2.3.1. Procedure.

The experimental procedure had a duration of approximately 90 minutes and was divided into 2 parts:

• First, the neutral condition, lasted 15 minutes and consisted of a « back-and-forth » passage, i.e. TDCs moved progressively from self to other, then from other to self while, during the same period of time, participants with AN experienced the opposite condition, i.e. they moved from other to self, then from self to other;
• The second part, called the sensorimotor stimulation condition, lasted 60 minutes and involved exposing participants to the same light intensity conditions as described in the neutral condition, i.e. a "back-and-forth » passage, but repeated 5 times with a different stimulus applied to the participants with AN during each passage.

Thus, the whole task consisted of 6 back-and-forth passages: one in neutral condition, then 5 with different successive sensory stimuli only for AN participants.

There was a 10-min pause between the first and second parts to allow attentional recovery.

First part: Neutral condition. This experimental protocol was inspired by the one previously used by Du Boisgueheneuc in Alzheimer’s disease patients, Keromnes and Tordjman in patients with schizophrenia [45] and Lavenne-Collot in patients with ASD [84]: the light intensity of the two LED sets were gradually and independently increased or decreased on the two sides of the mirror, such that the two individuals found themselves alternatively in the other condition or the self condition, as explained below.

Before starting the task, participants were seated on opposite sides of the device so that the eyes of both partners were strictly aligned in the mirror reflection. They were instructed not to make any body or facial movements (including facial expressions or grimaces) for the whole duration of the task.

At the beginning of the experiment, the patient was in the other condition, i.e., he starts by seeing the control subject’s face through the mirror without seeing his own face; then, the patient’s own image in the mirror becomes increasingly more apparent as a function of the light intensity.

Then, the patient is in the self condition, i.e., he begins to see his own face reflected in the mirror without seeing the control’s face; then, with changes in the light intensity, the control subject’s image in the mirror becomes increasingly more apparent.

At the same time, the control subject undergoes the same procedure, except that he starts in the self condition which transforms into the other condition.

Similar to Keromnes et al. [45], we called a “passage back and forth” the series of conditions that permitted each subject to return to the starting conditions after the conditions with identical light intensities was presented twice. Thus, one passage back and forth consists of the following sequence:

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

https://doi.org/10.1371/journal.pone.0309548.t002

The simultaneous alterations in light intensity on each side of the mirror experienced by the AN patient and his matched control are presented in Fig 2A. Importantly, since modifications in the light intensity on one side of the mirror modify its reflective properties on both sides, we should consider the relative light intensity that is actually perceived by participants on each side, as shown in Fig 2B.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2.

Simultaneous alterations in light intensity on each side of the mirror over time (Fig 2a + 2b). Fig 2a. Light intensity over time.

1. - At the beginning of the procedure, the light intensity was 100% for the HCs (mirror effect), whereas the light intensity was 0% for the AN patients (transparent window effect). -Then, the light intensity is progressively increased in 10% steps on the AN patient’s side while remaining at 100% for the HCs until 100% is reached on both sides.
2. - During this first step, the AN patient’s image progressively appears and becomes increasingly fused with the HC’s image.
3. - When both sides are at 100%, the light intensity is progressively decreased in 10% steps on the HC’s side, while it remains at 100% on the AN patient’s side.
4. - During this step, the HC’s image progressively fades until its total disappearance when the light intensity drops to 0%.
5. - Then, the reverse procedure is used to return to the initial configuration (100% on the HC’s side, i.e., a mirror effect for the HCs, and 0% on the AN patient’s side, i.e., transparent window effect for the AN patients).
6. - Therefore, a condition with identical light intensities is presented twice to the participants during this sequence called one “passage back and forth”.

Fig 2b. Relative light intensity perceived by each participant over time. Relative light intensity perceived by the participants taking into account that modifications of light intensity on one side modifies its reflective properties on both sides.

https://doi.org/10.1371/journal.pone.0309548.g002

After every change in light intensity, the participants were instructed to identify whether the presented image mostly corresponded to their own face or to the other’s face. Both subjects were asked a simple question: “who do you recognize most in the mirror?” The expected response was either “me” or “her/him” (to designate the other person), without any other alternative (i.e., there were no different answers and no answer was not permitted). The question was always addressed first to the individual with AN and then to the HC participant. Each stimulus was presented for 10 seconds each during a progressive sequence. The response time was unlimited. Once the answer was given by each participant, the next stimulus was presented. The type of response measured and recorded was a verbal response. The use of response button boxes was initially discussed as a way to control for bias in participants’ verbal responses. However, in addition to the difficulty of using manual buttons in totally dark lighting conditions, visual control might have been required to press the button, and thus, attention to the mirror image could have been disrupted or the movement could have induced a misalignment of the 2 faces. In addition, as in Keromnes [45], oral responses seemed to reassure these patients, as the task involved maintaining relationships with others and were more relevant in terms of identity reinforcement and assertive self-recognition.

Second part: Sensorimotor stimulation condition. In this second part, the same procedure was applied i.e. each stimulus of identical light intensity was presented twice to the participants during one passage back and forth as previously described in the neutral condition, but the postural or sensory stimulation conditions for the AN individuals were changed while the configuration for the HC individuals remained unchanged throughout the whole procedure. Indeed, our aim was to explore whether sensorimotor conditions reinforcing body image could improve the performance obtained by the AN participants. Accordingly, the tasks were used only for AN participants to investigate this compensatory effect, assuming that healthy control participants have intact bodily self-consciousness abilities (without the need for reinforcing supports)

As previously mentioned, body schema and self consciousness are both based on a multisensory integration process [8]. Therefore, we modified sensory inputs in participants with AN to assess their effect on mirror self-recognition and SOD. Five different tasks were administered, including different sensory modalities and locations, in agreement with psychomotor therapists, based on their knowledge of strategies used to reinforce body schema in usual care of patients with AN.

The objectives of these different tasks were: (1) to increase the awareness of each of the 2 hemi-bodies by successive weighting of its (1) right and (2) left sides, (3) to reinforce the median axis of the body by providing increased postural support at the level of both the back and the pelvis, (4) to stimulate the vestibular system with a cushion generating an unstable position, and (5) to provide intensive somatosensory stimulation with a neoprene jacket.

Participants were tested successively on each of these distinct tasks that were administered always in the same order:

1. Right hemi-body weighting (RHB) condition: Weights of 75 grams were placed around the AN patient’s right wrist and right ankle
2. Left hemi-body weighting (LHB) condition: Weights of 75 grams were placed around the patient’s left wrist and left ankle
3. Back- and Foot- Rest (BFR) condition: Patients with AN were no longer seated on a stool but in a chair with a backrest and a step was placed under their feet to allow pelvic flexion
4. Unstable Cushion (UC) condition: Patients with AN were sitting on an unstable cushion installed on the stool
5. Neoprene (NP) condition: Patients wore a neoprene jacket

Debriefing occurred at the end of the experiment during an informal and friendly period of time that did not include data collection. A snack with cakes and drinks was offered during this period. The participants were asked about their feelings regarding the task and their awareness about differences in stimulation, asked whether they found face recognition and discrimination difficult, and whether they were familiar with this type of morphing (as some teenagers regularly use morphing applications).

2.3.2. Data analysis.

The main outcome was the light intensity levels at which critical changes in self/other identification occurred:

1. Level M1 was the critical perceptual threshold corresponding to the first time the individual recognized himself when his own image progressively appeared in the mirror during the other-to-self morphing sequence.
2. Conversely, level M2 was the critical perceptual threshold corresponding to the first time the individual recognized the other’s image in the mirror when this image progressively appeared in the mirror during the self-to-other morphing sequence.

To summarize, the lower the M1 or M2 levels were, the larger the proportion of self was in the image (and the smaller the proportion of the other), which reflected a difficulty for the individual to uncouple from his own image.

Determination of M1 and M2 levels as well as a synthesis of the experimental procedure are shown in Fig 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Synthesis of the experimental procedure during one “passage back and forth”.

During the task, the image the AN patient sees undergoes a morphing from other to self and then from self to the other. Simultaneously, the image the HC sees undergoes a morphing from self to other and then from other to self. The threshold M1 corresponds to the light intensity at which the perceptual shift from other to self occurs. The M2 threshold corresponds to the light intensity at which the perceptual shift from self to other occurs. The procedure is then repeated a second time. Thus, for each subject, 2 M1 and 2 M2 thresholds are obtained.

https://doi.org/10.1371/journal.pone.0309548.g003

2.3.3. Statistical analysis.

After each step (change in light intensity), the participant’s verbal responses were recorded. The analysis of the main outcome variable was conducted by comparing the M1/M2 levels (expressed as percentage of light intensity) between individuals with AN and HCs. As this variable was not normally distributed, the nonparametric one-tailed Mann–Whitney test was used to compare M1 and M2 levels between the two groups.

The effects of differents sensorimotor tasks on the M1 and M2 levels were assessed by using the non-parametric Wilcoxon test to compare for the AN patients the results obtained in the neutral condition and the results obtained in the 5 sensorimotor conditions (RHB, LHB, BFR, UC, NP).

4.1.1. Self- consciousness.

The results of the present study support that AN individuals show a difference in mirror face recognition thresholds compared to HCs on the double mirror task. These results are consistent with previous research showing that self-recognition abilities may be impaired in AN [59, 77]. Furthermore, in line with previous research showing the value of the double mirror device in studying bodily self-consciousness [46, 97], our results support the notion that self-disorder is an important dimension of AN [98, 99] and, in particular, that the embodied aspects of the self could be disturbed [5, 100].

4.1.2. Comparison with previous research using static images.

Our results showed that the AN individuals required images that contained less proportion of self than controls to recognize themselves in the reflected image, regardless of the direction of the morphing (i.e., in the "self" or "other" conditions). These results contrast with those found by Hirot et al. [77] during a self/other morphing task using static images that showed that patients required images containing more ’’self’’ to recognize themselves. Several arguments support this discrepancy: First, (i) contrary to photo, looking at ourselves in the mirror gives us access to our own image, besides the proprioceptive, tactile and motor sensory cues that are necessary for the representation of one’s own face [101]; (ii) studies have found that mirror self-recognition emerges prior to photo self-recognition [102]; (iii) different neural responses have been found when comparing mirror and photo self-processing [103] and; (iv) preserved self-face recognition on photographs despite incapacity in mirrors has been described in some neurological patients showing a kind of agnosia for recognition of their own reflected image [104]. Moreover, our experimental design differs from that of Hirot’s study because they used an aleatory morphing protocol thought we presented the morphs incrementally from 0% self/other to 100% self/other in two directions separately.

4.2.1. Implications regarding spatial cognition: Privileged egocentric reference frame.

The results obtained in the present study showing earlier self-recognition and delayed other recognition in the AN individuals, compared to the HCs, could be interpreted as a difficulty to inhibit their own perspective during the face recognition task and adopt the reference frame of the other, resulting in a privileged use of the egocentric reference frame.

Indeed, unlike when one is faced with a simple photograph, the mirror requires processes of perspective shifting and spatial transformations beyond pure self-recognition. In particular, self-recognition in the mirror requires matching one’s sensorimotor experience (1st person perspective) with the object seen in the mirror (3rd person perspective), thereby identifying the "I" with the "me" and representing that self as an object to others and to oneself [103]. Moreover, the face-to-face postural configuration of both subjects during the Alter Ego double mirror task requires a 180° mental rotation of one’s own body. In particular, Thirioux et al. [83] emphasized the role of visuospatial abilities on self-recognition and SOD within this task. Thus, our results are consistent with previous studies showing alterations in spatial cognition in ANs and, in particular, a privileged use of the egocentric frame of reference during spatial performance tasks [14, 55]. Moreover, given the implications of impaired spatial references for social cognition and perspective shifting or empathy [54, 56–58], this raises the question of the ability of patients with AN to infer mental states other than their own. To date, the nature of the links between alterations in body schema and alterations in spatial cognition remains unknown [14]: is it the alteration of the body signal that may participate in the disruption of spatial cognition (bottom up effect)? Or is it the alteration of spatial cognition that may participate in the alteration of body schema (top down effect)? Overall, analysis of the results under sensorimotor conditions tends to support the second hypothesis (see below).

4.2.2. Implications for multisensory processing: Overinclusion of the other’s face.

Our results showed that the AN individuals, compared to the HCs, judged the morphs to be more like the self in both directions of morphing. These findings can be interpreted as a greater level of assimilation of the other’s face in the representation of one’s own face among the AN individuals than the HCs. Such misattribution of others’ facial features to oneself has been reproduced in nonclinical populations through simple psychophysical manipulations in a procedure involving sensory processing called interpersonal multisensory stimulation [29]. Indeed, studies using body ownership illusions have shown that, under certain conditions, the sense of self can be manipulated to include a fake or another person’s body part, for instance, the hand [105], face [101, 106, 107] or whole body [108].

The double mirror task involves a simultaneous visual perceptual conflict between the visual input (“I see the other”) and the proprioceptive experience (“I expect to see myself in the mirror”). This conflict is reinforced by the everyday experience we may have developed that, when we look in the mirror, we see ourselves and not another person. As specified by Thirioux [83], here we studied the effect of a unisensory (visual) conflict, whereas previous studies have typically studied the effect caused by multisensory (especially visual-tactile) illusions [29].

Several studies have revealed abnormalities in the multisensory integration process in patients with anorexia nervosa [9]. In particular, people with anorexia nervosa have been found to exhibit greater sensitivity to the Rubber Hand Illusion [105], indicative of a greater susceptibility to appropriating and embodying an object external to their own body, i.e. a more plastic bodily self [13, 109]. Yet conflicting results have shown similar levels of sensitivity in full-body illusion tasks compared to controls [110, 111], but only when a highly emotionally valenced body part, such as the abdomen, was stimulated to promote the illusion [9].To our knowledge, this greater susceptibility to the embodiment illusion for body parts (contrasting with full-body) has never been explored at face level in individuals with AN, as assessed through paradigms such as the “enfacement illusion” [106, 112]. However, both body plasticity and face representation have been shown to rely on the same multisensory integration processes [8, 101]. Thus, the results of the present study are consistent with the generalization to the face of a greater tendency to the embodiment illusion for body parts in AN individuals.

Moreover, our results indicated that people with AN continued to recognize themselves in the mirror despite the increase in the proportion of the other person’s face in the image, regardless of the direction of the morphing. These results may indicate that, in the AN individuals, the proprioceptive system would be less vulnerable to bias originating from visual information (i.e., AN individuals depend more heavily on proprioceptive information than on visual information when incongruent). This finding is in line with previous studies that pointed out an overreliance on proprioception in the presence of competing signals from other modalities [17]. Moreover, they are also consistent with reduced integration of visual and proprioceptive information in AN, also supported by the ubiquitous clinical finding in these patients that their visual perception of the body, especially in a mirror, does not correct their distorted body image [113].

4.2.3. Comparison between the self and other conditions: Imbalance mechanisms of projection / simulation.

Our results showed that the AN individuals, compared to the HCs, judged the morphs to be more like the self in both directions of morphing. Importantly, the comparison between self and other conditions shows that, unlike the HCs, the AN individuals switched from other to self and from self to other at different thresholds. This discrepancy indicated that the confusion between self and other may have been determined by a unidirectional overinclusion of the other’s attributes by the AN individuals. Conversely, the HCs may have exhibited compensatory mechanisms with bidirectional self- and other-facial feature attribution. i.e., they projected their facial features onto the other’s face as much as they introjected their partner’s facial features onto their own face. This understanding is supported by Thirioux’s findings using a double mirror in healthy subjects [83] and human interpersonal understanding based on balanced mechanisms of self-projection and simulation [41, 114]. Indeed, as noted by Freud “the boundaries of the self are not constant” [115]. That is, both processes also transformed the margins between the self and the other, either by externalizing facial features of the self into the other (projection) or by alienating facial features that do not belong to the self but to the other (introjection) [116]. Thus, our findings may be consistent with psychoanalysts who described anorexia as a failure of identification and a pathology of object relations [117] particularly marked by deficiencies in the balanced introjection/projection mechanisms that are necessary to establish a sense of identity, leading to characteristic forms of pathological projective identification often evident in these patients [118].

4.2.4. Switching between the abstracts of self and other.

Our experimental design requires switching from self to other and from other to self as the morphing gradually shifts from 0% to 100% between self and other and vice versa. Thus, the double mirror Alter Ego task is consistent with recent conceptualizations of SOD as being achieved through the ability to switch between representations of self and other and to inhibit the representation that is not relevant in a given situation [119]. In our study, differences in M1 and M2 thresholds between the AN individuals and HCs were consistent with AN being associated with an inadequate control of the self/other control or switch [37].

Another hypothesis explaining our results is that anorexic participants may have difficulties matching their own actions with their sensory inputs, i.e. difficulties in predicting the sensory outcome of their own action. This hypothesis is consistent with the theory of active inference, which postulates that the brain uses an internal model to continuously infer and control external or internal signals associated with the body [120, 121]. This theoretical framework is attracting growing interest in understanding alterations in sense of self [65], particularly in AN [122].

Indeed, although our task does not involve active movements, since participants were instructed not to move, we cannot exclude minimal movements such as eye blinking or respiratory movements. In any case, the double mirror task involves motor imagery, especially mental rotation [83] which implies predicting the sensory consequences of the imagined movement [123]. Usually, the brain weights and selects the most informative signals it receives to minimize prediction error and reduce surprise, notably by attenuating self-generated percepts that are highly predictable [124] through a copy of the motor signal called efference copying [125]. In AN, subjects have been shown to rely on incorrect predictions, i.e. on the unverified representation of their own body. In particular, some authors have highlighted a lack of integration of interoceptive and exteroceptive signals leading to an allocentric lock-in state, where body memory is not updated according to new sensory evidence relating to the body’s actual shape [122, 126]. Interestingly, with regard to the distinction between self and others, prediction errors have been demonstrated at the tactile level in subjects with AN, the lack of neural attenuation of tactile sensations produced by self making them more similar to those produced by others [74]. This finding is particularly interesting considering the crucial role of tactile cues in the construction of self-consciousness from fetal life [63, 66, 67].

4.3.1. Back and foot support condition.

Comparison between neutral and BFR conditions. Our results showed a significant increase in the M1 threshold in the BFR condition compared to the neutral condition. This indicated that, when they were in the BFR condition, the AN participants showed a delayed self-recognition, i.e., they continued to perceive the other longer while their own image was gradually appearing.

Back support is an essential early developmental modality for the construction of one’s identity, and in particular one’s body. Indeed, during fetal development, the uterine wall provides the primary back support that allows for the development of oral activities (swallowing, etc.). Even after birth, back support remains fundamental, especially in the sensorimotor model developed by Bullinger [127] that argues that postural control and tonic-sensory balance allow infants to interact with their environment and develop social involvement [128]. Notably, in a pilot study, Rahme et al. showed a significant increase in social attention, in the back support condition compared to no back support in children with neurodevelopmental disorders [129]. Similarly, the results of our study are consistent with the idea that the BFR condition may have helped the AN individuals to be more other-centered.

Comparison between AN and HC groups. In contrast to the results obtained in the neutral condition, when the AN participants were settled in the BFR condition, there was no more significative difference between the HC and AN groups. These results suggest -that the implementation of postural supports against the medial axis would have compensated for the alterations in self-recognition and SOD in the AN participants.

4.3.2. Other sensorimotor conditions.

As far as other sensorimotor conditions are concerned, both left and right hemibodies ballasting, unstable cushion and neorprene conditions had a greater effect on M2 threshold than on M1, including a decrease in M2 threshold. This indicates that these conditions had no significant effect on self-recognition, but were significantly associated with delayed other-recognition, suggesting that they may have helped the AN individuals to be more self-centered. These results are in line with previous research using double mirror in schizophrenia patients that showed that tactile and kinesthetic tasks helped to be more self-centered [45].

Furthermore, our results did not show any significant difference regarding the right- or left-lateralized tasks, which did not support a preferential input lateralization effect on face recognition and SOD thresholds in the AN individuals.

Regarding the Neoprene condition, previous research conducted in a single case study conducted by Grunwald showed that intensive somatosensory stimulation with a neoprene diving suit led to improved body image in the AN participants [130]. However, the neoprene suit was worn for one hour three times a day whereas, in our study, this condition may have been applied for too short a time to induce a change in self-recognition patterns.

4.4.1. Relationship between bodily self-consciousness and socio cognitive functioning.

In summary, our results highlight: (i) disturbances in mirror self-recognition suggesting bodily self-consciousness impairments in AN; (ii) an egocentric bias in self/other disctinction found in AN participants compared to HCs and; (iii) changes in self-recognition and SOD thresholds were observed when sensorimotor or postural conditions were modified in individuals with AN.

Although our preliminary results must be put into perspective given the small sample size, they support the hypothesis that more fragile bodily self-consciousness in AN participants may impair self-referential processes (bottom up effect). This hypothesis is consistent with the results obtained by Thirioux [83] who showed an alteration of self-referential processes by inducing self-fragility through self/other fusion via the double mirror in healthy subjects.

4.4.2. More fragile median body axis.

Our results support fragilities of the body schema at the level of the median axis in AN. This hypothesis is consistent with clinical observations showing, in these patients, a misalignment of the whole body, especially with respect to this axis, a restricted breathing pattern coupled with muscle rigidity having a major negative impact on body stability [131], problems with stabilization of the back and pelvis in a neutral position and difficulties in performing pelvic movements [132] as well as gait abnormalities with biomechanical differences primarily in the pelvis and hip [133]. Interestingly, while studies on the effects of body therapies aimed at increasing body-awarness (often referred to as ‘body awareness therapies’) are still rare, the authors of a recent study on the lived body in patients with anorexia found that by allowing themselves to lean on the seat and back of the chair (during physiotherapy consultations), instead of perching on the edge of the chair, anorexic subjects felt more attentive and present in the here and now, indicating an increase in self-consciousness through the addition of medial postural support, consistent with our findings [134].

Furthermore, our results did not find a preferential effect of lateralized tasks contrasting with previous research supporting similarities found in AN patients with hemineglect patients [135] but with divergent results, just as the existence of a hemispheric predominance is still debated [82, 135]. More than a hemineglect similar to the one found in neurological patients, our results may raise the hypothesis of a functional cleavage or split at the level of the median axis that is supposed to connect the right and left parts of the body.

4.4.3. Further therapeutic and research implications.

As they highlight the prominence of abnormalities in body self-consciousness in understanding alterations in social-cognitive and interpersonal functioning in AN, our results also support the value of physical treatments in this condition. In particular, our results support that psychomotor therapy focused on strengthening and stabilizing the body axis could improve self-awareness, which could then have training effects on social-cognitive functioning.

In addition, the fragilities found in the median axis of the body raise the hypothesis of disturbances in the very early stages of body construction in AN patients, as observed in other neurodevelopmental disorders [127, 129]. Therefore, it would be interesting to look for possible early signs of vulnerability in the sensorimotor domain, following the example of retrospective studies using home movies performed for such disorders [136].

Finally, using the same protocol respectively in patients with schizophrenia and also ASD versus healthy controls, Keromnes and Tordjman [45] and Lavenne-Collot [84] found similar results to the present study (i.e., earlier recognition of self and later recognition of others). Interestingly, given the overlap in symptomatology between ASD and AN, it would have been interesting to collect participants’ autism spectrum quotient (AQ) scores, but this goes beyond the scope of this preliminary study. Overall, these results underline the value of considering SOD disturbances as a transdiagnostic dimension shared with other psychiatric or neurodevelopmental disorders and supports the interest of further research using the Alter Ego in a double mirror at the diagnostic but also therapeutic level.