Abstract
The ability to integrate semantic information into the context of a sentence is essential for human communication. Several studies have shown that the predictability of a final keyword based on the sentence context influences semantic integration on the behavioral, neurophysiological, and neural level. However, the architecture of the underlying network interactions for semantic integration across the lifespan remains unclear. In this study, 32 healthy participants (30-75 years) performed an auditory cloze probability task during functional magnetic resonance imaging (fMRI), requiring lexical decisions on the sentence's final words. Semantic integration demands were implicitly modulated by presenting sentences with expected, unexpected, anomalous, or pseudoword endings. To elucidate network interactions supporting semantic integration, we combined univariate task-based fMRI analyses with seed-based connectivity and between-network connectivity analyses. Behavioral data revealed typical semantic integration effects, with increased integration demands being associated with longer response latencies and reduced accuracy. Univariate results demonstrated increased left frontal and temporal brain activity for sentences with higher integration demands. Between-network interactions highlighted the role of task-positive and default mode networks for sentence processing with increased semantic integration demands. Furthermore, increasing integration demands led to a higher number of behaviorally relevant network interactions, suggesting that the increased between-network coupling becomes more relevant for successful task performance as integration demands increase. Our findings elucidate the complex network interactions underlying semantic integration across the aging continuum. Stronger interactions between various task-positive and default mode networks correlated with more efficient processing of sentences with increased semantic integration demands. These results may inform future studies with healthy old and clinical populations.