The Branching Layers of Memory Updating

Memory is built through recurrent learning and updating events. Reconsolidation, triggered by mismatches between retrieved memories and current circumstances, destabilizes stored traces and induces gene expression required for restabilization. During this labile state, memories can incorporate new information or lose content if restabilization fails. Since memories can undergo multiple reconsolidation events, tracking the effect of impairing restabilization across successive reactivations allows studying the architecture of updated memories. Implementing this strategy with two reactivation sessions in the spatial object recognition task, in rats, we observed that each reconsolidation event added new information without merging it into a single trace, but by building a scaffold of partially overlapping information layers: disrupting an updated layer spared basal ones, whereas impairing the learning layer also affected subsequent one. Adding a third reactivation session confirmed these phenomena and showed that impairing the learning layer extends amnesia to the two following layers. Yet, the layer scaffolding does not always follow a simple temporal sequence. If the remainder associates with the learning and its first update, the new update links to the original learning information rather than to its update. Thus, memory organizes in layers but depending on the reminder, these layers seem to branch from different memory representations, rather than a linear stack over time