Abstract
<p>This paper advances a new account of linguistic typology grounded in the hypothesis of neural reuse, namely that languages selectively recruit pre-existing neurocognitive systems for grammatical purposes. The proposal is tested through a case study of the Mandarin classifier zuò (座), combining cognitive semantic analysis with fMRI evidence. The semantic analysis shows that zuò forms a highly structured polysemy network organized around scene-based construal, progressing from anchored entities and perceptual salience to bounded environments, landmarks, navigable spaces, and goal-directed schemas.To examine whether this organization reflects underlying neural systems, an fMRI experiment contrasted zuò with a non-scene classifier. Processing zuò alone—prior to any scene-denoting noun—activated regions associated with scene perception and navigation, including the parahippocampal place area, retrosplenial cortex, and occipital place area. This indicates that a grammatical morpheme can recruit a scene-selective cortical system, supporting the view that polysemy may be constrained by shared neural architecture.At the same time, the full semantic range of the classifier cannot be derived from neural structure alone. Several extensions depend on lived experience and culture-specific knowledge, suggesting that neural reuse provides a scaffold that is subsequently elaborated through usage and cultural models. The findings support a typological perspective in which languages differ in whether and how they recruit available neural systems, thereby shaping the emergence of grammatical categories and polysemy networks.</p>