Graph-like organization of non-spatial knowledge about social closeness in movie narratives

Research in cognitive science has progressively highlighted the notion of geometric representations (map and graph-like structures) for storing and organizing knowledge in both spatial and non-spatial domains. It's unclear, however, whether these representations aid in organizing knowledge from unconstrained and naturalistic episodic encoding and whether it is possible to identify objective episodic parameters that support the implicit construction of structural-semantic knowledge. Here, we investigated how statistical regularities in a movie narrative contribute to the generalization process underlying social knowledge's gradual construction and organization. Using the narrative of a TV series, participants watched five episodes and then performed a retrieval task in which they rated the strength of social relationships between characters (i.e., social closeness). An objective graph of social closeness, based on parameters extracted from the teleplay, was compared to subjective social graphs derived from participants' judgments. The results revealed a strong correlation between the two graphs and highlighted the importance of physical co-occurrence in shaping social representations. The generalization process was independent of awareness of task demands, suggesting an implicit mechanism. Additionally, increased episodic exposure improved both the inter-subject stability and the coherence of the social graphs, supporting the notion that repeated episodic experiences enhance semantic representations. These findings emphasize the role of episodic statistical regularities as building blocks for the organization of non-spatial, conceptual knowledge in graph-like structures.