Complex kanji characters in Japanese are characterized by the conjunction of submorphemic components (or radical) and their "companions." Two experiments investigated the roles of radical token-frequency (i.e., the number of occurrences of the radical in written materials) and of radical neighborhood size (i. e., the number of kanji members containing the radical, which correspond to the number of companions of the radical, Nc) in the recognition of kanji characters. Participants were briefly presented with two left-right kanji (e. g., 研 and 秒), each of which consisted of two radicals, and asked first to write one of the radicals and then its companion (accordingly, to write one of the two characters}. Characters consisting of high token-frequency radicals were reported more accurately than characters with low token-frequency radicals (Experiment 1). This suggests that the radical token-frequency increases the activation of the presented character and its recognition accuracy. When the radical's Nc was manipulated so that both average radical token-frequency and character frequency were matched across conditions (Experiment 2), participants reported characters containing a left radical with low-Nc more accurately than characters containing a left radical with high-Nc; however, such an effect of the radical's Nc was not clearly found for the right radical. This pattern suggests that representations of neighborhood characters sharing the same left radical are activated as candidates via radical processing in competition with the representation of the presented character. All the results in this study provide evidence that recognition of logographic characters are explained by a model such as the interactive-activation model for alphabetic scripts.

Key words: visual word recognition, type-frequency, token-frequency, kanji character-word, radical-companion