Definitions of 20 words represented by kana-letters (Japanese alphabet) and Chinese characters were read to 300 students. In all, 222 partial reports were treated as in the tip-of-the-tongue state and analyzed by Rubin (1975)'s method. The results indicated that the major clusters were composed of one to three kana-letters, or one-letter Chinese characters. These results partially support the Rubin's conclusion, but they also suggest the possibility that the memory unit for the Japanese word is not the morpheme but the “pseudo-letter morpheme”.