Metacognition and executive functions facilitate adaptable and goal-oriented behavior throughout the lifespan. In detail, metacognition (MC) is conceptualized as the monitoring (metacognitive monitoring; MC-M) and control (metacognitive control; MC-C) of one's neurocognitive processes (Nelson and Narens, 1994; Flavell, 1979). The maturation of these components is non-uniform along growth and shows temporal discrepancies (Roebers, 2017). On the other hand, executive functions (EF) are key processes that support goal-directed decision making, strategy adaptation, and problem solving (Diamond, 2013). Similar to MC, the development of EF subcomponents (inhibition, working memory, cognitive flexibility) shows a non-uniform maturation (Best and Miller, 2010; Diamond, 2013). Notably, recent reviews (Roebers, 2017; Marulis et al., 2020; Fleming, 2024) highlight that MC and EF strongly interact both throughout development and during adulthood. Furthermore, both MC and EF are related to academic achievement (Roebers, 2017), career success (Chan et al., 2021; Jain et al., 2023), and wellbeing (Luerssen and Ayduk, 2017; Varshney and Barbey, 2021). Therefore, the literature strongly emphasizes the importance of developing early assessments and interventions that consider both educational and clinical contexts. Unfortunately, the majority of psychological and neuropsychological tests target EF (e.g., the Stroop task; MacLeod, 2005) or metacognition (e.g., perceptual decision-making task; Rahnev, 2021), thus failing to provide a satisfactory measure of both constructs and their interaction. This limitation forces researchers and clinicians to exploit multiple test batteries, thus making developmental experiments and assessments time-consuming and hard to interpret (Howieson, 2019). Notably, the Metacognitive Wisconsin Card Sorting Test (meta-WCST; Koren et al., 2004) introduces a modified version of the most widely used test of cognitive flexibility (WCST; Berg, 1948; Heaton et al., 1993), thereby examining the relationship between executive functioning and MC sub-components (MC-M and MC-C). Unlike the basic WCST, this version has only been validated in adults (Koren et al., 2004, 2005, 2006; Arbel et al., 2013; Marucci et al., 2019; Tercero et al., 2021; Badal et al., 2023b,a; Gorora et al., 2024; Jones et al., 2021; Quiles et al., 2014; Bampa et al., 2024; Fan et al., 2025) and teenagers (Koren et al., 2017), with no validation in children with typical and atypical development (e.g., ADHD, SLD, autism). This methodological gap significantly limits the use of the test in developmental research, educational programs, and clinical evaluations. We address these issues by conducting a comprehensive review of studies involving the meta-WCST administration and/or its data analysis. This review has resulted in an evaluation of the task's adaptability to developmental contexts, leading to the proposal of specific directions for this adaptation. Overall, our evaluations have implications for several fields of developmental neuropsychology, thus offering insights for the development of an assessment tool for educational and clinical contexts. Furthermore, they stimulate the improvement of task-related analysis, enhancing its usability for research applications with populations of all ages.

The Meta-WCST (Koren et al., 2004, 2006) is an alternative version of the WCST (the gold-standard neuropsychological test of executive functions; Berg, 1948; Heaton et al., 1993). The basic version requires participants to sort 128 cards with items showing a combination of three characteristics/sorting rules (color, shape, number). After each sort, the operator returns feedback (“Yes”/“No”) without informing the participant of the correct rule. Therefore, participants infer the correct rule based on the feedback. Noteworthy, the sorting rule changes without warning after 10 correct sorts, so the participant must adjust their sorting based on the changed feedback. The Meta-WCST introduces the assessment of metacognitive functions (MC-M and MC-C; Flavell, 1979). Unlike the WCST, participants are asked to rate their confidence level in their sorting choice and decide whether that response should be counted or discarded for the final score. Therefore, a single trial consists of (1) a sorting phase, (2) a metacognitive monitoring phase—confidence assignment, (3) a metacognitive control phase—double choice, and (4) a feedback phase (Figure 1). Participants are presented with only 64 cards due to the additional metacognitive task phases.

This review focuses on studies involving the administration and/or data analysis of the meta-WCST and was supported by a structured process (Figure 2). Therefore, we performed parallel searches on several databases (Google Scholar, PubMed, Web of Science, Scopus) using the keywords “Meta-WCST” and “Metacognitive Wisconsin Card Sorting Test”. We selected these keywords because our review focused only on the metacognitive version of the WCST, regardless of the population to which it was administered. The search was limited to articles published between January 2004 (the first introduction of the meta-WCST) and May 2025. The eligibility criteria included English-language publications with the following features: (1) being published in peer-reviewed journals, (2) the presence of meta-WCST administration/analyses. The online search was exhaustive, and 14 studies met our inclusion criteria (Table 1). Among them, although several studies included different clinical groups, the majority involved participants with schizophrenia (SCHZ; n = 10) and bipolar disorder (BD; n = 5). Other clinical conditions were investigated, each in only one study, such as anorexia nervosa (AN), risk of psychosis (APS), mild cognitive impairment (MCI), and schizoaffective disorder (SCHZAF). Overall, healthy controls were included in three studies, and one study involved only healthy individuals. Finally, 13 studies focused on adult populations (38.37 ± 8.15 years), one study on adolescents (15.85 ± 0.07 years), and none on children.

Koren et al. (2004) investigate insight in SCHZ patients. The study takes into account several cognitive indices (QS, CC, PE, perseverative responses—PR, and trials-to-first category—TFC) and all metacognitive indices (AS, FIM, GM, MG, MR, CS). They detected a correlation between poor insight and metacognitive deficits, highlighting that metacognitive measures improve insight prediction without affecting cognitive measures. Although this is the study that proposes the meta-WCST, it shows methodological limitations such as the lack of a healthy control group and a small, unbalanced experimental group.

Koren et al. (2005) investigate the “competence to consent” in SCHZ patients, taking into account the same indices as the previous study. In particular, they find a correlation between poor competence to consent and metacognitive deficits. Furthermore, they again observe that metacognitive measures improve the prediction of competence to consent without compromising cognitive measures. Unfortunately, the study shares the methodological limitations of the previous one (no healthy control group, small and unbalanced experimental group).

Koren et al. (2006) integrate the two previous studies by investigating both insight and competence to consent in schizophrenic patients. This study takes into account the same indices as the previous one. Furthermore, it shows the same correlations between insight and consensus competence with metacognitive deficits, confirming that cognitive measures are not compromised. This study shares the same limitations as the previous ones.

Arbel et al. (2013) investigate insight in anorexic patients. This study takes into account only three cognitive indices (QS, CC, PR) and excludes the MG from the six metacognitive indices. They detected that insight impairment is more correlated with metacognitive deficits than with cognitive deficits. Furthermore, the relationship between insight, metacognition, and cognition is mediated by IQ in the meta-WCST. Although these interesting findings, the study shows methodological limitations, such as a clinically heterogeneous and gender-unbalanced sample and the inclusion of few cognitive indices.

Quiles et al. (2014) investigate the impact of metacognitive measures on cognitive measures in a healthy population. They administered a modified version of the WCST (Nelson, 1976) and a modified version of the meta-WCST, thus taking into account only CC and TE and no metacognitive index. Unexpectedly, they detected no significant differences between the basic and metacognitive condition. However, the study considered only two cognitive indices (CC, TE) and ignores other indices that may help to investigate cognitive load and distraction (e.g., NPE).

Koren et al. (2017) investigate the relationship between self-disturbance (SD), cognition, and metacognition in help-seeking adolescents (15.85 ± 0.07 years) with and without attenuated psychotic symptoms. They take in consideration only one cognitive index (CR) and five metacognitive indices (AS, FIM, GM, MR, CS). Unexpectedly, they find a positive correlation between metacognition and SD that is not influenced by APS. However, they hypothesize that the hyper-reflectivity of SD individuals may explain the results. This study lacks a healthy control group and an adequate set of cognitive indices.

Marucci et al. (2019) investigate the relationship between subjective (SME) and objective (OME) metacognitive evaluations in SCHZ patients. Therefore, they administer the meta-WCST to collect OME measures (PE, AS, FIM, GM, MG, MR, CS) and the Subjective Scale to Investigate Cognition in Schizophrenia (SSTICS; Stip et al., 2003) to collect SME measures. Unexpectedly, they detected no correlation between the SSTICS and meta-WCST indices. However, they found a strong negative correlation between PE and the metacognitive indices of the meta-WCST. In addition to common sample limitations (e.g., no healthy control group), the inclusion of only PE severely limits this study.

Tercero et al. 2021) investigate Introspective Accuracy (IA) and Introspective Bias (IB) in SCHZ patients and BD patients. They administer a modified version of the meta-WCST that shows different requests and scores. Therefore, they use only the CR index of the meta-WCST and elaborate ad-hoc metacognitive indices. They detect that both groups show impaired IA and positive IB; therefore, they overestimate their performance. However, BD patients perform significantly better than SCHZ patients, probably also because the former show higher feedback sensitivity. On the other hand, the metacognitive indices of SCHZ patients were better correlated with each other than with cognitive performance. Similar to other studies, this one has sample limitations (no healthy control group, unbalanced and heterogeneous experimental groups) and limited data analysis (they only calculated CR as a cognitive index). Finally, they significantly modified the meta-WCST, thus preventing a clear comparison with the original studies.

Jones et al. 2021) investigate the relationship between sadness and IB in SCHZ and BD patients, taking in consideration only CR of the meta-WCST. They found that meta-WCST performance was significantly worse in the never-sad and occasionally-sad SCHZ patients compared to the BD group. This study shares the methodological limitations of others (no control group, unbalanced experimental group, few cognitive and metacognitive indices, a modified meta-WCST that prevents clear comparison with the original studies).

Badal et al. (2023b) investigate the IA in SCHZ and BD patients. They adopt a modified version of the meta-WCST, thus computing the CR and ad-hoc metacognitive indices with decomposition analyses. They detect that BD patients modulate their confidence rate based on feedback, while SCHZ patients do so very poorly, and this deficit is distinct from their perseverative behavior. This study shows the same limitations as many others (no healthy control group, clinically heterogeneous and unbalanced samples, and modified meta-WCST poorly comparable with the original studies).

Badal et al. (2023a) investigate the temporal relationships between IA, accuracy judgments, confidence, and feedback processing in SCHZ patients, BD patients, and healthy controls (HC). They adopt a modified version of the meta-WCST, thus calculating CR and ad-hoc metacognitive indices with network analyses. They find a stronger correlation between confidence and feedback processing in BD patients and HC compared to SCHZ patients. Moreover, past confidence ratings show a negative correlation with future performance in SCHZ patients. Although this study uses new methods (network analyses), it has the same limitations as the previous one.

Bampa et al. (2024) investigate the efficacy of a Metacognitive Training Program (MTP) in patients with MCI. Therefore, they include the meta-WCST administration into a follow-up design (start, 3 months, 6 months) with a control group (untrained patients) and an experimental group (trained patients). Notably, they take into consideration several cognitive indices (CR, TE, PE, NPE, PR, CC, TFC, FMS), a few metacognitive indices (AS, GM), and ad-hoc metacognitive indices. After 3 months, the experimental group showed fewer PR than the control group, while after 6 months, the differential improvement also involved CR and PE. The same trend was observed in metacognitive measures. This is the only study that includes the meta-WCST in a follow-up design, although it has a small sample and a single-blind design.

Gorora et al. (2024) applied new analyses to data collected by (Tercero et al. 2021), thereby measuring how trial accuracy and accuracy judgments predict the trial correctness and accuracy judgment in SCHZ and BD patients. They exploit a time-series analysis including several cognitive indices (CR, CC, TFC), standard metacognitive indices (AS), and ad-hoc metacognitive indices. They detected that BD patients show faster improvements in trial accuracy and correct accuracy judgments. In addition, these accuracy judgments strongly influence the subsequent trial correctness and accuracy judgments. In contrast, trial correctness poorly predicts subsequent trial correctness in SCHZ patients, and there is no correlation between previous and subsequent accuracy judgments. Overall, the study suggests that SCHZ patients do not take into consideration external feedback when making decisions. Although the study applies innovative analyses, it inherits both the dataset and the methodological limitations (no healthy control group, clinically heterogeneous samples, few cognitive indices, a modified version of meta-wcst, not easy to compare with the original one).

Fan et al. 2025) investigate the efficacy of brain stimulation (transcranial direct current stimulation; tDCS) applied to a frontal structure (right rostrolateral prefrontal cortex; rIPFC) in order to enhance IA in SCHZ and SCHZAF patients. They only considered the QS index and confidence levels. They observed a performance improvement over time in the experimental group (target stimulation) compared to the control group (no target stimulation). Furthermore, SCHZAF patients performed better than SCHZ patients. Although this is the first study to combine brain stimulation with meta-WCST administration, it has limitations similar to other studies, such as clinically heterogeneous samples, limited cognitive and metacognitive indices, and a version of the meta-WCST that is not fully comparable to the original.

We began addressing literature limitations through theoretical, methodological, and experimental approaches. For example, we recently proposed a theoretical framework and a related computational model of the meta-WCST, thereby validating it with healthy and clinical populations (Granato et al., 2025b). In our study, we introduced a modeling approach called the ‘neuropsychological digital-twin method', which supports the creation of digital twins for neuropsychological profiling and intervention predictions. On the other hand, we are developing automated neuropsychological pipelines (task administration, scoring, modeling), whose theoretical-computational proposals are validated with several human populations in healthy, neurodivergent, and clinical conditions (Granato et al., 2020; Granato and Baldassarre, 2021; Granato et al., 2022; Granato and Baldassarre, 2024; Granato et al., 2025a,b). Notably, these pipelines contain trial-by-trial analyses and are also being integrated with EBRAINS (Granato, 2025), a widely distributed platform that fosters integration within the European research ecosystem (Schirner et al., 2022; Appukuttan et al., 2023). Finally, we are developing and validating a “developmental meta-WCST” to conduct experimental investigations with children showing typical and atypical development. We aim to administer this task to four groups of children (control, ADHD, SLD, and mixed profiles) to generate normative and clinical data. Specifically, we will extract several new indices (e.g., reaction times for each trial) to identify suitable markers that may distinguish clinical from typical developmental conditions. These data should provide experimental validation for the task and serve as an effective tool for researchers and clinicians.