Code-mixing, which denotes switches between languages as well as a phenomenon reflecting grammars of both languages in interpersonal interactions simultaneously, is a universal language-contact phenomenon present in both individual bilingualism and societal bilingualism, and individual differences exist in both frequency and complexity of code-mixing out of multifarious factors. The present corpus-based longitudinal study investigates impacts of the variable of language dominance on Cantonese-English bilingual children's code-mixing. Spontaneous speech data with critical case sampling were collected from the Hong Kong Bilingual Child Language Corpus, where code-mixing identified in participants' utterances were analyzed both quantitatively and qualitatively. Bilingual children in the current study are discovered to code-mix more frequently and in a more complicated fashion in their weaker language with more intra-sentential switches in the dominant language involving incorporation of language structures of the dominant language with higher syntactic complexity as well as semantic value into utterances of the weaker language. In light of the correlation between language dominance and code-mixing, patterns of code-mixing can plausibly be capitalized upon to formatively assess kindergarten toddlers' bilingual development.