Paradoxical replay can protect contextual task representations from destructive interference when experience is unbalanced.

We provide an explicit normative explanation and simulations of the experimentally observed puzzle of "paradoxical replay", which we show can serve to protect certain task representations from the destructive effects of unbalanced experienceWe confirm with new analyses the main prediction of the theory, that "rich" task representations, measured using representational distance in the rodent hippocampus, show more paradoxical replay compared to "lazy" task representationsOur theory refines the notion of consolidation in complementary learning systems theory in showing that not all task representations benefit equally from interleaving, and provides an example of how the use of replay in artificial neural networks can be optimized.