When we looked at the model weights, we found that voxelwise peaks were distributed across all directions (like a compass), but the voxelwise troughs tended to be aligned with the main axis of the space. The latter result was unexpected and we think it’s pretty cool!

These heading codes generalized across perceptually different versions of the city, different locations in the city, and different phases of the taxi-cab task—thus fulfilling the expected characteristics of a “neural compass”.

We found heading-related response in retrosplenial complex (RSC) and superior parietal lobe (SPL). This is nicely consistent with previous studies that observed heading codes in these regions using trial-based designs.

How do we keep track of our sense of direction during navigation? Head direction cells in rodents are well-studied, but what about humans? To answer this question, we used an encoding model to analyze fMRI responses while participants performed a “taxi-cab” task in a VR city.

Turnout at the Philly @standupforscience.bsky.social march was incredible! I'm not great at measuring crowds, but this was easily >1k. Wish I could have gotten an elevated view.