Current thinking and research on the nature of formative evaluation of replicable forms of instruction are summarized. Methodologies appropriate for other evaluation roles (e.g. summative) are shown to be often inappropriate for the formative role. This paper attempts to classify sources of information concerning the merit of educational materials and to suggest appropriate methodologies of formative evaluation. The data sources are of three broad types: external sources, internal sources, and contextual sources. Several types of data under each heading are examined to illustrate the classification system. (Author)