Statistical learning of radical configuration in Chinese character recognition

Poster Session A - Saturday, April 13, 2024, 2:30 – 4:30 pm EDT, Sheraton Hall ABC

Tzu-shin Lin^1,2 (ts.lin@g.ncu.edu.tw), Chia-ying Lee^1,2,3; ¹Academia Sinica, ²National Central University, ³National Chengchi University

Chinese characters are comprised of radicals within a constant square-shaped space. Previous studies have demonstrated that statistical learning is the core mechanism for extracting orthographic regularities, such as radical position, radical combinability, and phonetic consistency, in learning to read Chinese. However, radicals can be arranged with different spatial configurations. This study analyzed the distribution of radical configuration and found that 60％ of compound characters contain two radicals, which are horizontally arranged from left to right (i.e., 柏). However, only 20% of them are vertically arranged from top to bottom (i.e., 空). Participants were invited to perform a lexical decision task in a priming paradigm to examine whether readers would process the radical configuration information in Chinese character recognition. The prime character would either be the dominant (left-right) or non-dominant (top-bottom) configurations. The target would either be a real character or a pseudo-character with a congruent or incongruent configuration with the prime. The event-related potentials (ERPs) to the target characters indicate an interaction between configuration dominancy and congruency on N170 with posterior sites and P200 with frontal regions, in which incongruent condition elicits greater amplitude than congruent condition for the non-dominant configuration. Meanwhile, a three-way interaction effect is found on N400 with posterior sites, revealing that pseudo-characters show a greater N400 negativity than real characters at incongruent conditions in targets with the dominant configuration. These findings support that the radical configuration also serves as an important orthographic information for Chinese character recognition.

Topic Area: LANGUAGE: Other