Background
To establish and maintain everyday interpersonal relationships, individuals must accurately represent their own and others’ mental states, encompassing intentions and beliefs. These intricate processes are collectively referred to as mentalizing or theory of mind (ToM), enabling individuals to comprehend and anticipate the behaviour of others.
However, mentalizing is a multifaceted ability comprising various subcomponents, such as detecting intentions and reasoning about mental states. Furthermore, it relies on other cognitive processes, including emotion recognition and social knowledge. Recent theories emphasize that mentalizing encompasses both declarative and procedural processes (Duclos, Desgranges, Eustache, & Laisney, Reference Duclos, Desgranges, Eustache and Laisney2018). Consequently, mentalizing is hypothesized to possess both explicit and implicit aspects (Apperly & Butterfill, Reference Apperly and Butterfill2009; Butterfill & Apperly, Reference Butterfill and Apperly2013). Implicit mentalizing is believed to be present during the early, preverbal period, suggesting it is a fundamentally biologically rooted skill (Heyes & Frith, Reference Heyes and Frith2014). It operates autonomously, reflexively, fast, and unconsciously, independent of verbal abilities. In contrast, the explicit dimension is slower but reflective and conscious, primarily relying on verbal skills. Explicit mentalizing emerges as a result of cultural learning processes (Heyes & Frith, Reference Heyes and Frith2014).
The term ‘spontaneous’ is frequently employed as a synonym for ‘implicit’ in mentalizing research (Csulak & Herold, Reference Csulak and Herold2021). However, Senju (Reference Senju2013) posits that spontaneous mentalizing differs from implicit mentalizing in several ways. Unlike implicit mentalizing, spontaneous mentalizing is not as obligatory as automatic processing. It is not necessarily unconscious and can interfere with competing tasks. Additionally, spontaneous mentalizing is conceptualized as an immediate, real-time reasoning ability that requires a rapid decoding of biological motion and action. It also serves as a prerequisite for explicit and verbal ToM activity (Koelkebeck et al., Reference Koelkebeck, Hirao, Miyata, Kawada, Saze, Dannlowski and Murai2013). Consequently, spontaneous mentalizing reflects the preparedness of the mentalizing capacity for recruitment. Although implicit, spontaneous, and explicit mentalization are complex and interconnected in nature, it is crucial to note that the measurement paradigms differ significantly in all three cases. Measures of explicit mentalizing require conscious recognition and description of another’s mental state. Implicit mentalizing tasks use indirect approaches, such as nonprompted behaviors. Indirect performance indicators include reaction time without verbal responses. Spontaneous mentalizing is assessed through tasks that do not explicitly instruct interpreting mental states, but it is measured by the spontaneous use of mental state terms.
Several meta-analyses have confirmed the presence of mentalizing deficits in schizophrenia (Bora & Pantelis, Reference Bora and Pantelis2013, Reference Bora and Pantelis2016; Bora, Yucel, & Pantelis, Reference Bora, Yucel and Pantelis2009; Sprong et al., Reference Sprong, Schothorst, Vos, Hox and Van Engeland2007). This impairment is observed both during the acute phase (Bora et al., Reference Bora, Yucel and Pantelis2009; Martin et al., Reference Martin, Robinson, Dzafic, Reutens and Mowry2014) and during remission (Fekete et al., Reference Fekete, Pótó, Varga, Csulak, Zsélyi, Tényi and Herold2020; Herold, Tényi, Lénárd, & Trixler, Reference Herold, Tényi, Lénárd and Trixler2002). Studies on the correlation between symptomatology and ToM show heterogeneity, with no clear link between positive, negative, or disorganization/cognitive symptom dimensions. A recent meta-analysis found a small to moderate association between cognitive/disorganization symptoms and negative symptoms but a weaker association with positive symptoms (Thibaudeau et al., Reference Thibaudeau, Rae, Raucher-Chéné, Bougeard and Lepage2023). ToM is generally linked to neurocognitive function, with moderate correlations between ToM and executive functions, memory, attention, and other cognitive domains (Thibaudeau et al., Reference Thibaudeau, Achim, Parent, Turcotte and Cellard2020). However, ToM often emerges independently of these functions, suggesting a distinct social-cognitive impairment (Parola et al., Reference Parola, Salvini, Gabbatore, Colle, Berardinelli and Bosco2020). Furthermore, it can be detected in the asymptomatic first-degree relatives of patients (Bora & Pantelis, Reference Bora and Pantelis2013; Herold et al., Reference Herold, Varga, Hajnal, Hamvas, Berecz, Tóth and Tényi2018; Janssen, Krabbendam, Jolles, & Van Os, Reference Janssen, Krabbendam, Jolles and Van Os2003). Individuals clinically high-risk for psychosis and those with first-episode psychosis also exhibit lower mentalizing performance (Bora & Pantelis, Reference Bora and Pantelis2013). Longitudinal studies have demonstrated a deterioration in social functionality in schizophrenia (Velthorst et al., Reference Velthorst, Fett, Reichenberg, Perlman, van Os, Bromet and Kotov2017). Mentalizing emerged as a robust predictor of functional outcomes, demonstrating the strongest associations with productive activities, such as vocational and educational pursuits, and performance-based functional measures. (Fett et al., Reference Fett, Viechtbauer, Penn, van Os and Krabbendam2011; Thibaudeau, Cellard, Turcotte, & Achim, Reference Thibaudeau, Cellard, Turcotte and Achim2021).
The majority of studies examined the explicit abilities of individuals with schizophrenia, and most of these studies have revealed a marked decrease in explicit mentalizing functions (Csulak et al., Reference Csulak, Hajnal, Kiss, Dembrovszky, Varju-Solymar, Sipos and Herold2022). However, increasingly more studies are focusing on the involvement of implicit and spontaneous mentalizing in schizophrenia. In a previous article, we summarized the results of implicit mentalizing research (Csulak et al., Reference Csulak, Hajnal, Kiss, Dembrovszky, Varju-Solymar, Sipos and Herold2022). We demonstrated in a meta-analysis that implicit mentalizing is also impaired in schizophrenia but not to the same extent as explicit abilities. Furthermore, neurocognitive deficits may significantly limit the efficiency of implicit skills. In our meta-analysis, we focused on those studies that measure implicit mentalizing with automatic behavioural signs without verbal responses.
The primary objective of this meta-analysis is to investigate the nature of spontaneous mentalizing performance in individuals diagnosed with schizophrenia. Some researchers argue that explicit tasks may not accurately reflect everyday social interactions because they are inherently problem-solving in nature (Klin, Reference Klin2000). Furthermore, task performance can also be influenced by neurocognitive abilities (Thibaudeau et al., Reference Thibaudeau, Achim, Parent, Turcotte and Cellard2020). Conversely, paradigms employing indirect instructions have been demonstrated to effectively identify impairment in patients, even when executive function is not compromised (Langdon, Connors, & Connaughton, Reference Langdon, Connors and Connaughton2017). The tasks designed to assess spontaneous mentalization necessitate the unprovoked activation of mentalizing abilities. Several paradigms have been developed to measure mentalizing abilities indirectly through indirect instructions (see Appendix for detailed description). Perhaps the most widely used test is the animated triangle task (ATT) (Abell, Happe, & Frith, Reference Abell, Happe and Frith2000) It contains animations of two triangles performing three types of movements: random (movements without purpose), goal-directed (GD), and scripts involving mentalizing interactions (i.e., coaxing, mocking). Random and GD animations serve as control tasks. The performances in each animation are scored based on intentionality (the extent to which mental state attribution is expressed in describing the interactions) and appropriateness (the degree to which the description accurately reflects the intended meaning). Intentionality performance is particularly well-suited to assess mentalizing abilities (Bliksted, Ubukata, & Koelkebeck, Reference Bliksted, Ubukata and Koelkebeck2016). An earlier meta-analysis of studies employing ATT revealed significantly lower intentionality and appropriateness scores, accompanied by substantial effect sizes, in individuals with schizophrenia (Bliksted et al., Reference Bliksted, Ubukata and Koelkebeck2016). In addition to ATT, other paradigms have also been developed. The modified versions of Heider and Simmel’s (Heider & Simmel, Reference Heider and Simmel1944) animation are also employed in schizophrenia research (Bell, Fiszdon, Greig, & Wexler, Reference Bell, Fiszdon, Greig and Wexler2010; Langdon, Connors et al., Reference Langdon, Connors and Connaughton2017). The Social Attribution Task – Multiple Choice version (SAT-MC) is based on this animation and was designed to reduce the reliance on verbal function with multiple-choice questions after the Heider and Simmel animation (Bell et al., Reference Bell, Fiszdon, Greig and Wexler2010). The joke appreciation task utilizes cartoons, where the comprehension of a joke hinges on spontaneously recognizing false beliefs (Langdon, Flynn, Connaughton, & Brüne, Reference Langdon, Flynn, Connaughton and Brüne2017). The majority of studies conducted with individuals diagnosed with schizophrenia have indicated that patients employ less appropriate social terms. Consequently, it is plausible that they exhibit reduced receptivity to various cues that trigger spontaneous mentalizing (Langdon, Connors et al., Reference Langdon, Connors and Connaughton2017).
In addition to our meta-analysis on implicit mentalizing, we consider it imperative to summarize the findings on spontaneous mentalizing as well. The primary objective of the present study was to assess the extent of spontaneous mentalizing performance in schizophrenia through a meta-analysis. We postulated that individuals with schizophrenia would exhibit impairments in tasks evaluating spontaneous mentalizing but not in control conditions assessing GD or random interactions. To achieve this, we included case–control studies employing indirect instructions that measure spontaneous verbal descriptions of social interactions.
Methods
This meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Page et al., Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann, Mulrow and Alonso-Fernández2021). The review protocol was registered on PROSPERO (CRD42022318909), and no protocol deviations were observed.
Search strategy
A comprehensive systematic search was conducted in four prominent databases: MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials (CENTRAL), and Web of Science. The search date was March 8, 2022. The search query employed the following keywords: ((implicit) OR (spontaneous)) AND ((theory of mind) OR (mentalizing) OR (mentalization)) AND (schizophrenia). The search encompassed all fields and textual content within each database. No restrictions or filters were applied to the search.
Selection and eligibility criteria
The search results were synthesized using reference manager software (EndNoteX9; Clarivate Analytics, Philadelphia, Pennsylvania). After both automatic and manual duplicate removal, a screening process was conducted based on title, abstract, and full text. Subsequently, the references and citations of the full-text screening records were reviewed. The selection process was conducted by two independent researchers (AH and TC). Disagreements were resolved by an independent third investigator (RH). Reference lists and publication citations (Google Scholar search engine) of the included studies were screened to identify additional studies.
We included studies examining spontaneous mentalization in patients with schizophrenia (and patients with schizoaffective and schizophreniform disorders). Studies from which we could not extract data of sufficient quality were excluded. Articles concerning patients with any other major psychiatric disorders (e.g., bipolar affective disorder) have been excluded.
Data extraction
From each of the eligible studies, the following data was extracted: first author, publication year, study design, country, number of centers, studied population, gender distribution, age distribution, number of patients, accuracy, intentionality (during random, GD, and mentalizing tasks separately), and rate of mentalizing terms during mentalizing tasks. Data were extracted by two independent review authors. Disagreements were resolved by an independent third investigator.
Risk of bias assessment
The ‘Quality In Prognosis Studies’ (QUIPS) methodology was employed, as per the recommendations of The Cochrane Prognosis Methods Group (PMG). Two researchers were involved in the assessment process. Any disagreements were resolved through a third review author (Supplementary Figures 1–7).
Statistical analysis
For continuous outcomes, weighted mean differences (WMDs) or when different scoring scales or paradigms had been employed to measure the outcomes, and standardized mean differences (SMD) were computed with 95% confidence intervals.
In terms of appropriateness, SMD was calculated for both mentalizing and control tasks, as the studies employed distinct tasks (ATT, SAT-MC, joke appreciation). Additionally, SMD was computed for GD tasks, given the utilization of varying scoring scales. Given that several studies utilized the same paradigm (ATT) and scoring scale, we also calculated WMD within appropriateness. ATT is the sole paradigm for which intentionality was additionally measured, necessitating the calculation of WMD for this outcome.
We applied the Hedges method to estimate the SMD since a few studies had sample sizes less than 20. Random effect models were utilized to pool the effect sizes. A p-value less than 0.05 was deemed statistically significant. Additionally, we applied the restricted maximum likelihood estimator and the Knapp–Hartung adjustments (Knapp & Hartung, Reference Knapp and Hartung2003) to calculate the heterogeneity variance τ 2 and the confidence interval of the overall effect, respectively. The results of the meta-analyses were presented in forest plots.
Heterogeneity was assessed using the I2 statistics. Following the guidelines provided in the Cochrane Handbook (Chandler et al., Reference Chandler, Cumpston, Li, Page and Welch2019), I2 values were categorized as follows: 0%–40%, 30%–60%, 50%–90%, and 75%–100%. These categories correspond to the following interpretations: ‘Might not be important,’ ‘Moderate,’ ‘Substantial,’ and ‘Considerable,’ respectively. Heterogeneity was considered significant when the p-value was less than 0.1. For outcomes involving at least 10 studies, we conducted a funnel plot analysis and performed Egger’s test to assess potential publication bias. All analyses were conducted using the R environment (R Core Team, 2021; R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. R version 4.1.2, 2021-11-01).
Furthermore, we investigated the association between intentionality and appropriateness score during mentalizing tasks and mean age and mean years of education for both the schizophrenia and control groups. This was achieved by fitting a linear meta-regression model. (Supplementary Figures 14–21).
Results
Systematic search and selection
The systematic search yielded 648 records. After the automatic and manual removal of duplicates, 576 records remained. The flowchart of the publication selection process is presented in Figure 1. After checking the records and conducting citation searching, 15 studies were included in the quantitative synthesis (Bell et al., Reference Bell, Fiszdon, Greig and Wexler2010; Beck et al., Reference Beck, Simonsen, Wang, Yang, Zhou and Bliksted2020; Bliksted et al., Reference Bliksted, Fagerlund, Weed, Frith and Videbech2014, Reference Bliksted, Frith, Videbech, Fagerlund, Emborg, Simonsen and Campbell-Meiklejohn2019; Das et al., 2012; Horan et al., Reference Horan, Nuechterlein, Wynn, Lee, Castelli and Green2009; Koelkebeck et al., Reference Koelkebeck, Hirao, Miyata, Kawada, Saze, Dannlowski and Murai2013; Langdon & Ward, Reference Langdon and Ward2009; Langdon, Flynn et al., Reference Langdon, Flynn, Connaughton and Brüne2017; Langdon, Boulton, Connaughton, & Gao, Reference Langdon, Boulton, Connaughton and Gao2020; Lee et al., Reference Lee, Corbera, Poltorak, Park, Assaf, Bell and Choi2018; Lugnegård, Hallerbäck, Hjärthag, & Gillberg, Reference Lugnegård, Hallerbäck, Hjärthag and Gillberg2013; Pedersen et al., Reference Pedersen, Koelkebeck, Brandt, Wee, Kueppers, Kugel and Ohrmann2012; Roux, Smith, Passerieux, & Ramus, Reference Roux, Smith, Passerieux and Ramus2016; Veddum, Pedersen, Landert, & Bliksted, Reference Veddum, Pedersen, Landert and Bliksted2019). The meta-analysis included 526 patients and 536 controls in its analysis. The characteristics of the included studies are presented in Table 1.
PRISMA flowchart for the study selection process (Page et al., Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann, Mulrow and Alonso-Fernández2021).
The characteristics of the included studies
Intentionality
For the purpose of investigating intentionality during ToM animations, data from ten studies (all employing the ATT) were analysed (Supplementary Figure 8). These studies involved a total of 269 patients and 282 controls. A statistically significant difference was observed between the two groups, with a p-value less than 0.001. The mean difference (MD) was −0.99, and the confidence interval (CI) ranged from −1.39 to −0.59. The between-study heterogeneity, expressed as I2 value, was 88.2%, indicating substantial heterogeneity.
Given the divergent findings of the Das et al. study compared to the others, we conducted a calculation by excluding this data (Figure 2). Consequently, the distinction between the two groups remains statistically significant (MD: −0.81; CI (−1.00; −0.61), p < 0.001), albeit with a substantial reduction in heterogeneity (I 2 = 25.28%).
Intentionality Score during ToM tasks (Omitted by Das et al., Reference Das, Lagopoulos, Coulston, Henderson and Malhi2012).
A total of eight studies were selected for analyses of intentionality during random tasks, involving 231 patients and 246 controls (Supplementary Figure 9). The mean difference (pooled effect size) between the two groups was 0.034, with a p-value of 0.5395, and the confidence interval ranged from −0.09 to 0.16, which indicates that there is no statistically significant difference in the impact of the two groups.
For intentionality during GD animations, data from four studies (all employing the ATT paradigm) were utilized (Figure 3). These studies included 136 patients and 152 controls. A significant difference in reaction time was observed between the two groups (MD: −0.31; 95% CI (−0.53; −0.08); p = 0.022). Furthermore, the between-study heterogeneity, expressed as I 2 value, was 0, indicating that 0% of the variance in observed effects can be attributed to true effects rather than sampling error.
Intentionality in goal-directed animations.
Appropriateness
A total of 14 studies (utilizing various methodologies) were selected for appropriateness during ToM tasks, involving a combined sample of 526 patients and 536 controls (Figure 4). A significant difference (SMD = −0.97; 95% CI (−1.16; −0.79); p < 0.0001) was observed between patients with schizophrenia and controls with moderate statistical heterogeneity (I2 = 37.4%). Schizophrenic patients exhibited a substantially weaker performance compared to controls, with an effect size of −0.97, which is regarded as a large effect.
Appropriateness in mentalizing tasks.
We conducted an analysis of the appropriateness of tasks during the mentalizing tasks using the ATT study paradigm (Supplementary Figure 10). We included six studies that examined 186 patients and 190 controls. There was a significant difference between the two groups (MD: -0.63; 95% CI (−0.90; −0.35); p = 0.0021). The heterogeneity was considerable (I2 = 81.35%). The high heterogeneity may be attributed to the varying lengths of illness among the subjects. Additionally, some studies included patients with schizoaffective disorder, and most of the examined patients were also taking medication, which may have influenced the results.
A total of nine studies (utilizing diverse spontaneous mentalizing tasks) were selected for analyses of appropriateness during random tasks involving a combined sample of 274 patients and 293 controls (Supplementary Figure 11). No statistically significant difference was observed between the two groups (SMD: -0.09; 95% CI (−0.47; 0.30); p = 0.624). The I 2 value was 68.11%, indicating substantial heterogeneity.
To assess the appropriateness during random tasks in studies using ATT, data from five studies were analysed (Supplementary Figure 12). These studies involved 163 patients and 168 controls. No significant difference was found between patients with schizophrenia and controls (MD = −0.10; 95% CI (−0.40; −0.20); p = 0.398). The statistical heterogeneity was substantial (I 2 = 76.84%). The results are presented in Figure 9 in the supplement.
Given the outlier status of the results presented by Veddum et al. (Reference Veddum, Pedersen, Landert and Bliksted2019)), we conducted a leave-one-out analysis by excluding this observation (Supplementary Figure 13). Despite this, no significant difference was detected between the groups (MD = −0.18; 95% CI (−0.39; 0.02); p = 0.0634). However, this approach resulted in a substantial reduction in heterogeneity (I 2 = 20.07%).
To ensure appropriateness during GD animations, data from four studies (which employed the same paradigm but utilized distinct scoring scales) were analysed. These studies involved a total of 136 patients and 152 controls. A significant difference (SMD = −0.55; 95% CI (−0.97; −0.13); p = 0.025) was observed between patients with schizophrenia and controls, with negligible statistical heterogeneity (I 2 = 14.90%). Schizophrenic patients exhibited a weaker performance compared to controls, with an effect size of −0.55, which is classified as a medium effect. These findings are presented in Figure 5.
Appropriateness during goal-directed animations.
Meta-regression
Meta-regression analysis was conducted for both schizophrenia and control groups, with age (in years) and time spent in education as covariates. In the control group, a significant negative correlation (p = 0.0492) was observed between age and appropriateness in mentalizing tasks. Similarly, a trend towards a negative correlation was observed in the schizophrenia group, although it did not reach statistical significance. No other significant effects were detected in either group. However, tendencies were observed in both groups, suggesting a potential improvement in intention during ToM tasks with increased time spent in education. Additionally, both groups exhibited lower, albeit insignificantly lower, intentionality scores during ToM tasks as a function of age.
Assessment of the quality of evidence (GRADE Approach)
The overall assessment of the quality of evidence is presented in Table 2. The overall quality of evidence for the results was assessed as low or very low. The outcomes were downgraded due to the inconsistency and imprecision of the results. In most outcomes, the heterogeneity of the results (as determined by the I2 statistic) resulted in considerable inconsistency. A substantial imprecision was caused by the relatively small sample sizes and wider confidence intervals, which further contributed to the downgrades. Regarding the risk of bias, no serious limitations were identified. In the studies included, no problems compromising the directness of the evidence were observed.
Assessment of the quality of evidence
Risk of bias assessment
The overall risk of bias was low to high in the included studies. Detailed results of the quality assessment are available in the Supplementary Materials (Supplementary Figures 1–7).
Conclusions
In our study, we reviewed the research results pertaining to spontaneous mentalization in schizophrenia. In accordance with our hypothesis, we observed impaired spontaneous mentalization skills in individuals with schizophrenia. Conversely, performance in random movement tasks did not differ significantly from that of normal control groups. However, contrary to our hypothesis, we also identified significant differences in the interpretation of GD interactions between subjects with schizophrenia and normal controls.
Patients with schizophrenia exhibited significantly lower scores in both intentionality and appropriateness in interpreting ToM interactions. Our meta-analysis revealed impaired intentionality with a large effect size, although the heterogeneity was substantial. By excluding one outlying study, the heterogeneity decreased considerably, but the difference between the groups remained significant. Furthermore, significant differences were also observed between the groups in appropriateness with a large effect size and low heterogeneity. These findings suggest a substantial deficit in spontaneous ToM in schizophrenia, particularly evident in patients who performed equally to healthy controls in the control condition with random animations. These results imply that patients with schizophrenia are less likely to appropriately capture situations requiring spontaneous mentalizing activity and exhibit a lower degree of mental state attribution in describing the interactions.
Effect sizes (intentionality: −0.99, appropriateness: −0.97) in our meta-analysis indicate that the severity of impairment in spontaneous mentalizing is comparable to the level of impairment reported in explicit mentalizing meta-analyses (e.g., Bora et al., Reference Bora, Yucel and Pantelis2009; Bora & Pantelis, Reference Bora and Pantelis2013). Furthermore, we observed significantly longer reaction times with a comparable effect size (0.89) in our previous meta-analysis on implicit ToM (Csulak et al., Reference Csulak, Hajnal, Kiss, Dembrovszky, Varju-Solymar, Sipos and Herold2022). Reaction time serves as a proxy for the interference effect of intention detection, making it a suitable measure to assess implicit ToM in indirect task paradigms (Gardner & Buchanan, Reference Gardner and Buchanan2023). Overall, the meta-analyses conducted on various aspects of ToM skills in schizophrenia reveal a pervasive deficit in mentalizing abilities.
Additionally, meta-regression revealed a negative correlation between age and mentalizing performance in control subjects, which aligns with the findings suggesting that older adults tend to experience a decline in mentalizing abilities (Cassidy, Hughes, & Krendl, Reference Cassidy, Hughes and Krendl2021; Greenberg, Warrier, Abu-Akel, & Baron-Cohen, Reference Greenberg, Warrier, Abu-Akel and Baron-Cohen2023). A similar trend was observed in patients, although the results were not statistically significant.
Patients also performed significantly worse in GD conditions, although the effect sizes were lower in detecting and describing GD animations compared to the mentalizing condition. However, due to limited data, we could not analyze the relationship between mentalizing and goal-oriented animations. Only four studies were included from the available seven due to data limitations, reducing the generalizability of the findings. The lower performance in the GD animation condition suggests a deeper deficit, potentially indicative of general cognitive dysfunction. Pedersen et al. (Reference Pedersen, Koelkebeck, Brandt, Wee, Kueppers, Kugel and Ohrmann2012)) emphasized the significance of reaction time in understanding social cognition. In their functional imaging study, similar brain areas were activated in patients and control individuals during the observation of triangles, but the activation occurred later in the patients. Given that the processing of ATT stimuli entails an implicit mentalizing component (Bliksted et al., Reference Bliksted, Ubukata and Koelkebeck2016), we can also posit that cognitive impairments may hinder the performance in ATT through the implicit processing. Furthermore, data on implicit mentalizing suggest that impaired neurocognition (e.g., reaction time) may play a limiting role in implicit processing in schizophrenia (Csulak et al., Reference Csulak, Hajnal, Kiss, Dembrovszky, Varju-Solymar, Sipos and Herold2022). Lower performance in GD and ToM animations may indicate impaired visual processing of moving objects. The processing of biological motion by living organisms is crucial to social cognition, including mentalizing (Kim et al., Reference Kim, Norton, McBain, Ongur and Chen2013; Pavlova, Reference Pavlova2012). A meta-analysis found that patients with schizophrenia perform poorly on basic biological motion tasks and emotion recognition tasks (Okruszek & Pilecka, Reference Okruszek and Pilecka2017). However, the equal performance in random tasks in our study contradicts this assumption. It seems a more plausible explanation that GD and mentalizing tasks require more cognitive effort than random tasks. The lower performance in GD may also reflect an inappropriate mentalizing activity. A study conducted by Russell, Reynaud, Herba, Morris, and Corcoran (Reference Russell, Reynaud, Herba, Morris and Corcoran2006)) revealed that patients provided more mentalizing responses in the GD condition compared to controls. This raises the question of whether the GD condition accurately represents a GD interaction without the mentalizing aspect. GD task can also be interpreted as a fundamental mentalizing task (Roux et al., Reference Roux, Smith, Passerieux and Ramus2016). These uncertainties led to a recent meta-analysis of autism spectrum research underscoring the low-level construct validity of ATT (Wilson, Reference Wilson2021).
The lower performance in schizophrenia may indicate patients’ restricted ability to assign narratives to moving stimuli. Linguistic research suggests that narrative language production problems in schizophrenia are linked to semantic deficits and cognitive impairments, resulting in less coherent and clear spoken life narratives (Marini & Perlini, Reference Marini and Perlini2013, Lundin, Cowan, Singh, & Moe, Reference Lundin, Cowan, Singh and Moe2023). They even exhibit impaired understanding of the narrative content of literary fictions and use fewer mental state terms spontaneously to describe social events (Fekete et al., Reference Fekete, Pótó, Varga, Csulak, Zsélyi, Tényi and Herold2020). Narrative processing requires synchronized cooperation of mentalizing, language, and neurocognitive skills (Delgado et al., Reference Delgado, Rodríguez-Ortiz, Lopez-Soto, Garcia-Lopez and Salguero-Lamillar2024). Longitudinal studies reveal the interplay between these abilities (Ebert, Reference Ebert2020; Shahaeian, Haynes, & Frick, Reference Shahaeian, Haynes and Frick2023). Spontaneous mentalizing performance also correlated with reasoning, verbal IQ, and verbal memory in schizophrenia (Koelkebeck et al., Reference Koelkebeck, Pedersen, Suslow, Kueppers, Arolt and Ohrmann2010). Roux et al. (Reference Roux, Smith, Passerieux and Ramus2016)) and Koelkebeck et al. (Reference Koelkebeck, Hirao, Miyata, Kawada, Saze, Dannlowski and Murai2013)) employed a scoring scale to measure utterances during the ATT task. Patients’ descriptions of tasks were comparable to controls’, but they used fewer intentional and action-describing terms, suggesting repetitions and descriptions of physical reality rather than actions. Majority of the research indicates that patients with schizophrenia also tend to provide shorter responses and use mental state terms less frequently (Andreasen et al. Reference Andreasen, Calage and O’Leary2008; Bliksted et al., Reference Bliksted, Ubukata and Koelkebeck2016; Langdon, Connors et al., Reference Langdon, Connors and Connaughton2017). Furthermore, some studies have suggested a weak correlation between the WAIS Vocabulary test and schizophrenia (Bell et al., Reference Bell, Fiszdon, Greig and Wexler2010; Lugnegård et al., Reference Lugnegård, Hallerbäck, Hjärthag and Gillberg2013). Consequently, we propose that vocabulary, including mental state terms, may also be compromised in schizophrenia. Recent longitudinal studies emphasize the role of vocabulary learning in mentalizing development (Delgado, Reference Delgado, Rodríguez-Ortiz, Lopez-Soto, Garcia-Lopez and Salguero-Lamillar2024; Ebert, Reference Ebert2020). Regrettably, we were unable to analyze the results on verbal performance during the spontaneous mentalizing tasks due to data incompatibility difficulties. In this regard, it would have been beneficial if sufficient data had been available to assess the impact of general intelligence. Some studies lacked IQ measurement (Bell et al., Reference Bell, Fiszdon, Greig and Wexler2010; Das et al., Reference Das, Lagopoulos, Coulston, Henderson and Malhi2012; Koelkebeck et al., Reference Koelkebeck, Hirao, Miyata, Kawada, Saze, Dannlowski and Murai2013), conducted only subtests (Horan et al., Reference Horan, Nuechterlein, Wynn, Lee, Castelli and Green2009; Lugnegård et al., Reference Lugnegård, Hallerbäck, Hjärthag and Gillberg2013; Pedersen et al., Reference Pedersen, Koelkebeck, Brandt, Wee, Kueppers, Kugel and Ohrmann2012; Veddum et al., Reference Veddum, Pedersen, Landert and Bliksted2019; Beck et al., Reference Beck, Simonsen, Wang, Yang, Zhou and Bliksted2020), or measured premorbid estimated intelligence levels (Langdon et al., Reference Langdon, Boulton, Connaughton and Gao2020; Langdon & Ward, Reference Langdon and Ward2009). Studies that measured IQ yielded conflicting results. Some found a correlation with spontaneous mentalization (Bliksted et al., Reference Bliksted, Fagerlund, Weed, Frith and Videbech2014; Lee et al., Reference Lee, Corbera, Poltorak, Park, Assaf, Bell and Choi2018), while others concluded that IQ does not account for diminished performance in schizophrenia (Roux et al., Reference Roux, Smith, Passerieux and Ramus2016).
There was no adequate data for meta-regression in the case of symptomatology either. Some studies found no significant relationship between symptoms and mental performance (Bliksted et al., Reference Bliksted, Fagerlund, Weed, Frith and Videbech2014; Das et al., Reference Das, Lagopoulos, Coulston, Henderson and Malhi2012; Langdon, Reference Langdon, Connors and Connaughton2017; Roux et al., Reference Roux, Smith, Passerieux and Ramus2016; Bell et al., Reference Bell, Fiszdon, Greig and Wexler2010; Lee et al., Reference Lee, Corbera, Poltorak, Park, Assaf, Bell and Choi2018). Others described correlations with negative symptoms (Horan et al., Reference Horan, Nuechterlein, Wynn, Lee, Castelli and Green2009), referential ideation, but not with persecutory ideation (Langdon et al., Reference Langdon, Boulton, Connaughton and Gao2020), or with the higher levels of both positive and negative symptoms (Bliksted et al., Reference Bliksted, Ubukata and Koelkebeck2016). Interestingly, Bliksted et al. (Reference Bliksted, Frith, Videbech, Fagerlund, Emborg, Simonsen and Campbell-Meiklejohn2019) described different neural activation during intentional scenes depending on symptoms. However, not all studies reported direct data on the correlations of symptoms and mentalizing.
In summary, individuals with schizophrenia exhibit challenges in spontaneously activating their mentalizing capacity in social contexts. This highlights the significance of enhancing spontaneous mentalization skills. However, as schizophrenia is characterized by a pervasive mentalization deficit, interventions providing compensatory strategies for explicit mentalization are necessary, but supporting spontaneous mentalization is also crucial. Meta-analytic evidence supports the utility of broad-based social cognition training in enhancing explicit mentalizing abilities in individuals with schizophrenia, though its durability remains uncertain (Nijman, Veling, van der Stouwe, & Pijnenborg, Reference Nijman, Veling, van der Stouwe and Pijnenborg2020). However, limited evidence exists for improving spontaneous mentalization. Attention training (Langdon, Connors et al., Reference Langdon, Connors and Connaughton2017) or imitation–inhibition training that facilitates perspective-taking abilities may be considered as a potential approach (Simonsen et al., Reference Simonsen, Mahnkeke, Fusaroli, Wolf, Roepstorff, Michael and Bliksted2020). Enhancing mental state vocabulary, for instance, through reading literary fiction, could also contribute to improved spontaneous mentalization performance (Fekete et al., Reference Fekete, Pótó, Varga, Hebling, Herold, Albert and Herold2023). Mentalization-based therapy offers a novel approach, but its efficacy in schizophrenia spectrum disorders is still uncertain. Studies suggest it helps patients with schizophrenia understand social causality (Weijers et al., Reference Weijers, Kaam, Selten, Winter and Kate2023) and may benefit recent-onset patients (Weijers et al., Reference Weijers, Kate, Viechtbauer, Rampaart, Eurelings and Selten2021).
Limitations
Our meta-analysis has several limitations, the primary one being the relatively low number of eligible studies. Regrettably, several studies were excluded due to insufficient data for meta-analysis. However, it is important to note that, according to the Cochrane Handbook for Systematic Reviews, a meta-analysis is the statistical combination of results from two or more separate studies. Furthermore, as asserted by Valentine, Pigott, and Rothstein (Reference Valentine, Pigott and Rothstein2010)), at least two studies are necessary for a meta-analysis because it remains the most transparent and valid method of synthesizing research data. Additionally, there was heterogeneity in the application of research paradigms. Among the 15 studies included in the meta-analysis, various approaches were employed to assess ToM: 10 studies utilized the ATT, three articles employed the joke appreciation task, and two utilized the SAT-MC. Moreover, not all aspects of the original task were consistently utilized across different research studies. For instance, not all studies of ATT employed the three types of tasks (random, GD, and mentalizing). The heterogeneity of the research designs employed significantly restricts the generalizability of our findings. This limitation is somewhat mitigated by the fact that the analysis of data employing solely the ATT paradigm yielded comparable results. Heterogeneity was also apparent in the sample selection. In most studies, the average duration of illness was up to 10 years, while others included first-episode patients. Another limitation of our study pertains to meta-regression analysis. Due to data incompatibility (utilization of distinct testing tools), we were unable to examine the association between mentalizing performance during ToM tasks and IQ, nor the presence of symptoms, as data from at least 10 studies are needed to perform meta-regression (Higgins & Green, Reference Higgins and Green2011). Both dimensions are crucial mediators of mentalizing performance. For instance, higher verbal intelligence is associated with enhanced activation of brain regions involved in mentalization (Tantchik et al., Reference Tantchik, Green, Quidé, Erk, Mohnke, Wackerhagen and Walter2023), while lower intelligence is linked to more pronounced deficits in mentalizing tasks (Sahl et al., Reference Sahl, Rognlien, Andreassen, Melle, Ueland and Vaskinn2022).
In conclusion, this meta-analysis shows that schizophrenia patients have significant impairments in spontaneous mentalizing. They have reduced sensitivity to bottom-up signals that trigger mental state attributions and diminished intentionality and appropriateness in describing social interactions (Langdon, Connors et al., Reference Langdon, Connors and Connaughton2017). These impairments disrupt the automatic mobilization of mentalizing skills, which are crucial for effective social functioning. While performance deficits were pronounced in mentalizing tasks, patients showed no significant differences in random animation tasks. However, they had impairments in GD tasks, suggesting a more generalized cognitive dysfunction. The findings also highlight the pervasive nature of mentalizing deficits, encompassing explicit, implicit, and spontaneous dimensions. These findings have implications for developing tailored remediation strategies to enhance social cognition and functional outcomes for individuals with schizophrenia.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725100755.
Funding statement
The study was supported by the FIKP-IV and TINL projects.
Competing interests
The authors declare that they have no conflicts of interest.
Appendix. Paradigms for assessing spontaneous mentalizing
ATT: During the ATT (Abell et al., Reference Abell, Happe and Frith2000), participants observe 34–45 second animations depicting the movements and interactions of a large and a small red triangle. These animations are categorized into three types: random movements (e.g., bouncing), goal-oriented movements (e.g., chasing), and animations that incorporate mentalizations (e.g., surprising, mocking). Participants’ responses are evaluated based on several criteria. These include the extent to which their appreciation of mental states (intentionality) is evident, the accuracy of the animations’ description (appropriateness), and, in some trials, the length of their responses.
Heider and Simmel’s Animation: Langdon, Connors et al. (Reference Langdon, Connors and Connaughton2017) utilized the Heider and Simmel’s animation, which incorporates the social interaction of geometric shapes (two triangles and a circle). This test elicits intentionality and emotional mental status. Inspired by Klin’s (Reference Klin2000) study, participants were asked, ‘What happened there?’ Subsequently, a cue instruction was applied, personifying geometric shapes without explicitly representing them as intentional agents. During the test, patients were evaluated for spontaneous responses, total word count, accuracy of basic and social expressions, and the utilization of various mental state expressions (perception, emotion, goal-driven intentions).
Social Attribution Task – Multiple Choice version (SAT-MC): During the SAT-MC examination (Bell et al., Reference Bell, Fiszdon, Greig and Wexler2010), participants are presented with a 64-second animation (Heider and Simmel’s animation) depicting a social interaction involving two triangles and a circle. This recording is viewed twice. Subsequently, 19 questions are posed, requiring participants to select the correct answer from four possible options.
Joke Appreciation Task (Langdon & Ward, Reference Langdon and Ward2009): Two sets of cartoons are presented, one of which necessitates mentalizing, and the other does not. Participants are then requested to elucidate the joke, and their responses are subsequently evaluated on a three-point scale.
Open access
