Past research on classroom environment was extended to science laboratory class settings in an investigation of associations between student outcomes and classroom environment, The sample consisted of 1,594 senior high school chemistry students in ! X2 classes. The Science Laboratory Environment Inventory was used to assess student cohesiveness, openendedness, integration, rule clarity, and material environments in the laboratory dass, and student outcomes encompassed two inquiry skill and four attitude measures. Simple, multiple, and canonical analyses were conducted separately for two units of analysis (student scores and class means) and separately with and without control for general ability. Past research was replicated in that the nature of the science laboratory classroom environment accounted for appreciable proportions of the variance in both cognitive and affective outcomes beyond that attributable to general ability.he strongest tradition during the previous quarter of T a century of classroom environment research has involved investigation of the predictability of students' cognitive and affective learning outcomes from their perceptions of classroom environment (Fraser, 1989; Fraser, in press;Fraser & Walberg, 1991). Past research provides convincing and consistent support for the predictive validity of student perceptions in accounting for appreciable amounts of variance in learning outcomes, beyond that attributable to student characteristics such as general ability. This pattern has been replicated in numerous countries and for a variety of outcomes and classroom environment instruments.The present study of associations between students' outcomes and their perceptions of classroom psychosocial environment was distinctive in several ways. First, we used the Science Laboratory Environment Inventory (SLEI) in the first study of outcome-environment relationships specifically in laboratory classroom settings. Second, to permit comparison with results from method-
BARRY J. FRASER Curtin University of Technologyologically diverse past studies, we analyzed our data in several different ways (i.e., simple, multiple, and canonical correlation analyses). Third, we compared the magnitudes of environment-outcome relationships for two units of analysis, namely, the individual student's score and the class mean score. Fourth, we performed all analyses separately with and without control for student general ability.