AR system.

We used an HM-A1 (Canon) video see-through-type head-mounted display (HMD) with a resolution of 1280 × 1024 pixels per eye and a total field of view of 41° horizontally and 31° vertically displayed at 60 Hz. We implemented the AR environment using the AR platform system (Canon, MP-110). Head tracking was performed with Polhemus 3 SPACE FASTRAK. Fig 1A presents the experimental set-up.

Download:

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

Fig 1. Experimental environment and manipulation of the perception.

(A) Experimental set-up. (B) Appearance of the limb on the HMD with the alpha-value set at 0.25. (C) Appearance of the limb on the HMD with the alpha-value set at 0.5. (D) Virtual stick passing through a participant’s limb. The alpha value was 0.25. (E) Touching a participant’s wrist with a real wooden stick with an alpha value of 1.0.

https://doi.org/10.1371/journal.pone.0268618.g001

Manipulations.

We manipulated the opacity of the participants’ limbs to alter their body representation. Toward this end, the scene was acquired from the participants’ perspective with the camera on the HMD, whereas the limb was extracted from the scene based on its color. We set the value of the alpha channel of the extracted area to 0.25 in Experiment 1. However, the level of body ownership decreased greater than we expected according to the change in the alpha value. Therefore, we used 0.50 for the alpha channel in Experiment 2. The limb was presented on a virtual black-and-white background (Fig 1B and 1C). In the condition where the alpha value was changed, the alpha value was decreased from 1 to 0.50 or 0.25 over 2 to 3 s after a cue from the investigator. The participants observed the change, which permitted them to conceive that their limb was transforming.

Two additional manipulations were used to alter the sense of body ownership and enforce the intangible nature of body representation. First, we used the pass-through manipulation, where the participants observed a virtual stick passing through their limb (Fig 1D). This step increased the feeling that the limb is intangible. First, we showed the participants the virtual stick and confirmed that they observed it. We moved a virtual stick through the participant’s wrist from upper to lower and from lower to upper and repeated this process five times. The investigator carefully adjusted the position of his/her hand to avoid displaying it on the HMD.

The second manipulation aimed to recover the decreased level of body ownership. In Experiment 1, the participants were instructed to move a finger and observe it (movement manipulation). Previous studies demonstrated that visuo-motor consistency can increase the feeling of body ownership of the observed limb [24]. First, the participants bent their wrist backward to avoid touching the desk when moving their fingers. The experimenter then called out the name of one of the fingers, such as the index finger, in random order. In response, the participants bent the named finger toward their palm at the metacarpophalangeal joint and placed it back to the normal position. Each finger was called twice in random order for each participant. Finally, they returned their wrist angle to straight.

The results for Experiment 1 indicated that the movement manipulation was unable to recover the body ownership of the participants. In Experiment 2, we used the touch manipulation, where the participants’ wrist was touched 10 times using a real wooden stick (Fig 1E). The participants were instructed to continue observing the limb and as it was being touched by the stick. Visuo-tactile synchronous stimulation is a very popular method for increasing the sense of body ownership [18, 25]. We adjusted the range of the color extraction to enable the observation of the stick. The experimenter held the stick carefully so that his hand was not displayed on the HMD.

We omitted the abovementioned manipulations from the following conditions: alpha = 1.0 and the without the addition of each manipulation. No other alternative task or observation time was prepared. The time duration for observing the limb and the observing material (the wooden and virtual sticks) differed across conditions. The participants were allotted 10 s to observe the (manipulated) limb before filling out the questionnaire or receiving the pain stimulus under all conditions.

We investigated eight conditions, namely, 2 (alpha value = 1 or 0.25) × 2 (with or without finger movement) × 2 (with or without pass-through) in Experiment 1. In Experiment 2, wrist touch replaced the finger movement as follows: 2 (alpha value = 1 or 0.5) × 2 (with or without wrist touch) × 2 (with or without pass-through).

Electrical stimulation.

We used an electrical stimulus to induce pain, because we targeted the analgesic effect in case of brief acute painful sensations, such as the sting of a needle (e.g., injection). The stimulus was a direct current boosted by a Cockcroft–Walton circuit and controlled by a laptop computer through an input/output board (RBIO-2U; Kyohritsu Electronic Industry Co., Ltd.). We then fixed a conductor (diameter: 0.12 mm, 10 cores) to a 1-mm thick rubber sheet and applied the current to the conductor in order to trigger pain. The intensity of the electrical stimulation was 320 V at a current of 1.8 mA, which was generally deemed tolerable, with a pulse width of 0.15 s. We used a fixed intensity to avoid multiple stimulations for determining a subjectively equivalent point. We attached the device at the midpoint between the ulnar styloid and the epicondylus medialis on the radial side of the forearm.

Questionnaire.

We collected data on body representation using a questionnaire. Table 1 presents the items, which included those based on previous works and our pilot study. In the latter, the participants who did not participate in Experiment 1 or 2 were asked about how they felt about their limb at various alpha values. Based on the answers, we included items about transparency and intangibility. We treated the two constructs separately, because we expected that transparency and intangibility would decrease perceived pain more than only the transparency feeling is high.

Download:

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

Table 1. Items in the questionnaire.

https://doi.org/10.1371/journal.pone.0268618.t001

Items 9 and 20 directly asked about the feeling of their limb becoming transparent. Item 2 pertained to the sparsity of the limb. If an object’s density is sparse, we can see the scene behind it identical to a transparent object. Items on intangibility focused on the feeling that the limb could not be touched: pass through (item 5) and empty (item 18). As representative instance of an object that cannot be touched, item 12 “I feel as if I were a ghost” was added.

As previously described, previous studies demonstrated that the sense of body ownership influenced the subjective perception of pain intensity. Thus, we assessed their level of ownership over an observed arm using items from Botvinick and Cohen [18] and Kalckert and Ehrsson [26] (items 6, 16, and 19). We replaced the term “rubber hand” with “observed arm.” We also added items to assess the perception of the reality of the observed limb (items 3 and 11).

Anxiety is also known as a feeling that increases the intensity of perceived pain and decreases pain threshold [27, 28]. To measure anxiety, we used several terms from the Anxiety Sensitivity Index [29] and State–Trait Anxiety Inventory [30]: anxious, ill (antonym of comfortable), calm (inverted scale), and fear (antonym of calm and secure). Items 8, 10, 13, and 15 evaluated how strongly each construct was perceived upon observing the manipulated limb.

Additionally, Senna et al. [4] demonstrated that feelings of hardness and heaviness of the body, which are described as stiffness similar to stone, influenced pain perception. We adapted items from the said study on hardness, sensitivity (inverted scale), numbness, and heaviness (items 4, 7, 14, and 17, respectively). Item 1 pertains to the strength of the limb and was added, because a stone (marble) is stronger than human arms. We named the construct that represented these items as “stiffness” after Senna et al. [4].

All items were prepared in the Japanese language. Items were rated using a seven-point Likert-type scale (from 1: not at all to 7: extremely). The order of items was indicated in Table 1.

Questionnaire phase.

In this phase, a questionnaire was used to investigate body representation. The participants wore the HMD and placed the non-dominant forearm on a desk (Fig 1A). At this point, all participants observed their forearm with an alpha value of 1.0 under all conditions. Afterward, the manipulations were conducted according to the condition of the trial. In Experiment 1, the order of manipulations is as follows: change in alpha value, pass-through manipulation, and movement manipulation. We followed this sequence because pass-through manipulation may influence the sense of body ownership. In Experiment 2, the touch manipulation was followed by the alpha-value manipulation, whereas the pass-through manipulation was performed at the last. We used this sequence to avoid conflict between the touch and pass-through manipulations.

After all manipulations, the participants observed their (manipulated) forearm for 10 s and answered the questionnaire. Table 1 presents the questionnaire items. We repeated this procedure for all conditions, which resulted in eight evaluations for each participant. The order of the conditions was randomized.

Pain evaluation phase.

Pain evaluation was conducted after a more than five-minute rest. At the start of the pain evaluation phase, we verified whether the participants felt pain with the electrical stimulation. The participants observed their limbs without the HMD and were given the electric stimulus, which is the same as that used in the experiment. If the participants reported that they did not feel pain, then the experiment was terminated and he/she was excluded from the study. In Experiment 2, the participants also evaluated pain intensity induced by this test stimulation using the same method for pain evaluation, which was used as the baseline evaluation (for details, see the Analysis section).

In the evaluation, the procedure was identical to that in the questionnaire phase up to the moment where the participants observed their manipulated limb. Before the HMD shut down, an electrical stimulus was applied simultaneously with a verbal cue. The participants removed the HMD and assessed pain intensity using the visual analog scale. We prepared a 100-mm line with the left and right parts representing “no pain” and the “worst possible pain.” An x was used to mark the point that reflects the level of perceived pain. All the participants experienced all conditions once in random order with a rest of approximately 2 min between trials.

Both phases were completed in one session of 90 min.

Ethical approval.

The Ethics Review Committee for Medical and Health Research Involving Human Subjects of the Ritsumeikan University approved the study (BKC-人医-2018-033). The individual in this manuscript has given written informed consent (as outlined in PLOS consent form) to publish these case details and their images (Fig 1) alongside this manuscript.

Modification of body representation.

Data from one participant who did not complete all conditions were excluded from the following analyses. Fig 2 displays the changes in the scores for each feeling and each condition in Experiment 1. We conducted a 2 (alpha value) × 2 (movement manipulation) × 2 (pass-through manipulation) ANOVA. The results indicate that the three-way interaction was non-significant for any feeling (Fs (1, 19) < 1.425, ps > 0.248, ).

Download:

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

Fig 2. Averaged scores for each feeling and each condition in Experiment 1 (n = 20).

Each point is the average score of all items for each feeling. On the x-axis, data are grouped according to the alpha-value (1.0 or 0.25), pass-through (with or without pass-through), and movement (with or without movement) manipulations. Error bars represent the standard errors.

https://doi.org/10.1371/journal.pone.0268618.g002

Ownership. The two-way interaction between the alpha-value and pass-through manipulations was significant (F(1, 19) = 14.696, p = 0.001, ). The decrease in alpha value decreased the ownership score for the pass (F(1, 19) = 13.460, p = 0.002, ) and no-pass (F(1, 19) = 39.268, p < 0.001, ) conditions. Moreover, the pass-through manipulation decreased the ownership score only at an alpha value of 1.0 (alpha value = 1.0: F(1, 19) = 11.739, p = 0.003, ; alpha value = 0.25: F(1, 19) = 0.609, p = 0.445, ). The main effect of the alpha value was significant (F(1, 19) = 27.753, p < 0.001, ). No other interaction or main effect was significant (Fs (1, 19) < 3.265, ps > 0.085, ).

Transparency. Only the main effect of the change in alpha value was significant (F(1, 19) = 355.627, p < 0.001, ). As the alpha value decreased, the participants experienced a stronger feeling of transparency. The other effects and interactions were non-significant (F(1, 19) < 3.450, ps > 0.075, ).

Intangibility. The interaction between the alpha-value and pass-through manipulations was significant (F(1, 19) = 5.794, p = 0.026, ). The small alpha value significantly increased the intangibility score regardless of the pass-through manipulation (F(1, 19) > 29.000, ps < 0.001, ). Moreover, the pass-through manipulation increased intangibility less at an alpha value of was 0.25 (F(1, 19) = 12.865, p = 0.002, ) compared with 1.0 (F(1, 19) = 46.894, p < 0.001, ). The score for intangibility resulted in a significant main effect on the alpha-value (F(1, 19) = 48.063, p < 0.001, ) and pass-through (F(1, 19) = 50.905, p < 0.001, ) manipulation. No other main effect or interaction reached significance (F(1, 19) < 0.125, ps > 0.730, ).

Anxiety. A similar tendency was observed for anxiety in terms of the interaction between the alpha-value and pass-through (F(1, 19) = 13.050, p = 0.002, ) manipulations, which is significant, as well as the main effect of the two factors (alpha-value manipulation: F(1, 19) = 17.406, p < 0.001, ; pass-through manipulation: F(1, 19) = 7.140, p = 0.015, ). However, the simple main effect of the alpha-value manipulation differed from that of intangibility, that is, alpha value = 0.25 increased anxiety only without pass-through manipulation (F(1, 19) = 23.102, p < 0.001, ). With the pass-through manipulation, the score remained high and uninfluenced by the alpha value (F(1, 19) = 2.801, p = 0.111, ). The pass-through manipulation increased anxiety only at alpha value = 1.0 (F(1, 19) = 14.363, p = 0.001, ) and not at alpha value = 0.25 (F(1, 19) = 0.008, p = 0.930, ). We found no significant main effect or interaction (Fs (1, 19) < 2.190, ps > 0.155, ).

Stiffness. No interaction or main effect was significant for the stiffness score (Fs (1, 19) < 2.560, ps > 0.125, ).

These results indicate that the change in alpha values largely influenced nearly all feelings, with the strongest effects observed for transparency (; others: ). The alpha-value and pass-through manipulations influenced the scores for ownership, intangibility, and anxiety. The decrease in alpha value also decreased the ownership score, whereas the pass-through manipulation decreased the ownership score at alpha value = 1.0. The intangibility score increased with the decrease in alpha value, where the pass-through manipulation further increased intangibility. The anxiety score at alpha value = 1.0 and without pass-through manipulation was less than those under the other conditions.

Pain.

Fig 3 depicts the pain scores under each condition. If the intensity of perceived pain was identical to that obtained under the without manipulation condition (i.e., alpha value = 1.0 without the pass-through and movement manipulations), the score reached 0. We performed a 2 (alpha value) × 2 (move manipulation) × 2 (pass-through manipulation) ANOVA on the pain score. The three-way interaction did not reach significance (F(1, 19) = 0.064, p = 0.803, ), whereas the two-way interaction between the alpha-value and movement manipulations was significant (F(1, 19) = 5.151, p = 0.035, ). Alpha value = 0.25 significantly decreased the pain score when the participants did not move their finger (with movement manipulation: F(1, 19) = 1.827, p = 0.192, ; without movement manipulation: F(1, 19) = 12.991, p = 0.002, ). The movement manipulation did not exhibit a significant effect regardless of alpha value (Fs (1, 19) < 3.655, ps > 0.070, ). The main effect of the alpha value was significant (F(1, 19) = 9.984, p = 0.005, ), whereas those of the two other manipulations (Fs (1, 19) < 0.940, ps > 0.345, ) and two-way interactions were statistically non-significant (Fs (1, 19) < 2.651, ps > 0.115, ).

Download:

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

Fig 3. Pain scores under each condition of Experiment 1 (n = 20).

Each bar represents the average pain score. When the evaluation was identical to baseline, the pain score reached 0. Baseline evaluation was conducted without manipulation (i.e., the left-most bar in the graph). A negative score indicates that perceived pain decreased than the baseline evaluation, whereas a positive score denotes a score for perceived pain that exceeded that at baseline evaluation. On the x-axis, data are grouped according to alpha-value (1.0 or 0.25), pass-through (with and without pass-through), and movement (with and without movement) manipulations. Error bars constitute standard errors.

https://doi.org/10.1371/journal.pone.0268618.g003

Analyses of the questionnaire score did not demonstrate a significant interaction between the alpha-value and movement manipulations. However, pain perception seemed to be the most affected when modifying the alpha value (Fig 3). Therefore, transparency may be the feeling that is most related to perceived pain.

We calculated the repeated measures correlation between the scores for each feeling and the pain score. Table 2 displays the correlation coefficients for all combinations. A significantly positive correlation between body ownership and pain score was detected (r(139) = 0.198, p = 0.019), whereas a significantly negative correlation was noted between transparency and the pain score (r(139) = −0.280, p < 0.001). The strongest correlation was obtained between transparency and pain score. The analysis also revealed a correlation between the majority of feelings, with a particularly strong correlation between body ownership and transparency (r(139) = −0.707, p < 0.001).

These results suggested a decrease in perceived pain and an increase in transparency. However, examining the effect of transparency independently from body ownership was difficult, because the alpha-value manipulation exhibited a strong effect on both feelings. Then, we modified the experimental set-up in Experiment 2 to reveal differences in the scores for body ownership with alpha-value manipulation.

Download:

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

Table 2. Correlation coefficients between the feelings and pain score in Experiment 1 (n = 20).

https://doi.org/10.1371/journal.pone.0268618.t002

Modification of body representation.

Fig 4 indicates the changes in the scores for each feeling and each condition in Experiment 2. In summary, changes in the alpha-value and pass-through manipulations nearly equally influenced the scores for ownership and intangibility. The alpha value exhibited a very strong effect on transparency, which was also slightly influenced by the pass-through manipulation. Each manipulation influenced anxiety in a complex manner. In general, the participants felt anxious when the alpha value was set to 0.5. At alpha value = 1.0, the pass-through manipulation increased anxiety. No manipulation exerted a significant effect on stiffness.

Download:

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

Fig 4. Averaged scores for each feeling and each condition in Experiment 2 (n = 26).

Each point is the average score of all items for each feeling. On the x-axis, data are grouped according to alpha-value (1.0 or 0.5), pass-through (with or without pass-through), and touch (with or without touch) manipulations. Error bars represent standard errors.

https://doi.org/10.1371/journal.pone.0268618.g004

Ownership. The main effects of the alpha-value (F(1, 25) = 36.056, p < 0.001, ) and pass-through (F(1, 25) = 17.890, p < 0.001, ) manipulations were significant. The decrease in alpha value and the addition of the pass-through manipulation decreased the score for ownership. No other significant main effect or interaction was detected (Fs (1, 25) < 2.950, ps > 0.095, ).

Transparency. The three-way interaction was non-significant (F(1, 25) = 1.391, p = 0.249, ). The interaction between the alpha-value and touch manipulations reached significance (F(1, 25) = 4.254, p = 0.0497, ). Although the simple main effect of the alpha value was smaller with the touch manipulation, the decrease of the alpha value greatly increased the score for transparency with (F(1, 25) = 221.068, p < 0.001, ) and without (F(1, 25) = 361.890, p < 0.001, ) the touch manipulation. The effect of the touch manipulation was non-significant for both alpha-value conditions (Fs (1, 25) < 1.300, ps > 0.270, ). Moreover, the main effects of the alpha-value (F(1, 25) = 309.560, p < 0.001, ) and pass-through (F(1, 25) = 8.774, p = 0.007, ) manipulations were significant. All other effects were non-significant (Fs (1, 25) < 4.145, ps > 0.052, ).

Intangibility. The three-way interaction did not reach significance (F(1, 25) = 2.874, p = 0.103, ), whereas the interaction between the alpha-value and pass-through manipulations was significant (F(1, 25) = 7.462, p = 0.011, ). The pass-through manipulation increased the score for intangibility significantly regardless of alpha value, although its effect was smaller at alpha value = 0.5 (F(1, 25) = 28.432, p < 0.001, ) than alpha value = 1.0 (F(1, 25) = 43.039, p < 0.001, ). The effect of the alpha value was smaller when associated with the pass-through manipulation, but the decrease in alpha value significantly increased the scores for intangibility with (F(1, 25) = 17.636, p < 0.001, ) and without (F(1, 25) = 60.501, p < 0.001, ) the pass-through manipulation. The significant main effect indicated that the alpha-value (F(1, 25) = 42.149, p < 0.001, ) and pass-through (F(1, 25) = 55.921, p < 0.001, ) manipulations influenced the score for intangibility. The main effect of the touch manipulation and other interactions were non-significant (Fs (1, 25) < 2.875, ps > 0.102, ).

Anxiety. The three-way interaction was significant (F(1, 25) = 11.075, p = 0.003, ). Table 3 summarizes the results of the follow-up analyses. In general, the score for anxiety remained high with the pass-through manipulation. In contrast, however, this score remained high at alpha value = 0.5 without the pass-through manipulation. At alpha value = 1.0 without the pass-through manipulation, the score for anxiety was lower than those observed under other conditions. The effect of the touch manipulation was small: at alpha value = 0.5, the pass-through manipulation did not increase the anxiety score without the addition of the touch manipulation.

Download:

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

Table 3. Results of follow-up analyses for the three-way interaction in anxiety in Experiment 2.

https://doi.org/10.1371/journal.pone.0268618.t003

The interaction between the alpha-value and pass-through manipulations (F(1, 25) = 18.065, p < 0.001, ) and their main effect (F(1, 25) = 8.834, p = 0.007, and F(1, 25) = 47.148, p < 0.001, , respectively) reached significance, whereas the opposite is true for the others manipulations (Fs (1, 25) < 0.176, ps > 0.678, ).

Stiffness. Similar to the result of Experiment 1, no significant interaction or main effect was observed in the score for stiffness (Fs (1, 25) < 3.915, ps > 0.059, ).

Pain.

Fig 5 displays the pain scores for each condition. If the score was zero, then the intensity of perceived pain under a given condition was identical that evaluated with full ownership sense and no transparency without the HMD. A three-way ANOVA revealed the significant main effect of the alpha-value manipulation (F(1, 25) = 5.691, p = 0.025, ). The pain score decreased at alpha value = 0.25. All other effects were non-significant (Fs (1, 25) < 2.750, ps > 0.110, ).

Download:

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

Fig 5. Pain scores for each condition in Experiment 2 (n = 26).

Each bar represents the average pain scores. When the evaluation was identical to baseline, then the pain score reached 0. Moreover, the baseline was measured without the HMD prior to all experimental conditions. A negative score indicates that perceived pain decreased more than the baseline evaluation, whereas a positive score denotes that perceived pain increased more than the baseline evaluation. On the x-axis, data are grouped according to the alpha-value (1.0 or 0.5), pass-through (with and without), and touch (with and without) manipulations. Error bars denote standard errors.

https://doi.org/10.1371/journal.pone.0268618.g005

Considering its empirical uses, examining whether perceived pain increased, decreased, or remained from the baseline, which is measured without the HMD, is important. Toward this end, we compared the pain score for each condition to a score of 0. The reason for this notion is that if the pain intensity did not differ from baseline, then the pain score should be 0. The pain score was significantly less than 0 at alpha value = 0.5 with the touch manipulation and without the pass-through manipulation (t(25) = 3.255, p = 0.003, d = 0.638). In other conditions, the pain score did not significantly differ from 0 (ts (25) < 1.755, ps > 0.090, ).

The results indicated that the alpha-value manipulation influenced the pain score, which also mainly affected body ownership, transparency, and intangibility. The pass-through manipulation exhibited a strong effect on the scores for ownership and intangibility but not the pain score. Moreover, the alpha-value manipulation influenced anxiety to a certain extent. However, the pass-through manipulation exhibited a stronger effect on anxiety. The results suggested that transparency exerted the strongest effect on the intensity of perceived pain.

We calculated the correlations between the scores for each feeling and the pain score (Table 4). The strongest negative correlation was observed between transparency and pain score (r(181) = −0.268, p < 0.001). The other feelings exerted significant but weaker correlations to the pain score compared with transparency, that is, intangibility and anxiety were negatively correlated to pain (intangibility: r(181) = −0.175, p = 0.018; anxiety: r(181) = −0.167, p = 0.024), whereas ownership and stiffness exhibited a positive correlation with pain score (ownership: r(181) = 0.154, p = 0.038; stiffness: r(181) = 0.168, p = 0.023). The correlation among the feelings, except for stiffness, was strong.

Download:

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

Table 4. Correlation coefficients between the feelings and pain score in Experiment 2 (n = 26).

https://doi.org/10.1371/journal.pone.0268618.t004

Additionally, the pain score was smaller than that at baseline when alpha value = 0.5 with the touch manipulation and without the pass-through manipulation. As shown in Fig 4, under this condition, the scores for ownership and transparency were relatively high. This result suggested that transparency decreased perceived pain, particularly when the participants felt that the observed limb belonged to their body. In contrast, the effect of transparency seemed to be weakened with the decrease in body ownership.

Observing one’s body.

Previous studies demonstrated that observing one’s body decreased perceived pain [31–33]. The participants also felt less pain when they perceived that an artificial or virtual body/body part belonged to their body (i.e., high ownership; [19–21, 34–39]). This effect implies a negative correlation between pain and body ownership. However, in the current study, the alpha-value manipulation decreased the scores for ownership and perceived pain. Thus, the analgesic effect triggered by observing one’s body does not explain the results of the experiment.

However, this analgesic effect may have occurred in all conditions because the participants observed their limb throughout the experiments. The pain score in Experiment 2 did not increase from baseline values, which were obtained without the HMD, that is, when the participants observed their bodies with full ownership. The analgesic effect was undetected due to the lack of a control condition, that is, the participants did not observe their limbs. The investigation of the analgesic effect of the observation of one’s body is beyond the scope of the study but may constitute an important topic of investigation in the future.

Immersion in a virtual environment.

Evidence exists that immersive and interactive virtual reality environments can decrease perceived pain [40, 41]. For example, Hoffman et al. [40] demonstrated this effect in the treatment of burn injuries. The study found that the analgesic effect was stronger when the VR environment was more immersive. In other words, the patients allocated more attention to the immersive environment and were distracted from the painful situation in the real world.

Our experiments did not measure the feeling of immersion. However, the score for body ownership may reflect this feeling. The alpha-value manipulation rendered the situation less plausible, which resulted in low levels of body ownership. This result suggested that the world became less immersive. However, in Experiment 2, the pass-through manipulation influenced body ownership but not the pain score. When the limb was touched without pass-through manipulation, a certain level of ownership and immersion were retained but did not influence the pain score. Therefore, the attentional factor discussed by previous studies [40, 41] did not offer a suitable explanation for the current results. Additionally, the participants observed their limb in the experiments. If they focused their attention to the scene being observed, that meant that they focused on their biological limbs. Therefore, they were not distracted from the painful stimulus. Nevertheless, we assume that the attentional factor may be related to the analgesic effect in the current study, which will be discussed later.

Body state.

Previous studies also demonstrated the effect of body state on pain perception. For instance, the sensation of pain was increased for an injured virtual limb [42, 43], whereas the pain threshold to heat was decreased for a virtual limb with red skin [10]. In addition, many researchers investigated the effect of size [12, 13]. At the level of the nervous system, the GSR to an approaching threatening stimulus changed when the participants believed that their limb was made of marble based on sound manipulation [4]. In the current experiment, three scores represented the mental body state, namely, transparency, intangibility, and stiffness. We added stiffness following Senna et al. [4]. However, the manipulations in the experiment did not target stiffness, which remained the same under all conditions. The results indicated that transparency was the feeling that is most suitable for explaining perceived pain. As the participants’ feeling of transparency decreased, perceived pain decreased. However, intangibility did not increase the analgesic effect.

The sense of body ownership was deemed to modulate the effect of transparency. Indeed, the decrease in perceived pain was greater with high levels of body ownership. We hypothesized that transparency decreased perceived pain, whereas low levels of body ownership decreased the analgesic effect induced by observing the limb, which resulted in high levels of perceived pain. In addition, the possibility exists that body representation was less modified in participants with low levels of body ownership because they less considered the observed limb as their own, which decreased the effect of transparency on perceived pain. In a previous study, the level of pain was low for an injured virtual limb if no body illusion procedure was conducted [43]. Romano et al. [13] illustrated that only when a virtual body was embodied can body size influence skin conductance response. Thus, further investigation is required to reveal the relationship between body ownership and change in the mental body model.