In current interaction design research there is a widespread belief that situated action and embodied interaction should replace mental representations in the theoretical account of human cognition. This exclusion of representation is however diagnosed as a sign of representation-phobia by Anderson (2003) who claims that it is misguided. This paper aims to show why and how it can be overcome. Initially, a literature review will show how representation-phobia manifests itself through two different versions in HCI research. On the basis of this I argue that representationphobia leads to a theoretical dead end. Then, by drawing on semiotics and recent findings from cognitive research, I argue that we cannot understand the rich complexity of embodied interaction unless we furnish our thinking with a dynamic notion of representation.