Although specific procedures for "participant observation" (PO) vary, it is always based on the principle that while the observer (O) will adopt preconceived structural outlines and dimensional scales in the course of his summary, he--rather than any outlines or scales--is the instrument for inferring data. Applications of modified PO approaches were made on selected Head Start and elementary school classes in connection with 2 concurrent projects. Eleven experienced teachers received observer training in PO theory. Films were made of a stratified sample of classes for use in training and for analysis of observational reports and ratings. In one project 8 scales were used in rating 16 teachers; rater agreement was 92% and interscale correlations varied between .60 and .90. In the other project an observational survey of 38 classes over a 5-month period produced intercorrelation between 10 scales ranging between .60 and .80; the internal consistency of the scales is well documented. Since complete data is not yet available, this report is incomplete. (Included are a 6-page discussion of the rationale of various strategies for observing teacher behavior, with notes on the comparative validity and uses of "objective" and "subjective" methods of abstracting teacher behavior into data; a theoretical model for "participant observation"; a 7-item bibliography; the scales used for rating PO reports; and other materials used in training O's). (JS)