To test if specific correlations exist between cognitive measures and psychotic dimensions in schizophrenic subjects and if similar correlations, between cognition and schizotypal dimensions, are present in non-psychotic subjects.

We administered the same battery of cognitive tests (Source Monitoring, Verbal Fluency [VF] and Stroop tests) to schizophrenic subjects (N = 54), their first-degree relatives (N = 37) and controls (N = 41). Scores of negative, positive and disorganisation dimensions were derived from the Signs and Symptoms of Psychotic Illness scale in schizophrenic subjects, and from the Schizotypal Personality Questionnaire in relatives and controls.

In schizophrenic subjects, as hypothesised, the negative dimension correlated with performance on VF and disorganisation with performance in the Stroop test. The positive dimension did not correlate with any cognitive measure.

With only one exception, the significant correlations observed in non-psychotic subjects did not match correlations seen in schizophrenic subjects. In non-psychotic subjects greater disorganisation was associated with more clustered words in VF suggesting that excessive automatic spreading of activation in semantic networks could underlie this dimension.

As a whole, data lent partial support to our hypothesis of specific cognitive–clinical correlations in schizophrenic subjects but did not support the existence of similar correlations in non-psychotic subjects.

Although many patients with schizophrenia are impaired in mental states attribution abilities, a significant number perform within normal or near-normal ranges in mental state attribution tasks. No studies have analysed cognitive or behavioural differences between patients with – to some extent – preserved mental state attribution skills and those with poor mentalising abilities.

To examine characteristics of “poor” and “fair” mentalisers, 58 patients with schizophrenia performed a mental state attribution task, a test of general intelligence, and two executive functioning tests. “Poor” and “fair” mentalising skills were defined according to a median-split procedure; the median score in the patient group was also within two standard deviations of the control group. In addition, patients’ social behavioural skills and psychopathological profiles were rated.

Patients performing within normal or near normal ranges on the mental state attribution task had fewer social behavioural abnormalities than patients with poor mentalising abilities (even when controlled for intelligence), but did not differ in executive functioning. Fair mental state performers showed less disorganisation and excitement symptoms than poor performers. The degree of disorganisation mediated the influence of mental state attribution on social behavioural skills.

Schizophrenia patients with (partially) preserved mentalising skills have fewer behavioural problems in the social domain than patients with poor mentalising abilities. Conceptual disorganisation mediates the prediction of social behavioural skills through mentalising skills, suggesting that disorganised patients may require special attention regarding social-cognitive skills training.

Poor social cognition is prevalent in schizophrenia spectrum disorders. Some authors argue that these effects are symptom-specific and that socio-cognitive difficulties (e.g. theory of mind) are strongly associated with thought disorder and symptoms of disorganisation.

The current review tests the strength of this association.

We meta-analysed studies published between 1980 and 2016 that tested the association between social cognition and these symptoms in schizophrenia spectrum disorders.

Our search (PsycINFO, MEDLINE and Web of Science) identified 123 studies (N = 9107). Overall effect size as r = −0.313, indicating a moderate association between symptoms and social cognition. Subanalyses yielded a moderate association between symptoms and theory of mind (r = −0.349) and emotion recognition (r = −0.334), but smaller effect sizes for social perception (r = −0.188), emotion regulation (r = −0.169) and attributional biases (r = −0.143).

The association is interpreted within models of communication that highlight the importance of mentalisation and processing of partner-specific cues in conversational alignment and grounding.

None.