A question behind our study of anticipatory postural control

In everyday life, we continuously adjust our posture without even being aware of it in order to maintain balance. In particular, when we can anticipate an upcoming disturbance, such as the floor tilting or the body being pushed, we often begin to prepare before the disturbance actually occurs. This process is known as anticipatory postural control, and it is one of the key functions that supports stable posture and movement. Even before raising an arm, taking a step, or reaching for an object, the body has already begun to make adjustments in preparation for the next action. But how is this posture-adjusting mechanism, which allows us to prepare in advance for future disturbances, actually generated?

Where this study began

This study grew out of our earlier work using a rat model. We had developed an experimental paradigm in which rats stood bipedally on their hindlimbs and were exposed to a disturbance after a warning cue. In that work, we showed that the postural responses of rats in this task could be reproduced by dynamic simulations based on model predictive control. Because the center-of-mass responses observed in rats repeatedly exposed to the same disturbance could be explained within a framework that predicts future states and optimizes control inputs, we began to think that prediction and optimization might provide an effective way to understand anticipatory postural control.

At the same time, the rat experiments had important limitations. In rats, the center-of-mass changes during the anticipatory period were relatively small, and it was not easy to measure muscle activity during upright stance without interfering with posture itself. For that reason, we decided to build a human experimental paradigm in order to examine in more detail how the body moves during the anticipatory period and what roles individual muscles play. We designed a task in which participants stood on a platform that tilted after a warning cue, and we measured both center-of-mass motion and muscle activity during the anticipatory period.

A result that was not what we expected

At the beginning of the study, we had a fairly clear expectation. If the center of mass moved forward during the anticipatory period, then we thought this movement should be generated by activation of the tibialis anterior, which produces a forward torque at the ankle. This seemed like a natural hypothesis, and previous studies that informed our experimental design had also reported tibialis anterior activity.

However, the actual measurements told a different story. Although the center of mass moved forward before the perturbation, there was very little activity in the tibialis anterior. Instead, what increased was the activity of the gastrocnemius. Because the gastrocnemius produces a backward torque at the ankle, its activation during a phase in which the center of mass was moving forward was difficult to explain with a simple mechanical interpretation. We therefore had to examine this result carefully, considering not only the data themselves but also whether experimental conditions or analytical procedures might have influenced the outcome.

What simulation revealed

A key step in interpreting this result was musculoskeletal simulation based on model predictive control. Using the human experiments as a basis, we constructed a musculoskeletal model with ankle and hip joints and ran simulations in a framework that determines control inputs by predicting future body states. The simulations reproduced the same gastrocnemius-dominant activity pattern that we observed experimentally during the anticipatory period. This suggested that the muscle activity pattern was not simply a measurement artifact, but could instead be understood as a consequence of control based on prediction and optimization.

We then reconsidered how the forward movement of the center of mass was actually being generated. What emerged was the following postural strategy. Rather than pushing the center of mass forward using muscle force alone, the body first exploits gravity to generate a natural forward-leaning motion, thereby shifting the center of mass forward. Then, to prevent the center of mass from moving too far forward or the body from becoming unstable, the gastrocnemius becomes active and adjusts that motion. In other words, to achieve both forward center-of-mass displacement and postural stabilization at the same time, the body appears to use gravity while relying on minimal muscle activity.

To test this interpretation further, we changed the weight assigned to control input during the anticipatory period in the cost function of the model predictive controller. As we increased the emphasis on keeping control input small, the experimentally observed gastrocnemius-dominant activity pattern became more likely to appear. This indicates that the postural strategy was not a special-case phenomenon or an accidental outcome, but rather emerged naturally from the requirement to maintain future stability while minimizing control effort. Although the direction of center-of-mass movement and the direction suggested by muscle action appear inconsistent at first glance, the behavior can in fact be understood as a rational strategy in which the body makes effective use of gravity as an external condition.

Looking ahead

This work is not only about understanding human postural control; it is also closely connected to the rat studies we are continuing. In our rat experiments, we are applying the same kind of anticipatory perturbation task while inactivating the cerebellum or motor cortex, in order to investigate the neural mechanisms involved in anticipatory postural control. Ultimately, we want to clarify how the “prediction” and “optimization” represented in our model are implemented in the nervous system itself. By linking a control-theoretic framework with neurophysiological findings, we hope to approach the neural basis of the postural strategy identified here.

We are also applying a similar task to patients with cerebellar disorders in order to examine how anticipatory postural strategies change in disease. Anticipatory postural control is known to be impaired in disorders such as cerebellar disease and Parkinson’s disease, and if these changes can be described from the perspective of control theory, this may contribute to a better understanding of disease mechanisms and to the design of rehabilitation strategies. In addition, understanding how humans stabilize posture efficiently against disturbances may provide useful insights for designing robotic systems with more adaptive, human-like control.

Our study suggests that postural control is not merely a reflexive response, but a predictive and optimized strategy constructed in anticipation of future body states. It also suggests that the body does not generate movement through muscle force alone, but actively takes advantage of gravity as part of the environment. We hope that this perspective will help deepen our understanding of anticipatory postural control and provide clues for thinking about both its neural basis and its broader applications.