This talk addresses the question: how does memory help us learn?  Aronowitz begins by re-thinking the epistemic problem that memory systems solve in light of memory successes and failures in humans, rodents, and artificial systems. Rather than merely functioning to store information or to preserve justification, Aronowitz argues that the core function of any memory system is to support accurate and relevant retrieval. This problem formulation has consequences for which structures and mechanisms make up a memory system. In brief, memory systems are modeling systems. This means that they generate, update and manage a series of overlapping, simplified, relational representations that map out features of the world. Succeeding at building and maintaining models requires the kind of active knowledge generation traditionally associated only with deliberative reasoning.