Previous studies have shown that prosodic information plays an important role in spoken word recognition by L2 learners (Cooper et al., 2002). However, it still remains unclear exactly how and when learners use prosodic cues in the word recognition process. This study aims at answering the question whether Korean (Seoul dialect) learners of Japanese utilize prosodic information incrementally by investigating exactly how and when L2 learners use accentual information in the course of word recognition process. We conducted eye-tracking experiments with a visual world paradigm, using artificial Japanese words. The results indicated that the learners couldn’t use prosodic cues incrementally in word recognition because of the interference by perceptual experience with L1 phrase-level prosody.