THE FEEDBACK

By Arieh Schwartz (London School of Economics) and Nir Fresco (Ben-Gurion University of the Negev)

Abstract

Cognitive maps play a crucial role in mammalian navigation. They provide the organism with information about its own location and the locations of landmarks within known environments. Cognitive maps have yet to receive ample attention in philosophy. In this article, we argue that cognitive maps should not be understood along the lines of conceptual mental states, such as beliefs and desires. They are more plausibly understood to be non-conceptual. We clarify what is at stake in this claim, and offer two empirically-informed arguments in its favor. Both arguments submit that cognitive maps are probably non-conceptual because their representational structure seems to differ from that of conceptual mental states.

You can find Arieh and Nir’s article in Synthese here: Mapping Content: Why Cognitive Maps are Non-conceptual Mental States (2025).

Commentary

In 2014, John O’Keefe, and May-Britt and Edvard Moser shared the Nobel Prize in Physiology and Medicine for discovering place cells and grid cells–central cellular components of the cognitive maps by which humans and other mammals are thought to navigate known environments. Cognitive maps are internal allocentric representations of space which encode landmark locations, distances, self-position, and heading. While cognitive map theory is a pillar of navigation neuroscience, philosophers have been relatively quiet about these representational posits. It remains unclear, for example, whether cognitive maps fit within the traditional philosophical framework for understanding beliefs, desires, and other conceptual mental states.

            On the traditional framework, conceptual mental representations are structured, much like language, from discrete repeatable concepts combined predicatively. This representational structure is meant to explain, among other things, why thinking is systematic, and how inference can be rational. Thinking seems systematic in that if one has the resources to understand one thought, (e.g., that John loves Mary,) it is all but guaranteed that one can understand others, (that John loves someone, that Mary loves John, and so forth). Because conceptual mental states have this structure, their contents can also be combined by logical operators and stand in deductively rational relations to one another. That said, not all mental states seem to have this representational structure; cognitive maps arguably do not.

            There are some influential arguments that cognitive maps are non-conceptual. For example, while one can easily apply logical operators to thoughts, (e.g., it is not the case that Albert is in Antarctica,) it is unclear how one might negate or disjoin cognitive map contents. We doubt that such arguments are decisive, though we agree with their conclusion. They rest on an analogy between cognitive maps and the more familiar sort of cartographic maps which can be found in an atlas. In ‘Mapping Content: Why Cognitive Maps are Non-Conceptual Mental States’ (Synthese), we aim to move the debate onto firmer empirical ground. Whereas conceptual mental states are thought to represent relations with discrete representational elements, (i.e., concepts,) we argue that cognitive maps have a different structure. They are best understood as structural-, or S-representations, which represent (some) relations between entities (e.g., relative spatial position, or relative heading,) not with discrete representational elements but via structural relations between such elements. 

            Some spatial relations seem to be encoded by means of the temporal structure in neural activity. Consider, for example, the phenomenon of phase precession. As an animal moves through a place cell’s place field, the timing of that cell’s spikes occurs at progressively earlier phases of the theta cycle so that temporal relations in neural activity systematically covary with spatial relations. This provides a mechanism by which spatial relations can be encoded via relations between neural events rather than discrete representational constituents. If cognitive maps are S-representations, as this and other examples suggest, then their representational structure differs from that traditionally ascribed to conceptual mental states. They are non-conceptual. 

            If correct, our view has important metaphysical and epistemic upshots. It makes it plausible that we share navigational mental states with creatures that lack our concepts. It also suggests that cognitive maps do not justify beliefs in the ways beliefs are often understood to justify each other (e.g., deductively). 

            More broadly, philosophical accounts of mental representation have often assumed a specific picture of representational structure, and a simple division between conceptual and non-conceptual mental states. As neuroscience provides increasingly detailed evidence about neural representations, we will likely need a more fine-grained taxonomy of representational structures.

Editor’s note: Please feel encouraged to share your thoughts and questions about this commentary and/or the article in the comment box below!