Comparative studies aim to identify and explain differences of homologous phenomena in two or more contexts. Comparative studies of mathematics teaching and learning are undertaken with a variety of purposes and methods, and their results and interpretations remain the subject of fierce debates, especially in the case of large-scale quantitative surveys such as PISA. Even if the measurement of student performance is not its central preoccupation, didactic research is certainly concerned with these debates. This paper first proposes a model for clarifying the levels at which comparison is done, based on the anthropological theory of didactics. Using this model, we then propose a synthetic analysis of a selected body of international comparative research on mathematics education (from large-scale surveys to small-scale binary comparisons in doctoral projects) with the particular aim of understanding how the different studies differ and relate.