The current study evaluated the relation between Spanish and English vocabulary. Whereas previously reported correlations have revealed strong differences among types of vocabulary measures used and the ages of the students tested, no prior study had used a multilevel model to control for classroom-level differences. The current study used multiple measures of vocabulary--picture vocabulary and narrative production tasks--in multilevel models of 1,300 Spanish-speaking students in 247 kindergarten and 1st-grade classrooms in English immersion and bilingual transitional programs. The current results highlight the need to separate classroom effects from student effects, since for vocabulary measures, student-level correlations were strongly biased toward zero when classroom-level correlations were opposite in direction from student-level correlations. Most important, the current results support a strong distinction between types of vocabulary measure (e.g., picture vs. narrative) and suggest sizable influence of instruction for questions of bilingual performance. (Contains 3 figures and 5 tables.)