Invited Symposium 1 - Cognitive functions of replay

Invited Symposium 1: Sunday, March 30, 2025, 10:00 am – 12:00 pm EDT, Grand Ballroom

Chair: Anna Schapiro¹; ¹University of Pennsylvania
Presenters: Anna Schapiro, Matthijs van der Meer, Nathaniel Daw, Helen Barron

During periods of sleep, awake rest, and pauses amidst active behavior, the brain reactivates memories of previous experiences. What is the function of this offline replay? Replay appears to be far from a simple recapitulation of recent experience, both in the way that experiences are prioritized for replay and in the content of the reactivation. Observations about the complex characteristics of replay have led to diverse theories of replay’s function, including that it plays important roles in memory consolidation, planning, decision making, and reinforcement learning. The speakers in this symposium will lay out a range of recent views on the functions of replay, with some contrasts for discussion.

Presentations

Replay for transformation and integration

Anna Schapiro¹; ¹University of Pennsylvania

The canonically-assumed function of replay is to strengthen recent memories for long term storage and use. But a simple strengthening account does not explain how memories change over time or how they are integrated with our existing knowledge. I will present empirical evidence that replay drives memory transformation as opposed to just strengthening, that it results in knowledge abstracted away from the superficial features of memories, and that it can drive integration of related new and old memories. I will present computational models accounting for ours and others’ empirical results that support a memory-focused interpretation of replay’s function.

Unbalanced training regimes, task representations, and the function of replay

Matthijs van der Meer¹; ¹Dartmouth College

Despite many compelling experimental and theoretical results, we still don’t know the internal logic of what experiences are prioritized for replay. After a brief critical review where I evaluate the leading theories against the data, I will examine one particular puzzle: on tasks that feature unbalanced training regimes, rats paradoxically replay the less-experienced trajectory. To understand why, we simulated a feedforward neural network using either rich (structured representations tailored to task demands) or lazy learning (unstructured, task-agnostic representations). Rich, but not lazy, representations degraded following unbalanced experience, an effect that could be reversed with paradoxical replay. To test if this computational principle can account for the experimental data, we examined the relationship between paradoxical replay and learned task representations in the rat hippocampus. In two different data sets, we found an association between the richness of learned task representations and the paradoxicality of replay. Taken together, these results suggest that paradoxical replay specifically serves to protect rich representations from the destructive effects of unbalanced experience, and more generally demonstrate a novel interaction between the nature of task representations and the function of replay in artificial and biological systems.

Prioritized replay: theory and practice

Nathaniel Daw¹; ¹Princeton University

There are many hypotheses – but rather little direct evidence – about the function(s) of replay. We have suggested that which items the brain “chooses” to replay in different circumstances should be revealing about the goals driving those selections, and therefore bear on the question of function. I review several examples of this theoretical framework and recent results supporting the notion that nonlocal activity in hippocampus is judiciously allocated with respect to the animals’ ongoing goals.

Building internal models during periods of rest and sleep

Helen Barron¹; ¹University of Oxford

Every day we make decisions critical for adaptation and survival. We repeat actions with known consequences. But we can also infer associations between loosely related events to infer and imagine the outcome of entirely novel choices. In the first part of the talk I will show that during periods of rest, co-activation of hippocampal cells in sharp-wave/ripples represent inferred relationships that include reward, thereby “joining-the-dots” between events that have not been observed together but lead to profitable outcomes. I will show how this hippocampal mechanism appears to propagate to other brain regions, to build a hierarchical internal model. Building on these findings at the cellular level, I will then show the implications of this neural mechanism for behaviour. I will show that memory reactivation during periods of rest facilitates participants’ ability to perform novel inferences, with no effect observed for directly learned information. Together these studies suggest that computing new mnemonic links during rest/sleep provides an important mechanism to support adaptive behaviour.