Structural and Semantic Speech Graph Analysis of Dream Reports in Congenitally and Late Blind Individuals

Background: Visual input is thought to influence spatial cognition and language. While blind individuals often rely on egocentric spatial representations, it remains unclear whether and how visual deprivation affects dream-related language. This study applied speech graph analysis (SGA) to investigate linguistic differences in dream reports from congenitally blind (CB), late blind (LB), and sighted control (SC) individuals. Methods: We retrospectively analyzed 333 dream reports from an open-access database (DreamBank): 118 from CB, 75 from LB, and 140 from SC individuals. Graph-theoretical metrics of structural and semantic speech organization were extracted using validated NLP and SGA pipelines, including recurrence (L2/L3), connectivity (LSC), and lexical diversity (nodes). Results: Compared to SC, both CB and LB groups showed significantly reduced lexical diversity and increased long-range recurrence (LSC), suggesting greater linguistic cohesiveness. LB reports showed a specific increase in short-range recurrence cycles (L2, L3), not observed in CB. Spectral analysis supported these group differences, indicating altered graph-wide connectivity properties in blind groups. Conclusions: Blind individuals demonstrate distinct structural and semantic features in dream-related language, consistent with more egocentric narrative construction. These findings support a potential role of sensory experience in shaping cognitive-linguistic encoding. Further prospective studies are needed to explore underlying neural mechanisms and developmental trajectories.