Data were collected in a teacher effectiveness study on the interaction of teachers and first grade students while the students participated in reading groups within intact classes. Eighteen teacher behavior variables drawn from the data were then related to student achievement using three units of analysis: class means, reading group means within classes, and student scores within reading groups. The results were then compared across the three sets of analyses. It was found that the pattern of results differed depending upon the unit used in the analyses. The relationships of teacher behaviors to student achievement were frequently inconsistent across within-class and between-class analyses. Relationships that occurred at the class level of analysis did not always occur within classes, and relationships that occurred within classes at the student-within-reading-group level of analysis frequently did not occur at the reading-group-within-class or class levels of analysis. It was concluded that a combination of both within-class analyses and between-class analyses provided a better understanding of the relationship of teacher-student interactions to student achievement than did either within-class or between-class analysis alone. (Author/FL)