Variations in classroom instruction were observed, and related to student achievement in reading and mathematics. A six-construct model was applied to data on students and instructional practices in sixty second grade classrooms at nine geographic sites, seven of which were sponsored by Project Follow-Through. The constructs were: (l) initial cognitive ability of the students, (2) opportunity to learn criterion material, (3) motivators supporting student learning, (4) curriculum structure, (5) quality of instructional events, (6) end-of-year student achievement on the math and reading subtests of the Metropolitan Achievement Test. Two of the three major instruments in the Classroom Analysis Package (CAP)--the Teacher Interview/Questionnaire and the Verbal Instructional Behaviors coding scheme--and a videotape of teacher behavior in the classroom provided data on instructional practices. The results of the data analysis are presented in two stages: variance in end-of-year student achievement using (1) the first five constructs as predictors and (2) selected measures of the CAP as predictors. The construct model contributes to teacher effectiveness and evaluation research because it identifies which practices seem relevant for which gains, rather than simply stating that one program is more successful than another. (CP)