This report describes a study of the relationship of instructional process and program organization to pupils' learning in Title I compensatory education projects, as measured by the Stanford Reading Test. This is the first attempt to apply economic input/output methodology to compensatory education. Personnel in 42 projects in 37 California school districts were interviewed to obtain detailed data on teaching strategies, individual instruction time per pupil, intensity of instruction, patterns of coordination of project personnel, and other variables. Variables were related to pupils' monthly gain in grade equivalents via multiregression techniques, holding program length and beginning score constant. Results contradict reports that compensatory education is ineffective. Individual instruction by trained reading specialists was consistently related to gains. Less strongly related were staff planning time and individual instruction by classroom aides. The six best projects averaged at least 1.25 months' learning per month of instruction. None were large or urban, all had small group instruction by specialists, high ratio of managers to pupils, and planning coordination. (MBM)