Association of Schizotypy With Dimensions of Cognitive Control: A Meta-Analysis

Abstract

Schizotypy is defined as a time-stable multidimensional personality trait consisting of positive, negative, and disorganized facets. Schizotypy is considered as a model system of psychosis, as there is considerable overlap between the 2 constructs. High schizotypy is associated with subtle but fairly widespread cognitive alterations, which include poorer performance in tasks measuring cognitive control. Similar but more pronounced impairments in cognitive control have been described extensively in psychosis. We here sought to provide a quantitative estimation of the effect size of impairments in schizotypy in the updating, shifting, and inhibition dimensions of cognitive control. We included studies of healthy adults from both general population and college samples, which used either categorical or correlative designs. Negative schizotypy was associated with significantly poorer performance on shifting (g = 0.32) and updating (g = 0.11). Positive schizotypy was associated with significantly poorer performance on shifting (g = 0.18). There were no significant associations between schizotypy and inhibition. The divergence in results for positive, negative, and disorganized schizotypy emphasizes the importance of examining relationships between cognition and the facets of schizotypy rather than using the overall score. Our findings also underline the importance of more detailed research to further understand and define this complex personality construct, which will also be of importance when applying schizotypy as a model system for psychosis.

Introduction

Schizotypy is a time-stable multidimensional personality trait consisting of positive facets, such as unusual perceptual experiences and ideas of reference, negative facets, such as no close friends and flat affect and disorganized facets, such as odd speech and eccentric behavior.^1–3 In addition, schizotypy can be quantified as an overall score derived from the combination of these subfacets (“mixed schizotypy”).⁴

The scientific investigation of schizotypy is mainly motivated by 2 aspects. Firstly, following a personality-based approach allows the exploration of the underlying mechanisms of schizotypy to further our understanding of this complex cluster of traits.^5,6 Secondly, following a clinical approach allows to examine clinical implications of high levels of schizotypy and to evaluate the role of schizotypy within the spectrum of psychosis.⁷ For instance, it has been shown that high schizotypy is associated with self-reports of poor quality of life⁸ and higher drug use.⁹ Moreover, as there is considerable overlap in the clinical picture of schizotypy and schizophrenia,^7,10,11 it has been suggested that schizotypy can serve as a model system of psychosis, allowing the investigation of psychosis in the absence of potential confounds such as medication or hospitalization.^10,11 Following this approach, variation in schizotypy as well as its covariation with cognitive, neural, and behavioral measures has been studied with the objective to elucidate the etiology of psychosis.^3,12–15

Evidence from both correlational (making full use of variance in schizotypy scores) and categorical (allocating individuals to discrete groups on the basis of schizotypy scores) studies suggests subtle yet fairly widespread cognitive alterations in high schizotypy,⁶ including poorer performance in tasks measuring cognitive control.^16–25 Cognitive control, or executive function, involves the domain-independent representation and maintenance of task goals, thereby supporting the flexible adaptation of information processing and behavior in a changing environment. Importantly, cognitive control refers to a heterogeneous set of functions. An influential model argues that cognitive control involves 3 overlapping yet separable dimensions, namely shifting, updating, and inhibition.²⁶Shifting involves the flexible switching of attention between tasks or mental sets. Updating refers to the flexible updating as well as monitoring of information in working memory. Inhibition refers to the ability to withhold a nondesirable response that is dominant or prepotent.

Impairments in cognitive control are a cardinal feature of schizophrenia,²⁷ with recent research efforts focusing on the measurement of cognitive control functions to improve both the classification of psychopathology²⁸ and pharmacological treatment.^29,30 Given the importance of cognitive control deficits in psychosis,^28–30 they are a key target for research aiming to validate schizotypy as a model system.^7,15

Although there is evidence of cognitive control impairments in high schizotypy,^16,31 some studies could not replicate these findings.^22,32,33 A recent meta-analysis by Chun et al⁸ concluded that high schizotypy in university samples was associated with somewhat poorer performance in set shifting (shifting) and working memory (updating) tasks, while other cognitive features, such as attention and memory, were intact.⁸

Extending this previous work, we carried out comprehensive meta-analyses of studies reporting associations between schizotypy and inhibition, shifting, and updating. We restricted our definition of schizotypy to psychometric instruments¹⁰ such as the SPQ,³⁴ but did not include clinical symptom rating scales, eg, Paranoia Checklist.³⁵ To avoid bias due to sampling strategies, we included all studies that investigated these cognitive control dimensions in relation to schizotypy, irrespective of whether they used correlational or categorical designs. Moreover, we included studies irrespective of whether participants were drawn from university or general community samples to avoid selection bias.

Method

Literature Search

Following the PRISMA guidelines,³⁶ we identified relevant published studies by searching the electronic database PubMed for combinations of the keywords “schizotypy” or “psychosis proneness” and their variations with the keywords “cognition” or “neuropsychology” and their variations as well as more specific terms such as “working memory,” “response inhibition,” “set shifting,” and “executive functions.” The full search term can be found in the supplementary materials. Relevant studies that were published before July 19, 2017 were included in the current analyses.

Study Selection

For studies to be included in the meta-analyses, the following inclusion criteria had to be fulfilled. First, they were required to report performance on a task measuring shifting (eg, Wisconsin Card Sorting Test; WCST, Trail Making Test; TMT), updating (eg, n-back, spatial working memory tasks), or inhibition (eg, Stroop task, antisaccade task, Stop-Signal task). Second, sufficient data needed to be reported or made available upon request to allow the quantification of the relationship between performance and schizotypy (standardized mean difference or correlation). Review articles not reporting original data or articles reporting single case studies were excluded. Third, schizotypal personality scores had to be obtained via psychometric questionnaires (eg, Schizotypal Personality Questionnaire,³⁴ SPQ; Chapman psychosis-proneness scales^37–39; Oxford-Liverpool Inventory of Feelings and Experiences,¹ O-LIFE) irrespective of how commonly the questionnaire was used. Fourth, the sample had to consist of healthy participants. Thus, for patient studies only healthy control groups were included. Fifth, the sample had to comprise adults only. Sixth, the article had to be published in an English language journal with peer review.

Our conceptualization of schizotypy should be distinguished from the symptom oriented approach proposed by van Os and Reininghaus,⁴⁰ although the 2 approaches may in fact be related.⁴¹ To maximize homogeneity in the conceptualization of schizotypy in our analyses, we included only studies that employed psychometric instruments for measuring schizotypy,¹⁰ eg, the SPQ³⁴ and did not include clinical symptom rating scales, eg, the Paranoia Checklist.³⁵

To provide a comprehensive summary of the literature, we did not limit inclusion of tasks measuring cognitive control to those used by Miyake et al,²⁶ but included other tasks that we considered to measure the relevant dimensions. We decided in consensus which tasks were to be included in the analyses. An exhaustive list of all tasks included can be found in the supplementary materials.

Data Extraction and Computation of Effect Sizes

The main outcome measure was the relationship between cognitive task performance and schizotypal personality scores. Data extraction was performed by M.S. and I.M. and verified by K.F. and P.K. (see Acknowledgments section). When studies compared performance between individuals with high schizotypy and individuals with low or medium schizotypy, relevant data were extracted to enable a computation of the standardized mean effect Hedges’ g.⁴² When studies reported a relationship between performance scores and schizotypal personality scores, correlation coefficients were extracted and converted to Hedges’ g.⁴² The variance of g was estimated for all included studies.⁴² Computation of Hedges’ g and its variance was carried out blindly by purpose-written routines in Matlab (MathWorks, version 2014a).

For studies that reported categorical and correlative outcomes, the group results were included in the meta-analyses. For studies with categorical designs that investigated more than 2 groups (ie, low, average, high) the most extreme groups were selected for further computations.

To assure the validity of effect size conversion from correlation coefficients to Hedges’ g, we tested the conversion using studies that provided both correlation coefficients and Hedges’ g measures. To assure the validity of combining effect sized from studies employing different designs (correlational, categorical), we performed moderator-analyses (separate meta-analyses for categorical and correlational design studies) for the variable study design as well as sensitivity analyses. Results of the sensitivity analyses are reported in supplementary table 1. Despite this potential risk of bias, effect size conversion is the preferred strategy in comparison to the exclusion of studies using an alternate metric which would be associated with a systematic loss of information.

We contacted the authors of articles that did not report the necessary data. Studies were excluded from our analysis whenever data to calculate effect sizes were not available. Furthermore, where reported, the following additional information was extracted: year of publication, schizotypy questionnaire, and sample type (community, college).

Computation of Meta-analyses and Moderator Effects

Following Viechtbauer,⁴³ the overall effect size was calculated using a random-effects-model with the restricted maximum-likelihood method to estimate heterogeneity in the population effect sizes. Between-study heterogeneity was evaluated using the Q-test.⁴⁴ As the 3-parameter selection model has been reported to outperform other bias correcting methods,⁴⁵ the presence of publication bias was explored by the weight-function model of Vevea et al.^46,47 All analyses were performed using the R software for statistical computation (version 3.4.1)⁴⁸ and the software packages metafor⁴⁹ and weightr.⁴⁷

Separate meta-analyses were conducted for each of the 3 cognitive domains (updating, shifting, inhibition) and each of the 3 schizotypy dimensions (positive, negative, disorganized). Additionally, we conducted meta-analyses for the 3 cognitive domains and schizotypy measured as an overall score (comprised of a combination of the schizotypy facets). Whenever a study had more than one outcome for one of the investigated cognitive domains or schizotypy dimensions, the corresponding effect sizes were averaged. We assumed K >8 studies as a sufficient size to conduct a meta-analysis. We chose this conservative threshold on the basis of the potential heterogeneity (tasks, schizotypy measures employed) and previous meta-analyses⁸ to avoid potential misinterpretations.

To further assess the influence of year of publication, study design (categorical, correlation), and sample type (community, college), these variables were included into the model as moderators. For information about which study employed which study design and sample type refer to supplementary table 4. If more than one dimension of schizotypy was significantly associated with a cognitive domain, we conducted an additional moderator analysis as follows: We computed a meta-analysis for all studies of these schizotypy dimensions with that cognitive domain and used the schizotypy dimension as a moderator.

Results

Literature Search

The detailed selection process and all exclusion criteria can be found in figure 1. Our literature search identified K = 1324 potentially relevant articles. After excluding K = 1132 articles based on their abstract and title, the full texts of the remaining K = 192 articles were assessed for eligibility. This procedure resulted in K = 86 articles that were included in the meta-analyses. Some studies could be assigned to more than one of the cognitive dimensions and some studies could be assigned to more than one of the schizotypy dimensions. An overview of the results for all conducted meta-analyses can be found in figure 2. For an overview of the tasks used in each analysis, please refer to supplementary table 3. For an overview of all questionnaires employed in the included studies, please refer to supplementary table 2.

Fig. 1.

The flow diagram shows the process of study inclusion and exclusion for the meta-analysis, including all exclusion criteria.

Fig. 2.

Overview of all conducted meta-analyses, y-axis displays mean Hedges’ g and 95% confidence intervals (CI).

Updating

Thirty-three studies investigated the relationship between schizotypy and updating. We conducted separate meta-analyses for positive, negative, and overall schizotypy. There were not enough studies to conduct a meta-analysis for disorganized schizotypy (K = 8).^{19,33,50–55} Results are reported in table 1 and illustrated in figure 3.

Table 1.

Overview of Meta-analytic Results

	Number of Studies Included (K)	Number of Subjects Included (N)	Hedges’ g (SE)	z (P)	95% CI		Q total (P)	χ2 (P)
					Lower	Upper
Updating
Positive schizotypy	14	1894	0.11 (0.06)	1.90 (.06)	−0.004	0.23	16.66 (.22)	0.83 (.66)
Negative schizotypy	12	1772	0.11 (0.05)	2.03 (.04)*	0.004	0.21	15.83 (.15)	2.13 (.34)
Overall schizotypy	19	2075	0.08 (0.05)	1.50 (.13)	−0.02	0.18	24.82 (.13)	6.57 (.04)
Shifting
Positive schizotypy	18	1692	0.18 (0.08)	2.30 (.02)*	0.03	0.34	37.40 (.003)	0.14 (.93)
Negative schizotypy	19	1846	0.32 (0.07)	4.56 (<.0001)*	0.18	0.46	30.83 (.03)	0.95 (.62)
Overall schizotypy	21	1708	0.06 (0.07)	0.84 (.40)	−0.08	0.20	37.70 (.01)	1.41 (.49)
Inhibition
Positive schizotypy	31	4284	0.15 (0.08)	1.91 (.06)	−0.004	0.31	100.04 (<.0001)	2.29 (.32)
Negative schizotypy	20	3503	0.07 (0.07)	1.05 (.29)	−0.06	0.20	35.67 (.01)	1.81 (.40)
Overall schizotypy	19	2605	0.15 (0.08)	1.79 (.07)	−0.01	0.31	35.84 (.007)	0.12 (.94)
Disorganized schizotypy	12	2900	0.08 (0.09)	0.85 (.40)	−0.10	0.26	27.17 (.004)	0.72 (.70)

Note: SE, standard error of the mean; CI, confidence interval.

Fig. 3.

Forest plots for the meta-analyses for the relationship between positive, negative, and overall schizotypy and updating and shifting, respectively. The figure shows effect size (Hedges’ g) and confidence intervals for all included studies. Studies are ordered alphabetically by last name of first author. Sample size is illustrated through the size of the square. The 95% confidence intervals are reflected in the length of the horizontal lines. Source: Chan (2011)a65; Chan (2011)b⁸⁴.

The SPQ was used in 20 studies, the Chapman psychosis-proneness scales were used in 7 studies, the O-LIFE was used in 6 studies, and the Cognitive Slippage Scale and the Schizotypal Personality scale (STA) were used in 1 study each. For an overview of the tasks used in the analyses, please refer to supplementary table 3.

Positive Schizotypy

Fourteen studies^{19,22,33,50–52,55–62} including 1894 participants were included. There was no significant effect of positive schizotypy on updating, with an average effect size of g = 0.11 (z = 1.90, P = .06). There was no publication bias (P = .66) and none of the moderators had a significant effect (all P ≥ .05).

Negative Schizotypy

Twelve studies^{17,19,22,33,50–52,55,57,62–64} including 1772 participants were included. There was a significant effect of negative schizotypy on updating, yielding an average effect size of g = 0.11 (z = 2.03, P = .04). There was no publication bias (P = .34) and none of the moderators had a significant effect (all P ≥ .06).

Overall Schizotypy

Nineteen studies^{4,8,17,18,33,50,51,55,65–75} including 2075 participants were included. There was no significant effect of overall schizotypy on updating, with an average effect size of g = 0.08 (z = 1.50, P = .13). There was evidence for a potential publication bias (P = .04). Publication year (P = .007), but none of the other moderators, had a significant effect (all P ≥ .30).

Shifting

Thirty-eight studies investigated the relationship between schizotypy and shifting. We conducted separate meta-analyses for positive, negative, and overall schizotypy. There were not enough studies to conduct a meta-analysis for disorganized schizotypy (K = 6).^{19,32,33,51,76,77} Results are reported in table 1 and illustrated in figure 3.

The SPQ was used in 22 studies, the Chapman psychosis- proneness scales were used in 12 studies, the O-LIFE was used in 4 studies, the STA and the Schizophrenism scale were used in 1 study each and a composite score of several schizotypy scales was used in 2 studies. For an overview of the tasks used in the analyses, please refer to supplementary table 3.

Positive Schizotypy

Eighteen studies^{19,21,22,24,32,33,51,57,58,62,76–83} including 1692 participants were included. There was a significant effect of positive schizotypy on shifting, yielding an average effect size of g = 0.18 (z = 2.30, P = .02). There was no publication bias (P = .93) and none of the moderators had a significant effect (all P ≥ .27).

Negative Schizotypy

Nineteen studies^{19,21,22,24,32,33,51,54,57,62,63,76–80,82,84,85} including 1846 participants were included. There was a significant effect of negative schizotypy on shifting, yielding an average effect size of g = 0.32 (z = 4.56, P < .0001). There was no publication bias (P = .62). The study design (P = .04), but none of the other moderators, had an effect on the analysis (all P ≥ .26). If only categorical studies were included in the analysis, the effect increased to g = 0.44 (z = 5.78, P < .0001) and if only correlative studies were included, the effect decreased to 0.18 (z = 1.79, P = .07).

Overall Schizotypy

Twenty-one studies^{18,21,33,51,55,65,72,73,82,84,86–96} including 1708 participants were included. There was no significant effect of overall schizotypy on shifting, at an average effect size of g = 0.06 (z = 0.84, P = .40). There was no publication bias (P = .49) and none of the moderators had a significant effect (all P ≥ .09).

Comparison Between Negative and Positive Schizotypy

There was no significant moderator effect of schizotypy dimension (positive, negative) for the association between schizotypy score and shifting (P = .20).

Inhibition

Forty-four studies investigated the relationship between schizotypy and inhibition. We conducted separate meta-analyses for positive, negative, disorganized, and overall schizotypy. Results are reported in table 1 and illustrated in figure 4.

Fig. 4.

Forest plots for the meta-analyses for the relationship between positive, negative, and overall schizotypy and inhibition. The figure shows effect size (Hedges’ g) and confidence intervals for all included studies. Studies are ordered alphabetically by last name of first author. Sample size is illustrated through the size of the square. The 95% confidence intervals are reflected in the length of the horizontal lines. Source: Chan (2011)a65; Chan (2011)b⁸⁴.

The SPQ was used in 22 studies, the Chapman psychosis-proneness scales were used in 13 studies, the STA was used in 6 studies, the Rust Inventory of Schizotypal Cognitions was used in 4 studies, the O-LIFE was used in 3 studies, the Cognitive Slippage scale, the Oviedo Schizotypy Assessment Questionnaire-Abbreviated, the Personality Syndrome Questionnaire, the Restrictive Emotional Expression scale, and the Schizophrenism scale were used in 1 study each, and a composite score of several schizotypy scales was used in 1 study. For an overview of the tasks used in the analyses, please refer to supplementary table 3.

Positive Schizotypy

Thirty-one studies^{23,24,32,52,57,77,80,82,83,97–118} including 4284 participants were included. There was no significant effect of positive schizotypy on inhibition, yielding an average effect size of g = 0.15 (z = 1.91, P = .06). There was no publication bias (P = .32) and none of the moderators had a significant effect (all P ≥ .20).

Negative Schizotypy

Twenty studies^{23,24,32,52,57,77,80,82,84,99,101,102,104,106–108,113,115–117} including 3503 participants were included. There was no significant effect of negative schizotypy on inhibition, yielding an average effect size of g = 0.07 (z = 1.05, P = .29). There was no publication bias (P = .40) and none of the moderators had an effect on the analysis (all P ≥ .07).

Disorganized Schizotypy

Twelve studies^{23,32,52,77,99,101,102,104,108,113,116,117} including 2900 participants were included. There was no significant effect of disorganized schizotypy on inhibition, yielding an average effect size of g = 0.08 (z = 0.85, P = .40). There was no publication bias (P = .70) and none of the moderators had a significant effect (all P ≥ .05).

Overall Schizotypy

Nineteen studies^{65,72,75,82,84,87,90,94,95,99,101,115–117,119–123} including 2605 participants were included. There was no significant effect of overall schizotypy on inhibition, yielding an average effect size of g = 0.15 (z = 1.79, P = .07). There was no publication bias (P = .94) and none of the moderators had a significant effect (all P ≥ .16).

Discussion

We investigated the association between schizotypy and dimensions of cognitive control in a comprehensive series of meta-analyses. To do so, we carried out separate analyses for combinations of the updating, shifting, and inhibition dimensions with positive, negative, disorganized, and overall schizotypy. In summary, negative schizotypy was associated with significantly poorer performance on shifting (g = 0.32) and updating (g = 0.11). The association between negative schizotypy and shifting was more pronounced for studies employing a categorical design than for those with a correlative design (g = 0.44 vs g = 0.18). Positive schizotypy was associated with significantly poorer performance on shifting tasks (g = 0.18). There was no significant difference in the strength of the association of positive schizotypy and shifting and negative schizotypy and shifting. There were no significant associations between schizotypy and inhibition. Despite significant heterogeneity in most of the analyses, the moderators that we included (study design, publication year, sample type) had only small effects on the results.

The differences in updating and shifting are in line with findings by Chun et al.⁸ Importantly, our analyses extend those by Chun et al⁸ beyond student populations and to a more comprehensive assessment of cognitive control and schizotypy. The deficits in cognitive control that we observed were most pronounced in relation to negative schizotypy. This finding is in line with literature discussing the negative dimension as a primary feature of schizotypy.¹²⁴ Negative schizotypy is the most heritable dimension and has been associated with poorer quality of life, poorer well-being, and higher levels of perceived stress.^6,124,125 The divergent results for the schizotypy facets reported here emphasize the importance of examining relationships between cognitive variables and the 3 facets of schizotypy rather than combining them into an overall score. While positive and negative schizotypy were associated with poorer performance in updating and shifting, we found no such association for overall (“mixed”) schizotypy. It should be noted that overall schizotypy is characterized by a mix of features from all domains and, therefore, may comprise a different population which seems, on the basis of our analyses, not to be associated with impairments in cognitive control. In terms of clinical implications, high schizotypy has been shown to be associated with impoverished quality of life.^8,126 It stands to reason that the small impairments on shifting and updating found here are not severe enough to explain this relationship. Some evidence suggests that impaired social cognition, such as emotional face recognition, may have a larger impact on quality of life.¹²⁷

With regards to schizotypy as a psychosis model system, in the present analysis individuals with high schizotypy showed somewhat similar deficits as those reported in schizophrenia, albeit to a much lower extent. Average effect sizes reported for schizophrenia patients on updating, shifting, and inhibition range from g = 0.86 to g = 0.99.¹²⁸ This is in line with findings that most features that overlap between schizotypy and schizophrenia are less pronounced in schizotypy.⁷ This further underlines schizotypy as a useful tool to study etiological mechanisms of psychosis as well as possible protective factors,¹¹ without the confounds such as medication and hospitalization¹⁴ typically encountered in patient samples.

Between-study heterogeneity was moderate to high for all analyses of the inhibition dimension, in line with findings that associations between tasks measuring inhibition are rather low.¹²⁹ This may make the detection of associations between this heterogeneous construct and schizotypy more challenging. In terms of schizotypy as a model system this suggests that schizotypy does not mimic all features of psychosis. To some extent this is true for all model systems.¹³⁰ However, considering the complexity of psychosis, it might be challenging for one single model to mimic all features of this heterogeneous disorder.¹³¹ Defining the limitations of existing model systems can improve our understanding of specific etiological mechanisms underlying psychotic symptoms. Additionally, it would be desirable to distinguish subfacets of inhibition and to investigate their association with schizotypy.

Our findings of somewhat overlapping impairments in schizophrenia and schizotypy are not of course proof of similar etiological factors. Therefore, the alternative hypothesis, that schizotypy and schizophrenia are 2 completely unrelated phenomena that both independently, via different molecular and cellular mechanisms, lead to impairments in cognitive control, has to be considered carefully.

A strength of our study is its comprehensiveness: We comprehensively quantified the influence of schizotypy facets as well as its total score on 3 key dimensions of cognitive control. To avoid bias due to sampling strategies, we included all studies that investigated these cognitive processes in relation to schizotypy, irrespective of whether they used correlational or categorical designs. Moreover, we included studies irrespective of whether participants were drawn from university or general community samples to avoid selection bias.

One limitation of our analyses might be that we included different measures of schizotypy. We did this to avoid sampling bias; however, this comes with the risk of increased heterogeneity of the studies included in our analyses. A second limitation is that we combined effect sizes from correlational and categorical studies. While this approach has the benefit of precluding systematic loss of information due to study exclusion, it may potentially lead to bias in our analyses. We addressed this potential bias by computing moderator analyses for the study design variable. Study design was only a significant moderator in the association between negative schizotypy and shifting, where the association was stronger for categorical than correlational studies. To further address this issue, we conducted separate analyses for categorical and correlational studies (sensitivity analysis, results reported in supplementary table 1), and found that effects tended to be higher for categorical than correlational studies overall. This is an important methodological implication for the interpretation of studies reporting associations between schizotypy and cognitive control. Finally, our meta-analysis did not of course provide a characterization of all cognitive impairments in schizotypy. Instead, we focused, on the basis of prior work,³⁰ on cognitive control as a central human cognitive function. Further meta-analyses are required to provide similar estimations of the magnitude of impairment in schizotypy in other domains and stages of information processing.

In summary, our analyses revealed subtle cognitive control deficits in high schizotypy. These were most pronounced in negative schizotypy, which was associated with impairments in the updating and shifting dimensions. The divergent results for positive, negative, and disorganized schizotypy reported here emphasize the importance of examining relationships between cognitive variables and the 3 facets of schizotypy rather than combining them to an overall score. Effects on cognitive control similar to those reported here have been described extensively in psychosis, although effects in patients generally are larger. Our findings underline the importance of more detailed research to further understand and define this complex personality construct, which will also be of importance when applying schizotypy research as a model system for psychosis. Such research may focus on direct comparisons between psychosis patients and schizotypy, comparisons between the different dimensions of schizotypy as well as a differential approach to executive function and its features.

Supplementary Material

Supplementary data are available at Schizophrenia Bulletin online.