Teacher attitudes (inputs) and classroom practices (processes) were used to predict residual student achievement scores (outputs). A combination of three approaches were used to control for student background characteristics and isolate the effects of school variables: (1) a "residual" measure of achievement; (2) an "input-output" or "production function" using both student and school background characteristics as inputs and achievement scores as outputs; and (3) "process-outcome" or "process-product" controls. A causal modeling effort in Phase I of the Seattle School Effectiveness Project in 1979-80 produced a "school effectiveness index," designed to be free of influence from a few major student characteristics (family income, initial achievement and ethnic classification) which are beyond the school's control, yet relate to student achievement gains. Phase II identified school characteristics which are correlated with the index. Phase III findings in 1981-82 were compared with Phase II results in 1980-81, and an explanatory causal model is proposed which examines reading, mathematics and language achievement gains. (Author/CM)