Cortical systems for local and global integration in discourse comprehension

To understand language, we integrate what we hear or read with prior context. This research investigates the neural systems underlying this integration process, in particular the integration of incoming linguistic information with local, proximal context and with global, distal context. The experiments used stories whose endings were locally consistent or locally inconsistent. In addition, the stories' global context was either relevant or irrelevant for the integration of the endings. In Experiment 1, reading latencies showed that the perceived consistency of an ending depended on its fit with the local context, but the availability of a relevant global context attenuated this effect. Experiment 2 used BOLD fMRI to study whether different neural systems are sensitive to the local consistency of the endings and the relevance of the global context. A first analysis evaluated BOLD responses during the comprehension of story endings. It identified three networks: one sensitive to consistency with local context, one sensitive to the relevance of the global context, and one sensitive to both factors. These findings suggest that some regions respond to the holistic relation of local and global contexts while others track only the global or the local contexts. A second analysis examined correlations between BOLD activity during listening of the story endings and subsequent memory for those endings. It revealed two distinct networks: Positive correlations in areas usually involved in semantic processing and memory for language, and negative correlations in sensory, motor, and visual areas, indicating that weaker activity in the latter regions is conducive to better memory for linguistic content. More widespread memory correlates were found when global context was relevant for understanding a story ending. We conclude that integration at the discourse level involves the cooperation of different networks each sensitive to separate aspects of the task, and that integration is more successfully achieved when the processing of potentially distracting information is reduced.