A Bayesian Network Approach to Social and Nonsocial Cognition in Schizophrenia: Are Some Domains More Fundamental than Others?

Abstract

Social and nonsocial cognition are defined as distinct yet related constructs. However, the relative independence of individual variables-and whether specific tasks directly depend on performance in other tasks-is still unclear. The current study aimed to answer this question by using a Bayesian network approach to explore directional dependencies among social and nonsocial cognitive domains. The study sample comprised 173 participants with schizophrenia (71.7% male; 28.3% female). Participants completed 5 social cognitive tasks and the MATRICS Consensus Cognitive Battery. We estimated Bayesian networks using directed acyclic graph structures to examine directional dependencies among the variables. After accounting for negative symptoms and demographic variables, including age and sex, all nonsocial cognitive variables depended on processing speed. More specifically, attention, verbal memory, and reasoning and problem solving solely depended on processing speed, while a causal chain emerged between processing speed and visual memory (processing speed → attention → working memory → visual memory). Social processing variables within social cognition, including emotion in biological motion and empathic accuracy, depended on facial affect identification. These results suggest that processing speed and facial affect identification are fundamental domains of nonsocial and social cognition, respectively. We outline how these findings could potentially help guide specific interventions that aim to improve social and nonsocial cognition in people with schizophrenia.