Recognition memory enables us to discriminate whether an event has occurred in the past, and is widely interpreted to reflect the conscious retrieval of episodic traces or familiarity. Non-conscious mnemonic influences, such as repetition priming, are thought to have a negligible effect on standard tests of recognition memory. A major difficulty with this conclusion is that it is exclusively based on the results from experimental protocols that use stimulus materials available to conscious perception. In eight experiments (N = 144), we tested the necessity of mechanisms related to conscious perception for accurate recognition memory by manipulating observers' awareness of either the encoded event and/or the retrieval cues. Remarkably, observers made accurate objective and subjective recognition memory-guided judgments without visual awareness of the encoded events, retrieval cues or, most strikingly, both. These results demonstrate that non-conscious processes can drive accurate recognition memory, and are a significant challenge to neurobiological accounts centered on the conscious retrieval of episodic traces or familiarity.