Students of advanced Japanese are expected to be able to read authentic Japanese texts, but when doing so, they often encounter unknown kanji words. The present study aims at providing better understanding of these students' ability to infer unknown kanji words while reading Japanese texts not tailored for non-native readers.
The theoretical and pedagogical issues examined in this study are situated in the wider field of L2 reading or psycholinguistic research: (a) L2 lexical inferencing, (b) kanji word processing and retention, (c) lower-level processing in L2 reading, and (d) heritage language learners' reading skills.
The subjects (N=42) were English L1 students enrolled in advanced Japanese classes at an American University. They were divided into two proficiency groups based on their levels of comprehending two authentic Japanese texts. The subjects' kanji inferencing ability was measured by (a) the decontextualized kanji test (a test of kanji words in isolation), and (b) the contextualized kanji test (a reading and thinking aloud task).
Major findings were:
1. The subjects could identify the meanings of kanji words in context significantly better than when the same words were given as isolated words; at the same time, they frequently made erroneous guesses or failed to guess at all. In general, more proficient readers could use the contextual clues better than less proficient readers. Among the more proficient subjects, the bilingual heritage language learners could not utilize the contextual clues for guessing any better than their non-heritage counterparts.
2. Contextual clues did not help the subjects read kanji words any better. However, when they could pronounce the given semantically unknown kanji words fully or partially while reading, such phonological knowledge proved to be related to successful inferencing in context. The use of contextual clues in guessing the meanings of completely unpronounceable kanji words did not succeed very often.