditions. If on the other hand body illusions are influenced by the availability of synchronous information in an additional sensory modality, we should expect the self-reported strength of the illusion to increase and or the onset time to shorten in the condition where participants are stroking, relative to waving. In the first experiment we explored the hypotheses under investigation by having participants induce the FBI using both self-generated movement and self-generated stroking in a 3PP. The second experiment was intended as a replication of Experiment 1. In the third1
experiment we tried to replicate the findings of Experiments 1 and 2 and additionally looked into the illusion onset times. Finally, in the fourth experiment we controlled for transfer effects between both methods of inducing a FBI and examined if self-generated movement would also be effective in inducing a FBI when presented independently from self-generated stroking. Hypotheses, sample sizes and planned analyses were preregistered for Experiments 2, 3 and 4 to create a clear timestamp of the decisions that were made before the experiments were conducted.2
The methods for Experiment 3 were very similar to the methods of Experiments 1 and 2. Only the differences will be described.
A total of 76 participants took part in the experiment (Mage
= 22.1, range = 18–28, 20 males, 1 unknown, 3 left-handed). To determine our sample size, we made use of the PANGEA app (v2.0) by Jake Westfall (jakewestfall.org/pangea/) which is suitable for calculating the power for designs that make use of linear mixed-effects models. Our design consisted of the factor participants (random) and the factor induction method with two levels (stroking, movement). We used an expected effect size d of 0.56 (based on Kalckert & Ehrsson, 2017
), two replicates (two measurements of onset time per induction method), 32 as our sample size and the default values for var(error) 0.5 and var(P*I) 0.167. For a justification of these default values see Westfall (2016
). The power calculation indicated that this should result in a power of 0.81. A cautious interpretation from the results of our second study suggested that only 35% of the participants reports the illusion for both methods (see also SOM3). For this reason, we ended up testing 76 participants in total to arrive at 32 participants for whom illusion onset can be measured for both induction methods.
Five participants were excluded due to misunderstandings of the questionnaire (e.g. scoring high on one or more of the illusion statements but reporting not to have had the experiences during the illusion check, see below).
Head-mounted display set-up
In addition to the regular illusion blocks, we added four practice blocks to familiarize participants with the experiences they may have during the illusion task. We had one practice block for each combination of induction method and video condition. The practice blocks preceded the test blocks of each induction method and were not followed by the illusion or control statements. The practice blocks were presented in the same counterbalanced order as the test blocks that followed.
In addition to the two statements that were administered in Experiment 3, participants completed an additional statement regarding the experienced location of their body in the current experiment (see Table 1
). This statement was chosen as an alternative to S3 in Experiment 1 because we wanted to have an additional test of the degree to which participants identified with the virtual body and this alternative item had been used successfully in previous research by other groups (Debarba et al., 2015
; Galvan Debarba et al., 2017
; Kokkinara & Slater, 2014
In addition to the three illusion statements we also added three control statements to control for the possibility that some participants may be inclined to answer affirmatively to any question on bodily experience (see Table 1
An illusion check was performed in which the scores on the illusion statements were inspected and the participants were asked about their experiences. The goal of this check was twofold: (1) the check was performed to determine whether participants’ interpretation of the experience corresponded with the way they used the illusion scale and answered the illusion statements and (2) to check which participants affirmed experiencing the illusion. The check was performed for each induction method separately after both the practice and the test blocks had been completed and the illusion statements had been rated. First the illusion scores were inspected. Next, in a semi-structured interview participants were asked to describe their experiences, to indicate whether they believed an illusion occurred and to indicate whether the illusion was stronger for the condition with a moving video image or the condition with a static video image. If they expressed an experience that did not correspond with the way they scored it on the illusion statements they were asked further questions to clarify the discrepancy. The following criteria were used to determine if a participant experienced the illusion: (1) participants had a higher average score on the illusion statements in the synchronous condition compared to the static control condition for this method and (2) the interview confirmed that the participant experienced the full-body illusion for this method. This resulted in a binary outcome variable for illusion onset which indicated for each participant whether the illusion was experienced for that induction method or not.
To gain insight into the temporal development of the illusion and potential differences between the two induction methods, the illusion onset time was measured. Illusion onset was only measured for participants who affirmed experiencing the illusion according to the criteria described under illusion check. To log the time of illusion onset, participants completed the synchronous condition again with the same induction method as in the test block they had just completed. They were instructed to signal the experimenter the moment at which they first started experiencing the illusion again. The experimenter then logged the time of illusion onset. This procedure was repeated once more for a better estimate of the illusion onset time (Kalckert & Ehrsson, 2017
; Metral et al., 2017
Two questionnaires were completed for exploratory purposes as part of a student project. Empathy was measured with the Toronto Empathy Questionnaire (Spreng et al., 2009
). Fantasy proneness was measured with the Creative Experiences Questionnaire (Merckelbach et al., 2001
). The questionnaires were administered at the end of the experiment and will not be further discussed.
In Experiment 3 the illusion tasks for each induction method consisted of (1) Two practice blocks, one for the synchronous and one for the static condition, (2) two test blocks, one for the synchronous and one for the static condition, each followed by the illusion and control statements, (3) an illusion check in which the scores on the illusion statements were inspected and the participants were asked about their experiences and (4) an illusion onset measurement (optional) that was only completed in case the participants met the criteria described under illusion onset.
Participants were instructed to take off the headset after completion of each block and to take some time to stretch their arm before moving on to the next block. This allowed a potential illusion to subside before the next block commenced, making sure that there were as little carry-over effects as possible. Instead of four minutes, each block only lasted three minutes in Experiment 3 to minimize the load for participants. This decision was based on previous research in our lab in which we found that it takes up to 96.3 ± 69.3 s on average to induce the illusion. This is still well under three minutes. The experiment ended with the exploratory questionnaires and a demographics questionnaire.
To analyse the illusion statements we first used separate linear mixed-effects models6
using the lmer function of the lme4 package (version 1.1.17; Bates et al., 2015
) in R (R Core Team, 2015
). Our model included a fixed intercept and a fixed effect for the factors induction method (stroking, moving), video condition (synchronous, static) and their interactions (all coded using sum-to-zero contrasts). The repeated measures nature of the data was modelled by including a per-participant random adjustment to the fixed intercept (“random intercept”). To determine p-values we computed Type 3 bootstrapped likelihood ratio tests (using 1000 simulations) as implemented in the mixed function of the package afex (Singmann et al., 2017
), which in turn calls the function PBmodcomp of the package pbkrtest (Halekoh & Højsgaard, 2014
). To explore potential suggestibility effects, the same model was used for the separate control statements. To further explore the potential suggestibility effects, we conducted an additional linear mixed-effects model. This model included the additional factor statement type (illusion, control) and its interactions with the factors induction method and video condition.
To analyse the illusion onset times, we made use of the same procedure as was used for the analysis of the statements. The model included a fixed intercept and a fixed effect for the factor induction method (stroking, moving; coded using sum-to-zero contrasts) and a per-participant random adjustment to the fixed intercept. The factor video condition was redundant as the onset times were only measured in the synchronous condition and was, therefore, left out of the model. The analysis was conducted on the 32 participants who reported the illusion for both induction methods.
Additionally, Bayesian paired samples t-tests were conducted using JASP (JASP Team, 2017
) on the difference scores for S1, S2 and S3 as described for Experiment 1 and on the average onset times of the two induction methods. We used a default Cauchy prior width of 0.707.
In the third experiment, we replicated the findings of Experiments 1 and 2. Again, we found that the illusion strength was equal for both induction methods. Furthermore, a similar effect was found in the group of participants who reliably reported the FBI with both induction methods. This latter result further corroborates the AoN hypothesis that body illusions operate as an all-or-nothing phenomenon whereby synchronous activity in an additional modality does not further enhance the strength of the illusion. More precisely, by selecting participants who experience the illusion in both respective conditions we can unambiguously show that inclusion of an additional synchronous modality on top of a confirmed illusion does not further deepen the illusion experience.
Importantly, the findings in Experiment 3 also showed that the onset times of the FBI were identical with both induction methods. This rules out the possibility as suggested by the SE hypothesis that adding an additional sensory modality will increase the speed at which evidence for an alternative common cause is collected such that a switch in self-location is experienced at an earlier time. Instead, the finding in this experiment suggests that not so much the amount of synchronous activation is what is driving the speed of onset of body illusions, but rather the length of time in which sensory and/or motor signals are found to be synchronised (Kokkinara & Slater, 2014
). The finding that the onset speed of the FBI is not influenced by the method of induction is furthermore corroborated by the analysis of the number of participants who experienced the illusion with either induction method. Analysis of the probability of the illusion indicated that there was no significant difference in the amount of participants who experienced a FBI as a consequence of self-movement or self-stroking in a 3PP. This latter finding suggests that both methods were similarly potent in inducing the FBI.
In Experiment 3 we also investigated the possibility that the FBI as measured with the questionnaire items could reflect enhanced suggestibility or social desirability of participants. To this end, three control items were included that inquired about body perceptions on which no effects were expected. The fact that we found a significant main effect of video condition on two of the control statements suggests that indeed some suggestibility or social desirability was present in our sample. However, a significant video condition * statement type interaction indicated that the effects on the FBI items were much larger than the effects on the control items, ruling out suggestibility as an explanation for the FBI as reported in the current study.
It does seem like we obtained a somewhat stronger illusion in Experiment 3 compared to the previous experiments. Due to the practice sessions, participants may have felt more confident in what they experienced and what not and may have scored their experiences higher than they would have done without the practice.
In the final experiment, we moved attention away from the comparison between the two induction methods and focussed more selectively on the method in which we used active self-generated movements to induce the FBI.
In the present paper, we investigated the basic functional mechanisms that are responsible for the construction of (illusory) body representations. More specifically, we asked the question whether the availability of synchronous information in an additional sensory modality would increase the strength of (illusory) embodiment or shorten the time at which (illusory) embodiment sets in. Two opposing perspectives were contrasted. The AoN perspective (Kalckert & Ehrsson, 2014
) suggests that the (illusory) embodiment functions as an all-or-nothing phenomenon and predicts that adding synchronous sensory information will not enhance the illusion strength or facilitate the onset of the illusion. Oppositely, the SE perspective (Samad et al., 2015
) proposes that illusory embodiment varies as a function of the available sensory evidence and predicts that the illusion strength will be enhanced and the time of onset will be shortened with additional synchronous input. In three experiments, we contrasted both perspectives by comparing FBI ratings between two illusion induction methods, one in which the FBI was induced by asking participants to stroke the side of their neck, and another method in which participants were asked to execute the same action while maintaining a 10 cm distance between their hand and neck. Results of the first three experiments unambiguously supported the AoN hypothesis and disconfirmed the SE hypothesis, by revealing that both induction methods were equally strong, equally fast and equally potent in inducing a FBI. In short, the inclusion of extra synchronous tactile information does not have any stimulating effect on the illusion. In Experiment 2, we furthermore confirmed that the FBI can be induced through the perception of simple movements in a 3PP, and that this outcome does not depend on transfer effects from the induction method that included touch.
The current findings provide clear support for the idea that adding synchronous information from an extra sensory modality does not further enhance the level or onset speed of the illusion. The findings in Experiments 1, 2, and 3 matched the predictions of the AoN hypothesis which states that synchrony between a minimum of two (sensory or motor) channels rather than the sum of synchronous information in multiple channels is what is driving (illusory) embodiment. Although the current results are straightforward when it comes to the ineffectiveness of including additional synchronous sensory evidence for illusory embodiment, it is still the question if this also implies that body illusions indeed function as an all-or-nothing phenomenon in the broader sense.
Importantly, and as outlined in the introduction, previous studies have compared ratings of body illusions that were induced by active movements, passive movements and static induction methods (Brugada-Ramentol et al., 2019
; Dummer et al., 2009
; Kalckert & Ehrsson, 2012
; Ma & Hommel, 2015
; Pyasik et al., 2019
; Riemer et al., 2013
). Some of these studies have found illusion ratings to increase when the illusion was induced via active movements by the participants (Dummer et al., 2009
; Ma & Hommel, 2015
). These findings clearly conflict with the AoN principle. As pointed out, actively induced body illusions are typically accompanied by a stronger sense of agency, which in turn could have influenced body ownership ratings. It is currently unclear whether agency is to be considered a natural factor in body ownership as has been suggested by (Ma & Hommel, 2015
) or if body ownership and agency represent two independent psychological functions (Kalckert & Ehrsson, 2014
) that could potentially interact or influence each other. Further research is necessary to disentangle the relationship between agency and body ownership and to determine if the strength of a body illusion (e.g. as reflected in proprioceptive drift, or the location at which participants report to feel their body) can be estimated independently from the sense of agency.
Furthermore, future studies should not only report condition averages but should also pay attention to (the distribution) of individual data points. It could, for instance, be the case that condition effects (e.g. active vs. static induction) are driven by individual participants who do not experience an illusion in the one condition (e.g. in the static condition) and do experience the illusion in another condition (e.g. the active condition). The consequence would be that on a group level it would appear as if the illusion increases in strength, whereas in reality, there is simply a larger number of participants who experience the illusion in the condition with the active induction. Indirect support for this suggestion is presented in SOM3, where we show that the distribution of FBI ratings follows a bimodal distribution which suggests that the average scores on the illusion statements reflect the middle ground between individuals who do experience and individuals who do not experience the illusion. This also explains the relatively low illusion ratings in our experiments. In sum, the current data indicate that additional synchronous evidence does not enhance the strength of body illusions which is consistent with the hypothesis that (illusory) embodiment functions as an all-or-nothing phenomenon. We do realize, however, that more research is necessary to clarify the inconsistent findings that have appeared in the literature. Part of the solution may be to pay closer attention to the data of individual participants and the distribution of data points that make up group averages.
A prerequisite for addressing the main question of the current study was that the self-movement (waving) condition would be effective in inducing a FBI. In line with previous studies that successfully managed to induce body illusions such as the RHI, the VHI, and full-body ownership illusions from a first-person perspective (1PP) using active movements by the participant (Debarba et al., 2015
; Galvan Debarba et al., 2017
; Gorisse et al., 2017
; Kalckert & Ehrsson, 2012
; Rognini et al., 2013
; Romano et al., 2015
; Sanchez-Vives et al., 2010
; Walsh et al., 2011
), the present study (to the best of our knowledge) is the first to demonstrate that a FBI can be induced by simply observing self-generated movements from a 3PP. We consistently found that the FBI was evoked by self-movement in three consecutive experiments. Results of the fourth experiment furthermore indicated that the self-movement FBI was effective on its own and does not depend on transfer effects (Hohwy & Paton, 2010
). This finding goes beyond recent studies that have reported that the FBI can be self-generated through tactile self-stimulation (Hara et al., 2014
; Swinkels et al., 2020
). More precisely, our findings demonstrate that touch is not a necessary element to induce the FBI, and that the illusion can be induced just as effectively through self-movement.
The current findings indicate that the perception of a self-generated movement in 3PP suffices to induce a FBI and that addition of a tactile component does not offer any advantage with regard to the strength of the body illusion, the speed of illusion onset or the likelihood that the illusion will be induced. These findings may be relevant for developers of virtual role-playing applications that aim to induce illusory embodiment of a virtual 3PP avatar and embodied presence in the virtual environment. We believe that our findings may easily translate to virtual reality applications considering that several body illusions have already been successfully induced in virtual reality (Kilteni et al., 2015
). Furthermore, motion capture and via-point animation of avatars may offer a more natural (Slater et al., 2009
) and feasible approach to induce virtual embodiment (Spanlang et al., 2014
) than continuous self-stimulation through stroking. Considering that self-movement appears to be similarly effective as self-stroking to induce the FBI, we recommend the former approach. Notably, this should not mean that the development of sensory feedback devices should be abandoned (Tactical Haptics, 2017
). Considering that action-effect feedback of goal-directed actions has been found to be effective in enhancing agency and illusory ownership (Choi et al., 2016
; Riemer et al., 2019
; Wen et al., 2016
), further development of sensory feedback devices supporting action-effects may be an efficient approach to enhance embodied presence.
The onset times that we obtained in Experiment 3 may furthermore be interesting to game developers and illusion researchers alike because they can be used to establish the minimally required duration of synchronous stroking or movement to establish a FBI. We demonstrated that participants need on average 33 s before they experience an illusion over the body that they see in front of them if a 3PP is used and that 95% of the participants experience the illusion within 95 s. However, it is important to note that these numbers come from individuals who have at least some experience with the illusion. Future research could investigate whether the onset times are similar in a truly virtual environment where gamers use their real body to control the movements of their avatar in a more natural and less repetitive manner. Kalckert and Ehrsson (2017
) found that participansts were a bit faster to signal the RHI in an active induction condition (21 s) than with passive induction (24 s). In their study 95% of the participants felt the illusion within the first minute.
Although the present study contributed several new insights in the nature of illusory embodiment, several limitations may be noted. First, there is a remote possibility that tactile information was covertly activated in the movement condition. Although participants did not touch their necks in the movement condition and waved their hand up and down at a distance of approximately 10 cm from the neck, it is known that visuotactile neurons in parietal and premotor regions of the cortex respond to objects that loom in peripersonal space (Graziano et al., 1997
). It should be noted however that these bimodal neurons are predominantly activated by objects that move towards the body, and much less so or not by objects that move away from the body (Canzoneri et al., 2012
; Clery et al., 2015
; Graziano & Cooke, 2006
; Kandula et al., 2015
), or that are on a trajectory that is not likely to impact the body (Huijsmans et al., 2020
). Considering that the movements in the waving condition were parallel to the body it is questionable if these movements triggered covert tactile activations in this system.
A second limitation of the current study is that tactile stimulation in the stroking condition may be considered as an action effect that is known to enhance agency (Choi et al., 2016
; Riemer et al., 2019
; Wen et al., 2016
) and that could strengthen the induced body illusion. However, the repetitive stroking movement was intransitive and no discrete goal or effect was obtained in executing the stroking. As such it is unlikely that the stroking condition generated more agency than the movement condition. This interpretation is furthermore supported by the fact that no significant differences were found in the strength of the FBI between conditions.
A third limitation is that no measure of agency was included in the current experiments. Previous research suggested that the experience of agency might enhance the strength of the ownership illusion (Ma & Hommel, 2015
; Ma et al., 2017
). In the design of the current study we controlled for potential differences in agency between conditions by comparing two methods that both employed active movements to induce the illusion. Although we cannot exclude the possibility that differences in agency may have existed between conditions, we reckon this to be unlikely considering the strong similarity in illusion strength between conditions and the matched design.
A final limitation is that the dependent measures in this study were mostly subjective, consisting of ratings on statements that captured the essence of the FBI and self-reported onset times of the FBI. Consequently, it could be argued that participants’ answers may not have objectively captured if they experienced the illusion and to what degree. It has to be noted though that the brief interviews after Experiment 2 provide us with some clues about who experienced the illusion and who did not and Experiment 3 even included extensive systematic interviews to identify participants who experienced the illusion with both induction methods. Moreover, we included several control questions in Experiment 3 to rule out the possibility that participants answered in a socially desirable manner about their experience of the FBI. Finally, measures about the time of illusion onset were repeated twice to increase their reliability.
In conclusion, the current study consistently found over multiple experiments that added synchronous sensory information does not increase the strength of the FBI, its speed of onset, or the probability for the illusion to occur. These findings are in line with the all-or-nothing principle that has been proposed to underlie (illusory) embodiment. Further research is necessary, however, to disentangle the interactions between self-reported agency and body illusion strength, and to validate the all-or-nothing principle in various conditions of illusory and impaired embodiment.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.