This study examined single-word code-mixing produced by bilingual preschoolers in order to better understand lexical choice patterns in each language. Analysis included item-level code-mixed responses of 606 five-year-old children. Per parent report, children were separated by language dominance based on language exposure and use. Children were assigned to a no-risk or at-risk for language impairment group based on individual performance from an English-Spanish screening battery. Data analysis compared the prevalence, frequency, and accuracy of code-mixed responses on expressive semantic items across participants' language dominance and risk status. Language dominance and risk status impacted children's code-mixing patterns. The correct number of code-mixed responses on the English screener was influenced by risk status, whereas language dominance determined the number of correct code-mixed answers on the Spanish screener. Lexical choice and language selection depend on linguistic knowledge and skill. During this emergent stage of bilingualism, preschoolers demonstrate the use of code-mixing as a compensatory strategy to fill lexical gaps. Consistent with previous studies, findings indicate that code-mixing necessitates linguistic competence in more than one language. (Contains 3 tables and 3 figures.)