Classroom Observation Protocol

More than one. Generalizability analyses, including from the Measures of Effective Teaching project, show that a single lesson scored by one rater is too unreliable for high-stakes use because much variance comes from the particular lesson and the particular observer rather than the teacher. Reliable measurement typically requires several lessons scored by multiple trained raters; the exact number depends on the instrument and the desired reliability, which G-theory decision studies estimate directly.

Because an observation score reflects several sources of variation at once — the teacher, the rater, the lesson/occasion, and their interactions. Generalizability theory partitions the score variance into these facets and estimates how dependable scores are under different designs (more raters, more lessons). It then guides design through decision studies, telling you how many observations and raters are needed to reach a target reliability. This is essential because the quantity of interest — the teacher's typical practice — must be separated from rater and occasion noise. See the related Generalizability Theory entry.

They embody somewhat different conceptions of teaching and serve different uses. CLASS focuses on the quality of teacher-student interactions across emotional support, classroom organization, and instructional support, and is heavily used in early-childhood and research contexts. The Danielson Framework for Teaching is a broader four-domain framework (planning and preparation, classroom environment, instruction, professional responsibilities) widely adopted in K-12 teacher evaluation. Both are standardized, rubric-based, and require trained raters, but they differ in scope, structure, and typical application.