This study discusses the duration of English consecutive stops across word boundaries. When two stops are adjacent across a word boundary, one is a word-final stop and the other is a word-initial stop. They share the stop closure. It is known that closure durations of voiceless word-final stops are longer than those of voiced stops in English and the places of articulation of both word-final and word-initial stops do not affect their closure durations. But the VOTs of the word-initial stops are known to be affected by the voicing and their places of articulation. In this study, the shared closure duration of the consecutive stops followed by the VOT of the word-initial stop was measured as the duration of the two consecutive stops. Both English and Korean speakers produced English nonsense word pairs that contain the consecutive stops. The results showed that English speakers’ production was affected by both stops; however, they relied more on the voicing and the places of the word-initial stops than on those of word-final stops. But the Korean speakers did not rely on the voicing and the places of articulation of the word-final stops at all due to the influence of the well-known coda neutralization in Korean phonology. They relied only on those of the word-initial stops. The average durations of the consecutive stops produced by the Koreans were longer than those by the English speakers.