This study examined how the voice onset time (VOT) affected the intelligibility of English voiceless stops produced by 20 native Japanese speakers. Five native English speakers evaluated the intelligibility of three English voiceless stops in the word-initial position of the words,“ pet,”“ Ted,” and“ keck” when pronounced by native Japanese speakers. The VOT for each production was measured, and its relationship with intelligibility was examined. The results showed that /t/ and /k/ have the lowest and highest intelligibility, respectively. When voiceless stops were pronounced by native Japanese speakers with lower intelligibility, the native English speakers heard most of them as voiced. An acoustic analysis indicated that Japanese English stops with lower intelligibility had a shorter VOT than did those with higher intelligibility. Japanese English voiceless stops were highly intelligible when the VOT was longer than 30 ms for /p/, around 50 ms or longer for /t/, and about 55 ms or longer for /k/. Intelligibility tended to be low when the VOT was shorter than 20 ms for the three voiceless stops.