This paper relates the recent proposal that modeling is a distinct theoretical strategy based on indirect representation and analysis to the current discussion on scientific representation. I will argue that both the thesis of indirect representation and the pragmatist approach to scientific representation effectively decouple the model-target dyad, which has been the basic unit of analysis in the discussion on models and representation. As a result of this decoupling models could be considered independent objects with no predetermined ties to real-life systems. This, in turn, poses the question of how we are supposed to gain understanding and knowledge through models. I will suggest that this characteristically happens through indirect reasoning, in other words reasoning from the results derived from the models to the underlying mechanisms or structures of interest. I argue that this also accounts for the modal nature of models.