The Multilink model applied successfully to non-alphabetic scripts The case of Japanese kana

Abstract

Multilink is a localist-connectionist computational model that accurately simulates Dutch-English bilingual word processing times in several linguistic tasks (Dijkstra et al., 2018). The model can help us greatly to understand the bilingual human mind, but until now it has only been applied to languages that use an alphabetic script. Many other languages across the world use a variety of non-alphabetic scripts, and many bilinguals are fluent in languages with differing scripts. The current project extends the model’s performance to non-alphabetic scripts, namely Japanese katakana and hiragana. The model was indeed able to process these new scripts and it could accurately simulate human lexical decision studies involving Japanese-English bilinguals. This suggests that the mechanisms underlying human word processing may be universal and extend across writing systems.