Physical activity and learning: what the research shows

Every school timetable is a zero-sum ledger. When test pressure rises, minutes move from somewhere to somewhere else. And for two decades the minutes have moved in one direction: out of physical education and recess, into seat time. The logic feels airtight. Children learn maths in maths lessons, not on the playground. So more lesson, more learning. The trade rests on one assumption — that moving and learning compete for the same minutes and give nothing back to each other.

That assumption is testable, and it has now been tested at every level science has. A meta-analysis — a study that pools many studies into one estimate — put a number on the link between physical activity and children’s cognition back in 2003 (Sibley & Etnier, 2003). Lab studies then measured what a single 20-minute bout of walking does to a child’s attention (Hillman et al., 2009). Classroom trials measured what short movement breaks do to on-task behaviour (Mahar et al., 2006). A randomized trial dosed exercise like a drug and watched executive function respond (Davis et al., 2011). And in 2016 a full position stand graded the entire evidence base (Donnelly et al., 2016).

The verdict is worth stating plainly, because both camps overclaim. Physical activity is not a cognitive miracle. It will not raise a school’s league-table position by itself. But the link is real, it runs in the helpful direction, and — the finding with the most policy weight — taking time from lessons for movement does not cost achievement. The trade schools keep making has the sign wrong. This article walks the evidence in the order it arrived.

The trade schools keep making

Start with how common the trade is. In a large US cohort of eight- and nine-year-olds, roughly three children in ten had little or no daily recess at all (Barros, Silver & Stein, 2009). The pattern was not random. Children in poorer schools, and in schools under the most test pressure, were the most likely to lose their break. The children most likely to need a reset were the least likely to get one.

The implicit theory behind the cut is that attention is a tap: keep the child in the seat and learning keeps flowing. The rival theory says attention is a tank. It drains with use, and movement is one of the few things that refills it during a school day. A second rival theory works on a slower clock: regular aerobic activity changes the brain itself — the tissue and circuits that attention and memory run on. Both rival theories make measurable predictions. One predicts short-lived gains right after activity. The other predicts that fitter children think differently in ways that build over months.

What follows is the evidence for each, in sequence. The honest summary up front: the effects are real and repeatedly found, and they are modest. Anyone selling movement as a substitute for teaching is wrong. So is anyone treating it as stolen time.

The first pooled answer

The modern literature starts with Sibley and Etnier’s 2003 meta-analysis. They pooled more than 40 studies relating physical activity to cognition in school-age children, spanning designs from simple correlations to true experiments (Sibley & Etnier, 2003). The overall relation came out at roughly 0.32 standard deviations — positive, reliable, and moderate-to-small. In plain terms: more active children tended to score somewhat better on cognitive measures than less active ones.

The pattern across outcomes mattered as much as the average. Perceptual and motor-linked skills showed the largest relations, at roughly half a standard deviation. Measures of IQ and of academic achievement sat near a third. Maths and verbal test relations were smaller. And memory — the outcome parents most want moved — showed essentially no relation at all (Sibley & Etnier, 2003). That profile recurs through the rest of this article: movement helps the attention-and-control end of thinking more than the storage end.

The 2003 result carried two caveats its authors flagged. Most pooled studies were correlational, so the number describes a relation, not a proven cause. Active children differ from inactive children in many ways — family income, sleep, health — and those differences could carry some of the link. And the studies varied widely in quality. The meta-analysis therefore set an agenda rather than settling one: the relation exists; now isolate the cause. The next fifteen years of work did exactly that, in three settings — the lab, the classroom, and the randomized trial.

One bout, twenty minutes, measurable

The cleanest causal test is the smallest one: take a child, add one bout of exercise, and measure thinking right after. Hillman and colleagues ran the canonical version. Nine- and ten-year-olds came to the lab twice. On one visit they walked on a treadmill for 20 minutes at a moderate pace; on the other they sat and rested. After each, they took a computerised attention task and short academic tests in reading, spelling and arithmetic (Hillman et al., 2009).

After walking, the same children were more accurate on the attention task — specifically on the trials that required ignoring distraction. Their brain response told the same story. The researchers recorded event-related potentials, small voltage changes the brain produces while working. A component called P3, which tracks how much attention a person allocates, was larger after exercise (Hillman et al., 2009). And performance on the academic tests was better after walking than after sitting, most clearly in reading.

Three things make this small study load-bearing. It is causal: same child, both conditions, order controlled. It is cheap: the “intervention” is a brisk 20-minute walk. And it is time-limited: the window is minutes to an hour, not the afternoon. A single bout buys a sharper next lesson, nothing more. The mechanism sits at the arousal end of the story — exercise transiently raises the neurochemicals that tune alertness, which is why the effect lands on attention rather than knowledge (Hillman, Erickson & Kramer, 2008). The slower, structural story comes later. First, the same experiment run where it matters: in classrooms.

Movement breaks and recess in real classrooms

Lab effects often shrink on contact with school. This one survived. The Energizers programme trained teachers to run short activity breaks — about ten minutes of movement woven into lessons — and observers coded pupils’ on-task behaviour before and after the programme, interval by interval (Mahar et al., 2006).

On-task behaviour rose by roughly 8 percentage points after the breaks were introduced. The distribution of the gain is the finding schools should care about. For the pupils who were least on-task at baseline — the fidgeters, the drifters, the ones every teacher is thinking of right now — the improvement was on the order of 20 points (Mahar et al., 2006). The break did little for children who were already attentive. It bought back the attention of the children who had none to spare. A ten-minute break that returns eight points of on-task time across a class is not lost instruction. It is instruction recovered.

Recess shows the same direction with weaker causal footing. In the US cohort mentioned earlier — several thousand eight- and nine-year-olds — children who got at least 15 minutes of daily recess were rated better behaved in class than children who got none, on teachers’ own ratings (Barros, Silver & Stein, 2009). This is an association, not a trial. Schools that protect recess differ in many other ways, and the authors said so. But the association ran the same way in a design where the deck was stacked against finding it: the no-recess children were disproportionately in schools already fighting for attention. Cutting their break did not appear to be helping.

Fitness, the brain, and the dose question

Everything so far is the fast clock: one bout, one break, one sharper hour. The slow clock asks what months of regular aerobic activity do to the machinery itself. The background evidence here is broad. Aerobic exercise reliably changes brain structure and function in animal studies and in older adults — more of the growth factors that support new connections, better-preserved memory structures, stronger frontal control networks (Hillman, Erickson & Kramer, 2008). In children, higher-fit and lower-fit groups differ the same way: fitter children tend to show stronger performance on tasks that demand cognitive control, and corresponding differences in brain measures. Those comparisons are correlational. The causal test needs a randomized trial.

Davis and colleagues ran it like a drug trial, with a dose. They randomized 171 sedentary, overweight seven- to eleven-year-olds to one of three arms for around 13 weeks: no programme, 20 minutes of aerobic games after school each day, or 40 minutes (Davis et al., 2011). The outcome was executive function — the family of skills that includes planning, holding a goal in mind, and resisting impulse — plus academic achievement. Executive function is glossed here because it is the load-bearing term of this whole literature: it is the cluster movement keeps landing on.

The result was a dose–response curve, the signature causation leaves. Planning scores improved with exercise, and improved more with 40 minutes than with 20. Maths achievement moved the same way. Reading did not move, and broad vocabulary did not move — the effect was specific, not a general rise in cleverness (Davis et al., 2011). A brain-imaging subgroup showed increased activity in prefrontal regions, the seat of exactly the control functions that improved. The study population was specific — sedentary and overweight children, where headroom is largest — so the size of the gain will not transfer to every classroom. The shape of it, more dose, more effect, is the point.

The position stand and the honest meta-analysis

By the mid-2010s the literature was large enough to grade as a whole. The American College of Sports Medicine convened the field’s senior researchers — including authors of most of the studies above — to write a position stand: a formal, evidence-graded review (Donnelly et al., 2016). Its conclusions are the fairest one-paragraph summary the field has.

The panel found the links between physical activity, fitness and children’s cognition to be real and positive, with the strongest and most consistent evidence sitting exactly where this article has kept landing: fitness and executive-type control functions, and brain structure and function. On academic achievement the grade was more cautious. Results were positive to null, rarely negative, and too mixed in quality to declare cause and effect settled (Donnelly et al., 2016). The panel was blunt that many studies were small, short, and loosely controlled. And it underlined the floor result that matters most for timetables: across the trials that moved lesson time into physical activity, achievement did not drop.

A year later, a meta-analysis of the intervention studies alone — only designs where activity was added and cognition was measured against a comparison group — put the causal number in place. Pooled across trials, physical-activity programmes improved children’s cognition and metacognition by a small margin, on the order of 0.2 standard deviations, with benefits spread across attention-and-control functions and thinking-about-thinking measures (Álvarez-Bueno et al., 2017). Set the three numbers side by side and the literature snaps into focus. The correlational relation is roughly 0.32. The causal, intervention-driven effect on cognition is roughly 0.2. The causal effect on achievement itself is smaller still and not yet pinned down — positive in direction, unproven in size (Donnelly et al., 2016). Correlation shrinks on the way to causation, as it almost always does. It does not vanish. And nowhere on the path does movement cost learning.

What the evidence doesn’t show

The movement literature earned its positive verdict, and it is routinely oversold anyway. The limits below are as much a part of the finding as the headline.

• No promise of smarter children. The causal effect on cognition pools at roughly 0.2 SD — worth having, visible in a class, invisible in a single child (Álvarez-Bueno et al., 2017). Programmes promising transformation are quoting the correlational number, or nothing.
• The chain to grades is the weak link. Activity to attention is well evidenced; attention to achievement is where the designs thin out. The position stand explicitly declined to call achievement effects settled (Donnelly et al., 2016).
• Memory barely moves. The outcome profile has been lopsided since 2003 — perceptual and control skills respond, memory measures sit near zero (Sibley & Etnier, 2003). Movement is not a retention strategy.
• Dose and type are under-specified. One trial shows 40 minutes beats 20 (Davis et al., 2011), but the field has no curve — no established minimum, optimum, or best modality for cognitive benefit.
• Acute effects are short-lived. The single-bout gains sit in a window of minutes to an hour after activity (Hillman et al., 2009). A morning run does not carry the afternoon.
• Small studies, soft controls. Many trials are brief, under-powered and loosely controlled, and null results are the kind that go unpublished — cautions the reviewers themselves raise (Donnelly et al., 2016).

Where the evidence stops

1. 1No promise of smarter children
2. 2The chain to grades is the weak link
3. 3Memory barely moves
4. 4Dose and type are under-specified
5. 5Acute effects are short-lived
6. 6Small studies, soft controls

© 2026 FUTURE PROOF™

Movement by the evidence

Read as one body of work, the literature converts into a short operating manual for schools, and none of it requires new budget.

Stop trading physical education for test preparation. The best-graded finding in the field is the floor result: moving lesson time into activity did not lower achievement in the trials that tried it (Donnelly et al., 2016). The cut buys nothing measurable and spends the one input with a demonstrated attention payoff.

Protect recess as an attention budget, not a treat. Withholding break correlates with worse classroom behaviour, not better (Barros, Silver & Stein, 2009). Recess-removal as a punishment takes the reset away from exactly the child who needs it most — the child the movement-break data says gains roughly 20 points of on-task time from moving (Mahar et al., 2006).

Place movement before the demanding block. The single-bout window is minutes to an hour (Hillman et al., 2009). A ten-minute break before maths is timed medicine; the same break at day’s end is just a break. Teachers can run this as a within-class experiment: same lesson, break versus no break, count the redirections.

Treat fitness as slow infrastructure. The dose–response trial says sustained, repeated aerobic activity is what moves executive function, and more dose moves it more (Davis et al., 2011). That is a term-length commitment, and its likeliest beneficiaries are the least active children — the group standing furthest from the ceiling.

Measure attention, not next week’s test scores. The causal effects land on attention and control first (Álvarez-Bueno et al., 2017); achievement, if it follows, follows slowly through them. A school evaluating a movement programme against one term of test results is measuring the wrong variable at the wrong lag. Count on-task behaviour, as the classroom studies did (Mahar et al., 2006).