This paper describes the technical adequacy and potential uses of an observation system used to measure the quality of literacy instruction in kindergarten classrooms. The Classroom Observations of Student-Teacher Interactions (COSTI) documents the frequency of four student-teacher interactions during beginning reading instruction: explicit teacher demonstrations, student independent practice, student errors, and teacher corrective feedback. Data were collected during kindergarten reading instruction, and the analyses address reliability, stability of the coded teaching behaviors, and predictive validity. Results indicated that data could be collected reliably and that teachers' provision of opportunities for independent student practice was stable across the school year. Student independent practice opportunities also predicted gains in several important reading outcomes. Implications are discussed, including potential uses of the instrument for providing teachers with feedback on their literacy instruction and for extending the knowledge base on effective literacy instructional practices. (Contains 5 tables.)