A structural model of mathematics achievement and attitude was tested with a national probability sample of 3,116 young adolescents from the Longitudinal Study of American Youth using structural modeling. A three-wave longitudinal design incorporated data from students, teachers, and parents to construct a prespecified theoretical model of mathematics achievement and mathematics attitude. A revised model provided a better fit than the original specification and than regression models used in 23 previous studies. Supported in cross-validation, the model revealed a complexity of direct and indirect effects not apparent in previous studies. Prior achievement and home environment influenced subsequent achievement most powerfully; motivation, exposure to extramural reading media, peer environment, and instructional exposure also had significant influences on achievement. Previous attitude had the most powerful influence on subsequent attitude, although the direct effects of instructional quality and the indirect effects of motivation and home environment were also notable. Appropriate teacher use of instructional time, thorough textbook coverage, and daily introduction of new material, although educationally alterable, are themselves influenced by previous student achievement. Similarly, instructional practices are significant alterable influences on mathematics attitudes, but such practices are themselves influenced by students' initial attitudes. Considerable research has sought to identify students' in- and out-of-school experiences that influence achievement and related outcomes, especially those that are alterable by policy intervention. Comparative studies continue to show the poor performance of U.S. students in mathematics and science, particularly in junior and senior high school. Large-scale surveys of the International Association for the Evaluation of Educational Achievement (McKnight et al., 1987) and the International Assessment of Educational Progress (Lapointe, Mead, & Phillips, 1989; Dossey, Mullis, Lindquist, & Chambers, 1988; Olson, 1990) confirm poor U.S. standings and also suggest underlying causes. Walberg (1981, 1986) has attempted to specify the chief productive factors in school learning suggested by several research literatures. The nine-factor specification, derived from the Cobb and Douglas (1928) theory of national economic productivity, guided many quantitative syntheses of classroom research and secondary analyses of large national surveys. The present study builds on this work by testing the productivity model with structural equation modeling. It employs national, longitudinal data on middle-school students' growth in mathematics achievement and attitude as related to variations in their exposure to instruction and to educationally stimulating psychological environments within and outside school.