The production of English word-final stops by Mandarin and English speakers in five contexts was examined. While studies show that the phonetic realizations of segments in native English speech are context sensitive [P. A. Luce and J. Charles-Luce, J. Acoust. Soc. Am. 78, 1949--1957 (1985)], such contextual effects have not been investigated in non-native speech. Six Mandarin speakers of English and six American speakers read 10 carrier sentences containing the test words back and `bag in utterance-final position or followed by words beginning with /s/, /t/, /n/, and the vowel /(inverted vee)/. Three repetitions of the words in each context were measured for the vowel duration and the word-final stop closure duration. The closure/vowel ratio was calculated, and the two subject groups were compared. Results show that for both groups, the vowel and closure durations varied with contexts: while utterance final position showed the strongest effect, among the other contexts, /n/ and /t/ triggered longer vowel and closure durations than /s/ and /(inverted vee)/ did for both speaker groups. The non-native speakers' closure durations and the closure/vowel ratios were not nativelike and their articulatory gestures for the velar stops differed from those of native English speakers.