A Meta-Analysis of Mentalizing Impairments in Adults With Schizophrenia and Autism Spectrum Disorder

Abstract

Mentalizing has been examined both in autism spectrum disorder (ASD) and schizophrenia (SCZ) primarily by either cognitive-linguistic (referred to as verbal) or emotion recognition from eyes (referred to as visual) mentalizing tasks. Each type of task is thought to measure different aspects of mentalizing. Differences in clinical features and developmental courses of each disorder may predict distinct patterns of mentalizing performance across dis orders on each type of task. To test this, a meta-analysis was conducted using 37 studies that assessed mentalizing either verbally or visually in adults with SCZ or ASD. We found that the estimated effect sizes of impairments in verbal and visual mentalizing tasks for both clinical groups were statistically large and at a similar level (overall Hedges’ g = 0.73-1.05). For each disorder, adults with SCZ showed a trend towards larger impairments on verbal (overall Hedges’ g = 0.99) than on visual mentalizing task (overall Hedges’ g = 0.73; Qbet = 3.45, p =.06, df =1). Adults with ASD did not show different levels of impairment on the verbal versus visual tasks (Qbet = 0.08, p =.78, df =1). These results suggest that both clinical groups share, at least in part, some common cognitive processing deficits associated with mentalizing impairments.

Introduction

Mentalizing is a high-level component of social cognition defined as the essential ability that allows us to infer the mental states of others, such as beliefs, thoughts, and emotions.^1,2 The development of successful mentalizing proceeds through different stages from childhood to adolescence into adulthood.^3,4 Given that schizophrenia (SCZ) and autism spectrum disorders (ASD) are characterized by impaired social functioning,^5,6 mentalizing has been frequently examined both for SCZ and ASD using identical mentalizing tasks in independent studies. However, mentalizing impairments commonly observed in the 2 disorders may reflect different causal factors. Comparing mentalizing impairments across the 2 disorders may help us understand whether there are common mentalizing impairments that cut across diagnostic categories⁷ or could help differentiate each disorder in a way that might enhance the identification of disorder-specific endophenotypic markers. We speculate that there might be disorder-specific patterns of mentalizing performance deficits on visually based measures of emotion recognition from eyes vs cognitive linguistic–based measures of intention and/or belief inference due to different clinical features and developmental courses for each disorder, as described in more detail below. Thus, the goal of the current meta-analysis was to determine if adults with SCZ vs ASD show similar or different levels of impairments on emotion recognition from eyes vs intention/belief inference tasks.

ASD is a neurodevelopmental disorder with an early onset of symptoms and disrupted development of cognitive and social communication skills.⁸ Core features of ASD are characterized by social impairments due to the failure of coordinating visual social cues on the basis of information from eyes.^9,10 A body of literature has found that individuals with ASD show deficits of eye-gaze processing at different levels such as eye contact,¹¹ gaze following,^12,13 and joint attention,^12,14–16 which provide crucial information about emotions and mental states for successful mentalizing.¹⁷ Eye-gaze processing is an early-developing mentalizing skill found in infancy and early childhood before the acquisition of vocal language.^17,18 It allows one to learn social signals as early as the first month after birth.¹⁹ At 12–18 months of age, a typically developing infant attends to the object of another’s attention and is able to infer another’s mental state from an eye gaze at 36–48 months.^17,20 The typical function of eye-gaze processing in an early stage of development is an essential precursor to the development of language and high-order mentalizing abilities in later life.²¹ For example, there is consistent evidence showing that the ability to initiate joint attention among preschool-aged children with ASD predicted later language ability at age 3–4 years.²² Similarly, early ability to understand and infer emotion in others predicts later high-order mentalizing abilities.^23–25

Although successful processing of eye gaze in early development is critical, it is not sufficient for the development of advanced intention and/or belief inference, which usually emerges around 8–9 years of age.²⁶ Intention and/or belief inference also requires the ability to recognize that another’s knowledge state is different from one’s own and integrate this information with situational context and advanced language understanding to determine “real” intentions and meanings beyond the literal meaning.²⁷ Given this sequence of mentalizing development, early developmental disruptions in inferring others’ emotion from eyes in childhood may contribute to a cascade of later mentalizing deficits in adulthood of ASD.⁹

In contrast to ASD, SCZ has been frequently considered an adulthood disorder, with the median ages of 20–35 years at onset of the first psychotic symptoms (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition; DSM-IV)⁸ although some subgroup of patients with SCZ presents their psychotic symptoms in early life.²⁸ Some impairments in social and nonsocial cognition can precede the onset of psychotic symptoms in many individuals,²⁹ leading some to think of SCZ as a neurodevelopmental disorder. A body of social cognitive literature suggests that mentalizing impairments occur at or prior to the onset of psychosis and continue throughout the course of SCZ, with evidence of verbally measured mentalizing deficits even in high-risk populations (though these may be less severe).^6,30 Nonetheless, different from the ASD literature, there is no compelling evidence suggesting abnormal precursors of mentalizing, especially, emotion inference on the basis of information gained from eyes in early childhood or prodromal phase of SCZ.^31,32

Interestingly, emerging evidence suggests that context processing, which has been long considered a core cognitive function thought to be supported by the dorsolateral prefrontal cortex (PFC), mediates, at least partly, mentalizing impairments in SCZ.^33,34 Here, context processing refers to the ability to actively represent task- or situation-relevant information in working memory to guide ongoing processing.³⁵ It should be noted that context processing in SCZ may reflect different psychological or neurobiological mechanisms from that of the ASD literature, which indicates “weak central coherence” on perceptual context processing tasks. In the SCZ literature, an increasing number of studies have shown the relevance of context processing to impairments on cognitive- linguistic mentalizing tasks in SCZ. For example, context processing, as measured by the AX-continuous performance test,³⁶ partially mediated mental state inferences in patients with SCZ.³⁷ This line of research suggests that mentalizing impairments in SCZ may be due to inefficient or failed context processing.³⁸ As such, we speculate that at least some mentalizing impairments in SCZ may be secondary to other cognitive deficits, such as context processing deficits, and may emerge later in development, potentially providing a relative sparing of earlier developing mentalizing abilities such as inferring others’ emotions from eyes.

As described above, the different developmental courses of ASD and SCZ may predict different patterns of mentalizing impairments on tasks that measure emotion recognition from eyes vs cognitive-linguistic functions as measured by intention/belief inference tasks. Considering abnormal precursors of mentalizing in early life such as eye contact and gaze orientation for ASD,³⁹ it is predicted that in the case of ASD, abnormal precursors to mentalizing in early life may lead to a cascade of later mentalizing deficits in adulthood such as intention/belief inference that may build upon these more basic functions. In contrast, mentalizing impairments in SCZ may stem from domain-general deficits, such as context processing, rather than from basic function in early development. Thus, we hypothesized that adults with ASD should show similar levels of impairment on emotion recognition from eyes vs intention/belief inference tasks, whereas individuals with SCZ should show greater impairments on intention/belief inference tasks than on emotion recognition from eyes. Further, we hypothesized that adults with ASD will show greater impairments on emotion recognition from eyes compared with those with SCZ. However, only a few studies have directly addressed these questions, with mixed results. For example, patients with ASD showed worse performance on a facial recognition test than did patients with SCZ,⁴⁰ which is consistent with the hypothesis outlined above. In contrast, other work found that adults with SCZ and those with Asperger’s syndrome performed worse on the Hinting test⁴¹ and the “Reading the Mind in the Eyes” test (Eyes)¹⁴ relative to healthy controls (HCs), but that performance was not significantly different from each other on either task.⁴² Other research⁴³ has shown that children with high-functioning ASD were more impaired compared with those with SCZ on a deception task⁴⁴ though not on a false-belief task.⁴⁵

The purpose of this study was to compare less cognitive-linguistic (referred to as visual) vs cognitive-linguistic (referred to as verbal) aspects of mentalizing across disorders by conducting a meta-analysis of studies with SCZ and ASD samples. This meta-analysis will provide a synthesized effect size estimate that has more power than individual studies. The verbal tasks included the Strange Stories⁴⁶ and Faux Pas tests²⁶, and the visual task was the Eyes task,⁴⁷ which is hypothesized to require less demand on cognitive-linguistic abilities. We predicted that adults with ASD would show similar impairments in both types of tasks, but that adults with SCZ would show greater impairments in verbal tasks rather than the visual task. Further, we predicted that adults with ASD would show greater impairments in visual task compared with adults with SCZ but potentially similar deficits in the cognitive-linguistic tasks. Despite previous meta-analyses,^6,48 mixed findings exist regarding the influence of clinical and demographic characteristics on mentalizing in SCZ. For example, it is still not clear whether mentalizing deficits are moderated by general intelligence deficits, antipsychotic medications, positive/negative symptom severity and other demographic features such as age and gender. Similarly, in the ASD literature, despite early theory suggesting female superiority of emphasizing in ASD,⁴⁹ there is no empirical evidence whether gender moderates mentalizing performance for adults with ASD. Therefore, the second aim of this meta-analysis was to examine moderating effects of positive and negative symptom severity of SCZ, IQ, antipsychotic mediations, and other demographic characteristics including age and gender on mentalizing impairments in the 2 disorders.

Methods

Data Sources and Literature Search

The flowchart of the present meta-analysis is presented in figure 1.

Fig. 1.

Schematic procedure for studies included and excluded in meta-analysis.

ASD, autism spectrum disorders; HC, healthy controls; SCZ, schizophrenia.

^aA literature search covered published papers up to December 2011 using the databases Pubmed, Medline, EMBASE, or Sciencedirect to identify the relevant studies.

^bOf these 37 studies, 2 included both individuals with SCZ and those with ASD.

^cThe total sample included 634 adults with SCZ (80% male), 362 ASD (82% male), and 926 HC (65% male).

Inclusion Criteria and Coding of the Variables

In terms of SCZ studies, consistent with prior review studies,⁶ we limited the sample to studies with SCZ disorder participants, including schizoaffective disorder or first-episode psychosis, but excluding any personality-related disorder. Studies involving the contrast between individuals with SCZ and HCs and/or between ASD and HC were included. The details of inclusion criteria and coding variables are presented in table 1.

Table 1.

Inclusion Criteria and Coding Variables

Inclusion Criteria

1. Studies should be published in peer-reviewed journals in English.

2. Studies should report mentalizing abilities in adults with diagnoses of schizophrenia, schizoaffective disorder or first-episode psychosis (SCZ) or ASD according to Research Diagnostic Criteria, DSM-III-R, DSM-IV or the International Classification of Diseases criteria.

3. Mentalizing tasks should be originally developed in autism and adopted to SCZ literature and used at least in 5 independent studies either for SCZ or for ASD.

4. Studies should also include healthy subjects as a comparison group.

5. For the comparison between adults with SCZ and ASD, the age of participants who performed those tasks should be between 18 and 65 years.

6. Studies should report means and standard deviations, or F or t values, or exact P value so that standardized mean differences could be calculated. Measures with dichotomous outcomes were excluded.

Coding of the Variables

1. Name of the first author and year of the publication.

2. Number of participants and percentage of males in patients and healthy control groups (eg, schizophrenia and autism spectrum disorder).

3. Mean and SDs for demographic variables (eg, age, duration of education), patients variables (eg, age of onset, duration of illness), percentage of medicated patients, and IQ assessment.

4. Mean, SDs and calculated effect sizes of the individual tasks. A total mentalizing score that was calculated by averaging effect sizes of individual tasks.

Note: ASD, autism spectrum disorders; SCZ, schizophrenia.

Mentalizing Tasks

Cognitive-Linguistic Tasks.

The typical cognitive- linguistic mentalizing task includes verbal passages in which participants have to infer a character’s mental state with given specific situational details. Without considering situational details, a person cannot infer the character’s mental states appropriately. As such, the correct responses in these tasks require participants to make explicit use of cognitive reasoning to provide situational explanations that go beyond the literal meaning about an interaction that has been described in verbal passages.⁴⁶ Performance on cognitive-linguistic mentalizing tasks demand language understanding and may also require some aspects of executive function, as evidenced by prior research showing correlations among verbal-type mentalizing performance, impaired metaphor and/or proverb understanding, and executive function.^50,51 We chose the Strange Stories⁴⁶ and Faux Pas tasks^26,52 because they have both been used in a sufficient number of studies with either SCZ or ASD (or both, see online supplementary methods). Although a few other studies used additional tasks that might measure similar constructs, we did not combine these tasks so as to minimize potential confounds arising from combining tasks with potentially different psychometric properties.

Emotion Recognition From Eyes (Less Cognitive-Linguistic Tasks). “Reading the Mind in the Eyes test” (the Eyes)^47,53 measures recognition of emotions and mental state in others from eyes.⁵⁴ Unlike the cognitive- linguistic tasks described above, the Eyes test does not include any situational details that should be integrated for correct responses. Therefore, relative to the cognitive-linguistic mentalizing tasks, performance on the Eyes test may depend on less cognitive function and language understanding.⁵³

Statistical Analyses

Meta-Analyses Procedure.

The meta-analysis software, Comprehensive Meta-analysis,⁵⁵ was used to analyze the data. The mean and standard deviation for each comparison (between adults with SCZ and HC: HC-SCZ, between adults with ASD and HC: HC-ASD) were used as a measure of the effect size. Homogeneity of the resulting mean-weighted effect sizes was tested with the Q test. We used a random-effects model for the meta-analyses. Publication bias was investigated using funnel plots, Egger’s test, and the Fail Safe number (see online supplementary methods for details).

Effect Size Estimation.

We estimated effect sizes using Hedges’ g. ⁵⁶ For individual studies, we computed a standardized effect size by subtracting the mean of the patient group from that of the HC (eg, HC-SCZ or HC-ASD comparison). We then divided the results by the pooled standard deviation of the 2 groups with Hedges’ g correction for bias⁵⁶ in small samples. We reported between-group differences as positive effect sizes when patient groups performed poorer than the HC. For studies that reported more than 1 mentalizing task, a pooled effect size was calculated.

Selecting Moderator Variables.

The list of moderating variables and summary data is reported in tables 5 and 6. The effects of demographic, clinical variables including symptoms severity, and general intelligence on group differences in mentalizing impairments were examined using a meta-regression method.

Table 5.

Meta-Regression of Potential Moderators of Effect Sizes in Schizophrenia

	Family of Variables	Descriptive
Task Type	Variable Name	Ka	M (SD)	Unit	P-value
Verbal	Patient status (all outpatient=1, all inpatients=0)	7	0.78 (0.39)	Mean proportion	.15
	BPRS or PANSS overall psychopathology	7	0.32 (0.22)	Severity index 0–1	.71
	Positive symptoms	11	0.14 (0.08)	Severity index 0–1	.07
	Negative symptoms	11	0.22 (0.13)	Severity index 0–1	.43
	Age of onset	7	24.08 (1.34)	Years	.88
	Illness of duration	9	11.76 (6.57)	Years	.14
	Proportion of medicated patients	8	0.99 (0.03)	Mean proportion	.09
	Difference in male proportion	11	−0.09 (0.11)	Difference of mean proportion	.54
	Difference in age	11	−0.08 (0.28)	Hedges’ g	.47
	Difference in FIQ reported by NART or WAISa	6	0.85 (0.40)	Hedges’ g	.49
Visual	Variable name	Ka	M (SD)	Unit	P-value
	Patient status (all outpatient =1, all inpatients=0)	8	0.94 (0.17)	Mean proportion	>.05
	BPRS or PANSS overall psychopathology	8	0.26 (0.21)	Severity index 0–1	.85
	Positive symptoms	11	0.14 (0.09)	Severity index 0–1	.24
	Negative symptoms	11	0.23 (0.11)	Severity index 0–1	.26
	Illness of duration	11	12.48 (3.80)	Years	.21
	Proportion of medicated patients	10	0.983 (0.05)	Mean proportion	.96
	Difference in male proportion	15	−0.06 (0.08)	Difference in mean proportion	.59
	Difference in age	16	−0.14 (0.35)	Hedges’ g	.75
	Difference in FIQ reported by NART or WAIS	8	0.73 (0.31)	Hedges’ g	.44
	Difference in education	12	0.54 (0.55)	Hedges’ g	.69

Note: PANSS, Positive and Negative Syndrome Scale;

Table 6.

Meta-Regression Analyses of Potential Moderators of Effect Sizes in Autism Spectrum Disorders

Task Type	Family of Variables	Descriptive
Verbal	Variable Name	Ka	M (SD)	Unit	P Value
	Difference in male proportion	10	−0.08 (0.14)	Difference of mean proportion	.08
	Difference in age	9	−0.11 (0.23)	Hedges’ g	.67
	Difference in FIQ reported by NART or WAIS	7	0.47 (0.51)	Hedges’ g	.96
Visual	Variable Name	Ka	M (SD)	Unit	P Value
	Difference in male proportion	10	−0.08 (0.11)	Difference of mean proportion	.71
	Difference in age	11	−0.08 (0.50)	Hedges’ g	.33
	Difference in FIQ reported by NART or WAIS	8	0.26 (0.34)	Hedges’ g	.18

Note: FIQ, full IQ; NART, National Adult Reading Test; WAIS, Wechsler Adult Intelligence Scale.

Results

Descriptive Information

The characteristics and effect sizes of studies included in the analysis are presented in figure 1 and tables 2 and 3.

Table 2.

Characteristics and Effect Sizes of Studies of Schizophrenia Included in the Meta-Analysis

Cognitive-Linguistic Mentalizing Tasks
Study	Sample (Male)	Matched For	Age (M)	Hedges’ g
de Achával et al. (2010)	20 (13) SCZ	Age, sex	30.9	0.84
	20 (13) HC		28.2
Gavilan et al. (2011)	22 (18) SCZ	Age, education, sex, premorbid IQ	42.82	1.80
	22 (18) HC		41.95
Herold et al. (2009)	18 (11) SCZ	Age,sex	28.7	0.82
	21 (11) HC		27.4
Hooker et al. (2011)	21 (17) SCZ	Age, sex	44.33	1.46
	17 (13) HC		43.75
Langdon et al. (2010)	35 (23) SCZ	Age, gender, IQ	35.9	0.52
	34 (26) HC		32.0
Martino et al. (2007)	21 (12.18) SCZ	Age, education, IQ	32.66	1.30
	15 (6) HC		34.96
Pijnenborg et al. (2009)	46 (34) SCZ	No matched for age, gender, IQ	27.4	0.68
	53 (24) HC		31.1
Riveros et al. (2010)	15 (12) SCZ	Age, gender, education	37.57	0.92
	18 (11) HC		40.5
Stanford et al. (2011)	13 (6) SCZ	Age, sex, education	33.6	1.03
	14 (6) HC		29.1
Shur et al. (2008)	26 (17) SCZ	Age, education	32.58	0.90
	35 (20) HC		29.00
Zhu et al. (2007)	40 (18) SCZ	Age, education, sex	30.2	1.27
	31 (9) HC		29.97
Emotion Recognition From Eyes Task
Study	Participants	Matched For	Age (M)	Hedges’ g
Bailey and Henry (2010)	28 (14) SCZ	Age, education	40.3	0.60
	30 (15) HC		36.4
Bora et al. (2008)	91 (60) SCZ	Education	31.1	0.72
	55 (34) HC		35.6
Couture et al. (2008)	26 (22.88) SCZ	No matched for age and gender	24.9	0.43
	41 (38.13) HC		23.0
Couture et al. (2010)a	41 (34) HC	Differ in age and IQ	22.9	0.76
	44 (39) SCZ		27.5
Craig et al. (2004)a	16(11) SCZ	Age, IQ	31.69	1.64
	16 (11) HC		29.44
De Achával et al. (2010)	20 (13) SCZ	Age, gender	30.9	0.80
	20 (13) HC		28.2
Hirao et al. (2008)	20 (10) SCZ	Age, gender, and education	36.7	1.89
	20 (10) HC		35.0
Irani et al. (2006)	10 (7) SCZ	Gender, ethnicity	34.00	0.94
	10 (5) HC		38.00
Kington et al. (2000)	16 (13) SCZ	Age, gender, ethnicity	34.07	0.55
	16 (13) HC		34.75
Mcglade et al. (2008)	73 (49) SCZ	Age, gender	41.4	0.38
	78 (45) HC		38.3
Riveros et al. (2010)	15 (12) SCZ	Age, gender, education	37.57	0.47
	18 (11) HC		40.5
Russell et al. (2000)	5 (5) SCZ	Age, gender, education	36.00	1.37
	7 (7) HC		40.00
Schimansky et al. (2010)	40 (28) SCZ	Age	38.0	0.71
	40 (22) HC		34.4
Shur et al. (2008)	26 (17) SCZ	Age, education	32.58	0.53
	35 (20) HC		29.00
Stanford et al. (2011)	13 (6) SCZ	Age, gender, education	33.6	0.79
	14 (6) HC		29.1
Tso et al. (2010)	33 (22) SCZ	Age, gender, education	38.5	0.72
	33 (23) HC		38.2

Note: HC, healthy control; SCZ, schizophrenia.

^aTwo studies included both individuals with schizophrenia and those with autism spectrum disorders.

Table 3.

Characteristics and Effect Sizes of Studies of Autism Spectrum Disorders Included in the Meta-Analysis

Cognitive-Linguistic Mentalizing Tasks
Study	Participants (Male)	Matched For	Age (M)	Hedges’ g
Adler et al. (2010)	16 (15) HFA	Age, education, gender	21.87	0.69
	21 (20) HC		22.90
David et al. (2008)	24 (14) HFA	Age, gender, IQ	32.3	0.30
	24 (13) HC		30.6
Happe et al. (1994)	18 (13) ASD	Age	20.6	2.88
	10 (5) HC		20.5
Heavey et al. (2000)	16 (15) HFA	Age	34.7	1.56
	15 (15) HC		30.7
Hill et al. (2004)	15 (12) ASD	Age	34.40	1.13
	15 (7) HC		34.00
Jolliffe & Baron-Cohen (1999)	17 (15) HFA	Age, gender, IQ	30.71	1.73
	17 (15) ASD		27.77
	17 (15) HC		30.00
Ponnet et al. (2004)	19 (14) ASD	Age, IQ	21.06	0.04
	19 (14) HC		21.93
Roeyers et al. (2001)	24 (22) ASD	Gender, age, IQ, education	23.8	0.32
	24 (22) HC		23.1
Spek et al. (2010)	32 (27) HFA	Age, gender, IQ	42.1	0.81
	29 (25) ASD		43.67
	32 (24) HC		38.68
Zalla et al. (2009)	15 (12) ASD	Age, gender, IQ, education	28.00	1.72
	15 (11) HC		27.80
Emotion Recognition From Eyes Task
Study	Participants (Male)	Matched For	Age (M)	Hedges’ g
Adler et al. (2010)	16 (15) HFA	Age, education, gender	21.87	0.70
	21 (20) HC		22.90
Baron-Cohen et al. (1997)	16 (13) HFA	IQ, age	28.6	1.74
	16 (13) HC		30.0
Baron-Cohen et al. (1999)	6 (4) ASD	IQ, age, education	26.3	1.40
	12 (6) HC		25.5
Baron-Cohen et al. (2001)	15 (15) HFA/ASD	Age, IQ	29.7	1.69
	14 (14) HC		28.0
Couture et al. (2010)a	36 (29) HFA	Differ in age and IQ	20.9	0.61
	41 (34) HC		22.9
Craig et al. (2004)a	16 (11) HC	Age, IQ	24.12	1.43
	17 (15) ASD		29.44
David et al. (2008)	24 (14) HFA	Age, gender, IQ	32.3	1.23
	24 (13) HC		30.6
Kleinman et al. (2001)	30 (21) HFA	No matched for age, IQ, gender	31.43	0.42
	24 (10) HC		22.33
Ponnet et al. (2004)	19 (14) ASD	Age, IQ, gender	21.06	0.21
	19 (14) HC		21.93
Roeyers et al. (2001)	24 (22) ASD	Gender, education	23.8	0.11
	24 (22) HC		23.1
Spek et al. (2010)	32 (27) HFA	Age, gender, IQ	42.1	0.32
	29 (25) ASD		43.67
	32 (24) HC		38.68

Note: ASD, autism spectrum disorders; HC, healthy control; HFA, high-functioning autism; SCZ, schizophrenia.

^aTwo studies included both individuals with schizophrenia and those with autism spectrum disorders.

Impairments on Cognitive-Linguistic Tasks in SCZ and ASD

Meta-analysis of the cognitive-linguistic tests demonstrated highly significant mentalizing impairments in adults with SCZ and ASD compared with HC (table 4). Effect sizes of HC-SCZ and HC-ASD comparisons were quite large (overall Hedges’ g = 0.99, 1.05, respectively, both P < .001). Further, the magnitude of the deficits for HC-ASD and HC-SCZ were not significantly different according to the Q-test (Q _bet = 0.00, df = 1, P = .96). The distribution of the effect sizes of cognitive-linguistic mentalizing impairments for HC-SCZ showed a trend for heterogeneity according to the Q-test (Q _w = 18.59, P = .05, df = 10) but were significantly heterogeneous for the HC-ASD comparison (Q _w = 45.38, P = .00, df = 9). Egger’s regression test showed no publication bias for the cognitive-linguistic tasks in either comparison (HC-ASD: b = 3.35, P = .19; HC-SCZ: b = 2.06, P = .12). The fail-safe number of missing studies required to make the group difference nonsignificant is also quite large (n = 239 and n = 449 for the comparison for HC-ASD and HC-SCZ, respectively) (see figures 2 and 3 for mean effect sizes and confidence intervals of SCZ and ASD samples).

Table 4.

Overall Effect Sizes for Group Differences on Cognitive-Linguistic and Eye-Gaze Mentalizing Tasks

Task Type	Comparisons	Ka	Hedges’ g	95% CI	z	P	Q-test P
Verbal	HC-SCZ	11	0.99	0.78–1.20	9.15	.00**	.05
	HC-ASD	10	1.05	0.60–1.50	4.58	.00**	.00**
Visual	HC-SCZ	16	0.73	0.56–0.90	8.36	.00**	.06
	HC-ASD	11	0.81	0.48–1.14	4.83	.00**	.00**

Note: ASD, autism spectrum disorders; HC, healthy control; SCZ, schizophrenia.^aThe number of studies included.p < .01

Fig. 2.

Forest plot with mean effect size (g) and confidence intervals for cognitive linguistic–based mentalizing tasks in patients with schizophrenia.

Fig. 3.

Forest plot with mean effect size (g) and confidence intervals for cognitive linguistic–based mentalizing tasks in adults with autism spectrum disorders.

Impairments on the “Reading the Mind in the Eyes Task” in SCZ and ASD

Meta-analysis of the Eyes test also demonstrated significant mentalizing impairments in both clinical groups. The overall effect sizes of HC-SCZ and HC-ASD comparisons were 0.73 and 0.81 (both P < .001). However, contrary to our predictions, the magnitude of effect sizes of HC-SCZ and HC-ASD comparisons was not significantly different according to the Q-test (Q _bet = 0.27, P = .60, df = 1). Again, the distribution of the effect sizes of the Eyes test for HC-SCZ tended to be heterogeneous at a trend level (Q _w = 23.73, P = .06, df = 15), whereas the distribution of the effect sizes of the Eyes test for HC-ASD comparisons was significantly heterogeneous according to the Q-test (Q _w = 37.75, P < .001, df = 10) (table 4).

Egger’s regression test suggested that there might be some publication bias for performances on the Eyes test in adults with ASD (b = 4.66, df = 9, P = .01). However, the fail-safe number of missing studies needed to make the group difference nonsignificant was large (n = 217), considering the relatively small number of studies included in the analysis (k = 11 studies). Similarly, Egger’s regression test for SCZ also showed a possibility for publication bias (b = 2.38, P = .01). However, the fail-safe number of missing studies was again quite large (n = 473) (see figures 4 and 5 for mean effect sizes and confidence intervals of SCZ and ASD samples).

Fig. 4.

Forest plot with mean effect size (g) and confidence intervals for visual mentalizing tasks in patients with schizophrenia.

Fig. 5.

Forest plot with mean effect size (g) and confidence intervals for visual mentalizing tasks in adults with autism spectrum disorders.

Comparisons of Task Types Within Each Diagnostic Group

In SCZ, the magnitude of mentalizing impairments on the cognitive linguistic tasks was larger than that of mentalizing deficits on the Eyes test at a trend level (HC-SCZ: Q _bet = 3.45, P = .06, df = 1). In the ASD group, the magnitude of mentalizing impairments on the Eyes test was not significantly different from the cognitive-linguistic mentalizing tasks (HC-ASD: Q _bet = .08, P = .78, df = 1).

Meta-Regression Analyses

To further investigate the sources of the variance in effect sizes, we examined potential moderator effects on mentalizing impairments.

Cognitive-Linguistic Mentalizing Tasks

Schizophrenia.

No demographic characteristics or clinical symptoms had an impact on the magnitude of mentalizing impairments in SCZ (see table 5).

Autism Spectrum Disorder.

Due to the small number of studies reporting these variables, we included only 3 moderator factors: age, IQ, and the difference in the proportion of males in the ASD vs HC groups in meta-regression analysis (see table 6). Group differences in age (k = 9 studies) and full IQ scores (k = 7 studies) had no impact on the magnitude of mentalizing impairments in adults with ASD. However, differences in the proportion of males in the HC and ASD groups tended to moderate the magnitude of impairments on the cognitive-linguistic mentalizing tasks, such that a larger relative percentage of males in the ASD group was associated with greater impairments at a trend level (k = 10 studies, B = 1.67, SE = 0.96, Z = 1.73, P = .08).

Eyes Task

Schizophrenia and Autism Spectrum Disorder.

No moderator variables significantly moderated the variance in mentalizing impairments on the Eyes task in either group.

Discussion

The purpose of the current meta-analysis was to compare adults with SCZ and ASD on impairments in emotion recognition from eyes (visual) and intention/belief inference (cognitive-linguistic) tasks. The current meta-analysis showed that adults with SCZ and ASD exhibited large effect size deficits on both types of mentalizing tasks (SCZ: overall Hedges’ g = 0.56–1.20, ASD: overall Hedges’ g = 0.48–1.50). Impairments in cognitive-linguistic mentalizing tasks in SCZ were somewhat larger than on the Eyes test at a trend level but did not differ across tasks in ASD. Similarly, the magnitude of impairments on both types of tasks did not differ across groups.

The current meta-analysis findings indicating a similar pattern of mentalizing impairments in adults with SCZ and ASD are consistent with 2 studies that directly compared adults with SCZ and those with ASD regarding mentalizing performance. For example, Craig et al⁴² compared patients with paranoid delusions and those with Asperger syndrome. Their results showed that both clinical groups exhibited impaired performance on the Hinting task⁴¹ and the Eyes tests compared with the HC but that the scores for the 2 patient groups did not differ from each other. Consistent with our results, the direct comparison study between ASD and SCZ by Couture et al⁶⁹ also found that both individuals with (High-Functioning Autism) HFA (n = 36) and SCZ (n = 44) showed impairments on the Eyes test compared with the HC. Importantly, there were no clinical group differences on the Eyes test⁴⁷ after controlling for IQ.⁶⁹

Contrary to our predictions, adults with ASD did not show greater impairments than adults with SCZ on the Eye’s test. The Eye’s test uses still photographs of the eyes. It is possible that subtle deficits on eye-gaze processing in HFA with above-average IQ would be more pronounced in natural settings that include less predictable and more complex online social interactions rather than in well-structured experimental settings.^87,88 In addition, as previous reviews have pointed out,^48,89 limited psychometric properties such as low discriminatory power might have also led to a failure to discriminate differentially impaired aspects of mentalizing performance in ASD vs SCZ.

It should be noted that the majority of the ASD studies we included had HFA and Asperger Syndrome participants with at least average intelligence (average IQ = 108.54). Therefore, generalizing current results showing similar pattern of mentalizing deficits between SCZ and ASD to full-blown autism with an intellectual disability (IQ < 70) may be limited, and we might have seen greater differences had we been able to include more studies with lower functioning ASD. Second, the average age for adults with SCZ (M = 46.55) was older compared with those with ASD (M = 29.65) and HC (M = 31.71). We cannot rule out the possibility that age may have influenced both verbally (Strange Stories) and visually measured (the Eyes test) mentalizing performance. However, the current meta-analysis found no age moderator effect on either type of mentalizing performance (tables 5 and 6). Further, consistent with our results, emerging evidence suggests there is no relationship between age and mentalizing abilities.^90–92

To further investigate the source of heterogeneous effect sizes, we examined the effect of putative moderators. Consistent with a previous review of studies with SCZ samples,⁴⁸ the current meta-regression results showed no moderating effects on the effect size of mentalizing impairments. In terms of the relationship between symptom severity and mentalizing impairments, converging evidence suggests the possibility that mentalizing impairments are a trait factor in SCZ rather than state-dependent factor, considering that remitted patients⁶⁸ and first-degree relatives of patients with SCZ show mentalizing impairments.^48,71,93

Regarding gender effects on mentalizing abilities, we found that a larger relative percentage of males in the ASD, but not in then SCZ group, tended to be associated with greater impairments in the cognitive-linguistic mentalizing tasks (k = 10 studies, B = 1.67, SE = 0.96, Z = 1.73, P = .08), consistent with “female superiority” in mentalizing, as argued by Baron-Cohen et al.⁹⁴ However, except for early studies by Baron-Cohen et al^26,47,49,94 later studies from different research groups have failed to find a significant gender difference on the Eyes test among adults with ASD,⁸⁶ adults with SCZ,⁴⁸ relatives of patients with SCZ,^57,95 and other clinical groups.⁹⁶ Thus, further research is needed to determine whether there are strong and consistent gender differences in mentalizing performance among individuals with ASD.

Despite limited information about the psychometric properties of mentalizing tasks, comparing cognitive-linguistic verbal mentalizing with visual mentalizing, especially for adults with ASD and SCZ, is a crucial attempt to tease apart multifaceted mentalizing abilities into specific aspects of impaired psychological or neural mechanisms in each disorder. The results of this meta-analysis suggest that both SCZ and ASD disorders share at least some common impairment in mentalizing. Emerging evidence from neuroimaging studies consistently indicates that several neural regions show altered activation in both disorders during mentalizing tasks. Such findings may suggest a common neural mechanism underlying the observed behavioral deficits in the 2 disorders. For example, a recent neuroimaging meta-analyses⁹⁷ found that adults with SCZ and ASD both exhibited reduced engagement of medial PFC, thought with a greater degree of impairment in patients with ASD during facial emotional processing and the Eyes test⁴⁷ compared with SCZ.

The current results may have been limited for the following reasons. The major limitation is that this study compared effects sizes across diagnostic groups from different studies, rather than from studies that directly compared SCZ and ASD, though our results are consistent with previous studies that directly compared between ASD and SCZ.^42,69 Further, despite no significant impact of proportion of medicated patients on the magnitude of mentalizing impairments in SCZ, we cannot rule out the possibility that antipsychotic medication may have affected mentalizing abilities in SCZ, which led to similar degree of verbal-type mentalizing impairments between ASD and SCZ. However, there is little empirical evidence suggesting that antipsychotic medication improves mentalizing ability in SCZ. For example, Mizrahi et al⁹⁸ examined drug-induced improvement in psychotic symptoms of SCZ and mentalizing ability measured by the Hinting task. They found positive scores and mentalizing ability in SCZ improved after the first 2 weeks of antipsychotic treatment. However, the changes observed in psychotic symptoms and the percentage improvements in mentalizing performance were not significantly associated, and they could not rule out the possibility that the improvements on the Hinting task reflected practice effects, though different versions were used across testing sessions. Further, the mentalizing tasks chosen in the current meta-analysis were quite selective. However, the comparison of cognitive-linguistic vs emotion recognition from eyes is consistent with a distinction made in the literature between mental state decoding (the Eyes test) and reasoning (cognitive-linguistic mentalizing tasks).^99–102 To support this, a distinctive neural mechanism has been found to be involved in performance of the different sets of tasks. Whereas left ventromedial PFC regions appear to be involved in cognitive “reasoning” about mental states,^103,104 the right inferior frontal and anterior temporal lobes may be relatively crucial for “decoding sensory processing” based mentalizing.^100,104 Importantly, it should be noted that because most mentalizing tasks included in the current meta-analysis have been used in experimental settings, most of them have not been standardized, which led to investigators’ arbitrary choice of items and/or modifications for a given task. For example, in terms of the Strange Stories task, Gavilan et al⁵⁸ reduced the number of passages considering patients’ cognitive difficulty and developed 3 types of story passages from the original version of the Strange Stories. Also, 2 versions of the Eyes test (eg, Baron-Cohen et al^14,47) existed in the current analysis. However, given that each clinical group and HCs used the same mentalizing task in an individual study, observed mean difference of mentalizing impairments cannot simply reflect psychometric artifacts related to the use of slightly varying versions of the tasks instead of distinct aspects of mentalizing performance.

In summary, in the current meta-analysis, adults with SCZ and ASD exhibited similar levels of mentalizing impairments in the 2 types of mentalizing tasks. These results suggest at least in part, common cognitive behavioral deficits in mentalizing for these 2 clinical disorders. However, the current mentalizing tasks used in this field may not be sufficiently sensitive to distinguish visually (emotion recognition from eyes) vs verbally (cognitive-linguistic) measured mentalizing performance due to their uncertain psychometric properties. As mentioned in the Introduction, mentalizing impairments in ASD may evolve from impairments in deficits starting early in life, including alterations in eye contact and gaze orientation.³⁹ On the other hand, mentalizing impairments in SCZ may arise later in life in patients with SCZ and may reflect greater contributions from cognitive deficits. As such, longitudinal studies that compare the evolution of mentalizing abilities across the 2 disorders may help identify and understand disorder-specific features of mentalizing. Furthermore, along with establishing psychometric properties of mentalizing tasks, future studies are required to develop novel tasks that tap into specific psychological constructs or neural circuitry, which potentially differentiate ASD and SCZ.

Supplementary Material

Supplementary material is available at http://schizophre niabulletin.oxfordjournals.org.