Abstract
Several studies have shown that communicative language production as compared to non-communicative language production recruits parts of the mentalizing or theory of mind network, yet the exact role of this network in communication remains underspecified. In this study, we therefore aimed to test under what conditions the mentalizing network contributes to communicative language production. We were especially interested in distinguishing between situations in which speakers have to consider which information they do or do not share with their addressee (common vs. privileged ground information). We therefore manipulated whether speakers had to distinguish between common and privileged ground in order to communicate efficiently with the listener, in addition to comparing language production in a communicative and a non-communicative context. Participants performed a referential communicative game in the MRI-scanner as well as a similar, non-communicative task. We found that the medial prefrontal cortex, a core region of the mentalizing network, is especially sensitive to communicative contexts in which speakers have to take their addressee's needs into account in order to communicate efficiently. In addition, we found neural differences between the communicative and the non-communicative settings before speakers started to plan their utterances, suggesting that they continuously update common ground in a communicative context.