Between planning and map building: Prioritizing replay when future goals are uncertain.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• The computational function of hippocampal replay remains debated, with competing "value" (goal-directed planning) and "map" (environmental representation) hypotheses.
• Empirical data supporting either hypothesis are abundant, but differentiating them is challenging due to a lack of formal specificity, particularly for the map hypothesis.
• Recent findings of replay lagging rather than leading goal achievement pose challenges to the pure planning hypothesis.

Purpose of the Study:

• To formally extend a prominent planning theory to incorporate routes to future goals, effectively creating a "map" component.
• To reconcile the "value" and "map" hypotheses of hippocampal replay by proposing a unified framework.
• To investigate how the estimated likelihood of future relevance influences replay's preference for current versus future goals.

Main Methods:

• Extension of a prominent computational theory of planning.
• Inclusion of routes to future goals within the planning framework to simulate a map-like representation.
• Formal modeling to assess the conditions under which replay would favor current versus future goals based on predicted relevance.

Main Results:

• The extended model suggests that replay's preference depends on the estimated likelihood of future relevance.
• This framework unifies the planning and map hypotheses by demonstrating how they can arise from a single underlying mechanism.
• The model reconciles existing empirical data, including observations where replay appears to lag rather than lead goal achievement.

Conclusions:

• Hippocampal replay likely serves a dual function, integrating goal-directed planning with the construction of environmental representations.
• The apparent distinction between "value" and "map" functions may stem from the temporal relevance of potential future goals.
• A unified computational account reveals a deep relationship between planning and mapping in the hippocampus.