Abstract: To explore whether the boundedness of words could influence the code-switching processing and whether this influence was present across different phrase structures (language specific and language non-specific), 72 Mandarin-dialect bilinguals and 68 Mandarin-English bilinguals were recruited to complete the tasks of classifier phrases production in both code-switching and non-switching conditions.There are two major findings:1) The switch from strong language (Mandarin) to the weak one (Dialect & English) incurred cost,but the switch from weak languages to the strong one did not; 2) Only when classifiers phrases were language non-specific, did the typologies (general classifiers & special classifiers) of classifiers have influence on the producing RT of weak language (Dialect), and the magnitude of switch cost from weak language to the strong one.The findings suggest that code-switching mechanism is weak language biased, that the boundedness of words in language non-specific structure can affect code-switching processing.