The 2003 PISA survey focused on mathematics literacy skills and problem solving in the numerous participating countries. The aim was not only to assess students' ability to perform mathematical operations, but rather to assess how well they can recognize, formulate and tackle mathematical problems in the context of real life. In Turkey, it was observed that between school variation is so large among the schools that it exceeds twice the OECD average. When the data were analyzed in more depth, remarkable performance differences were found between different types of school. This stimulated research into possible explanations for the differences in terms of school and student characteristics, looking for connections between student attainment and various factors explored in the school and student questionnaires: school climate, teachers' expectations, student motivation to learn, etc. The results of the study are expected to provide relevant information to education policy makers in Turkey, to help close the gap among the schools, and eventually to enhance the overall quality of the low performing schools. The presentation will describe the results of application of a discriminant function used to identify the variables that characterize the high and low performing schools, complemented by supportive evidence from the school questionnaires. Results will include both descriptive information and interpretations of the discriminant function characterizing the school differences.