How Many Steps for Teen Mental Health: No Plateau to 12,000

A parent standing in a kitchen at eleven at night, having just read that walking more is good for a child's mental health, has a narrower question than the headline answers. Not whether movement helps, which has been settled for years, but how much, at what age, and whether the round number printed on a fitness tracker has anything to do with the number in the study. The coverage offered three tidy bands: under 5,000 steps for the most symptoms, around 8,000 for fewer, 12,000 or more for the fewest. The paper underneath it has five bands, and no boundary at 8,000 anywhere in it.

That paper is a cohort study by Qingyang Li, Amanda Paluch and Divyangana Rakesh, published in JAMA Network Open on 1 October 2026 as Physical Activity and Adolescent Brain Structure, Mental Health, and Academic Outcomes. It asks what a plain daily step count predicts about an adolescent's mental health a year later, which is a more ordinary question than it sounds: not a fitness test, not a questionnaire about exercise, just the number a cheap wrist device counts while a child walks to school.

The data come from the Adolescent Brain Cognitive Development Study, release 5.1, which runs across 21 sites in the United States. The analysis included 5,515 adolescents, 2,640 of them (47.9%) female, with a mean age of 11.93 years (SD 0.65). Each wore a Fitbit Charge 2 on the wrist for three weeks before the study's two-year follow-up visit, the earliest point at which step data exists in the cohort. Step and imaging data were collected between November 2018 and November 2020, the analysis was run between May 2025 and August 2026, and the analysis plan was preregistered on the Open Science Framework.

Mental health a year later was measured with the Brief Problem Monitor, a 19-item participant-reported questionnaire covering internalizing, externalizing and attention symptoms, and an abbreviated form of the Child Behavior Checklist. Academic performance was the average of parent- and youth-reported school grades on a 12-point scale, reverse-coded so that higher meant better. Both were taken at the three-year follow-up, roughly a year after the step counters came off.

One design choice does most of the work. The models adjusted for each adolescent's own earlier score on the same measure, so the question is not whether active children are doing better than inactive ones, but whether they moved further from where they already were. Household socioeconomic status is the usual control, and this analysis added neighbourhood factors as well, including neighbourhood disadvantage and access to green space, which shape how walkable a child's daily life is before any choice about walking is made. Imaging models adjusted for intracranial volume, and multiple comparisons were handled with false-discovery-rate correction within each family of tests.

Treated as a continuous measure, a higher step count was associated with fewer total mental health problems a year later: a coefficient of −0.080, standard error 0.015, t = −5.497, P < .001. The categorical analysis is the one built to answer the parent's question, because its purpose was to test whether the benefit stops anywhere. Step counts were grouped into five bands — under 5,000, 5,000 to 7,499, 7,500 to 9,999, 10,000 to 11,999, and 12,000 or more — with boundaries chosen to sit close to the sample's own quartiles of 7,366, 9,323 and 11,446 steps.

Measured against the under-5,000 group, every higher band had a lower symptom score, and the gaps widened band by band: 0.048 points lower for 5,000 to 7,499 (95% CI, −0.089 to −0.007; P = .02), 0.058 lower for 7,500 to 9,999 (−0.098 to −0.018; P = .004), 0.073 lower for 10,000 to 11,999 (−0.115 to −0.032; P = .001), and 0.102 lower for 12,000 or more (−0.144 to −0.059; P < .001). A post hoc comparison put the most active band below the two middle bands as well, by 0.053 points (95% CI, 0.014 to 0.092; P = .002) and 0.043 points (0.009 to 0.078; P = .005). Adding a quadratic term did not improve the model's fit, which is the statistical way of saying the line did not bend.

So there is a genuine dose-response gradient with no ceiling inside the observed range, and the total span from the least to the most active band is about a tenth of a point on a 19-item questionnaire. Both halves of that sentence are true, and a reader who keeps only one of them has the story wrong. The shape of this finding is more interesting than its size: whatever is happening did not run out at 10,000 steps, a figure that has anchored public-health messaging for decades without ever having been derived from evidence about children.

The brain results are narrower than the reporting suggested. A higher step count was associated with greater cortical thickness in two regions, the paracentral (0.057, SE 0.016; P < .001) and the rostral anterior cingulate (0.052, SE 0.016; P = .001), and was not associated with average cortical thickness, total surface area or total cortical volume. The effect is regional rather than a generally thicker brain. Paracentral thickness then statistically mediated the link between steps and later mental health, with an indirect effect of −0.002 (SE 0.001; 95% CI, −0.0045 to −0.0004; P = .02) that held after adjustment for prior mental health. No other mediation path survived correction, and the authors themselves note that choosing mediators in the same sample can inflate exactly this kind of estimate.

Two further places where the paper and its coverage part company are worth naming. The press materials reported no link between steps and school performance; the paper reports a flat continuous association plus one isolated categorical result, with the 10,000-to-11,999 band scoring 0.262 higher than the under-5,000 band at P = .044. A single significant cell out of four, at that threshold, is closer to nothing reliable than to nothing tested, and the distinction matters to anyone deciding what the study showed. And the widely repeated 8,000-step figure does not appear in the paper at all: its middle bands are 5,000 to 7,499 and 7,500 to 9,999. A number that reaches the public ought at least to be one the study measured.

What makes this work useful is less the effect size, which is small, than what the exposure was. Previous research in this area leaned on self-reported activity or on accelerometer windows as short as 24 hours, in samples of roughly a hundred adolescents, and adjusted for household circumstances while leaving the neighbourhood out. A plain step count from a wrist device worn for three weeks, in 5,515 children across 21 sites, with each child's earlier symptom score already controlled for, is a more measurable and more ordinary thing. Ordinary is the point: steps accumulate from the walk to school and from messing about outside, not from anything a family has to buy or join.

The authors' own framing is the modest one. Daily steps are an accessible target associated with better mental health and brain development, not a treatment and not a threshold to clear. The absence of a plateau is the part most likely to interest clinicians, because it argues against the premise built into every round-numbered step goal: that there is a point at which a child has done enough, and beyond which more movement stops counting for anything.

The number worth keeping from this paper is not a step target but the absence of one. Across five bands, symptom scores kept falling — through 10,000, the figure that has anchored a generation of public-health messaging despite never having come from evidence about children, and on to 12,000 and above, where the line still had not bent. That shape is the finding, and it is a more useful thing to know than any single threshold would have been. The size deserves the same honesty: a tenth of a point on a 19-item questionnaire between the least and most active adolescents in the cohort is a population signal, not a change anyone would see in one child, and the brain-mediation path that drew the most enthusiastic coverage rests on an indirect effect of two thousandths that the authors themselves caution may be inflated. What this study does better than its predecessors is measure the exposure plainly, with a wrist device rather than a questionnaire, in 5,515 children rather than a hundred, with the neighbourhood as well as the household accounted for. The distance between that and the three tidy step bands that reached the public — one of which, 8,000, appears nowhere in the paper — is where this news cycle mislaid the actual result.