The mammalian hippocampus is responsible for storage and retrieval of episodic experiences and for computing their novelty. Memory formation requires novelty detection to conserve capacity and reduce oversampling artifacts. Novelty detection requires episodic memory to generate predictions for the future. The neural implementations of perceptions are patterns of neural spikes in which the spatial (across neurons) and temporal (across time) dimensions are both important. To understand how the structure of the hippocampus implements its function, models need to be constructed that process realistic patterns in realistic ways. This prospectus outlines a verbal theory of novelty detection and episodic memory based on the anatomy and physiology of the hippocampus. Recurrent collaterals within CA3 implement a resonance network. Novelty is computed within CA3 by drops in activity produced by mismatches between predictions and perceptions. CA1 "reads out" CA3 trajectories and implements episodic memory in the forward projection to the subiculum. Consideration is given to the importance of neuromodulator activity, temporally asymmetric learning rules, synaptic unreliability, and timing issues in "direct-indirect" architectures. The focus of the dissertation will be to translate the verbal theory into a mathematical model that can make quantitative psychological and neuroscientific predictions.