We model CA3 as a recurrent autoencoder recalling sensory experiences from noisy and partially occluded observations. In this formulation, an agent’s experiences during exploration define a low-dimensional manifold embedded in a high dimensional experience space. We show that training a network to pattern-complete these experiences causes spatially localized firing fields, i.e., place cells, to emerge. The emergent place fields reproduce key aspects of hippocampal phenomenology: a) remapping (maintenance of and reversion to distinct learned maps in different environments), implemented via repositioning of experience manifolds in the network’s hidden layer, b) orthogonality of spatial representations in different arenas, c) robust place field emergence in differently shaped rooms, with single units showing multiple place fields in large or complex spaces, and d) slow representational drift of place fields. These results arise because continuous traversal of space makes sensory experience temporally continuous. We make testable predictions: a) rapidly changing sensory context will disrupt place fields, b) place fields will form even if recurrent connections are blocked, but reversion to previously learned representations upon remapping will be abolished, c) the dimension of temporally smooth experience sets the dimensionality of place fields, including during virtual navigation of abstract spaces. In this formulation of the place system, a change of behavioral or environmental context causes the encoded experience manifold to reposition within the network activity space. Thus the context can be decoded from these manifolds provided they do not intersect. We show that this leads to a tradeoff between location resolution and contextual capacity.