Spatiotemporal dynamics of reading Kana (syllabograms) and Kanji (morphograms)

Abstract

Reading engages complex neural networks integrating visual, phonological, and semantic information. The dual-stream model posits ventral and dorsal pathways for lexical and sublexical processing in the left hemisphere and is well-supported in alphabetic languages. However, its applicability to non-alphabetic scripts remains unclear. The Japanese writing system, comprising Kana (syllabograms) and Kanji (morphograms) with distinct orthographic, phonological, and semantic properties, provides a unique framework to investigate neural dissociation between phonological and orthographic-semantic processing. Previous studies suggest that Kanji relies on the ventral route for whole-word recognition and semantic processing, whereas Kana depends mainly on the dorsal route for phonological decoding via grapheme-to-phoneme conversion; however, their spatiotemporal dynamics remain unknown. Using high-gamma power analysis from electrocorticography recordings in 14 patients with epilepsy and subdural implants, we examined the spatiotemporal neural dynamics of Kana and Kanji reading. Participants completed a visual lexical decision task with Kana and Kanji words and pseudowords. Across 912 electrodes, differential high-gamma power analysis showed that Kanji activated bilateral occipitotemporal fusiform regions early (120–550 ms) and the left inferior temporal gyrus (150–240 ms). Conversely, Kana showed prolonged late activation (270–750 ms) in the left-lateralised superior temporal, supramarginal, and inferior frontal gyri, especially during pseudoword processing. These findings indicate that Kanji relies on bilateral ventral stream earlier, while Kana depends on the left dorsal stream, with slower processing reflecting the extra grapheme-to-phoneme conversion. This underscores the value of non-alphabetic languages in elucidating both universal and script-specific neural mechanisms, advancing a cross-linguistic understanding of the reading network.