Conference on Cognitive Computational Neuroscience

Gridlike patterns have been implicated in human medial prefrontal cortex (mPFC) and entorhinal cortex for representing conceptual knowledge. It has been proposed that abstract concepts are organized in a spatial coordinate system, similar to those for mapping physical environment. However, constructing multi-dimensional concept spaces often involves determining the metric relationship between distinct, even incomparable, concepts (e.g., continuous measures for fruit diameters versus tartness). One possibility is that, rather than using real world metrics, the brain organizes continuous concepts with "adaptive scales", adjusting for the statistical distribution of concepts and maximizing representational precision with limited coding capacity. Using fMRI, we formally evaluate this possibility against a range of alternatives, by deriving the implications of concept scaling in the well-established fMRI signatures of gridlike codes for cognitive map. Our data showed that, despite differences in the actual concept metrics, the mPFC exhibited the highest gridlike effect when the 2D concept space was built on the putative adaptive scale, in comparison to alternative scales. These results point to an organization principle for flexible, context-dependent construction of abstract cognitive maps.