Thread by @jacobbellmund

I just published a paper in @sciencemagazine that I have been working on with Peter Gärdenfors, Edvard Moser & @doellerlab. Since I love other people’s threads summarizing their science, I’ll try to return the favor: [1/n] science.sciencemag.org/content/362/64…

We connect a theory from cognitive science with findings from research on how the hippocampal-entorhinal region maps the space around us for navigation

Okay, this gif might be slightly misleading. It's more this kind of #hippocampus

Place cells in the #hippocampus signal an animal’s position in space.
Image: Wikimedia Commons

Grid cells in the entorhinal cortex are at least as amazing. They fire in this regular, hexagonal pattern. These cells were discovered @KISNeuro in the lab of Edvard Moser and @MayBrittMoser. @neurojosh and Z. Nadasdy have also recorded them in humans.
Image: Wikimedia Commons

To measure grid-like activity non-invasively in humans, we use hexadirectional signals in fMRI as a proxy measure

There are a number of studies on this by @doellerlab, @NeilBurgess10, @wolberslab, @epstein_lab, @behrenstimb, @alexa_tweety, @NauMatt , @T_N_Schroeder & @MatStangl (Plus others not on twitter)

Now, the amazing work by @alexa_tweety & @behrenstimb shows these grid-like signals also when the human brain processes trajectories through a bird space where the two dimensions are the length of the birds’ legs and neck

Likewise, findings from research in animals demonstrate that place and grid cells can encode other things than just space for navigation.

This includes time, sound frequency, positions in visual space or the positions of other animals in space.
Too many great scientists involved to name all here, I'm afraid

We apply a theory from cognitive science to these findings. It was originally developed by our co-author Peter Gärdenfors from @lunduniversity in Sweden

Cognitive spaces are spanned by feature dimensions. These dimensions need to satisfy some geometric constraints, for example 'betweenness'. This allows axioms like: If B is between A and C and C is between B and D, then B has to be in between A and D

In a cognitive space, each object has a position based on its values along the feature dimensions. So similar objects are located close together and dissimilar objects are far apart

Now the idea is that place cells and grid cells provide a neural code to map the feature dimensions of cognitive spaces, similarly to how they map physical space for navigation

Such a spatial representational format to organize knowledge and experience can be very useful. We can place new stimuli in the cognitive space based on their features and infer how we should respond to them

There are more coding principles of the hippocampal-entorhinal region that could be useful for cognitive spaces:

There is a gradient in how fine the code of place cells and grid cells is. This might help us to remember both the details and think big picture

Place are very flexible: They can remap to create maps for new spaces

Another idea we describe based on sequential coding in the hippocampus: We can simulate trajectories through cognitive space. For example, to plan which action takes us to a goal

If this sparked your interest, you can read the paper here science.sciencemag.org/content/362/64… or check the press release on the @doellerlab website: doellerlab.com/cognitive-spac…