Abstract This study examines how environmental features shape Chinese cape toponyms formed with five body-part generics: Zui (嘴 ‘mouth’ ) , Jiao (角 ‘horn’ ) , Tou (头 ‘head’ ) , Bi (鼻 ‘nose’ ) , and Wei (尾 ‘tail’ ) . Using a national gazetteer of Chinese headlands, we model the association between dialect regions and geomorphic parameters (slope, elevation, area, apex angle) and the choice of generics. Dialect region is the strongest predictor, followed by slope, elevation, and apex angle, whereas area shows no effect. Threshold analyses identify parameter ranges under which particular generics are most likely to be used. Classifiers trained on these features achieve solid performance (e.g., CNN F1 = 0.70), supporting the robustness of the predictors. We also compare “geographical similarity” (prototype 3D terrain reconstructions) with “semantic similarity” (embedding-based distances) and find they only partially align: terms that are semantically close in general language can map to distinct landform profiles in toponyms. Methodologically, the study demonstrates how integrating quantitative modeling, terrain visualization, and semantic analysis provides a reproducible framework for exploring language-environment interactions. Theoretically, it deepens our understanding of how natural landscapes influence lexical choice and metaphorical extension, showing that linguistic categorization reflects perceptual and ecological constraints.
Hu et al. (Mon,) studied this question.