Resolving valid multiple model inferences activates a left hemisphere network

Abstract

Abstract Resolving multiple model syllogisms is more difficult than resolving single model syllogisms. Mental model theory predicts that visuospatial processing is critical for resolving syllogisms, and that demands on visuospatial processing systems will increase as reasoning problems become more difficult. An alternative account, the mixed-model approach, postulates that linguistic representations may augment visuospatial representations in multiple model problems. To test these competing hypotheses, we reorganized published archival fMRI data into single and multiple model problems, and reanalyzed it along this dimension. The critical comparison of multiple model versus single model problems revealed activation in both the left superior parietal spatial system and left frontal and temporal language areas, indicating that as reasoning problems become more difficult, reasoners augment any visuospatial model that they may have constructed with linguistic representations. This result is consistent with the mixed-model approach.