“Cognitive maps as a medium for thought“
- Cognitive maps and propositional representations (the ‘language of thought’) are different from each other but share some key characteristics, such as compositionality.
- To be useful for thinking, cognitive maps must interface with language of thought-like representations. This may involve the same operations that are used to convert cognitive maps into routelike representations during navigation.
- Dynamic neural mechanisms such as replay, theta sequences, and vector computation are possible candidates for mediating this interface.
Cognitive maps are essential for navigation, but to what extent are they involved in thinking more broadly?
In this opinion article, we examine whether cognitive maps constitute a genuine vehicle for thought and how they relate to the notion of a language of thought. Integrating work from philosophy, psychology, and neuroscience, we suggest that map-like representations offer a distinct, flexible medium for encoding information and exhibit compositionality in both spatial and nonspatial domains.
We then consider how dynamic neural mechanisms—replay, theta sequences, and vector computation—may transform map-based representations into sequential routelike formats that interface with cortical systems supporting propositional thinking.
Cognitive maps may be complementary to symbolic cognition, jointly enabling flexible human reasoning.

(A) Different types of representations are plotted with respect to the fidelity of their semantic constituents (i.e., faithful to the physical world vs. arbitrary) and how those constituents syntactically combine (i.e., spatial isomorphism vs. abstract structure).
(B) A series of propositions describing the spatial layout of an area of Philadelphia (top) is compared with a representation of the same area in the form of a map (bottom). The spatial relationships among the elements are difficult to determine from the propositions but are readily apparent from the map.
(C) A depiction of a translation from an abstract map to a specific planned route is shown. The map represents the allocentric relationships between all the elements in the environment; the route represents the sequence of elements that one encounters and the spatial choices (e.g., turns) that one must make to reach a specific destination.

Replay, theta sequences, and vector computation are mechanisms for cognitive map readout in the spatial domain.
Each mechanism exhibits some evidence for compositionality, suggesting that they are all candidates for supporting compositional thinking (in collaboration with cortical mechanisms). However, it remains to be seen whether any of them are necessary or sufficient for compositional thought.
Cognitive maps are a critical tool in our cognitive toolkits—an instrument for representing complex, multidimensional information in an efficient, interpretable manner.
In this way, we believe that cognitive maps are a proper medium for thought, distinct from but interfacing with a putative LoT (Language of Thought).
A growing body of work is taking this possibility seriously, and we have attempted to provide a succinct overview of this literature here. While uncertainty remains about how maps support compositional thinking and how they interact with other media of thought, future work in cognitive science, particularly in cognitive neuroscience, may do well to examine this interface directly.
