Advertisement

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Compensation Related Utilization of Neural Circuits Hypothesis (CRUNCH) of Electrocortical Data during Walking on Uneven Terrain

  • Erika M. Pliner ,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing – original draft, Writing – review & editing

    erika.pliner@utah.ed

    ☯ Co-first authors.

    Affiliations Department of Mechanical Engineering, University of Utah, Salt Lake City, Utah, United States of America, Rocky Mountain Center for Occupational and Environmental Health, Salt Lake City, Utah, United States of America, Department of Biomedical Engineering, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Chang Liu ,

    Roles Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

    ☯ Co-first authors.

    Affiliations Department of Biomedical Engineering, University of Florida, Gainesville, Florida, United States of America, Department of Kinesiology and Nutrition, University of Illinois at Chicago, Chicago, Illinois, United States of America

    ⨯
  • Jacob S. Salminen,

    Roles Data curation, Formal analysis, Investigation, Project administration, Validation, Writing – review & editing

    Affiliation Department of Biomedical Engineering, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Patricia A. Reuter-Lorenz,

    Roles Conceptualization, Writing – review & editing

    Affiliation Department of Psychology, University of Michigan, Ann Arbor, Michigan, United States of America

    ⨯
  • Ryland D. Swearinger,

    Roles Data curation, Investigation, Project administration, Writing – review & editing

    Affiliation Brain Rehabilitation Research Center, Malcom Randall VA Medical Center, Gainesville, Florida, United States of America

    ⨯
  • Jungyun Hwang,

    Roles Data curation, Investigation, Project administration, Writing – review & editing

    Affiliation Department of Neurology, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Arkaprava Roy,

    Roles Formal analysis, Methodology, Writing – review & editing

    Affiliation Department of Biostatistics, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Natalie Richer,

    Roles Data curation, Investigation, Methodology, Project administration, Writing – review & editing

    Affiliation Department of Kinesiology and Applied Health, University of Winnipeg, Winnipeg, Manitoba, Canada

    ⨯
  • David J. Clark,

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Writing – review & editing

    Affiliations Brain Rehabilitation Research Center, Malcom Randall VA Medical Center, Gainesville, Florida, United States of America, Department of Neurology, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Todd M. Manini,

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Writing – review & editing

    Affiliation Department of Health Outcomes and Biomedical Informatics, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Chris J. Hass,

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Writing – review & editing

    Affiliations Department of Applied Physiology and Kinesiology, University of Florida, Gainesville, Florida, United States of America, Norman Fixel Institute for Neurological Diseases, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Yenisel Cruz-Almeida,

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Writing – review & editing

    Affiliations Pain Research and Intervention Center of Excellence, University of Florida, Gainesville, Florida, United States of America, Department of Community Dentistry and Behavioral Science, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Rachael D. Seidler,

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing – review & editing

    Affiliations Department of Applied Physiology and Kinesiology, University of Florida, Gainesville, Florida, United States of America, McKnight Brain Institute, University of Florida, Gainesville, Florida, United States of America

    ⨯
  • Daniel P. Ferris

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing – review & editing

    Affiliations Department of Biomedical Engineering, University of Florida, Gainesville, Florida, United States of America, McKnight Brain Institute, University of Florida, Gainesville, Florida, United States of America

    ⨯

Abstract

Mobility limitations are common in older age, but the association of age-related changes in the neural control of walking are not well understood. We investigated whether the Compensation Related Utilization of Neural Circuits Hypothesis (CRUNCH) – a model that describes age differences in brain activity with increasing cognitive demand – applies to walking under varying levels of task difficulty. We assessed the electrocortical data of younger and older adults walking on a treadmill with varied levels of uneven terrain (i.e., varying difficulty levels). We observed CRUNCH effects in the posterior parietal regions, which are thought to be associated with planned movements, spatial reasoning, and attention. We found older adults displayed greater alpha desynchronization than younger adults at low terrain unevenness, and younger adults displayed greater range in utilized spectral power in the alpha band across levels of terrain unevenness than older adults. These results demonstrate both over-recruitment and ceiling effects, respectively, of neural resources in older adults compared to younger adults, as predicted by the CRUNCH model. These findings provide novel insights into age differences in the neural control of mobility and may lead to new interventions for maintaining function in older age.

Citation: Pliner EM, Liu C, Salminen JS, Reuter-Lorenz PA, Swearinger RD, Hwang J, et al. (2026) Compensation Related Utilization of Neural Circuits Hypothesis (CRUNCH) of Electrocortical Data during Walking on Uneven Terrain. PLOS Aging Health 1(1): e0000017. https://doi.org/10.1371/journal.page.0000017

Editor: Joaquin U. Gonzales, Texas Tech University, UNITED STATES OF AMERICA

Received: January 12, 2026; Accepted: May 18, 2026; Published: July 21, 2026

Copyright: © 2026 Pliner et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data that support the findings of this study are publicly available on OpenNeuro with the identifers: ds004625, ds006095, and ds007713######.

Funding: This study was supported by the National Institute of Health (U01AG061389) for all authors. National Institute of Health grants F32AG072808 and T32AG062728 supported author EMP. American Heart Association Fellowship (23POST1011634, doi.org/10.58275/AHA.23POST1011634.pc.gr.161292) partially supported author CL. DPF was also supported by National Institutes of Health (R01NS104772). A portion of this work was performed in the McKnight Brain Institute, which is supported by National Science Foundation Cooperative Agreement No. DMR- 1644779 and the State of Florida, and in part by an NIH award, S10 OD021726, for High End Instrumentation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

1. Introduction

Our ability to move within the environment (e.g., home, neighborhood, global community) is integral to our quality of life [1,2]. Unfortunately, the incidence of mobility limitation rises with increasing age [2,3]. Approximately 23 million older adults (aged 65+) in the United States have a mobility limitation, accounting for 40% of the older adult population [4]. Mobility limitations are associated with adverse individual and community outcomes, including increased fall risk, hospitalization, depression, dependence and mortality [3,5]. Hospital admissions associated with reduced mobility are costly to the individual and stress the healthcare system [6,7]. Thus, there is a societal and economical need to understand the mechanisms of mobility loss with increased age.

Physical health is a primary risk factor of mobility limitation [2,8,9]. Specifically, decreases in muscular strength and power are robust predictors of mobility [10–12]. This has led to physical performance metrics (e.g., gait speed, time-up-and-go) as the standard assessment of mobility [2]. Yet, physical performance metrics do not explain all mobility declines. Mobility decline is also linked to sensory loss [2] and control deficits [13,14], which are not well captured in physical performance metrics. A robust capture of mobility decline would be characterized by an integration of sensory input, computational processing, and motor execution, a combination of processes that can all be assessed from brain activity. Thus, understanding the neural correlates of mobility decline has the potential to transform mobility healthcare.

Brain activity with increasing levels of task difficulty may follow a similar utilization framework across mobility and cognitive task contexts. In the cognitive domain, the modulation of brain activity in response to varying levels of task demand has been associated with individual and age-differences in performance, as well as benefits from cognitive training [15,16]. These findings have been understood in terms of the Compensation Related Utilization of Neural Circuits Hypothesis (CRUNCH). The CRUNCH framework has been observed in various cognitive tasks with varying demands (perceptual, motoric, mnemonic, verbal and spatial) [17–19]. CRUNCH characterizes the over-recruitment and ceiling effects of brain activity (measured with functional Magnetic Resonance Imaging; fMRI) across multiple levels of task difficulty in older adults relative to younger adults. Specifically, older adults recruit more brain resources for simple tasks to achieve similar performance outcomes of younger adults (compensation) [20]. As the task increases in difficulty, older adults reach a ceiling in brain activity at a lower level of task demand than younger adults, which is related to reductions in performance. These findings suggest a restricted dynamic range of brain responsivity and an association limitation in neural resources for brain activity of older adults (Fig 1a.).

thumbnail
Download:
Fig 1. Graphical representation of the Compensation Related Utilization Neural Circuits Hypothesis (CRUNCH) framework (a) for younger (teal) and older (purple) adults.

The best modality measure to assess brain activity for fMRI and fNIRS data is a Blood-Oxygen-Level Dependent (BOLD) signal (b). The best modality measure to assess brain activity for electrocortical data in the theta and alpha/beta frequency bands is spectral power (c). These plots showcase the CRUNCH recruitment of cognitive resources by imaging modality. Regardless of imaging modality, CRUNCH reflects recruitment of cognitive resources to be higher in older adults at low difficulty levels and to reach a plateau sooner than younger adults.

https://doi.org/10.1371/journal.page.0000017.g001

CRUNCH has largely been investigated with cognitive tasks, mostly due to a focus on magnetic resonance technology. Specifically, fMRI studies have excellent spatial resolution for capturing brain activity, but mobility during fMRI studies is not possible in the bore of the magnet [17,19]. Recent innovations in brain imaging technology enable the assessment of brain activity in mobile paradigms. Functional near-infrared spectroscopy (fNIRS) and electroencephalogram (EEG) permit the mobile recording of metabolic and electrocortical data, respectively. fNIRS is commonly used to assess hemodynamic responses in the prefrontal cortex [21]. Mobile EEG has excellent temporal resolution and good spatial resolution when using a high number of electrode recording sites (i.e., high-density recording) [22]. Advanced mobile EEG processes capture the signature of cortical regions that are contributing to the strongest electrocortical signal (i.e., far field potentials of neural activity) [23–27]. While each of these brain imaging modalities has inherent limitations, parallel investigation across these modalities can help to overcome imaging modality restrictions and comprehensively investigate whether CRUNCH applies to mobility tasks.

This study is part of a larger project investigating whether and how CRUNCH applies to the neural control of mobility in older adults [28]. The objective of the present study was to determine whether CRUNCH describes electrocortical data collected via mobile EEG during an uneven terrain walking task of varying difficulty with younger and older adults. CRUNCH has been observed from the Blood-Oxygen-Level Dependent (BOLD) signal of fMRI [15,17,29–31], where greater changes in the BOLD signal are reflective of increases in cognitive resources (Fig 1b). We might expect CRUNCH patterns to manifest differently in electrocortical signals, which can be decomposed into multiple signals of various oscillation frequencies. The amplitude of each signal characterizes the power of that frequency in the composite signal. Electrocortical data is characterized by spectral power changes. The complexity of electrocortical data interpretation stems from both increases and decreases in spectral power to suggest increases in brain activity. Previous literature on mobile brain imaging supports the interpretation of EEG by frequency band and cortical region [22]. That is, increases and decreases in the spectral power of electrocortical data are frequency and cortical area dependent. In line with mobile EEG literature, this study will investigate CRUNCH in the theta (4–8 Hz), alpha (8–13 Hz), and beta (13–30 Hz) frequency bands of EEG across the cortex.

We developed the hypotheses for this study based on previous fMRI literature pertaining to CRUNCH and EEG literature while walking. Over-recruitment of cognitive resources has been demonstrated in the prefrontal cortex for many cognitive tasks [16,18,32]. Executive function controls like attentional processing of other frontal regions, such as, the anterior cingulate [31,33] can provide further compensation to task demands [18]. Results from electrocortical recordings during challenged walking (i.e., stepping along a balance beam, varied treadmill speed) is associated with increases in theta power in the anterior cingulate and prefrontal cortical areas compared to treadmill walking [27,34]. Further, reestablishing stable foot contact after a loss of balance from a balance beam step elicits decreases in alpha power in the anterior cingulate, prefrontal, and parietal areas [27]. During gait adaptation to a split-belt treadmill paradigm, widespread beta power decreases were observed across the cortex, including in the anterior cingulate [34]. Given these findings, we hypothesized that older adults would show CRUNCH-like patterns of activity in the prefrontal and anterior cingulate brain regions. That is, we predicted that 1) older adults would exhibit greater recruitment of frontal and cingulate cortices at low levels of terrain unevenness in comparison to younger adults (over-recruitment); and 2) older adults would demonstrate reduced activation at higher levels of terrain unevenness than younger adults (i.e., a ceiling is reached at a lower level of task difficulty for older adults than younger adults). We will assess greater brain activity by increases in theta power and decreases in alpha and beta power, reflecting CRUNCH through inverted and standard parabolic curves (Fig 1c).

2. Results

2.1. Participant characteristics

Analyses comprised 31 younger adults and 71 older adults (Table 1). In the overground walking clinical assessments, older adults walked significantly slower than younger adults during the 400-meter test and across bouts of the uneven terrain conditions. This resulted in older adults being prescribed a slower treadmill speed (set to 75% of the participant’s average walking speed across over-ground walks of each uneven terrain condition). Older adults also scored lower in the Short Physical Performance Battery test compared to younger adults. Cognition measured via the Montreal Cognitive Assessment was not significantly different between younger and older adults.

thumbnail
Download:
Table 1. Participant demographics. Average (standard deviation) age, treadmill speed, Short Physical Performance Battery score, 400 m walk time and speed, and Montreal Cognitive Assessment score for younger and older adults.

https://doi.org/10.1371/journal.page.0000017.t001

2.2. Walking performance

The coefficient of variation for step time and medial-lateral sacral excursion increased with increasing terrain unevenness in both younger and older adults (Fig 2). When characterizing step time by best-fit linear lines on the coefficient of variation across terrain height variability, older adults displayed greater initial variation in step time and faster increases in variation (steeper slope) than younger adults (χ23,194 = 4.9; p = 0.026). For medial-lateral sacral excursion younger adults displayed greater initial variation than older adults, but older adults still displayed faster increases in variation with increasing terrain height variability (χ23,194 = 14.3; p < 0.001). The best-fit linear lines were also found on the coefficient of variation across terrain height variability for other gait parameters (i.e., gait cycle, stance phase, swing phase, single support time, double support time, anterior-posterior sacral excursion, superior-inferior sacral excursion; Supplementary Material 1; S.Table 1 in S1 File).

thumbnail
Download:
Fig 2. Coefficient of variation in step time (left column) and medial-lateral (ML) sacral excursion across terrain unevenness.

The average coefficient of variation of step time (left column) and ML sacral excursion (right column) for each younger adult (left side data per condition, darker shades) and older adult (right side data per condition, lighter shades) with the corresponding distribution displayed for each terrain condition (top row of plots). The best-fit linear lines on average coefficient of variation per younger adult (teal lines) and older adult (purple lines) is plotted across terrain height variability for step time and ML sacral excursion (middle row of plots). The coefficient of variation for younger and older adults is represented by a bold line and showcased without participant variability and after rescaling (last row of plots). The coefficient of variation of step time and ML sacral excursion increases at a greater rate with terrain unevenness in older adults than younger adults. Asterisks denote lines that are significantly different from one another.

https://doi.org/10.1371/journal.page.0000017.g002

2.3. EEG source analysis

Nine brain source clusters met the criteria for data analysis (Fig 3; Table 2). The centroid location of these brain clusters comprised the left sensorimotor, right sensorimotor, left posterior parietal, right posterior parietal, left pre-supplementary motor, right premotor, precuneus, mid-cingulate, and temporal regions. Two brain source clusters (lingual, right temporal regions) were excluded for not retaining clusters from more than half of all participants, resulting in less than 16 brain sources among younger adults and less than 31 brain sources among older adults. Contrary to our hypothesis, brain source clusters were not identified in the prefrontal and anterior cingulate cortices. Given the higher order processing in the posterior parietal cortices (planned movements, spatial reasoning, and attention) [35], the linked relationship between the prefrontal cortex and posterior parietal cortex [36], the known association of proprioceptive function with our uneven terrain walking paradigm [37], and the importance of electrocortical processing in the posterior parietal cortices during challenged walking [27,34,38], we adapted our analysis to assess CRUNCH in these cortical areas.

thumbnail
Download:
Table 2. Brain source location by cluster. Brain sources plotted in the Montreal Neurological Institute (MNI) template [39]. MNI coordinates and anatomical atlas labels are included as denoted by the cluster centroid. Younger and older adult brain sources are differentiated by color shade.

https://doi.org/10.1371/journal.page.0000017.t002

thumbnail
Download:
Fig 3. Individual (top row) and centroid (bottom) brain source location for each identified cluster in the sagittal (Left), coronal (middle) and axial (right) planes.

The brain sources are plotted in the Montreal Neurological Institute template and are differentiated by color. These plots provided transparency and context on the location and clustering of brain sources in this study.

https://doi.org/10.1371/journal.page.0000017.g003

2.4. EEG CRUNCH analysis

Spectral power of electrocortical data in the left posterior parietal cortex displayed a general increase in the theta band and general decrease in the alpha and beta bands with increasing terrain unevenness in both younger and older adults (Fig 4). When characterizing spectral power changes by the individual through best-fit parabolic curves of average spectral power data across terrain height variability, some participants displayed an inverted parabolic curve in the theta frequency band. However, this curve characterization did not reflect the average for younger and older adult participants. Parabolic curves characterizing the average spectral power in the theta band for the younger and older adults were not significantly different from one another (χ23,379 = 4.5; p = 0.476). In the alpha band, the average spectral power across increasing terrain height variability displayed a standard decreasing parabolic curve for younger and older adults. Younger adults displayed a curve with a greater range in average spectral power across terrain height variability than older adults. Older adults displayed greater average spectral power at lower terrain height variability and less range in spectral power than younger adults (χ23,399 = 72.3; p < 0.001). Similarly, in the beta band, the average spectral power across terrain height variability displayed a standard decreasing parabolic curve for younger and older adults. While the curve for younger adults showcases an approach to a more depressed vertex and greater range in average spectral power across terrain height variability than older adults, these curves were not significantly different from one another (χ23,405 = 8.5; p = 0.132).

thumbnail
Download:
Fig 4. Spectral power of electrocortical data in the left posterior parietal cortex.

The average spectral power per frequency band (left column: theta; middle column: alpha; right column: beta) for each younger (left side data per condition, darker shades) and older (right side data per condition, lighter shades) adult with the corresponding distribution is displayed for rest and each terrain condition (top row of plots). The best-fit quadradic curve of average spectral power per younger (teal lines) and older (purple lines) adult is plotted across terrain height variability per frequency band (middle row of plots). The average spectral power for younger and older adults is represented by a bold line and showcased without participant variability and after rescaling (last row of plots). The curves between younger and older adults for spectral power in the alpha band is statistically different. Specifically, older adults display a reduced range in utilized spectral power and greater desynchronization in alpha spectral power at low terrain unevenness than younger adults. Asterisk denotes curves that are significantly different from one another.

https://doi.org/10.1371/journal.page.0000017.g004

Spectral power of electrocortical data in the right posterior parietal cortex displayed a general increase in the theta band and general decrease in the alpha and beta bands with increasing terrain difficulty in both younger and older adults (Fig 5). When characterizing spectral power changes by the individual through best-fit parabolic curves of average spectral power data across terrain height variability, older adult participants displayed an inverted parabolic curve in the theta frequency band whereas younger adults displayed and standard increasing parabolic curve. This resulted in statistically different parabolic curves of average spectral power in the theta band for the younger and older adults (χ23,322 = 17.8; p = 0.003). In the alpha band, the average spectral power across terrain height variability displayed a standard decreasing parabolic curve for younger and older adults. Younger adults displayed a curve with a greater range in average spectral power across terrain height variability than older adults; while older adults displayed greater average spectral power at lower terrain height variability and less range in spectral power than younger adults than older adults (χ23,327 = 15.8; p = 0.007). In the beta band, the average spectral power across terrain height variability displayed a standard decreasing parabolic curve for younger and older adults. However, these curves were not significantly different from one another (χ23,326 = 7.2; p = 0.207).

thumbnail
Download:
Fig 5. Spectral power of electrocortical data in the right posterior parietal cortex.

The average spectral power per frequency band (left column: theta; middle column: alpha; right column: beta) for each younger (left side data per condition, darker shades) and older (right side data per condition, lighter shades) adult with the corresponding distribution is displayed for rest and each terrain condition (top row of plots). The best-fit quadradic curve of average spectral power per younger (teal lines) and older (purple lines) adult is plotted across terrain height variability per frequency band (middle row of plots). The average spectral power for younger and older adults is represented by a bold line and showcased without participant variability and after rescaling (last row of plots). The curves between younger and older adults for spectral power in the theta and alpha bands are statistically different. Specifically, older adults display a reduced range in utilized spectral power and greater synchronization and desynchronization in the theta and alpha spectral power bands, respectively, at low terrain unevenness than younger adults. Asterisk denotes curves that are significantly different from one another.

https://doi.org/10.1371/journal.page.0000017.g005

The best-fit quadradic curves were found for the average spectral power of electrocortical data across terrain height variability for the other brain sources (Supplementary Material 2; S.Table 2 in S1 File). All identified brain sources in our analysis displayed significantly different curves between younger and older adults in at least one spectral band. However, only the precuneus and left temporal brain sources show curves that are visually representative of CRUNCH. These results are not discussed in detail in this manuscript to focus our analysis on high order processing regions. In addition, supplementary analyses were performed to assess the robustness of our findings and explore alternative approaches for capturing CRUNCH. Specifically, CRUNCH curves were assessed for participants whose electrocortical data displayed a good fit to a quadratic curve (R2 > 0.5, 0.75, 0.9; S.Table 3 in S1 File) and whose performance did worsen with increasing terrain height variability (S.Table 4 in S1 File). CRUNCH curves were assessed when controlling for walking speed (S.Table 5 in S1 File) and when considering a subset of younger and older adult participants whose walking speeds matched (S.Table 6 in S1 File). CRUNCH curves were assessed with respect to the average spectral power activity in this study and other quantifying metrics (minimum, maximum, range) of electrocortical activity (S.Table 7 in S1 File). Lastly, CRUNCH curves were assessed between high and low functioning older adults (S.Table 8; Supplementary Materials 3 in S1 File).

3. Discussion

The objective of this study was to characterize the electrocortical activity via mobile EEG during uneven terrain walking, testing whether the Compensation Related Utilization of Neural Circuits Hypothesis (CRUNCH) provides a fitting explanatory framework. We identified nine electrocortical brain sources across younger and older adults for uneven terrain walking: left posterior parietal, right posterior parietal, left sensorimotor, right sensorimotor, left pre-supplementary motor, right premotor, precuneus, mid-cingulate, and temporal regions. While our analysis did not yield brain sources in the prefrontal and anterior cingulate as hypothesized, alternative higher order processing areas were identified in the left and right posterior parietal regions. In alignment with our hypothesis, we observed task-related changes in EEG that are consistent with the CRUNCH model. Specifically, we observed that with increasing terrain unevenness, there was an increase in theta (4–8 Hz) band and reduction in the alpha (8–13 Hz) band spectral power of the posterior parietal regions. Furthermore, older adults displayed greater theta synchronization and alpha desynchronization than younger adults at the low terrain unevenness level and younger adults displayed a greater range in utilized spectral power in the theta and alpha bands across all levels of terrain unevenness than older adults.

Older adults displaying greater theta synchronization and alpha desynchronization than younger adults at low terrain unevenness is consistent with over-recruitment of cognitive resources. To date, the benefit or consequence of over-recruitment for cognitive function is not well understood [16,18]. Over-recruitment may be a beneficial compensatory mechanism to achieve age-equivalent performance at low levels of task demand due to declining efficiency in the corresponding brain area or other brain regions. Alternatively, or in addition to, over-recruitment may be compensating for increased noise or reduced precision in cognitive processes [18]. Consequences of over-recruitment may be due to insufficient cognitive strategies, lack of inhibition in brain regions, or inability to select and specify neural processes [18]. To interpret over-recruitment, cognitive processes and performance data should be assessed across multiple difficulty levels [16]. Our supplementary investigation of CRUNCH between younger and older adults with matched walking speeds suggests some benefit of over-recruitment in this study. That is, a portion of older adults could navigate the uneven terrain at similar walking speeds to younger adults, displaying greater cognitive resources at low terrain unevenness. While over-recruitment was beneficial to maintain walking speed in older adults, the utilized resources did not avoid increases in gait parameter variability. Older adults showed greater synchronization or desynchronization at lower task demands and greater gait variability compared to younger adults. Thus, additional investigation is needed to assess gait stability in parallel with over-recruitment.

Younger adults displayed a greater range of average theta and alpha spectral power across terrains than older adults. This may indicate a limit (i.e., ceiling/floor effects) in available cognitive resources for older adults. Such limits are thought to be linked to recruitment impairment or brain atrophy [18]. Limits at higher terrain unevenness may have driven the greater increases in gait parameter variability for older adults than younger adults. That is, older adults may have not had the available resources to compensate for the more difficult terrain, resulting in greater increases in gait parameter variability than younger adults at greater terrain difficulty. This agrees with previous CRUNCH literature on cognitive tasks that has shown ceiling effects to be accompanied by poorer performance [17,18].

The lack of a prefrontal cortex cluster in our data of walking across uneven terrain makes it difficult to directly compare our results to previous CRUNCH literature on cognitive tasks [29,40]. A recent fNIRS analysis from this Mind in Motion project found greater prefrontal activity in older adults compared to younger adults walking on uneven terrain [41]. Specifically, in the same cohort, greater changes in the prefrontal oxyhemoglobin were observed in older adults than younger adults who shared similar walking abilities [41]. While more activation in the prefrontal area was displayed in older adults compared to younger adults via hemodynamic responses, these changes were not captured via our EEG analysis. We employed independent component analysis of scalp EEG, which extracts the strongest signals from the brain during the task. Prefrontal brain activity that is primarily arrhythmic, discrete bursts of neuronal firing, would likely have difficulty competing with the signal strength of ongoing rhythmic electrocortical fluctuations that are dominant throughout the brain during steady state gait. Sporadic prefrontal brain activity would generate an fNIRS signal indicating increased neural recruitment but may not produce high amplitude far field potentials. The prefrontal activity ongoing in our participants may not have been strong enough in comparison to the synchronized electrocortical rhythms in the brain during walking. This is one of the key differences in metrics used by fNIRS vs. EEG [22]. Past mobile EEG studies that examined large discrete perturbations to gait such as loss of balance during beam walking [27] or loss of vision [42] have found prefrontal and anterior cingulate clusters of activity. Those type of perturbation events are inherently different than steady state walking on uneven terrain. Gait perturbations generate event related potentials of high amplitude that propagate to many scalp electrodes with far field potentials. Because we did not obtain clusters of independent components from the prefrontal and anterior cingulate with our data, we could not directly compare our data with CRUNCH experiments using fMRI and fNIRS with cognitive tasks. However, failure to detect clusters in the prefrontal cortex and anterior cingulate with the analysis techniques used in this study does not imply an absence of activity in these regions. Region of interest analysis techniques may be better suited for capturing activity in the prefrontal and anterior cingulate regions.

There were other limitations to our current study. In this experiment, participants were required to walk on a treadmill at a set speed. While this speed was tailored to participant walking ability level, participants were not afforded the opportunity to disengage from the task (i.e., they must match their walking speed to the treadmill speed). While this is a strength in our experimental design to ensure brain activity was not confounded by changes in participant speed, it is also an important distinction from cognitive tasks where participants can disengage without consequence to their safety (i.e., a downturn in brain activity at higher task demand levels). Future research should investigate CRUNCH in mobility tasks that provide the participant with greater autonomy in sacrificing performance (e.g., overground walking). This work focused on assessing the potential of CRUNCH in mobility tasks. While the change in electrocortical data was accompanied by decreases in walking performance stability, this study did not directly link electrocortical data to participant ability and walking performance. However, secondary analyses on utilized cognitive resources in relation to participant ability and walking performance are presented in the supplementary materials. Additional knowledge could be gained from future work that links functioning ability and performance metrics during a broader range of mobility tasks. Further, this work investigated CRUNCH in the posterior parietal cortex. Supplementary analyses in other brain areas are provided to assist in guiding future work.

In conclusion, this is the first study to apply CRUNCH as a framework for mobility via electrocortical data. We observed patterns of brain activity consistent with CRUNCH in the posterior parietal cortex when walking on a treadmill with increasing terrain unevenness. Older adults showed greater activity than younger adults at low terrain unevenness, suggesting over-recruitment of cognitive resources, and an associated restricted range of activity at the higher terrain unevenness, suggesting limits of cognitive resources in the theta and alpha bands of the posterior parietal regions compared to younger adults. Quantifying the cognitive processing mechanisms of mobility is essential to understanding the complexity of mobility limitations. Future work can leverage this knowledge to improve mobility among older adults.

4. Materials and methods

4.1. Participants

Thirty-five healthy younger adults (19 females, 24 ± 4 years of age) and 96 older adults (58 females, 75 ± 7 years of age) participated in this study. Exclusion criteria comprised rheumatoid arthritis, neurological disorders that impair muscle function or mobility, dementia diagnosis or medication use, severe cardiovascular disorders that limit mobility, terminal illness, visual impairment that cannot be corrected, and implants with contraindication to Magnetic Resonance Imaging (MRI). Additional inclusion criteria for older adults comprised independence of daily living, signs of onset of physical ability decline (e.g., using hands for assistance when rising from a chair), absence of mild cognitive impairment assessed through the Montreal Cognitive Assessment (MoCA ≥ 26), and walking disability (i.e., inability to complete the 400 m walk test in less than 15 minutes without assistance) [28]. All participants provided written informed consent before participating in this study, which was approved by the University of Florida Institutional Review Board (IRB 201802227). Twenty-five older and four younger adults were excluded due data collection concerns that have been previously reported [43].

4.2. Data collection

4.2.1. Study visits.

Younger and older adult participants completed five baseline visits. Older adults completed follow-up visits every 6 months (up to 3.5 years). Baseline visits comprised: 1) an onsite eligibility screening, informed consent, and battery of clinical assessments; 2) an assessment of brain activity via EEG during actual and imagined walking on smooth and uneven treadmill surfaces; 3) a biomechanics assessment of various mobility tasks; 4) an assessment of brain activity via fNIRS during actual and imaged walking on smooth and uneven treadmill surfaces; and 5) an assessment of brain structure and function with multiple MRI sequences. This paper presents the cross-sectional results (i.e., baseline data) of clinical assessments and brain activity in younger and older adult participants via EEG during walking on smooth and uneven surfaces.

4.2.2. Onsite eligibility screening & clinical assessment.

Participants completed a battery of clinical assessments to ensure their eligibility for the study. Specifically, participants completed a 400 m walk test and the MoCA. Participants were excluded if they were unable to complete the 400 m walk test in less than 15 minutes without assistance or scored lower than a 26 on MoCA. Participants also completed the Short Physical Performance Battery (SPPB) to better characterize mobility limitation in this cohort. In addition, participants completed over-ground walks over uneven terrains. These walks were timed to tailor the treadmill speed for uneven terrain walks to the participant’s ability level.

4.2.3. Electroencephalogram setup for mobile brain imaging.

Participants wore a dual-layer EEG system, comprising of two caps and 248 electrodes (Brain Products GmbH, Germany) [24,26,44]. Electrodes recorded electrical activity at the scalp (120 electrodes), electrical activity of neck muscles (8 electrodes), and noise from electrode movement and environment artifact (120 electrodes). These electrodes are referred to as scalp, muscle and noise electrodes, respectively. The scalp electrodes followed a 10−05 system for electrode placement of 128 electrodes [45]. Eight of these electrodes (TP9, P9, PO9, O9, O10, PO10, P10, and TP10) were re-purposed for muscle electrodes. The muscle electrodes had a bipolar electrode placement on the sternocleidomastoid and upper trapezius muscles on both sides of the neck. A noise electrode was mechanically coupled to each scalp electrode (Fig 6). The scalp electrode faced the scalp, whereas the noise electrode faced the reverse direction of the scalp electrode. The primary cap in the dual-layer system was placed on the participant’s head. This cap secured each scalp-noise electrode couple in place. A secondary cap comprising conductive fabric was wrapped over the dual-layer EEG systems, to facilitate an artificial skin for the noise electrodes and minimize electrode movement. An abrasive electrolyte-gel and high-viscosity electrolyte-gel was used for the scalp and noise electrodes, respectively. The impedance of scalp electrodes was targeted below 15 k ohm prior to data collection. The impedance of ground and reference electrodes was targeted below 5 k ohm. Ground and reference electrodes were located at Fpz and CPz, respectively. The location of electrodes was digitized via a structural scanner (ST01, Optical Inc., San Francisco, CA, USA). EEG data was recorded from four LiveAmp 64 amplifiers at 500 Hz.

thumbnail
Download:
Fig 6. Bottom (a) and side (b) view of a coupled scalp and noise electrode.

These mechanically coupled electrodes in a dual configuration [24,44] improves removal of motion artifact from electrocortical data.

https://doi.org/10.1371/journal.page.0000017.g006

4.2.4. Wearable setup for mobility assessment.

Participants wore insole sensors (loadsol 1–184 sensor, Novel Electronics Inc., St. Paul, MN, USA) and an inertial measurement unit (IMU, Opal APDM Inc., Portland, OR, USA) to capture gait kinetics and kinematics, respectively. The insole sensors captured ground reaction forces at 200 Hz. The IMU was strapped around the participant’s waist and placed at the sacrum to measure waist kinematics at 128 Hz. Ground reaction force and IMU data were synchronized offline [46]. Ground reaction force data were synchronized via pulses that occurred continuously every 5 seconds. IMU data were synchronized via pulses at the start and stop of each trial. Participants wore a safety harness attached to an overhead beam during walking trials. The harness did not provide support during walking but would prevent the participant from falling to the ground if they were to trip

4.2.5. Mobility protocol.

Participants completed a 48-minute walking protocol on smooth and uneven treadmill surfaces (PPS 70 Bari-Mill, Woodway, Waukesha, WI, USA; 70 cm x 173 cm walking surface). The walking protocol consisted of two sections. One section varied speed conditions and the other section varied terrain conditions on the treadmill. Electrocortical data was processed from the combined speed and terrain conditions, enabling a robust dataset for data processing methods [47,48]. Speed conditions comprised walking on the flat treadmill (no terrain) at 0.25, 0.5, 0.75, and 1.0 m/s (this data is not reported in this manuscript and the effects of speed on cortical activity is reported in Salminen et al. 2025). Terrain conditions comprised walking on rigid foam disks of flat, low, medium, and high terrain (Fig 7). Disks were 12.7 cm in diameter, covering 43% of the walking terrain, and varied in color and height variability by terrain condition. The flat terrain comprised green disks painted on the treadmill (0 cm in height; 0 cm height variability). The low terrain comprised yellow disks of a singular height (100% at 1.3 cm in height; 0.6 cm height variability). The medium terrain comprised orange disks of two heights (50% at 1.3 cm in height; 50% at 2.5 cm in height; 1.0 cm height variability). The high terrain comprised red disks of three heights (30% at 1.3 cm in height; 20% at 2.5 cm in height; 50% at 3.8 cm in height; 1.6 cm height variability). Low, medium, and high terrain disks were rigid disks made from polyurethane and attached to the treadmill via hoop-and-loop fastener. Walking speed during the terrain trials was set to 75% of the participant’s slowest walking speed across over-ground walks of each uneven terrain condition. This participant-tailored speed was tested and adjusted, if needed, before starting uneven terrain walking trials. Participants completed two 3-minute walking trials of each speed and terrain condition. The order of trial type was pseudorandomized (selected from a set of eight orders) within each speed and terrain section. Between the speed and terrain sections, participants completed a 3-minute recorded sitting rest period.

thumbnail
Download:
Fig 7. Experimental setup of treadmill walking across speed and terrain conditions.

The terrain conditions were established by varying the heights of disks attached to the treadmill. The percentages indicate the proportion of disks at each specified height. This distribution is essential in quantifying the difficulty of walking terrain condition via terrain height variability. Participants wore an electroencephalography (EEG) system, inertial measurement unit (IMU) and insole force sensors. Participant safety was maintained with a safety harness that was connected to a fall arrest system.

https://doi.org/10.1371/journal.page.0000017.g007

4.2.6. Magnetic resonance imaging for customized head models.

Brain structural MRI data were collected using a T1-weighted magnetization prepared rapid gradient echo (MPRAGE) sequence via a Siemens 3T MAGNETOM Prisma scanner with a 64-channel head coil. The imaging parameters comprised a repetition time = 2000 ms, echo time = 2.99 ms, flip angle = 8°, voxel resolution = 0.8 mm3, and field of view = 256 × 256 × 167 mm2 (4:22 minutes of scan time).

4.3. Data analysis

4.3.1. Electrocortical data preprocessing.

EEG data was processed with custom scripts in Matlab (R2020b), utilizing EEGLAB [49] and the BeMoBIL pipeline [50]. Preprocessing was consistent with processing procedures for mobile brain imaging [47,48,51], comprising high-pass filtering, line noise removal, channel rejection, average referencing, and artifact rejection. Details of our preprocessing procedures are as followed with EEGLAB functions and plugins italicized. Drift was removed from all scalp, noise and muscle channels through a 1 Hz high-pass filter (−6 dB at 0.5 Hz; eegfiltnew). An additional 20 Hz high-pass filter was performed on muscle channels. Line noise was removed at 60 Hz and 120 Hz (CleanLine). Scalp and noise channels were rejected if the average value of the channel was more than 3 standard deviations away from the average value of all scalp and noise channels, respectively (select). All scalp, noise and muscle channels were referenced to the average scalp, noise and muscle channels, respectively (reref). Data that was highly correlated with noise channel data was removed from scalp channel (R2 = 0.65; 4-second moving window) and muscle channel (R2 = 0.4; 4-second moving window) data (iCanClean) [52,53]. Additional noisy channels and time frames were removed (clean_artifacts). Scalp channels were referenced to the average again. Preprocessing parameters were identified from a preliminary analysis on a subset of data that minimized the number of channels and time frames rejected while maximizing the number of brain components identified by ICLabel [51,54]. To identify the electrocortical source signal, an adaptive mixture independent component analysis (AMICA) was performed on the preprocessed scalp channel data [55]. This analysis decomposed the scalp data (electrocortical data at the channel level) into statistically independent components. The signatures of these independent components determined the locations of these components in the cortex (i.e., brain source or electrocortical data at the source level).

4.3.2. Participant-specific volume conduction head model and source localization.

Participant-specific head models were created to obtain accurate brain source localization. Specifically, we modeled the volume of the various tissue layers of the head from the T1-weighted MRI scan for each participant. Each tissue type was assigned to standardized conduction properties. Processes for this model were performed in Fieldtrip (v. 20210910). The head image was resliced to perform tissue segmentation using headreco from SimNIBS toolbox (v 3.2). The head image was segmented into six tissues layers (scalp, skull, air, cerebrospinal fluid, gray matter and white matter). Hexahedral meshes were generated across the tissue layers [51]. To co-register the head image to the EEG electrode locations (align the coordinate systems), the fiducial locations (left/right preauricular, nasion) on the MRI image were digitized and aligned with the digitized EEG electrode location space with marked fiducials. Then the leadfield matrix was computed using the SIMBIO toolbox with a 5 mm separation distribution of source position in the gray matter.

Brain source localization was performed with equivalent dipole fitting (ft_dipolefitting function in Fieldtrip toolbox). Dipole locations were wrapped to the Montreal Neurological Institute (MNI) template [39] for both younger and older adults via ANTs normalization [56]. Independent components were classified as brain components (as opposed to other physiological signals) if the source met the following criteria: had a 50% or greater probability of brain classification (ICLabel) [54]; a negative power spectral density slope between 2 and 40 Hz; a residual variance of dipole fitting less than 15%; and the dipole was located inside the brain. The average number of sources classified as brain components per person was 13. Younger adults (15 ± 5 brain components) showcased a greater number of brain components than older adults (12 ± 5 brain components) (). Ten older adult participants were excluded from our analysis for displaying fewer than 5 total brain components.

4.3.3. K-means clustering of brain sources.

Brain components were clustered by dipole location via k-means clustering (k = 11) in EEGLAB. Clusters with more than half the younger adults (n ≥ 16) and half of the older adults (n ≥ 31) were retained for further group analysis. Components greater than three standard deviations away from the centroid of the cluster were identified as outliers and excluded. If multiple components per participant were found within a cluster, only the component with the highest brain classification from ICLabel was retained.

4.3.4. Electrocortical data postprocessing.

Postprocessing was performed on electrocortical data to correct for muscle artifacts in power spectral density (PSD) plots. Specifically, spectral principal component analysis (sPCA) [57] was employed to remove muscle artifact from brain components. Details of these procedures are described in Salminen et al. 2025.

4.3.5. Computing power spectral density for each cluster.

The power spectral density (PSD) and the flattened PSD was calculated for each cluster. Specifically, the resting electrocortical data was applied with the fitting oscillations and one-over frequency (FOOOF) to separate the aperiodic and periodic components of the PSD curve [58]. FOOOF parameters were set to a range of power spectra of 3–40 Hz; peak width limits of 1–8; minimum peak height of 0.05; and maximum number of peaks at 2. The flattened PSD was obtained by subtracting the aperiodic component from the original PSD. The average, minimum, maximum, and range of power was calculated for each frequency band of interest: theta (4–8 Hz), alpha (8–13 Hz) and beta (13–30 Hz) from the flattened PSD.

4.3.6. CRUNCH analysis.

The electrocortical data was fitted to a quadratic curve to assess the validity of the CRUNCH framework for brain activity changes with increasing terrain unevenness. To create these curves, terrain unevenness (ordinal data) was transformed into a continuous metric from the quantified variability in height of the distributed disks per terrain condition, referred to as terrain height variability. Quadratic curves were estimated from the average spectral power across terrain height variability. A quadratic curve (Equation 1) was estimated for each participant by frequency band: theta (4–8 Hz), alpha (8–13 Hz) and beta (13–30 Hz) within each brain cluster. The electrocortical data was normalized for each participant by subtracting the average spectral power in the flat walking terrain condition from the flat (setting the spectral power to zero), low, medium and high walking terrain conditions. All quadradic models were forced through the zero spectral power data at the flat terrain condition (height terrain variability = 0), resulting in no intercept value among curves. After this curve normalization, the intercept was set to the average spectral power in the flat walking condition for each participant. Curve characteristics were extracted from each best-fit quadradic equation (sum of least squared errors). Specifically, the a-coefficient, b-coefficient, c-coefficient (intercept), terrain height variability at the vertex, curve-quantified electrocortical power at the low terrain condition, and the curve-quantified power at the curve vertex were extracted. The terrain height variability at the vertex is representative of CRUNCH (i.e., difficulty level where cognitive resources peak) and is referred to hereafter as the CRUNCH parameter. The curve-quantified power at the low terrain condition and vertex are representative of over-recruitment and ceiling effects, respectively, and are referred to as so hereafter. Extracted curve parameters that fell outside 3 times the interquartile range by age group were excluded from statistical analysis (extreme outliers)

Eq.(1)

To perform sub-analyses, curve characteristics were also extracted from best-fit quadratics of the minimum, maximum, and range of power from the flattened PSD by frequency band, participant, and cluster.

4.3.7. Walking performance analyses.

Kinematic and kinetic data was extracted from insole force sensors and an IMU above the sacrum to characterize walking performance. Foot strike and foot off were defined by the ground reaction force exceeding and falling below a 20 N threshold, respectively [46]. The variation in step time and medial-lateral sacral excursion were extracted gait parameters for analysis of walking performance in this study due to previous work showcasing these parameters to capture challenged walking in uneven terrains [46]. Step time was quantified from the time between foot strike of one foot to the foot strike of the contralateral foot. Medial-lateral sacral excursion was quantified from IMU data [46]. Outliers beyond ± 2.5 standard deviations were excluded. The variability of these performance metrics was quantified from the coefficient of variation (standard deviation over mean), where a greater variability is indicative a worse walking performance. To assess if walking performance varied between younger and older adults across terrain height variability, best-fit linear regressions were found for each participant. Linear regressions were utilized to model walking performance to align with performance decline trends in previous CRUNCH literature [15,17–19,31]. Line characteristics were extracted from each best-fit linear regression equation (sum of least squared errors). Specifically, the a-coefficient (slope) and the b-coefficient (intercept) were extracted. An increasing slope would indicate walking performance to worsen with increasing terrain height variability. To perform sub-analyses, line characteristics were also extracted from best-fit linear regressions on the variation of other gait parameters.

4.4. Statistical analysis

To test our hypothesis, general least squared models were applied on the quadratic parameters (a-coefficient, b-coefficient, c-coefficient, CRUNCH parameter, over-recruitment, ceiling effect) of the average electrocortical activity data across terrain height variability by frequency band and brain area. Age group across all parameters was added to the model to investigate if the curves varied by age group. Similarly, to assess walking performance, general least squared models were applied to the estimated linear regression parameters (a-coefficient, b-coefficient) of variation in gait parameters across terrain height variability. Age group across all parameters was added to the model to investigate if the linear trends in walking performance varied by age group. To ensure normality, a square-root transform after a positive shift was applied to the assessed data. In addition, to assess differences in participant demographics across age groups, Welch t-tests were performed for normally distributed data and Wilcoxon Ranked Sum tests for non-normally distributed data.

Supporting information

S1 File. S1 Data.

Links to available data on OpenNeuro. S1 Material. Average coefficient of variation by gait parameter. S1 Fig. Coefficient of variation across terrain for step time, gait speed, and stance phase. S2 Fig. Coefficient of variation across terrain for swing phase, single support time, and double support time. S3 Fig. Coefficient of variation across terrain for medial-lateral excursion, anterior-posterior excursion, and superior-inferior excursion. S1 Table. Linear line comparison between younger and older adults via a General Least Squared (GLS) model for all coefficient of variance of gait parameters. S2 Material. Average spectral power of electrocortical data by brain source. S4 Fig. Spectral power of electrocortical data across terrain for the left posterior parietal cortex. S5 Fig. Spectral power of electrocortical data across terrain for the right posterior parietal cortex. S6 Fig. Spectral power of electrocortical data across terrain for the left sensorimotor cortex. S7 Fig. Spectral power of electrocortical data across terrain for the right sensorimotor cortex. S8 Fig. Spectral power of electrocortical data across terrain for the left pre-supplementary motor cortex. S9 Fig. Spectral power of electrocortical data across terrain for the right motor cortex. S10 Fig. Spectral power of electrocortical data across terrain for the precuneus cortex. S11 Fig. Spectral power of electrocortical data across terrain for the mid-cingulate cortex. S12 Fig. Spectral power of electrocortical data across terrain for the left temporal cortex. S2 Table. CRUNCH curve comparison between younger and older adults via a General Least Squared (GLS) model for all brain source clusters by frequency band. S3 Table. CRUNCH sub-analysis: Extraction of people whose electrocortical data has a poor fit to a quadric curve by R2 value. S4 Table. CRUNCH sub-analysis: Extraction of people whose performance does not worsen, as define by a negative slope in walking performance variation. S5 Table. CRUNCH sub-analysis: Including walking speed as a covariate in the model. S6 Table. CRUNCH sub-analysis: Assessed for age-equivalent performance – based on walking speed. S7 Table. CRUNCH sub-analysis: Curves with respect to the average and other quantifying metrics of electrocortical activity (minimum, maximum, range). S8 Table. CRUNCH sub-analysis: Assessed for high and low functioning older adults. S3 Material. Average spectral power of electrocortical data across the left and right posterior parietal brain sources for high and low functioning older adults. S13 Fig. Spectral power of electrocortical data across terrain for the left posterior parietal cortex in high and low functioning older adults. S14 Fig. Spectral power of electrocortical data across terrain for the right posterior parietal cortex in high and low functioning older adults.

https://doi.org/10.1371/journal.page.0000017.s001

(ZIP)

Acknowledgments

This study was supported by the National Institute of Health (U01AG061389) for all authors. National Institute of Health grants F32AG072808 and T32AG062728 supported author EMP. American Heart Association Fellowship (23POST1011634, doi.org/10.58275/AHA.23POST1011634.pc.gr.161292) partially supported author CL. DPF was also supported by National Institutes of Health (R01NS104772). A portion of this work was performed in the McKnight Brain Institute, which is supported by National Science Foundation Cooperative Agreement No. DMR- 1644779 and the State of Florida, and in part by an NIH award, S10 OD021726, for High End Instrumentation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

  1. 1. Webber SC, Porter MM, Menec VH. Mobility in older adults: a comprehensive framework. Gerontologist. 2010;50(4):443–50. pmid:20145017
  2. 2. Rantakokko M, Mänty M, Rantanen T. Mobility decline in old age. Exerc Sport Sci Rev. 2013;41(1):19–25. pmid:23038241
  3. 3. Brown CJ, Flood KL. Mobility limitation in the older patient: a clinical review. JAMA. 2013;310(11):1168–77. pmid:24045741
  4. 4. Musich S, Wang SS, Ruiz J, Hawkins K, Wicker E. The impact of mobility limitations on health outcomes among older adults. Geriatr Nurs. 2018;39(2):162–9. pmid:28866316
  5. 5. Brown CJ, Friedkin RJ, Inouye SK. Prevalence and outcomes of low mobility in hospitalized older patients. J Am Geriatr Soc. 2004;52(8):1263–70. pmid:15271112
  6. 6. Kabiri M, Brauer M, Shafrin J, Sullivan J, Gill TM, Goldman DP. Long-Term Health and Economic Value of Improved Mobility among Older Adults in the United States. Value Health. 2018;21(7):792–8. pmid:30005751
  7. 7. Hardy SE, Kang Y, Studenski SA, Degenholtz HB. Ability to walk 1/4 mile predicts subsequent disability, mortality, and health care costs. J Gen Intern Med. 2011;26(2):130–5. pmid:20972641
  8. 8. Freiberger E, Sieber CC, Kob R. Mobility in Older Community-Dwelling Persons: A Narrative Review. Front Physiol. 2020;11:881. pmid:33041836
  9. 9. Maresova P, Krejcar O, Maskuriy R, Bakar NAA, Selamat A, Truhlarova Z, et al. Challenges and opportunity in mobility among older adults - key determinant identification. BMC Geriatr. 2023;23(1):447. pmid:37474928
  10. 10. Hicks GE, Shardell M, Alley DE, Miller RR, Bandinelli S, Guralnik J, et al. Absolute strength and loss of strength as predictors of mobility decline in older adults: the InCHIANTI study. J Gerontol A Biol Sci Med Sci. 2012;67(1):66–73. pmid:21546582
  11. 11. Visser M, Goodpaster BH, Kritchevsky SB, Newman AB, Nevitt M, Rubin SM, et al. Muscle mass, muscle strength, and muscle fat infiltration as predictors of incident mobility limitations in well-functioning older persons. J Gerontol A Biol Sci Med Sci. 2005;60(3):324–33. pmid:15860469
  12. 12. Lord SR, Lloyd DG, Li SK. Sensori-motor function, gait patterns and falls in community-dwelling women. Age Ageing. 1996;25(4):292–9. pmid:8831874
  13. 13. Bouça-Machado R, Maetzler W, Ferreira JJ. What is Functional Mobility Applied to Parkinson’s Disease?. J Parkinsons Dis. 2018;8(1):121–30. pmid:29480225
  14. 14. Sutliff MH. Contribution of impaired mobility to patient burden in multiple sclerosis. Curr Med Res Opin. 2010;26(1):109–19. pmid:19916707
  15. 15. Iordan AD, Cooke KA, Moored KD, Katz B, Buschkuehl M, Jaeggi SM, et al. Neural correlates of working memory training: Evidence for plasticity in older adults. Neuroimage. 2020;217:116887. pmid:32376302
  16. 16. Fettrow T, Hupfeld K, Tays G, Clark DJ, Reuter-Lorenz PA, Seidler RD. Brain activity during walking in older adults: Implications for compensatory versus dysfunctional accounts. Neurobiol Aging. 2021;105:349–64. pmid:34182403
  17. 17. Cappell KA, Gmeindl L, Reuter-Lorenz PA. Age differences in prefontal recruitment during verbal working memory maintenance depend on memory load. Cortex. 2010;46(4):462–73. pmid:20097332
  18. 18. Reuter-Lorenz PA, Cappell KA. Neurocognitive Aging and the Compensation Hypothesis. Curr Dir Psychol Sci. 2008;17(3):177–82.
  19. 19. Mattay VS, Fera F, Tessitore A, Hariri AR, Berman KF, Das S, et al. Neurophysiological correlates of age-related changes in working memory capacity. Neurosci Lett. 2006;392(1–2):32–7. pmid:16213083
  20. 20. Cabeza R, Albert M, Belleville S, Craik F, Duarte A, Grady C, et al. Cognitive neuroscience of healthy aging: Maintenance, reserve, and compensation. Nature Reviews Neuroscience. 2018;19(11):701.
  21. 21. Leff DR, Orihuela-Espina F, Elwell CE, Athanasiou T, Delpy DT, Darzi AW, et al. Assessment of the cerebral cortex during motor task behaviours in adults: a systematic review of functional near infrared spectroscopy (fNIRS) studies. Neuroimage. 2011;54(4):2922–36. pmid:21029781
  22. 22. Richer N, Bradford JC, Ferris DP. Mobile neuroimaging: What we have learned about the neural control of human walking, with an emphasis on EEG-based research. Neurosci Biobehav Rev. 2024;162:105718. pmid:38744350
  23. 23. Gwin JT, Gramann K, Makeig S, Ferris DP. Electrocortical activity is coupled to gait cycle phase during treadmill walking. Neuroimage. 2011;54(2):1289–96. pmid:20832484
  24. 24. Nordin AD, Hairston WD, Ferris DP. Human electrocortical dynamics while stepping over obstacles. Sci Rep. 2019;9(1):4693. pmid:30886202
  25. 25. Peterson SM, Ferris DP. Differentiation in Theta and Beta Electrocortical Activity between Visual and Physical Perturbations to Walking and Standing Balance. eNeuro. 2018;5(4):ENEURO.0207-18.2018. pmid:30105299
  26. 26. Richer N, Downey RJ, Hairston WD, Ferris DP, Nordin AD. Motion and Muscle Artifact Removal Validation Using an Electrical Head Phantom, Robotic Motion Platform, and Dual Layer Mobile EEG. IEEE Trans Neural Syst Rehabil Eng. 2020;28(8):1825–35. pmid:32746290
  27. 27. Sipp AR, Gwin JT, Makeig S, Ferris DP. Loss of balance during balance beam walking elicits a multifocal theta band electrocortical response. J Neurophysiol. 2013;110(9):2050–60. pmid:23926037
  28. 28. Clark DJ, Manini TM, Ferris DP, Hass CJ, Brumback BA, Cruz-Almeida Y, et al. Multimodal Imaging of Brain Activity to Investigate Walking and Mobility Decline in Older Adults (Mind in Motion Study): Hypothesis, Theory, and Methods. Front Aging Neurosci. 2020;11:358. pmid:31969814
  29. 29. Cabeza R, Daselaar SM, Dolcos F, Prince SE, Budde M, Nyberg L. Task-independent and task-specific age effects on brain activity during working memory, visual attention and episodic retrieval. Cereb Cortex. 2004;14(4):364–75. pmid:15028641
  30. 30. Gerver CR, Neely KA, Kurkela KA, Diaz MT, Goodman JT, Blouch S, et al. Shared neural recruitment across working memory and motor control tasks as a function of task difficulty and age. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn. 2020;27(6):864–79. pmid:31877068
  31. 31. Van Ruitenbeek P, Santos Monteiro T, Chalavi S, King BR, Cuypers K, Sunaert S, et al. Interactions between the aging brain and motor task complexity across the lifespan: balancing brain activity resource demand and supply. Cereb Cortex. 2023;33(10):6420–34. pmid:36587289
  32. 32. Reuter-Lorenz PA, Park DC. How does it STAC up? Revisiting the scaffolding theory of aging and cognition. Neuropsychol Rev. 2014;24(3):355–70. pmid:25143069
  33. 33. Heuninckx S, Wenderoth N, Debaere F, Peeters R, Swinnen SP. Neural basis of aging: the penetration of cognition into action control. J Neurosci. 2005;25(29):6787–96. pmid:16033888
  34. 34. Jacobsen NA, Ferris DP. Electrocortical activity correlated with locomotor adaptation during split-belt treadmill walking. J Physiol. 2023;601(17):3921–44. pmid:37522890
  35. 35. Whitlock JR. Posterior parietal cortex. Curr Biol. 2017;27(14):R691–5. pmid:28743011
  36. 36. Cavada C, Goldman-Rakic PS. Posterior parietal cortex in rhesus monkey: II. Evidence for segregated corticocortical networks linking sensory and limbic areas with the frontal lobe. J Comp Neurol. 1989;287(4):422–45. pmid:2477406
  37. 37. Shah VA, Cruz-Almeida Y, Roy A, Cenko E, Downey RJ, Ferris DP, et al. Correlates of gait speed changes during uneven terrain walking in older adults: differential roles of cognitive and sensorimotor function. Exp Brain Res. 2025;243(3):72. pmid:39976706
  38. 38. Liu C, Downey RJ, Salminen JS, Rojas SA, Richer N, Pliner EM, et al. Electrical brain activity during human walking with parametric variations in terrain unevenness and walking speed. Imaging Neurosci (Camb). 2024;2:imag–2–00097. pmid:39989610
  39. 39. Mazziotta J, Toga A, Evans A, Fox P, Lancaster J, Zilles K, et al. A probabilistic atlas and reference system for the human brain: International Consortium for Brain Mapping (ICBM). Philos Trans R Soc Lond B Biol Sci. 2001;356(1412):1293–322. pmid:11545704
  40. 40. Reuter-Lorenz PA, Jonides J, Smith EE, Hartley A, Miller A, Marshuetz C, et al. Age differences in the frontal lateralization of verbal and spatial working memory revealed by PET. J Cogn Neurosci. 2000;12(1):174–87. pmid:10769314
  41. 41. Hwang J, Liu C, Winesett SP, Chatterjee SA, Gruber AD, Swanson CW. Prefrontal cortical activity during uneven terrain walking in younger and older adults. Frontiers in Aging Neuroscience. 2024;16:1389488.
  42. 42. Symeonidou ER, Ferris DP. Visual occlusions result in phase synchrony within multiple brain regions involved in sensory processing and balance control. IEEE Transactions on Neural Systems and Rehabilitation Engineering. 2023;31:3772–80.
  43. 43. Liu C, Pliner EM, Salminen J, Downey RJ, Hwang J, Roy A, et al. Age differences in electrocortical dynamics during uneven terrain walking. Imaging Neurosci (Camb). 2025;3:IMAG.a.1039. pmid:41322366
  44. 44. Nordin AD, Hairston WD, Ferris DP. Dual-electrode motion artifact cancellation for mobile electroencephalography. J Neural Eng. 2018;15(5):056024. pmid:30074489
  45. 45. Oostenveld R, Praamstra P. The five percent electrode system for high-resolution EEG and ERP measurements. Clin Neurophysiol. 2001;112(4):713–9. pmid:11275545
  46. 46. Downey RJ, Richer N, Gupta R, Liu C, Pliner EM, Roy A, et al. Uneven terrain treadmill walking in younger and older adults. PLoS One. 2022;17(12):e0278646. pmid:36534645
  47. 47. Liu C, Downey RJ, Salminen JS, Rojas SA, Richer N, Pliner EM, et al. Electrical brain activity during human walking with parametric variations in terrain unevenness and walking speed. bioRxiv. 2023.
  48. 48. Salminen J, Liu C, Pliner EM, Tenerowicz M, Roy A, Richer N, et al. Gait speed-related changes in electrocortical activity in younger and older adults. J Neurophysiol. 2025;133(6):1761–94. pmid:40272781
  49. 49. Delorme A, Makeig S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods. 2004;134(1):9–21. pmid:15102499
  50. 50. Klug M, Jeung S, Wunderlich A, Gehrke L, Protzak J, Djebbara Z, et al. The BeMoBIL Pipeline for automated analyses of multimodal mobile brain and body imaging data. BioRxiv. 2022:2022.09. 29.510051.
  51. 51. Liu C, Downey RJ, Mu Y, Richer N, Hwang J, Shah VA, et al. Comparison of EEG Source Localization Using Simplified and Anatomically Accurate Head Models in Younger and Older Adults. IEEE Trans Neural Syst Rehabil Eng. 2023;31:2591–602. pmid:37252873
  52. 52. Downey RJ, Ferris DP. iCanClean removes motion, muscle, eye, and line-noise artifacts from phantom EEG. Sensors. 2023;23(19):8214.
  53. 53. Gonsisko CB, Ferris DP, Downey RJ. iCanClean Improves Independent Component Analysis of Mobile Brain Imaging with EEG. Sensors (Basel). 2023;23(2):928. pmid:36679726
  54. 54. Pion-Tonachini L, Kreutz-Delgado K, Makeig S. ICLabel: An automated electroencephalographic independent component classifier, dataset, and website. Neuroimage. 2019;198:181–97. pmid:31103785
  55. 55. Palmer JA, Kreutz-Delgado K, Makeig S. AMICA: An adaptive mixture of independent component analyzers with shared components. Swartz Center for Computational Neuroscience, University of California San Diego. 2012.
  56. 56. Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage. 2011;54(3):2033–44. pmid:20851191
  57. 57. Seeber M, Scherer R, Wagner J, Solis-Escalante T, Müller-Putz GR. High and low gamma EEG oscillations in central sensorimotor areas are conversely modulated during the human gait cycle. Neuroimage. 2015;112:318–26. pmid:25818687
  58. 58. Donoghue T, Haller M, Peterson EJ, Varma P, Sebastian P, Gao R, et al. Parameterizing neural power spectra into periodic and aperiodic components. Nat Neurosci. 2020;23(12):1655–65. pmid:33230329