This study investigated the perception and production of L2 English and L1 Portuguese stops in initial position by analyzing the acquisition of voice onset time (VOT) categories. 36 Brazilian Portuguese (BP) learners of English and 36 monolingual BP speakers, all of them living in Brazil, participated in this study. There were two language sessions, English and Portuguese (monolinguals took part in the Portuguese session only). In each language session, participants took part in a production and a perception task in the respective language (the tasks were mirror-images of each other). To elicit the production data, participants took part in a delayed repetition task. To elicit the perception data, participants took part in a two-alternative forced-choice identification test. The analysis of the data showed that improvement in L2 may entail improvement in L2 perception of stops. On the other hand, the lack of effects of proficiency in L2 production of stops may suggest occurrence of learning stabilization of L2 VOT categories. L1 phonetic drift was observed in the production of the Portuguese /b/, /k/, and /g/ (but not for /p/). However, no effects of L2 learning on L1 was observed for the perception of Portuguese /b/-/p/. L2 learners who had learned English in their L1 country and in formal settings demonstrated that they were able to form new phonetic categories for the production of /p/, /b/, and /g/. The higher-proficiency group (but not the lower-proficiency group) demonstrated that they developed language-specific phonetic strategies for /p/-/b/ since they were able to process the same set of sounds on a continuum from /b/ to /p/ as either L1 or L2 stops as a function of language mode. The perception study showed that language-specific phonetic strategies, which had been observed among highly fluent bilinguals, could also be possible for this population.