Skip to content
AutoMark AutoMark
Log in Join the waiting list
Explore AutoMark

Evidence

Handwriting earns its place, within limits

The popular case for handwriting starts with brain scans and ends with pedagogy, and it fails at the first step. The case that survives a sceptical reader is narrower, mostly Australian, and still enough to act on. This is that case, along with three claims we'd suggest you stop repeating.

By AutoMark · Published

What holds up

Start with the study closest to home. In Valcan and colleagues' 2024 study of Year 2 children in Perth public schools, students wrote nearly twice as many words by hand as on a keyboard in the same ten minutes, and their handwritten stories were rated higher for quality.

Government research in New South Wales points the same way. The NSW Centre for Education Statistics and Evaluation (CESE, 2021) tested students in Years 2 to 5 in both modes and found they scored lower on the computer-based writing test than on paper. On that evidence CESE recommended that Year 3 NAPLAN writing stay on paper, and Year 3 students still sit the writing test on paper while the older years sit it online.

The pattern isn't a local quirk. A meta-analysis of international studies from Foundation to Year 6 by the same Australian research group (Malpique and colleagues, 2024) found that primary students produce better-quality writing by hand than by keyboard.

Teaching handwriting works, too. A meta-analysis of handwriting instruction across the school years (Santangelo and Graham, 2016) found that explicit teaching improves both legibility and fluency. The Australian Education Research Organisation (AERO, 2022) recommends that students from Foundation to mid-primary get about fifteen minutes of daily instruction, practice and feedback in handwriting.

For children who can't yet read, forming letters by hand helps them recognise letters better than typing them does, in short training studies, according to a meta-analysis by Araújo, Domingues and Fernandes (2022). That finding is about recognising letter shapes, not about learning to read.

Paper also has a small edge when students read informational text. Two meta-analyses (Delgado and colleagues, 2018; Clinton, 2019) found that students comprehend informational text slightly better on paper than on screen. Both findings concern informational text, not stories.

What doesn't

Three claims turn up in nearly every staffroom conversation about handwriting. We'd rather you didn't quote them in a staff meeting, because someone will check.

"The pen is mightier than the keyboard." Mueller and Oppenheimer's 2014 study reported that university students who took lecture notes by hand understood the material better than students on laptops. It is one of the most-quoted studies in the field. The paper later needed a published correction, and a preregistered direct replication (Urry and colleagues, 2021) found no handwriting advantage. The largest recent meta-analysis (Flanigan and colleagues, 2024) does find a small test advantage for handwritten notes, but only in university students, and it is contradicted by Urry's analysis and by Morehead, Dunlosky and Rawson (2019), which found none.

"Handwriting builds brain connectivity for learning." Van der Weel and van der Meer's 2024 EEG study travelled through the world's news sites as proof that handwriting helps children learn. It recorded the brain activity of university students writing familiar words with a digital pen on a touchscreen, or typing them with one finger, and it measured no learning at all. A published commentary (Pinet and Longcamp, 2025) points out that the protocol included no learning and that stretching it to children in classrooms is hard to justify. Different brain activity is not better learning.

"AI is building cognitive debt." "Your Brain on ChatGPT" (Kosmyna and colleagues, 2025), the source of the cognitive-debt headlines, is a preprint: a small study of adults that has not been through peer review, and whose method and reporting have themselves been criticised. It may turn out to be right. Today it isn't evidence.

None of the claims in the section above depends on any of these three.

It's fluency, not the pen

The handwriting advantage is an advantage of fluency. Young children write better by hand because the pen is the tool they are fluent with, not because a pen does something a keyboard can't. In the UK, Connelly, Gee and Walsh (2007) found that primary students' typed scripts lagged behind their handwritten ones, and the authors put that down to missing typing instruction rather than to the keyboard itself.

That gap is not fixed: it narrows through upper primary. In a Dutch study (van Weerdenburg, Tesselhof and van der Meijden, 2019), teaching touch-typing to upper-primary students improved their typed stories and their spelling.

That is why AERO's 2022 review recommends teaching both: instruction in typing and frequent chances to compose on a device, alongside handwriting instruction and handwritten composition. AutoMark sits on the "teach both" side, not the "ban screens" side. It doesn't argue against devices. It gives the teacher a checked first pass on a handwritten test, so the paper can stay on paper.

The argument that applies to tests, including in secondary

Most of the writing evidence above comes from Foundation to Year 5. If you teach secondary, the composition research isn't really your case. This is.

Unsupervised written work no longer shows what an individual student can do, because three recent studies found that neither markers nor software can reliably tell when a model wrote it. In a real university examination system, Scarfe and colleagues (2024) found that AI-written answers submitted for marking went largely undetected. Fleckenstein and colleagues (2024) found that teachers could not reliably tell AI-generated student essays from student-written ones. A test of detection tools by Weber-Wulff and colleagues (2023) found them neither accurate nor reliable.

How students use AI while they practise matters as well. In a large trial with high-school maths students, Bastani and colleagues (2025) found that unguided access to a chatbot during practice lowered students' scores on the exam they later sat without it, while a version built as a tutor, with guardrails, did not.

So a supervised handwritten task is the cheapest evidence there is of what a student can do unaided. There's no device to lock down, no detector to argue with, and no log to audit. The paper in front of you is the student's own work.

Maths has a reason of its own. Two studies of algebra tutoring, Anthony, Yang and Koedinger (2007) and de Morais and Jaques (2022), found that students write equations and working faster by hand than by keyboard. Neither found that it taught more. The case is speed, and keeping the working on the page where you can mark it.

Secondary is also where AITSL's workforce data put the heaviest marking load, which brings us to the problem.

The catch is the marking

Keeping work on paper means someone reads every page. In the OECD's TALIS 2024 survey, reported for Australia by ACER, too much marking was the second most-cited source of stress for Australian lower-secondary teachers: about half of them named it, and they reported several hours a week spent marking and correcting.

National workforce data show where the load falls. AITSL's 2025 reporting of its 2023 survey found that full-time secondary classroom teachers spend more hours marking than primary teachers do, and that roughly a quarter of full-time secondary teachers mark for ten or more hours a week.

Here is the part nobody can yet prove. It is plausible that marking load is part of what pushes teachers toward typed and auto-marked formats: a multiple-choice quiz on a screen marks itself, and a stack of handwritten papers doesn't. Nobody has measured that in Australia. We're stating it as a hypothesis, because that is what it is.

Where AutoMark sits

AutoMark marks the structured parts of a handwritten paper: short written answers, numbers, multiple choice, true or false, fill-the-blanks, tables cell by cell, matching and ordering. You build the test and its answer key, print the papers, and scan the completed class set. AutoMark reads each answer, marks it against your key, checks every reading against the scan, and flags anything it can't verify with the student's handwriting beside it. You confirm every mark before it's recorded.

It does not mark essays today, and the strongest handwriting evidence above is about exactly that kind of writing. Set the long-answer question aside as hand-marked and the rest of the paper still marks, so the extended response gets your full attention and AutoMark gives you a checked first pass on the rest. Essay marking is on our roadmap, and when it arrives it will sit behind the same rule: a proposal you confirm, never a mark recorded on its own.

We don't quote a time saving. Nobody, including us, has measured marking time or accuracy on Australian school-age handwriting, and we won't invent a number to fill the gap. More on how the reading works is in marking handwritten tests, and the difference between this and essay grading is in automated marking vs AI grading.

Sources

In the order they appear above.

If you keep one thing: students in the primary years write more and better by hand, a supervised handwritten task shows what a student can do unaided, and marking is one of the most-cited stressors for Australian teachers. AutoMark gives you a checked first pass on the structured parts of a handwritten paper, so the paper can stay on paper.

Keep the paper. Get a checked first pass.

The beta is full for now — leave your address and we'll tell you when a place opens. Free for two months when it does.

Join the waiting list