Abstract

It is traditionally thought that integration in cognitive science requires combining different perspectives, elements, and insights into an integrated model or theory of the target phenomenon. In this paper I argue that this type of integration is frequently not possible in cognitive science due to our reliance on using different idealizing and simplifying assumptions in our models and theories. Despite this, I argue that we can still have integration in cognitive science and attain all the benefits that integrated models would provide, without the need for their construction. Models which make incompatible idealizing assumptions about the target phenomenon can still be integrated by understanding how to draw coherent and compatible inferences across them. I discuss how this is possible, and demonstrate how this supports a different kind of integration. This sense of integration allows us to use collections of contradictory models to develop a consistent, comprehensive and non-contradictory understanding of a single unified phenomenon without the need for a single integrated model or theory.