The cognitive-behavioral missing link in myotonic dystrophy type 1: Rethinking social cognition assessment Authors: Emanuele Maria Costantini, Silvi Cadri, Alice Ciattaglia, Michela Coccia, Valeria A. Sansone Published: 2026 ABSTRACT BackgroundMyotonic dystrophy type 1 (DM1) is associated with central nervous system involvement and cognitive–behavioral alterations, including deficits in social cognition. However, standard neuropsychological assessments based on accuracy may underestimate clinically relevant dysfunction.ObjectivesTo determine whether task completion times provide a more sensitive index of social-cognitive performance than accuracy-based metrics in DM1.MethodsThirty-three individuals with genetically confirmed DM1 and 40 healthy controls completed the Reading the Mind in the Eyes Test (RMET) and the Story-Based Empathy Test (SET). Both accuracy and task completion times were analyzed. Executive functions and processing speed were evaluated using standard neuropsychological measures. Group comparisons and regression analyses were performed adjusting for age, education, and executive performance.ResultsDM1 patients showed different patterns for accuracy and completion time across social-cognitive tasks. RMET accuracy was reduced, whereas SET accuracy was comparable between groups. In contrast, completion times were consistently prolonged across all social-cognition tasks in DM1. These differences remained significant after controlling for executive function. SET completion time showed stronger discriminative performance than SET accuracy measures, whereas RMET completion time and accuracy showed comparable discriminative performance.ConclusionsSocial cognition in DM1 is characterized by prolonged completion times, while accuracy is relatively preserved on several measures. This discrepancy suggests that conventional accuracy-based assessments may not fully capture subtle differences in social-cognitive performance. Temporal metrics may reveal aspects of social cognition performance not detected by accuracy-based assessments alone and contribute to a more comprehensive neuropsychological evaluation. 1. INTRODUCTION Myotonic dystrophy type 1 (DM1) is an autosomal dominant, highly pleomorphic multisystem disorder caused by CTG repeat expansion in the DMPK gene. Although muscle involvement is the hallmark of the disease, central nervous system abnormalities and the related cognitive-behavioral phenotypes are increasingly recognized as a major contributor to disability, with implications for autonomy, social participation, and quality of life, as well as treatment adherence.1–4 A significant proportion of DM1 patients exhibit a behavioral pattern nuancing from reduced social engagement and inefficient interpersonal interactions to atypical personality traits,5–8 often characterized by apathy,2,3 and sometimes accompanied by obsessive–compulsive or schizoid manifestations, spanning subtle social uneasiness to overt psychiatric manifestations. 7 Neuroimaging and clinical-neuropathological studies have demonstrated widespread alterations involving both white and gray matter, particularly within fronto-parietal and fronto-temporo-insular networks,5,9–17 while neuropsychological studies have consistently reported deficits in executive functioning, processing speed, visuospatial abilities, and memory.15–19 These neural alterations involve regions that are critically implicated in behavioral and emotional regulation, closely mapping the Social Cognition domain.9,11,12,19–21 Social cognition encompasses a continuum of abilities, ranging from empathy—the capacity to perceive, understand, and share others’ emotional states—to more cognitively demanding, higher-order processes collectively referred to as Theory of Mind (ToM), or mentalizing. To mentalize means to interpret one’s own and others’ behavior in terms of underlying mental states, including beliefs, intentions, desires, emotions, and knowledge. Accordingly, ToM refers to the ability to attribute such mental states to oneself and others in order to understand, predict, and respond appropriately to behavior. It is commonly divided into two partially dissociable components: cognitive ToM, which involves inferring others’ beliefs, intentions, thoughts, and knowledge, and affective ToM, which concerns inferring and understanding others’ emotions and feelings.22–24 Previous studies have reported impairments in these domains in DM1, including reduced sensitivity to emotional expressions and impairments in ToM-related processes. However, findings remain heterogeneous, with some investigations documenting significant deficits and others reporting performance within normal or near-normal ranges on standard tasks.8,25–31 Furthermore, only a few have directly examined ToM,20,21,25,28 and even fewer have assessed emotion recognition and ToM within the same sample 21 making it difficult to determine whether deficits in emotion decoding and higher-order mentalizing co-occur or reflect partially independent dysfunctions. This heterogeneity may partly reflect methodological limitations of current assessment approaches that may not fully capture the dynamic and adaptive nature of social cognition competences. Most tasks rely primarily on accuracy scores, implicitly assuming that correct responses adequately reflect intact processing. However, successful social interactions depend not only on making accurate social inferences but also on generating them rapidly enough to support the tightly timed dynamics of interpersonal exchange. Conversational turns are typically separated by only a few hundred milliseconds, requiring listeners to anticipate turn endings and begin preparing their responses before the current speaker has finished; moreover, faster conversational response times are associated with greater perceived social connection. Accordingly, delays in interpreting emotional cues or inferring others’ intentions may disrupt conversational fluency and adaptive social responding even when the eventual inference is correct.32–34 Nevertheless, the temporal dimension of social-cognitive performance has received little attention in DM1, and no previous studies have systematically examined whether task completion times provide additional information beyond conventional accuracy-based measures. Although more dynamic and context-enriched tools have been developed to capture real-world social functioning and already tested in other clinical populations, such as schizophrenia,35,36 they have not been clinically validated nor available in non-English language. Furthermore their construct validity as real-word social functioning proxy is debated. 37 Therefore, the aim of this study was to evaluate whether completion times in largely available and clinically grounded tasks could provide a more sensitive and informative index of social-cognitive performance in DM1 than accuracy-based measures.We hypothesized that temporal metrics may capture an additional dimension of performance that is not reflected by accuracy alone and that could represent a clinically relevant aspect of social cognition. To test this hypothesis, we examined both accuracy and completion times in established socio-cognitive tasks, while controlling for executive functioning, in order to determine whether slowed performance represents a measurable feature of DM1, whether it is explained by general cognitive slowing or reflects a domain-specific impairment, and whether completion times provide information beyond conventional accuracy-based measures. If temporal measures capture an additional dimension of social-cognitive performance that is not reflected by conventional accuracy-based assessments, incorporating them into neuropsychological evaluation may contribute to a more comprehensive characterization of social-cognitive performance in DM1. 2. METHODS This cross-sectional case-control study included adults with genetically confirmed non-congenital DM1 and healthy controls (HC). Patients were consecutively recruited from a tertiary referral center for neuromuscular diseases during the study period. The diagnosis of DM1 was confirmed by molecular genetic testing demonstrating a pathogenic CTG repeat expansion in the DMPK gene. Individuals with other neurological disorders or conditions precluding reliable neuropsychological testing were excluded. HC were recruited among hospital staff and members of the local community. They had no history of neurological or major psychiatric disorders and no known cognitive impairment. Because no previous studies had examined social-cognitive task completion times in DM1, the expected difference was based on preliminary clinical observations obtained during study planning. We anticipated a mean SET completion time of 6 minutes (360 seconds) in individuals with DM1 and 4 minutes (240 seconds) in healthy controls, with a common standard deviation of 150 seconds, corresponding to Cohen’s d = 0.80. With a two-sided alpha of 0.05 and 90% power, 33 participants per group were required. The same standardized effect size (Cohen’s d = 0.80) was considered applicable to both co-primary completion-time outcomes. Healthy controls were recruited to provide a comparable reference group for between-group analyses. Global cognitive functioning was screened using the Montreal Cognitive Assessment (MoCA). 38 Executive and processing speed-dependent abilities or abilities contributing to domain-general processing efficiency were assessed using phonemic and semantic verbal fluency, 39 the Stroop color–word test (recording both total time and interference), 40 and the Trail Making Test (TMT) A and B, 41 with the TMT B–A difference score computed to isolate set-shifting while minimizing visuomotor and psychomotor demands. Visuoconstructive ability was measured using the Rey–Osterrieth Complex Figure copy (ROCF-copy). 42 Demographic variables included age and years of education. CTG repeat expansion size was reported according to the laboratory classification as E1 (50–150 CTG repeats), E2 (151–1,000 CTG repeats), or E3 (>1,000 CTG repeats). 43 Age at disease onset was classified according to the five-phenotype DM1 framework as congenital (<1 month of age), childhood-onset (1 month–9 years), juvenile-onset (10–19 years), adult-onset (20–40 years), or late-onset (>40 years).44,45 Different aspects of social cognition were evaluated using two complementary instruments: the Reading the Mind in the Eyes Test (RMET),46,47 commonly used to assess emotion recognition and affective mental-state inference from subtle facial cues, and the Story-Based Empathy Task (SET), 48 which assesses cognitive and affective components of mentalizing (i.e. inferring, representing, and interpreting the mental states of oneself or other people) through context-rich narrative sequences. The revised adult version of the Reading the Mind in the Eyes Test (RMET) was administered to assess the ability to infer complex mental states from restricted facial cues. The task consists of 36 black-and-white photographs depicting only the eye region of male and female faces. For each item, participants are presented with four mental-state descriptors and are required to select the term that best represents what the depicted individual is thinking or feeling. One point is awarded for each correct response, yielding a total score ranging from 0 to 36, with higher scores indicating more accurate mental-state attribution. A glossary of the mental-state terms may be consulted to minimize the influence of unfamiliar vocabulary. The RMET was originally developed as a 25-item, two-alternative forced-choice task and was subsequently revised to increase its psychometric sensitivity by expanding both the number of items and the number of response alternatives.49,50 The Italian version of the RMET has shown satisfactory test-retest reliability and evidence of convergent and discriminant validity. Nevertheless, recent reviews have questioned the consistency of its construct-validity evidence, suggesting that performance may reflect multiple processes, including affective mental-state inference, emotion recognition, lexical knowledge, and visuoperceptual abilities. 51 Accordingly, we interpreted the RMET as a task-specific measure of mental-state inference from facial cues reflecting both emotion recognition and affective ToM competences. The RMET was scored for affective mental-state inference accuracy and binary sex identification accuracy (RMET-M/F). This measure was included as a low-level visuoperceptual control condition, allowing us to distinguish impairments in emotional inference from potential deficits in basic perceptual processing. The Story-based Empathy Task (SET) is a non-verbal social-cognition task designed to assess both cognitive and affective components of Theory of Mind. It comprises 18 cartoon-based trials, divided into three conditions of six items each: intention attribution (SET-IA), requiring participants to infer a character’s underlying intention; emotion attribution (SET-EA), requiring inference of the character’s emotional state; and causal inference (SET-CI), a control condition requiring the identification of an outcome based on the physical properties of objects and human bodies, without mental-state attribution. In each trial, participants view a three-vignette story sequence and subsequently select the most appropriate conclusion from three alternative endings. One point is awarded for each correct response, producing subscale scores ranging from 0 to 6 and a global score ranging from 0 to 18. Higher scores indicate more accurate inference of intentions, emotions, or physical causality, depending on the condition. 48 SET performance was analyzed across the three subcomponents: SET-CI, SET-IA, and SET-EA. For both tasks, RMET and SET, accuracy and task completion time indices were analyzed. For each task, completion time indices were computed as the total completion time rather than stimulus-locked reaction times, and were intended to capture processing efficiency at a behavioral level. SET subcomponent times were also examined to explore potential differences between cognitive-abstract components (SET-IA) and the more affective component (SET-EA), as well as for total completion time (SET_total_time). The SET control task completion time (SET-CI) was also noted. For consistency throughout the manuscript, the term completion time is used to indicate the total time required to complete each task. In the present study, completion time was considered a behavioral index of processing efficiency, which we define as the overall efficiency with which social information is processed and translated into a response. As such, completion time should not be interpreted as a pure measure of reaction time, but rather as an integrated measure reflecting the combined contribution of multiple cognitive processes involved in task performance. Task completion times were derived from response latencies recorded by a trained examiner using a stopwatch according to a standardized administration protocol. All assessments were conducted in a controlled clinical setting, with identical instructions and procedures across participants. To minimize variability, all completion time measurements were recorded by the same examiner, who was specifically trained in the administration and timing procedures prior to the study. Although this approach does not provide millisecond-level precision, it allows a consistent and reliable estimation of relative differences in processing efficiency between groups under standardized conditions. This method has been adopted to preserve feasibility within a clinical neuropsychological setting and to provide a standardized estimate of overall task completion efficiency. To reduce the potential influence of domain-general cognitive functions on completion-time measures, these indices were interpreted in conjunction with prespecified executive and related cognitive measures, including inhibitory control (Stroop), cognitive flexibility (TMT B-A), and visuoconstructive performance (ROCF copy). Although these measures are not pure indices of processing speed, they were selected among the others tested, because we considered these cognitive processes substantially contributing to social cognition tasks completion time.52,53 Although ROCF copy is conventionally classified as a measure of visuoconstructional praxis, we selected it as a non-verbal index of higher-order cognitive resources potentially contributing to social-cognitive performance. In fact, successful reproduction of the ROCF requires not only visuoperceptual analysis and graphomotor execution, but also global–local integration, identification of the figure’s hierarchical structure, planning and sequencing of construction, monitoring and correction of ongoing performance, inhibition of fragmented or perseverative responses, and the temporary maintenance and updating of spatial relationships across successive drawing steps.42,54,55 These characteristics make ROCF copy particularly relevant as a control measure in the present study, in which both social-cognition tasks were predominantly non-verbal and required the efficient selection, integration, and manipulation of complex visual information. Moreover, visuospatial and visuoconstructional dysfunction represents one of the most robust cognitive abnormalities reported in DM1. A systematic review and meta-analysis identified visuospatial perception as the domain showing the largest pooled impairment relative to healthy controls and ROCF copy as one of the individual measures yielding the most consistent between-group differences with acceptable heterogeneity. 19 More recently, Garmendia et al. 56 confirmed impaired copy accuracy in DM1 with a large effect size, together with abnormalities in configural accuracy, planning, organization, and perseveration, further supporting its sensitivity to the characteristic non-verbal cognitive dysfunction of this population. Because of this considerations, collectively, we will refer to the abilities implied in the selected cognitive tasks as executive-loading functions. The control condition, SET-CI, was also used as covariate in separate regression models to isolate the inference competence specific for social cues (i.e. mentalization), as detailed below. The primary outcome was RMET and SET total completion time, defined as the total time required to complete the test, comparing DM1 patients and healthy controls. Secondary outcomes included completion times of the SET subcomponents (SET-CI, SET-IA, SET-EA), as well as accuracy-based measures of social cognition (RMET accuracy, RMET-M/F accuracy, and SET total and subcomponent accuracy scores). Secondary analyses evaluated the relationship between completion-time measures and general cognition, and whether temporal metrics provided incremental information beyond conventional accuracy-based measures. 2.1. Statistical Analysis Statistical analyses were performed using JASP (Version 0.95.4). All variables were tested for normality (Shapiro–Wilk) and homoscedasticity (Levene). Between group differences were evaluated using Student's t-tests when normality and homoscedasticity assumptions were met, Welch’s t-tests when normality was met but variances were unequal, and Mann-Whitney U tests otherwise. Categorical comparisons used χ2 or Fisher’s exact tests as appropriate. Effect sizes were summarized as Hedges’ g (parametric) or rank biserial r (non-parametric). To examine within-group associations, Spearman correlations were computed among social-cognitive measures (RMET, SET) and executive/visuoconstructive variables. These analyses were considered exploratory and aimed at describing patterns of association rather than testing specific a priori hypotheses. Differences in correlation strength between groups were assessed with Fisher’s Z. To test the moderation effect of the clinical diagnosis on the relationship between executive functions and social cognition processing speeds, hierarchical multiple linear regressions were performed for SET and RMET completion times. The models were built in three steps: demographic variables (age, education) and one executive function measure were entered in Step 1; Group status (DM1 vs. Controls) in Step 2; and the interaction term between the executive function and Group status in Step 3. This approach also aimed to reduce model complexity relative to sample size and to limit the risk of overfitting. To avoid multicollinearity issues in models involving interaction terms, continuous executive function predictors (Stroop time, ROCF copy score, and TMT B-A time) were grand-mean centered prior to analysis. To evaluate the independent diagnostic predictive value of social cognition measures beyond general executive dysfunction, hierarchical logistic regressions were employed with diagnosis as the binary dependent variable. Age, education, and one baseline executive measure (Stroop time, ROCF Copy, or TMT B-A time) were entered in Step 1 as covariates, followed by specific social cognition metrics (time or accuracy) in Step 2. Odds ratios (ORs) with 95% confidence intervals (CIs) were reported. Additionally, to isolate the specific drivers of social cognition deficits and directly compare the predictive weights of SET versus RMET measures, backward stepwise logistic regressions were conducted incorporating SET subcomponents and combined SET/RMET scores. Collinearity diagnostics, including Tolerance and Variance Inflation Factor (VIF), were examined for all regression models to ensure values remained within acceptable limits. To further examine whether the between-group differences in mentalizing-task completion times were independent of the general cognitive and task-related demands captured by SET-CI, a hierarchical set of complementary analyses was performed. First, two separate multiple linear regression models were fitted, with SET-EA time and SET-IA time entered as dependent variables. Group (DM1 vs. HC) and SET-CI completion time were simultaneously entered as predictors. These initial additive models assessed whether group status remained associated with completion time in each mentalizing condition after accounting for individual variability in the non-mentalizing control condition. Second, the assumption that the relationship between SET-CI time and mentalizing-task time was comparable across groups was formally tested using separate analyses of covariance. SET-IA time and SET-EA time were entered as dependent variables, Group as a fixed factor, and grand-mean-centred SET-CI time as a continuous covariate. The Group × centred SET-CI-time interaction was included in each model. Centring was performed using the mean SET-CI completion time calculated across the entire sample. A significant interaction was interpreted as evidence that the association between control-condition and mentalizing-condition completion times differed between DM1 participants and controls, thereby precluding the interpretation of a single common adjusted group effect across the full range of SET-CI performance. Third, a repeated-measures ANCOVA was conducted to compare the two mentalizing conditions directly. Subtest, with two levels corresponding to SET-IA time and SET-EA time, was specified as the within-subject factor; Group was entered as the between-subject factor; and grand-mean-centred SET-CI time was included as a covariate. The Subtest × Group interaction tested whether the magnitude of the DM1–control difference differed between intention and emotion attribution. The Subtest × Group × SET-CI-time interaction tested whether any group-dependent association between SET-CI and mentalizing completion times differed across the two mentalizing conditions. Finally, following the geometric framework proposed by Dodich et al., originally developed for accuracy scores, the same vectorial analysis was adapted to completion-time measures. 57 Two two-dimensional Cartesian spaces were constructed for each participant: SET-CI time on the x-axis and SET-IA time on the y-axis, and SET-CI time on the x-axis and SET-EA time on the y-axis. Cartesian coordinates were transformed into polar coordinates to derive vector length, (d), and vector angle, (α). Because the present analysis was based on completion times rather than accuracy scores, greater vector length represented greater combined completion time, and therefore lower overall processing efficiency, across the control and mentalizing conditions. The vector angle represented the relative proportional distribution of completion time between the mentalizing and control conditions. Separate one-way ANOVAs were conducted for the vector length and angle derived from each configuration, with Group as the between-subject factor. A group difference in vector length without a corresponding angular difference was interpreted as evidence of a predominantly quantitative increase in completion time without a disproportionate imbalance between mentalizing and causal-inference processing. Finally, the discriminative clinical accuracy of social cognition tasks without covariate adjustment was assessed using univariate Receiver Operating Characteristic (ROC) curves. The Area Under the Curve (AUC), sensitivity, and specificity were calculated based on the optimal cut-off points determined by the Youden index. Statistical significance was set at α = 0.05 (two-tailed). 3. RESULTS A total of 33 DM1 (Female=19) and 40 HC (Female=24) participants were included. Groups did not differ in age (DM1: 42.61 ± 13.22 years; HC: 44.20 ± 17.68 years; p = 0.670) or years of education (DM1: 12.03 ± 3.0; HC: 13.15 ± 4.18; p = 0.116). MoCA performance was comparable between groups (DM1: 25.52 ± 3.35; HC: 25.45 ± 3.37; p= 0.875) (Table 1). There were no significant differences in the gender proportions between the groups (χ2=0.044, p=0.834).Table 1.Sample demographics and global cognition.​HC (n=40, F=24)DM1 (n=33, F=19)StatisticpAge (y) mean (SD)44.20 (17.68)42.61 (13.22)T=0.4280.670Education (y) mean (SD)13.15 (4.18)12.03 (3.01)U=796.0000.116MoCA mean (SD)25.45 (3.37)25.52 (3.35)U=606.0000.875Note. For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. HC = healthy controls; DM1 = myotonic dystrophy type 1; F = female; MoCA = Montreal Cognitive Assessment; SD = standard deviation; T = Student's t-test; U = Mann-Whitney U test. Table 1.Sample demographics and global cognition.​HC (n=40, F=24)DM1 (n=33, F=19)StatisticpAge (y) mean (SD)44.20 (17.68)42.61 (13.22)T=0.4280.670Education (y) mean (SD)13.15 (4.18)12.03 (3.01)U=796.0000.116MoCA mean (SD)25.45 (3.37)25.52 (3.35)U=606.0000.875Note. For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. HC = healthy controls; DM1 = myotonic dystrophy type 1; F = female; MoCA = Montreal Cognitive Assessment; SD = standard deviation; T = Student's t-test; U = Mann-Whitney U test. Sample demographics and global cognition. Note. For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. HC = healthy controls; DM1 = myotonic dystrophy type 1; F = female; MoCA = Montreal Cognitive Assessment; SD = standard deviation; T = Student's t-test; U = Mann-Whitney U test. Regarding the clinical and genetic profile of the DM1 cohort, the mean disease duration was 24.0 years (SD = 8.6 years), with the age at disease onset ranging from 15 to 58 years, including n=6 juvenile-onset, n=24 adult-onset and n=3 late onset patients. Age at the time of evaluation was predictably different between these groups (F(2,30)=6.744, p=0.004), but education was similar (F(2,30)=0.571, p=0.571). Molecular analysis of the CTG repeat expansion in the 3′-untranslated region (3′-UTR) of the DMPK gene stratified the patients into three established expansion size ranges: E1 (n = 5), E2 (n = 21), and E3 (n = 3). In four cases CTG repeat expansion was confirmed, but the expansion size was not determined. No differences were observed between these groups in terms of age (F(1,26)=3.205, p=0.057) or education (F(1,26)=0.570, p=0.572). Disease severity and motor impairment, as assessed by the Muscular Impairment Rating Scale (MIRS), revealed the following sample distribution: grade 1 (MIRS1, n = 1), grade 2 (MIRS2, n = 12), grade 3 (MIRS3, n = 13), and grade 4 (MIRS4, n = 7). As expected, more severely affected patients were older, with statistically significant differences between MIRS2 (mean age 33.76±9.98 years) and both MIRS 3 (mean age 45.85±10.63 years) and MIRS4 patients (mean age 54.29±11.63 years; Tukey-adjusted p=0.020 and p<0.001, respectively). No differences in education were observed (F(2,29)=0.328, p=0.723). 3.1. Between-groups comparison General cognitive tasks (Table 2) revealed a consistent pattern of cognitive slowing and reduced fluency in DM1. Patients showed poorer performance on Stroop, TMT, verbal fluency, and ROCF-copy measures (Figure 1).Table 2.Group comparison – general cognitive performances.​DM1HCStatisticEffect Size (SE)pMean (SD)Mean (SD)Stroop_Time25.28 (10.97)19.87 (15.82)U=398.000-0.378 (0.137) 0.006 Stroop_Interference2.29 (1.95)0.75 (1.34)U=362.500-0.434 (0.137) < 0.001 TMT A46.41 (15.80)29.2 (12.12)T=-5.122-1.235 (0.277) < 0.001 TMT B126.27 (58.40)72.43 (27.91)U=240.000-0.576 (0.142) < 0.001 TMT B-A83.76 (50.68)44.31 (22.44)U=321.500-0.431 (0.142) 0.003 Phonemic Fluency33.03 (13.74)41.25 (12.99)T=2.6010.610 (0.244) 0.011 Semantic Fluency41.75 (10.95)52.42 (10.52)U=998.0000.559 (0.137) < 0.001 ROCF_Copy27.53 (6.34)35.00 (1.88)U=1113.0000.739 (0.137) < 0.001 Note. Independent Samples T-Test or Mann-Whitney U-test - For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation (r). HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A; ROCF = Rey-Osterrieth Complex Figure; T = Student's t-test; U = Mann-Whitney U test. Bold values indicate statistically significant results (p < 0.05).Figure 1.Group comparison between DM1 and HC in performances in executive and visuospatial/visuoconstructive tasks (general cognition). ROCF= Rey Osterrieth Complex Figure. TMT_B-A= Trail Making Test B-A. Table 2.Group comparison – general cognitive performances.​DM1HCStatisticEffect Size (SE)pMean (SD)Mean (SD)Stroop_Time25.28 (10.97)19.87 (15.82)U=398.000-0.378 (0.137) 0.006 Stroop_Interference2.29 (1.95)0.75 (1.34)U=362.500-0.434 (0.137) < 0.001 TMT A46.41 (15.80)29.2 (12.12)T=-5.122-1.235 (0.277) < 0.001 TMT B126.27 (58.40)72.43 (27.91)U=240.000-0.576 (0.142) < 0.001 TMT B-A83.76 (50.68)44.31 (22.44)U=321.500-0.431 (0.142) 0.003 Phonemic Fluency33.03 (13.74)41.25 (12.99)T=2.6010.610 (0.244) 0.011 Semantic Fluency41.75 (10.95)52.42 (10.52)U=998.0000.559 (0.137) < 0.001 ROCF_Copy27.53 (6.34)35.00 (1.88)U=1113.0000.739 (0.137) < 0.001 Note. Independent Samples T-Test or Mann-Whitney U-test - For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation (r). HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A; ROCF = Rey-Osterrieth Complex Figure; T = Student's t-test; U = Mann-Whitney U test. Bold values indicate statistically significant results (p < 0.05). Group comparison – general cognitive performances. Note. Independent Samples T-Test or Mann-Whitney U-test - For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation (r). HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A; ROCF = Rey-Osterrieth Complex Figure; T = Student's t-test; U = Mann-Whitney U test. Bold values indicate statistically significant results (p < 0.05). Group comparison between DM1 and HC in performances in executive and visuospatial/visuoconstructive tasks (general cognition). ROCF= Rey Osterrieth Complex Figure. TMT_B-A= Trail Making Test B-A. Social-cognition measures showed different patterns for accuracy and completion time between DM1 and HC. (Tables 3 and 4; Figures 2 and 3). RMET accuracy was lower in DM1, whereas SET total and subscale accuracy scores were comparable between groups. Completion times were longer in DM1 across the RMET and all SET measures.Table 3.Group comparison – Social cognition accuracy performances.​DM1HCStatisticEffect Size (SE)pMean (SD)Mean (SD)RMET19.84 (3.75)24.12 (3.86)T=4.7371.111 (0.265) < 0.001 SET-tot14.09 (3.39)15.32 (2.49)U=814.0000.2330.136SET-IA4.70 (1.38)5.18 (0.84)U=760.5000.152 (0.136)0.242SET-EA4.91 (1.21)5.20 (0.99)U=742.5000.125 (0.136)0.330SET-CI4.36 (1.52)4.92 (1.25)U=819.0000.241 (0.136)0.067RMET M/F34.97 (1.03)34.83 (2.25)U=676.0000.127 (0.140)0.338Note. Independent Samples T-Test or Mann-Whitney U-test - For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; RMET = Reading the Mind in the Eyes Test; SET = Story-based Empathy Task; IA = intention attribution; EA = emotion attribution; CI = causal inference; RMET M/F = Reading the Mind in the Eyes Test, male/female discrimination control condition; T = Student's t-test; U = Mann-Whitney U test. Bold values indicate statistically significant results (p < 0.05).Table 4.Group comparison – Social cognition completion times.​DM1HCStatisticEffect Size (SE)pMean (SD)Mean (SD)RMET time462.81 (154.51)326.95 (127.11)U=279.500-0.563 (0.140) < 0.001 SET total time405.24 (186.88)240.30 (107.24)U=247.000-0.626 (0.136) < 0.001 SET-EA time139.42 (70.52)90.62 (45.30)U=327.500-0.504 (0.136) < 0.001 SET-IA time141.91 (79.06)73.77 (31.71)U=208.500-0.684 (0.136) < 0.001 SET-CI time123.85 (55.78)75.80 (40.20)U=271.500-0.589 (0.136) < 0.001 Note. For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. Completion times are reported in seconds. HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; RMET = Reading the Mind in the Eyes Test; SET = Story-based Empathy Task; IA = intention attribution; EA = emotion attribution; CI = causal inference; T = Student's t-test; U = Mann–Whitney U test. Bold values indicate statistically significant results (p < 0.05).Figure 2.Group comparison between DM1 and HC in accuracy scores on Social Cognition tasks. SET= Story Based Empathy Test; -EA= Emotion Attribution; -IA= Intention Attribution; -CI= Causal Inference; RMET= Reading the Mind in the Eyes (accuracy, i.e. number of expressions correctly identified); RMET M/F=number of correctly assigned gender in the RMET set of figures. ns= not significantFigure 3.Group comparison between DM1 and HC in task completion times on Social Cognition tasks. SET= Story Based Empathy Test; -EA= Emotion Attribution; -IA= Intention Attribution; -CI= Causal Inference; RMET= Reading the Mind in the Eyes. Table 3.Group comparison – Social cognition accuracy performances.​DM1HCStatisticEffect Size (SE)pMean (SD)Mean (SD)RMET19.84 (3.75)24.12 (3.86)T=4.7371.111 (0.265) < 0.001 SET-tot14.09 (3.39)15.32 (2.49)U=814.0000.2330.136SET-IA4.70 (1.38)5.18 (0.84)U=760.5000.152 (0.136)0.242SET-EA4.91 (1.21)5.20 (0.99)U=742.5000.125 (0.136)0.330SET-CI4.36 (1.52)4.92 (1.25)U=819.0000.241 (0.136)0.067RMET M/F34.97 (1.03)34.83 (2.25)U=676.0000.127 (0.140)0.338Note. Independent Samples T-Test or Mann-Whitney U-test - For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; RMET = Reading the Mind in the Eyes Test; SET = Story-based Empathy Task; IA = intention attribution; EA = emotion attribution; CI = causal inference; RMET M/F = Reading the Mind in the Eyes Test, male/female discrimination control condition; T = Student's t-test; U = Mann-Whitney U test. Bold values indicate statistically significant results (p < 0.05). Group comparison – Social cognition accuracy performances. Note. Independent Samples T-Test or Mann-Whitney U-test - For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; RMET = Reading the Mind in the Eyes Test; SET = Story-based Empathy Task; IA = intention attribution; EA = emotion attribution; CI = causal inference; RMET M/F = Reading the Mind in the Eyes Test, male/female discrimination control condition; T = Student's t-test; U = Mann-Whitney U test. Bold values indicate statistically significant results (p < 0.05). Table 4.Group comparison – Social cognition completion times.​DM1HCStatisticEffect Size (SE)pMean (SD)Mean (SD)RMET time462.81 (154.51)326.95 (127.11)U=279.500-0.563 (0.140) < 0.001 SET total time405.24 (186.88)240.30 (107.24)U=247.000-0.626 (0.136) < 0.001 SET-EA time139.42 (70.52)90.62 (45.30)U=327.500-0.504 (0.136) < 0.001 SET-IA time141.91 (79.06)73.77 (31.71)U=208.500-0.684 (0.136) < 0.001 SET-CI time123.85 (55.78)75.80 (40.20)U=271.500-0.589 (0.136) < 0.001 Note. For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. Completion times are reported in seconds. HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; RMET = Reading the Mind in the Eyes Test; SET = Story-based Empathy Task; IA = intention attribution; EA = emotion attribution; CI = causal inference; T = Student's t-test; U = Mann–Whitney U test. Bold values indicate statistically significant results (p < 0.05). Group comparison – Social cognition completion times. Note. For the Student t-test (T), effect size is given by Hedges’ g. For the Mann-Whitney test (U), effect size is given by the rank biserial correlation. Completion times are reported in seconds. HC = healthy controls; DM1 = myotonic dystrophy type 1; SD = standard deviation; RMET = Reading the Mind in the Eyes Test; SET = Story-based Empathy Task; IA = intention attribution; EA = emotion attribution; CI = causal inference; T = Student's t-test; U = Mann–Whitney U test. Bold values indicate statistically significant results (p < 0.05). Group comparison between DM1 and HC in accuracy scores on Social Cognition tasks. SET= Story Based Empathy Test; -EA= Emotion Attribution; -IA= Intention Attribution; -CI= Causal Inference; RMET= Reading the Mind in the Eyes (accuracy, i.e. number of expressions correctly identified); RMET M/F=number of correctly assigned gender in the RMET set of figures. ns= not significant Group comparison between DM1 and HC in task completion times on Social Cognition tasks. SET= Story Based Empathy Test; -EA= Emotion Attribution; -IA= Intention Attribution; -CI= Causal Inference; RMET= Reading the Mind in the Eyes. 3.2. Correlational analyses The correlation pattern differed between HC and DM1 in magnitude and geometry, indicating distinct cognitive architectures (Figure 4). In HC, the strongest associations emerged between semantic fluency, ROCF-copy, and affective mental-state decoding. In DM1, the broad coupling between semantic fluency and social-cognitive accuracy weakened, while phonemic fluency emerged as a stronger correlate. Stroop time showed extensive associations in DM1 across both accuracy and timing indices of social cognition. By contrast, TMT-B and TMT-B–A were weakly and inconsistently related to social-cognitive measures in DM1.Figure 4.Correlation matrix of executive-visuospatial/visuoconstructive performances and Social Cognition metrics both in accuracy and timing, in HC and DM1 patients. SET= Story Based Empathy Test; -EA= Emotion Attribution; -IA= Intention Attribution; -CI= Causal Inference; RMET= Reading the Mind in the Eyes. Note. Spearman’s correlation coefficient ρ. * p<0.05 ** p<0.01, ***p<0.001. Correlation matrix of executive-visuospatial/visuoconstructive performances and Social Cognition metrics both in accuracy and timing, in HC and DM1 patients. SET= Story Based Empathy Test; -EA= Emotion Attribution; -IA= Intention Attribution; -CI= Causal Inference; RMET= Reading the Mind in the Eyes. Note. Spearman’s correlation coefficient ρ. * p<0.05 ** p<0.01, ***p<0.001. Between-group comparisons of correlation magnitudes (Fisher’s Z) yielded only a limited number of robust interaction effects at the present sample size. RMET accuracy correlated more strongly with phonemic fluency in DM1 than in controls, while SET-IA accuracy correlated more strongly with ROCF-copy in HC than in DM1. 3.3. Regression and ROC analyses Across all models, DM1 diagnosis emerged as the strongest independent predictor of social-cognitive completion times. Among social cognition measures, completion-time metrics showed greater discriminative performance than accuracy-based measures for SET, whereas RMET completion time and accuracy demonstrated comparable diagnostic value. Predictors of Social Cognition task completion times Hierarchical multiple linear regressions assessed the predictive roles of executive functions and clinical diagnosis on social cognition processing speeds. SET Total Time. Baseline models showed that executive functions initially predicted SET total time (Stroop: p = 0.005; ROCF copy: p < 0.001; TMT B-A: p <0.001). However, accounting for clinical diagnosis yielded a highly significant model improvement (ΔR2 range: 0.132–0.205, all p < 0.001), absorbing the direct predictive power of executive-loading functions (Stroop: p =0 .061; ROCF: p = 0.787; TMT: p =0 .162). Group status emerged as the strongest independent predictor (p < 0.001), indicating that DM1 patients were substantially slower regardless of the executive-loading measure entered in Step 1 of the hierarchical regression model. No significant disease-driven moderation effects were observed (all interaction p > 0.130; see Table 5).Table 5.Summary of hierarchical linear regression models predicting social cognition processing speeds.Dependent VariableExecutive-loading CovariateModel 1 (M0)model 2 Group MAIN EFFECT (M1)model 3 Interaction (M2)Model R2ΔR2Group B (SE)Group pInteraction ΔR2 (p)SET Total TimeStroop Time0.305***0.205***138.82 (26.20) < 0.001 0.017 (0.131)ROCF Copy0.346***0.138***145.88 (34.44) < 0.001 0.007 (0.351)TMT B-A0.395***0.132***113.15 (26.96) < 0.001 0.011 (0.220)RMET TimeStroop Time0.0510.163***120.87 (32.62) < 0.001 0.000 (0.864)ROCF Copy0.0850.138**143.72 (41.94) 0.001 0.009 (0.389)TMT B-A0.0690.142**123.56 (37.01) 0.001 0.008 (.428)Notes. B = unstandardized regression coefficient for group status (DM1) from step 2 (M1). All models included age and education as covariates. Executive covariates were grand-mean centered. *p < 0.05; **p < 0.01; *** p < 0.001. SE = standard error; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A; ROCF = Rey–Osterrieth Complex Figure. Bold values indicate statistically significant results (p < 0.05). Table 5.Summary of hierarchical linear regression models predicting social cognition processing speeds.Dependent VariableExecutive-loading CovariateModel 1 (M0)model 2 Group MAIN EFFECT (M1)model 3 Interaction (M2)Model R2ΔR2Group B (SE)Group pInteraction ΔR2 (p)SET Total TimeStroop Time0.305***0.205***138.82 (26.20) < 0.001 0.017 (0.131)ROCF Copy0.346***0.138***145.88 (34.44) < 0.001 0.007 (0.351)TMT B-A0.395***0.132***113.15 (26.96) < 0.001 0.011 (0.220)RMET TimeStroop Time0.0510.163***120.87 (32.62) < 0.001 0.000 (0.864)ROCF Copy0.0850.138**143.72 (41.94) 0.001 0.009 (0.389)TMT B-A0.0690.142**123.56 (37.01) 0.001 0.008 (.428)Notes. B = unstandardized regression coefficient for group status (DM1) from step 2 (M1). All models included age and education as covariates. Executive covariates were grand-mean centered. *p < 0.05; **p < 0.01; *** p < 0.001. SE = standard error; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A; ROCF = Rey–Osterrieth Complex Figure. Bold values indicate statistically significant results (p < 0.05). Summary of hierarchical linear regression models predicting social cognition processing speeds. Notes. B = unstandardized regression coefficient for group status (DM1) from step 2 (M1). All models included age and education as covariates. Executive covariates were grand-mean centered. *p < 0.05; **p < 0.01; *** p < 0.001. SE = standard error; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A; ROCF = Rey–Osterrieth Complex Figure. Bold values indicate statistically significant results (p < 0.05). Additive linear regression models. In the preliminary additive model predicting SET-EA completion time, Group was not a significant independent predictor after adjustment for SET-CI time (t = 0.288, p =0.774), whereas SET-CI time was significantly associated with SET-EA time (B=0.816, 95% CI (0.782, 1.136), t=10.814, p<0.001). Thus, the initial between-group difference in emotion-attribution completion time was substantially attenuated after accounting for variability in the control condition. Conversely, in the preliminary model predicting SET-IA completion time, Group remained a significant predictor (t = 2.412, p=0.018), although the main effect of SET-CI time was also significant (B=706, 95% CI (0.689, 1.083), t=8.985, p<0.001) This additive model initially suggested that prolonged intention-attribution time in DM1 could not be entirely accounted for by performance in the control condition. However, this result was subsequently qualified by the significant Group × SET-CI-time interaction. Separate ANCOVA models. For SET-EA time, the Group × centred SET-CI-time interaction was not significant, (F(1,69)=1.170), (p=0.283), indicating no evidence that the association between SET-CI and SET-EA completion times differed between DM1 participants and healthy controls. The assumption of homogeneous regression slopes was therefore considered acceptable for the EA model. Consistent with the preliminary additive regression, SET-CI time accounted for a substantial proportion of EA-time variability, whereas no independent adjusted group effect was detected. For SET-IA time, the Group × centred SET-CI-time interaction was significant, (F(1,69)=5.882), (p=0.018), (η2=0.038). Thus, the relationship between SET-CI and SET-IA completion times differed between groups. Accordingly, the effect of Group on IA time could not be represented by a single adjusted mean difference applicable across the entire range of SET-CI times. The main effect of Group in this interaction model represents the DM1–control difference specifically at the overall mean SET-CI time, whereas the magnitude of the between-group difference changes as a function of SET-CI performance. Repeated-measures ANCOVA. The Subtest × centred SET-CI-time interaction was not significant, (F(1,69)=0.809), (p=0.372), indicating that, when averaged across groups, the overall relationship between SET-CI time and mentalizing completion time did not differ significantly between the IA and EA conditions. The Subtest × Group interaction did not reach the conventional threshold for statistical significance, although the observed effect approached it, (F(1,69)=3.785), (p=0.056). Therefore, the magnitude of the between-group difference could not be shown to differ significantly between intention attribution and emotion attribution. At the between-subject level, the Group × centred SET-CI-time interaction was significant, (F(1,69)=5.074), (p=0.027), indicating that the relationship between SET-CI time and the completion time averaged across the two mentalizing conditions differed between DM1 participants and controls. Crucially, the Subtest × Group × centred SET-CI-time interaction was not significant, (F(1,69)=1.433), (p=0.235). Thus, although the Group × SET-CI-time interaction was significant in the separate IA model but not in the EA model, the difference between these two interaction effects was not itself statistically significant. The analyses therefore did not provide formal evidence that the group-dependent relationship between control-condition and mentalizing-task completion times was selectively altered for intention attribution. Vectorial analysis. In the SET-CI–SET-IA configuration, a significant effect of Group was found for vector length, (F(1,71)=24.92), p<0.001), η2=0.260). Participants with DM1 showed greater vector lengths than healthy controls, reflecting a greater combined completion time across causal inference and intention attribution. In contrast, vector angle did not differ significantly between groups, (F(1,71)=2.198), p=0.143, η2=0.030). A comparable pattern emerged for the SET-CI–SET-EA configuration. Vector length differed significantly between groups, (F(1,71)=16.53, p<0.001), η2=0.189), with greater values in the DM1 group, whereas no significant group difference was found for vector angle, (F(1,71)=1.321, p=254, η2=0.018). Taken together, the vectorial analyses indicated an overall increase in completion time in DM1, without evidence of a systematic group-level alteration in the proportional balance between mentalizing and causal-inference processing. RMET Time. Baseline demographic and executive models failed to predict RMET processing speed (all p > 0.110). Consistent with the SET findings, adding Group status significantly improved all models (ΔR2 range: 0.138–0.163, all p ≤ 0.005), acting as the sole significant predictor (p ≤ 0.001). Similarly, no significant interaction effects were found (all p> 0.380; see Table 5). Independent Diagnostic Predictive Value of Social Cognition Hierarchical logistic regressions were performed to evaluate the unique diagnostic value of social cognition impairments beyond general executive dysfunction. RMET. Executive-loaded measures entered in Step 1 of the hierarchical regression models strongly predicted clinical status, with the ROCF Copy serving as the most robust general cognitive benchmark (Nagelkerke R2 range: 0.695–0.738, p < 0.001). Crucially, introducing RMET performance significantly improved model fits. Longer RMET completion time provided unique diagnostic utility beyond the Stroop, TMT B-A, and ROCF covariates (all Δχ2 p ≤ 0.001). Lower RMET accuracy similarly improved model fits beyond all three executive measures (all Δχ2 p ≤ 0.014). In their respective final models, both RMET time and accuracy remained significant independent predictors of a DM1 diagnosis (all ps ≤ 0.029; see Table 6).Table 6.summary of hierarchical logistic regression models predicting DM1 diagnosis.Social Cognition PredictorExecutive-loading CovariateModel FitpPredictor IndependencepModel AUCΔχ2Odds Ratio (95% CI)RMET TimeStroop Time13.06 < 0.001 1.008 0.003 0.821ROCF Copy10.59 0.001 1.012 0.005 0.965TMT B-A11.12 < 0.001 1.009 0.004 0.888RMET AccuracyStroop Time14.09 < 0.001 0.733 0.001 0.806ROCF Copy6.10 0.014 0.695 0.029 0.971TMT B-A6.66 0.010 0.770 0.016 0.866SET Total TimeStroop Time26.45 < 0.001 1.016 < 0.001 0.866ROCF Copy11.93 < 0.001 1.019 0.006 0.967TMT B-A17.91 < 0.001 1.016 < 0.001 0.897SET AccuracyStroop Time0.370.5440.9400.5460.716ROCF Copy0.980.3221.1760.335--TMT B-A0.010.9111.0140.9110.808Note. All models included age and education as covariates in step 1. Executive covariates were grand-mean centered. AUC= area under the curve; RMET = Reading the Mind in the Eyes Test; ROCF = Rey–Osterrieth Complex Figure; SET= Story-based Empathy Task; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A. Bold values indicate statistically significant results (p < 0.05). Table 6.summary of hierarchical logistic regression models predicting DM1 diagnosis.Social Cognition PredictorExecutive-loading CovariateModel FitpPredictor IndependencepModel AUCΔχ2Odds Ratio (95% CI)RMET TimeStroop Time13.06 < 0.001 1.008 0.003 0.821ROCF Copy10.59 0.001 1.012 0.005 0.965TMT B-A11.12 < 0.001 1.009 0.004 0.888RMET AccuracyStroop Time14.09 < 0.001 0.733 0.001 0.806ROCF Copy6.10 0.014 0.695 0.029 0.971TMT B-A6.66 0.010 0.770 0.016 0.866SET Total TimeStroop Time26.45 < 0.001 1.016 < 0.001 0.866ROCF Copy11.93 < 0.001 1.019 0.006 0.967TMT B-A17.91 < 0.001 1.016 < 0.001 0.897SET AccuracyStroop Time0.370.5440.9400.5460.716ROCF Copy0.980.3221.1760.335--TMT B-A0.010.9111.0140.9110.808Note. All models included age and education as covariates in step 1. Executive covariates were grand-mean centered. AUC= area under the curve; RMET = Reading the Mind in the Eyes Test; ROCF = Rey–Osterrieth Complex Figure; SET= Story-based Empathy Task; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A. Bold values indicate statistically significant results (p < 0.05). summary of hierarchical logistic regression models predicting DM1 diagnosis. Note. All models included age and education as covariates in step 1. Executive covariates were grand-mean centered. AUC= area under the curve; RMET = Reading the Mind in the Eyes Test; ROCF = Rey–Osterrieth Complex Figure; SET= Story-based Empathy Task; TMT = Trail Making Test; TMT B-A = completion time of Part B minus completion time of Part A. Bold values indicate statistically significant results (p < 0.05). SET. SET total completion time significantly improved prediction of DM1 status, whereas SET accuracy did not. Including SET total time significantly improved prediction models over the executive covariates entered at Step 1 (all Δχ2 ps ≤ 0.001), with slower times independently predicting a DM1 diagnosis. Conversely, SET accuracy did not improve model fits beyond the executive function measures entered at Step 1 (all Δχ2 > 0.320), suggesting DM1 primarily affects SET processing efficiency rather than accuracy under these task conditions (see Table 6). In backward stepwise logistic regressions, among the SET sub-components completion times, SET-IA time was the sole surviving significant predictor of DM1 diagnosis (Odds Ratio = 1.065, p < 0.001). Furthermore, when directly comparing SET tasks versus RMET performance models (adjusted for the ROCF benchmark), both SET total time and (p = 0.045), RMET accuracy (p = 0.025) maintained a significant classification effect. Discriminative Diagnostic Accuracy (ROC Curve Analysis) Unadjusted ROC curve analyses assessed the pure discriminative ability of social cognition tests against the established ROCF Copy benchmark (Figure 5). The ROCF Copy yielded the highest discriminative accuracy (AUC = 0.870; optimal cut-off: 32.50, sensitivity: 71.9%, specificity: 87.5%). Within social cognition variables, SET processing speed measures yielded higher AUC values than accuracy-based measures. SET total time yielded good discriminative ability (AUC = 0.813; cut-off: 326.5 sec). In line with the backward regression findings, the Intention Attribution subtest (SET-IA time) emerged as the most accurate social cognition classifier, reaching an AUC of 0.842 (cut-off: 95.5 sec) and closely approaching the reference executive measure. RMET tasks demonstrated moderate-to-good discriminative accuracy, with RMET accuracy (AUC = 0.787; cut-off: 21.5) and RMET time (AUC = 0.782; cut-off: 412 sec) showing comparable clinical value. Conversely, overall SET accuracy yielded poor discriminative ability (AUC = 0.617), confirming its lack of clinical utility for identifying DM1 patients compared to processing speed metrics (see Table 7).Figure 5.ROC curves. ROCF= Rey Osterrieth Complex Figure. SET= Story Based Empathy Test; -EA= Emotion Attribution; -IA= Intention Attribution; -CI= Causal Inference; RMET= Reading the Mind in the Eyes.Table 7.discriminative diagnostic accuracy of clinical metrics (univariate roc analysis).Clinical MeasureAUCAccuracySensitivitySpecificityOptimal Cut-off General cognition Benchmark ROCF Copy0.87080.6%71.9%87.5%32.50 Social Cognition: Theory of Mind SET Accuracy0.61756.2%36.4%72.5%--SET Total Time (sec)0.81371.2%60.6%80.0%326.5SET-IA Time (sec)0.84272.6%75.7%80.0%95.5SET-EA Time (sec)0.75261.6%51.2%75.0%117.0SET-CI Time (sec)0.79471.2%57.6%85.0%101.5 Social Cognition: Emotion Recognition RMET Accuracy0.78769.4%71.9%67.5%21.5RMET Time (sec)0.78268.1%53.1%80.0%412.0Notes. Sensitivity and Specificity are reported for the optimal cut-off point (Youden’s Index). Cut-offs for time metrics indicate that scores higher than the cut-off predict DM1. Cut-offs for accuracy metrics indicate that scores lower than the cut-off predict DM1. AUC = Area Under the Curve; ROC = receiver operating characteristic; SET = Story-based Empathy Task; IA = intentionality attribution; CI = cognitive inference; EA = emotion attribution; RMET = Reading the Mind in the Eyes Test. ROC curves. ROCF= Rey Osterrieth Complex Figure. SET= Story Based Empathy Test; -EA= Emotion Attribution; -IA= Intention Attribution; -CI= Causal Inference; RMET= Reading the Mind in the Eyes. Table 7.discriminative diagnostic accuracy of clinical metrics (univariate roc analysis).Clinical MeasureAUCAccuracySensitivitySpecificityOptimal Cut-off General cognition Benchmark ROCF Copy0.87080.6%71.9%87.5%32.50 Social Cognition: Theory of Mind SET Accuracy0.61756.2%36.4%72.5%--SET Total Time (sec)0.81371.2%60.6%80.0%326.5SET-IA Time (sec)0.84272.6%75.7%80.0%95.5SET-EA Time (sec)0.75261.6%51.2%75.0%117.0SET-CI Time (sec)0.79471.2%57.6%85.0%101.5 Social Cognition: Emotion Recognition RMET Accuracy0.78769.4%71.9%67.5%21.5RMET Time (sec)0.78268.1%53.1%80.0%412.0Notes. Sensitivity and Specificity are reported for the optimal cut-off point (Youden’s Index). Cut-offs for time metrics indicate that scores higher than the cut-off predict DM1. Cut-offs for accuracy metrics indicate that scores lower than the cut-off predict DM1. AUC = Area Under the Curve; ROC = receiver operating characteristic; SET = Story-based Empathy Task; IA = intentionality attribution; CI = cognitive inference; EA = emotion attribution; RMET = Reading the Mind in the Eyes Test. discriminative diagnostic accuracy of clinical metrics (univariate roc analysis). Notes. Sensitivity and Specificity are reported for the optimal cut-off point (Youden’s Index). Cut-offs for time metrics indicate that scores higher than the cut-off predict DM1. Cut-offs for accuracy metrics indicate that scores lower than the cut-off predict DM1. AUC = Area Under the Curve; ROC = receiver operating characteristic; SET = Story-based Empathy Task; IA = intentionality attribution; CI = cognitive inference; EA = emotion attribution; RMET = Reading the Mind in the Eyes Test. 4. DISCUSSION The present study contributes to characterizing the social-cognitive phenotype in DM1, delineating a non-uniform profile characterized by reduced accuracy in affective mental-state inference on the RMET, relatively preserved accuracy on the SET, and a pervasive reduction in processing efficiency across both tasks. These findings are consistent with previous reports of impaired facial emotion recognition27,30,31 and altered Theory of Mind in DM1,9,17,18,25,28,58 including recent evidence of a preferential impairment of affective rather than cognitive mentalizing. 58 At the same time, they may help reconcile and understand both clinical and research observations reporting normal or near-normal performance on some conventional social-cognition measures.17,18,28 Preserved SET accuracy should not necessarily be interpreted as evidence of fully intact processing, as patients may still reach the correct inference through slower, more effortful, or compensatory processing and the observed pattern may reflect a shift in the balance from relatively efficient to more controlled social-cognitive processing. Dual-process accounts propose that mental-state attribution can be supported by a rapid but relatively inflexible route and by a slower, more flexible route that places greater demands on cognitive control.59–61 Reduced efficiency of the former could therefore increase reliance on executive processes, including inhibition, updating, and cognitive flexibility, which contribute to explicit mentalizing and RMET performance.52,53 Such compensatory executive resources allocation might allow patients to maintain response accuracy at the cost of longer completion times, producing a form of processing inefficiency that accuracy-based scores alone may fail to detect.62,63 Previous studies have emphasized the contribution of executive control, mental flexibility, and social-contextual processing,18,28 and neuroimaging investigations have associated impaired social cognition with abnormalities involving distributed frontal, temporal, insular, and connecting white-matter networks.9,18,58 The present findings support an intermediate account: general executive and processing-speed limitations contribute to social-cognitive performance, but do not appear to fully explain the observed temporal inefficiency. Although DM1 is characterized by well-established impairments in completion-time performance, several observations argue against a purely domain-general explanation of the present findings. Importantly, the observed slowing remained significant after adjustment for executive measures, suggesting that it represents a distinct aspect of social cognitive performance. The analyses incorporating the SET-CI control condition nevertheless indicate that this temporal impairment is not entirely independent of general task-related cognitive demands. In the vectorial analysis applied, DM1 participants showed greater vector lengths but preserved vector angles in both the CI–IA and CI–EA configurations. This pattern suggests a predominantly quantitative increase in completion time, affecting mentalizing and causal-inference conditions in broadly proportional ways, rather than a clear group-level imbalance between mentalizing and non-mentalizing processing. In fact, for EA, the group effect was no longer significant after accounting for CI time, and the Group × CI interaction was not significant, indicating that emotion-attribution slowing was largely shared with the perceptual, inferential, and response-selection demands captured by the control condition. A more complex pattern emerged for IA, for which the relationship between CI and completion time differed between DM1 participants and controls. This result suggests that intention attribution may show an altered dependency on general processing resources or reduced automaticity in DM1. However, neither the Subtest × Group interaction nor the Subtest × Group × CI interaction reached statistical significance. Therefore, although IA time was the strongest individual SET predictor of clinical status, the present data do not establish a formal dissociation between intention and emotion attribution or demonstrate that the altered CI–mentalizing relationship is selectively specific to IA. Taken together, these findings reconcile the strong independent discriminative value of social-cognitive completion times with the contribution of broader cognitive inefficiency. SET and RMET slowing cannot be reduced entirely to executive dysfunction, as temporal measures retained predictive value beyond established executive and visuoconstructive benchmarks. At the same time, the control-condition and vectorial analyses show that a substantial component of SET slowing is shared with general task demands. Furthemore, time measures showed distinct patterns of association with executive variables, suggesting that slowed performance reflects more than a simple global slowing. These findings are consistent with the presence of a partially domain-specific component of social cognitive processing inefficiency. The most plausible interpretation is therefore not a purely selective loss of Theory of Mind, but reduced efficiency and automaticity of social information processing, superimposed on broader cognitive slowing, with intention attribution representing the most sensitive—but not definitively selective—SET marker of DM1. These findings may help explain the frequently observed discrepancy between relatively preserved neuropsychological test performance and everyday social difficulties in DM1. Social-cognitive processes may remain sufficiently intact to support correct responses under structured testing conditions yet operate with reduced efficiency. From a clinical perspective, the present findings suggest that accuracy-based assessments may underestimate functionally relevant aspects of social-cognitive performance. Incorporating temporal measures into neuropsychological evaluation may therefore provide a more sensitive characterization of social-cognitive abilities and improve the detection of clinically meaningful impairment. Similar observations have been documented in other conditions. In schizophrenia, reduced accuracy and prolonged response times have been observed across multiple social-cognitive domains, and the integration of reaction time with accuracy and confidence measures has provided greater discriminatory information than accuracy alone.35,36 Similarly, studies in Parkinson’s disease have reported abnormalities in both the accuracy and speed of facial-emotion recognition, 64 while Theory of Mind performance in multiple sclerosis has been closely associated with general information-processing speed. 65 Subtle or compensated impairments may remain undetected when evaluation relies exclusively on accuracy-based paradigms. The integration of temporal measures may therefore enhance the sensitivity of neuropsychological tools and improve the detection of clinically meaningful dysfunction. Some limitations should be acknowledged. RMET and SET were originally developed as accuracy-based instruments rather than as experimental reaction-time paradigms. Consequently, completion times likely reflect the integrated contribution of multiple processes involved in task performance, including comprehension, decision-making, response selection, and motor execution. Nevertheless, our regression analyses showed that group differences in completion time remained significant after accounting for executive and psychomotor measures, suggesting that the observed temporal delay was not entirely attributable to general cognitive inefficiency. Moreover, the documented discrepancy between RMET and SET accuracy may reflect differences in task structure and cognitive demands. The RMET relies on fine-grained decoding of subtle facial cues with limited contextual support, making it more sensitive to subtle impairments in affective mental-state inference. In contrast, the SET provides richer contextual and narrative information, which may facilitate compensatory strategies. A further limitation concerns the construct validity of the social-cognition measures. Although the RMET and SET are widely used and have been validated in Italian samples, the SET relies on static and simplified social scenarios that do not reproduce the reciprocal and rapidly changing nature of everyday interactions. Participants may therefore reach correct responses through deliberate reasoning, potentially masking difficulties that would emerge in more dynamic contexts. Therefore, completion times should not be interpreted as pure measures of social-cognitive reaction speed. Accordingly, completion times may provide a clinically informative index of the overall efficiency with which social information is processed and translated into a response, but this interpretation remains preliminary, because no ecological or real-world measures of social functioning were available to determine whether prolonged completion times are associated with everyday interpersonal difficulties. Social cognition relies on multiple interacting processes that operate dynamically as facial, prosodic, gestural, and contextual cues unfold over time.66,67 Consistent with recent international recommendations, contemporary approaches to social-cognitive assessment increasingly emphasize the need for new multidimensional frameworks integrating different social-cognitive domains and complementary performance measures, rather than relying on isolated accuracy-based scores alone (SIGNATURE initiative). 68 The present findings add to this literature by showing that, in DM1, completion-time measures obtained from conventional social-cognitive tasks may provide additional information about reduced processing efficiency that is not fully captured by accuracy scores. Nevertheless, adding temporal indices to established instruments such as the RMET and SET represents only a partial methodological advance, because these tasks remain highly structured, non-interactive, and only moderately representative of everyday social exchanges.66,69 Future studies should therefore combine accuracy and response-time measures with more dynamic and context-rich instruments, such as the Awareness of Social Inference Test, 70 the Movie for the Assessment of Social Cognition, 71 and the Edinburgh Social Cognition Test, 72 as well as with genuinely interactive paradigms 73 together with measures of real-world social functioning, to enhance the ecological validity and clinical relevance of social-cognitive assessment. No formal estimate of general or premorbid intellectual functioning was available. Therefore, although participants were characterized using a multidomain neuropsychological battery, the potential contribution of baseline intellectual ability to group differences cannot be fully excluded. Additionally, completion time measures were obtained using a stopwatch rather than computerized millisecond-precision recording. Although this approach may introduce measurement variability, all assessments were conducted under standardized conditions by a single trained examiner, ensuring methodological consistency across participants. Moreover, the observed group differences were robust and consistent across tasks, supporting the reliability of the findings. In addition, the cross-sectional design does not allow conclusions about the longitudinal evolution of these alterations, and the relatively small sample size may limit generalizability. Finally, the relatively large number of statistical analyses performed in relation to the available sample size represents an additional limitation. Although the primary hypothesis focused on between-group differences in social-cognitive completion times, several secondary analyses, including correlation analyses, regression models, and ROC analyses, were exploratory. Consequently, no formal adjustment for multiple comparisons was applied, and the risk of type I error and model overfitting cannot be excluded. Sensitivity analyses indicated that the achieved sample size was adequately powered to detect moderate-to-large effects, whereas smaller effects may have remained undetected. These findings should therefore be interpreted cautiously and require replication in larger independent cohorts. Overall, these findings support the hypothesis that the temporal inefficiency observed in DM1 is not merely a byproduct of broader executive or psychomotor dysfunction but represents a distinct aspect of social-cognitive performance. Ultimately, however, the primary value of these temporal metrics lies in their ability to capture an aspect of social-cognitive efficiency that is not fully represented by accuracy measures alone. Regardless of the precise neurocognitive mechanisms involved, measuring the overall delay in social-cognitive task performance may provide clinically relevant information that is not fully captured by examining its underlying cognitive components separately. Real-world social interactions do not depend on isolated social-cognitive operations, but rather on the efficient integration of these multiple cognitive processes within a limited time frame. Therefore, although completion times may not be considered pure measures of social-cognitive reaction speed, they may provide a clinically meaningful index of the overall efficiency with which social information is processed and translated into behavior. 5. CONCLUSIONS Overall, patients with DM1 showed relatively preserved accuracy across several social cognition measures, although reduced accuracy was observed on the RMET. In contrast, prolonged completion times were consistently detected across social cognition tasks. These findings suggest that completion-time metrics capture clinically relevant aspects of performance that may not be detected by conventional accuracy-based assessments alone. Incorporating temporal metrics into social cognition assessment may improve the detection of subtle differences in performance and may provide additional clinically relevant information regarding cognitive functioning in DM1.