A novel ecological approach to assess theory of Mind and social norm understanding for social cognition phenotyping in multiple sclerosis

Abstract

Social cognition (SC) impairment is a hallmark of the cognitive profile in Multiple Sclerosis (MS). However, there is a still no consensus on recommended measures for SC phenotyping in this clinical population, limiting our understanding of underlying mechanism and the development of targeted intervention. This study evaluated the suitability of the Edinburgh Social Cognition test (ESCoT), a novel ecologically valid tool assessing affective (ToM_A) and cognitive (ToM_c) Theory of Mind, as well as interpersonal (SNU_INTER) and intrapersonal (SNU_INTRA) social norm understanding, for SC phenotyping in MS. Seventy-one participants (39 with MS and 32 healthy controls) completed a neuropsychological battery including both non-social (Stroop test, Digit Span) and socio-cognitive (Eyes Test, ESCoT) measures. ESCoT demonstrated strong diagnostic validity (AUC = 0.803), outperforming the Eyes test (AUC = 0.653). Notably, ESCoT captured the peculiarity SC impairment profile in MS, marked by a dissociation between affective and cognitive ToM (ToM_A< ToM_c; group*ToM domain interaction: F = 21.87, p < 0.001), and between interpersonal and intrapersonal SNU (SNU_INTER< SNU_INTRA: group*SNU domain interaction: F = 22.76, p < 0.001). In the MS group, ToM_A and ToM_c scores were associated with the working memory, while SNU_INTRA score was linked to executive functioning. These findings support ESCoT as a valid tool for SC phenotyping in MS.

Introduction

Social cognition (SC) in Multiple Sclerosis (MS) represents a neurocognitive domain frequently impacted by the disease^1,2, with significant consequences for social connectedness and the overall quality of life^3–5. Research shows that people with MS frequently struggle to recognize emotions from others’ facial expressions, infer the affective states, thoughts, and intentions behind others’ behaviors, engage in effective social communication, and apply moral reasoning in daily interactions^6–9. The investigation of the extent and specificity of SC impairment in MS has only recently attracted the attention from clinicians and researchers, making this topic still in its infancy, especially compared to other, non-social, neurocognitive domains. This delayed interest is evident in the lack of a standard assessment battery to detect SC deficits in MS, a limited understanding of the mechanisms underlying these deficits, and a shortage of targeted intervention strategies^10,11.

Regarding SC assessment, there is currently no consensus on recommended measures for phenotyping social cognition level in MS. Overall, heterogeneous SC tools have been adopted, and evidence of their reliability, validity, and diagnostic capacity in this population is lacking^8,12,13. Moreover, traditional neuropsychological batteries commonly adopted in MS clinics primarily target other neurocognitive domains rather than SC (e.g.,¹⁴. Importantly, when SC is assessed, it is often fragmented into its components, resulting in lengthy sessions with multiple tests, each assessing a specific SC subdomain, such as emotion recognition, theory of mind, and empathy. Additionally, most used SC measures fail to capture the complexity of social stimuli encountered in daily interactions with others. For instance, many tools use static stimuli, such as photographs or vignettes (e.g.,^15–17. or require individuals to reason about verbal stories depicting complex social scenarios (e.g.,¹⁸. These types of stimuli, due to their static and decontextualized nature, limit the capacity of the tests to accurately reflect real-world social operations, which involve dynamic, embedded, and culturally nuanced stimuli. Prior evidence supports the use of movie-based stimuli to better elicit SC processes as they occur in daily life^19,20.

To address this issue, the Edinburgh Social Cognition Test (ESCoT) measures social cognition using dynamic, contextually embedded cartoon-like video clips that depict everyday social interactions within a social norm framework. It assesses individuals’ ability to understand affective and cognitive mental states, known as Theory of Mind (ToM), as well as their knowledge of social norms and their interpersonal and intrapersonal understanding²¹. Specifically, it examines reasoning about crystallized knowledge of implicit social boundaries prescribing what people should or should not do in certain circumstances, both interpersonally, understanding the implicit social norms that guide others’ behavior, and intrapersonally, those that guide one’s own behavior. Originally designed and normed for the Scottish culture, the ESCoT has been validated in clinical populations^22,23, showing good diagnostic validity in autism spectrum disorder, and high sensitivity in traumatic brain injury. More recently, the test has been translated and adapted for use in the Italian context²⁴. Notably, the ESCoT stimuli have also been adopted in an fMRI task to examine their effectiveness in eliciting neural activity in social brain networks among healthy Italian participants, demonstrating their validity in evoking the SC neural pathway²⁵. In the ESCoT, participants watch silent movies that depict social interactions between characters, mirroring everyday exchanges and illustrating social dynamics that align with or contrast with social norms. After viewing, they answer questions related to the characters’ mental states and the appropriateness of their behavior with respect to social norms.

The present study aims to evaluate the suitability of the ESCoT for phenotyping SC abilities in individuals with MS. Specifically, we examine the sensitivity of the ESCoT in capturing SC phenotype in MS and its diagnostic validity in comparison to other conventional tools. Finally, we deepen the link between the ESCoT performance, clinical characteristics, and non-social cognitive measures in people with MS. Given the ecological nature of the ESCoT stimuli, we expect to appreciate a higher capacity of the test to capture the SC impairment in MS than other measures, allowing for a precise and clear classification efficiency of people with MS and healthy individuals.

Results

Participants

In total, 71 participants took part in the research, 32 people in healthy condition and 39 individuals with MS. Table 1 illustrates the demographics, clinical, and neuropsychological characteristics of the two groups. The MS group reported a mean disease history of 16 years and a mild level of disability (median EDSS score = 3.5). Neuropsychological assessment results suggest an MS lower performance compared to the HC group in working memory and executive function measures (see Table 1). Considering SC, participants with MS underperformed on the Eyes test and the ESCoT test, with decreased ToM_A and SNU_INTER scores compared to HC (see Fig. 1).

Table 1.

Participants’ characteristics and groups comparison results.

Demographics	MS N = 39	HC N = 32	p value
Sex, n females (%)	28 (71.8)	20 (62.5)	0.563
Age (mean, sd)	51.87, 9.40	49.03,13.01	0.290
Education (mean, sd)	13.56, 3.25	14.25, 2.75	0.347
Clinical profile
EDSS, (median, 25th − 75th )	3.5, 3-5.5	-	-
Disease duration (mean, sd)	16.76, 8.58	-	-
Clinical phenotypes, n (%)
RR	34 (87.2)	-	-
SP	5 (12.8)	-	-
Disease-modifying therapy, n (%)		-	-
Yes	29 (74)
No	10 (26)
Immunomodulatory		-	-
Interferon beta, n (%)	2(7)
Glatiramer acetate, n (%)	4(14)
Fingolimod, n (%)	2(7)
Ozanimod, n (%)	1(3)
mAb		-	-
Ocrelizumab, n (%)	6(21)
Ofatumumab, n (%)	2(7)
Natalizumab, n (%)	4(14)
Other		-	-
Teriflunomide, n (%)	4(14)
Dimetilfumarate, n (%)	4(14)
Neuropsychological profile
Short-term memory Forward Digit Span (mean, sd) Working memory	5.50, 1.20	6.03, 1.31	0.082
Backward Digit Span (mean, sd)	4.18, 1.29	4.95, 1.20	0.013
Executive functions
Stroop
Interference index errors (mean, sd)	0.27, 0.73	0.35, 1.23	0.718
Interference index time (mean, sd)	21.15, 11.51	14.94, 8.01	0.013
Theory of Mind
RME (mean, sd)	24.68, 3.82	26.69, 4.04	0.037
Theory of Mind and Social Norm Understanding
ESCoT (mean, sd)	62.03, 7.58	71.12, 6.30	< 0.001
ToMc (mean, sd)	16.10, 1.87	15.72, 1.78	0.383
ToMA (mean, sd)	14.79, 4.55	18.84, 3.98	< 0.001
SNUINTER (mean, sd)	11.10, 2.56	15.69, 1.87	< 0.001
SNUINTRA (mean, sd)	20.03, 2.84	20.87, 2.51	0.191

RME = The Reading the Mind in the Eyes test; ESCoT= Edinburgh Social Cognition Test; ToMC= cognitive Theory of Mind; ToMA= affective Theory of Mind; SNU_INTER= Interpersonal Social Norm Understanding; SNU_INTRA= Intrapersonal Social Norm Understanding; MS= Multiple Sclerosis; HC= Healthy Control; sd= standard deviation; EDSS= Expanded Disability Status Scale; RR= Relapsing-Remitting; SP= Secondary-Progressive; n= number.

Fig. 1.

Group comparison in social cognition tests. RME = Reading the Mind in the Eyes Test; ESCoT_TOT = ESCoT total score; TOMA = affective Theory of Mind; TOMC = cognitive Theory of Mind; INTER = Interpersonal Understanding of the Social Norm; INTRA = Intrapersonal Understanding of the Social Norm. HC = Healthy Control group; MS = Multiple Sclerosis group.

By testing the differences between ESCoT subscales within and between groups, a trend of differences between ToM_A than ToMc was highlighted (F = 3.67, p = 0.06) and a statistically significant decreased SNU_INTER than SNU_INTRA score was found (F = 324.73, p < 0.001) (Fig. 2). Moreover, by exploring the interaction effect between group (MS vs. HC) and ToM (affective vs. cognitive) as well as SNU (interpersonal vs. Intrapersonal) domains in a repeated measures model, a significant interaction was observed (group*ToM domains: F = 21.87, p < 0.001, η² = 0.241, ω = 0.996; group*SNU domains: F = 22.76, p < 0.001, η² = 0.248, ω = 0.997) (Fig. 3). In particular, people with MS showed an opposite pattern than HC in ToM domains, with a lower performance in ToM_A than ToM_C, and a more pronounced difference in SNU_INTER and SNU_INTRA (SNU_INTER< SNU_INTRA) than HC.

Fig. 2.

Interaction effect between group and ToM domain (panel A) and group and SNU domain (panel B). MS= Multiple Sclerosis; HC= Healthy Control.

Fig. 3.

ROC curves assessing the diagnostic validity of the ESCoT total scores compared to the Eyes Test. ROC curves assessing the diagnostic validity of the ESCoT total scores compared to the Eyes Test (panel A) and the ESCoT subscores (panel B).

Figure 2. Interaction effect between group and ToM domain and group and SNU domain.

Correlations between social cognition, demographics, clinical and other neuropsychological measures

Bivariate correlations did not show statistically relevant associations between the ESCoT scores, the Eyes test, demographics, clinical, and other neuropsychological tests in the HC group.

Instead, in the MS group, the ESCoT total score was correlated with the Digit Span backward performance (r = 0.498, p = 0.001), as well as its ToM_A (r = 0.446, p = 0.037) and ToMc (r = 0.447, p = 0.041) subscores, with a higher ESCoT performance related to a greater working memory level. Also, the ESCoT SNU_INTRA score showed an association with the Stroop interference time (r=−0.396, p = 0.014) and error performance (r=−0.483, p = 0.002), with a higher ESCoT performance linked to a lower Stroop time response and number of errors. Similarly to the HC group, no correlations between ESCoT scores and the Eyes test were found.

ROC curves

By investigating the diagnostic capacity of the ESCoT, we observed a good diagnostic sensitivity and specificity of ESCoT (AUC = 0.803), which was higher than that of the Eyes test (AUC = 0.653, see Fig. 3 panel A). By comparing the two AUCs, a trend of statistically significant difference was found (z = 1.85, p = 0.06). A cut-off value of 67 was identified for the ESCoT total score. Considering the ESCoT subscores, only ToM_A, but not ToMc, showed an acceptable diagnostic ability (AUC = 0.744) with a relevant different AUC (z = 5.08, p < 0.001), from and the SNU_INTER, differently the SNU_INTRA, reported an outstanding diagnostic validity (AUC = 0.918) (z = 4.95, p < 0.001) (see Fig. 3, panel B). The suggested criterion values for the ToM_A and SNU_INTER were 20 and 14, respectively. Table 2 details the ROC analysis results.

Table 2.

ROC curve analysis.

	Sensitivity	Specificity	AUC	p value	Cutoff (J)
ESCoT total	0.78	0.74	0.807	-	67 (0.52)
ToMA	0.53	0.90	0.744	< 0.001	20 (0.43)
ToMC	0.03	1.00	0.362		20 (0.03)*
SNUINTER	0.87	0.85	0.918	< 0.001	14 (0.72)
SNUINTRA	0.56	0.56	0.543		21 (0.13)*

AUC, Area Under the Curve; J, Youden index; ESCoT= Edinburgh Social Cognition Test; ToM_C= cognitive Theory of Mind; ToM_A= Affective Theory of Mind; SNU_INTER= Interpersonal Social Norm Understanding; SNU_INTRA= Intrapersonal Social Norm Understanding. *Due to the reduced value of J score, the cut-off score is not considered valid and suitable for using in clinical practice.

Discussion

The present research aimed to test the suitability of the ESCoT, a comprehensive social cognition tool using dynamic, ecological, and culturally nuanced stimuli, to sensitively capture the socio-cognitive phenotype in MS as compared to conventional tools, such as the Eyes test.

Our first finding revealed the potential of the ESCoT to discriminate the MS social cognition phenotype from healthy individuals with higher sensitivity and specificity (AUC > 0.80) than those of a conventional tool. This evidence supports the use of the test in both clinical and research practice. With a great diagnostic capacity, ESCoT outperformed the Eyes test, one of the most used social cognition tools in the clinical neurology field²⁶. The superior sensitivity of ESCoT compared to the Eyes test might be related to the nature of the stimuli employed to investigate the quality of social reasoning. Depicting daily social interaction dynamics through video clips presents a complex scenario, which is closer to what individuals process in everyday social interactions, and, therefore more ecological and relevant to the deployment of socio-cognitive abilities in daily life^19,20. Even if the Eyes test presents real photographs of people expressing complex mental states, the stimuli remain decontextualized and may only partially target cognitive processing in everyday social interactions.

Examining more closely the ability of the ESCoT to fully grasp the specificity of the social cognition deficit in MS, we found that the tool effectively captures the nature of socio-cognitive difficulties in this clinical population. Thanks to its comprehensive assessment of social abilities, the ESCoT allowed us to observe dissociation between the components of ToM in individuals with MS, who, as reported in prior meta-analyses^27–29, typically show a selective decline in affective ToM in the early phases of the disease. Notably, this pattern contrasts significantly with healthy controls, who showed a reverse trend, demonstrating better performance in affective rather than cognitive ToM. In this sense, the dissociation between ToM domains, as supported by our findings and prior studies^8,29, serves as a marker of MS social cognitive phenotype. In this regard, when comparing the diagnostic capacity of affective and cognitive ToM ESCoT scores, consistent with the ToM dissociation in MS, we found that only the affective score showed good discriminatory power to distinguish MS from the HC group. Crucially, the ESCoT enables the exploration of distinct ToM domains using the same dynamic scenarios, unlike earlier studies that often used two separate dedicated tools (e.g³⁰).

Moving to the social norm understanding scores, the ESCoT assessment highlighted a further dissociation between interpersonal and intrapersonal social norm comprehension in MS. In particular, individuals with MS performed worse than healthy controls only in the interpersonal domain, specifically when they were asked to explain a character’s behavior in terms of what he or she should have done and why. They tended to explain the social behavior not in terms of underlying norms but more often by referring to other aspects such as personality traits or accidental circumstances. On the contrary, when they were asked to predict their behavior by imagining living the same situation as the character, they easily linked their conduct to social norms. This finding highlights a discrepancy between social expectations attributed to others and those applied to oneself. Although a slight discrepancy between inter- and intra-personal social norm understanding was also reported in the HC group, the participants with MS showed a more pronounced pattern, as evidenced by the significant interaction effect between group and social norm domain in the repeated measures model. Considering the diagnostic ability of the ESCoT social norm scores, we found excellent discriminative capacity (MS versus healthy participants) of the interpersonal social norm understanding score (AUC > 0.90), unlike the intrapersonal score, which showed low sensitivity. This finding suggests that moral reasoning, which is currently an underinvestigated social cognition domain in MS^7,13,31, merits further research, as it may represent a distinctive feature of the socio-cognitive profile of this population. To date, few studies have explored social norm understanding in MS, and, among these, Ehrlè et al.³¹ proposed the so-called “ultramorality” phenomenon, which describes a confusion between conventional norms, i.e., implicit societal norms, such as the ones depicted in the ESCoT scenarios, and the moral norms, i.e., behaviors causing serious physical or psychological damage. In this context, people with MS likely consider violations of conventional social norms as carrying severely immoral implications, which is not always the case. It is plausible that the MS disease history and the experience of disability, inevitably related to defeat and stigma, contribute to a harsher moral lens, likely creating a perception of rupture with society, and, in turn, feeling that one’s values and norms differ from society’s. Even if social norms still represent crystallized knowledge, they are no longer seen as the foundation of social behavior, but as morally charged elements integrated into personal values. As a result, individuals with MS may interpret prosocial behavior as driven by a person’s kind heart rather than an implicit social norm, and view social norm violations as stemming from malevolent intent. Further research is needed to explore this hypothesis by investigating more deeply the relationship between perceived social support, trust in society, disability experience, and social cognition.

Another aspect to be considered when explaining the dissociation between different domains of social norms understanding is that interpersonal comprehension may depend on the ability to predict others’ intentions, while intrapersonal comprehension may rely more on predicting one’s own intentions. It is possible that reasoning on intentions behind others’ social behavior and relating it to social norms constitutes a more complex task than reflecting on one’s conduct in clinical populations. Neural models of social cognition³² posit distinct neural networks implicated in recognizing one’s own and others’ intentions. To address this issue, future studies may further investigate the relationship between social norm understanding and ToM abilities in MS.

When exploring the association between the ESCoT scores and other clinical, demographic and cognitive variables, we found a relationship only with ToM scores and working memory, as well as between intrapersonal social norm scores and executive functions. This aligns with prior literature, theoretical models, and empirical evidence suggesting that ToM performance relies on working memory resources to maintain and manipulate socially relevant information, whereas social norm understanding draws on executive processes involved in behavioral regulation and cognitive control^8,30,33. In this regard, a previous work suggested that the recognition of social norm break is related to executive function, and that difficulties in this ability are especially related to executive dysfunctions in the clinical population, such as individuals with frontotemporal dementia and Alzheimer’s disease³⁴. In contrast, the lack of an association between the ESCoT total score and the performance on the Eyes test may reflect the fact that ESCoT captures more specific and multifaceted aspects of SC, as well as differences in task demands and stimulus characteristics.

This study is not without limitations. First, our sample size was relatively small, and further studies with a larger population are needed to confirm our results. However, this is the first research exploring the suitability of ESCoT in MS cognitive phenotyping, reporting novel insights into the MS social cognition profile. Second, we collected data only on a conventional social cognition tool, such as the Eyes test, without including other tests that may be nearer to the comprehensive approach of ESCoT, such as the Yoni task^25,35 or the Movie for the Assessment of Social Cognition test³⁶. Future studies should therefore consider integrating broader and more specialized assessment batteries, including standardized screening tools (e.g., BICAMS) and more extensive measures of executive functioning. Finally, in line with recent conceptualizations of multiple sclerosis as a clinical and biological continuum³⁷, phenotype-based stratification was not a primary focus of the present study. Although information on disease course (relapsing–remitting vs. secondary progressive) is clinically relevant, as highlighted by previous studies³⁸, subgroup analyses were not performed due to the limited number of participants with a progressive disease course, which would have resulted in insufficient statistical power. Future investigations with larger and phenotypically balanced samples may explore potential differences in social and cognitive functioning across the disease continuum or among diagnostic categories.

In conclusion, this study proposes the ESCoT as a valid tool to capture the social cognition phenotype of MS, addressing the urgent priority to provide robust and efficient SC measures for clinic purposes, as highlighted in the recent literature¹⁰. Future studies may adopt this tool to further explore socio-cognitive underlying mechanisms, especially the ones related to the ToM and social norm understanding domains dissociations in this clinical population.

Methods

This prospective cross-sectional study was conducted at the IRCCS Don Gnocchi Foundation in Milan from June 2023 to December 2024. The study protocol received approval from the Local Ethics Committee IRCCS Fondazione Don Carlo Gnocchi (protocol number 08_23/02/2022). The research adhered to the ethical guidelines outlined in the 2024 revised Helsinki Declaration. All participants read the information sheet and signed the informed consent form and voluntarily participated in the research without receiving any monetary compensation.

Participants

Individuals with MS and a group of sex- age- and education-matched healthy controls were enrolled in the study.

Participants with MS were consecutively recruited at the MS Centre of the IRCCS Don Carlo Gnocchi Foundation by a neurologist during follow-up clinic visits. They met the following inclusion criteria: (i) diagnosis of MS according to the 2017 McDonald’s criteria³⁹; (ii) age between 25 and 85 years; (iii) at least 5 years of schooling; (iv) preserved cognitive level as assessed by the Montreal Cognitive Assessment (MoCA) test (MoCA score > 15.50⁴⁰;; (v) no severe psychiatric illnesses or auditory/visual impairments able to interfere with the test performance; (vi) at least 3 months since the last relapse.

Healthy adults were also enrolled from the IRCCS Don Gnocchi Foundation clinic in Milan. These individuals included researchers, volunteers, and administrative staff. They were considered eligible based on the following inclusion/exclusion criteria: (i) absence of neurological and/or major psychiatric conditions; (ii) age between 25 and 85 years; (iii) at least 5 years of schooling; (iv) no pharmacological treatment with antipsychotics, antidepressants, or antiepileptics; (v) no auditory/visual impairments able to interfere with the test performance.

In general, before accepting eligible participants for the study, a brief clinical interview was conducted to ensure compliance with the research study’s inclusion/exclusion criteria. Particularly, during the interview the clinical history (past and current pathologies, surgical operations, pharmacological treatments) was investigated.

Procedure and materials

Following their recruitment, participants participated in a 90-minute individual neurocognitive evaluation session at the clinic with a neuropsychologist. The neuropsychological assessment included a test battery designed to measure both non-social and social cognitive abilities.

Neuropsychological tools

Executive functions and memory neurocognitive domains were evaluated through the following tests:

• Stroop Color-Word Test⁴¹: a measure of executive functioning, including selective attention, cognitive flexibility, and cognitive inhibition. Performance times and error scores were collected and adjusted as outlined by Caffarra et al.⁴².

• Digit Span Test Forward and Backward⁴³: a measure of short-term and working memory. Forward and backward total scores were calculated and adjusted according to Monaco et al.⁴³.

Theory of Mind and Social Norm understanding were assessed using the following measures:

• Reading the Mind in the Eyes Test (RME¹⁵: a measure of affective ToM assessing the ability to recognize complex mental states through static stimuli (photographs of individuals’ eye region). The total score (range 0–36) was computed and a higher score was interpreted as a greater SC performance.

• Edinburgh Social Cognition Test (ESCoT^21,24: a measure of social cognition, evaluating affective and cognitive ToM, as well as interpersonal and intrapersonal social norm understanding. In detail, the culturally adapted Italian version administered in this study consists of eight 30-second cartoon-style silent videos, plus one practice clip, depicting everyday social interactions between two characters, who either adhere to or violate social norms (e.g., a young man who does not help an old woman needing assistance after her shopping bag bursts) (see Fig. 4). During the test, participants watch each animation and then answer five open-ended questions while a static storyboard summarizing the interaction is displayed on the screen. The questions focuses on the following aspects¹: scenario comprehension², characters’ cognitive mental states (cognitive ToM)³, characters’ affective mental states (affective ToM)⁴, interpersonal comprehension of the social norms underlying the social interaction⁵, intrapersonal comprehension of the social norms underlying the social interaction (see Fig. 4).

• To fully capture individual’s interpretation of social scenario, prompts are provided to encourage further explanation of the social dynamics when necessary.

• Each social cognition response (questions 2–5) is scored on a scale from 0 (minimum) to 3 (maximum). The total score for each scenario ranges from 0 to 12 points, with an overall ESCoT score ranging from 0 to 96. For each question, a score of 3 indicated a response that effectively extracts and integrates relevant social information and context related to the interaction depicted in the video. Apart from the total score, four subscores were calculated, summing the scores for cognitive ToM (ToM_C, 0–24), affective ToM(ToM_A, 0–24), interpersonal social norm understanding (SNU_INTER, 0–24), and intrapersonal social norm understanding (SNU_INTRA, 0–24).

To ensure a good quality of scoring, given that this is the first study which administered the Italian version of the ESCoT to a clinical population, a two raters double blind approach was adopted. The two raters were psychologists blind to the participants’ group (MS vs. HC) and the research purpose. The inter-rater reliability analysis showed a substantial agreement⁴⁴ between rater (Cohen k: M = 0.762, 95%CI = 0.732–0.791).

Fig. 4.

An example of ESCoT scenarios (Adapted from Isernia et al.,²⁴).

Statistical analysis

The R software (retrieved from https://www.r-project.org/) version 4.3.3 was used for the statistical analyses.

Before performing bivariate and multivariate models, the distribution of variables was checked by visually inspecting quantile-quantile and histogram plots, and descriptive statistics were run to characterize the participants in terms of demographics, clinical, and neuropsychological profiles. Differences between MS and HC groups in baseline profile were assessed by running unpaired comparisons with Student’s t, Mann-Whitney, and Chi-squared tests (R package: tableone). Moreover, the group effect (MS versus HC), the domain effect (ToM_A versus ToM_C, SNU_INTER versus SNU_INTRA), and the interaction effect between group and domain (group*ToM domain and group*SNU domain) was tested using a repeated measure ANOVA model. Bivariate correlations were run to explore convergent validity and the influence of other variables (demographics, clinical and neuropsychological measures) on ESCoT performance. Holm’s multiple comparison correction was adopted (R package: corrplot). Finally, to assess the diagnostic validity of ESCoT, Receiver Operating Characteristic (ROC) curves were computed on both the ESCoT total score and its subscores (ToM_A, ToM_C, SNU_INTER, SNU_INTRA) (R package: ROCit). To compare the diagnostic accuracy with the conventional SC assessment, the classification accuracy of MS from HC of ESCoT was compared to that of the Eyes test using DeLong’s test. This comparison was also performed considering the ToM components (ESCoT ToM_A versus ToM_C) and the SN domains (ESCoT SNU_INTER versus SNU_INTRA). According to the literature, the ROC area under the curve (AUC) was interpreted as a prediction accuracy corresponding to random chance when equal to 0.5 and to perfect accuracy when equal to 1. In detail, an AUC ranging from 0.7 to 0.8 was considered acceptable, from 0.8 to 0.9 as excellent, and more than 0.9 as outstanding⁴⁵. Potential cut-off of ESCoT scores was extracted, and the best criterion value was identified by considering the Youden Index (J). As reported in the literature⁴⁶ a J index close to 1 was interpreted as a high capacity of the test to correctly identify MS and HC conditions, with a J > 0.5 suggesting an acceptable classification ability and a J > 0.7 an optimal diagnostic capacity.

Author contributions

FBa and SI conceived the study; GS and SI performed formal analysis; FR and FBo enrolled participants and carried out neuropsychological evaluations; GS and SI wrote the first draft of the manuscript; FBa, FR and FBo reviewed and edited the manuscript. All authors approved the final version of the manuscript.

Funding

This work was supported by the Italian Ministry of Health – Ricerca Corrente program 2025–2027.

Data availability

Data can be obtained upon reasonable request to the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.