The PISA survey influences educational policies through an international competitive process which is not wholly rationally-oriented. Firstly, PISA league tables act normatively upon the definition of formal educational aims while the survey tests cannot evaluate the educational systems' relative strengths with regards to such aims. We argue that they measure a kind of academic potential of the students. Secondly, errors in interpreting national success or failure stem from the causal inferences developed from observing national cases. In order to give such comparisons a stronger basis, we distinguish five main educational models within the OECD countries and compare their PISA 2006 results to their population's socio-cultural levels as well as their performance recorded in an academic-programme-based survey such as TIMSS. Our conclusions contrast heavily with the usual lessons inferred from PISA, which follow OECD-promoted guidelines, and explain French students' weakening as well as Finnish students' success. (Contains 8 tables and 22 notes.)