The purposes of this study were (a) to identify how ESL learners make phonological changes in English words in a code-mixing context; and (b) to examine ESL learners' attitudes toward Cantonese-accented English and code-mixing in the classroom setting. Two groups of learners with high proficiency (HP) and mid-level proficiency (MP) were recruited to participate in the research. A specially designed codemixed script, an English translated version, and a list of isolated English words served as the tasks for collecting phonological data. A questionnaire survey was then used to examine participants' opinions on code-mixing and its effects on pronunciation learning. The results showed that HP and MP learners pronounced numerous words similarly in a Cantonese-accented manner; however, MP learners were less likely to switch back to the correct pronunciation when the context of code-mixing was changed to pure English or when given a list of isolated English words. The survey results found that MP learners tended to be slightly more positive toward Cantonese-accented English and the use of a mixed code in English as a medium of instruction classrooms. Nonetheless, the use of code-mixing was less preferred in English lessons for learners of both groups. More HP learners considered Cantonese-accented English as a symbol of identity as Hongkongers than MP learners.