A Unified Account of Concepts and Chunks
This study addresses the theoretical disconnect between research on concepts and chunking in cognitive psychology by proposing a modality-independent unified acquisition framework. Methodologically, it extends the Cobweb model to construct a unifying theory and implements the Trellis system, which is validated through computational modeling, context-free grammar testing, and synthetic grammar experiments. The results demonstrate that this system successfully represents syntactic knowledge, enabling cross-modal learning as well as sentence parsing and generation. Furthermore, it autonomously acquires compositional structures directly from input samples. By effectively bridging the theoretical gap between concepts and chunks, this work offers an integrated account of how structured linguistic knowledge can emerge without modality-specific constraints, advancing our understanding of cognitive acquisition mechanisms.