Orthographic Neighbourhood Size Effects in Chinese Character Recognition: Small, Inconsistent, and Theoretically Ambiguous

Existing orthographic neighbourhood size measures for Chinese words rely on the representation of words or characters as consisting of two components, which leads to coarse measurements. Recently, Wang and Keuleers (2024) proposed the use of Ideographic Description Sequences (IDSs) which encode the basic stroke patterns of characters and their spatial relationships to create more precise neighbourhood measures. In two studies, we compared the effect of different neighbourhood size measures on recognition of single Chinese characters, while controlling for character frequency and number of strokes. In Study 1, we replicated the conventional two-unit approach using neighbourhood size measures for semantic (N-sc) and phonetic (N-pc) components. We found no effect of N-pc or N-sc. In Study 2, we used IDS-based measures to represent orthographic similarity at a more detailed level. In Study 2a, we found that the effect of a neighbourhood size measure based on single-component substitution (N-ch) was inhibitory but weak. In Study 2b, we found that a neighbourhood density measure using weighted edit distance (WLD-10) showed facilitation. In Study 2c, we found that this facilitatory effect was retained with a normalised measure (WND-10). While conceptually the IDS-based measures are better neighbourhood measures and one might hope that this would also imply stronger neighbourhood effects, our results show that the effects of neighbourhood size on character recognition remained small and difficult to interpret theoretically. Our results suggest that if any meaningful effects of orthographic neighbourhood size exist for character recognition in Chinese, they are probably small and inconsistent.