Back

Hippocampal Place-like Signal in Latent Social Space

Schafer, M.; Kamilar-Britt, P.; Sahani, V.; Bachi, K.; Schiller, D.

2022-07-18 neuroscience
10.1101/2022.07.15.499827 bioRxiv
Show abstract

During navigation, the hippocampus represents physical places like coordinates on a map; similar location-like signals have been seen in sensory and concept spaces. It is unclear just how general this hippocampal place code is, however: does it map places in wholly non-perceivable spaces, without locations being instructed or reinforced and during navigation-like behavior? To search for such a signal, we imaged participants brains while they played a naturalistic, narrativebased social interaction game, and modeled their relationships as a kind of navigation through social space. Two independent samples showed hippocampal place-like signals in both region-based and whole-brain representational similarity analyses, as well as decoding and average pattern similarity analyses; the effects were not explained by other measures of the behavior or task information. These results are the first demonstration of complete domain generality in hippocampal place representation. One-Sentence Summaryhippocampal place-like signal in non-perceivable and unreinforced space during naturalistic navigational behavior. Significance statementThe hippocampus is a brain structure known to encode maps of physical spaces; this study shows that it also maps fully abstract, latent and uninstructed spaces. People played a naturalistic social interaction game while their brains were scanned. Hippocampal brain activity correlated with the fictional characters locations in an abstract social space framed by axes of affiliation and power, despite the participants never being exposed to a perceivable spatial representation. This mapping was present across multiple analyses and two samples, demonstrating that the brain system responsible for spatial mapping maps our social interactions too.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.