Measures of teaching are used internationally to understand and improve quality in early education with little consideration for ecological validity. In this study, we analyze videos gathered in 58 K-1 classrooms in Central Mexico to evaluate the validity and reliability of scores from an observational tool developed in the U.S. and used internationally: the Classroom Assessment Scoring System (CLASS). We use mixed methods to examine ecological validity to propose a revised, better-fitting three-factor solution of the CLASS in Mexico: Emotional Support (Positive Climate, Negative Climate), Social Relationships for Teaching (Teacher Sensitivity, Instructional Learning Formats, Behavior Management, Productivity), and Instructional Interactions (Regard for Student Perspectives, Concept Development, Quality of Feedback, Language Modeling). Generalizability study findings demonstrate moderate reliability across three sources of measurement error: raters, days, and video segments. These findings support the need for a 'unified approach to validity' to develop, adapt, and refine measures of teaching as a basis for enhancing teacher development and teaching quality in early education in Mexico and throughout Latin America.