抄録
We investigated the effect of existing words on the generation of non-words. Native Japanese speaking college students (n = 103) generated 2702 non-words consisting of two to ten Hiragana letters during five minutes. The generated non-words resembled existing, familiar, Japanese words in terms of letter frequencies and of bigrams (two adjacent letters in a word). The non-words were also constrained by a phonological rule (*DD-constraint; Ito & Mester, 1996) of the Japanese lexicon, which implies that the non-words were related to existing words. and could be included in the extreme periphery of the Japanese lexicon. Further analysis indicated that the selection of letters in the generated non-words was biased. The participants tended to repeatedly use the same letters, and moreover, frequencies of letters in non-words were more redundant than those in existing words and random combination of letters. These results suggest that the generated non-words were influenced by existing words, but also constrained by characteristics of the process of retrieving letters from memory.
