EXPRESS: Stroke Features in the Chinese character Recognition.

Abstract

While line vertices, terminations, and midsegments are critical for Roman letter identification, the diagnostic features of Chinese character strokes remain unclear. This study examines how local stroke-level features and global line-relation mechanisms contribute to Chinese character recognition. In Experiment 1, we applied the Bubbles classification image technique to native Chinese readers to identify diagnostic stroke features. Results revealed four key features: horizontal hooks, dots, vertical turnings, and raise. These features, while analogous to terminations in alphabetic systems, reflect unique dynamics of Chinese stroke production, marking stroke origins and terminations. Experiment 2 employed a delayed-segment paradigm to assess functional significance of these features. Greater degradation of vertices and midsegments significantly prolonged reaction times, and removal of stroke-based terminations (e.g., hooks) impaired recognition accuracy. Together, these findings support a two-tiered hierarchy in Chinese character recognition: stroke-specific terminals enable fine-grained feature discrimination, while line-relation features support global structural integration. The results affirm script-general principles (midsegments and vertices as perceptual anchors) and highlight language-specific adaptations, where stroke terminations function as dynamic positional cues.