The aim of the present study is to provide further evidence that the errors that arise from improper application of the linear model are not random and not easy to overcome. Using three different types of test, we attempt to show that the errors referred to in the literature as "pseudo-analogous" are the result of an epistemological obstacle--"linearity". The results of this study reveal that students' erroneous application of proportional reasoning in non-proportional problem situations is not a random phenomenon. On the contrary, this phenomenon resists, persists, and reappears regardless of students' grade and of the tests' settings. This is evidence that pseudo-analogous errors result from the epistemological obstacle of linearity. (Contains 3 tables and 4 figures.)