To investigate the processing of Japanese verb morphology, a new experimental task, segment naming task, was developed. Thirty undergraduate and graduate students participated in this study. The participants in Experiment 1 were asked to read aloud underlined bigrams embedded in five-letter Kana verb-phrases that consisted of two morphemes. Some bigrams were comprised of letters embedded within a morpheme, whereas others crossed the boundary of two morphemes. Naming latency was longer when a target bigram crossed the morphemic boundary than when the bigram letters came from within the morpheme. Experiment 2 showed that this boundary-crossing effect was related to familiarity. When phrases were familiar to readers, no boundary-crossing effect was observed, which suggests that such phrases were processed as one unit. When phrases were less familiar, a clear boundary-crossing effect was observed, suggesting that such phrases were processed as two units. A model of the processing Japanese verb morphology is proposed.

Key words: morphology (linguistics), Japanese, naming task, processing unit, mental lexicon