Description: The recent international assessment studies of mathematics -TIMSS 1999, PISA 2000 and PISA 2003 - have offered an opportunity to analyse the Finnish educational system in a variety of ways. The assessment results have revealed that our comprehensive school system can produce mathematics performances which are highly equal and of a relatively high standard. An examination of factors linked to students' performance enables us to identify strengths and weaknesses of our mathematics education and, hopefully, also to discover ways of improving it. This study will explore the relationships between relevant background characteristics from student and school level and Finnish students' performance in mathematical literacy based on the PISA 2003 data. Some comparisons regarding the role of different background factors will also be made between Finnish-speaking and Swedish-speaking students. Conceptually, the research is based on theories and models about educational effectiveness (e.g. Reezigt et al., 1999). Two research questions will be explored: 1) What are the factors describing student and home background and school environment to be related to students' mathematical literacy performance? and 2) What kind of differences there are when examining the explanatory models of the Finnish-speaking and Swedish-speaking student groups? Multilevel modelling techniques have become increasingly available to researchers for analysing hierarchically structured data (Bryk & Raudenbush, 1992; O'Dwyer, 2005). The strength of multilevel models in dealing with school data is based on the feature that they not only take account of the inherent structure of the data but also treat variables of different levels simultaneously within the same model. In recent years, hierarchical linear modelling (HLM, MLwiN) techniques have been taken into wide use (e.g. Howie, 2002). In this study the HLM software are used to formulate a two-level regression model. Further, there is an extensive empirical research that can be applied when selecting the potential factors related to mathematics performance (e.g. Bos & Kuiper, 1999; Fullarton & Lamb, 2000). The Finnish data comprises test answers and background information from a total of 4589 Finnish-speaking students (15-year-olds) and 150 school principals, and 1207 Swedish-speaking students and 47 school principals. We expect that secondary analyses of the PISA 2003 data by means of multilevel modelling will further specify the field of explanations for students' mathematical literacy performance. Because in PISA 2003 more attention was paid to measuring students' attitudinal aspects there are strong reasons to believe that our analyses would reveal new explanatory factors and important connections. In addition, it is extremely useful to discuss the consequences of the findings from the developmental perspective of mathematics education.