Exercise May Sharpen Teen Executive Function, But Evidence Remains Weak

A large synthesis of randomized controlled trials suggests that sustained exercise programs lasting at least four weeks are associated with better performance on executive function tasks in adolescents aged 10 to 19. The pooled analysis, published in Frontiers in Psychiatry, combined data from 22 independent trials involving 1,599 unique participants and found an overall standardized effect of Hedges’ g = 0.55 (95% confidence interval 0.33 to 0.77; p < 0.001) favoring exercise over control conditions that involved no additional structured exercise. Yet the authors themselves urge caution: the evidence base is heterogeneous, mostly of low certainty, and the pooled figure masks wide variation across trials.

Executive function is an umbrella term for three higher-order cognitive processes: inhibitory control, which allows a person to resist impulses and distraction; working memory, which holds and updates information in mind; and cognitive flexibility, which permits shifting between rules or tasks. These abilities depend on frontostriatal and frontoparietal brain networks that continue reorganizing through adolescence, a developmental window when trajectories of self-regulation, learning behavior, and mental health begin to diverge. Prior observational work has linked better childhood self-control to health, economic, and social outcomes in adulthood, and most adolescents worldwide fail to meet recommended physical activity levels, making the question of whether structured exercise can boost cognition both scientifically and practically important.

The research team, led by Dongying Li and Lulu Niu of Yunnan Minzu University in Kunming, China, searched nine databases through 29 September 2026 and applied strict eligibility criteria. Only individually randomized, parallel-group trials of English-language, peer-reviewed reports were included; crossover designs, nonrandomized studies, acute single-session exercise, interventions shorter than four weeks, and control groups given additional structured exercise were excluded. The final set of 22 trials, published between 2015 and 2026, spanned nine countries, with eleven conducted in China, four in South Korea, and single trials in Tunisia, Spain, India, Brazil, Belgium, Türkiye, and Denmark. Twelve trials enrolled generally healthy adolescents, five involved adolescents with overweight or obesity, three involved adolescents with ADHD, and single trials involved adolescents with subthreshold depression or elevated suicide risk.

The interventions themselves varied enormously, lasting from 8 to 78 weeks and ranging from high-intensity interval training and resistance exercise to basketball, badminton, rope skipping, taekwondo, swimming, yoga, football juggling, small-sided games, coordination training, and classroom cycling. This diversity is one reason the authors resist any claim about an optimal exercise prescription. Control conditions were equally varied, spanning waitlist designs, non-exercise health education, usual school activities, and ordinary physical education classes; in some school-based trials both groups continued regular physical education while only the intervention group received additional structured exercise.

Methodologically, the review tackled a persistent problem in exercise-cognition research: multiple dependent outcomes within single trials. A single study often reports several tasks, conditions, response-time measures, and accuracy measures. Selecting one outcome arbitrarily discards information, while treating every outcome as independent overstates precision. The authors extracted 86 effect sizes from 27 exercise-control comparisons and synthesized them with three-level random-effects models that nested effect sizes within trials, distinguishing within-trial from between-trial heterogeneity. Because the original trials rarely reported the correlations among their outcomes, the team constructed an approximate block-diagonal covariance matrix assuming a within-trial outcome correlation of 0.50, and used cluster-robust CR2 inference with the independent trial as the cluster to reduce reliance on a perfectly specified working covariance model.

The domain-specific results were all positive. Inhibitory control, the most heavily represented domain with 47 of the 86 effect sizes from 17 trials, yielded g = 0.56 (95% CI 0.28 to 0.84). Working memory, based on 25 effect sizes from 10 trials, yielded g = 0.51 (95% CI 0.18 to 0.84). Cognitive flexibility, with 14 effect sizes from nine trials, yielded g = 0.43 (95% CI 0.30 to 0.57). The authors stress that the overall estimate is an outcome-weighted average over the observed task mix, not an equally weighted measure of three distinct abilities, and that the domain estimates should not be read as a ranking of which executive function component is most responsive to exercise.

Heterogeneity was substantial. Total I-squared reached 84.92% for overall executive function, 84.93% for inhibitory control, and 80.67% for working memory, with the between-trial variance component accounting for 64.89% of total variability in the overall model. Exploratory subgroup analyses found a pooled estimate of g = 0.46 in healthy adolescents, g = 0.72 in adolescents with overweight or obesity, and g = 1.14 in the ADHD subgroup, but the ADHD estimate was extremely imprecise (95% CI −1.41 to 3.70) because it rested on only three trials, and the three-group difference was not statistically supported (p = 0.530). Four univariable meta-regressions examining mean age, intervention duration, weekly frequency, and session duration found no moderator that remained significant after adjustment for multiple testing.

Sensitivity analyses broadly held the finding in place. Leaving out one trial at a time produced overall estimates between 0.49 and 0.58. Excluding results rated at high risk of bias by the Cochrane RoB 2 tool left the overall estimate unchanged at g = 0.55. Excluding 11 scores with ambiguous domain attribution—such as Posner cueing scores that also reflect spatial attention, Digit Span totals that may emphasize storage over manipulation, and Trail Making Test B times that include visual search and motor speed—left a positive overall estimate of g = 0.53. Varying the assumed within-trial outcome correlation from 0.30 to 0.70 barely moved the result, but varying the assumed pre-post correlation across the same range shifted the overall estimate from g = 0.47 to g = 0.69, showing that the precise magnitude depends materially on an assumption the original trials did not report.

The certainty of the evidence tempers enthusiasm. Using the GRADE framework, the authors rated certainty as low for overall executive function, inhibitory control, and working memory, owing to serious risk of bias and serious inconsistency, and moderate for cognitive flexibility, downgraded only for risk of bias. Across 45 trial-arm-domain results assessed with RoB 2, only eight (17.8%) were rated low risk, 33 (73.3%) raised some concerns, and four (8.9%) were rated high risk, with concerns concentrated in the randomization process and selection of the reported result. Funnel-plot tests found no statistical asymmetry, but with ten or fewer trials contributing to two of the domains, the authors note these nonsignificant tests cannot exclude publication bias or small-study effects.

The authors conclude that sustained exercise may improve performance on executive function tasks in adolescents under the studied conditions, but that heterogeneous interventions and controls, overlapping task constructs, and mostly low-certainty evidence limit the generality and real-world interpretation of the pooled effects. Improvement on laboratory tasks, they caution, does not by itself demonstrate better academic achievement, classroom behavior, self-regulation, or daily functioning. For future work, they call for individually randomized trials that pre-register one clearly defined primary executive function score per domain, report the correlations needed for multi-outcome synthesis, distinguish added exercise from background physical education, and include direct academic and daily-function outcomes alongside task measures. Until such trials arrive, the message for adolescents, parents, and schools is a measured one: additional structured exercise is plausibly beneficial for measured cognition, but no specific modality, dose, or schedule can yet be recommended with confidence.

Subject of Research: The effects of sustained exercise interventions on executive function in adolescents, based on a three-level meta-analysis of randomized controlled trials

Article Title: Effects of exercise interventions on executive function in adolescents: a systematic review and three-level meta-analysis of randomized controlled trials

Article References: Li, D., Li, S., Yao, M., Ye, S., Geng, K., Zhou, L., & Niu, L. (2026). Effects of exercise interventions on executive function in adolescents: a systematic review and three-level meta-analysis of randomized controlled trials. Frontiers in Psychiatry, 17, Article 1968917. https://doi.org/10.3389/fpsyt.2026.1968917

Image Credits: AI Generated

DOI: 10.3389/fpsyt.2026.1968917

Keywords: adolescents, executive function, exercise intervention, meta-analysis, inhibitory control, working memory, cognitive flexibility, randomized controlled trials, GRADE certainty, heterogeneity, physical activity, cognition