One Month of Coding Changed How Six-Year-Olds Think. Not About Coding.

Seventy-six first graders in Padova, Italy, spent one month learning to code. When researchers tested them afterwards, the children hadn’t just learned to code. They had become measurably better at planning: the ability to look at a complex problem, break it into steps, and execute them in order. Not coding problems. Any problems.

The control group, who spent the same month on standard STEM activities, did not show the same improvement.

This was not a long intervention. Not an intensive programme. Not a study of gifted children or well-funded schools. It was four weeks of curriculum-integrated coding with ordinary children, and it shifted something in how they approached difficulty.

What the study actually measured

The University of Padova team, led by Barbara Arfé, ran two studies. The first was a cluster-randomised controlled trial with 76 first graders, mean age 5.9, randomly assigned to either a coding intervention or standard STEM lessons. The second replicated the design with second graders and added a longitudinal follow-up.

They measured two specific executive functions before and after the intervention. Planning was assessed using two standard tests: the Tower of London, which asks children to rearrange coloured beads to match a target in the fewest possible moves, and the Elithorn Perceptual Maze, which requires mapping an efficient path through a grid. Response inhibition was assessed using the NEPSY-II and a Stroop-style task, both of which measure whether a child can override an automatic response in favour of a deliberate one.

The coding itself was simple. Children used visual block-based tools, dragging instructions into sequences to guide a character through progressively harder challenges. No typing. No syntax. Just: work out the order of steps, predict what will happen, test it, fix what didn’t work.

The gain that transferred out of the domain

After one month, the coding group showed significant improvements in planning accuracy on both the Tower of London and the Elithorn tasks. The effect held even after controlling for age, socioeconomic status, and baseline performance. The control group, receiving the same amount of instructional time on other STEM content, did not.

The second study replicated this with second graders and found the same pattern. It also tracked children longitudinally, and the gains in planning persisted after the intervention ended.

What makes this worth paying attention to is what planning actually is. It is not a coding skill. It is a general cognitive capacity: the ability to hold a goal in mind, consider multiple possible steps, evaluate which sequence is most efficient, and resist the urge to act impulsively before thinking it through. It is what a child uses when they organise a piece of writing, work through a multi-step maths problem, or decide how to approach a social situation that has gone wrong.

The coding was the training stimulus. The gain was architectural.

Why coding does this and other activities don’t

The Padova researchers proposed a specific mechanism. Computational thinking, which underpins coding, requires children to decompose problems into ordered sub-steps, define those steps unambiguously, predict outcomes before executing, and debug when the result doesn’t match the intention. Each of these demands maps directly onto the executive function of planning: the prefrontal processes responsible for goal-directed behaviour, sequencing, and strategic adjustment.

Other STEM activities involve some of these elements. Building a bridge from straws involves trial and error. But coding is unusual in how tightly it constrains the feedback loop. A block-based programme either does what you intended or it does not. There is no ambiguity. The child cannot fudge the result or half-succeed. That binary clarity forces a precision of thinking that more open-ended tasks allow children to sidestep.

This is consistent with a broader finding from Johns Hopkins University, published in the Journal of Neuroscience in October 2025. Using fMRI, researchers tracked the brain activity of 22 university students before and after they learned Python. When students read code after the course, fronto-parietal regions associated with logical reasoning activated in response to the programme’s meaning. But even before the students learned to code, when they read plain English descriptions of the same algorithms, the same neural populations were already active.

Coding does not appear to create new cognitive machinery. It recruits and strengthens what is already there: the logical reasoning architecture that every brain develops in early childhood. The Padova study suggests that this recruitment can begin far earlier than university, and that even a modest amount of structured coding activates and sharpens the planning systems children will rely on across every domain.

What this doesn’t prove

The Padova study was well-designed but small: 76 children in the first study, with a replication in the second. The coding intervention was delivered by trained researchers, not by regular classroom teachers, and the conditions were more controlled than a typical primary school timetable would allow. The researchers themselves noted that the effects of coding on older children and on children from lower socioeconomic backgrounds remain untested.

The Johns Hopkins fMRI study involved university students, not children. The inference that young children’s brains respond similarly is reasonable but not directly demonstrated by that study.

Executive function is influenced by many things: sleep, nutrition, emotional regulation, the quality of relationships. Coding is one input. It is not a substitute for the others, and the planning gains observed in these studies should be understood as promising, not guaranteed.

The honest costs of starting early

Structured coding for five- and six-year-olds typically involves a screen. That is worth acknowledging plainly, because many parents are simultaneously trying to reduce screen time and considering activities that require it. The Padova study used a block-based platform on tablets. That is screen time, regardless of what the child is doing with it.

Young children will also need an adult nearby. The coding platforms used in research are designed for guided use, not independent play. A parent or teacher who sets a child in front of a coding app and leaves the room is not replicating the conditions that produced these results.

Not every child will find it engaging immediately. The first graders in the Padova study were motivated, but they were also in a structured classroom environment with trained facilitators. At home, the first few sessions may feel more like persuasion than education. That is normal. It is also a reason to keep sessions short. The intervention used focused lessons, not marathon sessions.

What this changes about the usual advice

Most guidance to parents about coding focuses on future career relevance or “digital literacy.” The Padova research shifts the frame. The reason to introduce young children to coding is not that they might become programmers. It is that the act of coding, even at a very basic, visual, block-dragging level, exercises the executive functions that predict academic performance across subjects, emotional regulation, and the ability to manage complexity.

A meta-analysis of 21 samples involving 7,947 primary school children found that executive functions had a moderately significant effect on academic performance across the board, with the predictive power strongest in the early years. Planning and working memory were the most consistent predictors. If coding strengthens planning, and the Padova evidence says it does, then the downstream benefits are not confined to anything resembling a screen.

Your child programs a character to navigate a maze on a tablet. They get it wrong, rethink the sequence, try again. That is the same thinking they will use when a maths problem doesn’t make sense, or when an essay needs restructuring, or when a plan with a friend falls apart and they have to come up with a new one. The domain changes. The process doesn’t.

The first graders in Padova didn’t know they were training their executive functions. They thought they were making a bird walk through a grid. The planning improvements showed up on tests that had nothing to do with coding, in a cognitive domain the children didn’t know existed.

That is what makes the finding interesting. Not that coding teaches coding. That it quietly reorganises something more fundamental, and it can start doing so before a child learns to write in joined-up letters.

Research sources

Arfé, B., Vardanega, T., Montuori, C., & Lavanga, M. (2019). Coding in primary grades boosts children’s executive functions. Frontiers in Psychology, 10, 2713. Cluster-randomised controlled trial, 76 first graders (Study 1), replicated with second graders (Study 2). University of Padova, Italy. https://pmc.ncbi.nlm.nih.gov/articles/PMC6917597/

Liu, Y.-F., & Bedny, M. (2025). Learning to program “recycles” preexisting frontoparietal population codes of logical algorithms. Journal of Neuroscience. fMRI study, 22 university students, pre- and post-semester Python course. Johns Hopkins University. DOI: 10.1523/JNEUROSCI.0314-25.2025

Cortés Pascual, A., Moyano Muñoz, N., & Quílez Robres, A. (2019). The relationship between executive functions and academic performance in primary education: review and meta-analysis. Frontiers in Psychology, 10, 1582. Meta-analysis of 21 samples, N = 7,947. University of Zaragoza. https://pmc.ncbi.nlm.nih.gov/articles/PMC6638196/