This study applied a model of student academic motivation to an existing set of classroom data from 323 third graders in 17 classes from 2 Stanford, California elementary school districts. The model proposed that common forms of student-motivated behavior, such as task engagement, are systematically related to students' cognitive structures and their active inferences and interpretations as they initiate academic tasks. The application was a mechanism for increasing the understanding of a complex set of classroom variables and assessing hypothetical predictions from the model. The results showed that, within a given class, the higher the ability, the more self-esteem students exhibited relative to others in the class; that contextual factors had minimal influence on students' self-esteem; and, that changes in students' self-esteem did relate to their year-end academic performance. The analyses showed little variation across classes between pre- and post-test self-esteem. Results showed predicted relationships among student verbal ability, internality, self-esteem, and academic performance, as well as evidence that the average ability level of the class may alter such relationships. A more precise specification of cyclical models and an attempt to operationalize variables according to some real-world correspondence should permit more controlled assessments. (Author/PN)