Abstract: When transcribing words written in Kana into Kanji, Japanese readers frequently generate pseudo-compounds of two real Kanji characters (Saito & Tsuzuki^[1,2]). Kawakami, Saito, and Yanase^[3] investigated whether participants' evaluation of Kanji pseudo-compounds as real words is related to their knowledge of individual Kanji characters and real Kanji compounds. Participants were divided into three knowledge groups (high, medium, and low), according to their performance on a cued recall test of Kanji compounds words. The medium knowledge group consistently rated Kanji pseudo-compound words as less wordlike than real Kanji compounds, compared to the high and low knowledge groups. In contrast, the high knowledge group rated real Kanji compounds as more wordlike than the middle or low knowledge groups. The data reported here were separately analyzed with “known or unknown” responses to the pseudo-compound words. The results indicated the same pattern of false positive responses with the “unknown” pseudo-compound words, but not with the “known”. Thus, good readers with extensive knowledge of Kanji compounds were more likely to identify a compound as real, whether it was or not, especially when they recognized the compound as “unknown”. These results suggest that linguistic metaknowledge concerning Kanji compounds and their componential characters can influence wordlikeness judgments for both real and pseudo Kanji compounds.

Key words: Wordlikeness, Common wordlikeness filter, Pseudo-compound words, Subword activation, Japanese Kanji characters