van Beers, R. | bioRxiv Science

bioRxiv2024

Vestibular-derived internal models in active self-motion estimation

Self-motion estimation is thought to depend on sensory information as well as on sensory predictions derived from motor feedback. In driving, the vestibular afference can in principle be predicted based on the steering motor commands if an accurate internal model of the steering dynamics is available. Here, we used a closed-loop steering experiment to examine whether participants can build such an internal model of the steering dynamics. Participants steered a motion platform on which they were seated to align their body with a memorized visual target. We varied the gain between the steering wheel angle and the velocity of the motion platform across trials in three different ways: unpredictable (white noise), moderately predictable (random walk), or highly predictable (constant gain). We examined whether participants took the across-trial predictability of the gain into account to control their steering (internal model hypothesis), or whether they simply integrated the vestibular feedback over time to estimate their travelled distance (path integration hypothesis). Results from a trial series regression analysis show that participants took the gain of the previous trial into account more when it followed a random walk across trials than when it varied unpredictably across trials. Furthermore, on interleaved trials with a large jump in the gain, participants made fast corrective responses, irrespective of gain predictability, suggesting they also rely on vestibular feedback. These findings suggest that the brain can construct an internal model of the steering dynamics to predict the vestibular reafference in driving and self-motion estimation.

neuroscience↗

bioRxiv2020

A neural surveyor in somatosensory cortex

Perhaps the most recognizable sensory map in all of neuroscience is the somatosensory homunculus. Though it seems straightforward, this simple representation belies the complex link between an activation in a somatotopic map and the associated touch location on the body. Any isolated activation is spatially ambiguous without a neural decoder that can read its position within the entire map, but how this is computed by neural networks is unknown. We propose that the somatosensory system implements multilateration, a common computation used by surveying and GPS systems to localize objects. Specifically, to decode touch location on the body, multilateration estimates the relative distance between the afferent input and the boundaries of a body part (e.g., the joints of a limb). We show that a simple feedforward neural network, which captures several fundamental receptive field properties of cortical somatosensory neurons, can implement a Bayes-optimal multilateral computation. Simulations demonstrated that this decoder produced a pattern of localization variability between two boundaries that was unique to multilateration. Finally, we identify this computational signature of multilateration in actual psychophysical experiments, suggesting that it is a candidate computational mechanism underlying tactile localization.

neuroscience↗