In this paper, the authors report research focused directly on the validation of the Coding of Academic Teacher-Student interactions (CATS) direct observation instrument. They use classroom information gathered by the CATS instrument to better understand the potential mediating variables hypothesized to influence student achievement. Their study's purpose is to gather the kinds of validity evidences that match the proposed interpretations and uses of the CATS instrument (Kane, 2008; Messick, 1995). Therefore, they first explore the content aspect of construct validity by collecting information about the content relevance and representativeness of the observation instrument (Messick, 1995). Second, concerned about the consistency of the observation data collected across a number of independent observers, they measure inter-observer agreement. Finally, they focus on the criterion-predictive aspect of construct validity and investigate the relationship between student gains and the quantity of teaching. They anticipate that classrooms with higher rates of practice opportunities and teacher demonstrations will foster greater increase in classroom-level achievement than classrooms with fewer instructional interactions. To address this assumption, they explore curriculum based measurement (CBM) scores and information of observed teaching practice. The authors conducted the current research within the context of a federally funded randomized control efficacy trial, the Early Learning in Mathematics: Efficacy Trials in Kindergarten Classrooms (ELM-ETKC) study (Baker, Chard, Clarke, & Smolkowski, 2008). A total of 66 teachers and 1,450 kindergarten students participated in the first year of the ELM-ETKC study. Preliminary findings indicate that the CATS tool is reliable and valid for documenting the quantity and quality of student-teacher interactions in kindergarten mathematics classrooms. This study has several implications for the observation of classroom instruction. First, the recent advent of Response to Intervention calls for the delivery of high quality instruction to occur in the general education classroom across content areas. School personnel such as instructional coaches and school psychologists will need reliable and valid tools for determining whether students are receiving highly interactive math instruction in content that aligns with state standards. The CATS tool could potentially serve as such a mechanism. Additionally, generating estimates of rates of student math practice opportunities that are associated with strong student mathematics outcomes could be useful for the design and development of new mathematics curricula. Likewise, patterns of teacher models and academic feedback that appear to associate with strong student math outcomes could have implications for teacher professional development of early mathematics instruction. (Contains 1 figure.)