LAUSR

Much of higher cognition involves abstracting away from sensory details and thinking conceptually. How do our brains learn and represent such abstract concepts? Recent work has proposed that neural representations in the medial temporal lobe (MTL), which are involved in spatial navigation, might also support learning of higher-level knowledge structures (Behrens et al., 2018; Bellmund et al., 2018). Under this view, a range of MTL neurons such as place cells, grid cells, and headdirection cells may support the ability to mentally “navigate” through conceptual spaces. This extends the original proposal by Tolman (1948) that people construct “cognitivemaps” that support broad psychological functions, and offers the exciting potential of understanding the cognitive processes that underlie category learning, reinforcement learning, and spatial navigation under a single unified framework. These ideas are supported by findings that neural representations in the MTL, as well as the medial prefrontal cortex (mPFC), are involved in “navigation” of simple two-dimensional spaces of visual stimuli (Constantinescu et al., 2016; Theves et al., 2019, 2020), social spaces (Tavares et al., 2015; Park et al., 2020), and odor spaces (Bao et al., 2019). A recent study in the Journal of Neuroscience (Viganò and Piazza, 2020) takes this research further by suggesting that the entorhinal cortex (EHC) and the mPFC are capable of mapping not only sensory spaces, but also abstract semantic spaces. In this opinion piece, we first describe the paradigm and results of Viganò and Piazza (2020), as well as the importance of their findings for the field. We then raise several methodological concerns and suggest changes to the paradigm to address these issues. Finally, we discuss potential future research directions including experimental and modeling approaches to tackle outstanding questions in the field.