Micou, C. | Search bioRxiv

bioRxiv2026

Hippocampal brain-machine interface-based navigation reveals CA1 representations of intended actions

The hippocampus integrates external cues and self-motion to construct cognitive maps. Activating these maps independently of immediate sensory and motor signals can support navigation by predicting future locations. The neural basis of such internally driven activation remains poorly understood. To address this, we had mice use a brain-machine interface (BMI) to directly control navigation from real-time hippocampal activity. In this condition, CA1 responses encoding running movement were not navigationally relevant, and place codes rapidly reconfigured to form new representations that discarded locomotion-related signals. By comparing neural representations across BMI-controlled navigation, locomotion-controlled navigation, and passive playback of predetermined routes, we found evidence of response patterns that were specific to conditions in which neural activity could causally influence an animals travel. This suggests the existence of CA1 responses that are not tied to external stimuli or self-motion, and that are suppressed when animals are merely passive observers.

neuroscience↗

bioRxiv2024

Heavy-tailed statistics of cortical representational drift are advantageous for stabilised downstream readouts

Representational drift of fixed stimuli, learned tasks and familiar environments is observed in many brain areas, leading to reconfiguration of population codes over days to weeks. This raises the question of whether downstream brain regions employ mechanisms to track changes in population activity and thus preserve the fidelity of the information they extract. We show that the statistical properties of drift have a significant impact on such mechanisms. Over an extended period, a net change in population tuning due to drift can arise from an accumulation of small changes distributed across the population, or via abrupt jumps that affect smaller subsets of cells at each time point. We demonstrate that an adaptive readout can exploit the heavy-tailed statistics of abrupt jumps to maintain a more stable readout using a simple inference mechanism. Using experimental data, we investigate the extent to which heavy-tailed drift statistics are observed during representational drift in the posterior parietal cortex and visual cortex. We find that experimentally measured drift does not conform to a Gaussian random walk. Instead, we find sudden jumps in neural tuning that would be advantageous for a downstream observer adapting to changes in representation. These observations motivate future study to determine whether adaptive decoding mechanisms exist in the brain and to determine the physiological mechanisms that shape the statistics of representational drift.

neuroscience↗