I argue that perception is characteristically non-conceptual and non-theoretically dependent while it is granted that observational reports are inferential in character and theory laden. Backing this viewpoint requires revising one of the most influential theses of the meta-scientific analysis of the nineteen sixties, the thesis that observation is theoretically loaded. On the other hand, if the contents of perceptions are non-conceptual we would be bound to accept, following Sellars, that these isolated and on their own do not provide justification for our beliefs – not even our beliefs of perception, which are made explicit by perception reports. Thus, we would have new reasons to reject fundationalist approaches. At any rate, it would not be necessary to advocate a purely coherentist approach