Brain Training Improves Cognition and Boosts Precuneus Neural Activity in Neurological Disorders
This meta-analysis makes a stronger case for cognitive training as a supportive strategy than for any single “brain region fix.” The most clinically relevant finding is not simply that task-related precuneus activity increased, but that this increase tracked with better performance across studies. Even so, that relationship should be read as a useful neural correlate rather than proof that boosting precuneus activation directly causes cognitive improvement. Neuroimaging markers can help explain how training may work, but they do not automatically translate into meaningful day-to-day gains such as better medication management, conversation, planning or independence.
The practical signal here is modest but encouraging: structured training appears capable of producing moderate improvements across very mixed neurological conditions, despite major differences in diagnosis, disease severity and training design. That breadth is promising, but it also limits precision. “Neurological disorders” is not one condition, and readers should be cautious about assuming the same effect size applies equally to dementia, stroke, traumatic brain injury or multiple sclerosis. The fact that improvements were more closely linked to trained tasks than transfer tasks also matters. It suggests people may get better at the kinds of exercises they repeatedly practice, while broader real-world benefits may be less reliable.
For patients, families and clinicians, the takeaway is to view cognitive training as a potentially useful part of rehabilitation, not a standalone solution. Questions that still matter include whether benefits persist over time, which patients respond best, and whether improvements extend beyond testing situations into everyday functioning. Those are the outcomes that ultimately determine real wellbeing impact.
A systematic meta-analysis synthesizing twenty-one task-based neuroimaging investigations has revealed that structured cognitive training yields significant, moderate improvements in cognitive task performance among individuals diagnosed with neurological disorders, while concurrently producing a convergent functional signature of heightened task-evoked brain activation centered within the bilateral precuneus and the left precentral gyrus. Published in Behavioral and Brain Functions, the comprehensive investigation demonstrated that behavioral cognitive performance gains across studies were specifically tracked by heightened functional recruitment in the bilateral precuneus, identifying this regional activation change as a candidate neural marker of intervention-related functional plasticity. In contrast, although the left precentral gyrus similarly demonstrated statistically significant post-training activation increases relative to control cohorts, regional activation changes in this premotor area did not exhibit a statistically significant association with measured behavioral improvements across the synthesized trials.
Neurological disorders encompass a clinically diverse spectrum of conditions, including neurodegenerative diseases, cerebrovascular pathology, traumatic brain injury, and demyelinating syndromes that frequently converge on widespread disruptions of large-scale distributed cognitive networks. These pathological alterations are commonly accompanied by progressive, debilitating impairments in attention, executive function, and memory that substantially undermine independent daily functioning and quality of life, and may ultimately prove fatal. Dementia represents the most prevalent clinical manifestation of this cognitive decline, currently affecting an estimated 57.4 million individuals worldwide, with global prevalence projected to escalate to 153 million individuals by the year 2050. Beyond their devastating personal consequences, these conditions impose immense economic and societal strains, illustrated by unpaid dementia-related caregiving in the United States alone carrying an annual valuation exceeding 339.5 billion dollars, while available pharmacological therapies remain modest in efficacy and emerging anti-amyloid agents deliver only partial therapeutic benefits.
Because approved pharmaceutical interventions provide limited symptomatic relief across clinical populations, researchers have increasingly evaluated non-pharmacological behavioral approaches designed to preserve cognitive function and remediate network disruptions. Cognitive training, also termed cognitive remediation, engages individuals in repeated, structured cognitive exercises to bolster cognitive capabilities by promoting structural and functional neural plasticity. Prior empirical investigations examining task-related neural changes have yielded divergent conclusions across different clinical populations and cohorts, with some studies documenting post-intervention activation increases in frontal cortices while others reported enhancements in the left anterior temporal lobe. Theoretical frameworks conceptualize training-induced activation shifts through two primary mechanisms: enhanced recruitment representing increased neural engagement, and neural efficiency characterized by decreased task-related activation alongside superior task performance. In neurological populations characterized by constrained physiological reserves, the Compensation-Related Utilization of Neural Circuits Hypothesis suggests that training predominantly prompts compensatory regional hyperactivation to offset baseline network inefficiency.
To rigorously resolve these empirical discrepancies and evaluate whether cognitive remediation establishes a convergent neural activation signature across heterogeneous neurological conditions, the investigative team prospectively registered their systematic review protocol in the PROSPERO database under registration number CRD42024599205. Adhering to the methodological standards established by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement and authoritative neuroimaging guidelines, the authors conducted comprehensive literature searches across five primary electronic databases: PubMed, Web of Science, APA PsycINFO, MEDLINE, and Embase. The literature search was systematically updated through November 12, 2024, deploying specialized keyword strings focused on neuroimaging protocols and structured cognitive training. The primary literature search initially captured 4,659 records, which was reduced to 2,543 unique citations following the removal of 2,116 duplicate entries through EndNote software, while reference lists of eligible papers were manually examined to identify additional studies.
Screening procedures were executed independently by two investigators who reviewed all titles and abstracts against predetermined inclusion and exclusion criteria, subsequently evaluating 131 full-text publications for definitive inclusion, with disputes adjudicated through consensus or senior author consultation. Eligibility criteria mandated that included investigations: reported exact stereotactic three-dimensional coordinates of task-related functional activation changes elicited by cognitive training; utilized whole-brain imaging analyses; implemented within-subject pre-to-post or between-subject comparative designs containing active control comparators or passive waitlist controls; and examined human participants diagnosed with neurological conditions such as neurodegenerative diseases, stroke, brain trauma, or demyelinating illnesses. Ultimately, twenty-one peer-reviewed studies fulfilled all methodological requirements and were integrated into the primary neuroimaging meta-analysis, while a subset of thirteen studies providing adequate quantitative behavioral data was incorporated into a dedicated multivariate behavioral meta-analysis.
Standardized data extraction protocols captured participant demographic characteristics, clinical parameters, cognitive training dosages, control condition designs, and peak stereotactic coordinates alongside respective statistical test values, including t-statistics, z-scores, and p-values. Behavioral performance synthesis was executed in RStudio using the metafor and clubSandwich analytical libraries under a three-level hierarchical random-effects model that nested individual outcome metrics within participant samples and nested participant samples within distinct studies. Effect sizes were computed as standardized mean differences using Hedges’ g to correct for small-sample estimation bias, derived from between-group comparisons of pre-to-post performance change scores. Within-study correlation coefficients among cognitive measures were set at r = 0.5 following Cochrane Handbook recommendations, with model robustness subsequently verified across sensitivity analyses assuming alternative correlation values of 0.3, 0.7, and 1.0, while metrics such as error rates were reverse-coded so that positive effect sizes uniformly denoted performance improvement.
Coordinate-based meta-analysis of task-evoked neuroimaging alterations was conducted using Seed-based d Mapping software, version 5.15, standardizing disparate coordinates and statistical metrics into uniform stereotactic anatomical space. Image preprocessing implemented an isotropic Gaussian filter set at a 20-millimeter full-width at half-maximum and applied 500 permutations, combining multiple statistical contrasts within individual studies into unified composite maps using dedicated SDM functions. Statistical significance for brain activation clusters was established at an uncorrected threshold of p < .0001, an analytical cutoff empirically demonstrated to provide stringency comparable to conventional corrected thresholds in SDM investigations, with cluster significance requiring a peak SDM-Z score of at least 1.0 and a contiguous spatial extent of ten or more voxels. The reliability of identified neuroimaging clusters was formally evaluated through jackknife sensitivity analyses alongside Egger’s linear regression tests to assess potential small-study publication bias.
Across the twenty-one included neuroimaging investigations, the analyzed participant cohort comprised 502 individuals presenting with diverse neurological disorders, exhibiting a mean age of 56.95 years and a female representation of 52.72 percent. On average, structured cognitive training protocols spanned 15.08 training sessions totaling 834.31 minutes of cumulative intervention time, delivered at a mean scheduling frequency of 2.82 sessions per week with an average single session duration of 53.18 minutes and a high overall mean adherence rate of 92.3 percent. The multivariate behavioral meta-analysis of thirteen eligible studies confirmed that structured cognitive training generated a statistically significant, moderate enhancement in cognitive task performance relative to control interventions, yielding a pooled Hedges’ g of 0.451 with a 95 percent confidence interval spanning from 0.207 to 0.696, a t-statistic of 3.745, a significance level of p < .001, and moderate heterogeneity quantified at I² = 35.955 percent.
The coordinate-based SDM neuroimaging synthesis identified two spatially distinct, statistically significant clusters exhibiting elevated task-related functional activation following cognitive training compared against control conditions. The primary cluster emerged within the bilateral precuneus, demonstrating robust post-training activation enhancement characterized by a peak SDM-Z score of 2.932 across an anatomical volume of 1,112 contiguous voxels at p < .0001, with minimal between-study heterogeneity measured at I² = 7.445 percent. The secondary significant cluster appeared in the left precentral gyrus, presenting a corresponding elevation in task-evoked activation with a peak SDM-Z score of 2.753 encompassing 707 contiguous voxels at p < .0001 and low heterogeneity at I² = 1.821 percent. Meta-analytic regression linking regional neuroimaging changes to behavioral outcomes demonstrated that cognitive task gains were significantly correlated with activation increases in the bilateral precuneus (beta = 0.331, p = .021), whereas no statistically significant brain-behavior association was observed for activation changes within the left precentral gyrus.
Moderator analyses exploring study design features demonstrated that categorical training variables exerted selective regulatory influences over regional functional activation profiles. Fixed-difficulty training protocols elicited greater left precentral gyrus activation than adaptive-difficulty training paradigms (SDM-Z = 2.084; 104 voxels), whereas passive control conditions provoked stronger left precentral gyrus activation than active control comparators (SDM-Z = 1.608; 159 voxels). Furthermore, task modality significantly modulated regional engagement: transfer tasks assessing generalized cognitive abilities recruited the left precentral gyrus more extensively than trained tasks (SDM-Z = 1.604; 63 voxels), whereas trained tasks preferentially stimulated functional activation within the bilateral precuneus (SDM-Z = 2.517; 804 voxels). In contrast, comparative analyses revealed no significant activation moderation by training domain, training paradigm, delivery format, study design framework, or continuous participant characteristics including mean chronological age, gender proportion, single session duration, total program duration, weekly training frequency, and overall intervention adherence.
In interpreting these findings, the study authors highlighted the distinct functional roles played by the identified anatomical regions in neurocognitive adaptation. The precuneus serves as an essential associative hub within the frontoparietal control network and acts as a connector linking parietal and frontal cortices to facilitate large-scale information integration between the default mode network and task-positive systems. In line with the Scaffolding Theory of Aging and Cognition, upregulation of bilateral precuneus activity reflects compensatory recruitment of alternative neural circuits to maintain cognitive output in the presence of compromised neural efficiency. Conversely, increased activation in the left precentral gyrus likely reflects non-specific motor preparation and response demands associated with manual button-press tasks rather than cognitive remediation, explaining its dissociation from behavioral gains. Diminished precuneus recruitment during transfer tasks indicates that generalized tasks fail to sufficiently engage this gain-relevant associative hub, providing a neural mechanism for why cognitive improvements often remain restricted to directly trained tasks.
The authors addressed several methodological considerations and study limitations that contextualize the synthesized outcomes, noting that the modest volume of eligible neuroimaging trials precluded formal family-wise multiple-comparisons correction, although applying a stringent uncorrected threshold of p < .0001 effectively curtailed potential false-positive inflation. Additionally, several continuous moderators required binary categorization to avoid sparse subgroups during comparative analyses, and inconsistent reporting of clinical severity and imaging variables across original trials constrained nuanced confounder control. Jackknife sensitivity analyses demonstrated high replicability, with core activation findings persisting across all leave-one-out iterations, and Egger’s regression tests confirmed an absence of significant funnel plot asymmetry for both the bilateral precuneus and the left precentral gyrus. Ultimately, this rigorous synthesis establishes that cognitive training elicits meaningful behavioral improvements in neurological disorders, driven by a convergent activation signature wherein bilateral precuneus upregulation serves as a primary candidate neural marker of intervention-related neuroplasticity.
Subject of Research: Meta-analysis of cognitive training effects on cognitive performance and brain activation in neurological disorders
Article Title: Cognitive training in neurological disorders: meta-analytic evidence for behavioral gains and neural correlates
Article References: Li, G., Liu, Y., Liu, C., & Zhang, Y. (2026). Cognitive training in neurological disorders: meta-analytic evidence for behavioral gains and neural correlates. Behavioral and Brain Functions, 22(1), Article 25. https://doi.org/10.1186/s12993-026-00323-3
Image Credits: AI Generated
DOI: 10.1186/s12993-026-00323-3
Keywords: Cognitive Training, Neurological Disorders, Neuroimaging Meta-Analysis, Seed-Based d Mapping, Bilateral Precuneus, Left Precentral Gyrus, Multivariate Meta-Analysis, Brain-Behavior Associations, Neural Plasticity, Transfer Tasks, Adaptive Training, Frontoparietal Network, Default Mode Network, Scaffolding Theory of Aging and Cognition, CRUNCH Hypothesis, Cognitive Remediation, Task-Evoked Brain Activation
Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.
