Motor control or the art of betting

Abstract

The study by Zaback et al. (2025) in the present issue of The Journal of Physiology involved repeatedly exposing individuals to the same postural threat at the same time as probing vestibular and motor responses using stochastic vestibular stimulation. This made it possible to manipulate the emotional and autonomic state in the context of the same threat. Their results suggest that threat-related changes in vestibular and motor function are closely related to threat-induced emotional and autonomic changes, and are not an invariant response to specific contextual features of the threat. This study, which is very interesting in itself, raises some fascinating questions about sensorimotor transformations.

One of these concerns the status of sensory information in the CNS. Secondary vestibular neurons in the brainstem take their name from the fact that they receive primary vestibular afferents. However, as we later realized, these neurons also receive proprioceptive inputs, mainly from the neuromuscular spindles of the 48 neck muscles, and visual inputs via the accessory optic system. This explains why, when you are sitting on a train and the train next to you leaves, you feel your own train moving. As a result, second-order vestibular neurons actually develop an internal multimodal representation of head and body movement in space. This representation is used to stabilize gaze and posture, to generate an awareness of movement and to allow separation between one's own movement and externally generated movement. In this context, can we still talk about vestibular neurons? Should we ask ourselves whether the central nervous system preserves ‘vestibular, visual and proprioceptive’ information or whether it mixes these inputs in the service of a multimodal representation of the world for frequency tuned channel? In the light of electrophysiological work on animal models, several types of multimodal representation of our own movement probably coexist in the vestibular nucleus, with variable weighting for visual and proprioceptive vestibular afferents.

And that's not all. The so-called ‘vestibular’ neurons also receive cerebellar and cortical inputs, and their activity is modulated by several neuromodulators involved in vigilance, emotional and autonomic changes. This partly explains the results obtained by Zaback et al. (2025) and Cullen (2023), which show that we are a long way from a 19th century conception of postural control as the result of the sum of elementary reflexes. We are confronted with ultra-rapid representations elaborated by the CNS on the basis of a cocktail of sensory inputs, the actual level of vigilance, and the affective and probably social context. In around 200 ms, these are used to develop coherent representations of a fluctuating world, which are necessary to stabilize our gaze, control our posture, distinguish between someone pushing us and a movement we have initiated; in short, to avoid motor disasters.

In such critical situations, we might expect everyone's control of their gaze and posture to be relatively uniform in identical circumstances. This is not the case, as Zaback et al. (2025) and others have shown. All neurophysiologists who work on sensory-motor transformations, whether in animal models or in humans, learn that the more ecological the context in which they work and the more precise their measurements of behaviour, the greater the differences between individuals. In other words, the question is not just to understand how, on the basis of multimodal representations influenced by context, we reconstruct a reality that enables us to act. It is also about understanding how different people reach different conclusions when it comes to regulating their motor control. In other words, each of us has our own perceptual-motor style (for a review, see Vidal et al., 2021) and even more so in pathological states. The consequence is that the average of behaviours in a cohort can be misleading. Each person, when the control of gaze and posture is challenged, takes a different view of the risks to which she/he is exposed and therefore adopts different solutions to resolve the problem. This is why individual longitudinal monitoring is becoming a key element in the study of motor control.

Also, being a permanent biped is a risky game. Uncertain prediction and/or the adoption of poor muscular synergies lead to a fall to the ground 500 ms later, around 700 ms after the start of a postural perturbation. Furthermore, falls are the second leading cause of death by unintentional trauma in the world, not to mention the various injuries they cause and the loss of independence of the elderly. The study by Zaback et al. (2025) and previous others and previous studies show that control of gaze and posture, far from being a series of reflexes, is a question of ultra-rapid bets, not predictions, to assess the risks, and the optimal motor programs aiming to avoid them. These strategies have communalities between individuals (we rarely bet at random). They also define different styles for different individuals and unreasonable betting leads to neuroses (acrophobia, fear of falling). These considerations are probably even truer when pathological processes and their rehabilitation are involved. Again, individual longitudinal monitoring should become the rule.

Two final comments to conclude this editorial. First, to meet the challenge of gambling, various studies of the vestibular system have highlighted two constituent elements that could be instrumental at the neuronal level: vestibular habituation and vestibular adaptation. Second, the ubiquitous artificial intelligence does not seem to be able to answer the questions raised here: if there's one thing large language models (LLMs) do not do, it's bet and be anxious about it. Perhaps these two typically human behaviours explain why gaze and postural control became so effective in the permanent bipeds we have become, and why they ultimately failed with age.