How do people acquire new concepts? Most theorists (including Quine, Chomsky, Fodor, and many others) assume that childhood learning is a kind of theory-building. This picture implies that children acquire new concepts by deploying concepts they already possess, and that in turn implies what I call the Conceptual Mediation Thesis (CMT): that, in order to acquire any new concepts, a cognizer must first already have some concepts.

I argue that CMT is false. While CMT implies that at least some concepts are innate, it is widely accepted because it is thought to provide the only way to explain how concepts are acquired. However, I argue that the apparent explanatory virtues of CMT are in fact illusory. I then show how we can satisfy the explanatory goals that CMT was supposed to satisfy without postulating any innate concepts—indeed without any innate mental representations at all.

I distinguish between indicating states and representing states of cognizers. Indicating states differ from representing states in being stimulus-bound: only those tokens directly caused by what they indicate count as correct. I argue that perception produces states that indicate features of the environment. These indicating states serve as input to mechanisms that record these states. These recording devices, in turn, respond to the input of systematically similar indicating states by creating states that represent what those indicating states merely indicate. I describe some processes whereby these recording devices can create representational states without any representational input. I argue that this explanation requires no appeal to mental representations that the agent already possessed. Finally, I show that this approach to concept acquisition has the resources to explain a variety of psychological phenomena that traditional views struggle to accomodate.