Active Predictive Coding (APC) posits the neocortex as a predictive engine integrating sensory inputs and motor actions to anticipate environmental states and action consequences (14, 15). Prediction errors—discrepancies between expected and actual sensory inputs—are iteratively minimized across hierarchical cortical levels, driving adaptive behavior (14). Each cortical area contains both a state-prediction network estimating hidden sensory states and an action-prediction network generating actions based on these estimates. APC reframes perception and action as bidirectionally coupled and central to cognition. Visual search tasks illustrate how prior knowledge guides eye movements to optimize sensory acquisition and reduce uncertainty (1, 16).

Neurophysiological studies in rodents support APC: optogenetic and calcium imaging reveal that up to 40% of neurons in primary visual cortex (V1) encode predictive signals about voluntary movements, with task performance dropping by 60% upon optogenetic inhibition, highlighting early sensory-motor integration (17). While these findings may have limited direct generalizability to humans due to species differences, complementary computational and empirical studies show that neurons in cortical layers 2/3 integrate motor inputs with sensory feedback with 92% prediction accuracy, while layer 5 neurons refine predictions via feedback, reducing errors by 70% during sensory-motor tasks (1). Additionally, a 15-ms reduction in neuronal response latency during sensory-motor integration emphasizes the efficiency of feedback loops (1).

Oscillatory alpha-band activity has been implicated in APC, with traveling waves reflecting the recursive exchange of predictions and errors essential for temporal dynamics in adaptive sensory-motor integration (10).

The frontoparietal network is crucial for cognitive-motor tasks, enabling dynamic, context-sensitive modulation of motor outputs according to cognitive demands (18). DLPFC orchestrates these processes especially under multitasking and cognitive-motor interference conditions. Functional near-infrared spectroscopy (fNIRS) studies show increased DLPFC activity during dual-task scenarios, accompanied by enhanced connectivity to the inferior parietal cortex, reflecting heightened sensory-motor integration under cognitive load (18). Granger causality analyses reveal a top-down hierarchical flow from DLPFC to parietal regions, underscoring the network’s active role in coordinating predictive and executive functions necessary for adaptive behavior (18).

Research indicates that the frontoparietal network flexibly adapts to changing task demands by dynamically recoding task-relevant information on a trial-by-trial basis. This flexible coding, less stable during task switches, predicts behavioral performance and highlights the network’s central role in cognitive control and behavioral flexibility (5, 6). Predictive coding theory further attributes frontoparietal regions with generating higher-level, contextual predictions beyond those in sensory cortices, especially during complex tasks like working memory and language comprehension (7, 8). The network’s dynamic connectivity supports both top-down predictions and bottom-up integration and can be externally modulated via brain stimulation to enhance cognitive performance under demanding conditions (8, 9). Synchronization and oscillatory alpha-band activity are integral to these predictive coding mechanisms, facilitating recursive exchanges of predictions and prediction errors (10). Individual variability in frontoparietal network organization and synchronization significantly informs differences in cognition and susceptibility to psychopathology, with implications for personalized rehabilitation (11–13).

Parkinson’s disease (PD) serves as a valuable clinical model for investigating disruptions in cognitive-motor integration (19). The dopaminergic degeneration that is characteristic of PD has been shown to affect not only the basal ganglia but also the cortical circuits involved in the planning and execution of motor actions (20). As demonstrated by Wang et al. (21), patients diagnosed with PD exhibit a 25% increase in functional connectivity between the prefrontal cortex (PFC) and other cortical areas during dual-task walking, compared to healthy controls (p < 0.01). This heightened reliance on the PFC suggests that cognitive resources are increasingly recruited to compensate for motor deficits, indicating a shift from automatic motor control to more conscious, effortful processing (21). Such a shift has been shown to impose cognitive strain, particularly during complex tasks, as evidenced by the inverse correlation between PFC connectivity and walking speed (r = −0.52, p < 0.01).

PD impairs both motor planning and execution, worsening gait function. For instance, Wang et al. (21) reported a 30% reduction in stride length and a 20% decrease in walking speed among PD patients compared to healthy controls (p < 0.001). Furthermore, the gain of the soleus H-reflex—a critical spinal reflex for locomotion—is reduced under dual-task conditions. In a separate study, Al-Yahya et al. (22) found that this reduction was 25% in healthy older adults (p < 0.05), but significantly more pronounced in PD patients, who exhibited a 40% decrease compared to their single-task baseline (p < 0.01). The present findings highlight the complex relationship between cortical compensation and peripheral motor deficits, emphasizing the importance of interventions that target both levels of control (22).

Beyond core motor symptoms, a range of non-motor manifestations also characterizes PD, including executive dysfunction, attentional deficits, and impaired working memory—all of which affect movement control and adaptability. Set-shifting difficulties and cognitive inflexibility, for example, are commonly associated with freezing of gait, particularly in novel or cognitively demanding environments (23).

Consistent with these observations, dual-task paradigms have consistently demonstrated that PD patients experience greater postural instability and reduced gait performance when simultaneously engaged in cognitive tasks, suggesting that frontostriatal dysfunction contributes to both motor and cognitive deficits (24).

These converging findings support the integration of cognitive training into rehabilitation strategies alongside motor exercises. Techniques such as external cueing, task-switching protocols, and dual-task training may enhance outcomes by improving cognitive-motor coordination. Recent evidence highlights the role of cognitive reserve (CR)—a construct shaped by education, occupational complexity, and intellectual engagement—as a key modulator of symptom severity and rehabilitation outcomes in PD (25–27). Higher CR is associated with preserved executive function, reduced motor impairment, and greater responsiveness to cognitively demanding interventions such as virtual reality and tele-rehabilitation (25, 28, 29). Conversely, individuals with lower CR may benefit more from conventional approaches, underscoring the need to tailor rehabilitation to individual cognitive profiles (30). Given its multidimensional nature and buffering capacity against both motor and non-motor symptoms (27, 31, 32), CR should be considered a critical factor in the design of personalized, cognitively informed therapeutic strategies (26, 29, 33).

Cognitive-motor integration is often impaired in stroke survivors due to focal damage in cortical or subcortical structures. A hallmark of post-stroke recovery is functional reorganization, whereby preserved cortical regions assume compensatory roles. A recent meta-analysis by Wang et al. (34), summarize that stroke patients exhibited a 40% increase in PFC activation during dual-task walking compared to healthy controls (p < 0.01), measured using fNIRS. This shift marks a compensatory adaptation in brain function, reflected in measurable changes in gait dynamics and motor efficiency.

While plasticity enables reorganization, compensatory pathways are often inefficient, reducing motor performance. The same meta-analysis reported a significant negative correlation between PFC activation and walking speed (r = −0.47, p < 0.05), highlighting the trade-off between increased cognitive load and reduced motor efficiency. Moreover, stroke survivors with severe motor-related cortical damage exhibited a 30% reduction in stride length compared to healthy individuals (p < 0.001), further emphasizing the impact of elevated cognitive demands on motor task performance. These motor outcomes not only highlight the cost of compensatory cognitive control but also provide key targets for designing more effective rehabilitation strategies. Neural plasticity allows for the formation of compensatory circuits, but these pathways typically demand increased cognitive effort to support motor functions. This increased reliance on the PFC during motor tasks is indicative of an adaptive yet inefficient strategy, underscoring the necessity for interventions that optimize the balance between cognitive load and motor performance in stroke rehabilitation (34, 35).

Beyond motor impairments, stroke frequently results in cognitive deficits, particularly in domains such as attention, executive function, and memory—all of which play a pivotal role in motor recovery and rehabilitation outcomes.

Executive and attentional deficits disrupt motor learning and adaptation. For instance, stroke survivors with limited working memory capacity may struggle to consolidate and retain motor strategies over time (36). Functional neuroimaging studies have revealed that successful motor relearning is associated with increased activation in both the prefrontal and parietal cortices, suggesting that these regions may support compensatory cognitive processes.

Dual-task paradigms—commonly used to assess cognitive-motor interference—frequently demonstrate that stroke patients exhibit reduced gait stability and walking speed when performing a concurrent cognitive task. This supports the notion that compromised cognitive resources directly affect motor execution (37).

Therefore, tailoring rehabilitation interventions to individual cognitive profiles—particularly by incorporating training for attention and executive function—may enhance motor recovery. Recent biomarker-based models support this integrative approach, aiming to align neuroplastic mechanisms with personalized therapeutic strategies (38).

In aging populations, the neural basis of cognitive-motor integration undergoes significant changes, including a 25% reduction in connectivity within the frontoparietal net-work compared to younger adults (p < 0.001) and a 30% increase in activity in regions associated with compensatory processes, such as the PFC (18). These alterations are particularly pronounced in neurodegenerative disorders like Alzheimer’s disease, where the capacity to perform dual tasks is significantly impaired due to compromised cortical control. Miura et al. (18) demonstrated that individuals with early-stage Alzheimer’s dis-ease exhibit a 40% reduction in primary motor cortex activity during dual-task walking (p < 0.01), accompanied by a 35% increase in PFC activation (p < 0.01), reflecting an over-reliance on compensatory mechanisms. Cognitive decline limits efficient resource allocation for dual tasks. Decreased basal ganglia and motor cortex activity further burdens compensatory areas such as the PFC, leading to slower walking speeds (a 25% reduction compared to controls, p < 0.001) and diminished task-switching efficiency. These findings underscore the maladaptive nature of compensatory neural activity in impaired motor-cognitive integration. Interventions that target enhancing neural plasticity and connectivity show promise in mitigating these deficits. Miura et al. (18) highlighted the potential of VR-based rehabilitation, which improved dual-task performance by 20% (p < 0.05) after an eight-week intervention in participants with mild cognitive impairment. The findings of this study lend support to the hypothesis that combining such strategies with physical exercise and cognitive training targeting multitasking capabilities may further facilitate the re-establishment of motor-cognitive integration in aging and neurodegenerative populations.

To synthesize the discussed evidence, Table 1 delineates the distinct neurophysiological and behavioral profiles associated with each condition, emphasizing how variations in prefrontal activation, gait dynamics, and frontoparietal connectivity reflect compensatory or maladaptive responses with direct clinical relevance.

The present narrative review is subject to certain inherent limitations. Notably, the absence of systematic inclusion criteria and the lack of formal quality assessment may introduce selection bias in the sources consulted (53). While the narrative approach enables a flexible synthesis of emerging evidence across diverse domains, it inherently limits the replicability and generalizability of findings compared to systematic reviews or meta-analyses. Moreover, a substantial portion of the reviewed literature is derived from studies employing heterogeneous methodologies and involving diverse clinical populations. This variability may affect the consistency and robustness of the conclusions drawn, underscoring the need for future research employing standardized protocols and well-defined outcome measures.