Retrospective Grid Cells in the Medial Entorhinal Cortex Encode Past Paths During Spatial Navigation

Neurons dynamically switch between prospective and retrospective coding depending on task demands. Prospective coding optimizes future actions, while retrospective coding helps learn from past experiences. The grid cells in the medial entorhinal cortex (MEC) play a key role in spatial navigation, with periodic firing fields to encode current position. While their role in prospective coding has been reported recently, whether and how grid cells exhibit retrospective coding (encoding past paths) remains unclear. Here, we combined large-scale two-photon imaging in freely moving mice with well-established analytical approaches to identify and characterize retrospective grid cells in the MEC. We identified the retrospective grid cells, which encoded path history, showing maximal gridness score at backward-shifted positions (∼15 cm), independent of calcium transient kinetics. Retrospective grid coding showed reduced spatial precision in darkness and was disrupted during assisted movement, indicating that its stable expression depends on sensory and behavioral conditions. In simulations, incorporating retrospective information into a path-integration model reduced accumulated position error and improved model performance across varying movement speeds. These findings support a dual-coding framework where retrospective and prospective cells jointly optimize navigation accuracy.