opha

of Melbrsity o

a r t i c l e i n f o a b s t r a c t

Article history:Received 13 August 2008Received in revised form 6 December 2008Accepted 16 December 2008Available online 4 February 2009

Keywords:Theory of mindSchizophreniaMentalisingCognitive

1. Introduction attenuated but remained signicant in remitted patients

Schizophrenia Research 109 (2009) 19

Contents lists available at ScienceDirect

Schizophrenia

.e lsTheory of mind (ToM) decits in schizophrenia have beeninvestigated by a substantial number of studies since Frith(1992) proposed a model suggesting a relationship betweenmentalising (ToM) impairment and specic symptoms ofschizophrenia. Two systematic reviews (Brne, 2005; Har-rington et al., 2005a,b) and a recent meta-analysis (Sprong

(d=0.69). While this result suggests that acute psychosis hasa signicant moderating inuence on ToM performance, italso implies that mentalising decits might be trait-char-acteristics of schizophrenia.

While there is no doubt that schizophrenia patients performpoorly on ToM tasks, there are important issues that were notadequately considered in previous studies and reviews. Themainet al., 2007) showed that nearly all pubstudies reported ToM impairment. Spronreported a large overall effect size (d=1.25studies. Interestingly, the magnitude of the

Corresponding author. Alan Gilbert Building NNFAustralia. Tel.: +61 3 8345 5611; fax: +61 3 8345 561

E-mail addresses: emrebora@hotmail.com, boremre

0920-9964/$ see front matter 2008 Elsevier B.V.doi:10.1016/j.schres.2008.12.020There is now substantial evidence for Theory of mind (ToM) impairment in schizophrenia.Despite this, we know little about how dynamic (state) variables and broad clinical, cognitiveand medication characteristics moderate the precise magnitude of the observed ToM decitduring task performance. Meta-analyses were conducted using 36 studies that reportedcontinuous data regarding ToM performances of schizophrenia patients and healthy controlsubjects. These 36 studies included 1,181 (67% male) patients with schizophrenia and 936(58.3% male) healthy control subjects. Individual analyses were also conducted for the Hintingand the Eyes tasks. The effects of moderator variables were studied by both subgroup andmeta-regression analyses. The effect sizes (Cohen's d) for overall ToM performance and the individualtasks were large (d=0.901.08). In remitted patients, the degree of ToM impairment was lesspronounced than non-remitted patients (d=1.21) but it was still signicant (d=0.80).Moreover, the distribution of effect sizes was more homogeneous for the individual tasks,especially in remitted patients. General intellectual decits observed in schizophreniapatients contributed to their ToM impairment only in the remission phase of the illness. Whilestate variables and task specic differences explain a large degree of the heterogeneity of theToM ndings observed in previous studies, the persistence of ToM decits in remittedpatients suggests there are trait related mentalising impairments in schizophrenia. Our reviewalso suggests that future research should consider the potential moderating inuence of IQdecits on ToM performance in remitted patients, as well as the potential effects of residualsymptoms.

2008 Elsevier B.V. All rights reserved.Meta-analysisTheory of mind impairment in schiz

Emre Bora a,, Murat Yucel a,b, Christos Pantelisa Melbourne Neuropsychiatry Centre, Department of Psychiatry, The Universityb ORYGEN Youth Health Research Centre, Centre for Youth Mental Health, Unive

j ourna l homepage: wwwlished empiricalg et al. (2007)) for 29 publishedimpairment was

level 3, Carlton 3053,0.@gmail.com (E. Bora)

All rights reserved.renia: Meta-analysis

ourne and Melbourne Health, VIC, Australiaf Melbourne, VIC, Australia

Research

ev ie r.com/ locate /schresissues relate to (i) heterogeneity of ToM tasks and theirneurocognitive demands; (ii) psychometric properties of ToMtests; and (iii) the inuence of clinical and demographiccharacteristics on ToM performance. First, while the study ofSprong et al. (2007) provided the rst attempt to qualitativelyanalyze the available data, it did not adequately take theheterogeneity of ToM tasks into account. Previous studies haveused different ToM tests comprised of very different test

2 E. Bora et al. / Schizophrenia Research 109 (2009) 19characteristics. For example, different tasks rely on differentneurocognitiveprocesses anddecits of schizophreniapatients inworkingmemory, executive functions and attention,which coulddifferentially inuence (and interactwith) patients` performanceon various ToM tasks. Secondly, ToM tasks are not wellstandardized and their psychometric properties have not beenstudied. Forexample,while someof the false belief tasks includedonly single or a few stories, others included more stories. Insupport of this, Sprong et al. (2007) reported heterogeneity forfalse belief tasks that may be related to variability of these tasks.Finally, the moderating inuence of clinical and demographiccharacteristics on ToM dysfunction in schizophrenia may beunderestimated and confound interpretations. For example,general intellectual decits of the patients with schizophreniacould affect performance of mentalising tasks.

In this context, meta-analysis is a useful tool for system-atically combining all research in this area to identify themostrobust aspects of ToM impairment in schizophrenia and tounderstand how performance on such tasks is moderated bystate and trait factors. In this way, we may be better able tounderstand the inuence of moderating inuences on socialcognition disturbances that characterize schizophrenia. Ouraims were to: (i) to conduct meta-analyses for individual ormore homogeneous tasks that can estimate a reliable overalleffect size for the different aspects of ToM impairment; and(ii) to study the effects of potentially confounding factorsrelated to clinical and demographic characteristics on ToMimpairment in schizophrenia.

2. Method

2.1. Study selection

A literature search was conducted using the databasesPubmed, Medline, EMBASE, SCOPUS and PsycINFO to identifythe relevant studies (January 1990 to May 2008). The fol-lowing keywords were used: schizophrenia, psychosis,Theory of Mind, mentalising and social cognition. The earlyonline editions of the major psychiatric journals were alsochecked and reference lists of published reviews and studieswere used to identify additional relevant publications.

2.2. Inclusion criteria

1) Studies should be published in peer-reviewed journals inEnglish.

2) Studies should report ToM abilities in adults with diag-noses of schizophrenia, schizoaffective disorder or rst-episode psychosis according to Research Diagnostic Cri-teria, DSM-III, DSM-III-R, DSM-IV, ICD-9 or ICD-10 criteria.

3) Studies should also include healthy subjects as a compar-ison group.

4) Tasks should include at least 3 items.5) Studies should report means and standard deviations, or F

and t values so that standardized mean differences couldbe calculated. Measures with dichotomous outcomeswereexcluded.

6) Tasks should be used by at least three different researchgroups.

7) Studies should reect results of the independent samples.In the case of sample overlap, the study with larger samplesize was chosen for inclusion. Where necessary, authorswere contacted to provide additional information.

To be included in individual task analysis of the study,there was an additional inclusion criterion: tasks should beused in at least ve different studies.

2.3. Coding of the variables

The following variables were coded prior to meta-analyses:

a- name of the rst author and year of the publicationb- number of participants and percentage of males in both

groupsc- mean and SDs for demographic variables (Age, duration of

education), clinical variables (age of onset, duration ofillness), dosage of antipsychotics (chlorpromazine equiva-lents) and IQ assessment

d- mean, SDs and calculated effect sizes of the individualtasks. A total ToM score that was calculated by averagingeffect sizes of individual tasks.

2.4. Mentalising tasks

2.4.1. False-belief tasksMost common measures used to assess ToM were false-

belief understanding tasks. In their simplest form, these tasksmeasure the subject's ability to understand that someone canact on the basis of beliefs thatmisrepresent reality (rst-orderfalse belief). In a more complicated version of these tasks,participants have to infer the false belief of one characterabout the belief of a second character (Second-order falsebelief). Two types of false belief tasks were used in this meta-analysis:

False belief stories: this is a verbal measure of false beliefunderstanding (Frith and Corcoran, 1996). Participants mustunderstand and answer the questions based on a verbal story.A separate analysis was also conducted for rst-order andsecond-order false-belief tasks. There was some variability inthe stories chosen for different studies.

False belief picture sequencing: subjects are asked tocorrectly complete a set of pictures based on their under-standing of mental state inferences. Two different versions ofthis task were used in this meta-analysis (Brne andBodenstein, 2005; Langdon et al., 1997).

2.4.2. Hinting taskThis is a measure of indirect speech (Corcoran et al., 1995).

In this task, a series of statements are presented to participantswho are required to understand the real intended meaning.

2.4.3. Eyes testIn this task, subjects should infer complex mental states of

subjects from the pictures of their eyes (Baron-Cohen et al.,2001). Unlike false belief tasks and indirect speech tasks, thistest depends on more automatic decoding abilities ratherthan reasoning about mental states. Some authors (Bora et al.,2006; Sabbagh, 2004; Tager-Flusberg and Sullivan, 2000)suggest that this is a different kind of ToM ability (social-perceptual or mental state decoding).

Irony, faux pas, character intention inference tasks werealso used to calculate the total ToM score. However, theywere

Table 1Characteristics of studies included into the study and effect sizes for ToM tasks in these studies.

Study Sample Matched for Illness duration Medication ToM tasks Effect size

Corcoran et al., 1995 55 Sch Age, sex No information No information Hinting 0.8230 HC

Langdon et al., 1997 20 Sch Age, sex 10.3 years dur 336 mg cpz-eq FB-seq 0.7520 HC

Sarfati et al., 1997 24 Sch Age All AP+ CIT 1.3124 HC

Sarfati et al., 1999a 25 Sch 10.5 years dur 780 mg cpz-eq CIT 1.1715 HC

Sarfati and Harde-Bayle, 1999b 26 Sch Age, sex 11.5 years dur 888 mg cpz-eq CIT 1.1113 HC 25/26 AP+

Russell et al., 2000 5 Sch Sex, Edu 13 years dur All AP+ Eyes 1.417 HC

Kington et al., 2000 16 Sch Age 12 years dur 1181 mg cpz-eq Eyes 1.2926 HC All on AP+

Langdon et al., 2001 32 Sch Age, sex, edu 12.9 years dur 31/32 AP+ FB-seq 1.2324 HC 25/31 atypicals

Pickup & Frith, 2001 41 Sch 15 years dur 40/41 AP+ FB 1.6135 HC 888 mg cpz-eq

Langdon et al., 2002 25 Sch Age, sex, edu 9.6 years dur All on AP+ FB-seq 1.5420 HC 18/25 atypicals Irony 1.28

Brne, 2003 23 Sch Age 10 years dur FB 0.8612 HC

Brunet et al., 2003 25 Sch Sex, Age All on loxapine CIT 0.8825 HC

Randall et al., 2003 32 Sch Age, sex All on AP+ FB1 0.9318 HC FB2 1.58

Corcoran and Frith, 2003 59 Sch Age, sex, IQ 13.9 years dur 3/59 AP+ FB 1.4544 HC Hinting 0.99

Craig et al., 2004 16 Sch Age, sex Hinting 1.5816 HC Eyes 1.68

Zalla et al., 2004 40 Sch Age, sex, edu 16.8 years dur All AP+, FB-seq 2.040 HC Mostly atypicals

Harrington et al., 2005a,b 25 Sch Age,sex, edu 11 years dur All AP+, FB-seq 0.6538 HC 22/25 typicals FB1 0.55

FB2 0.73Marjoram et al., 2005 15 Sch Age 15 AP+ Hinting 2.13

15 HC 9/15 typicalsBrne and Bodenstein, 2005 31 Sch 11.8 years dur FB-seq 1.52

21 HC FB 1.64Kelemen et al., 2005 52 Sch Age, sex, IQ 35/52 AP+ Eyes 0.89

30 HC mostly atypicalsIrani et al., 2006 10 Sch Eyes 0.85

10 HCLangdon et al., 2006 34 Sch Age, sex, edu 12.5 years dur 32/34 AP+ FB-seq 1.41

21 HC 24/32 atypicalsPinkham and Penn, 2006 49 Sch 10.4 years dur 47/49 AP+ Hinting 0.62

44 HC 88% atypicals FB 0.62352 mg cpz-eq

Bonshtein et al., 2006 41 Sch Age 5.5 years dur All AP+ FB1 1.2522 HC FB2 1.70

Martino et al., 2007 21 Sch Age, edu All AP+ Faux pas 1.3315 HC 11/21 clozapine

Zhu et al., 2007 40 Sch Age 9.2 years dur All clozapine Faux pas 1.3431 HC

Brne et al., 2007 38 Sch Age 10 years dur FB-seq 1.029 HC FB 0.97

Bertrand et al., 2007 36 Sch First episode 32/36 AP+ Hinting 1.0927 HC 31/32 atypicals

Pousa et al., 2008 61 Sch Age, sex, IQ 11 years dur All AP+ FB-seq 0.1051 HC Mostly atypicals FB2 0.25

Kettle et al., 2008 13 Sch Edu First episode 10/13 AP+ Eyes 0.8116 HC 9/13 atypicals

Ba et al., 2008 16 Sch 9.5 years All AP+, FB1 0.7316 HC 14/16 atypicals FB2 1.92

194 mg cpz-eqMo et al., 2008 29 Sch Sex 19.3 years dur 210 mg cpz-eq Irony 1.41

22 HC

(continued on next page)

3E. Bora et al. / Schizophrenia Research 109 (2009) 19

not included in individual analyses since there were less thanve studies for these measures. Some other tasks like visualjokes, Gricean language maxims, and other ToM tasks that

clinical variables and IQ onToM impairmentwere studiedusingthe meta-regression method.

Table 1 (continued)

Study Sample Matched for Illness duration Medication ToM tasks Effect size

Tsoi et al., 2008 30 Sch Sex 17.3 years dur All AP+ FB-seq 1.1130 HC 22/30 atypicals

445 mg czp-eqCouture et al., 2008 26 Sch Age, sex, edu, b5 years dur Eyes 0.44

41 HCCorcoran et al., 2008 59 Sch 14 years dur 56/59 AP+ FB-seq 0.69

33 HC FB2 0.62Bora et al., 2008 91 Sch Sex, Edu 10.7 years dur All AP+ Eyes 0.73

55 HC Hinting 0.96

CIT: character intention inference, FB: False belief, seq: sequencing, AP = antipsychotic, dur = duration, Cpz-eq = chlorpromazine equivalent.

4 E. Bora et al. / Schizophrenia Research 109 (2009) 19used animated objects were not included in the current studysince studies that used these tasks did not meet studyinclusion criteria.

2.5. Statistical analyses

Meta-analyses were conducted with MIX software (Baxet al., 2006). We used the standardized mean differencemethod with Hedge's correction for bias in small samples.Whenever schizophrenia patients performed poorer thancontrols, we reported between-group differences as positiveeffect sizes. For studies that reportedmore than one ToM task, apooled effect size was calculated. Homogeneity of the resultingmean weighted effect sizes were tested with Q test. We used arandom effects model for the meta-analyses. Publication biaswas testedusing funnel plots and Egger's test. To reduce the riskof false positive results and to further investigate the source offunnel plot asymmetry, tasks with a signicant asymmetry(Egger's test, pb0.05) were further analyzed. Individualcharacteristics of the studies were further investigated and aFail Safe number (number of negative studies necessary tomake the group difference insignicant) was calculated. Asignicance level of pb0.05 was used for the random effectsmodel, homogeneity and publication bias analyses.

Two different methods were used to study the effect ofconfounding variables. The impact of state was studied bysubgroup analyses. We also calculated homogeneity statisticsQw and Qbet.Qbet was used to test the signicance of differencesin effect size magnitude between moderator subgroups,analogous to the F statistic. The effects of demographic and

Table 2Mean weighted effect sizes for ToM for patient-control differences.Test Study Patients Control D

Hinting 7 321 231 1.06Eyes 8 229 191 0.90FB-seq 11 395 327 1.08FB stories 11 475 363 1.06Total ToM 36 1181 936 1.10

Remitted patientsHinting 3 148 125 0.69Eyes 4 182 142 0.72FB-seq 3 124 105 0.70FB stories 6 192 174 0.70Total ToM 16 514 470 0.803. Results

Our extensive search strategy identied 73 studies thatinvestigated TOM impairment in schizophrenia. 17 of thesestudies did not include a healthy control group. Another 20were excluded since they did notmeet the inclusion criteria ofthe study (dichotomous variables, nonincluded tasks, sampleoverlap or insufcient data). The nal sample consisted of 36studies (Table 1). These 36 studies included 1181 (67% male)patients with schizophrenia and 936 (58.3% male) healthycontrol subjects. The study samples were matched for age(d=0.04, CI=0.070.15, Z=0.68, p=0.50, k=34).Patients with schizophrenia were less educated (d=0.40,CI=0.240.57, Z=04.70, pb0.0001, k=17) and had signi-cantly lower IQ than the control group (d=0.71, CI=0.550.87, Z=8.85, pb0.0001, k=28).

3.1. ToM impairment in patients

Meta-analyses of the individual ToM tasks and total ToMscore demonstrated highly signicant mentalising impair-ments in the schizophrenia group (Table 2). Effect sizes forthe between-group differences were large (0.901.08). Therewas heterogeneity for the distribution of effect sizes for FBstories (Fig. 1), FB-seq (Fig. 2) and total ToM score.Distribution of the effect sizes for the Eyes task (Fig. 3) washomogeneous and heterogeneity for the Hinting task (Fig. 4)was marginal compared to other tasks. Egger's test showedevidence of some publication bias for the Hinting, FB-seqtasks and total ToM, however fail-safe numbers of both tasks95% CI z P Q-test p Bias

0.781.34 7.3 b0.0001 0.04 0.040.641.17 6.8 b0.0001 0.19 0.150.721.43 5.9 b0.0001 b0.001 0.040.761.37 6.8 b0.0001 b0.001 0.060.951.25 14.8 b0.0001 b0.001 0.02

0.301.08 3.5 0.0004 0.15 0.240.480.95 6.0 b0.0001 0.63 0.950.011.42 1.9 0.05 0.001 0.300.401.0 4.5 b0.0001 0.09 0.050.571.03 6.4 b0.0001 0.004 0.67

Fig. 1. Forest plot of individual and pooled random effect estimates of the standardised mean differences between schizophrenia patients and controls for Falsebelief.

5E. Bora et al. / Schizophrenia Research 109 (2009) 19were quite high (221, 460 and 4450, respectively). When thecharacteristics of individual studieswere further investigated, itbecame clear that signicant correlations in Egger's testindicated there was heterogeneity of the studies rather than ale drawer effect. Sample sizes of the remitted patients werelarger compared to other studies and separate Egger's tests inremitted and non-remitted patients were not signicant.

For FB stories, additional analyses were conducted tocalculate separate effect sizes for rst-order and second-ordertasks. The patients with schizophrenia tended to be moreimpaired on second order tasks (d=1.09, CI=0.541.69)compared to rst-order tasks (d=0.87, CI=0.551.19). How-ever, this difference was not signicant according to Qbet

statistic (p=0.75). Finally, a subgroup analysis was conducted

Fig. 2. Forest plot of individual and pooled random effect estimates of the standardisto investigate the effect of antipsychotics on ToM impairment.Since, many studies did not report details of antipsychotictreatment, there was only one possible subgroup analysis(studies that reported that their patients were predominantlyusing atypicals). The effect size of the ToM impairment forthese studies (11 studies) was very similar to whole sample(d=1.08, CI=0.751.41).

3.2. ToM decits in remitted patients

Only few of the studies reported data from remittedpatients according to stringent criteria. Therefore, similar toKrabbendam et al. (2005), we dened remission as being

outpatients and/or inpatients just before discharge. Meta-

ed mean differences between schizophrenia patients and controls for FB-seq.

Fig. 3. Forest plot of individual and pooled random effect estimates of the standardisedmean differences between schizophrenia patients and controls for Eyes test.

6 E. Bora et al. / Schizophrenia Research 109 (2009) 19analyses of the ToM tasks in remitted patients revealedsignicant impairments compared to healthy controls(Table 2).While themagnitudes of these decits were smallerthan the results of total sample analyses, effect sizes were stillin the medium-large range (0.690.72). Effect size for TotalToM was slightly larger than effect sizes of the individualtasks. This difference reected the fact that remittedpatients had more severe impairments on irony and fauxpas tasks that were not analyzed individually.

Effect size distributions of three of the four ToM measureswere homogeneous in remitted patients. The only task thatshowed heterogeneity for distribution of its effect sizes wasFB-seq. There was also signicant heterogeneity for the totalToM score. There was no statistical evidence for publicationbias in remitted patients.

3.3. The impact of acute symptoms on ToM impairment

Effect size for the ToM impairment (d=1.21, CI=1.051.37, Z=14.42, pb0.0001, df=20) was larger than the decitreported for remitted patients (d=0.80). This difference

was highly signicant (Qbet=19.65, pb0.0001, df=1) showing

Fig. 4. Forest plot of individual and pooled random effect estimates of the standardistask.that state is an important moderator for ToM impairment inschizophrenia. However, there should remain some otherfactors contributing to the between-studies heterogeneity(i.e. heterogeneity between tasks, as above), since pooled Qwwas also signicant (Qw=68.9, p=0.0006, df=35).

3.4. Meta-regression analyses

Group differences in age (34 studies), sex (33 studies) andeducation (17 studies) had no impact on themagnitude of ToMimpairments. There was no effect of age at onset of illness (28studies), however, patients with longer duration of illnesstended to be more impaired on ToM tasks (k=28 studies,B=0.03, SE=0.01, Z=1.85, p=0.06). Antipsychotic dose(chlorpromazine equivalents) was not associated with ToMperformance, however only 10 studies reported data to calculatethis measure. The magnitude of the general intelligence decittended to increase the effect size of the ToM decit (k=28,B=0.35 SE=0.19 Z=1.87 p=0.06) in patients with schizo-phrenia. This association became signicant when the analysiswas restricted to remitted patients (k=13, B=071, SE=0.30

Z=2.40 p=0.01).

ed mean differences between schizophrenia patients and controls for Hinting

7E. Bora et al. / Schizophrenia Research 109 (2009) 193.5. First-episode schizophrenia and prodrome

Since only two of the studies (Bertrand et al., 2007; Kettleet al., 2008) included in the meta-analysis were rst-episodesamples, we were not able to calculate a summary score forthese studies. Both of these studies showed signicantimpairments (effects sizes 0.81 and 1.09) in the patientgroups. Together with the results of another study (Inoueet al., 2006) that was not included in the meta-analysis (sinceit reported non-parametric data); these preliminary resultssuggest ToM impairment is comparable to chronic patients.

To our knowledge, only two studies compared ToMperformances of subjects in prodrome phase with healthycontrols. While one of these studies reported impairments forstrange stories and false belief tests (Chung et al., 2008), thesecond study did not nd any signicant difference for theEyes test (Couture et al., 2008).

4. Discussion

Schizophrenia patients had mentalising impairments withlarge effect sizes observed for all tasks. The distribution ofeffect sizes was more homogeneous for individual tasks ofToM, especially in remitted patients. One of the mainndings of this study was that ToM impairment was stronglyinuenced by acute psychosis. Overall sample had prominentimpairment on all ToM tasks while remitted patients hadreduced but still signicant, ToM impairment. Contrastingwith the earlier views, this result suggests that ToM decitsmight be trait impairments. However, there is still thepossibility that ToM impairment in remitted patients maybe secondary to other factors like cognitive decits or residualsymptoms. Consistentwith this idea, in our study, impairementof general intelligence was signicantly contributing to ToMimpairment only in remitted patients.

Results of this meta-analysis showed that ToM impair-ments occur with large effect sizes in schizophrenia. In aprevious study, Sprong et al. (2007) also identied signicantToM impairment in patients with schizophrenia. However,their results indicated a more severe impairment than ourstudy. The current meta-analysis included more patients andthis might be one of the factors explaining the discrepancy.However, several methodological differences between thestudies are also likely to contribute to different results. InSprong et al. (2007) included overlapping samples in theiranalysis that may further reduce the actual sample size. Theyalso used the probit dmethod to estimate the effect sizes fromdichotomous variables. Since a skewed distribution of ToMtasks seems to be characteristic of these studies (in manycases, the control group scored 100% true), this estimationmethod is likely to give unreliable results. In Sprong et al.(2007) inclusion of multiple studies from overlappingsamples can also cause a bias in the mean effect size. Forthese reasons, we believe that the results of our study mayprovide a better estimation of actual effect sizes of ToMimpairment in patients with schizophrenia.

The current meta-analyses also examined two individualtasks of ToM that have been commonly used in the literature(i.e., Hinting and Eyes). We attempted to integrate differenttasks that assessed subjects' understanding of false-beliefs bystory comprehension (FB stories) or sequencing (FB-seq) inorder to increase homogeneity of the tasks. The estimated effectsizes for three of the tasks (Hinting, FB seq and FB stories) andtotal ToM score was very similar (1.061.08). The distributionsof effect sizes were much less heterogeneous for individualtasks compared to combined tasks and total ToM score. Thisresult supports the idea that heterogeneity of themethods usedto assess ToM abilities contributes to the inconsistencies in thereported ndings. This study also provided evidence for adifferent aspect of ToMability that ismeasuredby the Eyes task.The effect size for the impairment was moderately smaller forthe Eyes task compared to other tasks.

It is evident that ToM impairment is more severe in theacute phase of schizophrenia and differences in symptomcharacteristics of the various studiesmay therefore contributeto the heterogeneity of the reported ndings. However, ourresults suggest that ToM impairments in schizophreniapersist after the remission of acute psychosis. This resultcontradicts the earlier opinions of Frith (1992) and HardeBayle (Sarfati et al., 1997). Frith's concept of ToM decit inschizophrenia suggests a state related impairment. Thismodel proposed that psychotic symptoms in schizophreniamight be explained by mentalising impairment. Consistentwith this model, several studies of Frith and his colleaguesshowed intact ToM performance in remitted patients. Whilethe model of HardeBayle suggests an association of ToMimpairment with another symptom dimension of schizo-phrenia (disorganized thought), it also considers ToM decitas a state characteristic of schizophrenia. However, otherndings demonstrating ToM impairment in remitted patientsand in people at genetic risk of schizophrenia supports thenotion that ToM dysfunction may be a trait characteristic ofschizophrenia. Results of the current meta-analysis and theresults of Sprong et al also are consistent with this opinion.Insufcient power related to small sample size of somestudies may explain some of the previous reported negativendings. Differences in ToM batteries used and in criteria forremission may also explain the divergent ndings. Denitionof remission is clearly important to the proper interpretationof the ndings since residual positive and persistent negativesymptoms are commonly observed in stable patients withschizophrenia. Two recent studies investigated the inuenceof residual symptoms during remission. Pousa et al. (2008)reported a mentalising decit in patients with residualpositive symptoms but not in patients without such symp-toms. Bora et al. (2008) also demonstrated a more severe ToMimpairment in patients with residual positive and negativesymptoms. However, this study also showed a less pro-nounced but still signicant mentalising decit (ES=0.50) insymptom free patients.

The nding related to the effect of IQ decits on ToMimpairment in schizophrenia is one of the important resultsof this meta-analysis. A general and selective cognitive decitobserved in patients with schizophrenia is another importantfactor that can contribute to ToM impairment in symptomfree patients. However, previous studies gave inconsistentresults regarding the effect of non-ToM cognitive and generalintelligence decits on ToM impairment in schizophrenia. IQ,executive functions and memory abilities seem to becorrelated with ToM performances of schizophrenia patients,however previous reviews suggested that cognitive decitsand IQ cannot explain the ToM impairment in schizophrenia

8 E. Bora et al. / Schizophrenia Research 109 (2009) 19(Brne, 2005; Harrington et al., 2005a,b). Meta-analysis ofSprong et al. (2007) also reported that IQ did not affect meaneffect sizes for ToM impairment. However, our study showeda signicant inuence of IQ impairment on ToM decits whenthe regression analysis was restricted to remitted patients.In acute phases of the illness the impact of cognitive decitson ToM impairment may be masked by the relationshipbetween symptoms and mentalising ability. However, in thestable phases of illness general intellectual impairment andother cognitive decits can mediate the observed impair-ments of ToM ability. Consistent with this idea, the results ofBora et al. (2008) suggested that ToM impairment insymptom free patients could be explained by workingmemory decits of these patients. A recent study in rst-degree relatives of patients with schizophrenia also showedToM impairment that did not persist after correction for IQ(Pentaraki et al., 2008). These results suggest that, beforeconcluding that there is a trait related ToM impairment inschizophrenia, more careful research is needed to investigateToM abilities and the inuence of other cognitive decits insymptom free patients.

One limitation of the current study is the lack of availabledata for more detailed meta-regression analyses. One exam-ple for this is insufcient cognitive data that prevent us fromconducting further analyses. Besides IQ, other cognitiveabilities like executive functions and verbal memory impair-ment can inuence ToM impairment in schizophrenia.Unfortunately, we were only able to examine the effects ofIQ in the present meta-analysis. Second, while we didindividual analyses for the Hinting task and the Eyes task,therewas still between-studies task heterogeneity for the twofalse belief tasks included in the current meta-analysis.Finally, there was no sufcient data to calculate effect sizesfor the ToM impairment in rst-episode patients. This is alimitation since studies that investigate ToM impairment inearly phases and also before the onset of illness (i.e., prodrome)are important to understand the nature of ToM dysfunction inschizophrenia.

Our results lead to several recommendations for futurestudies. One important issue is to further test the role of ToMimpairment as a trait characteristic of schizophrenia. Futurework should carefully control the inuence of residualsymptoms and, importantly, the effect of other cognitiveimpairments in patients with schizophrenia. Studies in rst-degree relatives of patients and longitudinal studies in peopleat high risk for developing schizophrenia will also beimportant in understanding whether the ToM impairmentsin schizophrenia are enduring features of the disorder. Finally,another important step is to use ToM tasks with betterpsychometric properties.

Role of funding sourceNone.

ContributorsEmre Bora designed the study, collected and analysed the data and wrote

the rst draft of the manuscript. Murat Yucel and Christos Panteliscontributed to design of the meta-analysis. All authors contributed to andapproved the nal draft of the paper.

Conict of interestAll authors report no conict of interest.AcknowledgementMurat Ycel was supported by a National Health & Medical Research

Council (NH&MRC) Clinical Career Development Award (I.D. 509345).NHMRC.

References

Ba, M., Zanello, A., Varnier, M., Koelner, V., Merlo, M.C.G., 2008. Decits inneurocognition, theory of mind and social functioning in patients withschizophrenic disorders. Are they related? J. Nerv. Ment. Dis. 196,153156.

Baron-Cohen, S.,Wheelwright, S., Hill, J., Raste, Y., Plumb, I., 2001. The readingthemind in the eyes test revised version: a studywith normal adults, andadults with Asperger syndrome or high-functioning autism. J. ChildPsychol. Psychiatry 42, 241251.

Bax, L., Yu, L.M., Ikeda, N., Tsuruta, H., Moons, K.G.M., 2006. Development andvalidation of MIX: comprehensive free software for meta-analysis ofcasual research data. BMC Med. Res. Methodol. 6, 50.

Bertrand, M.C., Sutton, H., Achim, A.M., Malla, A.K., Lepage, M., 2007. Socialcognitive impairments in rst-episode psychosis. Schizophr. Res. 95,124133.

Bonshtein, U., Leiser, D., Levine, J., 2006. Naive theory impairment inschizophrenia. Is it domain specic? J. Nerv. Ment. Dis. 194, 753759.

Bora, E., Eryavuz, A., Kayahan, B., Sungu, G., Veznedaroglu, B., 2006. Socialfunctioning, theory of mind and neurocognition in outpatients withschizophrenia: mental state decoding may be a better predictor ofsocial functioning than mental state reasoning. Psychiatry Res. 145,95103.

Bora, E., Gokcen, S., Kayahan, B., Veznedaroglu, B., 2008. Decits of social-cognitive and social-perceptual aspects of theory of mind in remittedpatients with schizophrenia. Effect of residual symptoms. J. Nerv. Ment.Dis. 196, 9599.

Brunet, E., Sarfati, Y., Harde-Bayle, M.C., 2003. Reasoning about physicalcausality and other's intentions in schizophrenia. Cogn. Neuropsychiatry8, 129139.

Brne, M., 2003. Theory of mind and the role of IQ in chronic disorganizedschizophrenia. Schizophr. Res. 60, 5764.

Brne, M., 2005. Theory of mind in schizophrenia: a review of the literature.Schizophr. Bull. 31, 2142.

Brne, M., Bodenstein, L., 2005. Proverb comprehension reconsidered theory of mind and the pragmatic use of language in schizophrenia.Schizophr. Res. 75, 233239.

Brne, M., Abdel-Hamid, M., Lehmkamper, C., Sonntag, C., 2007. Mental stateattribution, neurocognitive functioning, and psychopathology: whatpredicts social competence in schizophrenia best? Schizophr. Res. 92,151159.

Chung, Y.S., Kang, D.H., Shin, N.Y., Yoo, S.Y., Kwon, J.S., 2008. Decit of theoryof mind in individuals at ultra-high-risk for schizophrenia. Schizophr.Res. 99, 111118.

Corcoran, R., Frith, C.D., 2003. Autobiographical memory and theory of mind:evidence of a relationship in schizophrenia. Psychol. Med. 33, 897905.

Corcoran, R., Mercer, G., Frith, C.D., 1995. Schizophrenia, symptomatology andsocial inference: investigating theory of mind in people with schizo-phrenia. Schizophr. Res. 17, 513.

Corcoran, R., Rowse, G., Moore, R., Blackwood, N., Kinderman, P., Howard, R.,Cummins, S., Bentall, R.P., 2008. A transdiagnostic investigation of theoryof mind and jumping to conclusions in patients with persecutorydelusions. Psychol. Med. 38, 15771583.

Couture, S.M., Penn, D.L., Addington, J., Woods, S.W., Perkins, D.O., 2008.Assessment of social judgements and complex mental states in earlyphases of psychosis. Schizophr. Res. 100, 237241.

Craig, J.S., Hatton, C., Craig, F.B., Bentall, R.P., 2004. Persecutory beliefs,attributions and theory of mind: comparison of patients with paranoiddelusions, Asperger's syndrome and healthy controls. Schizophr. Res. 69,2933.

Frith, C.D., 1992. The cognitive neuropsychology of schizophrenia. LawrenceErbaum Associates, Hove.

Frith, C.D., Corcoran, R., 1996. Exploring theory of mind in people withschizophrenia. Exploring theory of mind in people with schizophrenia26, 521530.

Harrington, L., Langdon, R., Siegert, R., McClure, J., 2005a. Schizophrenia,theory of mind and persecutory delusions. Cogn. Neuropsyhiatry 10,87104.

Harrington, L., Siegert, R.J., McClure, J., 2005b. Theory of mind inschizophrenia: a critical review. Cogn. Neuropsychiatry 10, 249286.

Inoue, Y., Yamada, K., Hirano, M., Shinohara, M., Tamaoki, T., Iguchi, H.,Tonooka, Y., Kanba, S., 2006. Impairment of theory of mind in patients inremission following rst episode of schizophrenia. Eur. Arch. PsychiatryClin. Neurosci. 256, 326328.

Irani, F., Platek, S.M., Panyavin, I.S., Calkins, M.E., Kohler, C., Siegel, S.J.,Schachter, M., Gur, R.E., Gur, R.C., 2006. Self-face recognition and theoryof mind in patients with schizophrenia and rst-degree relatives.Schizophr. Res. 88, 151160.

Kelemen, O., Erdelyi, R., Pataki, I., Benedek, G., Janka, Z., Keri, S., 2005. Theory ofmindandmotionperception in schizophrenia.Neuropsychology19,494500.

Kettle, J.W.L., O'Brien-Simpson, L., Allen, N.B., 2008. Impaired theory of mindin rst-episode schizophrenia: comparison with community, universityand depressed controls. Schizophr. Res. 99, 96102.

Kington, J.M., Jones, L.A., Watt, A.A., Hopkin, E.J., Williams, J., 2000. Impairedeye expression recognition in schizophrenia. J. Psychiatr. Res. 34,341347.

Krabbendam, L., Arts, B., Van Os, J., Aleman, A., 2005. Cognitive functioning inpatients with schizophrenia and bipolar disorder: a quantitative review.Schizophr. Res. 80, 137149.

Langdon, R., Michie, P.T., Ward, P.B., McConaghy, N., Catts, S.V., Coltheart, M.,1997. Defective self and/or other mentalising in schizophrenia: acognitive neuropsychological approach. Cogn. Neuropsychiatry 2,167193.

Langdon, R., Coltheart, M., Ward, P.B., Catts, S.V., 2001. Visual and cognitiveperspective-taking impairments in schizophrenia: a failure of allocentricstimulation. Cognit. Neuropsychiatry 6, 241269.

Langdon, R., Coltheart, M., Ward, P.B., Catts, S.V., 2002. Disturbed commu-nication in schizophrenia: the role of poor pragmatics and poor mind-reading. Psychol. Med. 32, 12731284.

Langdon, R., Corner, T., McLaren, J., Ward, P.B., Coltheart, M., 2006.Externalizing and personalizing biases in persecutory delusions: therelationship with poor insight and theory of mind. Behav. Res. Ther. 44,699713.

Marjoram, D., Gardner, C., Burns, J., Miller, P., Lawrie, S.M., Johnstone, E.C.,2005. Symptomatology and social inference: a theory of mind study ofschizophrenia and psychotic affective disorder. Cogn. Neuropsyhiatry 10,347359.

Martino, D.J., Bucay, D., Butman, J.T., Allegri, R.F., 2007. Neuropsychologicalfrontal impairments and negative symptoms in schizophrenia. Psychia-try Res. 152, 121128.

Mo, S., Su, Y., Chan, R.C.K., Liu, J., 2008. Comprehension of metaphor and irony

Murray, R., 2008. Theory of mind as a potential endophenotype ofschizophrenia: understanding false beliefs in families with schizophre-nia. Schizophr. Res. 98, 103 (Suppl).

Pickup, G.J., Frith, C.D., 2001. Theory of mind impairments in schizophrenia:symptomatology, severity and specicity. Psychol. Med. 31, 207220.

Pinkham, A.E., Penn, D.L., 2006. Neurocognitive and social cognitivepredictors of interpersonal skill in schizophrenia. Psychiatry Res. 143,167178.

Pousa, E., Duno, R., Brebion, G., David, A.S., Ruiz, A.I., Obiols, J.E., 2008. Theoryof mind decits in chronic schizophrenia: evidence for state dependence.Psychiatry Res. 158, 110.

Randall, F., Corcoran, R., Day, J., Bentall, R., 2003. Attention, theory of mindand casual attributions in people with persecutory delusions: apreliminary investigation. Cognit. Neuropsychiatry 8, 287294.

Russell, T.A., Rubia, K., Bullmore, E.T., Soni, W., Suckling, J., Brammer, M.J.,Simmons, A., Williams, S.C.R., Sharma, T., 2000. Exploring the social brainin schizophrenia: left prefrontal underactivation during mental stateattribution. Am. J. Psychiatry 157, 20402042.

Sabbagh, M.A., 2004. Understanding the orbitofrontal contributions to theoryof mind reasoning: implications for autism. Brain Cogn. 55, 209219.

Sarfati, Y., Harde-Bayle, M.C., Besche, C., Widlocher, D., 1997. Attribution ofintentions to others inpeoplewith schizophrenia: a non-verbal explorationwith comic stips. Schizophr. Res. 25, 199209.

Sarfati, Y., Harde-Bayle, M.C., Brunet, E., Widlocher, D., 1999a. Investigatingtheory of mind in schizophrenia: inuence of verbalization in disorga-nized and non-disorganized patients. Schizophr. Res. 37, 183190.

Sarfati, Y., Harde-Bayle, M.C., 1999b. How do people with schizophreniaexplain the behaviour of others. Psychol. Med. 29, 613620.

Sprong, M., Schothorst, P., Vos, E., Hox, J., Van Engeland, H., 2007. Theory ofmind in schizophrenia: meta-analysis. Br. J. Psychiatry 191, 513.

Tager-Flusberg, H., Sullivan, K., 2000. A componential view of theory of mind:evidence from syndrome. Cognition 76, 5990.

Tsoi, D.T.Y., Lee, K.H., Gee, K.A., Holden, K.R., Parks, R.W., Woodruff, P.W.R.,2008. Humour experience in schizophrenia: relationship with executivefunction and psychosocial impairment. Psychol. Med. 38, 801810.

Zalla, T., Bouchilloux, N., Labruyere, N., Georgieff, N., Bougerol, T., Franck, N.,2004. Impairment in event sequencing in disorganised and non-disorganised patients with schizophrenia. Brain Res. Bull. 68, 195202.

9E. Bora et al. / Schizophrenia Research 109 (2009) 19Psychiatry Res. 157, 2129.Pentaraki, A.D., Stefanis, N.K., Stahl, D., Kaliora, S.C., Roukas, D., Theleritis, C.,

Chatzimanolis, I., Smyrnis, N., Russell, T., Kravariti, E., Toulopoulou, T.,Zhu, C.Y., Lee, T.M.C., Li, X.S., Jing, S.C., Wang, Y.G., Wang, K., 2007.Impairments of social cues recognition and social functioning in Chinesepeople with schizophrenia. Psychiatry Clin. Neurosci. 61, 149158.in schizophrenia during remission: the role of theory of mind and IQ.

Theory of mind impairment in schizophrenia: Meta-analysisIntroductionMethodStudy selectionInclusion criteriaCoding of the variablesMentalising tasksFalse-belief tasksHinting taskEyes test

Statistical analyses

ResultsToM impairment in patientsToM deficits in remitted patientsThe impact of acute symptoms on ToM impairmentMeta-regression analysesFirst-episode schizophrenia and prodrome

DiscussionRole of funding sourceContributorsConflict of interestAcknowledgementReferences