Voiceless /p,t,k/ are implemented as aspirated stops in English, but as unaspirated stops in Dutch. We examined identification of a voice onset time (VOT) continuum ranging from /da/ to /ta/ in two language "sets" designed to induce native Dutch subjects to perceive the stimuli as if they were listening to Dutch or English. The effect of language set was highly significant, but the boundary shift was very small (2.1 ms longer in English than in Dutch) for three groups of subjects differing widely in English language proficiency. It nevertheless showed the subjects were aware of acoustic differences distinguishing Dutch and English /t/ and that the procedures were effective in creating differing language sets. Nearly every subject produced a longer mean VOT in English than Dutch /t/. The magnitude of the production shift was significantly greater for proficient than non-proficient subjects. Proficient Dutch speakers of English produced Dutch /t/ with shorter VOT values than non-proficient subjects, suggesting they formed a new category for English /t/. We speculate that the language set effect was small because subjects used their English /t/ category to identify stops in both sets. This was probably due to the fact that the synthetic stimuli, which were modeled on the English /t/-/d/ contrast, differed substantially from their Dutch /t/ category. © 1987.