The Neural Representation of Mental Models Across Content Type: A Common Spatial Structure

Mental models provide a cognitive framework that allows for organizing and manipulating information while reasoning about the world. Deductive reasoning with mental models is supported by a frontoparietal network, including regions of anterior prefrontal cortex associated with relational integration, and superior parietal regions associated with spatial cognition. Based in part on this evidence, mental models are often considered spatial representations. However, studies of transitive reasoning often rely on direct perception of stimuli that are inherently spatial in content, leaving open the possibility that the associated neural representations are specific to content that is inherently spatial or concretely perceivable. Here we directly test the hypothesis that the neural representation of mental models generated through transitive reasoning relies on this same frontoparietal network irrespective of the spatial nature of the stimulus content. Specifically, participants generated three distinct mental models through a transitive reasoning task. The content within the three models ranges from expressly visuospatial to entirely abstract. Moreover, all of the mental models participants generated were based on inferred relationships that were never directly observed. Multivariate representational similarity analysis was used to assess the correlation between these to-be-learned mental models and the patterns of neural activity elicited while viewing individual stimuli after training. Patterns representative of the mental models were revealed in both superior parietal lobule and anterior prefrontal cortex. Notably, these neural patterns were highly convergent across stimulus types. These results support the conclusion that, independent of content, relational reasoning using mental models relies on neural mechanisms associated with spatial processing.