A previous structural model of Walberg's theory of educational productivity (Reynolds & Walberg, 1991) was tested with a national probability sample of 2,535 10th graders for science achievement and attitude. Using data from the Longitudinal Study of American Youth, a 3-wave design incorporated information from students, teachers, and parents. Results indicated that a revised mediated-effects model fit the data best and accounted for substantial variance in Grade 11 science achievement (56%) and attitude (44%). The variables prior achievement, home environment, exposure to mass media through reading, and instructional time had the greatest total effects on science achievement. Prior attitude, home environment, motivation, and prior achievement made the greatest total contributions to science attitude. Although there were different weightings of the factors between the present study and Reynolds and Walberg (1991), both studies support a mediated-effects model of educational productivity.