PLACE AND VIEW RESPONSES IN THE PRIMATE HIPPOCAMPUS DURING VIRTUAL NAVIGATION

The role of the rodent hippocampus in spatial navigation has been one of the hallmark discoveries of modern neuroscience. Recently, a similar role has been established for the hippocampus of primates, where in addition to spatially tuned neurons, neurons tuned to parts of an animal's visual input have also been found.
Here we investigate the joint tuning to place and view of hippocampal neurons in NHPs engaged in virtual navigation. We found significant place tuning in 9.2% and significant view tuning in 9.5% of the cells, as well as significant head direction tuning in 10.7% of the cells. In addition, 65.3% of the cells with either place, view, or head direction tuning were selective to more than one variable. Most of these mixed selective cells (90.6%) were tuned to specific combinations of place and view features; we refer to these as conjunctive cells.
We created a simplified version of the task and trained RNN models to encode the spatial location and gaze positions at each timestep. Interestingly, we found that models trained to encode an arena with occlusions had lower proportions of conjunctive hidden units compared to models trained to encode an open arena, even though they exhibited more dependence between place and view coding in their outputs.
This suggests that the degree of conjunctive coding is affected by the type of environment being encoded, and hints at the potentially complex ways in which the presence of conjunctive cells is linked to the underlying network dynamics.