The Cost of Switching between Kanji and Kana while Reading Japanese

Abstract

Readers of Japanese must constantly switch between decoding two types of script, Kana and Kanji. Does this incur a measurable processing cost? In four discrete-trial reaction time experiments, with an inter-trial interval of 1s or 2s, Japanese readers had to switch predictably, every second trial, between reading words in Kanji and Kana (Hiragana in three experiments, Katakana in the fourth). The task was naming in two experiments, and semantic categorisation in two. In every case performance was significantly slower, by about 13 ms on average, and less accurate, on the trials following a change of script. In these and three further experiments we show that the cost does not arise from changes in spatial extent, number of characters, or the familiarity of words written in the two scripts, that neither the task nor the direction of switching have much impact on the cost, and that there is no cost for switching between naming words in the two Kana scripts, Hiragana and Katakana. We conclude that to decode Kana and Kanji draws on somewhat different resources, and speculate on the source of the switch cost.