Abstract -- The hippocampus plays a critical role in spatial look-ahead, single-trial learning, memory consolidation, and imagination.   Each of these learning dynamics depends on memory schemas.  We have developed a hierarchical Bayesian inference approach learns the structure of behavioral tasks or schemas in order to predict future task observations.  Using this general inference framework we show how the approach can account for (1) the hippocampal dependent of single trial learning, (2) variable time memory consolidation, (3) hippocampal place cell mapping and remapping, and (4) spontaneous object exploration in the rodent literature.  Finally, we suggest how these results can be applied in human behavior and imaging research.