Introduction
People are thought to store representations of their own body, so they can recognize themselves and discriminate themselves from others (Gallagher,
2000; Jeannerod,
2003). Popular methods to investigate how people represent their own and other bodies are paradigms producing the rubber hand illusion (RHI; Botvinick & Cohen,
1998) and the virtual hand illusion (VHI; Slater, Perez-Marcos, Ehrsson, & Sanchez-Vives,
2008; Sanchez-Vives et al.,
2010). In the original RHI paradigm, participants are facing a rubber hand lying beside/on top of their hidden real hand. If real and rubber hand are stroked synchronously (compared to a condition in which stroking is asynchronous), participants tend to attribute the felt stroking of their real hand to the rubber hand, judge the positions of their real hand as drifting towards the rubber hand and misperceive the rubber hand as their own. In the original VHI paradigm, participants are wearing a data glove on their real hand, which operates a virtual hand on a screen or virtual space in front of them. If virtual and real hand move in synchrony (compared to moving asynchronously), participants tend to perceive a sense of controlling the virtual hand, and misperceive the virtual hand as part of their own body. Hence, tight correlations between multisensory stimuli applied to, or produced by one’s own effector and an artificial effector close to one’s body seem sufficient to incorporate an artificial effector into one’s body representation.
The importance of such bottom-up factors notwithstanding, the aim of the present study was to test whether the ease or degree of incorporating an artificial effector into one’s own body representation depends on cognitive control. Two basic assumptions motivated our approach. First, we assumed that people represent different parts of their body the same way as they represent body-unrelated objects. This assumption is based on the Theory of Event Coding (TEC, Hommel, Müsseler, Aschersleben, & Prinz,
2001; Hommel,
2004), according to which the cognitive system represents both perceived and produced events by integrated networks of codes of the features of these events. Importantly, TEC does not discriminate between social and non-social or self-related and self-unrelated events (Hommel, Colzato, & van den Wildenberg,
2009), and it allows for various degrees of event integration, including the integration of self and other (Colzato, Zech, Hommel, Verdonschot, van den Wildenberg, Hsieh,
2012). Second, we assumed that the degree of event integration depends on the present metacontrol state. According to the metacontrol state model (MSM; Hommel,
2015), this state varies between the poles of extreme persistence (a state characterized by a strong top-down influence of the current action goal, high selectivity, and strong mutual competition between alternative representations, e.g., in decision-making) and extreme flexibility (a state characterized by weak top-down influence of the action goal, strong integration, and weak mutual competition). A bias towards persistence would be associated with the tendency to discriminate between given events whereas a bias towards flexibility would be associated with the tendency to integrate them. Importantly, this holds for both non-social and social events, including people and body parts (cf., Hommel & Colzato,
2017).
Applying these two assumptions to VHI, the paradigm we were using in the present study, motivates the prediction that a bias towards persistence would reduce, and a bias towards flexibility would increase the VHI: given that the illusion implies the integration of the representation of an artificial effector and the representation of one’s real body, a less integrative metacontrol state should indeed work against the illusion while a more integrative state should support it. As a method to bias participants towards persistence or flexibility, we used two kinds of creativity tasks: the Remote Associates Task (RAT; Mednick,
1962) and the Alternate Uses Task (AUT; Guilford,
1967). Both tasks call for some degree of top-down control (as they require a particular problem to be solved) and search (for the solution of this problem), and thus arguably require some degree of persistence and some degree of flexibility. And yet, the RAT requires the highly constrained search for one possible solution (convergent thinking in the terminology of Guilford,
1967) while the AUT requires a loosely constrained search for as many solutions as possible (divergent thinking)—which implies that the RAT relies more on persistence and less on flexibility than the AUT (Hommel,
2012).
Assuming that this leads people to establish a more persistence-heavy metacontrol state when working on the RAT and a more flexibility-heavy state when working on the AUT, studies have used these tasks to prime metacontrol towards persistence and flexibility, respectively, and to test whether the latter leads to more integration than the former. For instance, Colzato, van den Wildenberg, and Hommel (
2013) found evidence of more self–other integration in a joint Simon task when it was interleaved with an AUT than when it was interleaved with an RAT. Along the same lines, Sellaro, Hommel, de Kwaadsteniet, van de Groep, and Colzato (
2014) found more interpersonal trust when participants just completed an AUT than when they just completed an RAT. As these observations confirm the assumption that RAT and AUT are effective in biasing people’s metacontrol state towards persistence and flexibility, respectively, we used these two tasks as metacontrol primes in the present study as well. Hence, we not only assumed that performing the RAT and the AUT would lead participants to establish metacontrol states biased towards persistence and flexibility, respectively, but we also assumed that these states would be sufficiently inert to affect the following VHI induction (i.e., the synchrony manipulation) accordingly. To make sure that the induced metacontrol state would still be sufficiently strong during this induction, we interleaved creativity task and (a)synchrony presentation in such a way that participants switched repeatedly between the two.
Taken altogether, we thus predicted that the VHI would be more pronounced if carried out in the context of an AUT than if carried out in the context of an RAT. We further considered the possibility that being exposed to synchrony vs. asynchrony between one’s own and an artificial hand might also have an impact on metacontrol, which in turn might influence performance in the two creativity tasks. In particular, if being exposed to synchrony would bias metacontrol towards flexibility (i.e., integration), it is possible that this would be more beneficial for performance on the AUT than for performance on the RAT, while the opposite would hold for exposure to asynchrony. To test these possibilities, we thus looked into both the impact of engaging in a particular creativity task on VHI (as measured by comparing synchronous with asynchronous conditions) and the impact of experiencing synchronous and asynchronous conditions on performance in the creativity tasks. Our main measure referred to the subjective perception of ownership and agency (Botvinick & Cohen,
1998), but we also included an implicit measure (perceptual drift rates; Kalckert & Ehrsson,
2012) for explorative purposes.
Discussion
The findings are clear-cut. As in previous studies, participants were more likely to experience subjective agency and ownership for a virtual hand if it moved in synchrony with their own, real hand. As predicted, the size of this effect was significantly moderated by the type of creativity task in the context of which the illusion was induced. In the context of the RAT, which we assume to bias metacontrol towards persistence (and thus increase top-down control and event discrimination), the illusion was markedly smaller than in the AUT context, which we assume to bias metacontrol towards flexibility and integration. This suggests that the perception of body ownership in general, and the integration of candidate effectors in particular, are moderated by metacontrol. More specifically, persistence seems to support self–other segregation while flexibility promotes self–other integration. Given that persistence and flexibility was induced by means of tasks without any obvious social or self-related implications, this finding is consistent with TECs assumption that the cognitive representations of body parts are no different from cognitive representations of body-unrelated events and objects. The finding also fits with predictions from MSM, especially with respect to the impact of metacontrol on event integration.
It is interesting to note that the explicit measure of ownership, as assessed by the questionnaire, was sensitive to the metacontrol manipulation while the implicit measure, as assessed by the perceptual drift, was not. Like many previous dissociations between explicit and implicit measures of ownership (Holmes, Snijders, & Spence,
2006), this observation provides converging evidence that the two measures do not rely on exactly the same information. It may be that conscious self-perception of ownership is depending on more, or more integrated information than proprioceptive drift rates, which may rely more on multisensory discrepancy (Rohde, Di Luca, & Ernst,
2011).
The creativity tasks did not only serve to induce metacontrol-biases but they were also affected by the synchrony manipulation, which was also thought to be able to induce different metacontrol states. The outcome pattern seems to suggest that perceiving asynchrony between events promotes convergent, but impairs divergent thinking, while perceiving synchrony has little effect. The reason may be the relatively low illusion strength: the ownership ratings for synchronous conditions were not very high, and the ownership ratings for asynchronous conditions were quite low. This might suggest that synchrony did not induce strong flexibility but asynchrony induced strong persistence. If so, the asynchrony conditions would indeed be expected to improve performance in tasks that rely on persistence, such as the RAT, and to impair performance in tasks that rely on flexibility, such as the AUT.
It is important to keep in mind the fact that our present findings were obtained in a paradigm that strongly interleaved what we considered the task prime (i.e., the particular creativity task) and the induction of the VHI—the process we aimed to prime. The practical reason to do so was to increase the probability that the metacontrol state that the creativity tasks were hypothesized to induce or establish would be sufficiently close in time to the synchrony manipulation to have an impact on the thereby induced changes in self-perception. However, this implies that we are unable to disentangle the effects of the task prime proper and the effects of possible interactions between this task prime and the synchrony manipulation. There are indeed reasons to assume that such interactions are not unlikely to have occurred and that they would make perfect theoretical sense. The observation that the VHI was affected by the type of creativity task and performance in the creativity tasks was affected by the synchrony manipulation suggests some degree of overlap between the ways that engaging in particular creativity tasks and experiencing particular degrees of synchrony are able to bias perceived ownership and agency. In terms of our theoretical framework, this implies that engaging in divergent thinking biases metacontrol towards flexibility in similar ways as experiencing synchrony between one’s own movements and those of a virtual effector does, while engaging in convergent thinking biases metacontrol towards persistence as experiencing asynchrony does. What the present findings demonstrate is that both kinds of manipulation together bias the VHI in the predicted direction, but they do not allow to statistically or numerically separate and estimate the contribution that each of the two confounded manipulations might have made. Accordingly, the present findings should not be taken to provide conclusive evidence that priming tasks alone are able to change self-perception without being supported (and perhaps even enabled) by the experience of synchrony between proprioceptive and visual action feedback.