English
In this article two modes of knowing the social world are opposed, a descriptive one linked to realism of effects, and an explicative one linked to realism of causes. The first is based on models of total complex phenomena, without any consideration to the real processes at play. The second is based on models of real processes, with no aim to reproduce total observed phenomena. The practical impossibility to translate realistically, both causes and effects of real social processes permits one to clarify the problem of the criteria of validity of models in social sciences.