Bodies serve as the primary interface with the external world and constitute the foundation of the human sense of self. Body awareness arises from the ability to integrate multisensory information from the inside and outside of the body (Tsakiris et al., 2011), and the development of body awareness begins with the body ownership, which involves the primal recognition that one's body is the source of sensations, and the sense of agency (Tsakiris et al., 2006). Human sense of body ownership is dynamically influenced by both internal sensations and external interactions (Martel et al., 2016; Tsakiris et al., 2011), and distinguishing between them is a fundamental yet complex task (Zaidel and Salomon, 2023). Interoceptive sensibility can be considered as an expression of a high-level model or “belief” for generating predictions about information coming from inside the body (Barrett and Simmons, 2015; Critchley and Garfinkel, 2017; Pezzulo, 2014; Seth, 2013).

Attention plays a role in interoceptive processes by modulating the flow between top-down prior beliefs and bottom-up signal flow (Seth, 2013). This attentional process on the ongoing bodily experience can be adaptive, but proneness to rumination—i.e., maladaptive repetitive thinking about past events—on somatic cues can lead to maladaptive behaviors (Naraindas et al., 2023; Mehling et al., 2012, 2018). Beyond internal experience, expectations and beliefs also influence body awareness. Beliefs about the body are frequently associated with cultural-specific representations about a desired body (Cattarin et al., 2000; Ricciardelli et al., 2007; Thompson et al., 2005), which might lead to experiences of anxiety and avoidance of social situations.

Novel evidence indeed supports the idea of a deep interconnection between body awareness and self-awareness, and between social interaction and body-awareness (De Vignemont, 2023). Interoception gives the concept of self a solid foundation, enhancing its coherence over time by reducing its susceptibility to external influences (see Arnold et al., 2019 for a review). On the other hand, social situations may reduce interoceptive processing by shifting attention from internally- to externally-focused. Higher interoceptive abilities would mediate the ability to “self-regulate” in social situations by flexibly directing attention toward both internal and external information; the link between interoception and the social self is thus bi-directional (Monti et al., 2022).

The sound of voice is embroidered in human body schema, and the ability to discriminate between self- and non-self voice cues is fundamental for self-awareness and self-monitoring during verbal communication (Candini et al., 2018; Conde et al., 2018; Fernyhough and Russell, 1997; Graux et al., 2015; Xu et al., 2013). Rather than being just a vehicle of verbal communication, the voice can be considered a salient marker of individuality (Belin et al., 2004; Crow et al., 2021), and centers on a domain where the body intersects with cultural processes of socialization and identity formation (Zimman, 2018). The voice is both an inner bodily signal (i.e., the embodied experience of perceiving oneself speaking) and a means by which individuals transmit information about themselves to interact with others. As such, the voice serves as a bridge between the self and others.

Perceiving one's voice is part of human everyday experience. Specifically, the experience of speech production requires the multimodal integration of different sources of information: auditory, proprioceptive, tactile, and barometric (Kent, 2024; Ito et al., 2009). The internal model of speech, or somatorepresentation, reflects how we experience and monitor speech-related sensations (Haggard and de Boer, 2014). This system continually adjusts speech output by comparing expected and actual sensory input, updating its internal models to reduce mismatches (Kent, 2024). This is particularly important given that individuals experience their own voice in multiple, qualitatively distinct ways. These range from everyday overt speech, to the perception of one's voice during playback of recorded speech (e.g., audio messages), to the experience of interacting with themselves through inner speech (Alderson-Day and Fernyhough, 2015; Lœvenbruck et al., 2018; Morin, 2005), or silently reading (Kunz et al., 2025; Perrone-Bertolotti et al., 2012; Vilhauer, 2016; Yao et al., 2011). Despite the differing sensory and cognitive demands of these contexts, individuals typically experience a stable and unified sense of vocal identity.

Although the literature mentioned above has focused on the role of the voice in shaping self-identity and social interactions, not much is known about the mechanism underlying the perception of different instances of one's voice.

A recent survey of 1,522 U.S. participants revealed that nearly 40% of respondents reported discontent with their voice, with no correlation to voice disorders (Naunheim et al., 2023). A recent update (Naunheim et al., 2024) showed an increase in voice complaints from 2012 to 2022, suggesting a possible association between increased voice dissatisfaction and the increased habits of self-recording audio-video contents and remote work, which became widespread globally between and after the COVID-19 pandemic. This incongruence arises from the physical transformation that the voice goes through while speaking: hearing one's voice is mediated both by air and bone conduction, while the voice other people hear is only conducted through air (Orepic et al., 2023; Pörschmann, 2000; Stenfelt, 2016).

In order to define the connection between the embodied experience of having a voice and the auditory experience of hearing one's voice, Crow and colleagues introduced the concept of vocal congruence, defined as “…the extent to which one's voice is in alignment, or congruent, with one's sense of self” (Crow et al., 2021, p. 1). To measure this process, they designed a self-report measure of vocal congruence, the Vocal Congruence Scale (VCS). The instrument uses a five-point Likert scale to investigate the degree of individual identification with the voice and the beliefs about the reflection between the voice and the personhood. In addition, the authors showed through a heartbeat detection task (i.e., a classical interoceptive task in which participants are required to judge whether heartbeat sensations are simultaneous with external stimuli presented at different time delays) that the scale has moderate to low correlations with metacognitive judgments of interoceptive awareness and confidence, suggesting that voice perception is related to consciously focusing attention and reflecting upon bodily experience.

This scale captures different faces of the vocal congruence experience, ranging from voice ownership, agency, and control, to functional vocal use and expressiveness in social contexts. Crucially, it addresses identity-related dimensions of vocal congruence such as voice-identity congruence and alignment with one's gender identity. Finally, the scale also investigates metacognitive and evaluative components, including awareness and satisfaction and/or rumination with one's voice. Vocal congruence can thus be understood as a multifaceted phenomenon arising from the perceptual experience of receiving auditory vocal feedback that can match or mismatch with one's intended vocal production. This perceptual mismatch may trigger both an evaluative and metacognitive judgment, as well as an affective response to this experience.

Gender plays an important role in shaping the individual's bodily experience. Gender identity is typically conceptualized as a person's inherent sense of their own gender, and how they identify as a woman, a man, both, an alternative gender, or neither (American Psychological Association, 2015). Here, in line with previous literature, we use the label TGNC to refer to transgender and gender nonconforming individuals, i.e., people whose gender identity varies from assumptions based on their birth sex (American Psychological Association, 2015). Although this label is not comprehensive, we believe it provides a valuable and respectful shorthand for referring to a diverse range of gender identities.

TGNC individuals face continuous discrimination, inequality, and social stigma in all aspects of their lives (Carmel and Erickson-Schroth, 2016; Connolly et al., 2016; Drabish and Theeke, 2022; Pinna et al., 2022; Romani et al., 2021; Russell and Fish, 2016; Testa et al., 2015; Truszczynski et al., 2022). Previous research highlighted the higher prevalence of alexithymia, body-image disorders, somatization, body uneasiness, and emotion dysregulation within the TGNC population compared to cisgender peers (Budge, 2020; Hatzenbuehler, 2009; Kallitsounaki and Williams, 2023; McGuire et al., 2016; Mirabella et al., 2020, 2024a; Maniaci et al., 2024; Mazzoli et al., 2022; Testa et al., 2015). Body dissatisfaction among individuals with gender incongruence is a major source of distress, encompassing psychological, physical, and biological dimensions and leading to behaviors such as avoidance, body surveillance, and feelings of detachment from the body (Mirabella et al., 2020, 2024a). Discomfort may involve not just secondary sexual characteristics but also non-sexual body parts and may represent a profound concern about gender discrepancies and societal norms (Pu et al., 2025).

The voice is one of the bodily aspects most commonly reported by TGNC people as a source of incongruence and distress (Chadwick et al., 2022; Kennedy and Thibeault, 2020; James et al., 2016; Oestreicher-Kedem et al., 2024; Pu et al., 2025; van de Grift et al., 2016a,b; Ziltzer et al., 2023). Vocal cues can disclose information related to personal identity like provenience, age, or gender, and for TGNC people, they can represent possible unwanted episodes of exposure of their birth sex (Chang and Yung, 2021). TGNC individuals report the need to find and develop a voice and communication that reflect their individual's sense of gender, and often seek speech therapists since their outer voice does not match with their “inner” and “true” voice (Davies et al., 2015). In fact, gender dysphonia (i.e., the perception of one's gender identity as being inconsistent with the qualities of one's voice and communication; De Bruin et al., 2000; Kennedy and Thibeault, 2020), has been shown to impact the quality of life and everyday functioning of TGNC individuals (Hancock et al., 2011; Hancock and Pool, 2017; James et al., 2016; dos Santos Oliveira et al., 2024). These aspects intertwine with physiological impairments related to pitch, voice quality, inflections, resonance, and precision in the articulation, intensity, and prosody (Davies et al., 2015), as well as the psychological reactions to them (Hancock and Pool, 2017). Within people identifying as TGNC there is great variation in the extent to which voice changes are undertaken or desired. Some people within this community seek to develop both masculine and feminine speech patterns, either because they identify as bigender or due to external pressures preventing them from fully expressing their gender identity. Others may have a gender identity that does not fit within the traditional woman/man spectrum and desire a more flexible gender presentation.

Culturally inherited stereotypes about what is considered feminine or masculine are often enacted through language (Charlesworth et al., 2021; Jones et al., 2020; Lindqvist et al., 2019; Lewis and Lupyan, 2020; Sczesny et al., 2016; Skewes et al., 2018). Studies with priming paradigms showed that linguistic gender cues can impact both explicit and implicit associations and attitudes of stereotyped group members (Pesciarelli et al., 2019; Steele and Ambady, 2006), and cause self-perception and social behavior to become more congruent to gender stereotypes (Hundhammer and Mussweiler, 2012). However, to date the impact of gendered stereotypes conveyed by language on vocal congruence remains underexplored. Deepening the complex interweaving between the internal bodily experience and cultural constraints might be crucial for the understanding of the TGNC people's experience of gender.

Little is known about how voice self-perception affects social interactions and, reciprocally, how living in a socio-cultural milieu—where representations of gender identity within a cultural system can shape how individuals perceive themselves—affects the experience of perceiving one's own one voice (Hughes and Harrison, 2013; Peng et al., 2019). To our knowledge, no existing study tackles the relationship between inner and outer voice perception across different gender identities. Furthermore, no research has systematically examined how gender stereotypes regarding the expected sound of female or male voices influence vocal congruence across diverse gender identities. This is striking since the experience of voice incongruence, i.e., the mismatch between the voice perceived while speaking and the actual voice that other people hear (or one's own recorded voice from a voice message), is a common everyday experience also for cisgender people (Holzman and Rousey, 1966).

In this study, we investigate vocal congruence across populations with different gender identities (cisgender and TGNC) using a behavioral self-voice perception task. We build on existing studies of vocal congruence by introducing an additional condition—the perception of inner voice congruence—to previously studied scenarios, such as reading aloud and listening to one's own recorded voice (Crow et al., 2021; Welch and Helou, 2022). Importantly, participants are presented with excerpts of texts conveying either gender-stereotypical content (feminine vs. masculine) or gender-neutral content.

As our main prediction, we hypothesize differences in voice perception between Cisgender and TGNC participants. We expect the TGNC group to experience lower vocal congruence overall, with the lowest scores occurring in conditions that require focus on the externalized voice. Specifically, we expect TGNC participants to experience lower vocal congruence than cisgender participants during the Reading Aloud condition, which involves active speech motor control and simultaneous auditory-somatosensory feedback (including bone conduction). We also expect lower scores for TGNC participants in the Listening condition, which isolates the purely auditory-perceptual evaluation of the self-voice without motor intent. Conversely, we hypothesize that Silent Reading—requiring an engagement of auditory imagery without an external feedback—will elicit relatively higher congruence scores in the TGNC group, as it reflects a mental representation of vocal identity that is less mediated by the physical constraints or acoustic “mismatches” of externalized speech. A summary of the experimental conditions and their underlying processes is provided in Table 1.

Second, to investigate how gendered linguistic primes influence vocal self-perception, we manipulate the semantic content of the texts, presenting excerpts that are gender-stereotypical (masculine or feminine) and gender-neutral. We predict that the semantic content of the texts will interact with participants' gender identity, such that gendered primes will reduce vocal congruence for TGNC individuals by reinforcing stereotypical gender associations. Although a stricter test of the priming effect should have targeted specific gender identities within each group paired with the relevant linguistic stereotype, the scarcity of the sample prevented us from deepening this aspect. Nonetheless, we still expected the gendered semantic content of texts would specifically impact on TGNC participants' vocal congruence perception because it might draw participants' attention to the general gender conceptual dimension.

Prior research suggests that interoceptive sensibility may play an important role in voice perception (Crow et al., 2021; Orepic et al., 2022; Smeltzer et al., 2023). Building on this, and also considering the high prevalence of emotion regulation difficulties—such as alexithymia—among TGNC individuals (Kallitsounaki and Williams, 2023; Mazzoli et al., 2022; Reed et al., 2023), the present study examines a broader set of psychological factors that might influence the experience of vocal congruence across different gender identities. Specifically, we investigate the role of interoceptive sensibility in shaping vocal congruence, as well as the potential disruptive effect of alexithymia and emotion regulation difficulties. In addition we explore the relation between ontological beliefs about gender/sex and vocal incongruence. Finally, we also investigate whether experiences of gender-related discrimination in the TGNC sample influence vocal congruence perception.

All the materials, data, scripts, and analyses of the study are available at the OSF repository: https://osf.io/v2gtn/.

A total of 45 participants were recruited for the study. A sensitivity power analysis was conducted using G*Power software (Faul et al., 2007) to determine the minimum detectable effect size given the observed sample size. For the primary Condition × Group interaction, the study had 80% power to detect an effect size of f = 0.50 (η² =0.20) at α =0.05. Post-hoc approximations for between-group comparisons indicated adequate power to detect medium-to-large effects (d = 0.85, power = 87%). After excluding one participant due to failure to complete the questionnaires, the resulting sample size was N = 44 for all subsequent statistical analyses (M Age = 27.45; SD Age = 10.22; Age range = 18–66). We recruited participants through the involvement of an LGBTQIA+ rights local association and a hospital service dedicated to gender-affirmation processes. Trainer vocalists and individuals at advanced stages of the gender-affirmation process who had undergone voice training interventions were not eligible. This was ensured prior to participants' enrollment in the study. We asked for the following demographic information: age, birth sex, gender identity, sexual orientation, educational level, birth country, and languages spoken from childhood. Participants' demographic information is reported in Table 2. Most participants (57%, n = 25) were female at birth, whereas 41% (n = 18) were male at birth and one participant was intersex. Based on gender identity, we divided participants into two groups: TGNC participants (n = 22) and participants who identified as cisgender (n = 22). In the cisgender group, 55% of the participants identified as cisgender women (n = 12) and 45% as cisgender men (n = 10). In the TGNC group, 32% of the participants identified as transgender men (n = 7), 23% as transgender women (n = 5), and 23% as non-binary (n = 5). Four participants (18%) used the “other” response option and reported they identified as “genderfluid transgender man,” “non-binary transgender man,” “genderfluid,” “woman.” The two groups were comparable in terms of age, t_(36.33) = −0.087, p = 0.930 (M cisgender = 27.59; SD = 8.04; M TGNC = 27.31; SD = 12.21).

In the cisgender group, one participant completed only middle school (5%), 36% (n = 8) of participants reported having a high school diploma, one participant obtained a bachelor degree (5%), 36% (n = 8) a master's degree, and 18% (n = 4) a PhD title. In the TGNC group, 14% (n = 3) completed middle school, 50% (n = 11) of participants reported having a high school diploma, 36% (n = 3) had a bachelor's degree, 14% (n = 3) had a master's degree, and 18% (n = 4) a PhD title (Table 2).

The experimental procedure was divided into three parts, and participants completed it in two sessions. In the first session, participants completed the voice perception task to assess levels of vocal congruence. As an additional, separate part of the experimental procedure, participants also rated a set of words onn multiple semantic dimensions. However, these results are not discussed here, as they are not central to the aims of this study. Finally, they gave demographic information. In the second session, participants completed questionnaires investigating interoceptive sensibility, emotion regulation, alexithymia, gender identity, and gender discrimination. Questionnaires were administered 2 days after the completion of the first session to avoid fatigue and carryover effects of the voice perception task. The experimental procedure was implemented with Qualtrics, using an on-line questionnaire divided into sections that participants filled in a fixed order. The study obtained ethical approval from the Ethics Committee of the Sapienza University of Rome (Ethical Approval 0000856).

The distribution of VCS total scores was assessed using visual inspection of histograms and Q–Q plots alongside formal normality tests. The distribution appeared approximately unimodal and bell-shaped, with observations primarily concentrated in the mid-range. While the Shapiro-Wilk test (via “stats” R's Package) indicated a statistically significant deviation from normality (W = 0.9746, p < 0.001), visual diagnostics revealed that this deviation was primarily confined to the extreme tails (please refer to Supplementary materials), reflecting the bounded nature of the 10–50 sum score.

Given that linear mixed-effects models are robust to moderate departures from normality (Schielzeth et al., 2020), VCS total scores were analyzed using linear mixed-effects models (via the “lme4” package in R; Bates et al., 2015).

As a preliminary step, we conducted within-participant comparisons to assess whether the two gender-neutral texts differed in their effects on vocal congruence, and whether the two gendered texts differed from each other. Given the non-normal distribution of vocal congruence sum scores, we used Wilcoxon rank-sum tests with continuity correction. Results indicated no significant difference between the two gender-neutral texts (W = 8,529.5, p = 0.769), nor between the two gendered texts (W = 8,671.5, p = 0.948).

Because the primary level of analysis in the present study contrasts neutral vs. masculine vs. feminine texts, and only the neutral condition included two empirically equivalent texts, we averaged vocal congruence scores across the two neutral texts within participants. This approach preserves all available data while ensuring that the neutral condition contributes to a single, stable estimate per participant, comparable to the masculine and feminine conditions.

Prior to any statistical analysis, we examined distributional assumptions. Shapiro-Wilk tests revealed that the distribution significantly deviated from normality (W = 0.884, p < 0.001). Visual inspection of histograms confirmed a non-normal distribution with bimodal tendencies (peaks at ratings 1–2 and 3–4).

To further investigate vocal congruence as a function of Type of Text, we initially fitted a cumulative link mixed model with the “ordinal” R's package (Christensen, 2022) featuring scores of the Inner-Outer Voice Congruence Rating as dependent variable, Group (TGNC vs. cisgender), Type of Text (Gender-neutral vs. Stereotypically Feminine vs. Stereotypically Masculine), and their interaction as fixed factors, and participants as random intercepts. However, the model failed to converge due to quasi-separation issues in the data.

Therefore, we relied on non-parametric testing. To compare ratings between TGNC and cisgender participants across text types, we conducted Wilcoxon rank-sum tests with Bonferroni correction via the “rstatix” package (Kassambara, 2023).

To account for the potential difference in the questionnaires (MAIA, TAS-20, MULTI-GIQ, GMSR) scores between the two groups, we employed one-way ANOVAs. We further examined the role of each variable targeted by questionnaires (interoceptive sensibility, alexithymia, gender identity, gender-related discrimination) on vocal congruence. Given the exploratory nature of these analyses across multiple psychological domains, we ran separate linear mixed models for each subscale, using the same structure and methodology as the main model. To address the risk of Type I errors associated with multiple testing, we applied the Benjamini-Hochberg False Discovery Rate (FDR) correction to all p-values across the total set of models. Post-hoc contrasts were conducted with the “emtrends” function from the “emmeans” R package.

We investigated whether individual differences in the general ability to attend to and connect with bodily sensations—measured by the MAIA scale—and the ability to identify and express feelings—assessed via the TAS-20 scale—may moderate the effect of the experimental manipulation. Furthermore, we explored how aspects related to gender identity representation, as captured by the MULTI-GIQ, might affect judgments of vocal congruence. Finally, we examined whether experiences of gender discrimination, measured by the GMSR questionnaire, might specifically impair vocal congruence in TGNC participants. In the analyses we will only focus on the impact of each covariate on vocal congruence and its interaction with the different groups as a function of experimental conditions. For the full set of additional models please refer to the Supplementary materials.

Like for the previous vocal congruence model, we compared for each model featuring covariates the full model with the three-way interaction to a reduced model with only the two-way interaction between Condition and Group. Since the full model did not significantly improve the fit, χ(12)2 = 7.92, p = 0.791, and had higher AIC and BIC values (AIC = 2.504 vs. 2,488; BIC = 2,583 vs. 2,519), we retained the two-way interaction model for each covariate.

Density distributions of average VCS scores across Conditions and Groups suggest distinct performance profiles. Cisgender participants consistently exhibit slightly higher peak densities and appear to have somewhat more concentrated distributions around higher average scores. The TGNC group demonstrates more varied distribution patterns across modalities, with the Listening condition showing the most distinct separation between the groups (see Figure 1).

Our first research question investigates whether linguistic gender stereotypes affect the perception of vocal congruence in different conditions as a function of different gender identity experiences.

Looking at intra-class correlations, approximately 51.5% of the total variance in VCS Total scores was attributable to stable differences between participants.

The three-way interaction among Condition, Group, and Type of Text was not significant, χ(4)2 = 0.34, p = 0.986. There was, however, a significant two-way interaction between Condition and Group, χ(2)2 = 47.24, p < 0.001. Pairwise comparisons show that the two groups differed in the Silent Reading condition, β = −3.91, t = −2.250, SE = 1.74, p = 0.028, with TGNC participants reporting lower vocal congruence compared to cisgender participants (EMM TGNC = 34.9, SE = 1.22, UCL = 32.5, LCL = 37.4; EMM Cisgender = 38.8, SE = 1.23, LCL = 36.4, UCL = 41.3). Likewise, the two groups scores also differed in the Reading Aloud condition, β = –10.55, t = −6.070, SE = 1.74, p < 0.001, with TGNC participants reporting significantly lower vocal congruence compared to cisgender participants (EMM TGNC = 28.7, SE = 1.23, LCL = 26.2, UCL = 31.1; EMM Cisgender = 39.2, SE = 1.23, LCL = 36.8, UCL = 41.7). Finally, the two groups differed in the Listening condition too, β = –11.48, t = −6.606, SE = 1.74, p < 0.001, with TGNC participants reporting significantly lower vocal congruence compared to cisgender participants (EMM TGNC = 27.1, SE = 1.23, LCL = 24.6, UCL = 29.5; EMM Cisgender = 38.6, SE = 1.23, LCL = 36.1, UCL = 41.0). Finally, TGNC participants gave higher scores of vocal congruence in the Silent Reading compared to the Reading Aloud, β = 6.22, t = 7.334, SE = 0.84, p < 0.001, and Listening conditions, β = 1.61, t = 9.234, SE = 0.84, p < 0.001, while there was no difference between the Reading Aloud condition and the Listening condition, β = 1.61, SE = 0.84, p = 0.141 (EMM Silent Reading = 33.9, SE = 1.23, LCL = 32.5, UCL = 34.7; EMM Reading Aloud = 28.7, SE = 1.23, LCL = 26.2, UCL = 31.1; EMM Listening = 27.1, SE = 1.23, LCL = 24.6, UCL = 29.5). No significant difference emerged across conditions for the cisgender group, all p_s ≥ 0.701 (see Figure 2).

The model additionally indicated significant main effects of Condition, χ(2)2 = 48.52, p < 0.001, and Group, χ(2)2 = 29.44, p < 0.001, while Text was not significant, χ(2)2 = 2.08, p = 0.352. No other main effects or interactions reached significance (all ps > 0.218; please refer to Supplementary materials for report of model results).

Density distributions reveal that TGNC participants display greater variability compared to Cisgender participants, and reported consistently lower alignment between internal and external voice perception (Figure 3).

Wilcoxon rank-sum tests revealed significant differences between TGNC and cisgender participants across all text, all ps ≤ 0.002 (Bonferroni adjusted, see Table 5).

Descriptive statistics averaging the two groups ratings across conditions further evidenced the lower perceived congruence between TGNC's group inner vs. outer voice (M TGNC = 1.75; SD = 0.98; M Cisgender = 3.27; SD = 0.99).

Overall, the results tackling vocal congruence as emerging from the analyses of both the Vocal Congruence Scale and the Inner-Outer Congruence Rating suggest that TGNC participants experience vocal congruence in an idiosyncratic and subjective manner. Specifically, in the VCS assessing more general aspects connected to voice representation (e.g., control, adherence to identity, satisfaction, and awareness) the two groups differed in all conditions (i.e., Silent Reading, Reading Aloud, and Listening)—showing a decreasing perceived congruence for the TGNC group as a function of whether the condition required an auditory feedback or not. Hence, we found the starkest difference between the two groups in the Listening condition. The internal representation of their speech from the early stage (Silent Reading) to the latest stages of articulatory and/or auditory components (respectively, the Reading Aloud and Listening condition) would be perceived as not reflecting features of the bodily self. Moreover, the rating task investigating the perceived mismatch between one's own inner vs. outer voice highlighted the two groups systematically diverged in their assessments—with the TGNC group once again reporting lower congruence compared to the cisgender group, regardless of the experimental condition.

The previous analyses showed that vocal congruence varies across gender identities, with transgender and gender non-conforming (TGNC) participants systematically reporting lower congruence compared to cisgender participants. To better understand this phenomenon, we fitted separate models investigating the impact of specific aspects related to interoceptive and metacognitive-emotional components as well as aspects related to gender identity on vocal congruence (see Data Analysis). For space reasons, below we only report results from models in which the interaction between specific covariates, experimental condition, and group—where relevant—is significant. For the complete report of results please refer to Supplementary materials.