Knowing me, knowing you: a scoping review assessing the current field of social cognition in schizophrenia

Abstract Background Impaired social cognition (SC) occurs frequently in schizophrenia (SCZ), yet compared to autism spectrum disorder (ASD), the research on SC in SCZ is sparse and methodologically heterogeneous. To accurately assess between-group SC differences it is further necessary to establish the relationship between nonsocial cognition (NSC) and SC, particularly as this may not be identical across disorders. Purpose The present study aimed to map, index and assess the quality of research published between 2014–2021 investigating SC in SCZ, and to summarize existing limitations and recommendations for future research. Method Following the PRISMA guidelines for scoping reviews (PRISMA-ScR) fifteen (n = 15) case-control studies were identified and included across three electronic databases. Studies additionally utilizing ASD samples were included because of their clinical utility. Results Most studies reported significant SC impairments in SCZ relative to healthy controls (HC) with varying effect sizes. Significant differences between SCZ and ASD were not found in most studies including both samples. Weak-to-moderate correlations between SC and NSC were often found, although often only within patient samples. Across studies, SC tests were inconsistently described as measurements of "social cognition", "mentalization" and, most frequently and varyingly, “theory of mind”. Most studies lacked methodological transparency. Limitations related to sample sizes and test reliability were most frequently mentioned. Conclusions The current research on SC in SCZ is limited by conceptual and methodological uncertainties. Future research should focus on ensuring clear and valid definitions of key terms, evaluating and clarifying SC outcome measures and further untangling the relationship between SC and NSC.


Introduction
Impairments in social functioning have for a long time been considered a defining feature of schizophrenia (SCZ) [1][2][3] yet compared to autism spectrum disorder (ASD) the case-control study output has been relatively scarce.Moreover, we will argue that current knowledge of social cognition (SC), particularly in relation to SCZ, is somewhat limited by conceptual and methodological uncertainties.Consequently, the field in question may be in need of finding common ground regarding direction, methodology and language use.

What is social cognition (SC)?
A key aspect of SC consists of us surmising feelings, thoughts, intentions and beliefs of ourselves and others, enabling us to explain and predict behaviors [4].Although suggested that all cognition is somewhat social [5] several studies indicate that "social" and "nonsocial" cognitive domains only correlate to a small to moderate extent [6], and nonsocial cognitive abilities (NSC) have therefore been proposed as necessary yet not sufficient for SC [6,7].Several attempts at establishing operational definitions of SC have been attempted, oftentimes by either considering stimuli being processed or stimuli being produced [7].The former "input" perspective has enabled neuroimaging experiments using "social" versus "nonsocial" conditions [7], such as observing differences in test subjects' brain activity when presented with either picture of human faces or flowers.However, the philosophical question of how to categorize stimuli remains (e.g. can some stimuli be more social than others, and by what argument?).The "output" perspective has in turn inspired the development of standardized tests measuring the degree of social understanding (e.g.RMET [8,9] and TASIT [10]) often utilized in correlational studies for their practical value in allowing for between-group test comparisons.Nevertheless, social processes oftentimes imply subjectivity, making the comprehension of them intricate to study.

The straggly nomenclature of SC research
The term Theory of Mind (ToM) was coined in 1978 by Premack and Woodruff [11] to illustrate the multi-level reflective characteristics of social processing.Within developmental psychology the term ToM has been frequently applied in research on social competence, particularly in Autism Spectrum Disorder (ASD).The social impairments associated with ASD have been well-demonstrated [12][13][14][15]) and include eye gaze and joint attention disturbances [16], recognizing faces and emotions [17,18], participating in pretend play [19] and comprehending irony and subtle jokes [13,20].It was early suggested that children with ASD lack the ability to "read minds" due to a missing ToM mechanism [12,21], initially indicated by evidence of children with ASD not passing the Sally-Anne test at the anticipated age of 3-4 [12,21].Such false-belief tasks require the child to understand that their belief could differ from that of another person [12].Recently this false-belief operationalization of ToM as the be-all-end-all of ASD research has been criticized, the reason being that difficulties solving first-and second-order false-belief tasks appear neither specific for nor isolated in ASD [20,22].Using ToM as "synonymous" to SC has also been problematized as positing a risk for conceptual confusion and lack of standardization since SC includes a number of abilities unrelated to ToM [23][24][25].
The term mentalization has sometimes been used as an alternative to ToM when describing the "mind reading" or socially reflective aspects of SC [26].The term, originally associated with psychoanalytic theory [27] has been expanded upon by Peter Fonagy and colleagues to describe how individuals are able to "keep mind in mind" [27].
Attempts at identifying subdomains of SC using existing social cognitive outcome measures have also been conducted.Morrison et al. [28] presented evidence in favor of multiple, somewhat separate domains of SC through exploratory factor analysis, however, concluded that the number of domains may not be constant across populations (in this case, between ASD and HC).

SC research in SCZ
Past studies exploring SC in SCZ compared to HC have demonstrated emotion detection impairments for both facial [29][30][31] (for a review see [32]) and bodily cues [33].Some suggest these differences between HC and SCZ to be specific to fear and anger [32,33].Furthermore, SCZ patients tend to overestimate threat and danger [33,34] perhaps the result of impaired facial affect perception [34], or integration and appraisal difficulties [35].SCZ patients also tend to perform worse than HC on SC tasks [3,[36][37][38] unrelated to NSC [39,40].These difficulties have been associated with symptom severity [40], level of paranoia [36] and functioning capacity [41] (for a review see [2]).Moreover, SC impairments might trigger or worsen schizophrenic symptoms, as paranoia has been theorized to stem from erroneous mental state attributions [42], perhaps explaining the externalizing attribution and "hypermentalizing" biases found in some psychotic patients [42][43][44].Difficulties understanding own thoughts have also been suggested to cause delusions [3].

The link between SCZ and ASD
ASD and SCZ have long been intertwined, and were more or less "diagnostically conflated" until the 1980 publication of DSM-III [45].Like autistic children, those with childhood-onset schizophrenia have been seen to demonstrate early social comprehension difficulties [1,46], causing the hypothesis that the two disorders could be associated either through comorbidity or by constituting phenotypic variations within a common developmental disorder [1].Factor analytic studies have found a varying number of SC subdomains possible to derive between SCZ, ASD and HC samples [28,47].Between-study differences in the measurements used could be the source of these variations [47], or could they indicate that the domains associate differently across clinical groups [45].Although both disorders are associated with social impairments, differences between them have been established [48].To account for this, a "hyper" versus "hypo mentalizing" model of SC has been proposed [49,50].If ASD patients tend to mentalize less than appropriate, whereas SCZ patients do the opposite, this could possibly explain the between-group test accuracy similarities [50], as well as the tendency among SCZ patients to overestimate threats and become paranoid [42][43][44].Both hypo-and hypermentalizing could, however, coexist in SCZ [51], perchance depending on whether the task is social or nonsocial [52] suggesting that mentalizing impairments in SCZ could be caused by paradoxical neural responses.Disruptions in the development of SC might further have taken place earlier for those with ASD compared to SCZ [15,36,39,45], leaving SCZ patients with partially intact networks which they fail to functionally utilize (for a review see [53]).Feasibly in line with this, some studies have found between-group differences in ASD and SCZ only appearing for complex tasks [50], as well as impairment pattern differences [15].Since 2014 two studies assessing existing SC research on SCZ compared to ASD could be identified [54,55].Both concluded a similarity in SC ability between SCZ and ASD, resulting in difficulty differentiating between groups using the included outcome measures."Categories" or "domains" of SC were constructed by both [54,55] in order to compare different measures with the same intent.However, as stated by [54], both studies presented high between-study heterogeneity (as indicated by their large Q and I 2 values) and the comparability of the tests might therefore not be meaningful.Furthermore, both these meta-analytic attempts included studies with diverse diagnostic samples, including schizotypal personality disorder and first-episode psychosis.

Current study rationale
As presented above, we posit that there is reason to assume that some of the negative outcomes associated with SCZ (e.g.decreased interpersonal functioning and quality of life) may be attributed to impairments in SC, motivating further research on SC in SCZ.Moreover, SCZ and ASD are as previously mentioned connected by clinical tradition, commonalities in symptom presentation and potential underlying mechanisms.While specific research on SC in SCZ has yielded compelling results, these are difficult to understand in isolation, as others before us have already reckoned by additionally including ASD samples in their studies (e.g.[15,45,49,50,54,55]).However, earlier attempts at compounding such studies using meta-analytical approaches [54,55] have not been able to group the SC tests in ways similar enough for rigorous and meaningful statistical comparisons.We suspect that this could be a result of insufficiently demonstrated construct validity of the measures included and of between-study inconsistencies in the application of key terms and concepts.Because of these issues, we instead opted for a scoping review method, recommended to other systematic approaches when it is unclear whether a more precise question can be addressed [56,57] or when the available research is too broad and heterogeneous for meaningful quantitative comparisons [58].Such a method can aid research forward by pointing out patterns, gaps and weaknesses within the field in question, enabling future researchers to more clearly situate themselves within the field.By including both studies investigating SC in SCZ alone, and those additionally using ASD samples for between-group comparisons, our aim was to highlight the potential limitations of current assessment methods and inform the development of methods that may be better suited for individuals with both conditions.Including ASD in SC research on SCZ could potentially both be conceptually helpful within the field of SC research and of clinical utility in differential diagnostic dilemmas.

Objectives
The aim of this study is to conduct a scoping review on recent (2014-2021) research on SC among patients with SCZ following the PRISMA extension for scoping reviews (PRISMA-ScR; [58]).The scoping review framework was opted for because it is more appropriate than a systematic review for answering broader research questions [57,58].Our objectives are to: 1. Map and explore current knowledge on SC in schizophrenia 2. Assess the quality and risk of bias of the eligible studies 3. Present stated limitations and recommendations for future research within the scope and produce recommendations based on the mapping and assessment of studies

Protocol and registration
A protocol following the PRISMA guidelines for scoping reviews (PRISMA-ScR) was developed collaboratively by M.E.and K.J., and further assessed by a third party (P.J.) for revision feedback.The final protocol was preregistered with the Open Science Framework on the 25th of January 2022 https://osf.io/358x2/.

Information sources and literature search
The following databases were employed in the literature search: PubMed, MedLine, PsycInfo.An experienced librarian aided in the drafting of the search strategy, which was revised iteratively.The final search strategy can be found in Supplementary Appendix A.

Eligibility criteria
Papers were deemed eligible if they were (1)

Selection of reports
From the performed searches, priority was given to those that yielded the most specific and multileveled results, including several terms of interest within the same search string.All studies within these selected searches were screened for relevance and eligibility on the basis of title and abstract.Relevant matches were scanned for duplicates, and the remaining articles were assessed in their entirety by M.E.and K.J. Disagreements were resolved through internal discussions or with the help of a third party (P.J.).The final selection of articles deemed eligible was also checked and approved by P.J.

Data charting process
A data-charting form was developed through collaboration (M.E., K.J.) and evaluated by a third reviewer (P.J.) to decide which data items were to be extracted from each article.Manual data charting was performed independently by the two reviewers and later compared and discussed in order to compile all relevant data.Any discrepancies were resolved through the help of a third reviewer (P.J.).The data-charting form was updated iteratively when necessary.

Data items
From the eligible articles, we extracted (1) Administrative data (First author(s), place of origin), ( 2

Critical appraisal of individual sources of evidence
Articles were assessed for bias independently by K.J. and M.E.using The Scottish Intercollegiate Guidelines Network (SIGN) Methodology checklist for case controls [59] Results were compared in order to judge each study as either High, Acceptable or Low quality.Any discrepancies were resolved through discussion and/or by utilizing a third assessor granted a final say in the matter (P.J.).

Synthesis of results
Since all studies claimed to assess "social cognition" or "mentalization", we suspected that a variety of closely related yet slightly different social cognitive constructs was measured across the scope.We, therefore, intended to thematically categorize each outcome measure within a "domain" together with others aiming to assess similar constructs.For each measure, quotes describing its intended purpose were extracted from all studies employing it.If a clear or explicit purpose was not identified within the methods sections, quotes were sought for in other sections of the article.All quotes were compared for similarities and differences, where identical or similar descriptions (e.g."emotion" and "affect") guided the indexing.All outcome measures sharing similar purpose descriptions were grouped together under a domain name summarizing their stated intent, and domains were iteratively discarded and refined throughout the process of presenting additional outcome measures.Potential domains were primarily discarded if they lacked the ability to distinguish measures from each other (e.g.tests described as assessing "social cognition").Identified domains were then used as review subheadings to guide the data synthesis.If an outcome measure was not described sufficiently in regard to its purpose in any study employing it, its task procedure was evaluated to infer its suitable purpose.
In order to further recognize patterns and potential gaps, and to assess the quality of research across the scope, we developed a "Gap Map" inspired by EPPI-Mapper [60], combining the identified domains and the eligible studies with the results of the critical assessment.This was done to create an overview of existing research in terms of trends and quality across different areas of SC based on the choice of outcome measures.
Finally, the discussion section of each article was scanned for mentions of limitations and recommendations for future research, which was summarized and categorized by way of description similarities.Limitations or recommendations that could not easily be clustered together with other categories were fit into a broader category labelled "other" by way of convenience.

Selection of sources of evidence
The literature search generated 15 articles included for data extraction and critical assessment.See Figure 1 for an overview of the article selection.

Characteristics of sources of evidence
The characteristics of the included studies can be found in Table 1.

Critical appraisal within sources of evidence
The results of the critical assessment can be found in Figure 2.

Results of individual sources of evidence
The results of the data extraction can be found in its entirety in Supplementary Appendix B.

Key concepts used within the scope
The phenomena of interest were described using the following terms: "social cognition" (n = 13), "Theory of Mind"/"ToM" (n = 12) and "mentalizing"/"mentalization" (n = 8).11 studies used >1 term, and 6 studies employed all three."ToM" was used to describe measurements across three out of the five identified domains, whereas "mentalizing" and "social cognition" were both used once only for measurement descriptive purposes.

Identified domains
The thematic indexing resulted in the identification of five "domains", containing all SC measurements employed throughout the scope: 1) Social perception/Face recognition, 2) Emotion attribution, 3) Intention attribution, 4) Mental state/ Belief attribution and 5) Higher order/Complex SC.No outcome measure was deemed appropriate within >1 domain, although four measurements (RMET, TJAT, TFB-P, TFB-D, RAP) were either described inconsistently or insufficiently and therefore given an allocation after examining their respective task procedures.See Figure 3 for an overview of the domains and their contents.For a comprehensive list of SC outcome measures, see Supplementary Appendix C.

Gap map
The results of the Gap Mapping are displayed in Table 2.

Social perception/face recognition
Two studies [61,73] assessed social perception and/or face recognition abilities in SCZ, ASD and HC samples.Both employed the Benton Face Recognition Test (BFRT) [76] and found impairments only for ASD with effect sizes varying from small (d = 0.27-0.48) to medium and large (d = 0.67-0.96).Altschuler et al. [61] reported IQ level as unrelated to performance on BFRT across all groups.Pinkham et al. [73] found weak correlations between BRFT performance and NSC abilities assessed with the MATRICS battery in both ASD (r = 0.301) and SCZ (r = 0.339), however not in HC.One study [73] used the basic Biological Motion task (Bio Motion) [77] to assess social perception, without finding significant differences.The same study also employed the Relationships Across Domains test (RAD) [78] and identified equal impairments in both SCZ and ASD compared to HC with small effect sizes (d = 0.37-0.41).Weak to moderate associations were found between RAD performance and NSC ability in HC (r = 0.299), ASD (r = 0.491) and SCZ (r = 0.618).

Emotion attribution
Four studies [61,63,67,73] measured emotion attribution using the Reading the Mind in the Eyes Test (RMET) [8,9], out of which one [67] reported performance compared to HC, and two [61,73] to both HC and ASD.All three reported SCZ performing significantly worse than HC with varying effect sizes (d = 0.42-d = 2.32).Neither of the studies comparing SCZ to ASD found significant between-group differences, although between ASD and HC (d=.40-.89).Pinkham et al. [73] discovered moderate associations between NSC abilities and RMET performance in ASD (r = 0.520) and SCZ (r = 0.556), but not in HC.  [62] to examine the neural basis of toM impairments in patients with fEs versus hc scZ (n = 17) scZ patients showed impairments in toM compared to controls.Both hypo-and hypermentalization, the latter also identified using fMri hc (n = 17) Buck et al.
(2018) [63] to examine the automatic and controlled components that underlie an intentionality bias in scZ versus hc. to test a model of symptomatic domains which in psychosis could be impacted by an intentionality bias.
scZ + sca (n = 213) scZ patients showed a bias towards attributing intentionality, and a difficulty to effortfully control responses.these difficulties seem to be related to trait hostility and various functional outcomes, but not specifically to psychotic symptoms hc (n = 151) de la asuncion et al.
(2015) [64] to examine the interplay between fairness, intentionality, and emotional considerations during social decision-making in scZ versus hc scZ (n = 35) scZ patients rejected intentional unfair offers more likely than unintentional unfair offers, which points towards an intact sense of general fairness, but patients also rejected fair offers more often than healthy controls did.healthy controls seemed to be more affected by emotional facial cues than scZ participants.Four studies [63,69,73,75] assessed emotion recognition with the Penn Emotion Recognition test (ER-40) [79], out of which all but one [63] presented between-group comparisons.Both SCZ and ASD scored significantly lower than HC, with effect sizes ranging between d = 0.29-d = 1.24.Significant differences between the clinical groups regarding the recognition of specific emotions were reported.ER-40 performance correlated with NSC abilities in ASD (r=.260) and SCZ (r = 431) but not in HC [73].
Two studies ( [63,73]) assessed emotion recognition using the Bell Lysaker Emotion Recognition Test (BLERT) [80], out of Figure 2. overview of the critical appraisal of included studies.assessment of the included studies using the scottish intercollegiate Guidelines Network (siGN) checklist for case-control studies.White indicates "high quality" signifying clear methodological transparency and little to no bias.Grey indicates "acceptable quality", signifying some uncertainties and flaws associated with a risk of bias.Black indicates "low quality", signifying several flaws or a high risk of bias.as seen above, 47% (7/15) of studies were assessed as having an overall low quality in terms of methodological transparency.
which one [73] presented performance compared to HC.For total score and Sad recognition, HC outperformed both ASD (d = 0.26, d = 0.29) and SCZ (d=.40, d=.35) whereas clinical groups performed equally.BLERT performance correlated with NSC abilities in ASD (r = 0.442) and SCZ (r = 0.269) but not in HC.The same study [73] further assessed emotion recognition in ASD, SCZ and HC using the emotional biological motion task (EmoBio) [77], reporting HC significantly outperforming SCZ and ASD (d = 0.62, d = 0.40).Significant differences between clinical groups regarding the recognition of specific emotions were also reported.EmoBio performance correlated with NSC abilities in ASD (r = 0.385) and SCZ (r = 0.450) but not in HC.
One study [75] assessed emotion recognition in SCZ, ASD and HC using the Juslin and Laukka Auditory Emotion Recognition (JL-AER) [81].HC and ASD performed similarly, outperforming SCZ (d = 1.27, d = 0.58), however, differences were nonsignificant after accounting for sensory detection, ER-40 performance and processing speed.Another study [71] assessed emotion recognition in SCZ and HC using the Emotional Intelligence Scale-Faces [82] task without finding significant differences.Similarly, one study [65] found no between-group differences in raw scores in ASD, SCZ and HC when assessing emotion recognition using the Emotional face recognition task [65].
Finally, one study [74] assessed emotion recognition using their own Emotions in Context Task (ECT) and found HC significantly outperforming SCZ and ASD on Overall accuracy, % of correct scenes in isolation and response time for congruent context.SCZ performed significantly worse than both ASD and HC on % fixation time on face as well as accuracy for incongruent context, with no significant differences between ASD and HC.The authors found that premorbid IQ correlated with accuracy on each of the four conditions (r=.387-.516)for the SCZ group, but not for HC or ASD.

Intention attribution
Three studies [62,66,68] assessed intention attribution using the Animated Triangles Task (ATT) [83].All reported reduced performance in SCZ compared to HC, along with less attributed intentionality in SCZ than HC for the ToM condition.For the random condition, one of these studies also reported more intentionality attribution in SCZ than in HC [62], whereas two did not.Between-group differences could partly be explained by IQ in one study [62], whereas NSC was dismissed as an explanation by another [68].The latter study additionally included an ASD sample and reported no differences between ASD and SCZ on overall ATT performance.

Outcome measures: THT, TASIT, CTOM-I, MASC, TFB-P, TFB-D, TJAT
Higher order/Complex SC "Higher order", "Complex" + "Decision making", + "Social judgement" + "Sarcasm detecƟon" Outcome measures: UG, TT, RAP One study [66] used the Find the Chase Task to assess intentionality attribution in SCZ and HC, reporting no between-group differences.One study [63] used the Intentionality Bias Task (IBT) [84] to explore a potential intentionality attribution bias in SCZ compared to HC.They reported a higher overall tendency in SCZ to attribute intentionality, and a lower ability to control responses (η p 2 =.04), compared to HC.A higher degree of automatic bias for perceiving intentionality was also identified within SCZ compared to HC (η p 2 =.02) through the Process Dissociation Procedure (PDP) [85].No associations between IBT score and symptom severity or NSC were found.
One study [72] used its own modified version of IBT (The grammatical intentionality task).This study reported an intentionality bias in SCZ relative to HC, and further reported differences in functioning between the groups in regard to intentionality attribution depending on sentence type.Between-group differences were not, however, influenced by grammatical structure.
One study [71] assessed the attribution of intentionality using the Communicative Intentions Recognition Task (CIRT) [86].This study reported lower performance in SCZ compared to HC, unrelated to IQ.

Mental state/belief attribution
Three studies [63,67,73] used The Hinting Task (THT) [3] to assess mental state attribution.Two reported performances compared to HC [67,73] both found differences between SCZ and HC (d = 1.47, d = 0.62 respectively).One of these [73] reported no differences between SCZ and ASD in performance, concluding equal impairments across clinical conditions.In their study, THT performance was weakly associated with the NSC abilities in ASD (r = 0.400) and SCZ (r = 0.474) but not in HC.
Two studies [63,73] assessed mental state attribution using The Awareness of Social Inference Test (TASIT) [10], the latter reporting between-group comparisons.Lower total scores in ASD (d=.59) and SCZ (d = 0.72) compared to HC were found.TASIT total score was weakly positively associated with NSC abilities (r = 0.277-0.407) in all groups.
One study [66] used The False-Belief Picture Sequencing task (TFB-P) [87] and The False-Belief/Deception comprehension task (TFB-D) [88] to assess mental state attribution.Significantly lower scores were reported in SCZ than in HC on the ToM condition for both tasks.The strongest predictor of accurate ToM reasoning was a combination of NSC performance and ATT ToM Intentionality performance (the tendency to attribute intentionality, independently of accuracy).The same study used The Jokes Appreciation Task (TJAT) [89] and reported lower performance for SCZ than HC on the ToM condition, whereas no between-group differences were found on the random condition.One study [70] assessed mental state attribution using the Movie for the Assessment of Social Cognition (MASC) [90].They reported significantly more correct attributions in SCZ than in ASD, and fewer correct attributions in both SCZ and ASD compared to HC.
One study [73] measured mental state attribution for ASD, SCZ and HC using the Cartoon Theory of Mind task-Intentions subscale (CTOM-I) [91] and reported no significant between-group differences.They did however find weak to moderate associations between CTOM-I performance and NSC abilities (r = 0.347-0.571) in all groups.

Higher order/complex SC
One study [64] compared SCZ to HC on Higher order/Complex SC using their own modified version of the Ultimatum Game (UG).More fair offers were rejected in SCZ than HC, whereas no between-group differences were found in the acceptance of unfair offers.Emotional cues affected decision-making in HC, but not in SCZ.
One study [73] used the Trustworthiness Task (TT) [92] to assess complex SC in SCZ and ASD compared to HC. Significant between-group differences were found on the Trustworthy condition, but not on the total score or on the Untrustworthy condition.SCZ and ASD did not differ significantly from one another.No correlations were found between TT performance and NSC.
One study [75] assessed "sarcasm detection" in SCZ, ASD and HC using the Ross Attitudinal Prosody battery -Sarcasm (RAP) [93] and reported that SCZ performed worse than ASD, with both clinical groups impaired compared to HC.

Mentioned limitations and recommendations for future research within the scope Limitations.
Samples.Sample related limitations were mentioned by 8 studies, either stating a too small of a sample [65,66,70,71], that sample sizes differed between groups [68] or that the sample was unrepresentable [64,66,69,70,73].
Other limitations.Two studies mentioned a lack of resources [61,67], and one [62] mentioned the downside of not using a pilot study, as well as attrition issues.Five studies mentioned limitations directly related to the statistical methods applied [65,67,69,73,75].
Measurements.Three articles [61,63,71] called for research on measures and constructs related to SC, one stating that "more basic science work is needed to examine the construct validity of tasks" [82].
Aims.Three studies suggested identifying underlying mechanisms for SC difficulties [61,62,69], and another three encouraged the search for predictors, associations and mediators [67,74,75].Two studies encouraged research on SC interventions, specifically targeting face recognition [61] and social decision-making [64].Three advocated for research on SC in relation to neighbouring areas, such as attention [65], mindfulness and language abilities [67] and broader cognitive functioning [75].

Summary of evidence
In this scoping review, we identified 15 studies comparing SCZ to HC and/or ASD on SC outcome measures across five thematically identified domains.Most studies employed one SC measure, although this number varied from 1-5 per study.Emotion attribution measurements were most commonly used following Intention attribution measurements, whereas comparatively few employed measurements within the remaining three.Impairments in both SCZ and ASD compared to HC were reported with few exceptions across all domains.
Differences between SCZ and ASD were however not as consistently found.When identified, effect sizes were small or moderate.Most studies lacked methodological transparency according to SIGN [59], and approximately half achieved low quality.Notably, a majority failed to state how they ensured their HC samples' control status, e.g. by administering a clinical interview to dismiss psychiatric illness.Only three studies mentioned blinding to reduce expectation bias.Additional limitations were also identified.Authors reappeared across the scope, suggesting a limited number of active researchers.This could lead to a compounding of errors and decreased generalizability due to sample overlaps [94], and to "paradigmatism" by increasingly dogmatic views on theories and concepts.Three studies included authors from the SCOPE studys' [95,96] "RAND panel".Out of the remaining studies, eleven cited ≥1 RAND members whereas only one [72] featured neither.Samples were male dominated, which although clinically representative implies less applicability for female patients.Conversely, most studies excluded patients with concurrent substance abuse, eliminating the risk of confusing disorders [97,98], while also contradicting the clinical reality [99].
With an understanding of NSC as necessary-yetnot-sufficient for SC, the former must be assumed to play some part in SC performance.However, only one of the nine studies measuring it matched samples based on present IQ [61].Four studies matched samples using premorbid IQ, which although a valuable metric visavi the progression of SCZ falls short in accounting for whether present SC performance could be attributed to NSC.Two studies used MATRICS, a NSC battery specifically designed for patients with SCZ and not necessarily fit for comparisons with HC, and three did not mention measuring NSC.Studies including NSC in their analyses (irrespective of assessment method) generally reported weak to moderate associations to SC performance.Such associations were often only found in the clinical samples, perhaps suggesting NSC status being of greater importance for those with ASD and SCZ during social processing.
Another issue consists of studies modifying existing or developing novel SC outcome measures, rather than employing already available outcome measures with known psychometric properties.Not only do new measurements complicate validity estimations, but furthermore hinder meta-analyses, since their comparability to other SC measurements is unknown-perhaps indicated in the between-study heterogeneity observed in previous meta-analytic attempts [54,55].
The present study also aimed to compile stated limitations and recommendations in order to summarize collective obstacles and their proposed solutions.A common grievance was sample sizes, and several suggested increasing samples although without addressing how this may be achieved.Commonly mentioned confounds included antipsychotic medication and IQ/NSC status.For the former, suggested solutions included matching or using more discriminatory study designs.For the latter, matching samples based on NSC performance was advised.Another common concern related to outcome measures in terms of quantity, quality, validity and reliability, and conducting more basic research to ensure construct validity was advocated.Several authors raised the need to address their research questions outside of correlational studies.

Conclusions and implications for clinical practice
Although substantial research has explored SC in SCZ, few studies directly comparing SCZ to HC and/or to ASD on measures of SC were found.Several conceptual and methodological issues identified in the scope have been discussed, in particular those related to the operationalization and assessment of SC.Several authors mentioned them not entirely covering the range of potential confounders, which regardless might be impossible since many such attempts counteract each other.It should however be considered an issue that no authors discussed the risk of spillover between groups resulting from not fully ensuring controls are entirely healthy at study inclusion.In accordance with current criticism [23][24][25] our thematic indexing indicated that the ToM term lacked discriminatory power.Most studies within the scope reported impairments in SCZ and/or ASD compared to HC across several domains of SC, and most studies including an ASD population reported similar impairments in SCZ and ASD, in line with previous evidence.SC abilities have previously been linked to clinical outcomes in SCZ [40,41] (for a review, see [2]).Considering the SC impairments associated with SCZ, accounting for these during assessment could thus be of clinical importance, as also suggested by [41].Within Mentalization-Based Treatments (MBT) [100], the assessment of mentalizing domains is used to aid treatment by identifying the patients' skills and needs (for a comprehensive list of rating scales, see [101]).An understanding of how the patient perceives and appraises social information might further allow healthcare workers to share and discuss treatment-related information adapted to the individuals' unique needs and conditions [38].

Limitations
Search strategies risk being limited by publication bias as well as researchers' preexisting notions.We minimized these risks by using multiple databases, receiving a librarians' guidance and exploring terms of interest using several controlled vocabularies.
In this study, current understandings of SC measurements were utilized for indexing.While offering insight in where the field currently rests, this indexing should not be considered a reflection of any actual SC "domains", particularly since the conceptual confusion evident in the field [23,24] likely obstructs a consistent use of key terms.Meaningful comparisons between our descriptive domains and those previously established via factor analysis [28,47] are therefore impossible to make.
The data charting may additionally be subject to human error, and albeit not included in the PRISMA-ScR requirements, measuring inter-rater reliability could have increased our studys' transparency further.Lastly, we digressed from our preregistered protocol twice: the initial method of critical assessment was changed since SIGN was better geared towards assessing case-control studies, and the cataloging of stated limitations and recommendations was added ad hoc.

Recommendations for future research
Few studies within our scope adequately accounted for NSC despite the seeming consensus regarding NSC performance as a relevant factor for SC performance [6,7].Future studies ought to include such measures in order to ensure that what is being operationalized as "social" cognition is in fact distinguishable from NSC, and in order to further explore how it potentially varies between as well as within groups.Current SC measures seem unable to differentiate between ASD and SCZ on scoring alone, making their clinical utility questionable.However, current evidence suggests that patterns of performance may vary between the two.Future development of SC outcome measures ought to focus on trying to discriminate between clinical groups, perhaps by accounting for the type, rather than amount, of mistakes made.A remaining issue revolves around how to best divide social cognitive outcome measures into statistically comparable "domains", as theory-driven attempts at clustering SC measurements have not resulted in sufficiently homogenous groups.Future research, in the vein of previous attempts mentioned [28,47], should therefore aim to establish valid subdomains of SC.

Figure 3 .
Figure 3. thematic indexing and assessments for five domains of social cognition.
Figure 1.flowchart of the literature searches and screening of studies.

Table 1 .
aim, population and main findings for all n = 15 included studies.= 42) Both clinical groups (asd and scZ) showed impairments in affective toM compared to controls, but did not differ from each other significantly.face recognition ability is related to affective toM ability in asd, but not in scZ.