Unlike many communities studied by variationists in which SLI means convergence of spoken and written language, for Cantonese speakers, language ideology historically meant the opposite. In this presentation, I show how traces of this ideological distinction between written and spoken codes (cf. Snow 2004) remain present in both Hong Kong and in Toronto. I focus on two sets of sound correspondences found with (Standard) Mandarin cognates: Cantonese /i/ to Mandarin /ə/ and Cantonese /y/ to Mandarin /u/). Only Cantonese /y/ shows convergence, but only for second-generation Toronto speakers who are least likely to speak Mandarin. Toronto English influence, thus, better accounts for this change. The lack of standard Chinese influence on vowel pronunciation can be understood in terms of a historic distinction between codes and in terms of how speakers understand these distinctions (as I show from sociolinguistic interview excerpts) even as these distinctions continue to evolve under Western influence.