Japanese two-kanji compound words (Jukugo: e.g., 病院 "hospital") have two sets of neighbors: Rear-neighbors that share the same front-kanji and vary in the rear-kanji, and Front-neighbors that share the same rear-kanji and vary in the front-kanji of the Jukugo. We investigated the role of semantic activation of word neighbors in Japanese kanji word recognition. The effects of neighborhood frequency for Rear-neighbors and Front-neighbors (Experiment 1) and semantic similarity between a Jukugo and its Rear-neighbor (Experiment 2) were examined using the lexical decision task. In Experiment 1, when a Jukugo and its neighbors were semantically similar, Rear-neighbors of higher frequency produced longer reaction times, whereas Front-neighbors of higher frequency produced shorter reaction times. In Experiment 2, semantic similarity between a Jukugo and its neighbor of higher frequency produced longer reaction times, although no delay was observed for semantically dissimilar Jukugo and neighbors. The results are interpreted within the companion-activation model.

Key words: two-kanji compound word, semantic similarity, neighborhood frequency, word identification, companion-activation model.