One of the most striking features of the hippocampal network is its ability to self-generate neuronal sequences representing temporally compressed, spatially coherent paths. These brief events, often termed "replay" in the scientific literature, are largely confined to non-exploratory states such as sleep or quiet rest. Early studies examining the content of replay noted a strong correlation between the encoded spatial information and the animal's prior behavior; thus, replay was initially hypothesized to play a role in memory formation and/or systems-level consolidation via "off-line" reactivation of previous experiences. However, recent findings indicate that replay may also serve as a memory retrieval mechanism to guide future behavior or may be an incidental reflection of pre-existing network assemblies. Here, I will review what is known regarding the content of replay events and their correlation with past and future actions, and I will discuss how this knowledge might inform or constrain models which seek to explain the circuit-level mechanisms underlying these events and their role in mnemonic processes.