In this study we compared the duration ratios of native speakers of Cantonese, English, French, and Japanese who produced non-native phonemic quantity contrasts in Japanese (two-way) and Estonian (three-way). These four L1 backgrounds differ in terms of the extent to which duration is used to mark quantity contrasts (e.g. short vs. long), ranging from non-phonemic (i.e. French) to systematic two-way (i.e. Japanese). A shadowing task was used to elicit participants’ production. Estonian and Japanese stimuli (N = 360) were played in two separate blocks. The participants wore a pair of earphones in a quiet room and repeated the word they heard. The results showed that all participant groups were able to tell apart the quantity conditions, though for the Estonian target words Short vs. Long were better differentiated than Long vs. SuperL. The unexpectedly good performance of the French speakers in discrimination and identification (Lee et al., 2023) was replicated, as was the relatively poor performance of the Cantonese speakers. The theoretical implications of these findings are discussed.