Neuroscience offers more than new empirical evidence about the details of cognitive functions such as language, perception and action. Since it also shows many functions to be highly distributed, interconnected and dependent on mechanisms at different levels of processing, it challenges concepts that are traditionally used to describe these functions. The question is how to accommodate these concepts to the recent evidence. A recent proposal, made in "Philosophical Foundations of Neuroscience" (2003) by Bennett and Hacker, is that concepts play a foundational role in neuroscience, that empirical research needs to presuppose them and that changing concepts is a philosophical task. In defending this perspective, "PFN" shows much neuroscientific writing to be dualistic in nature due to our poor grasp of its foundations. In our review article we take a different approach. Instead of foundationalism we plead for a mild coherentism, which allows for a gradual and continuous alteration of concepts in light of new evidence. Following this approach it is also easier to deal with some neurological conditions (like blindsight, synaesthesia) that pose difficulties for our concepts. Finally, although words and concepts seem to seduce us to thinking that many skills and tasks function separately, it is language skill that--as neuroscientific evidence shows--co-emerges with action/perception cycles and thus seems to require revision of some of our central concepts.