Self beyond the body: task-relevant distal cues modulate performance and body ownership

The understanding of Body Ownership (BO) largely relies on the Rubber Hand Illusion (RHI) where synchronous stroking of real and Rubber Hands (RH) leads to an illusion of ownership of RH provided physical, anatomical, postural and spatial plausibility of the two body-parts. RHI also occurs during visuomotor synchrony, in particular, when the visual feedback of virtual arm movements follows the trajectory of the instantiated motor command. Hence BO seems to result from a bottom-up integration of afferent and efferent proximal multisensory evidence, and top-down prediction of both externally and self-generated signals, which occurs when the predictions about upcoming sensory signals are accurate. In motor control, the differential processing of predicted and actual sensory consequences of self-generated actions is addressed by, the so-called, Forward Model (FM). Based on an efference copy or corollary discharge, FM issues predictions about the sensory consequences of motor commands and compares them with the actual outcome. The discrepancies (Sensory Prediction Errors, SPEs) are used to correct the action on the consecutive trial and provide new estimates of the current state of the body and the environment. Here, we propose that BO might be computed by FMs, and therefore, it might depend on their consistency, specifically, in contexts where the sensory feedback is self-generated. Crucially, to reduce SPE, FMs integrate both proximal (proprioceptive) and distal (vision, audition) sensory cues relevant to the task. Thus, if BO depends on the consistency of FMs, it would be compromised by the incongruency of not only proximal but also distal cues. To test our hypothesis, we devised an embodied VR-based task where action outcomes were signaled by distinct auditory cues. By manipulating the cues with respect to their spatiotemporal congruency and valence, we show that distal feedback which violates predictions about action outcomes compromises both BO and performance. These results demonstrate that BO is influenced by not only efferent and afferent cues which pertain to the body itself but also those arising outside of the body and suggest that in goal-oriented tasks BO might result from a computation of FM.

Humans and other species simultaneously acquire and integrate both self-generated 2 (reafferent) and externally-generated (exafferent) information through different sensory 3 channels [1]. Imagine performing a goal-oriented task such as an Air Hockey where the 4 objective is to score points by hitting a puck into the goal. During the task, the brain 5 continuously stores and updates both (1) action-driven proximal (ie. proprioceptive) 6 and distal (ie. visual or auditory) sensory cues relevant to the task such as the 7 kinematics of the arm or the location of the puck [2,3], and (2) action-independent 8 signals driven by external contingencies such as wind [4]. Thus the ability of the 9 nervous system to determine the source of a given sensation and the boundaries of an 10 embodied self, commonly referred to as the sense of Body Ownership (BO) [5,6], is 11 fundamental in adaptive goal-oriented behavior [7,8]. However, mechanisms driving BO 12 in action contexts, which require manipulation of the environment and therefore 13 integration of distal cues (ie. Air Hockey) remain elusive. 14 Our understanding of BO largely relies on, the so-called, Rubber Hand Illusion 15 (RHI) paradigm where the participants experience the illusion of ownership towards a 16 Rubber Hand (RH) during externally-generated synchronous, but not asynchronous, 17 stroking of both real and artificial hands [5]. The illusion generalizes to distinct 18 body-parts including fingers, face or a full body [9][10][11]. Interestingly, further studies 19 show that while BO strongly depends on the spatiotemporal alignment of the 20 exteroceptive inputs, it also requires physical, anatomical, postural and spatial 21 plausibility of the two body-parts [3,[12][13][14]. Hence, this empirical framework proposes 22 the notion that BO is a sensory state which does not require motor commands and 23 relies on two complementary processes. In particular, (1) a bottom-up accumulation 24 and integration of afferent, proximal, multisensory (proprioceptive, tactile) evidence, 25 and (2) top-down comparison between the novel sensory stimuli (RH) and self-specific 26 priors based on the internal model of body representation [12,15,16]. 27 Interestingly, a body of recent studies extends the traditional view on the 28 mechanisms driving BO by demonstrating that it does not exclusively stem from the 29 integration of afferent sensory signals. Specifically, it has been shown that BO can also 30 emerge in a top-down manner as (1) a consequence of pure expectation of correlated 31 exafference and (2) an accurate prediction of sensory consequences following 32 self-generated movement. The former finding comes from the study of Ferri and 33 colleagues [17], were subjects experience BO when observing the experimenter's hand 34 approaching the RH and anticipating the sensory event in the absence of actual touch. 35 The latter, in turn, is supported by the results from Virtual Reality (VR)-based 36 experiments investigating the effects of visuomotor (a)synchrony on BO during 37 goal-oriented behavior [18][19][20]. Here, the authors report high BO in the condition where 38 movements of the real and virtual hands are spatiotemporally congruent. In particular, 39 when the actual sensory feedback (visual and proprioceptive) of the motor commands 40 matches the expected one. This evidence further highlights the crucial role of the 41 prediction-driven top-down processes in the modulation of BO [16,21,22]. 42 In the contemporary theoretical framework of motor control, the differential 43 processing of the predicted and the actual sensory consequences of self-generated actions 44 is addressed by, the so-called, internal Forward Model (FM) [23][24][25] which issues 45 predictions about sensory consequences of current motor commands. According to this 46 interpretation, in the Air Hockey task at every trial, the motor system prepares and 47 generates those commands which are most likely to cause the desired outcome. 48

2/15
Simultaneously, an Efference Copy (EC) [26] or a Corollary Discharge (CD) [27,28] of 49 these motor signals is sent to the FM. Based on a history of sensorimotor contingencies 50 and the current EC [29], the FM simulates and predicts both the proximal 51 (proprioceptive feedback from the upper limb muscles) and distal (visual or auditory 52 feedback of the puck hitting the goal spatiotemporally aligned with the trajectory) 53 sensory consequences of the efferent movement [7,30]. Upon action completion, the 54 predicted and the actual information from the sensory systems are compared and the 55 discrepancies constitute error signals, the so-called Sensory Prediction Errors (SPE) [31]. 56 FM integrates them to (1) correct the motor command on the consecutive trial to 57 maximize future performance [31,32], and (2) provide new estimates of the current state 58 of the body and the environment allowing self-recognition [16,21,22,31,33]. 59 Grounded in the discussed neurophysiological evidence about the nature of the 60 internal FMs and their functions in goal-oriented behavior as well as the novel insights 61 about BO, here, we propose that BO might be a product of the computation of FMs 62 and therefore it might depend on their consistency, specifically, in action contexts where 63 sensory signals are self-generated. Crucially, FMs are not limited to the bodily feedback 64 exclusively, but rather they integrate across both proximal and distal sensory 65 predictions which pertain to the interactions of an agent within an environment in a 66 goal-oriented manner [30]. We will refer to these signals as Task aligned with its trajectory that depends on the direction of the arm movement. In case 70 the actual location of the sound of the puck hitting the goal does not correspond to the 71 efference copy or corollary discharge, it would reflect on errors of FMs [23,24,29,34,35]

81
After providing written informed consent, sixteen healthy participants were recruited for 82 the study, eight males (mean age 24.0 ± 2.65) and eight females (mean age 22.64 ± 2.25). 83 All subjects were right-handed (handedness assessed using Edinburgh Handedness 84 Inventory) [36], had normal or corrected-to-normal vision and reported normal hearing. 85 They were pseudorandomly assigned to two experimental groups following a 86 between-subjects design, which prevented habituation to the ownership measures, distal 87 auditory cues, manipulations and fatigue. We used stratified randomization to balance 88 the conditions in terms of age, gender and previous experience with VR. All   The experimental setup (Fig 1 A) comprised a Personal Computer (PC), a motion Vive, www.vive.com) and headphones. Similar to others [19,37], here we used VR 96 method as a tool to study the modulation of BO. The protocol was integrated within 97 the Virtual Environment (VE) of the Rehabilitation Gaming System (RGS) [38]. 98 During the experiment, while seated at a table, participants were required to complete a 99 goal-oriented task that consisted in hitting a virtual puck into the goal (air hockey, Fig 100  1 A, B1). Throughout the experiment, participants' arm movements were continuously 101 tracked and mapped onto the avatar's arm, such that the subjects interacted with the 102 VE by making planar, horizontal movements over a tabletop (Fig 1 A, B). To prevent 103 repetitive movements, at the beginning of every trial, the puck pseudorandomly 104 appeared in one of the three Starting Positions (SP; left, center, right) (Fig 1 B2). The 105 frequency of appearance of every SP was uniformly distributed within every 106 experimental session. Participants received instructions to place their hand in an 107 indicated SP and to execute the movement to hit the puck when the color of the SP 108 changed to green ("go" signal). Each trial consisted of one "hit" which could end in   We used three measures to quantify performance: scores, directional error, and Reaction 137 Times (RTs). Scores were calculated as the percentage of successful trials (the puck 138 enters the gate), while the directional error equaled the absolute angular deviation from 139 the straight line between the starting position of the puck (left, central or right) and the 140 center of the gate (Fig 1 B3). We computed RTs as time intervals between the 141 appearance of the puck and action initiation. Since the task did not impose a time limit, 142 we expected neither significant differences in RTs between conditions nor In the experimental block, they were randomly split into two conditions: Congruent "C" (blue), and Incongruent "I" (black). At trial 151, all participants went through the threatening event which served to measure GSR responses. The same color-code ("C"-blue, "I"-black) is used throughout the manuscript. (E) TRDC auditory manipulations-temporal, spatial and semantic. Upper panel: congruent condition; lower panel: incongruent condition.

5/15
speed-accuracy trade-offs. We predicted, however, that the auditory TRDC 144 manipulations in "I" might alter scores and directional accuracy as compared to "C". threat [39]. GSR was recorded and stored throughout the experiment. For the analysis, 151 we calculated the mean and the standard deviation of the integral of baseline-subtracted 152 GSR signal per condition in a non-overlapping time windows of 9s. In particular, we 153 expected an increased GSR following the threatening stimulus in "C" as compared to instance [19]. Specifically, they were asked to point to the location of the tip of their left 158 index finger with the right index finger with no visual feedback available. The error in 159 pointing [12] was computed as the distance between the two locations (the actual 160 location of the tip of the left index finger and the pointing location) and measured in 161 centimeters. We subtracted baseline responses from post-experimental errors for each 162 participant. We expected stronger proprioceptive recalibration, and therefore, higher 163 pointing errors in "C" as compared to "I".

171
To test our hypothesis that action-driven distal cues which pertain to the task 172 contribute to Body Ownership (BO), we used a VR-based experimental setup (Fig 1 A, 173  B) where subjects were to complete a goal-oriented task by performing horizontal 174 (planar) movements, and manipulated the congruency of TRDC action outcomes (Fig 1 175  E). The experimental protocol (Fig 1 D) consisted of three phases: Training Block (TB), 176 (2) Experimental Block (EB) in either congruent ("C") or incongruent ("I") condition, 177 and (3) Threatening Event (TE) (Fig 1 D, C). To quantify BO, for each experimental   (Fig 2 A). To explore the effects of 187 the congruency of TRDC on performance, we compared both conditions in terms of 188 directional errors (Fig 2 B). In particular, a T-test indicated that the errors were   (Fig 2 C). To further investigate the relationship 191 between the quality of the TRDC and performance, we averaged and compared the 192 directional errors following the three types of auditory manipulations (Fig 2 D). 193 Interestingly, we found no difference between the distinct TRDC including spatial

226
In the present study, we asked the question of weather Body Ownership (BO) depends 227 on the consistency of FMs driving goal-oriented action and the proximal signals received 228 8/15 from the body (Task-Relevant Proximal Cues, TRPC) as well as distal ones triggered by 229 events outside of the body (Task-Relevant Distal Cues, TRDC). In particular, we 230 investigated the influence of TRDC on performance and BO using an embodied, 231 VR-based, goal-oriented task where action outcomes were signaled by distinct TRDC in 232 the auditory domain. We hypothesized that (in)congruency of TRDC would affect both 233 performance and BO. Our results confirm this hypothesis and demonstrate that both 234 are reduced when the TRDCs are incongruent.

235
The plasticity of BO relative to the congruency of proximal interoceptive and 236 exteroceptive (both efferent and afferent) signals is well accepted [15,22,[41][42][43]. In 237 particular, using the standard method of ownership manipulation (RHI) [5] or VR-based 238 visuomotor (a)synchrony paradigms, where BO is established by either 239 externally-generated visuotactile or self-generated visuomotor synchrony, 240 neurophysiological and behavioral studies have demonstrated that BO results from 241 bottom-up and top-down multisensory matching [15,42]. Hence, changes in self-specific 242 priors (incorporating a fake body part into self-representation) are driven by statistical 243 correlations when conflicts between visual, tactile, and proprioceptive inputs tend to 244 zero [5]. Current theoretical action-perception framework, grounded in predictions, 245 interprets this underlying mechanism driving BO as minimization of Sensory Prediction 246 Errors (SPE), which during RHI leads to bias in a multisensory generative 247 model [16,22,43]. Similar interpretations have been proposed in contexts when the 248 sensory cues are self-generated suggesting that, when acting, BO might result from a 249 computation of an internal Forward Model (FM) [23]. 250 Crucially, however, the discussed paradigms [18][19][20] do not include manipulation of 251 the environment, which constrains our understanding of BO to self-and 252 externally-generated proximal cues (TRPCs) exclusively. To study whether the 253 plasticity of BO depends on the consistencies of internal FMs and therefore not only 254 TRPC but also TRDC following volitional goal-oriented behavior, we used a paradigm 255 which required the participants to perform actions that trigger TRDC and manipulated 256 their congruency. We predicted that BO scores might be lower in the condition where support that TRDC which pertain to the task and violate predictions about the action 260 outcomes compromise BO. Specifically, we found that the BO scores were significantly 261 higher in the congruent compared to the incongruent condition in all analyses.

262
Subsequent correlations between the proposed measures (Fig 3) further confirmed the 263 consistency of the obtained results within three dimensions of ownership quantification 264 including physiological response, behavioral proprioceptive recalibration, and a 265 conscious report [40].

266
Similarly to the asynchronous stroking condition in the standard RHI paradigm [5] 267 or visuomotor mismatch [19], here we interpret the obtained low-ownership outcome in 268 the incongruent condition (Fig 2 Upper panel: body Ownership) as a consequence of 269 high sensory SPE of FMs or the efference copy [16,28]. In our case, however, the 270 sensory conflicts are driven by a discrepancy between the predicted and actual distal 271 auditory signals which do not pertain to the body. In particular, the manipulation of 272 TRDC might have reflected on the errors of the FMs which drive both performance and 273 BO. This could suggest TRDC might influence TRPC establishing a feedback loop such 274 that any (in)congruent relationship between them which pertain to the goal will affect 275 BO and even define the boundaries of the embodied self. To the best of our knowledge, 276 this result proposes for the first time that BO might be driven by bottom-up integration 277 and top-down prediction of not only proximal sensory cues deriving from the body but 278 also distal ones occurring outside of the body. This would support results from other 279 studies showing that BO is coupled to the motor systems [37,44] and that it might 280 9/15 depend on FMs. Consistent with other studies [20], the differences in the perceived BO 281 were not influenced by a lack of agency (Fig 2 J). In particular, the participants 282 experienced control over the virtual hand in both groups, probably due to the congruent 283 mapping of the proximal cues. The visual feedback of the movement of the arm always 284 followed the desired trajectory, which is one of the crucial questions addressed in the 285 standard self-reported agency assessment [40]. forward models) [16,22,45].

301
What is the role of TRDC in goal-oriented behavior? Our results demonstrate that 302 performance, measured as overall scores (Fig 2 A) and directional errors (Fig 2 B, C), 303 was significantly hampered in the incongruent compared to the congruent condition.

304
Importantly, these results did not depend on a difference in RTs (Fig 2 E) suggesting no 305 influence of possible attentional biases (ie. distractions) in either of the groups. On the 306 one hand, this outcome might be interpreted from a computational motor control 307 perspective. The reported differences in performance between the conditions could have 308 been influenced by the discrepancies between the efference copies of distal events and 309 the actual action outcomes. Indeed, results from motor control studies support the 310 notion that learning (progressive reduction of error) depends on both proximal and 311 distal sensory prediction errors that allow for adjustments and anticipation of possible 312 perturbations deriving from the body and environment [30,35,[46][47][48][49]. As a result, 313 inputs from all the sensory modalities are transformed into error signals while an action 314 is being executed, updating the forward model and, consequently, future 315 behavior [29,31,50,51]. In our experiment, the directionality of the error indicated by 316 the spatial distribution of the sound (left or right) as well as its temporal characteristics 317 could both be interpreted as error signals which supervise corrective commands. Thus 318 TRDC in the incongruent condition might have influenced performance, which, in turn, 319 affected body ownership. For example, clinical studies provide evidence that patients 320 suffering from hemiparesis, whose motor function is reduced due to stroke, progressively 321 stop using the paretic limb: the so-called learned non-use phenomenon [52]. In this and 322 other neurological cases, permanent lack of use (low performance) often causes 323 disturbances in the sense of ownership and agency [53] supporting a hypothesis that 324 there might be a causal effect between performance and body ownership. The present 325 design, however, which includes three types of sensory manipulations pseudorandomly 326 distributed within each block, does not allow us to disambiguate between the specific 327 contribution of each of TRDC. A systematic study on the influence of individual sensory 328 signals, including the three manipulations, would help to better understand the 329 mechanisms accounting for low-performance scores in the incongruent condition.  The FM, in turn, generates and updates predictions relative to both the body and the 334 environment during voluntary actions [44], reinforcing the history of sensorimotor 335 contingencies. In particular, we find evidence that BO is involved in generating 336 body-specific predictions about the sensory consequences of voluntary actions thus 337 determining somatosensory attenuation [44]. This outcome is consistent with another 338 study using standard RHI in VR [37], which shows that the degree of ownership 339 correlates with motor performance in a decision-making task. Contrary to the previous 340 discussion, in this case, ownership would have a modulatory effect on performance.

341
At the current stage, we cannot disambiguate between the two alternative hypotheses 342 and determine whether TRDC influence BO and performance in parallel or sequentially 343 and what is the directionality. We demonstrate, however, that both depend on TRDC, 344 which supports that both depend on the consistency of FMs driving goal-oriented action 345 and both proximal and distal cues which this action generates pertaining to both the 346 body and the environment [16,22]. We expect that this outcome will allow for the 347 advancement of our understanding of the mechanisms underlying body ownership.  acquired brain lesions including neglect [58], anosognosia for hemiplegia [59] or 355 somatoparaphrenia [60].

357
The authors declare no competing financial interests.