The research team, led by Philipp Bauer and João R. Vaz of the Egas Moniz School of Health and Science in Portugal, together with collaborators at the University of Essex, George Mason University and Monash University, recruited 38 healthy older adults aged 60 to 87 and 46 healthy young adults aged 18 to 34. Each participant completed a single laboratory session involving two very different motor tasks: a stationary ankle-force test and twelve minutes of continuous treadmill walking at their preferred speed. The key question was whether the finesse with which the nervous system modulates force at the ankle is statistically linked to the finesse of the stepping pattern itself, and whether that link depends on age.

To measure finesse, the researchers turned to a nonlinear signal analysis tool called sample entropy. Unlike conventional statistics that summarize a signal with an average or a coefficient of variation, sample entropy quantifies the temporal structure of a time series: it estimates the inverse probability that short sequences of data points will repeat themselves across the record. A value approaching zero describes a highly predictable, almost mechanical signal, in which similar patterns recur relentlessly. A high value describes a signal so irregular that no pattern ever repeats. Neither extreme is healthy. The most capable physiological systems occupy a middle ground, balancing enough regularity to remain controlled with enough unpredictability to adapt flexibly to stress. The researchers note that this framework has precedent in cardiology, where reduced heart-rate entropy distinguishes patients with congestive heart failure from people with normal sinus rhythm.

In the force task, participants sat in a reclined isokinetic dynamometer with hip and knee flexed at 60 degrees, their dominant foot strapped to a locked lever arm. After maximal voluntary contractions established each person’s strength ceiling, they performed 30-second isometric pushes at 40 percent of their maximum, tracking a target line on a screen. The steady portion of the torque signal was down-sampled to 25 hertz and analyzed for entropy. In the walking task, heel strikes were detected from vertical ground reaction forces on an instrumented treadmill, and the intervals between consecutive heel strikes of the same foot were compiled into a stride-interval series, also scored for sample entropy. Crucially, the two tasks were performed separately, so any shared signal between them could not be a mechanical artifact of simultaneous measurement.

The results were striking in their asymmetry. Among older adults, torque entropy and gait entropy were positively correlated, with a Kendall’s tau of 0.245 and a Bayes factor of 3.985, indicating moderate evidence for the association. Among young adults, the correlation was essentially absent, at tau of −0.079 with a Bayes factor of 0.257, which itself constitutes moderate evidence for no link. A Bayesian comparison of the two correlations showed that the association was significantly stronger in the older group, with a difference in tau of 0.324 and a Bayes factor of 3.456. In plain terms: in youth, being good at modulating force tells you nothing about the texture of someone’s stride; in older age, the two seem to rise and fall together.

The authors interpret this through the lens of network physiology, a framework that views the organism as a web of interacting subsystems rather than a collection of isolated parts. In this view, healthy function is characterized by flexible, appropriately weighted interactions among components, and by high network dimensionality, meaning that many independent degrees of freedom are available for control. Aging, according to the framework, compresses this network: interactions become stronger and more rigid, and the effective number of dimensions shrinks. One measurable consequence is that motor outputs which once varied independently across individuals begin to co-vary. That is precisely the pattern the study observed. It is also consistent with recent work showing that aging produces a breakdown and stratification in networks of inter-muscular coordination, suggesting muscles that once operated with task-specific flexibility become locked into less adaptable configurations.

The researchers are careful about what the finding does and does not prove. Because torque and gait were measured in separate tasks, the study cannot demonstrate direct, time-resolved coupling between the force-producing and gait-controlling subsystems. Nor does the cross-sectional design establish causation: it remains possible that torque complexity causally shapes gait complexity, or that both are downstream of shared upstream factors such as motor unit discharge properties, neuronal loss, altered synaptic characteristics, or unmeasured variables like physical activity level and general health. The effect size itself is modest, and the credible intervals are wide, extending close to zero. The older cohort, with a mean age of 70.4, was relatively young and healthy, with a mean BMI within recommended ranges, which may mean the more dramatic network changes of advanced aging were not yet present. The achieved statistical power was roughly 74 percent, slightly below the targeted 80, and the treadmill setting may not generalize to overground walking in daily life.

Why does any of this matter? Because falls remain a leading cause of injury, disability and loss of independence in older adults, and gait entropy has emerged as a marker of fall vulnerability. As biological inputs to the locomotor system decline with age, stepping patterns become simplified and rigid, an overly regular system with impaired capacity to respond to the environment, whether that means a uneven curb, a slippery floor or a stumble. The ankle plantar flexors are central to this story: they stabilize the body and generate forward propulsion during stance, and their strength correlates with walking speed and with the ability to recover from a trip. Computational modeling suggests that it is not only the magnitude of ankle force but the temporal characteristics of the ankle response that determine whether a trip becomes a fall.

The most consequential implication is that torque complexity may be a modifiable factor. If the entropy of ankle force output and the entropy of gait genuinely co-vary in aging, then interventions that restore richness to force control, such as resistance training, might in principle ripple outward into more adaptable walking and lower fall risk. Testing that possibility requires longitudinal and interventional designs in which torque control is deliberately manipulated and gait entropy is tracked in parallel. The authors also point toward future studies using simultaneous signal acquisition and time-delay stability analysis to test whether the neuromuscular subsystems of older adults are truly over-synchronized, and toward clinical populations where the loss of network dimensionality is expected to be more advanced. Aging itself cannot be reversed, the researchers conclude, but its neuromuscular manifestations may be modifiable, and restoring complexity to the way the body generates force could be one of the levers worth pulling.

Subject of Research: Age-related covariation of plantar flexor torque complexity and gait complexity in neuromuscular aging

Article Title: Complexity of plantar flexor torque and gait covaries in older adults: a network physiology approach to neuromuscular aging

Article References: Bauer, P., Gomes, J. S., Marques, T., Lourenço, S., Oliveira, J. H., Tallent, J., Cortes, N., & Vaz, J. R. (2026). Complexity of plantar flexor torque and gait covaries in older adults: a network physiology approach to neuromuscular aging. Frontiers in Network Physiology, 6, Article 1947848. https://doi.org/10.3389/fnetp.2026.1947848

Image Credits: AI Generated

DOI: 10.3389/fnetp.2026.1947848

Keywords: neuromuscular aging, sample entropy, gait complexity, plantar flexor torque, network physiology, fall risk, motor control, nonlinear dynamics, resistance training, older adults, physiological complexity, Bayesian statistics