Background. Social metacognition is still poorly understood in schizophrenia, particularly its neuropsychological basis and its impact on insight and medication adherence. We therefore quantified social metacognition as the agreement between objective and subjective mentalization and assessed its correlates in a sample of individuals with schizophrenia spectrum disorders. Methods. Participants consisted of 143 patients with schizophrenia or schizoaffective disorders who underwent a metacognitive version of a mentalization task, an extensive neuropsychological battery, and a clinical evaluation to assess their insight into illness and medication adherence. We studied potential interactions between confidence judgments and several neuropsychological and clinical variables on mentalization accuracy with mixed-effects multiple logistic regressions. Results. Confidence judgments were closely associated with mentalization accuracy, indicative of good social metacognition in this task. Working memory, visual memory, and reasoning and problem-solving were the three neuropsychological dimensions positively associated with metacognition. By contrast, the two measures of medication adherence were associated with poorer metacognition, whereas no association was found between metacognition and clinical insight. The multiple regression model showed a significant positive impact of better working memory, older age at onset, longer duration of hospitalization, and worse medication adherence on social metacognition. Conclusions. We discuss possible mechanisms underlying the apparent association between social metacognition and working memory. Adherence should be monitored when remediating social metacognition, and psychoeducation should be given to patients with a high level of awareness of their capacity to mentalize.