This review was conducted as a structured narrative review to provide a comprehensive synthesis of current evidence on dual-task walking interventions after stroke. Although not designed as a systematic review, the literature identification, screening, and selection processes were guided by the Scale for the Assessment of Narrative Review Articles (SANRA) to enhance methodological transparency, reduce selection bias, and improve reproducibility. This framework facilitated a structured appraisal of the neurological mechanisms and functional outcomes associated with dual-task gait rehabilitation.

A comprehensive literature search was performed in PubMed/MEDLINE and Scopus to identify studies published between January 2015 and December 2025. The search strategy combined Medical Subject Headings (MeSH) and free-text terms related to stroke, dual-task training, cognitive-motor interference, gait performance, and neuroplasticity using Boolean operators (AND, OR). The search strategy was as follows: (“stroke” OR “cerebrovascular accident” OR “hemiparesis” OR “brain infarction”) AND (“dual-task training” OR “dual-task walking” OR “cognitive-motor interference” OR “concurrent training”) AND (“gait performance” OR “walking speed” OR “stride parameters” OR “mobility” OR “fall risk” OR “neuroplastic change”). To maximize retrieval, the reference lists of eligible articles, relevant systematic reviews, and meta-analyses were manually screened to identify additional studies not captured through the electronic database search.

Studies were selected according to predefined eligibility criteria to ensure the clinical relevance and methodological consistency of the evidence. Eligible studies included randomized controlled trials, non-randomized clinical trials, and quasi-experimental studies investigating dual-task interventions in adult stroke survivors (≥18 years) with ischemic or hemorrhagic stroke across the acute, subacute, or chronic stages. Interventions were required to incorporate simultaneous motor-cognitive or motor-motor dual-task paradigms and to compare these with single-task training, conventional rehabilitation, or no intervention. Studies were also required to report quantitative measures of gait performance, including gait speed, stride length, cadence, dual-task cost, or other neurophysiological outcomes, and to be published as full-text articles in peer-reviewed English-language journals.

Conference abstracts, editorials, case reports, book chapters, dissertations, and animal or pediatric studies were excluded. Studies evaluating isolated cognitive or motor training without concurrent dual-task components, as well as those lacking quantitative gait-related outcome data, were also excluded.

Study selection was performed in two sequential stages. First, two reviewers independently screened the titles and abstracts of all retrieved records according to the predefined eligibility criteria. Potential disagreements were resolved through discussion and, when necessary, consultation with a third-party neurorehabilitation expert. Full-text articles meeting the initial screening criteria were subsequently assessed independently for final inclusion. Data extraction included study characteristics, participant demographics, stroke subtype, intervention protocols, comparator interventions, outcome measures, and principal findings related to gait performance and neurophysiological adaptations. The extracted evidence was synthesized narratively to characterize the effects of dual-task walking interventions and to integrate current knowledge regarding their underlying neurorehabilitation mechanisms. The results were summarized in Table 1.

Post-stroke gait impairment is primarily driven by motor deficits such as muscle weakness, spasticity, and impaired coordination, which are direct consequences of upper motor neuron lesions. Hemiparesis and reduced force generation in the lower extremity diminish propulsion and compromise gait symmetry, leading to reduced walking speed and step length. In addition to motor weakness, sensory and proprioceptive dysfunction contributes significantly to gait abnormalities after stroke. Impaired proprioception, the loss of accurate afferent feedback about limb position and movement, disrupts corrective adjustments during the stance and swing phases, leading to instability and increased dependence on vision for balance control. Observational studies have shown that deficits in sensory reweighting, including proprioceptive processing, are associated with greater walking impairment severity in chronic stroke survivors, indicating that disrupted somatosensory integration undermines postural control during gait [17,18].

Stroke-related proprioceptive loss often co-occurs with diminished tactile and vibration sensation, further degrading the central nervous system’s ability to regulate timing and coordination during locomotion. Reduced somatosensory input from the affected limb limits anticipatory postural adjustments and reduces the precision of foot placement, thereby increasing variability in step length and gait rhythm. Thus, sensory pathway damage—especially involving proprioceptive circuits—has been implicated as a key contributor to maladaptive gait patterns that persist even when motor strength partially recovers [18]. Beyond sensorimotor impairments, cognitive and executive dysfunctions play a substantial role in post-stroke gait disturbance. Patients with greater cognitive impairment often exhibit poorer gait performance and slower rehabilitation progress, suggesting that effective ambulation depends not only on motor capacity but also on higher-order neural processes that guide adaptive locomotion [19]. Deficits in these cognitive domains exacerbate gait variability and increase fall risk, indicating that post-stroke walking is not an automated motor act but a cognitively demanding task. This cognitive-motor interaction underscores the need to incorporate attention and executive function assessments into gait rehabilitation [20].

After stroke, walking in real-world environments rarely occurs in isolation: survivors must ambulate while talking, scanning traffic, or manipulating objects, and such dual-task situations reliably elicit disproportionate decrements in gait speed, spatial-temporal regularity, and cognitive performance compared with age-matched adults without stroke, reflecting heightened cognitive-motor interference (CMI). Experimental work demonstrates that adding a concurrent cognitive task (for example, serial subtraction or auditory Stroop) during overground walking produces higher dual-task costs on gait and cognition in people with stroke than in neurologically healthy controls, indicating a reduced capacity to flexibly share limited attentional resources between locomotor and non-locomotor tasks [21]. From a theoretical perspective, dual-task behaviour after stroke can be understood through capacity-sharing and bottleneck models of attention, in which damaged neural systems require greater voluntary control to maintain basic gait stability, leaving fewer resources for concurrent information processing [22,23].

Dual-task walking requires simultaneous engagement of cognitive and motor systems, particularly executive functions that allocate attention and regulate gait. Primary neuroimaging research in healthy adults showed that dual-task walking increases prefrontal cortex activation, demonstrating that higher-order cognitive control is essential for coordinating concurrent cognitive-motor demands [24,25,26]. In stroke survivors, lesions affecting frontal and parietal regions disrupt these executive control circuits, resulting in impaired coordination between cognitive task execution and gait modulation [27]. This disruption reduces adaptability during ambulation under cognitive load, reflected in slower gait speed and diminished stability during dual-task walking. Individual differences in motor and cognitive impairment further refine this theoretical model. In community-dwelling adults with chronic stroke, higher dual-task costs on gait are correlated with poorer lower-limb motor scores and slower usual gait speed, implying that those with greater baseline impairment operate closer to their attentional capacity limits even in single-task conditions [28]. Deficits in executive function and processing speed similarly magnify gait-related dual-task interference, indicating that the integrity of fronto-subcortical networks shapes the effectiveness with which patients allocate resources across concurrent tasks. The dual-task paradigm captures the interaction between residual motor control, cognitive reserve, and compensatory frontal recruitment after stroke.

Neurophysiological studies suggest that dual-task (motor-cognitive) practice after stroke engages a distributed cortical control network—with particularly strong contributions from the prefrontal cortex (PFC) (attentional/executive control) and premotor-supplementary motor areas (planning and gait regulation)—and that rehabilitation may work partly by reshaping how intensively this network must be recruited to maintain performance. Using fNIRS during dual-task walking, Lu and colleagues showed that adding a concurrent cognitive task increases activation across PFC and motor-planning regions (PMC/SMA), and that greater PMC/SMA activation covaries with gait metrics during interference, consistent with compensatory cortical upregulation to stabilize gait when automaticity is reduced post-stroke [29]. In subacute hemiplegic stroke, Nosaka et al. similarly found higher PFC oxygenation during dual-task stepping/walking than during single tasks, with a notable contralesional predominance across conditions and meaningful associations between PFC activation, frontal-lobe function scores, and dual-task interference—supporting the idea that post-stroke dual-tasking depends on executive “top-down” resources whose availability constrains motor-cognitive trade-offs [30]. Complementing hemodynamic measures, an intervention study leveraging EEG (FFT-based spectral analysis) reported that after a short course of dual-task training, participants showed reductions across multiple EEG bands (delta-beta) alongside clinical improvements, a pattern the authors interpret as training-related modulation of cortical dynamics (potentially reflecting reduced maladaptive slow-wave activity and/or greater neural efficiency). However, the small sample underscores the need for larger mechanistic trials linking spectral shifts directly to dual-task gait outcomes [31]. This evidence indicates that dual-task training after stroke promotes adaptive reorganization of cortical networks involved in executive control and motor planning. Training appears to modulate prefrontal and premotor-supplementary motor area activity, enhancing neural efficiency during cognitively demanding gait and supporting improved motor-cognitive integration.

When evaluating dual-task performance after stroke, the effects of the acute lesion should be interpreted within the context of pre-existing cerebral white matter pathology. Advancing age is frequently accompanied by chronic vascular risk factors, including hypertension, dyslipidemia, metabolic syndrome, and impaired glucose metabolism, which contribute to cerebral small vessel disease and the development of white matter hyperintensities. These microstructural alterations, reflected by reduced fractional anisotropy on diffusion imaging, are associated with slower information processing and impaired executive function [32]. Collectively, they disrupt fronto-subcortical networks that support attentional control and cognitive flexibility, thereby diminishing cognitive reserve before stroke onset [33]. Superimposition of an acute ischemic or hemorrhagic lesion on this pre-existing network vulnerability further compromises neural systems involved in attentional allocation and executive control. As a result, individuals with a greater burden of white matter pathology exhibit higher dual-task costs, poorer gait performance, and reduced capacity to allocate cognitive resources during complex walking tasks [34,35,36].

An important yet underexplored area of neurorehabilitation research is the differential response to dual-task interventions between lacunar and non-lacunar ischemic stroke. These stroke subtypes differ substantially in their underlying vascular pathology, lesion distribution, and functional consequences, which may influence rehabilitation outcomes. Lacunar stroke, resulting from cerebral small vessel disease, primarily affects deep subcortical structures and is often associated with relatively favorable early functional recovery but an increased risk of executive dysfunction and vascular cognitive impairment over time [37].

Non-lacunar stroke, particularly large-artery and cardioembolic stroke, is more frequently associated with territorial or large-vessel infarction that may involve cortical motor, sensory and higher-order associative networks. By contrast, lacunar infarcts occur within deep subcortical structures, including the internal capsule, basal ganglia, thalamus, corona radiata and pons, and may disrupt corticospinal, thalamocortical or fronto-subcortical pathways. These differences in lesion topology could produce distinct patterns of motor and cognitive impairment relevant to dual-task walking. Cortical or corticospinal injury may predominantly constrain motor execution. In contrast, disruption of fronto-subcortical connectivity and accompanying cerebral small vessel disease may increase reliance on attentional and executive resources during gait. However, direct evidence demonstrating systematically greater cognitive-motor interference in lacunar than in non-lacunar stroke remains limited. Comparative studies stratified by stroke mechanism, lesion location, white matter disease burden and baseline impairment are therefore needed to determine whether dual-task rehabilitation should be tailored according to ischemic stroke subtype [38,39,40,41,42].

In stroke rehabilitation, dual tasks are typically grouped into motor-motor (e.g., walking while carrying a tray) and cognitive-motor (e.g., walking while doing serial subtraction or verbal fluency), and the chosen secondary task strongly shapes the type and magnitude of interference observed [10]. Because task demands are not interchangeable, modern frameworks emphasize matching tasks to targeted real-world goals and documenting task features (domain, complexity, prioritization demands) to improve interpretability and comparability across studies [43]. Critically, task complexity and ecological validity matter: dual-task performance can change substantially when walking is tested in more context-rich, challenging environments (e.g., stepping targets or obstacles) rather than in straight, unobstructed corridors, supporting the idea that assessment—and training—should resemble daily-life walking to capture clinically relevant cognitive-motor control [44].

Cognitive-motor dual-task training typically combines walking with concurrent cognitive tasks, such as mental arithmetic, verbal fluency, or working memory challenges, to reduce cognitive-motor interference and improve gait adaptability in complex environments. Randomized trials in people with stroke show that cognitive-motor dual-task gait training yields greater gains in gait speed, balance, and dual-task cost than single-task gait training, particularly when tasks are progressively difficult and task-specific to community ambulation [5,11,45]. These interventions are most effective when therapists explicitly instruct patients how to prioritize tasks (gait first vs cognitive first) and when outcome measures capture both gait and cognitive performance under dual-task conditions [5,46]. Motor-motor dual-task training emphasizes simultaneous performance of two motor activities, for example, obstacle negotiation while carrying objects, manipulating the upper limb, or performing postural transitions. Studies in stroke and older adults demonstrate that adding secondary motor tasks to balance or gait exercises increases postural demand, improves dynamic balance, and enhances mobility. However, it may temporarily increase gait variability during training [47]. Experimental work on obstacle crossing shows that both physical and cognitive secondary tasks can degrade step length and clearance, underscoring the need to grade task difficulty and provide appropriate safety measures (e.g., harness, close supervision) when training obstacle negotiation under dual-task conditions [48].

Dual-task interventions can be implemented on treadmills, overground, or in combined paradigms, and the choice of modality influences both gait biomechanics and dual-task interference. Comparative work in stroke survivors suggests that treadmill walking can promote more automatic and symmetrical gait patterns, while overground training may better approximate real-world demands, especially during obstacle negotiation and turning [49,50]. Experimental dual-task studies indicate that overground walking often shows larger decrements in stride velocity and increased variability under dual-task conditions. In contrast, treadmill walking may maintain gait performance with reduced prefrontal activation, consistent with greater locomotor automaticity [51,52]. Dual-task training may be delivered individually or in group-based formats, each with distinct advantages for feasibility and adherence. Group programs have been shown in older adults to improve mobility, executive function, and dual-task performance, and they may enhance motivation through social interaction [53]. By systematically integrating cognitive-motor and motor-motor challenges across various delivery formats and environments, these interventions can enhance gait adaptability, safety, and real-world functional mobility. Future clinical practice and research should prioritize individualized task progression, standardized outcome measures, and long-term follow-up to optimize translation of dual-task training.

Randomized controlled trials evaluating dual-task training after stroke consistently indicate that embedding concurrent cognitive or motor challenges into gait practice can yield measurable gains in walking performance, particularly under dual-task conditions. In one of the earliest trials, Yang and colleagues randomly assigned individuals with chronic stroke to a dual-task exercise program incorporating concurrent cognitive and motor challenges during walking or to conventional exercise; the dual-task group showed greater improvements in gait velocity, stride length, and walking endurance, suggesting that practicing walking under attentionally demanding conditions enhances locomotor capacity beyond single-task therapy of similar duration. Treadmill-based cognitive dual-task paradigms provide more granular evidence that adding a concurrent cognitive task to gait training can modify spatiotemporal gait parameters and reduce dual-task cost. Kim and colleagues applied progressive treadmill cognitive dual-task training where participants walked while performing increasingly challenging mental tasks and showed significant gains in gait speed, stride length, and symmetry under both single and dual-task conditions relative to conventional treadmill training, indicating that the nervous system can be trained to manage cognitive-motor interference more efficiently [13]. Baek et al extended this approach in a randomized trial of chronic stroke survivors, comparing treadmill gait training delivered simultaneously with a cognitive task versus the same gait and cognitive components administered in separate blocks; only the simultaneous dual-task group achieved superior improvements in dual-task gait speed, stride length, variability, and dual-task cost, together with enhanced fall efficacy, implying that temporal co-occurrence of motor and cognitive demands is a critical ingredient for transfer to complex walking [14].

Other RCTs have emphasised overground and visually mediated dual-task paradigms, reinforcing the generalisability of these benefits across task formats. In a trial of visual cognitive dual-task training, Lee and colleagues asked chronic stroke patients to perform visual attention tasks while standing and walking with virtual reality, demonstrating improvements in balance and gait measures, including Timed Up and Go and Berg Balance Test, beyond those seen with conventional balance training, suggesting that targeting visuospatial attention during locomotion may be particularly relevant for obstacle negotiation in the community [54]. More recent randomized studies have embedded dual-task elements within broader multi-component rehabilitation programmes. Zhang et al evaluated a structured cognitive-motor dual-task training regimen in subacute and chronic stroke, reporting improvements in lower-limb motor scores and functional mobility that exceeded those of conventional therapy, with secondary analyses indicating favourable changes in gait speed and balance under challenging conditions [7]. Liang and colleagues tested multi-task balance training that combined stepping, postural transitions, and concurrent cognitive tasks in ambulatory stroke survivors, finding that this approach improved dual-task walking performance and dynamic balance more than balance exercises alone, which supports the notion that coupling postural control with concurrent information processing is essential for preparing patients for real-world ambulation [15].

Not all RCTs, however, demonstrate clear superiority of dual-task training over well-dosed single-task interventions, underscoring the importance of training intensity and patient selection. In a single-blind randomized clinical trial, Maeneja and colleagues compared a 12-week dual-task cognitive-gait programme with an aerobic exercise protocol (treadmill, cycling, and pedal exerciser) in people with subacute to chronic stroke; while both groups trained three times per week, aerobic exercise produced greater gains in cognitive attention and perceived exertion than dual-task walking, suggesting that when overall dose and cardiovascular load are high, generic aerobic training may match or exceed dual-task protocols for some outcomes [16]. Taken together, the RCT evidence indicates that dual-task training can improve walking performance, especially gait speed, step length, cadence, and dual-task cost after stroke, but its advantage over intensity-matched single-task or aerobic programmes is not uniform and appears to depend on task design, training dose, and the baseline level of cognitive-motor interference in the target population.

Dual-task training holds profound promise for restoring functional independence after a stroke; however, the field currently stands at a crossroads. Despite growing evidence supporting dual-task training for post-stroke rehabilitation, several methodological limitations continue to constrain the interpretation and clinical translation of current findings. Considerable heterogeneity exists across studies in intervention protocols, task characteristics, training intensity, session duration, and outcome measures. Recurrent limitations include small sample sizes, inadequate blinding, and insufficiently powered subgroup analyses, which reduce the certainty and generalizability of the available evidence. In addition, follow-up periods are generally short, limiting understanding of the long-term sustainability of improvements in gait and cognition, as well as their impact on community mobility and fall prevention.

Future studies should prioritize greater methodological standardization to improve comparability across trials and facilitate high-quality evidence synthesis. The development of standardized intervention protocols and core outcome sets would strengthen future meta-analyses and support the translation of research into clinical practice. Large, multicenter randomized controlled trials with adequate statistical power are also needed to evaluate treatment effects across clinically relevant subgroups, including stroke stage, lesion characteristics, and baseline functional impairment. Advances in neuroimaging and computational approaches may further enable a more individualized rehabilitation strategy. Integration of structural and functional neuroimaging modalities, including diffusion MRI, mobile functional near-infrared spectroscopy (fNIRS), and high-density electroencephalography (EEG), with machine-learning approaches may improve patient stratification according to lesion topology, white matter integrity, and network dysfunction. Such approaches have the potential to identify patients who are most likely to benefit from specific dual-task interventions. Technology-assisted rehabilitation, including virtual reality and exergaming, represents another promising direction by enabling ecologically relevant and progressively challenging cognitive-motor training. Integrating these technologies with personalized neuroimaging-based prediction models may facilitate precision rehabilitation, tailoring interventions to individual neurological profiles while improving long-term functional recovery and community participation.