Abstract
One view is that conceptual knowledge is organized using the circuitry in the medial temporal lobe (MTL) that supports spatial processing and navigation. In contrast, we find that a domain-general learning algorithm explains key findings in both spatial and conceptual domains. When the clustering model is applied to spatial navigation tasks, so-called place and grid cell-like representations emerge because of the relatively uniform distribution of possible inputs in these tasks. The same mechanism applied to conceptual tasks, where the overall space can be higher-dimensional and sampling sparser, leading to representations more aligned with human conceptual knowledge. Although the types of memory supported by the MTL are superficially dissimilar, the information processing steps appear shared. Our account suggests that the MTL uses a general-purpose algorithm to learn and organize context-relevant information in a useful format, rather than relying on navigation-specific neural circuitry.
Highlights
Previous work has explained a wide array of learning and memory phenomena in terms of clustering computations supported by the MTL23. This same basic account was shown to account for basic spatial navigation phenomenon, including place and grid cell-like response patterns
We showed that a learning mechanism that seeks to minimize error in the task-relevant feature space captures conceptual structure in concept learning tasks and spatial structure in two-dimensional navigation contexts, which lead to place and grid cell-like representations
Rather than spatial mechanisms providing a scaffolding for more abstract conceptual knowledge[3,10], the current results suggest that key findings in the spatial literature naturally arise as limiting cases of a more general concept learning mechanism
Summary
A common learning mechanism for space and concepts. As shown in Fig. 1a, the model when applied to categorizing animals as birds or mammals learns to segregate the items into two groupings. One might object that our account is inconsistent with conceptual learning brain imaging studies that find grid-like response patterns[12] These studies are consistent with the model because they follow the design principles of typical spatial studies—all feature combinations within a 2-dimensional stimulus space are sampled, which would lead to a hexagonal clustering solution (Fig. 2b). Krupic et al.[36] identified grid cells in rodent mEC in a square box, placed the animals in a trapezoid environment They found that activity maps of grid cells became less grid-like in the trapezoid and that the decline was greatest for responses elicited on the narrow side of the trapezoid. As in the empirical studies, the model’s overall grid scores declined in the trapezoid environment (Fig. 4e; trapezoid mean grid score: 0.058, bootstrap CIs [0.054, 0.061]; Fig. 4f; square minus trapezoid mean: 0.219, bootstrap CIs [0.214, 0.224]) and the grid scores were higher on the wide than on the narrow side of the trapezoid (Fig. 4g; wide minus narrow mean: 0.133, bootstrap CIs [0.127, 0.139]; see Tables S5–S7)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
