Best Exercise for Kids’ Working Memory? Evidence Says It Depends

What kind of exercise helps working memory fastest?

The honest answer is more specific than "exercise helps." A 2025 network meta-analysis by Guo et al. set out to rank long-term exercise types for working memory in healthy children and adolescents, not just compare exercise with doing nothing. Its reported framing is striking: dance appears to rank highest for working-memory accuracy, while aerobic exercise appears to rank highest for reaction time. If that holds up, the practical message is simple. The best training for getting answers right may differ from the best training for getting them out faster.

There is one problem, and it is not a small one. The available evidence here includes the study's design and rationale, but not the results table needed to verify the actual SUCRA rankings, effect sizes, confidence intervals, or pairwise significance. So the responsible version of this article is not "dance wins, case closed." It is "a well-designed comparison points that way, but the quantitative claim still needs confirmation from the full results." Science is less fun when it refuses to wear a cape, but it is more useful.

Why this study matters

Guo et al. focused on neurotypical youth, which matters because much of the exercise-and-cognition literature in children comes from ADHD research. That literature is valuable, but it answers a different question in a different population. The Guo et al. network meta-analysis included healthy children ages 6 to 12 and adolescents ages 13 to 18, and compared six long-term intervention categories: aerobic exercise, dance, mixed exercise, high-intensity interval training, sports games, and passive control.

A network meta-analysis is the right tool for this job. It can compare interventions even when most trials never tested them head to head. Instead of waiting forever for "dance versus jogging versus team sports versus HIIT" in one giant trial, the method combines direct and indirect evidence across randomized studies. That does not make the result infallible. It does make the question answerable.

The paper also treated working-memory accuracy and working-memory reaction time as separate outcomes. That is a smart choice. "Working memory" sounds like one thing, but accuracy and speed can move differently. A child may recall and manipulate information more precisely without getting faster, or get faster without becoming more accurate. In school, both matter. Reading comprehension and multistep math lean hard on accuracy. Fast updating matters more for tasks that reward quick shifts of attention.

The reported ranking, and why it is plausible

According to the study framing available here, dance ranks first for accuracy and aerobic exercise ranks first for reaction time. That split is biologically plausible. Dance asks the brain to coordinate timing, sequencing, spatial prediction, memory for movement patterns, and rapid error correction. Aerobic exercise, by contrast, has a stronger record for improving general arousal, processing efficiency, and executive function in broader clinical literature.

Mechanistic support for dance exists, but it is indirect. In a 2025 fNIRS study by Wang et al., young female street dancers showed greater prefrontal cortical activation during cognitive tasks than non-dancers. That fits the idea that dance may engage the prefrontal systems that support precise mental updating. It does not prove that dance improves working-memory accuracy in healthy schoolchildren. The study was observational, not randomized, and it used a cognitive-reserve framing rather than a working-memory accuracy trial.

A second hint comes from a 2025 intervention by Intawachirarat et al. in school-aged children with obesity. Contemporary dance training increased brain-derived neurotrophic factor, or BDNF, and improved executive function measures. That makes a neurotrophic pathway plausible. It still does not settle the main question because the population differed from the healthy children targeted in Guo et al. Obesity changes baseline metabolic and exercise-response profiles, so you cannot swap those children into a healthy-child analysis and pretend nothing changed.

What the ADHD literature shows, and what it does not

If you only read the ADHD literature, you would come away with a different top contender. In a 2026 network meta-analysis by Yang et al. and a 2025 network meta-analysis by Ouyang et al., aerobic and combined exercise repeatedly rank well for executive function outcomes in children with ADHD. Ouyang et al. pooled 26 randomized controlled trials. That is enough to take the pattern seriously.

But "executive function" is a bundle, not a single dial. A 2026 multilevel meta-analysis by Haozhi Wang et al. helps explain the mismatch: aerobic exercise's advantage in ADHD seems to map most strongly to inhibitory control. That is not the same endpoint as working-memory accuracy. A 2025 systematic review by Pengfei Wang et al. also supports aerobic exercise as a useful strategy for executive function in ADHD, but again in a clinical population. Liu et al. 2025 found that exercise effects in ADHD were moderated by exercise type, session duration, and symptom severity, and dance did not emerge as the standout option for working memory there.

That difference does not "contradict" the healthy-child ranking so much as narrow its scope. Clinical and neurotypical populations do not share the same baseline neurobiology. A training mode that works best for inhibitory control in ADHD does not have to be the best option for improving accuracy in healthy children.

The limits are serious

The biggest limit is the missing quantitative core. The available evidence does not provide the primary network meta-analysis's SUCRA values, standardized mean differences, or confidence intervals. Without those numbers, we cannot tell whether dance clearly outranked aerobic exercise, barely edged it out, or tied with it inside wide uncertainty bands. If the confidence intervals overlap heavily, the headline claim weakens a lot.

The reaction-time side is even shakier. Only eight trials reportedly contributed reaction-time data to the Guo et al. network. Sparse networks are vulnerable to single-study influence and unstable indirect comparisons. So even if aerobic exercise ranked first for speed, that ranking should be treated as preliminary until the full results confirm the margin and precision.

There is also a scope problem. The study coded six modalities, not eight. Any version of this story that inflates the comparison set is adding facts the evidence here does not support. The age range is broad as well, spanning 6 to 18 years. That covers major developmental differences in prefrontal maturation. A modality that works best in younger children may not lead in teenagers, and the pooled result cannot answer that without subgroup data.

Finally, adults should resist the urge to borrow this finding for themselves. There is no direct evidence here for healthy adults ages 25 to 65. The adult studies in this set concern older people or people with mild cognitive impairment. In the 2025 EXERT trial, 296 sedentary older adults with amnestic mild cognitive impairment were randomized to moderate-high intensity aerobic exercise or low-intensity stretching, balance, and range-of-motion work for 18 months. The intervention did not produce a between-group advantage on the primary cognitive outcome at 12 months; the reported regression was -0.078 with a standard error of 0.074 and p = 0.3. That does not tell us dance beats running in healthy 35-year-olds. It tells us, very specifically, that one aerobic program did not outperform a low-intensity comparator on global cognition in an older at-risk group.

What parents and schools can do with this

If the goal is to choose one activity for healthy children right now, the best evidence-based answer is conditional. The 2025 network meta-analysis appears to suggest that structured dance may be the best bet for working-memory accuracy, while aerobic exercise may be the best bet for reaction time. That is plausible, useful, and not yet fully verified at the level of effect size.

If a school or family can mix activities, that is the least fragile conclusion. Accuracy and speed are not redundant outcomes, and different training demands may help different parts of working-memory performance. If only one activity is practical, the choice should follow the goal. For precision-heavy academic tasks, dance has the more interesting signal. For cognitive speed, aerobic exercise has the broader support base, especially because the aerobic story also shows up across ADHD meta-analyses, even if mostly through inhibitory control.

The verdict is cautious by necessity: exercise type probably matters for working memory in healthy kids, and the early leader for accuracy may be dance rather than the usual aerobic default. But until the primary network meta-analysis's numerical results are verified, that ranking is a reported signal, not a settled fact.