The bounds of representation

Based on an endorsement of the hypothesis of extended cognition (and, more broadly, of distributed cognition), this paper proposes a criticism of the representationalist assumptions that still pertain to these contemporary models of cognition. I first rehearse some basic problems akin to any representationalist model of cognition, before proposing some more specific arguments directed against the necessity, the plausibility, and the coherence of the marriage between extended cognition and contemporary representationalism (not necessarily a symbolic one). Extended and distributed models of cognition have the resources to get rid of representationalism, and they should better do it. Their adherence to representationalism might be an (illusory) by-product of the extended character of the scientific study of cognition.