While previous studies have shown that pitch accent constrains lexical access in Japanese, the role of mispronunciation of it is less often addressed. Two types of pitch accent mispronunciations exist in Japanese. One implies that mispronouncing results in other lexical items; when jidoo HLL “children” is prosodically mispronounced as jidoo LHH, it is another real word (“automaticity”). The other implies that mispronunciation may result in nonwords; when tansu LHH “wardrobe” is mispronounced as tansu HLL, it is a nonword. The present study investigated the process of lexical access to prosodically mispronounced words using a priming paradigm. When a prosodically mispronounced nonword (e.g., tansu HLL as a mispronunciation of tansu LHH “wardrobe”) was presented as a prime, the response to a semantically related target word was delayed compared to when the prime was a semantically unrelated control word. However, when a segmentally mispronounced nonword (e.g., dansu LHH as a mispronunciation of tansu LHH) was presented, such a delay was not observed. These results indicated that the mispronunciation of pitch accent interfered with lexical access more significantly than the mispronunciation of a segment, supporting the view that pitch accent plays a crucial role in lexical access in Japanese spoken word recognition.