Cognitive training for ADHD has a long and mostly disappointing track record. Children get better at the trained task, reliably and measurably. What has proven far harder to demonstrate is that the improvement goes anywhere else. A large meta-analysis of randomized trials led by Samuele Cortese found that once you rely on blinded raters rather than the parents who knew which group their child was in, the evidence for cognitive training improving ADHD symptoms largely thins out. The pattern has repeated often enough that skepticism is the reasonable default. So a trial published in July in the Journal of Medical Internet Research is interesting less for what it found than for how it was built.

The study, led by Feilong Zhu of Beijing Sport University with collaborators at Beijing Normal University, Peking University Sixth Hospital, and several other institutions, randomized 66 children aged 6 to 10 with the inattentive presentation of ADHD. Thirty-three received a fine-motor training program delivered live online, three times a week, 60 minutes a session, for 12 weeks. Thirty-three were placed on a waitlist. Assessments were taken at baseline, immediately after the program ended, and again three months later.

Outcomes favored the intervention on parent-reported inattention, on selected executive function measures, and on fine motor skills. That is the headline. The more useful material is underneath it.

Why Fine Motor, and Why This Group of Children

Two choices in this trial need explaining before the results mean anything, because neither is obvious.

The first is fine motor skill. Motor difficulty is not part of the diagnostic criteria for ADHD and gets little attention in most discussions of it, yet estimates in the literature put the proportion of children with ADHD showing motor deficits somewhere between 30 and 70 percent. It is one of the most common features of the condition that almost nobody lists when describing it. Children who struggle to form letters, manage buttons and zips, or manipulate small objects accumulate a particular kind of daily friction, and handwriting difficulty in particular has obvious consequences in a classroom. Targeting fine motor skill in this population is therefore reasonable on its own terms, before any question of executive function enters.

The second is the specific population. This trial enrolled children with the inattentive presentation, not ADHD generally. That group is chronically under-served for a structural reason: they are not disruptive. A child who cannot sit still and interrupts the lesson gets referred. A child who is quietly elsewhere does not, or does so years later. The inattentive presentation is identified later, more often missed entirely, and studied less, which means interventions are frequently designed and validated on samples dominated by children whose difficulties look different. A trial that recruits this group deliberately is filling a real gap.

The combination also has a logic to it. If fine motor difficulty is common in ADHD, and if executive demands can be layered onto motor activity, then a single intervention can address something concrete and something abstract at once. Whether that logic holds is what the trial set out to test.

The Design Idea Worth Noticing

Most cognitive training asks a child to practice a cognitive operation in isolation. Hold four items in mind and reproduce them. Press the button for one shape and not another. Sort by color, then sort by shape. The operation is the whole task, stripped of context, delivered by software.

This program did something different. The activities were fine-motor tasks, the kind of hands-on manipulation a child would recognize as an activity rather than a test. The executive demands were built into them. Inhibition, sequencing, working memory, and switching requirements were embedded inside activities that had their own visible purpose and their own physical product.

That distinction is not cosmetic. The persistent failure of computerized training is a failure of transfer: the practiced operation does not carry into contexts that differ from the practice. One long-standing hypothesis about why is that stripping an operation out of context also strips out the retrieval cues, the motivation, and the situational structure that make the operation usable elsewhere. A child who has practiced inhibition only as a button-press has practiced it in a setting nothing in their life resembles.

Embedding the demand inside a meaningful activity is an attempt to address that directly. Whether it succeeds is a genuinely open question that this trial does not settle. But it is the right question, and it is a more interesting one than whether another training app produces another within-task gain.

What the Trial Actually Did

The methodological choices here are worth listing, because several of them are better than the field’s average.

Participants were randomized, which rules out the selection effects that plague single-arm designs. Assessors were blinded to group assignment. The program was manualized and delivered on a fixed schedule rather than improvised session to session. Attendance was tracked. Analysis used linear mixed-effects models, which handle missing data and repeated measures more gracefully than simpler approaches. The trial was prospectively registered with the Chinese Clinical Trial Registry, which constrains the temptation to redefine outcomes after seeing results.

And critically, there was a three-month follow-up. Most intervention studies in this space measure at the end of the program and stop, which tells you nothing about whether anything persists. This one looked again, and found that effects weakened on some outcomes over that interval. That is a less flattering result than the immediate post-test, and reporting it is to the authors’ credit.

The trial also sits within a small but growing body of work on remote delivery of physical interventions to children. A separate randomized trial published in 2025 delivered fine motor training by telerehabilitation to 70 children with ADHD aged 6 to 8, three days a week for eight weeks, and measured handwriting performance and motor proficiency with standardized instruments. That study reported improvements in handwriting. Neither trial is decisive on its own, but they are asking a coherent question from different angles, which is how an evidence base begins to form rather than how a single result gets overinterpreted.

The dose deserves attention too. Three sessions a week for 12 weeks is 36 hours of structured, live, one-to-one or small-group time. That is a substantial commitment for a family, and it matters for anyone thinking about translation. Whatever this program achieved, it achieved with a great deal of contact time.

The Waitlist Problem

Here is what a waitlist control can and cannot do.

It establishes that the treated group changed more than an untreated group over the same period. That rules out maturation, regression to the mean, and the ordinary drift of a school term. Those are real threats and the design handles them.

What it cannot do is separate the specific ingredient from everything that came bundled with it. The intervention children received 36 hours of attention from a trained instructor. Their parents were involved in setting up and supervising sessions three times a week for three months. Their families had a structured routine imposed on a recurring schedule. Everyone knew they were receiving the thing being studied and presumably expected it to help. The waitlist children received none of that.

So when parent-reported inattention improves more in one group than the other, the honest reading is that something in that bundle helped. Which part is an open question. A control group receiving an equally intensive activity without the embedded executive demands would answer it. This trial did not have one, and the authors say so.

The parent-report issue compounds this. Parents were the source of the primary symptom outcome, and parents knew perfectly well which group their child was in. A parent who has watched their child complete 36 sessions of structured activity has both an expectation and an investment. Blinded assessors handled the other measures, which is why the assessor-blinded outcomes carry more weight than the parent-reported one, even though the parent-reported one is the more intuitive headline.

Read the Gradient, Not the Summary

There is a useful way to read a study like this: rank the outcomes by how far each sits from what was actually trained, and let your confidence fall off with distance.

Fine motor skills sit closest. The children spent 36 hours practicing fine motor tasks and their fine motor skills improved. That is the least surprising result in the paper and the one requiring the least explanation.

Executive function outcomes sit further out, and the trial reports benefits on selected measures rather than across the board. That word is doing real work. When some measures within a domain move and others do not, the possibilities include a genuinely specific effect, a partial one, or noise in a sample of 66. Distinguishing among those requires either a larger trial or replication.

Parent-reported inattention sits furthest from the training and is the least well-controlled measure in the study. It is also, predictably, the result most likely to be quoted.

The gradient is not a reason to dismiss the study. It is the correct way to read any intervention trial, and it happens to be the reading the authors themselves invite.

The Mediation Analysis, and Why It Is Labeled Exploratory

The researchers also ran a mediation analysis, and the way they describe it is instructive.

Mediation asks not whether an intervention worked but how. If fine motor training improved inattention, did it do so by way of improved executive function, or by way of improved motor skill, or through some path not measured? Answering that would matter a great deal for anyone trying to build on the result, because it would identify the active ingredient rather than just the package.

The authors call their mediation analysis exploratory, and they conducted it using values taken at the 12-week post-intervention point. That combination is the limitation. Establishing that A caused B by way of C requires showing that C changed before B did, and measurements taken at a single moment cannot demonstrate that ordering. What such an analysis can show is that the variables are related in a pattern consistent with mediation, which is considerably weaker and is why the word exploratory belongs there.

This is a common feature of intervention research and not a failing peculiar to this study. It is worth understanding because mediation results are frequently reported downstream as though they had established a mechanism. Here the authors flagged the limitation themselves, in the abstract, alongside the other two.

Why the Authors’ Own Hedging Matters

Research papers are written by people who want their work to matter, and conclusions usually reflect that. This one is different in a way worth pointing out.

The authors describe the program as one that may be feasible and potentially beneficial as an adjunctive intervention. They then state directly that the findings should be interpreted cautiously, and name three specific reasons: the waitlist control design, the parent-reported symptom outcome, and the exploratory nature of the mediation analysis. That is the research team volunteering the three strongest objections to their own paper, in the abstract, where everyone will see them.

The word adjunctive is doing work too. They are not positioning this as an alternative to established ADHD treatment. They are positioning it as something that might sit alongside it.

When a research team is this careful about its own limits, the appropriate response is to take the finding at exactly the weight they assign it, rather than the weight a press release would.

What This Is Evidence For, and What It Is Not

This is not evidence for ADHD coaching. The population is children aged 6 to 10, the intervention is a structured motor-training protocol delivered by trained instructors, and the outcomes are symptom and skill measures. None of that transfers to coaching adults or adolescents, and anyone citing it that way is citing something the study did not examine.

It is also not evidence that fine motor activity in general improves attention. What was tested was a specific manualized program with specific embedded cognitive demands at a specific and substantial dose. Craft time is not the intervention. The intervention was 36 structured hours built around particular executive requirements.

What it is evidence for is narrower and more interesting: that embedding executive demands inside meaningful activity is a design approach worth testing further, and that a program built this way produced measurable change in a randomized trial with blinded assessors and a follow-up. That is a real result. It is also a starting point rather than a conclusion, which is roughly where the authors place it.

The comparison that would settle the question has not been run. Take two groups of children, give both an equally engaging hands-on activity at the same dose with the same instructor attention and the same parent involvement, and build the executive demands into only one of them. If the embedded-demand group still comes out ahead, the design principle is doing the work. If both groups improve equally, then what helped was 36 hours of structured attention, which would itself be worth knowing.

What Follows in Practice

For clinicians and educators, the immediate implication is modest and mostly about design rather than prescription. If you are constructing an activity intended to build executive capacity, the evidence base does not favor isolating the operation and drilling it. Building the demand into something the child is genuinely doing is at minimum a defensible alternative and possibly a better one.

The delivery finding is worth separating out. This program was delivered live online, in children’s homes, three times a week, with sufficient adherence to produce measurable results. Whatever one concludes about the mechanism, remote delivery of a hands-on intervention to young children with attention difficulties is feasible. That has practical value for families who cannot access in-person services, and it is the kind of thing that only gets established by someone running the trial.

And the follow-up finding is the one to remember. Effects weakened on some outcomes three months after the program ended. Programs that stop tend to produce gains that fade, which is an argument for maintenance rather than an argument against the program. It is also a reminder that a post-test taken on the last day of an intervention is measuring the intervention at its most flattering moment.