Hand, keyboard, AI: what neuroscience actually shows in 2024-2025

“Should we go back to handwriting?” It’s the question that keeps coming up at every conference on writing and cognition, in every mainstream article on the neuroscience of learning. The honest answer is neither “yes, drop the keyboard” nor “no, that’s folklore.” It’s more precise — and more useful.

Three reference studies now allow us to rank hand, keyboard, and AI by the brain activity they produce. This article synthesizes them, without dogma, and pulls the practical lessons for anyone who writes long-form at the keyboard.

What Van der Meer 2024 shows

The most cited study since its publication is by Audrey van der Meer and Ruud van der Weel at the Norwegian University of Science and Technology (NTNU), published in Frontiers in Psychology in January 2024: “Handwriting but not typewriting leads to widespread brain connectivity”.

The setup: 36 university students, high-density EEG (256 channels), two conditions — handwriting with a digital stylus on a tablet, typing the same words on a keyboard. Functional brain connectivity measured across several frequency bands (theta, alpha, beta, gamma).

The results are clear:

• In the handwriting condition, functional brain connectivity rises significantly in parietal and central regions, in theta and alpha bands — bands associated with learning and memory.
• In the typing condition, the same connectivity doesn’t appear, or appears at a much lower level.
• The effect is robust, statistically clean, and consistent with an older literature on motor memory.

The authors conclude — cautiously — that handwriting engages brain networks linked to learning that typing does not, and that this difference may explain the better conceptual memorization observed in handwritten note-takers.

A caveat: this study doesn’t measure long-form creative writing. It measures the writing of displayed words. The authors themselves flag the limit. Transposing it to long-form writing is a reasonable hypothesis, not a direct result.

The keyboard in the hierarchy: Mueller & Oppenheimer 2014, and what followed

The historical study on keyboard vs hand for note-taking is by Pam Mueller and Daniel Oppenheimer, published in 2014 in Psychological Science under the title “The Pen Is Mightier Than the Keyboard”.

Three experiments, more than 300 participants. Handwritten note-takers performed better on conceptual questions than keyboard note-takers, at comparable amounts of information retained. The authors’ explanation: handwriting forces synthesis and reformulation because it is slower. Typing allows verbatim transcription — less cognitively expensive, therefore less formative.

This result has been partially replicated and debated. The most cited replication is Morehead, Dunlosky & Rawson 2019, Educational Psychology Review, which confirms a note-taking mode effect with a more modest effect size. A 2022 meta-analysis by Voyer et al. finds a small but significant handwritten advantage on conceptual questions, null on factual questions.

Honest bottom line: the handwritten advantage exists, it is modest in recent studies, and depends on the cognitive task. It is neither null nor overwhelming.

AI at the bottom: Kosmyna MIT 2025

The most recent and direct study on AI-assisted writing is by Nataliya Kosmyna and colleagues at the MIT Media Lab, published in June 2025: “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing”.

54 participants, three groups — writing unaided, with a search engine, with ChatGPT — under EEG. I detailed this study in the article on AI-driven mental overload. The relevant findings here:

• The ChatGPT group shows the lowest brain connectivity during writing, in the same bands Van der Meer 2024 finds high in handwriting.
• 83 % of ChatGPT users cannot quote a sentence they had just “written.”
• The effect persists: when the ChatGPT group is later asked to write without an assistant, their neural engagement remains below the other two groups.

Stacking Van der Meer 2024, Mueller-Oppenheimer 2014 (and its replications), and Kosmyna 2025, we get a coherent hierarchy:

Handwriting > keyboard typing > AI-assisted writing

This isn’t a slogan. It’s three EEG or behavioral studies, on different setups, pointing in the same direction.

Why handwriting does better: the motor-trace theory

The mechanism proposed by neuroscientists is the motor trace. Karin James, at Indiana University, documented this in children as early as 2012 (“The Effects of Handwriting Experience on Functional Brain Development in Pre-Literate Children,” Trends in Neuroscience and Education): learning letters by handwriting them, rather than typing them or seeing them, activates a brain network that includes motor, visual, and memory areas simultaneously.

The central idea: writing by hand means producing a unique gesture for each letter. The brain encodes the shape via the gesture. At the keyboard, each key produces the same gesture, regardless of the letter. Multisensory integration is impoverished.

Anne Mangen and Jean-Luc Velay (2010, Advances in Haptics) had already formalized this: the hand is not a peripheral detail — it’s a cognitive partner to the brain in writing.

The implication: the more a tool flattens the gesture, the poorer the motor trace. Keyboard beats AI because at the keyboard the typing gesture is still tied to producing the sentence. With an assistant, you don’t even type anymore — you read, validate, re-read. The motor trace fades.

How to approach handwriting’s benefits at the keyboard

The practical part. If you write at length on a keyboard — out of necessity, habit, or flow — here are the levers that bring the experience closer to handwriting, without giving up the keyboard.

1. Slow the typing pace

Mueller & Oppenheimer already pointed to forced slowness as the engine of the handwritten advantage. At the keyboard, that’s doable: type at 60-80 % of your top speed. The brain gets time to reformulate, synthesize, choose the right word.

A simple trick: write on a mechanical keyboard with a slightly higher actuation force than average (50-60 g, see the article on writing posture). The brake is physical, so it’s easier to honor than a mental rule.

2. Kill autocomplete and aggressive autocorrect

Autocomplete splits the sentence in two: half comes from you, half from a model. The cognitive trace is shared with the tool. That’s exactly what Kosmyna 2025 measures under EEG: less engagement, less memory, less ownership.

Concretely: disable active suggestions, disable aggressive autocorrect, keep only the red underline for blatant typos. The software’s silence makes room for your voice.

3. Mark the end of every sentence with a pause

A mechanism close to breath: finish the sentence, lift your hands off the keyboard for a second, re-read in silence, then start the next one. It imposes a rhythm close to handwriting (~20-30 words per minute in reflective production) and naturally slows the typing.

4. One document, no tabs

Handwriting happens, by construction, in a single space: the page. At the keyboard, we juggle the editor, a browser, notes, a dictionary. Every switch costs ~23 minutes of recovered depth (Mark et al. 2008, CHI, already cited in the ADHD article).

A full-screen environment with no visible browser, no messaging, no anything — that’s exactly the logic of a silent editor like Draft_. Not magical: just, by construction, closer to the page.

What I’m not telling you

A few honest nuances:

• The handwritten advantage is modest on short tasks, more pronounced on long-term learning. For multi-hour creative writing, we still lack direct comparative studies.
• The kind of task matters. For note-taking in class, the handwritten advantage is documented. For novel writing, the data is anecdotal.
• Age and habit change things. An adult who has typed for twenty years won’t have exactly the same neural profile as a child learning to write.

The goal here isn’t to swap one tool for another. It’s to know where each tool sits in the cognitive hierarchy — so you can make conscious choices.

In short

• Van der Meer 2024 (EEG, NTNU): handwriting produces higher brain connectivity than typing, in bands linked to learning.
• Mueller & Oppenheimer 2014 (and 2019-2022 replications): handwritten note-takers memorize concepts better; modest but real effect.
• Kosmyna MIT 2025: AI-assisted writing produces the lowest connectivity of the three modes — the “cognitive debt.”
• Hierarchy: hand > keyboard > AI. Documented, not ideological.
• At the keyboard: slow down, kill autocomplete, pause between sentences, one document. That brings you functionally closer to handwriting.

Draft_ is built on that logic: one document, no autocomplete, no suggestions. The keyboard stays a keyboard, not a half-assistant. The typing gesture becomes a cognitive partner again, not just an input method.