The threat of an embodied virtual belly affects gastric feelings depending on somatic, interoceptive and alexithymic characteristics of healthy women

A pre-registered (see the Data availability statement section) a priori power analysis was performed based on an average effect size ( r = .28) computed from the effect of embodiment on the neural correlates of cardiac interoception 18 ( r = .37) and on the relationship between disembodiment and spontaneous sensations 59 ( r = .19). The analysis recommended a minimal sample size of 18 participants for a repeated-measures analysis of variance (ANOVA; 1 group, 4 measures, α = 0.05, p = .80). A larger minimal sample size was considered to anticipate potential outlier’s removal without compromising statistical power. Twenty-eight female students were recruited via a public online website recensing the experiments of the psychology department of the University and completed the experiment (age: 18–45 years; M  = 22.1; SD  = 6.8). None of them had ever suffered from diagnosed gastric disorders, eating disorders, endometriosis or medical phobia (e.g., syringe, injection), nor any medical or psychiatric condition contraindicated for the usage of virtual reality. Prior to the experimental session, all participants but two completed online anonymized questionnaires using LimeSurvey free online tool to measure psychological traits (Table  2 ): the State Trait Anxiety Inventory – YB version (STAI-YB 60 ), the Multidimensional Assessment of Interoceptive Awareness (MAIA-2 61 ), the Somatic Symptom Scale (SSS-8 62 ), and the Bermond-Vorst Alexithymia Questionnaire version B (BVAQ-B 63 ; see Questionnaire section for details). All protocol instructions and questionnaires were in French. The study was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and was approved by the local ethics committee of the Université Savoie Mont Blanc (USMB-CER 2023-13-BELIV). All participants provided their informed and written consent to participation after the experimental procedure was explained. Participants were granted with course credits. Table 2 Descriptive statistics of psychological traits ( N  = 26). Psychological traits Mean SD Range Internal consistency (McDonald’s ω) STAI-YB 49.7 11.3 32.0–67.0 0.93 MAIA-2 noticing 3.3 0.7 2.0–4.5 0.51 MAIA-2 non-distracting 2.2 0.8 1.0–3.5 0.79 MAIA-2 not-worrying 2.5 0.9 1.0–4.8 0.83 MAIA-2 Attention regulation 2.6 0.8 1.3–4.4 0.82 MAIA-2 emotional awareness 3.5 0.8 2.0–4.6 0.77 MAIA-2 self-regulation 2.6 1.0 0.5–4.3 0.81 MAIA-2 body listening 2.3 1.0 0.7–5.0 0.80 MAIA-2 trusting 3.3 1.1 1.3–5.0 0.85 SSS-8 14.4 5.9 4.0–26.0 0.83 BVAQ emotionalizing 8.9 2.7 4.0–15.0 0.49 BVAQ fantasizing 8.4 4.1 4.0–19.0 0.86 BVAQ identifying 11.4 3.1 5.0–17.0 0.62 BVAQ describing 14.4 5.9 4.0–19.0 0.85 BVAQ analyzing 12.3 4.3 4.0–11.0 0.59 Descriptives are based on the responses of twenty-six participants due to missing data from two participants. Descriptive statistics of psychological traits ( N  = 26). Descriptives are based on the responses of twenty-six participants due to missing data from two participants. The virtual reality environment was constructed with Unity software (Unity Technology, US). It consisted of a virtual room with an armchair for a semi-reclined position, similar to that in the experimental room. Three women avatars were constructed using Mixamo (Adobe, US) and Blender (Blender Foundation, NL): the embodied-avatar, the right-avatar, and the left-avatar (Fig.  1 a). The embodied avatar was seated on the armchair with elbows on the armrests. It wore shorts and a sports bra, allowing participants to see the avatar’s naked abdomen where they expected to see their own (Fig.  1 ). A head mounted display (HMD; Oculus Meta Quest 2; Meta Quest, US) was used for immersion in a first-person perspective with the embodied avatar. The HMD’s real-time tracking system enabled us to track the positions and orientations of the participants’ head and update the visual flow accordingly. The two additional avatars were standing on either side of the embodied avatar. The right-avatar held a wooden stick in the right hand and provided visual stimulations during the embodiment phase (Fig.  1 b). The right-avatar’s right arm was moved according to the experimenter’s right arm movements tracked by the controller she held. A wooden stick was attached to the controller for the experimenter to administer tactile stimulations on the participant’s abdomen. The experimenter’s movements animated the right-avatar’s right arm either (1) in real time for the virtual wooden stick to touch the avatar’s abdomen at the same time and place as the real wooden stick touched the participant’s abdomen (i.e., synchronous visuo-tactile condition), or (2) after a delay of 2 s (i.e., asynchronous visuo-tactile condition). The duration of the delay has been chosen based on unpublished attempts to create a noticeable a visuo-tactile mismatch. The left-avatar could hold either a teaspoon (i.e., spoon condition) or a syringe (i.e., syringe condition) in the left hand (Fig.  1 b). A pre-recorded animation caused the left-avatar to approach the object towards the embodied-avatar’s skin abdomen during the threatening phase. The STAI-YB was used to assess trait anxiety. The questionnaire is composed of twenty items listing feelings related to anxiety, and individuals indicate the frequency with which they experience those in general. Items are scored on a 4-point Likert scale ranging from 1 ( almost never ) to 4 ( almost always ). Responses are summed after coding the nine reversed items. Higher scores indicate stronger trait anxiety. The MAIA-2 was used to assess the eight dimensions of interoceptive awareness: noticing (the ability to notice pleasant and unpleasant body feelings; 4 items), non-distracting (the ability to ignore unpleasant body feelings; 6 items), not-worrying (the ability to not worry about unpleasant body feelings; 5 items), attention regulation (the ability to maintain attention on body feelings in disturbing contexts; 7 items), emotional awareness (the ability to relate emotional states to body feelings; 5 items), self-regulation (the ability to regulate body feelings to relax; 4 items), body listening (the tendency to spontaneously listen to one’s bodily sensations; 3 items), and trusting (the tendency to trust body feelings; 3 items). The 37 items are scored on a 6-point Likert scale ranging from 0 ( never ) to 5 ( always ). A mean score is computed for each dimension after coding the nine reversed items. The SSS-8 was used to assess the somatic symptom burden. The eight items refer to different somatization symptoms and are scored on a 5-point Likert scale ranging from 0 ( not at all ) to 4 ( very much ). Responses are summed to compute the total score. Higher scores indicate stronger somatization. The BVAQ-B was used to assess alexithymia. Twenty items scored on a 5-point Likert scale ranging from 1 ( totally disagree ) to 5 ( totally agree ) are used to assess five dimensions: emotionalizing, fantasizing, identifying, describing and analyzing emotions. Responses are summed for each scale. Higher scores indicate greater tendency to alexithymia. Prior to the experimental session, the participants completed the STAI-YB, the MAIA-2, the SSS-8, and the BVAQ-B online. During the experimental session, the participants were seated in a semi-reclined position on an armchair and wore the HMD. Once immersed, the participants were invited to explore the virtual environment and their virtual body they saw from a first-person perspective. Then, they had two virtual reality sessions (separated by a 5-minute break outside the virtual reality) that both included the following phases: an embodiment phase followed by a syringe phase (Fig.  1 b). During the embodiment phase, participants were asked to focus on their virtual abdomen. For 90 s, participants’ abdomen was repetitively stroked by the experimenter with a wooden stick (tactile stimulations) while they saw the virtual abdomen being touched by the right virtual experimenter with a wooden stick (visual stimulations). The experimenter administered tactile stimulations over the participant’s shirt following a predefined movement pattern and speed, which were identical for all participants. Given the coupling between the experimenter’s and the right-avatar’s right arms movements, the visuo-tactile stimulations were either synchronous (real-time coupling) or asynchronous (delayed coupling) depending on the experimental condition. All participants experienced both the synchronous visuo-tactile and the asynchronous visuo-tactile conditions in a counterbalanced order and spaced with a 5-minute break during which they exited immersion. After each 90-second embodiment phase, the left-avatar was programmed to approach either a syringe (threatening object) or a teaspoon (non-threatening object) near the embodied-avatar’s abdomen. All participants experienced two trials in both the syringe and the spoon conditions, in a randomized order. Each trial consisted of approaching the object three times towards the embodied-avatar’s abdomen, each approach phase lasting 12 s and 8 s apart. After each trial, the virtual room went black, and the participants were questioned regarding their gastric feelings, state anxiety, and sense of embodiment (see Questionnaire section). First, the participants were asked to indicate the level of intensity of gastric feelings they felt during the threatening phase. The participants responded based on an 11-point Likert scale ranging from 0 ( no feeling ) to 10 ( extremely strong feeling ). Then, the participants were asked to indicate the valence of the gastric feeling. They responded on an 11-point Likert scale ranging from − 5 ( extremely unpleasant feeling ) to 5 ( extremely pleasant feeling ), with 0 being a neutral valence. Additionally, they were asked to indicate the inner or outer nature of the sensations on an 11-point Likert scale ranging from − 5 ( totally visceral feeling ) to 5 ( totally cutaneous feeling ), with 0 being neutral. Then, forward and backward female silhouettes were presented to the participants in the virtual environment, and participants were given a controller. They were asked to use the controller to circle the body area on which they felt body sensations during the threatening phase. If they did not feel any sensations, they did not circle any area. During the analysis of the responses to this task, it was observed that instructions were not understood uniformly by participants (e.g., some participants circled their body sensations by visualizing themselves in mirror into the silhouettes, others made a mental rotation), thus, these responses were not considered in our analyses. For questions relating to the intensity, the valence and the viscerality of the felt body sensation, Likert scales were projected in the HMD and the participants responded verbally. Six items from the Peck & Gonzalez-Franco’s embodiment questionnaire 64 were used to assess the sense of embodiment for the embodied avatar. Items were selected to measure the three embodiment components (i.e., body ownership, agency, self-location; see 13 for a review) and were adapted to the present protocol as recommended by the authors (e.g., “During the experiment there were moments in which I felt that my own belly could be affected by the spoon / the syringe.”). The items were read aloud by the experimenter and the participants verbally responded on a 7-point Likert scale ranging from 1 ( strongly disagree ) to 7 ( strongly agree ) that was projected in the HMD. The sense of embodiment score was computed in each visuo-tactile synchronicity condition by averaging the responses ( McDonald’s ω : M = 0.85, Range = [0.72, 0.89]). State anxiety was measured with three items 25 : “How nervous/frightened/concerned did you feel when you perceived the object approaching the virtual abdomen”. Participants verbally responded on an 11-point Likert scale ranging from 0 ( not at all ) to 10 ( extremely ) that was projected in the HMD. A mean state anxiety was computed from the 3 items ( McDonald’s ω : M = 0.97, Range = [0.93, 0.99]). In sum, participants completed questionnaires related to clinical characteristics online prior to the experimental session (STAI-YB, MAIA-2, SSS-8 and BVAQ-B). During the session, each participant experienced 8 trials: 2 trials in each threatening condition (syringe versus spoon) for each visuo-tactile condition (synchronous versus asynchronous). After each trial, questions related to the gastric feelings (intensity, valence, viscerality, localization), the state anxiety and the sense of embodiment were asked (Fig.  1 b). For all statistical analyses, the significance level was set to alpha = 0.05. McDonalds’ ω was computed using JAMOVI software (version 2.3.28; The jamovi project (2025) [Computer Software]. Retrieved from https://www.jamovi.org ) to evaluate the internal consistency of each questionnaire used in this study. The ANOVA and mediation analyses were performed on R software (version 4.2.1). The partial least square correlation was performed on Python software (version 3.11.7). More details about the libraries and statistical functions used per analysis are mentioned in the following paragraph specific to each analysis. Two-way repeated measures ANOVAs were performed on R software with anova_test function to test the effect of the visuo-tactile synchronicity (within-subject factor), the threat (within-subject factor) and their interaction on the sense of embodiment, the state anxiety and the intensity, valence and viscerality of gastric feelings. The normality of the residuals for each statistical model was visually checked using performance package 65 . Paired comparison analyses were performed using two-tailed paired t-tests and p -values were Holm-Bonferroni corrected for a family of all four possible contrasts 66 . Mediation analyses were performed to test if the gastric feelings modulations by the visuo-tactile synchronicity and the threat were mediated by the changes in sense of embodiment and in state anxiety. Within-subject mediation analyses were performed on R software using the JSMediation package 67 . Mediation analyses were following the joint path analysis methodology 34 , 68 which stipulates that each following effects must be observed to reliably claim for the existence of a mediation: (1) the effect of visuo-tactile synchronicity or threat or their interaction on gastric feelings (i.e., total effect); (2) the effect of visuo-tactile synchronicity or threat or their interaction on state anxiety or sense of embodiment (i.e., effect of the factor on the mediator); (3) the effect of the changes in state anxiety or sense of embodiment between the visuo-tactile conditions or the threatening conditions on the changes of gastric feelings (i.e., effect of the mediator on the measure). Therefore, mediation analyses were only performed when both gastric feelings and state anxiety or sense of embodiment were modulated by visuo-tactile synchronicity or threat or their interaction. The 95% confidence internal of the estimate of the indirect effect, that is (1) the effect of visuo-tactile synchronicity or threat on the state anxiety or sense of embodiment (mediators) and then (2) the effect of the changes in the level of the mediator on gastric feelings, was estimated with 1000 Monte-Carlo simulations 67 . The absence of mediation effect cannot be rejected if 0 is included in the confidence interval. Then, if a mediation effect is indeed identified, it is qualified as “partial” if the effect of visuo-tactile synchronicity, threat or their interaction on gastric feelings after controlling for the effect of changes in the level of state anxiety or sense of embodiment (i.e., direct effect) decreases but remains significant, and it is qualified as “total” if such effect became not significant. As an exploratory analysis, a descriptive partial least square correlation analysis (PLS-C) was performed to identify whether gastric feelings modulations correlate with psychological traits. Let X be the psychological traits (trait anxiety, interoceptive awareness, alexithymia, somatization) and Y the gastric feelings (intensity, valence and viscerality and their changes between the experimental conditions). PLS-C is a statistical multidimensional method that consists of modeling latent variables for the X and Y datasets that share maximal covariance using singular value decomposition. It is well suited to datasets where the number of variables is larger than the sample size (see 69 , 70 for reviews and technical details about this method). The latent variables obtained with the singular value decomposition are a linear combination of the X and Y dimensions whose coefficients are given by the so-called “weights”. The loadings of each dimension in X and Y composing the latent variables corresponds to the Pearson’s correlation with their respective latent variable. In the present analyses, PLS-C was performed on dimensions of the STAI-YB, BVAQ-B, MAIA-2 and SSS-8 questionnaires resulting in 22 dimensions for the psychological variables. For the gastric feelings dimensions (i.e., intensity, valence and viscerality), we computed: (1) the overall average score, (2) the difference score between the syringe and the spoon conditions, (3) the difference score between the synchronous and asynchronous visuo-tactile conditions, and (4) the difference between the two differences expressed in points (2) and (3) (to obtain the interaction effect of visuo-tactile synchronicity and threat) resulting in 12 dimensions for the gastric feelings. The PLS-C was performed on Python with the behavioral_pls function from the pyls package 71 using the default parameters. The psychological traits and gastric feelings variables were standardized before applying the PLS-C in order to avoid any influence of the measuring scale on the variance to be captured by the latent variables. To determine if the association extracted by the PLS-C was relevant, the significance of the correlation between psychological and gastric feelings latent variables was tested compared to a distribution of correlations built by applying PLS-C under the null hypothesis based on 5000 sample permutations. To determine the loading reliability of each dimension in their respective latent variables, their 95% confidence interval was estimated using a bootstrapping approach with 5000 resampling. As a result, if significant correlations between the psychological traits and gastric feelings latent variables were identified, the most reliable dimensions of each latent variable could be determined.

Descriptive statistics supporting the analyses regarding the sense of embodiment, state anxiety and gastric feelings according to the experimental conditions are available in Table  1 . Table 1 Descriptive statistics of the sense of embodiment, state anxiety and gastric feelings ( N  = 28). Variables Visuo-tactile condition Threat condition Mean SD Range Sense of embodiment Synchrony Spoon 5.23 1.04 2.42; 6.67 Syringe 5.40 1.01 2.33; 7.00 Asynchrony Spoon 4.07 1.26 1.42; 6.67 Syringe 4.19 1.22 1.75; 6.75 State anxiety Synchrony Spoon 1.05 1.38 0; 4.17 Syringe 4.04 2.97 0; 8.67 Asynchrony Spoon 1.14 1.47 0; 5.50 Syringe 3.37 2.98 0; 8.34 Gastric feelings intensity Synchrony Spoon 2.55 1.95 0; 6.50 Syringe 4.09 2.71 0; 9.50 Asynchrony Spoon 2.46 2.14 0; 6.00 Syringe 3.14 2.50 0; 7.50 Gastric feelings valence Synchrony Spoon 0.07 1.11 − 2.00; 4.00 Syringe − 1.43 1.71 − 4.50; 4.00 Asynchrony Spoon 0.20 0.94 − 1.50; 3.00 Syringe − 0.91 1.59 − 3.50; 4.00 Gastric feelings viscerality Synchrony Spoon 0.63 2.06 − 3.50; 4.50 Syringe − 0.32 2.57 − 4.50; 3.50 Asynchrony Spoon 0.88 1.22 − 1.50; 4.00 Syringe − 0.14 2.25 − 4.50; 4.50 Descriptive statistics of the sense of embodiment, state anxiety and gastric feelings ( N  = 28). As expected, synchronous visuo-tactile stimulations applied on the participant’s and the avatar’s abdomen induced a significant higher sense of embodiment over the avatar compared to asynchronous visuo-tactile stimulations ( F (1, 27) = 37.0, p =  1.7 × 10 − 6 , η 2 G  = 0.221). However, there was no main effect of the threat condition ( F (1, 27) = 3.8, p =  .062, η 2 G  = 0.004) nor interaction effect between visuo-tactile synchronicity and threat on the sense of embodiment ( F (1, 27) = 0.17, p =  .68, η 2 G  < 0.001; Fig.  2 a). Additionally, the state anxiety was significantly higher in the syringe condition compared to the spoon condition ( F (1, 27) = 46.7, p =  2.5 × 10 − 7 , η 2 G  = 0.245). There was no main effect of the visuo-tactile synchronicity ( F (1, 27) = 0.73, p =  .399, η 2 G  = 0.004) nor interaction effect with threat on state anxiety ( F (1, 27) = 2.6, p =  .120, η 2 G  = 0.007; Fig.  2 b). Considering the efficiency of the protocol to modulate the sense of embodiment and the state anxiety, we then tested whether it affected the gastric feelings with repeated measure within-subject analyses of variance (ANOVA). There was a significant interaction effect between visuo-tactile synchronicity and threat on the intensity of the gastric feeling ( F (1, 27) = 7.4, p =  .011, η 2 G  = 0.009): intensity was higher in the syringe condition than in the spoon one in the synchronous visuo-tactile condition (Intensity[syringe – spoon] synchronous , t (27) = 4.36, p Holm = 6.9 × 10 − 4 ) but no significant difference was observed in the asynchronous one (Intensity[syringe – spoon] asynchronous , t (27) = 2.05, p Holm = 0.101). Moreover, the gastric feeling was significantly more intense in the synchronous visuo-tactile condition compared to the asynchronous one but exclusively when the virtual abdomen was threatened (Intensity[synchronous – asynchronous] syringe , t (27) = 2.85, p Holm = 0.025 ; Intensity[synchronous – asynchronous] spoon , t (27) = 0.29, p Holm = 0.776; Fig.  2 c). The gastric feeling was felt significantly more unpleasant in the synchronous visuo-tactile condition compared to the asynchronous one ( F (1, 27) = 4.63, p =  .040, η 2 G  = 0.014) as it was in the syringe condition compared to the spoon one ( F (1, 27) = 32.38, p =  4.8 × 10 − 6 , η 2 G  = 0.189). There was no significant interaction effect between visuo-tactile synchronicity and threat on the valence of the gastric feeling ( F (1, 27) = 2.94, p =  .098, η 2 G  = 0.005; Fig.  2 d). The gastric feeling was felt significantly more visceral than cutaneous in the syringe condition compared to the spoon one ( F (1, 27) = 12.3, p =  .002, η 2 G  = 0.054). In the spoon condition, the gastric sensation was perceived as significantly more cutaneous than neutral ( t (27) = 2.68, p =  .012, d  = 0.51) what was not observed in the syringe condition ( t (27) = -0.55, p =  .59, d  = 0.10). There was no main effect of the visuo-tactile synchronicity ( F (1, 27) = 0.79, p =  .383, η 2 G  = 0.003) nor interaction effect between visuo-tactile synchronicity and threat on the viscerality of the gastric feeling ( F (1, 27) = 0.03, p =  .875, η 2 G  < 0.001; Fig.  2 e). Fig. 2 Graphical representation of the ( a ) mean sense of embodiment, ( b ) mean state anxiety, ( c ) mean intensity of gastric feelings, ( d ) mean valence score of the gastric feelings, and ( e ) mean viscerality score of the gastric feelings as a function of the visuo-tactile synchronicity condition (synchronous versus asynchronous visuo-tactile stimulations) and the threat condition (syringe versus spoon). Error bars represent the 95% confidence interval. The “#”, “*”, and “X” symbols represent the significant effects of the visuo-tactile synchronicity, of the threat and of the interaction between both respectively. Statistical details about multiple comparisons are available in the text. # and X: p  < .05; **, p  < .01; *** and ###: p  < .001. Graphical representation of the ( a ) mean sense of embodiment, ( b ) mean state anxiety, ( c ) mean intensity of gastric feelings, ( d ) mean valence score of the gastric feelings, and ( e ) mean viscerality score of the gastric feelings as a function of the visuo-tactile synchronicity condition (synchronous versus asynchronous visuo-tactile stimulations) and the threat condition (syringe versus spoon). Error bars represent the 95% confidence interval. The “#”, “*”, and “X” symbols represent the significant effects of the visuo-tactile synchronicity, of the threat and of the interaction between both respectively. Statistical details about multiple comparisons are available in the text. # and X: p  < .05; **, p  < .01; *** and ###: p  < .001. Considering that gastric feelings were partly influenced by visuo-tactile synchronicity and threat, we then investigated to what extent gastric feelings differences were explained by changes in the levels of state anxiety and of sense of embodiment. Thus, joint path analyses were conducted to test whether state anxiety or sense of embodiment mediate gastric feelings. The state anxiety changes between the threatening conditions significantly predicted the gastric feeling intensity changes ( t (25) = 2.87, p =  .008). After controlling for this effect, the effect of threat on gastric feeling intensity changes was no longer significant ( t (25) = 0.67, p =  .510, indirect effect estimate: CI 95% [0.35, 2.18]). These results indicate that in the synchronous visuo-tactile condition, the more intense gastric feelings observed in the syringe condition compared to the spoon condition is entirely mediated by the increase in state anxiety (Fig.  3 a). The sense of embodiment changes between the visuo-tactile conditions significantly predicted the gastric feeling intensity changes ( t (25) = 5.54, p <  .001). After controlling this effect, the effect of visuo-tactile synchronicity on gastric feeling intensity changes was no longer significant ( t (25) = 1.42, p =  .167 indirect effect estimate CI 95% [0.78, 2.21]). Thus, in the syringe condition, the more intense gastric feelings observed in the synchronous visuo-tactile condition compared to the asynchronous one is entirely mediated by the increase in the sense of embodiment (Fig.  3 b). First, the sense of embodiment changes between the visuo-tactile conditions significantly predicted the gastric feeling valence changes ( t (25) = 2.69, p =  .012). After controlling for this effect, the effect of the visuo-tactile synchronicity on gastric feeling valence changes was no longer significant ( t (25) = 0.57, p =  .575 indirect effect estimate CI 95% [− 0.83, − 0.12]; Fig.  3 c). Second, the state anxiety changes between the threatening conditions significantly predicted the gastric feeling valence changes ( t (25) = 3.27, p =  .003). After controlling for this effect, the effect of the threat on gastric feeling valence changes was no longer significant ( t (25) = 1.06, p =  .298, indirect effect estimate: CI 95% [− 1.64, − 0.37]; Fig.  3 d). In sum, the more negative gastric feeling valence reported in the synchronous visuo-tactile condition compared to the asynchronous one and in the syringe condition compared to the spoon one are entirely mediated by the increase in sense of embodiment and the increase in state anxiety, respectively. The higher state anxiety observed in the syringe condition compared to the spoon one did not significantly predict the changes in gastric feelings viscerality, ( t (25) = 1.69, p =  .103; Fig.  3 e). Therefore, the state anxiety did not mediate the more visceral localization of gastric feelings in the syringe condition. Fig. 3 Graphical summary of the within-subject mediation analyses with the corresponding path estimate. ( a ) Mediation of the effect of the threat on gastric feelings intensity by state anxiety in the synchronous visuo-tactile condition. ( b ) Mediation of the effect of visuo-tactile synchronicity on gastric feelings intensity by the sense of embodiment in the syringe condition. ( c ) Mediation of the effect of the threat on gastric feelings valence by state anxiety. ( d ) Mediation of the effect of visuo-tactile synchronicity on gastric feelings valence by the sense of embodiment. ( e ) Absence of mediation of the effect of the threat on gastric feelings viscerality by state anxiety. Readers are encouraged to refer to Montoya and Hayes (2017) for further explanations. *: p  < .05; **: p  < .01; ***: p  < .001. Graphical summary of the within-subject mediation analyses with the corresponding path estimate. ( a ) Mediation of the effect of the threat on gastric feelings intensity by state anxiety in the synchronous visuo-tactile condition. ( b ) Mediation of the effect of visuo-tactile synchronicity on gastric feelings intensity by the sense of embodiment in the syringe condition. ( c ) Mediation of the effect of the threat on gastric feelings valence by state anxiety. ( d ) Mediation of the effect of visuo-tactile synchronicity on gastric feelings valence by the sense of embodiment. ( e ) Absence of mediation of the effect of the threat on gastric feelings viscerality by state anxiety. Readers are encouraged to refer to Montoya and Hayes (2017) for further explanations. *: p  < .05; **: p  < .01; ***: p  < .001. As an exploratory attempt to capture the multidimensional relationship between psychological traits (somatization, anxiety, alexithymia, and interoceptive awareness) and gastric feelings (intensity, valence, viscerality), we performed a descriptive partial least square correlation analysis. This analysis revealed a significant correlation between the psychological traits’ latent variable and the gastric feelings latent variable ( r  = .72, p <  .001, 1000 permutations; Fig.  4 a). The most stable and reliable dimensions composing each latent variable were selected if the 95% confidence interval of their loadings did not contain 0. Therefore, the latent variable of psychological traits was: (1) positively associated with the not-worrying dimension of interoceptive awareness and the analyzing dimension of alexithymia; (2) negatively associated with somatization traits (e.g., pain in the arms / legs / joints, thoracic pain or shortness of breath, back pain and headaches; Fig.  4 b). The latent variable including gastric feelings was: (1) positively associated with the difference in gastric feelings valence between the threatening conditions, with the overall valence of gastric feelings and with the difference in gastric feelings viscerality between the threatening conditions ; (2) negatively associated with the overall intensity of gastric feelings, with the difference in intensity between the threatening conditions as well as its interaction with visuo-tactile synchronicity (Fig.  4 c). Overall, as the psychological traits and gastric feelings latent variables were positively correlated, these results suggest the more the somatic symptoms, the worry about interoceptive feelings and the difficulty to analyze emotions, the more intense, unpleasant and visceral the gastric feelings are perceived overall or under threat of the virtual abdomen. A description of the loadings is displayed in Fig.  4 d. Fig. 4 Graphical summaries of the symmetric partial least square regression analyses performed between psychological traits and gastric feelings modulations by visuo-tactile synchronicity and threat. ( a ) Correlation between the psychological and gastric feelings latent variables. ( b ) Description of the loadings and 95% confidence intervals (1000 resampling) for the dimensions composing the psychological trait latent variable and ( c ) composing the gastric feelings latent variable. ( d ) Schematic of the most stable dimensions composing the latent variables. The “+” (“−“) symbol represents a positive (negative) influence between variables. Double arrows represent a bidirectional relationship, and single arrows represent a directed one. The rounded shaped variables are the latent variables, and the squared shaped ones are the most reliable dimensions that compose them. Graphical summaries of the symmetric partial least square regression analyses performed between psychological traits and gastric feelings modulations by visuo-tactile synchronicity and threat. ( a ) Correlation between the psychological and gastric feelings latent variables. ( b ) Description of the loadings and 95% confidence intervals (1000 resampling) for the dimensions composing the psychological trait latent variable and ( c ) composing the gastric feelings latent variable. ( d ) Schematic of the most stable dimensions composing the latent variables. The “+” (“−“) symbol represents a positive (negative) influence between variables. Double arrows represent a bidirectional relationship, and single arrows represent a directed one. The rounded shaped variables are the latent variables, and the squared shaped ones are the most reliable dimensions that compose them.

This protocol offers a novel, non-invasive way to study gastric interoception awareness through contextual and bodily self-consciousness manipulations. It paves the way for larger neurophysiological studies in both healthy and clinical populations, such as those with gastrointestinal symptoms or bodily self-consciousness disorders. For example, it may help understand conditions like irritable bowel syndrome 57 or schizophrenia-related gastrointestinal disturbances 58 . Thus, further work is needed to replicated and extend this protocol and the observed associations with psychological traits to clinical populations and men.

We provide the first empirical evidence that the intensity and valence of gastric sensations in healthy women can be modulated non-invasively through virtual embodiment and contextually induced state anxiety. We further provide evidence that gastric interoception is context-sensitive and psychologically shaped: gastric feelings are not just physiological—they are also constructed by perception, threat interpretation, and affective traits. Using a virtual reality paradigm, we successfully replicated the induction of embodiment via synchronous visuo-tactile abdominal stimulations, consistent with multisensory integration theories of body ownership 13 , 16 . This embodiment enabled us to modulate gastric feelings’ intensity and valence using virtual threats applied to the avatar’s abdomen with either a syringe or a spoon. Crucially, we examined a novel “viscerality” dimension of conscious gastric interoception—how deeply sensations are experienced within the body. Importantly, the manipulation of the intensity, unpleasantness, and perceived visceral nature of gastric sensations was positively associated with somatization tendencies, concern for interoceptive states, and alexithymia-related traits (especially with difficulties in analyzing emotions) enforcing the important top-down influence on gastric interoceptive processing by personality traits even in non-clinical populations. Overall, we have shown that our protocol can manipulate conscious gastric interoception (intensity, valence, viscerality). We hypothesized that gastric feelings could emerge without a specific gut physiological state change, as it has been previously proposed 5 , 35 , 36 . The central nervous system continuously receives and processes interoceptive afferences that remain mainly unconscious 1 . By raising the bodily consciousness and threatening the embodied abdomen, we aimed at inducing top-down influences facilitating the awareness of the processing of interoceptive afferences from the stomach. Indeed, anxiety has been found to increase with gastric sensitivity 37 and stronger afferent interoceptive processing has been found to be induced by the sense of embodiment 21 . By seeing the embodied virtual abdomen threatened by the syringe, participants may have anticipated and imagined the unpleasant feeling that would be caused by the sting. Moreover, this hypothesis is supported by interoceptive inference theories (e.g. the Insula Hierarchical Modular Adaptive Interoception Control model 38 , 39 ) , which suggest that conscious interoception (here, gastric feelings), can be formed through interactions between the insula, prefrontal cortex and striatum, which are partly influenced by past interoceptive experiences, context and visceral states. In this model, the insula plays a central role in gathering both visceral and contextual information from the prefrontal and limbic regions in order to predict the bodily states and compare them with their actual ones. The larger the resulting prediction error, the more likely it is that interoceptive feelings will emerge. Still, as we did not measure any visceral activity, we cannot exclude the possibility that our protocol induced visceral state changes that favor the awareness of gastric afferences and consequently explained the gastric feelings changes we observed. Indeed, emerging evidence suggests that embodiment in virtual reality 40 as well as emotional induction 6 can alter gastric physico-chemical states (e.g., pH, motility) but how these physiological states relate to interoceptive awareness remains unclear. As many possible visceral activity changes could hypothetically take part in the emergence of gut feelings, we believe this current protocol will help to further explore the possible physiological basis of interoceptive awareness. Previous studies have shown that threatening a virtual or prosthetic body part increases sympathetic nervous system activity as a defense mechanism, especially under synchronous visuo-tactile conditions 16 , 24 . We partially replicate these findings by showing that state anxiety increases under abdominal threat. However, our results diverge in that this increase occurred regardless of visuo-tactile synchrony, suggesting that the mere first-person perspective with the avatar might be sufficient to elicit anxiety, even under visuo-tactile conflict 41 – 43 . First-person perspective may promote a sufficient level of embodiment that allows for emotional responses to be modulated by threat, independently of multisensory congruence providing by synchronous abdominal visuo-tactile stimulations. We observed that the visuo-tactile synchronicity influenced gastric intensity and valence but not anxiety, while anxiety mediated those gastric feelings. Therefore, it appears that emotion and interoception are related but still differently influenced by embodiment. We then present a hypothesis that could provide a rationale for our findings. On the one hand, the link between anxiety and gastric feelings is supported by empirical studies showing overlapping networks including the Default Mode Network (DMN) and Salience Network for interoception and emotional regulation 44 . On the other hand, despite these similarities, recent work suggests distinct neural correlates for interoceptive intensity versus anxiety when focusing on interoceptive functions 45 . Interoceptive intensity is primarily explained by the cingulo-opercular task control (COTC) network—including anterior insula, anterior cingulate cortex, and thalamus 46 —that correlate with embodiment enhancement during synchronous visuo-tactile stimulation 24 , 47 and participate in gut interoceptive processing 48 . Conversely, anxiety involves more distributed network interactions 45 . We propose that visuo-tactile synchronous stimulation enhances embodiment through COTC activation known to facilitate top-down sustained attention which, in the current study, may be oriented towards interoceptive processes and consequently enhancing gastric sensation perception 45 . This embodiment-interoception pathway may operate independently from shared emotion-interoception networks, explaining why gastric feelings, but not anxiety, were influenced by embodiment in our study. While the unpleasantness and viscerality of gastric sensations increased in response to abdominal threat regardless of visuo-tactile synchrony, changes in the perceived intensity of these sensations occurred primarily when the abdomen was embodied—i.e., under synchronous visuo-tactile stimulations. This dissociation may reflect distinct neural mechanisms underlying different dimensions of interoceptive awareness. In particular, the intensity of gastric sensations may require congruent multisensory integration to be consciously accessed or amplified, whereas valence and viscerality may be more readily influenced by contextual threat alone. Recent neuroimaging work supports this distinction, showing that specific brain networks process the valence and the intensity of interoceptive experiences in pain and pleasure 49 . Specifically, valence is associated with limbic and DMN activity, while intensity is linked to the ventral attentional network. Our findings align with this model: the valence of gastric sensations (e.g., unpleasantness) was modulated by anxiety irrespective of multisensory synchronicity, whereas the intensity of those sensations depended on visuo-tactile congruence, potentially engaging attentional mechanisms directed toward the abdomen. Indeed, attention to bodily sensations is facilitated by embodiment, which may recruit the ventral attentional network involved in the conscious amplification of interoceptive intensity, more than the limbic regions associated with emotional valence 50 . The viscerality dimension—how deeply sensations are localized—was also modulated by threat but not mediated by state anxiety. Interestingly, the object used for threat (syringe vs. spoon) influenced these ratings, even though it never touched the avatar’s skin. Following the somatic marker hypothesis 51 , this suggests that participants’ expectations, shaped by real-world experiences with these objects, influenced how they interpreted their internal sensations. Participants may have inferred a deeper or more “penetrating” sensation from the syringe, which could be explained by a top-down bias process associated with their mental representation of this object. Women who somatize more, worry more about internal states, and struggle to analyze their emotions also reported that gastric sensations felt more visceral than cutaneous. Indeed, expected interoceptive states may be more likely to reflect an interoceptive discomfort, biased by past somatization experiences compared to the actual normal interoceptive states in an anxious context 5 . This interoceptive memory bias may explain why abdominal feelings are reported as more visceral in aversive context 5 . The current protocol provides a methodological tool for studying how psychological traits—such as somatization, interoceptive worry, and alexithymia—covary with interoceptive awareness without directly manipulating gastric states. In the context of healthy women, the observed inter-individual variability in psychological traits extends and aligns with the findings reported in the literature on gastrointestinal functional disorders. In these cases, correlations have been identified between gastrointestinal hypersensitivity and alexithymia traits 32 , 52 . Intriguingly, we found that associations with gastric sensations extended beyond gastrointestinal complaints tendencies, encompassing broader somatic symptoms (e.g., chest, joint, and limb pain) what raises the question of the sensitivity of gastric feelings modulations to general somatization.

The current study has several limitations and possible improvements can be made in future studies. The conclusions we drew can only be applied to healthy women. Considering the potential interoceptive sex differences 28 , future studies are required to evaluate how the current results generalize to men. We did not adjust the avatar’s visual appearance to the body mass index of participant. Even if the embodiment illusion can be successfully induced for avatars with strong visual appearance differences 53 – 55 , personalizing the avatar visual appearance could induce stronger sense of embodiment 56 and improve the gastric feelings modulation. Additionally, we recommend that future studies investigate the importance of the first-person perspective in virtual reality for enhancing gastric sensations, independently of abdominal visuo-tactile congruency. Furthermore, the influence of satiety states on gastric sensation still needs to be evaluated. We explored to what extent the gastric feelings modulation induced by our paradigm correlates with the MAIA-2 as a measure of trait interoceptive awareness. Yet, how the modulation of gastric feelings by threat and embodiment relates to gastric feelings in response to a stomach physico-chemical stimulation (e.g. Water Load Test II 7 ) remains to be evaluated. Further studies using measure of gut physiological states could be used to link conscious interoceptive feeling to actual measurement of gut activity. This would bring information on the biological basis and acuteness of the conscious interoception, namely a metacognitive information on the interoception of the participant.

Interoception refers to the process by which the nervous systems sense, integrate and regulate the inner body states 1 , 2 . Interoception allows the maintenance and the adaptation to different contexts of the inner physiological balance of each individual 3 . This process is also involved in cognition and the emergence of emotions 4 . Among interoceptive processes, gastric interoception refers to the bidirectional interaction between the gastrointestinal tract and the brain, known as the gut-brain axis, which allows the central nervous system to be informed about the state of the gastric, including mechanical distension or changes in temperature of the gastric wall. Although mostly unconscious, gastric interoception sometimes reaches consciousness, as illustrated by painful feelings of gastric pain, “butterflies” or satiety 5 . These gastric feelings may arise from a physicochemical state change of the stomach, yet, one can experience affective states such as anxiety, disgust or happiness that might influence different dimensions of interoceptive awareness 6 . To date, conscious gastric interoception has mainly been studied through methods that directly modify the physical state of the stomach. This includes non-invasive methods such as drinking different volumes of water, as in the Water Load Test (WLT) 7 ; drinking water at different temperatures 8 ; and using minimally invasive technologies such as with vibrating ingestible pills 9 . There are also invasive and painful techniques such as gastric mechanical distension using an inflatable balloon 10 , 11 . Therefore, it appears that every day-life moderately intense and unpleasant gastric feelings cannot be studied only through painful protocols or involving invasive gastric stimulation techniques modifying the gastric state. Yet, to the best of our knowledge, there is no validated protocol that allows the intensity and valence of gastric feelings to be modulated in a non-invasive and ecological way without physico-chemical modification of the gastric tract. Therefore, the aim of the present study was to develop and investigate the effectiveness of a non-invasive protocol based on avatar embodiment and threat in virtual reality to manipulate conscious gastric interoception. Embodiment refers to the experience of one’s own body that corresponds to the sense of owning the body, the sense of initiating movements and controlling the body, and the sense of being located inside the body 12 . Since this bodily experience relies partly on multisensory integration mechanisms of bodily signals such as tactile, proprioceptive, or interoceptive signals, experimental manipulation of sensory signals may affect the sense of embodiment 13 , 14 . More precisely, it is possible to induce embodiment illusions where a fake or virtual limb/body is experienced as one’s own (e.g., rubber hand, plastic mannequin, avatar 15 – 17 , or to influence the neural processing of one’s own internal state 18 . In original embodiment protocols, embodiment illusions tend to emerge when participants observe the fake limb/body being touched while they are touched in real time and in the same place (i.e., synchronous visuo-tactile stimulations) 15 , 16 . In contrast, this illusion is not observed when a delay is applied between visual and tactile stimulations (i.e., asynchronous visuo-tactile stimulations). Beyond the induction of a subjective embodiment illusion, sensory and emotional outcomes are observed. An increase in the integration of heartbeats by the central nervous system (neural representation of the processing of cardiac afferences) during synchronous visuo-tactile stimulations compared to asynchronous visuo-tactile stimulations has been shown 18 . These results suggest a stronger processing of cardiac afferences in a situation that promotes embodiment compared to a situation that impairs this experience. Conversely, raising awareness of one’s self-interoceptive states can facilitate the embodiment illusion, as it has been shown when the participant’s breathing rate is synchronized with that of the avatar 19 , 20 . By analogy and extension, we hypothesized that embodiment of an abdomen with synchronous visuo-tactile stimulations (compared to asynchronous) would facilitate the perception of gastric feelings. Drewing upon previous studies, we suggest that orienting the attention toward a virtual abdomen being stroked in synchrony with tactile stimulations applied on one’s own abdomen would induce an embodiment illusion of the virtual abdomen 16 , 21 . Then, we expected the sense of embodiment to enhance gastric feelings due to the emergence of a bodily self-consciousness applied to the virtual abdomen. In other words, experiencing the virtual abdomen as one’s own would involve processing its properties and contextual experiences as one’s own. In this sense, physiological and conscious experiences are indeed observed when threatening the fake limb/body following synchronous visuo-tactile stimulations but not following asynchronous visuo-tactile stimulations 22 , 23 . For example, neuronal activation related to anxiety and interoceptive awareness were found when an embodied rubber hand was threaten by a syringe 24 . Interestingly, previous correlational studies have demonstrated a link between anxiety and gastric sensations (e.g., gastric discomfort in children 25 , pain in functional gastro-intestinal disorders 26 , 27 . Considering the findings discussed here regarding embodiment and anxiety, as well as anxiety and gastric feelings, we propose that threatening an embodied virtual abdomen by approaching a virtual syringe should modulate perceived levels of anxiety and thus may affect gastric feelings. The present protocol aims to do so by applying a threat close to the skin. Because the skin is threatened rather than the viscera directly, we therefore investigated whether gastric sensations would be perceived as more cutaneous or more visceral, which we quantified along a new dimension of abdominal sensations called “viscerality”. As women tend to be more sensitive to their bodily sensations and link them more to their emotional states than men 28 , 29 , healthy women appears to be of particular interest for this proof-of-concept study. Considering previous studies showing that interoceptive feelings are disturbed or stronger for those: (1) who tend to be usually aware of their body feelings 30 ; (2) who tend to be more anxious 31 ; (3) who have difficulties in identifying their emotions 32 ; (4) who tend to experience period of somatic symptoms 33 , we also explored how interoceptive awareness 30 , anxiety 31 , alexithymia 32 and somatization 33 tendencies correlate with modulations of gastric feelings by our paradigm. In sum, we aimed at manipulating gastric feelings by the embodiment of virtual avatar’s abdomen and the induction of state anxiety by threatening the avatar’s abdomen and exploring how they relate to psychological traits. An overview of the protocol is displayed on Fig. 1 . Fig. 1 ( a ) Overview of the virtual environment. ( b ) General time sequence of the protocol and overview of the virtual environment displayed in the head-mounted display. ( a ) Overview of the virtual environment. ( b ) General time sequence of the protocol and overview of the virtual environment displayed in the head-mounted display.