Theory of mind is an aspect of social cognition that describes the ability to infer other people's mental state. A recent meta-analysis of almost 30 studies reported a large effect size for theory of mind impairment in schizophrenia. ^{Reference Sprong, Schothorst, Vos, Hox and van Engeland1} Investigations of these deficits have employed tasks that index false beliefs/deception (reasoning about the mental state of others), indirect speech (understanding irony, hints and ‘faux pas’), and mental state decoding (using information such as facial expression to infer mental state). Differences between tasks (at both behavioural and neuroanatomical levels) have resulted in a distinction between mental state reasoning ability and mental state decoding ability. ^{Reference Shamay-Tsoory, Shur, Barcai-Goodman, Medlovich, Harari and Levkovitz2,Reference Bora, Eryavuz, Kayahan, Sungu and Veznedaroglu3}

Following evidence linking affective recognition and social function, ^{Reference Addington, Saeedi and Addington4} Bora et al ^{Reference Bora, Eryavuz, Kayahan, Sungu and Veznedaroglu3} suggested that mental state decoding may be more important than mental state reasoning for social outcome in schizophrenia. They argue that this was due to mental state decoding tasks being based on more spontaneous/automatic inferential processing than theory of mind tasks involving effortful verbal processing. Furthermore, they suggest that decoding affective states from facial expressions (e.g. eyes) is likely to be closely related to empathy and the neural network underpinning empathy (inferior frontal/anterior temporal lobe function); ^{Reference Sabbagh, Moulson and Harkness5} empathy has previously been shown to be important to social function in schizophrenia. ^{Reference Shamay-Tsoory, Shur, Harari and Levkovitz6} Here we aimed to test whether mental state reasoning and mental state decoding ability differ in their ability to: (a) predict social function; and (b) mediate the relationship between neuropsychological and social function in chronic schizophrenia.

Method

After we received written informed consent, we assessed 73 patients (49 male) from a suburban Dublin psychiatric out-patient clinic using the Structured Clinical Interview for DSM–IV disorders ^{Reference First, Spitzer, Gibbon and Williams7} to confirm diagnosis of schizophrenia or schizoaffective disorder. Criteria for participation included being aged 16–65 years, having no history of intellectual disability, acquired brain injury resulting in loss of consciousness or substance misuse in the preceding 6 months (all based on chart review). Mean age was 41.4 years, mean duration of illness was 18.2 years, and almost all were prescribed atypical antipsychotic medication. Patients' theory of mind and neuropsychological performance was compared with the normal population using a sample of 78 age- and gender-matched healthy comparison participants recruited through the local media who satisfied the inclusion criteria and were free of psychiatric illness based on clinical interview (online Table DS1).

Mental state decoding ability was measured using the revised version of the ‘Reading the Mind in the Eyes’ test (‘Eyes task’). ^{Reference Baron-Cohen, Wheelwright, Hill, Raste and Plumb8} Participants were required to study 36 photographs of facial expressions and correctly identify the thought/feeling being portrayed based on four descriptors printed around each expression. Mental state reasoning ability was assessed using the ‘Hinting task’. ^{Reference Corcoran, Mercer and Frith9} This consisted of ten vignettes, each describing a social interaction between two characters, after which the participant is required to make inferences about the intent behind a hint dropped by one of the characters.

Neuropsychological function was assessed using tests from the Wechsler Adult Intelligence Scale ^{Reference Wechsler10} and the Cambridge Neuropsychological Test Automated Battery (CANTAB; Cambridge Cognition Ltd, Cambridge, UK), selected to index the general and specific domains of impairment commonly reported in schizophrenia, namely general cognitive function, attentional control, episodic memory and working memory (online Table DS1). ^{Reference Donohoe, Clarke, Morris, Nangle, Schwaiger, Gill, Corvin and Robertson11} Symptom severity was assessed using the Scale for the Assessment of Positive Symptoms ^{Reference Andreasen12} and the Scale for Assessment of Negative Symptoms. ^{Reference Andreasen13} Social functioning was measured using the problem-solving factor from the Independent Living Scale (ILS). ^{Reference Loeb14} Of the two factors yielded by the ILS (problem-solving and performance/information), the problem-solving factor (which includes items such as ‘what would you do if the taxi you ordered didn't come and it was getting late’), is reported to have good utility as a proxy measure for real-world functioning in schizophrenia. ^{Reference Revheim and Medalia15}

Results

Eyes task and Hinting task performance were significantly correlated (r=0.389; P=0.001). Despite this, patients only performed significantly below controls on the Eyes task (t=–2.3; d.f.=149; P=0.023). Eyes task performance correlated with verbal and performance IQ, and verbal and spatial working memory (P=0.003–0.000006) but not with verbal episodic memory. Hinting task performance by contrast was correlated with verbal episodic and working memory (P=0.008–0.02) but not IQ or spatial memory performance. Neither task correlated with attention control. Eyes task performance alone was moderately negatively correlated with positive symptom severity (r=–0.25; P=0.03) but not negative symptom severity. Neither task was associated with age, gender, duration of illness or medication dosage.

Scores on the ILS were significantly positively associated with Eyes task performance (r=0.41; P=0.001) but not with Hinting task performance (P>0.05). To test whether social functioning was predicted by mental state decoding performance (Eyes task performance) we performed a multiple regression analysis with ILS scores as the dependent variable. Symptom severity, which was significantly correlated with social function, was entered in the first step of the analysis, explaining 26.5% of variance in ILS scores (20% of which was contributed by negative symptom severity alone). This was followed by scores on the Eyes task and Hinting tasks on the second step. Eyes task performance explained a further 9.0% of the variance in social functioning (F _1,63 change=8.5; P=0.005); Hinting task performance was not a significant variable in the regression equation. The statistical significance of these results was unchanged by considering patients with schizophrenia and patients with schizoaffective disorder separately.

We performed two further stepwise multiple regression analyses. In the first analysis, when verbal IQ (the neuropsychological measure most highly correlated with ILS performance) was entered on the first step of the analysis followed by Eyes task performance on the second step, Eyes task performance continued to significantly predict variance in ILS (r ² change=0.09, F _1,68=7.26, P=0.009). In the second regression analysis, when Eyes task performance was entered on the first step and verbal IQ on the second, verbal IQ's ability to predict ILS scores was partly attenuated (r ² change reducing from 16.3% to 7.5%). Following Baron & Kenny's criteria ^{Reference Baron and Kenny16} for mediating variables this suggests that Eyes task performance partly mediated the relationship between verbal IQ and ILS scores. Again, the statistical significance of these results was unchanged by considering patients with schizophrenia and patients with schizoaffective disorder separately.

Discussion

This study provides further evidence that mental state decoding rather than mental state reasoning ability predicts social function in patients with chronic schizophrenia. Bora et al ^{Reference Bora, Eryavuz, Kayahan, Sungu and Veznedaroglu3} argue that the mental state decoding requires emotional perception and empathy. Perceptual deficits in social cognition (e.g. affect recognition) are reported as stable over time in schizophrenia, apparent in both patients with first-episode schizophrenia and those who are chronically ill, and mediate the relationship between basic cognition and social function. ^{Reference Addington, Saeedi and Addington4} Our findings suggest that decoding the mental state of others is similarly important to social function in schizophrenia. This view is consistent with the model of emotional intelligence developed by Mayer & Salovey ^{Reference Mayer, Salovey, Salovey and Sluyter17} in which emotional perception and emotional understanding are two of the four main abilities related to ‘emotional intelligence’ (together with emotional self-regulation and emotional facilitation of thinking), which contributes to higher social functioning.

Because of the general decline in cognitive function associated with schizophrenia, a challenge for studies of specific cognitive functions (either basic or social) is to establish whether any specific deficit has a unique or particular value in explaining outcome. In the regression analyses undertaken, mental state decoding performance (on the Eyes task) was able to explain variance in social function even after the variance explained by general cognitive ability (measured by verbal IQ) was already accounted for. Furthermore, accounting for the variance in mental state decoding led to attenuation in the amount of variance explained by general cognition. These results lead us to conclude that mental state decoding is not simply reflecting more general aspects of cognitive decline in schizophrenia, but instead is important in its own right in explaining social function. As further confirmation of these findings, and the earlier report by Bora et al, ^{Reference Bora, Eryavuz, Kayahan, Sungu and Veznedaroglu3} it may be helpful to investigate this relationship between mental state decoding and social function using multiple mental state decoding tasks within the same study design. Finally, for future studies, the ILS measure employed here, although previously validated in schizophrenia, ^{Reference Revheim and Medalia15} is a proxy measure for social function and outcome; further theory of mind studies may benefit from more direct outcome measures.

This research was supported by research grants from the Wellcome Trust (M.G., A.C.), Science Foundation Ireland (G.D.) and the Higher Education Authority of Ireland (M.G.).