What needs to be done with preschool programs is to define specifically their desired product (the behaviors we want in the children) and to determine the optimal process by which this product can be brought about. Using an instrument called the Observation of Substantive Curricular Input in 1967-68, a coding system based on a series of 3-minute scans of classroom activity, data were compiled for 151 Head Start classes. Analysis of the data indicated that while Head Start classes differed among themselves, certain generalizations could be drawn. Comparatively little time was spent on caretaking activities, aimless wandering, and teaching of pre-academic skills. A considerable amount of time was spent on structured waiting for the teacher, language development, dramatic play, and social interaction. Group vs individual activity differentiated classes and seemed dependent on teacher control. The next step will be to relate the obtained classroom characteristics to specific changes in children. With the causal relationships between process and product more clearly understood, more effective programs can be devised based on data rather than intuition. (MH)