A prevailing assumption underlying statewide performance-based assessments is that they not only serve as motivators in improving student achievement and learning, they also encourage instructional strategies and techniques in the classroom that are more consistent with reform-oriented educational outcomes. Given these high expectations, more comprehensive and direct evidence for the consequences of assessments (both negative and positive) need to be addressed. The purpose of this paper is to explore the relationship between changes in the scores from the Maryland School performance Assessment Program (MSPAP) science performance assessment from 1993 to 1998 and classroom instructional and assessment practices, student learning and motivation, students' and teachers' beliefs about and attitude towards the assessment, and finally, student characteristics. Using growth models estimated within a structural equation modeling (SEM) framework, several factors for each of these dimensions were observed to explain a significant amount of the variability in school performance. The paper discusses these factors as well as the design of evaluations that hope to study the impact of assessment programs on students, teachers, and schools. (Contains 3 figures, 3 tables, and 14 references.) (Author/SLD)