The interpretation and analysis of information is a crucial aspect in the use of school performance feedback (SPF) (Visscher, 2002). Examination of existing SPF systems reveals a lack of systematic research focusing on user comprehension in relation to the representation of feedback information. This is critical considering the fact that much SPF focuses on complex concepts like 'learning gains' and 'value-added'. Therefore, a study has been set up to study the impact of different ways to represent these concepts to a non-statistical target audience. The design principles, derived from Mayer's cognitive theory of multimedia learning (Mayer, 2001) and the representations used in existing SPF reports have influenced the design of the present study. Bloom's taxonomy (Anderson & Krathwohl, 2001) was used as framework for developing test items measuring user level of understanding. Six different ways to feedback achievement progress and value-added were evaluated in an experimental study. The entire population of freshmen in educational sciences at Ghent University (n = 346) was asked to take the role of a primary school principal and read a SPF report. Subjects were randomly assigned to one of the six experimental conditions. The SPF report was presented to each of the subjects individually online in a computer lab, allowing 50 students to participate simultaneously.The six experimental conditions reflect three different representation modes and two ways to explain the concept of value-added. The representation modes consisted of a baseline version using text and graphics, and two separate elaborated versions enriched with either tables or symbolic representations of underlying statistical concepts. Immediately after the presentation of the feedback the students completed an online test, assessing both their conceptual understanding and their skill in deriving information from graphs and tables. The test consisted of two parallel parts. In the first part it was not allowed to look back at the presentation, whilst they could do so in the second part. The experiment was organized over six sessions. Therefore parallel versions of the tests were developed and IRT-calibrated to obtain comparable test scores. Performance differences between conditions are tested using univariate and multivariate analysis of variance, controlling for potentially confounding variables (gender, age, maths level during previous educational experiences and attitude towards statistics). Additional analyses will focus on in-depth qualitative analysis of errors.The research data are currently being analysed and a full report will be available at the conference. The results of the study will answer the question of the differential effect of different ways of explaining concepts and use of different representation modes when feeding back school performance information to a non-statistical audience. Building on these results, future directions for research will be presented in the view of designing a new school performance feedback system.Anderson, L.W., & Krathwohl, D.R. (2001). (Eds.). A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom's Taxonomy of Educational Objectives. New York: Longman. Mayer, R. E. (2001). Multimedia learning. Cambridge, MA: University Press. Visscher, A. J. (2002). A framework for studying school performance feedback systems. In A. J. Visscher, & R. Coe (Eds.), School improvement through performance feedback. Lisse: Swets & Zeitlinger.