Associative theories of creativity posit that high-creativity individuals possess flexible semantic memory structures that allow broad access to varied information. However, the semantic memory structure characteristics and neural substrates of creative writing are unclear. Here, we explored the semantic network (SN) features and the predictive whole-brain functional connectivity (FC) associated with creative writing and generated mediation models. Participants completed two creative story continuation tasks. We found that keywords from written texts with superior creative writing performance were highly interconnected, and encompassed more semantic categories. Connectome predictive modeling (CPM) was conducted with resting-state (rs) functional magnetic resonance imaging (fMRI) data to identify whole-brain functional connectivity (FC) patterns related to creative writing. FC was found to mediate the relationship of creative writing and SN features. These results highlight how semantic memory structure and the internal thought-oriented brain functional connectivity pattern support creative writing performance. Our findings extend prior research on the role of semantic memory structure and the default mode network (DMN) in creativity, expand upon previous research on semantic creativity, and provide insight into the cognitive and neural foundations of creative writing.

Cross-domain analogical reasoning (CAR) is a potent cognitive tool that links seemingly unrelated knowledge do-mains, fostering creative thinking by identifying similarities across different fields. This study aimed to identify functional connectomes encoding individual variations in CAR abilities and to reveal the underlying brain-CAR mechanisms in predicting creativity. Participants included 69 typical university students who underwent resting-state brain MRI scans and behavioral tests. These tests assessed CAR and within-domain analogical reasoning(WAR) abilities using verbal analogy tasks in the A:B::C:D format and measured individual creativity levels using the Al-ternative Uses Test (AUT). We employed a connectome-based predictive modeling (CPM) approach, utilizing the Power264 brain atlas to identify functional connectomes supporting CAR abilities. CPM analysis indicated that both positive and combined network models could stably predict individual CAR scores, with functional connectiv-ity primarily distributed across the default mode, cingulo-opercular task control, salience, and motor networks. Moreover, mediation analysis suggested that CAR mediates the impact of brain functional connectomes on creativi-ty. Our research provides evidence for functional neural markers of CAR and reveals a potential neuropsychological pathway for predicting creativity, whereby brain functional connectomes support creativity through CAR.