Mentalising impairment (an impaired ability to think about people in terms of their mental states) has frequently been associated with schizophrenia.

To assess the magnitude of the deficit and analyse associated factors.

Method
Twenty-nine studies of mentalising in schizophrenia (combined n=1518), published between January 1993 and May 2006, were included to estimate overall effect size. Study descriptors predicted to influence effect size were analysed using weighted regression-analysis techniques. Separate analyses were performed for symptom subgroups and task types.

Results
The estimated overall effect size was large and statistically significant (d=–1.255, P<0.0001) and was not significantly affected by sample characteristics. All symptom subgroups showed significant mentalising impairment, but participants with symptoms of disorganisation were significantly more impaired than the other subgroups (P<0.01).

Conclusions
This meta-analysis showed significant and stable mentalising impairment in schizophrenia. The finding that patients in remission are also impaired favours the notion that mentalising impairment represents a possible trait marker of schizophrenia.

INTRODUCTION
‘Theory of mind’ and ‘mentalising’ refer to the cognitive ability to attribute mental states such as thoughts, beliefs and intentions to people, allowing an individual to explain, manipulate and predict behaviour. In 1992 Frith proposed a relationship between theory of mind and schizophrenia, and argued that several symptoms of schizophrenia could be explained by mentalising impairment (Frith, 1992). This led to a substantial body of research which has recently been critically reviewed twice (Brüne, 2005a; Harrington et al, 2005a). In both reviews it was concluded that theory of mind is impaired in individuals with schizophrenia. Although these reviews were executed thoroughly, they are limited to a qualitative description of the observed deficit, thus lacking important information on the magnitude of the effect. The purpose of this meta-analysis is to produce a synthesised effect size estimate that has considerably more power than the individual studies. In addition, effects of study characteristics on the findings are analysed.

DISCUSSION
The aim of this meta-analysis was to investigate the extent of mentalising impairment in people with schizophrenia. By combining 29 studies, a total sample size was created of over 1500 participants. The overall effect size was –1.1255, indicating that on average the theory of mind performance of participants with schizophrenia is more than one standard deviation below that of healthy controls. According to a widely used convention for appraising the magnitude of effect sizes this is considered a large effect (Cohen, 1988). Homogeneity analysis showed that the mean effect size of the combined samples is a good estimate of the typical effect size in the population. The large fail-safe number makes the `file drawer’ problem, which is a limitation of some meta-analyses, negligible.

The moderator variables IQ, gender and age did not significantly affect mean effect size. Thus, the impairment in theory of mind is robust and is not readily moderated by variables that may seem relevant. However, the effect of other potentially important moderator variables such as medication use and duration and severity of illness could not be analysed owing to a lack of information on these characteristics in many studies.

Participants with schizophrenia who had signs and symptoms of disorganisation were found to be significantly more impaired in terms of theory of mind than those in the other symptom subgroups. However, these results may also be explained by the composition of the disorganised symptom subgroup. The behavioural subgroup of the studies by Frith and colleagues was ranked highest in their hierarchical model. Thus, individuals in this group might also have had symptoms of the paranoid and/or passivity subgroup. This brings the risk that poorer performance in this group may be explained by having more severe and complex symptoms (Harrington et al, 2005a). Similarly, in two of the four studies by Sarfati and colleagues the disorganised subgroup had more general psychopathology, which might explain their poorer theory of mind performance (Sarfati & Hardy-Bayle’, 1999; Sarfati et al, 1999).

The mean effect size (d=–0.692) of mentalising impairment in patients in remission was smaller than in the other symptom subgroups, but is still considered to be medium to large (Cohen, 1988). Moreover, this effect did not differ significantly from the effect sizes of the disorganised and paranoid subgroups.

Unexpectedly – and despite apparent differences in type and difficulty of the theory of mind tasks – the mean effect sizes for different task types were found to be similar. An explanation might be that our method of grouping studies by task types was not correct. This is supported by the finding that two of the four task type analyses showed heterogeneity among effect sizes. However, since there is a lack of research on the psychometric properties of the tasks that were used, such as construct and concurrent validity, it is not yet possible to categorise these tasks objectively.

There was also no difference between the mean effect sizes of verbal and non-verbal tasks, which is consistent with the findings of Sarfati and colleagues (Sarfati et al, 1999, 2000). Thus, impairment of theory of mind does not to appear to be affected by verbalisation deficits that have been reported in people with schizophrenia.

Mirjam Sprong, MA, Patricia Schothorst, MD, PhD and Ellen Vos, MD

Department of Child and Adolescent Psychiatry, University Medical Centre Utrecht

Joop Hox, PhD
Faculty of Social Sciences, Department of Methodology and Statistics, Utrecht University

Herman Van Engeland, MD, PhD
Department of Child and Adolescent Psychiatry, University Medical Centre, The Netherlands

Correspondence: Ms M. Sprong, Department of Child and Adolescent Psychiatry, University Medical Centre, Heidelberglaan 100, HPA01.468, 3508 GA Utrecht, The Netherlands. Email: .(JavaScript must be enabled to view this email address)

Full text