The ability to differentiate one’s own body from others’ is commonly thought to rely on continuous body representations (Gallagher 2000
; Tsakiris, 2008
; Lenggenhager et al., 2007
), which, however, can be updated and adjusted to the present situation (e.g., Graziano & Botvinick, 2002
). For example, the rubber hand illusion (RHI) shows that people perceive ownership for a rubber hand lying in front of them if the rubber hand and their real hand are stroked synchronously (Botvinick & Cohen, 1998
). Similarly, the virtual hand illusion (VHI) demonstrates that people perceive ownership for a virtual hand if they can operate its movements by moving their own hand (Slater et al., 2008
). Even more important for our present purposes, people tend to perceive the face of another human as their own if it is stroked in synchrony with their own face—the enfacement illusion (Tsakiris, 2008
; Porciello, Bufalari, Minio-Paluello, Di Pace & Aglioti, in press).
Research suggests that embodying the body or face of another person does not only tend to diminish self–other discrimination, but also to invite what Ma, Sellaro, Lippelt, and Hommel (2016
) called “feature migration”. This hypothetical process is derived from recent attempts to apply the Theory of Event Coding (TEC: Hommel, Müsseler, Aschersleben & Prinz, 2001
) to the representation of self and others (Hommel, Colzato & van den Wildenberg, 2009
). The idea is that people represent themselves and others just like other perceptual events, namely, in terms of integrated networks of sensorimotor feature codes (event files: Hommel, 2004
) representing all discriminable features an event or person consists of, such as physical attributes, affective responses, control states, and covert and overt actions associated with a given event. Importantly, given that feature codes are integrated and bound together, they tend to be retrieved as a whole when one of the features of a given event is encountered. Moreover, the activation of feature codes is regulated by an “intentional weighting” process that gives more weight to attended or task-relevant features (Memelink & Hommel, 2013
). Accordingly, focusing on common or discriminating features can reduce or increase the degree to which self and other are perceived as different, respectively (Colzato, de Bruijn, & Hommel, 2012
). Importantly for present purposes, to the degree that the representation of self and other is perceived as part of the same event, features of one’s representation can “migrate” to the representation of the other (Treisman & Gelade, 1980
)—i.e., features of one event can become part of an “illusory conjunction” with the other.
Indeed, Ma et al. (2016
) observed that controlling the movements of a virtual face by moving one’s own lead to the migration of mood: participants were significantly happier and performed better in a mood-sensitive brainstorming task when controlling a smiling compared to a neutral face, but only if virtual and actual face moved in synchrony. In terms of TEC, experiencing a virtual happy face as being part of oneself caused participants to confuse their own features and states with the features and states of the virtual face to the extent that affective features of the virtual face (i.e., the smile) became assimilated with participants’ self-representation. Other studies suggest that embodying another person’s hand, face, or body affects people’s attitudes towards the other or the other’s social group (for a review, see Maister, Slater, Sanchez-Vives & Tsakiris, 2015
). For example, occupying a virtual child body facilitates combining the self with child-like attributes in an implicit-association test (Banakou, Groten, & Slater, 2013
), owning a dark-skinned rubber hand or a black avatar reduces implicit racial bias of light-skinned for dark-skinned people (Maister, Sebanz, Knoblich, & Tsakiris, 2013
; Farmer, Maister, & Tsakiris, 2013
; Peck, Seinfeld, Aglioti, & Slater, 2013
; but see Estudillo & Bindemann, 2016
), embodying avatars of old people, compared to young people, reduces negative stereotyping of the elderly (Yee & Bailenson, 2006
), and placing participants in avatars with a superhero ability promotes helping behavior (Rosenberg, Baughman, & Bailenson, 2013
). These observations suggest that feature migration can change both attitudes and behavior, which fits with the idea that feature codes are sensorimotor in nature, in the sense that they are activated by sensory information (i.e., perception) and can regulate both perception and overt/covert behavior (Hommel et al., 2001
So far, the impact of the embodiment of others was almost exclusively restricted to human bodies or body parts. Assuming a permanent body representation, this makes sense: candidate effectors should be accepted as possible body parts only to the degree that they are similar to parts that this representation entails (Tsakiris, Carpenter, James, & Fotopoulou, 2010
). However, there is growing evidence that perceived body ownership may be even more flexible than hitherto assumed. For instance, recent studies have shown that people can experience ownership over avatars that are shaped differently from them, including not only avatars of different gender (Slater et al., 2010
), race (Maister, Sebanz, Knoblich, & Tsakiris, 2013
; Farmer, Maister, & Tsakiris, 2013
; Peck, Seinfeld, Aglioti, & Slater, 2013
; Bufalari et al., 2014
) and age (Oh et al., 2016
; Banakou, Groten, & Slater, 2013
; Hershfield et al., 2011
; Yee & Bailenson, 2006
), but also avatars with very long arms (Kilteni et al., 2012
), with abnormally large and small bodies (van der Hoort, Guterstam, & Ehrsson, 2011
; see also Piryankova et al., 2014
), with tails (Steptoe, Steed & Slater, 2013
), and with amputated body parts (Kilteni et al., 2016
). Going a step further, we recently demonstrated ownership for visual objects without any obvious similarity to body parts, such as balloons and rectangles (Ma & Hommel, 2015a
Taken together, the aforementioned findings make it plausible to expect illusions of this sort to be also demonstrated across species. To the best of our knowledge, only one recent study has addressed this issue (Ahn et al., 2016
). Ahn et al. used immersive virtual reality to make participants viscerally experience the life through the eyes of a cow or coral in an acidifying reef. When embodying a cow, participants walked in a virtual pasture on all fours, while eating feed and drinking water, they were goaded with a virtual cattle prod and finally they were loaded onto a truck; when embodying a coral on a rocky reef, participants were let to experience coral suffering from ocean acidification by seeing, hearing and feeling the reef around them as well as their own body corrode. Results showed that participants embodying the cow or the piece of coral, compared to those who simply watched a video of these experiences, felt more connected with the nature and had more concerns about the environment.
In the present study, we were interested to see whether embodiment across species can be elicited also by simply letting participants control the movement of a virtual face of a member of another species. More importantly, we aimed to assess whether enfacing members of another species would induce self–other assimilation as observed for the embodiment of human body parts. To test that possibility, we designed an ape face, which is clearly discriminable from a human face but still keeps some degree of resemblance. To induce identification with this face we used a dynamic enfacement paradigm (see Ma et al., 2016
), in which participants could move the virtual face either synchronously or asynchronously.
Sforza et al. (2010
) reported findings showing that questionnaire ratings for enfacement illusions tend to be low on average, compared to ratings obtained with the RHI and embodiment illusions. This suggests that the subjective experience of enfacement can be harder to obtain for faces than for other body parts, probably because facial identity is at the core of the sense of the self (Tsakiris, 2008
). Being confronted with the face of a member of another species is not unlikely to cause feelings of strangeness and irritation, similar to the ones reported during virtual body illusions (Lenggenhager et al., 2007
), which might work against enfacement. Therefore, to reduce the risk of incurring in such irritations and to maximize the virtual enfacement illusion, we used a morphing procedure in which a synchronously- or asynchronously-moving virtual human face slowly morphed into an ape face. Note that we applied this technique to both synchrony conditions, so that possible effects of synchrony would be corrected for possible effects on the morphing procedure itself.
Building on earlier findings, we expected a synchronously-moving virtual face to increase perceived ownership, as measured by a standard enfacement questionnaire, and increase self–ape similarity. Crucially, we hypothesized that ownership, in turn, should facilitate feature migration between the representations of self and ape. Importantly, according to TEC logic, feature migration of this sort can involve all kind of features and can occur from the representation of other to the representation of oneself, and vice versa.
We tested two features that we considered particularly diagnostic. First, we had participants to undergo the Standard Progressive Matrices (SPM) test (Raven, 1938
), which assesses fluid intelligence. Previous studies have shown that people may conform to the expected attitudes and behavior of the avatar they are represented by, without necessarily being aware of that (i.e., the Proteus effect; e.g., Yee & Bailenson, 2007
; Yee, Bailenson, & Ducheneaut, 2009
; Peña, Hancock, & Merola, 2009
). For instance, participants have been found to behave more confidentially and friendlier with strangers or to behave more aggressively during a negotiation when controlling attractive and tall avatars, respectively (Yee & Bailenson, 2007
; see also Yee, Bailenson, & Ducheneaut, 2009
). Similarly, it has been found that participants represented by avatars dressed in Ku Klux Klan outfits reported more negative and aggressive thoughts than those dressed as doctors (Peña et al., 2009
). Taken together, these findings are consistent with previous literature showing that trait concepts and stereotypes become active automatically in the presence of relevant behavior or stereotyped-group features (i.e., through perception), and can cause people to behave consistently with the activated stereotypes (Chartrand & Bargh, 1999
). Therefore, considering the common stereotype that humans are more intelligent than non-human animals, we predicted that enfacing and, thus, in some sense becoming an ape should reduce intelligence and result in a lower score in the SPM test. Note that we relied on, but did not directly test this stereotype in our participants because the assessment would have primed the respective stereotype (cf., Memelink & Hommel, 2013
), thus introducing a confound and rendering our test less diagnostic.
Second, we tested participants by means of an emotion rating task assessing to which degree participants would attribute particular emotions to humans and apes (Demoulin et al., 2004
). As not all emotions are assumed to be shared among humans and other animals, we expected that more emotions would be attributed to humans than to apes, but that this difference would be reduced or eliminated after enfacing an ape. Indeed, there is evidence that embodiment illusions can affect people’s perception of and beliefs towards the embodied other, and that positive self-like associations can be extended to the embodied others (Maister et al., 2015
). Therefore, given that people are biased to maintain a positive self-image (Tajfel & Turner, 1986
), increased self–ape similarity should cause participants to attribute a higher ability to experience emotions to apes. In contrast, attribution of emotions to humans is not expected to be affected by the enfacement manipulation. Note that possible synchronicity-induced changes in the attribution of emotions to apes may be restricted to emotions that are considered to be not uniquely human (i.e., primary emotions; Epstein, 1984
), or may concern primary and secondary emotions alike. Whereas the former case would indicate that participants have judged apes from their human perspective, the latter case would indicate that they have judged apes from the perspective of an ape feeling like a human.
This study set out to test whether embodiment can be demonstrated across species and whether this would induce self–other assimilation (and feature migration) in terms of intelligence and emotion attribution—as implied by our application of TEC (Hommel et al., 2001
). The first question can be answered affirmatively: when the ape face moved in synchrony with the participants’ own face, they were more likely to perceive ownership for the former. It is true that ownership perception was not perfect, as the score fell into the midrange of the scale—a finding that is consistent with other studies of the virtual-hand (e.g., Ma & Hommel, 2015a
) and the virtual-face illusion (Ma et al., 2016
; Tajadura-Jiménez et al., 2012
; Sforza et al., 2010
). This limitation notwithstanding, the fact that a short practice over 3 min was sufficient to affect one’s identification with another species—which most participants were likely to have little experience with—must be considered notable. Converging evidence supporting the conclusion that our participants identified with the virtual ape comes from the IOS ratings, which confirm that synchrony increased the perceived overlap between participant and ape. Hence, humans are able to identify with a member of another species, probably in a similar way as they can embody the hand or face of another human. This result is not unanticipated as it fits with previous claims that people’s self-construal is dynamic and sensitive to situational and cultural biases (Colzato et al., 2012
; Kühnen & Oyserman, 2002
), and that people are rather flexible regarding which objects and events they consider as being part of their body—as long as they can control the behavior of these objects and events (Ahn et al., 2016
; Kilteni et al., 2016
; Ma & Hommel, 2015a
; Piryankova et al., 2014
; Steptoe et al., 2013
; Kilteni et al., 2012
; van der Hoort et al., 2011
). It is worth noting that, in the present study, to maximize the chance of eliciting a virtual enfacement illusion of an ape face, we made use of a morphing procedure in which a synchronously or asynchronously moving virtual human face slowly morphed into an ape face. As mentioned in the Introduction, such a choice was aimed at counteracting possible feelings of irritation and strangeness that could have arisen by confronting participants directly with the face of a member of another species. It remains to be established whether the morphing procedure was really necessary for the illusion to occur, or whether virtual enfacement of a member of a different species can also be obtained without such a procedure.
The second most important question can also be answered affirmatively: both the intelligence measure and the emotion attributions were affected by synchrony. As predicted from our TEC-based approach, perceiving oneself to own an ape face made people behave less intelligently; that is, participants tended to adopt the intellectual characteristics humans attribute to the species they perceived themselves to become a part of, thereby confirming previous findings documenting people’s tendency to adopt attitudes and behavior of the avatar they are represented by (i.e., the Proteus effect; e.g., Yee & Bailenson, 2007
; Yee, Bailenson, & Ducheneaut, 2009
; Peña, Hancock, & Merola, 2009
). We consider this a demonstration of feature migration in the sense of Ma et al. (2016
): increasing the overlap of self- and other-representation invites illusionary conjunctions, in which features of the other become features of oneself. Moreover, enfacing an ape tempted participants to attribute more emotions to apes, another case of feature migration: when becoming more like an ape, participants took their emotional capabilities with them, so to speak. Note that synchrony affected primary and secondary emotions alike. While primary emotions are considered to be shared among humans and other highly evolved animals, secondary emotions are commonly thought to be uniquely human (Epstein, 1984
). If our human participants would have judged apes from their human perspective, one might have expected that synchrony has a stronger impact on primary emotions attributed to apes than on secondary emotions. However, given that synchrony enhanced the attribution of both kinds of emotions, it seems that the emotion attribution in this condition relied on a more “insider perspective”, that is, from the perspective of an ape feeling like a human. In other words, when embodying another person (or a member of another species), self-related attributes can be extended to the embodied others who become more like the self. This fits with the observation that synchrony had no effect on the attribution of emotions to humans, which also rules out the possibility that the enfacement experience facilitated emotion attribution in general.
In the present study, we were interested to see whether feature migration can lead to the integration of features of another individual into the representation of oneself. However, we would like to emphasize that, theoretically speaking, features would be expected to migrate both ways: from other to self and from self to other. So, while our present study was tapping into the route from other to self, some recent studies have provided evidence that features may migrate from self to other as well. For instance, body-ownership illusions have been found to induce a more positive attitude toward social out-groups (e.g., Maister et al., 2015
), which from our theoretical view might suggest that a positive self-image can migrate to an embodied other. Indeed, considering that attributing emotions to outgroup members represents an attitude, our findings are consistent with, and can be seen as an extension of previous outgroup studies along the lines of Maister et al. (2015
). As already mentioned, synchronous stimulation of participants’ own hand and a rubber hand typical for a racial outgroup member increased perceived ownership for the rubber hand and induced a more positive attitude toward that outgroup (Farmer et al., 2013
). Similarly, Inzlicht, Gutsell, and Legault (2012
) reported that negative attitudes toward a racial outgroup were reduced by synchronizing one’s own movements with those of an outgroup member. It is true that such a modulation of implicit racial attitudes has not been observed in a recent study using a static enfacement paradigm (Estudillo & Bindemann, 2016
). This suggests that embodiment illusions are not always effective in biasing self- and/or other-perception, especially when concerning body parts that are strongly tied to the self-identity, just like faces. However, recent comparisons between dynamic and static hand illusion conditions have revealed that dynamic conditions, as used in the present study, are much more sensitive to manipulations and strongly increase the coherence between dependent measures (Ma & Hommel, 2015a
). It is thus possible that dynamic conditions, perhaps together with the morphing technique we used in the present study, will be more successful in demonstrating modifications of racial attitudes.
Note that in the present study we relied on explicit measures but did not use implicit measures of ownership, a decision we made for two reasons. First, explicit and implicit measures of body ownership have often been demonstrated to dissociate (e.g., Liepelt, Dolk & Hommel, 2017
; Ma & Hommel, 2015a
). This implies that explicit and implicit measures rely on different kinds of information, and so far, no theoretical account for such dissociations has been suggested. This renders it unclear what kind of information implicit measures would add and which theoretical implications the convergence or divergence with explicit measures would have. Second, the inclusion of implicit measures was unlikely to be successful in our study. Previous face-ownership studies have used self–other discrimination or recognition of self–other morphed faces (e.g., Sforza et al., 2010
; Tsakiris, 2008
; Tajadura-Jiménez et al., 2012
) as an implicit measure. The typical outcome was the tendency of participants to under-discriminate between pictures of oneself and of a very similar human other in the synchrony, compared to the asynchrony condition. This bias is commonly very small (around 5%), which renders it extremely unlikely to get anything measurable when comparing oneself against the picture of an ape.
Another important consideration pertains to the fact we did not assess participants’ stereotypes towards apes to verify whether and to what degree the common stereotype that humans are more intelligent than nonhuman animals was shared by our participants. As we pointed out, explicitly assessing this stereotype would have been likely to introduce unwelcome confounds. And yet, it would be interesting to test whether individual differences in the degree to which participants consider apes as less intelligent would predict the sizes of these synchrony-induced changes in intelligence and emotional competence judgments. Future studies might either assess these stereotypes in separate sessions and at considerable temporal distance and/or assess them in a more indirect fashion. Moreover, it would be interesting to see whether interventions of the sort investigated in the present study would lead to longer lasting changes in individual stereotypes—as suggested by the observations of Farmer et al. (2013
) and Inzlicht et al. (2012
). Notwithstanding these interesting open questions, our findings provide further evidence that the boundaries between perceived self and perceived other are rather flexible, and that representational self–other overlap invites illusory conjunctions of features from one representation to the other—including others of another species.