About

James Finley is a researcher whose work focuses on biomechanics, neurorehabilitation, and motor control, particularly in relation to walking and balance in individuals post-stroke and with neurological conditions such as Parkinson's disease. His research involves the development and application of virtual reality, biofeedback, and digital methods to assess and improve gait and balance, aiming to understand the underlying mechanisms of locomotor adaptation and stability. Finley's contributions include investigating the effects of asymmetry, external assistance, and balance control strategies on walking performance, as well as designing innovative mobility training tools and interventions. His work emphasizes the importance of understanding individual differences in locomotor function and leveraging external mechanical work to reduce metabolic costs and enhance rehabilitation outcomes.

Research topics

• Computer Science
• Artificial Intelligence
• Psychology
• Physical medicine and rehabilitation
• Neuroscience
• Medicine
• Cognitive psychology
• Anatomy
• Biology
• Physiology
• Ecology
• Gerontology
• Human–computer interaction
• Physical therapy
• Applied psychology
• Engineering

Selected publications

• Uncovering locomotor learning dynamics in people with Parkinson’s disease

  PLoS ONE · 2025-07-31

  articleOpen accessSenior authorCorresponding

  Locomotor learning is important for improving gait and balance impairments in people with Parkinson's disease (PD). While PD disrupts neural networks involved in motor learning, there is a limited understanding of how PD influences the time course of locomotor learning and retention. Here, we used a virtual obstacle negotiation task to investigate whether the early stages of PD affect the acquisition and retention of locomotor skills. On Day 1, 15 participants with PD (Hoehn and Yahr Stage 1-2) and 20 age-matched controls were instructed to achieve a specified level of foot clearance while repeatedly stepping over two different virtual obstacles on a treadmill. We assessed online performance improvement on Day 1 and overnight retention after at least 24 hours on Day 2. We used a hierarchical Bayesian state-space model to estimate the learning rate and the degree of interference between the two obstacles. There was a 93% probability that people with PD learned the locomotor skill faster than controls, but there was limited evidence of group differences in interference between the two heights of obstacles. Both groups improved their performance to a similar magnitude during skill acquisition and performed similarly during retention on Day 2. Notably, a slower learning rate was associated with greater online performance improvement, while lower interference was linked to better overnight retention, and this effect was strongest for the control group. These results highlight that people with early-stage PD retain the ability to use multisensory information to acquire and retain locomotor skills. In particular, our finding that people with early-stage PD learned faster than age-matched controls may reflect the emergence of compensatory motor learning strategies used to offset early motor impairments in people with PD.

  PublisherDOI

• Author response for "Aging modulates the effects of scene complexity on visual search and target selection in virtual environments"

  2025-09-28

  peer-reviewSenior author
  PublisherDOI

• Frequency and leg stiffness adaptation in human vertical hopping before, during and after added load

  Journal of Experimental Biology · 2025-10-27

  articleOpen access

  Terrestrial animal gaits often use spring-like mechanics to enhance movement economy through elastic energy cycling. Hopping is a relatively simple, constrained task, yet retains essential features of bouncing gaits, requiring cyclic regulation of limb stiffness and generation of high muscle forces to support body weight and enable elastic energy cycling. We investigated how humans adjust hopping frequency and leg stiffness before, during and after experiencing added load. Eighteen participants hopped bipedally for 90 s per trial, with hop frequency and height unconstrained, while kinematic, ground reaction force and ankle muscle electromyographic (EMG) data were collected. We analysed mechanics across four conditions: initial body weight (BWi), two added mass trials (+10% and +20% BW) and final body weight (BWf). With added mass, participants increased leg stiffness and maintained a consistent hopping frequency (∼2.15 Hz); yet, when returning to BWf, the elevated leg stiffness was maintained and hopping frequency increased (to ∼2.36 Hz) and reduced centre of mass (CoM) work per hop. BWf adaptations were driven by greater ankle stiffness, leading to less ankle work. Adaptation rates were consistent across trials, with steady-state mechanics reached in ∼30-40 s. Muscle coactivation decreased following BWi. Triceps surae mean EMG was unchanged with added mass and reduced in BWf. Similar patterns of adaptation were observed in bouncing without an aerial phase. Substantial inter-individual variability was observed in preferred hopping mechanics and adaptation strategy. Together, added mass and increased task familiarity led participants to recalibrate their hopping strategy. Based on literature evidence, the adaptations may align with reduced metabolic cost.

  PublisherOA PDFDOI

• Identification of Subtypes of Post-Stroke and Neurotypical Gait Behaviors Using Neural Network Analysis of Gait Cycle Kinematics

  IEEE Transactions on Neural Systems and Rehabilitation Engineering · 2025-01-01 · 4 citations

  articleOpen access

  Gait impairment post-stroke is highly heterogeneous. Prior studies classified heterogeneous gait patterns into subgroups using peak kinematics, kinetics, or spatiotemporal variables. A limitation of this approach is the need to select discrete features in the gait cycle. Using continuous gait cycle data, we accounted for differences in magnitude and timing of kinematics. Here, we propose a machine-learning pipeline combining supervised and unsupervised learning. We first trained a Convolutional Neural Network and a Temporal Convolutional Network to extract features that distinguish impaired from neurotypical gait. Then, we used unsupervised time-series k-means and Gaussian Mixture Models to identify gait clusters. We tested our pipeline using kinematic data of 28 neurotypical and 39 individuals post-stroke. We assessed differences between clusters using ANOVA. We identified two neurotypical gait clusters (C1, C2). C1: normative gait pattern. C2: shorter stride time. We observed three post-stroke gait clusters (S1, S2, S3). S1: mild impairment and increased bilateral knee flexion during loading response. S2: moderate impairment, slow speed, short steps, increased knee flexion during stance bilaterally, and reduced paretic knee flexion during swing. S3: mild impairment, asymmetric swing time, increased ankle abduction during the gait cycle, and reduced dorsiflexion bilaterally. Our results indicate that joint kinematics post-stroke are mostly distinct from controls, and highlight kinematic impairments in the non-paretic limb. The post-stroke clusters showed distinct impairments that would require different interventions, providing additional information for clinicians about rehabilitation targets.

  PublisherDOI

• Validity Of Between-limb Symmetry For Insole Sensors During Daily Activities

  Medicine & Science in Sports & Exercise · 2025-09-16

  article

  Restoring loading symmetry is often a target of recovery after neurologic and musculoskeletal injuries. Current understanding of lower extremity loading is limited to laboratory data collections using force plates. Understanding loading behaviors in natural environments is critical to translate laboratory findings to real life. PURPOSE: To assess the concurrent validity between force plates and insole sensors for between-limb loading symmetry across activities performed continuously that represent daily tasks including gait at various speeds. METHODS: Twelve participants (6 Female; 74.6 ± 18 kg; 1.73 ± 0.1 m) performed daily loading tasks on a treadmill instrumented with two force plates across two days while wearing loadsol insole sensors. Vertical ground reaction forces (vGRF) were recorded concurrently while participants performed a single limb stance (SLS; 30 second each limb) and a randomized series of tasks including walking, reaching high and low, shifting weight and standing. Body weight was calculated as the mean during the middle 10 seconds of SLS. Loading was analyzed across 12 minutes of repetitive tasks and was considered with and without normalization by body weight quantified by SLS insole data. Between-limb symmetry (BLS) was calculated as the ratio between left and right force impulse (area under the vGRF time curve) for both systems. Intraclass correlation coefficients (ICC3,k) were used to assess validity of SLS body weight, BLS, and normalized BLS (BLSN). Average between system differences and limits of agreement (LOA) were calculated. RESULTS: ICC results were excellent for SLS (0.99, p < 0.001), poor for BLS (0.46 p = .07) and moderate for BLSN (0.66, p = 0.006). Bias were 6.26 N, -0.014, and 0.02, and LOA were (-62, 74.5), (-0.11, 0.08), and (-0.05, 0.09) for SLS, BLS and BLSN, respectively. CONCLUSION: Loadsols are a valid tool for quantifying symmetry between limbs for loading throughout the day. These results validate the use of these insoles in ecological settings. Despite excellent agreement between systems for vGRF during SLS, LOA were up to 15% of body weight. Poor BLS ICC could be attributed to combined error between sensors. Improved BLSN suggests that normalizing insole vGRF by SLS values is necessary for interpretation of daily loading deficits.

  PublisherDOI

• Ageing modulates the effects of scene complexity on visual search and target selection in virtual environments

  Royal Society Open Science · 2025-11-01

  articleOpen accessSenior authorCorresponding

  Processing task-relevant visual information is important for many everyday tasks. Prior work demonstrated that older adults are more susceptible to distraction by salient task-irrelevant stimuli, leading to less efficient visual search. However, these studies often used simple stimuli, and less is known about how ageing influences visual attention in environments more representative of real-world complexity. Here, we test the hypothesis that ageing impacts how the visual complexity of the environment influences visual search. Young and older adults completed a virtual reality-based visual search task in environments with increasing visual complexity. As visual complexity increased, all participants exhibited longer times to complete the task, which resulted from increased time transferring gaze from one correct target to the next and increased delay between when correct targets were fixated and selected. The increase in time to completion can also be attributed to longer times spent re-fixating task-relevant objects and fixating task-irrelevant objects. These changes in visual search and target selection with increasing visual complexity were greater in older adults, and working memory capacity was associated with multiple performance measures in the visual search task. These findings suggest that visual search performance could be integrated into assessments of working memory in dynamic environments.

  PublisherDOI

• Within-session propulsion asymmetry changes have a limited effect on gait asymmetry post-stroke

  Journal of NeuroEngineering and Rehabilitation · 2025-01-22

  articleOpen access

  BACKGROUND: Biomechanical gait impairments, such as reduced paretic propulsion, are common post-stroke. Studies have used biofeedback to increase paretic propulsion and reduce propulsion asymmetry, but it is unclear if these changes impact overall gait asymmetry. There is an implicit assumption that reducing propulsion asymmetry will improve overall gait symmetry, as paretic propulsion has been related to numerous biomechanical impairments. However, no work has investigated the impact of reducing propulsion asymmetry on overall gait asymmetry. We aimed to understand how within-session changes in propulsion asymmetry affect overall gait asymmetry in individuals post-stroke, operationalized as the combined gait asymmetry metric (CGAM). We hypothesized that decreasing propulsion asymmetry would reduce CGAM. METHODS: Participants completed twenty minutes of biofeedback training designed to increase paretic propulsion. We calculated the change in propulsion asymmetry magnitude (Δ|PA|) and the change in CGAM (ΔCGAM) during biofeedback relative to baseline. Then, we fit a robust linear mixed-effects model with ΔCGAM as the outcome and a fixed effect for Δ|PA|. RESULTS: We found a positive association between Δ|PA| and ΔCGAM (β = 2.6, p = 0.002). The average Δ|PA| was -0.09, suggesting that, on average, we would expect a CGAM change of 0.2, which is 0.5% of the average baseline CGAM value. CONCLUSIONS: Reducing propulsive asymmetry using biofeedback is unlikely to produce substantial reductions in overall gait asymmetry, suggesting that biofeedback-based approaches to reduce propulsion asymmetry may need to be combined with other interventions to improve overall gait asymmetry. CLINICAL TRIAL REGISTRATION: NCT04411303.

  PublisherDOI

• Author response for "Aging modulates the effects of scene complexity on visual search and target selection in virtual environments"

  2025-10-03

  peer-reviewSenior author
  PublisherDOI

• Proactive adjustments to cued gait perturbations in people with and without chronic stroke

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-10-16

  preprintOpen accessSenior author

  Balance disturbances exist along a continuum from those that are fully unexpected to those that are predictable based on cues from the environment. When people experience predictable disturbances while walking, they may proactively adjust their gait to minimize losses of balance. However, one's ability to effectively implement these proactive control strategies may be impaired after a stroke due to a combination of motor and cognitive impairments that result from brain lesions. Here, we used explicit audiovisual cues to characterize the proactive and reactive control strategies implemented by people with and without stroke during unexpected versus expected gait perturbations. Following unexpected treadmill accelerations, both groups had smaller margins of stability on the recovery step than during unperturbed walking. When we provided audiovisual cues specifying the impending perturbation step, people without stroke performed less leg and joint work, especially at the ankle, during the cued perturbations and increased their subsequent margins of stability by approximately 3 cm on the recovery step. However, people post-stroke did not make these proactive adjustments. Instead, after any perturbation, they modified their stepping to maintain their center of mass position within their base of support, and this reactive strategy remained unchanged with audiovisual cues. Our findings suggest that people post-stroke rely on a general reactive control strategy rather than proactively modifying push-off work, even when given precise timing information about the impending gait perturbations.

  PublisherOA PDFDOI

• Perceived severity of forward versus backward balance disturbances during walking in young and older adults

  PLoS ONE · 2025-11-20

  articleOpen accessSenior author

  Falls, which often result from trips or slips, pose a major health concern, particularly among older adults. Experiencing falls or near-falls from balance disturbances can shape subsequent gait-related decisions, as individuals may avoid situations they perceive as risky or dangerous. Here, we explore whether perceptions of the severity of a gait disturbance are sensitive to the direction of the resulting loss of balance - forward or backward - and whether these perceptions change with age. Twenty young and twenty older adults walked on a split-belt treadmill while performing a two-alternative forced-choice task, where they indicated their preference between a forward-falling and a backward-falling treadmill perturbation. We varied the perturbation magnitudes using an adaptive staircase algorithm to obtain multiple forward-backward equivalence points, which reflect the points at which a forward and a backward perturbation are perceived as being equally severe. Using a mixed-effects linear model, we estimated the slope of this relationship between forward and backward treadmill perturbations, which quantified the direction and strength of the sensitivity to perturbation type. To assess reliability, we repeated the procedure on a second day. Additionally, we investigated two potential reasons underlying any observed sensitivity - 1) emotional responses measured by state anxiety, and 2) physical responses measured by peak center of mass velocity. We found that both young and older adults perceived backward-falling perturbations to be more severe than forward-falling ones, with no group difference in sensitivity. This sensitivity was moderately reliable across two days of testing, though most participants were less sensitive to perturbation direction on the second day. Neither state anxiety responses nor peak center of mass (CoM) velocity explained the directional sensitivity, though deviations in peak CoM velocity from unperturbed walking were higher during backward-falling than forward-falling perturbations for both age groups. These results suggest that the perceived severity of a gait disturbance is not determined solely by its magnitude, but also by its direction, thereby making direction an important component of gait-related decisions. Integrating a measure of relative perception of types of gait disturbances with the ability to recover from them may provide a more comprehensive assessment of fall risk and inform personalized training interventions tailored to individual perceptions and preferences.

  PublisherDOI

Recent grants

• Toward a Mechanistic Understanding of Optimization Principles Underlying Hemiparetic Gait

  NIH · $1.9M · 2018–2023

• Design and development of a mixed reality system for skilled locomotor training in individuals with Parkinson's disease

  NIH · $206k · 2016–2019

• Design and development of a mixed reality system for skilled locomotor training in individuals with Parkinson's disease

  NIH · $248k · 2016–2018

Frequent coauthors

• Amy J. Bastian

  Johns Hopkins University

  15 shared

• Natalia Sánchez

  15 shared

• Chang Liu

  University of Florida

  12 shared

• Nicolas Schweighofer

  Southern California University for Professional Studies

  12 shared

• Kristan A. Leech

  9 shared

• Eric J. Perreault

  Shirley Ryan AbilityLab

  9 shared

• Sarah A. Kettlety

  Southern California University for Professional Studies

  9 shared

• Yasin Y. Dhaher

  Southwestern Medical Center

  7 shared

Labs

• Locomotor Control LabPI

Education

• Ph.D., Biomedical Engineering

  Northwestern University

  2010

• B.S., Mechanical Engineering

  Florida A&M University-Florida State University College of Engineering

  2004