The study was conducted by Lucy Hancock, Kaneezah Begum and Ori Ossmy of Birkbeck, University of London, who recruited 100 young adults aged 20 to 39 and 100 older adults aged 60 to 88. Rather than dividing participants into predetermined categories such as “active” versus “sedentary,” the researchers administered a novel version of the Recent Physical Activity Questionnaire, a validated self-report instrument that captures the types, frequency and intensity of activities people perform in daily life. The questionnaire data were then fed into a k-means clustering algorithm, a form of unsupervised machine learning that identifies natural groupings within multivariate data without any prior labeling. This data-driven strategy, the authors argue, sidesteps a long-standing limitation of the field: the tendency to impose predefined activity classifications that may obscure meaningful individual differences in how people actually move.

Once the algorithm had partitioned the sample into distinct activity profiles, the researchers compared the cognitive performance of each group on two carefully chosen tasks. The first, the Towers of Hanoi, is a classic measure of planning and problem solving in which participants must move a stack of disks between pegs while obeying strict rules, requiring them to think several moves ahead. The second, Virtual Tools, is a modern computer-based task of physical reasoning in which participants must select and release virtual objects to achieve a goal, engaging their intuitive understanding of physics—gravity, momentum and collision. The Virtual Tools paradigm draws on recent computational work showing that humans simulate physical outcomes through rapid, trial-and-error mental modeling, making it a sensitive probe of how well the brain predicts the behavior of objects in the world.

The results were striking. Participants whose activity profiles were characterized by greater variety—engaging in many different types of physical activity—and by mindful, body-aware practices such as yoga or similar mind–body disciplines outperformed other groups on both planning and physical reasoning measures. This advantage held even though the sheer volume or intensity of activity did not uniformly predict better cognition. In other words, it was not the people who exercised hardest who reasoned best, but the people whose movement repertoire was most diverse and most attentive. The finding aligns with a growing body of evidence that cognitively enriched physical activity—movement that also demands coordination, learning and attention—produces stronger cognitive benefits than repetitive exercise alone.

Age, unsurprisingly, mattered. Older adults performed worse than younger adults overall on the cognitive tasks, consistent with decades of research documenting declines in executive function and fluid reasoning with typical aging. But a crucial nuance emerged in the statistical analysis: physical activity profile did not interact significantly with age group for any outcome. The relationship between varied, mindful movement and cognitive performance was statistically indistinguishable in the young and older samples. This absence of an interaction suggests that whatever benefit diverse and mindful activity confers on planning and reasoning, it appears to operate across the adult lifespan rather than being a special advantage reserved for one age group. For researchers of aging, that is an encouraging signal—it hints that the activity profiles associated with sharper cognition remain stable targets well into the eighties.

The study’s methodological approach deserves particular attention. Traditional studies of exercise and cognition typically categorize participants as meeting or not meeting physical activity guidelines, or compare specific interventions such as aerobic training against resistance training. These categorical approaches discard much of the richness of real-world behavior. By contrast, k-means clustering lets the data speak: the algorithm minimizes within-cluster variance and maximizes between-cluster separation across the full multidimensional space of self-reported activity, revealing profiles that no a priori taxonomy would have produced. The researchers also used bootstrapping—a resampling technique that repeatedly re-estimates cluster assignments to assess stability—following established statistical practice for validating k-means solutions. The choice of sample size was informed by power-analysis conventions recommending larger samples for reliable effects, and the two cognitive tasks were selected for their established reliability and validity as measures of executive planning and physical problem solving.

Why might varied and mindful movement be linked to planning and physical reasoning? The authors situate their findings within several converging theoretical frameworks. One possibility involves cognitive enrichment: activities that combine physical execution with strategic demands—such as dance, martial arts, climbing or racquet sports—simultaneously tax motor control, working memory and predictive reasoning, potentially strengthening shared neural circuitry. A second line of reasoning concerns mind–body practices specifically. Meta-analyses of meditation, yoga and tai chi have reported benefits for executive function in older adults, with proposed mechanisms ranging from stress attenuation and reduced cortisol to improved attentional control. Classic work has even shown improved performance on the Tower of London planning test following yoga practice, an intriguing precedent for the present findings. A third framework invokes the general physiology of exercise—increased brain-derived neurotrophic factor, enhanced vascular function and reduced inflammation—but the study’s data suggest these generic mechanisms alone cannot explain why variety and mindfulness, rather than volume, tracked cognitive performance.

The findings also speak to a persistent puzzle in the aging literature: the frequent failure of straightforward physical activity interventions to produce robust cognitive gains. Systematic reviews and meta-analyses of exercise trials in adults over 50 have found modest and heterogeneous effects, and some longitudinal studies have raised the possibility of reverse causation—that cognitively healthier people are simply more likely to stay active. The present study does not resolve the question of causality; its cross-sectional design cannot determine whether varied, mindful movement builds sharper reasoning or whether people with better reasoning gravitate toward richer activity repertoires. The authors are explicit on this point, calling for longitudinal and intervention studies to establish whether these activity profiles causally support cognitive function across adulthood.

Even so, the implications are tantalizing. If the associations hold up under experimental scrutiny, public health guidance might need to emphasize not just how much people move, but how they move. Encouraging older adults to diversify their activity—adding balance-based, skill-based and mindful practices to routine walking or gardening—could be a low-cost strategy for supporting the planning abilities that underpin everyday independence, from managing medications to navigating unfamiliar routes. The finding that the relevant profiles look similar in young and older adults also suggests that cultivating varied movement habits early in life may pay cognitive dividends decades later, framing physical diversity as a form of cognitive investment rather than merely a cardiovascular one.

The research also contributes to a younger scientific field: physical cognition, the study of how humans reason about objects, forces and tool use. Prior work from the same laboratory has examined how action concepts shape physical reasoning in late childhood and how physical reasoning declines with typical aging under both familiar and unfamiliar physics. The new results extend this program into the domain of lifestyle, proposing that the embodied experience of moving one’s body in diverse, deliberate ways may feed the same intuitive physics engine that the Virtual Tools task measures. On this view, the body is not just a vehicle the brain pilots; it is a training ground where the brain continuously learns the statistics of the physical world.

Limitations remain. Physical activity was self-reported, and questionnaires are known to be imperfect measures of energy expenditure, particularly in older populations where recall and interpretation of intensity categories can drift. Self-report also cannot capture the quality or cognitive load of an activity—two people may both report “yoga” while practicing at very different levels of attentional demands. The clustering approach, while flexible, yields group-level profiles that mask individual variability, and the sample, though large by laboratory standards, was recruited online and may not represent the full diversity of the adult population. The authors note that the underlying data will be made publicly available, enabling other researchers to test alternative models and replicating the profiles in independent cohorts.

For now, the study offers a fresh and data-driven lens on an old question. Instead of asking whether exercise is good for the brain, Hancock, Begum and Ossmy ask what kind of exerciser reasons best—and the answer, at least descriptively, is the one who moves in many ways and moves with the mind engaged. Whether prescribing variety and mindfulness can actually sharpen planning in a randomized trial is the obvious next experiment, and one that could reshape how clinicians, trainers and policymakers think about movement across the lifespan. In a field long dominated by step counts and heart-rate zones, the message that the brain may care more about the richness of movement than its raw quantity is a provocative and quietly viral idea—one that turns the daily workout from a metabolic chore into an opportunity for embodied learning.

The research team analyzed a vast cohort of 16,972 UK Biobank participants, capitalizing on long-term accelerometer data that objectively quantified participants’ physical activity over a full week. This method overcame limitations of self-reported exercise data commonly used in prior studies, which often suffer from recall biases and inaccuracies. Harnessing state-of-the-art machine learning algorithms, specifically the powerful Light Gradient Boosting Machine (LightGBM), researchers deciphered the complex nonlinear relationship between physical activity intensity and brain health, using over 1,400 image-derived phenotypes (IDPs) extracted from T1-weighted magnetic resonance imaging (MRI) scans.

Central to this research was the construction of a brain age prediction model that estimates “brain age” from neuroimaging data, adjusting for the confounding effect of chronological age. This brain age gap (BAG)—the difference between predicted brain age and actual age—serves as a quantifiable biomarker for neural aging and neurodegeneration. Leveraging LightGBM algorithms trained on hierarchical clusters of features selected through tree-based feature importance and SHAP (SHapley Additive exPlanations) values, the team achieved a refined model architecture minimizing redundancy while maximizing interpretability.

Their analyses uncovered a striking U-shaped association between physical activity intensity and BAG: both low and high intensities corresponded with accelerated brain aging, while moderate physical activity was neuroprotective. More specifically, moderate and vigorous physical activity (MPA and VPA respectively) correlated with meaningful reductions in BAG—VPA, for example, demonstrated a beta coefficient (β) of –0.27, indicating an inverse relationship with accelerated brain aging. This nuanced finding challenges the simplistic “more is better” dogma surrounding exercise, emphasizing balanced activity levels as pivotal for maintaining brain health.

To quantify physical activity, participants wore wrist accelerometers over seven consecutive days, allowing researchers to objectively categorize activity levels into light (LPA), moderate (MPA), vigorous (VPA), and moderate-to-vigorous physical activity (MVPA). Such precise measurement techniques ensured high-fidelity data acquisition, circumventing biases of subjective reporting and enhancing the clinical relevance of findings to public health guidelines.

An essential facet of this investigation was mediation analysis, which revealed that BAG partially mediates the beneficial effects of physical activity on cognitive functions, such as reaction time, and on brain-related disorders including dementia and depression. This mechanistic insight bridges behavioral activity with neurological outcomes, illustrating how modifiable lifestyle factors translate to tangible brain health metrics.

Neuroanatomical correlates uncovered by this study further elucidate the biological substrates underpinning these associations. Participants engaged in moderate physical activity exhibited reduced white matter hyperintensities—lesions often related to aging and cerebrovascular disease—alongside preserved gray matter volume in key subcortical regions including the cingulate cortex, caudate nuclei, and putamen. These brain areas are pivotal for cerebrovascular integrity and cortico-striatal circuitry, which underlie executive function and motor control, suggesting that exercise confers protection across multiple critical neural systems.

Associate Professor Xu emphasized the implications of these results: “Our study does not merely affirm a nonlinear relationship between objectively measured physical activity and brain aging at the population scale but also delivers a crucial public health message—more exercise isn’t always better. Finding the right intensity balance might be the key to preserving brain integrity and delaying neurodegeneration.”

Given the scale and rigor of this analysis, combining multimodal imaging with accelerometer-based activity quantification and advanced machine learning, this study sets a new benchmark in neuroscientific epidemiology. Moreover, it offers a template for integrating large-scale data analytics with longitudinal patient monitoring, an approach increasingly vital for precision medicine.

The research team’s future trajectory promises even deeper insights into aging biology. Plans are underway to develop a multi-scale aging framework integrating data on sleep patterns, sedentary behavior, diverse neuroimaging modalities, and omics (genomics, proteomics) profiles. This comprehensive approach aims to unravel complex interactions between lifestyle, brain structure and function, and molecular pathways that regulate aging.

Longitudinal intervention studies are also on the horizon to test how targeted behavioral changes might modulate brain aging trajectories. By combining genome-wide analyses with proteomic signatures, researchers hope to identify biomarkers predictive of individual responses to physical activity interventions, potentially guiding personalized exercise prescriptions to mitigate age-related cognitive decline and neurodegenerative diseases.

This pioneering study not only advances our understanding of the neurobiological benefits of moderate exercise but also challenges public perception around physical activity, urging a paradigm shift from indiscriminate intensity escalation to balanced, sustainable regimens that optimize brain health across the lifespan.

Subject of Research: The relationship between objectively measured physical activity and neuroimaging-based brain age estimation in a large population cohort.

Article Title: Accelerometer-Measured Physical Activity and Neuroimaging-Driven Brain Age

News Publication Date: 2-May-2025

Web References: DOI: 10.34133/hds.0257

Image Credits: Chen Han., et al, School of Public Health, Hangzhou Normal University

Keywords: Physical exercise, brain aging, accelerometer, neuroimaging, machine learning, LightGBM, brain age gap, cognitive function, white matter hyperintensities, cortico-striatal circuitry