The Short Answer: Is Handwriting or Typing Better for Notes?
Handwriting has a small, real advantage over typing for academic achievement: about a quarter of a standard deviation in the largest analysis of handwritten versus typed notes ever run on college students. That's a fraction of the effect the popular version of this story promises, and the original experiment behind the popular version failed to reproduce when researchers ran it again with tighter controls.
The advantage that does exist has nothing to do with your hand. It comes from what a pen forces you to do. Writing by hand is slow enough that you can't record everything, so you have to choose, and choosing is the part that teaches you. A keyboard removes that constraint, which is a gift when you're drafting and a trap when you're learning.
So the practical answer is not "use a pen." Whichever tool you pick, make yourself select, and make yourself come back. A typist who writes 60 words in their own phrasing and reviews them Thursday will beat a handwriter who fills four pages and never opens the notebook again.
The 2014 Study That Emptied the Lecture Hall
In 2014, Pam Mueller and Daniel Oppenheimer published "The Pen Is Mightier Than the Keyboard: Advantages of Longhand Over Laptop Note Taking" in Psychological Science. It became one of the most widely publicized education papers of the decade and one of the most quoted in faculty meetings.
The setup was clean. In the first experiment, 65 college students watched TED Talks and took notes either longhand or on laptops, then answered factual and conceptual questions. Laptop users produced more words and more text that overlapped the lecture verbatim. They also did worse on the conceptual questions, the ones that ask you to reason with the material rather than repeat it.
Mueller and Oppenheimer had a mechanism, and it's a good one. Typing is fast enough to let you act as a stenographer. Transcription is shallow processing: the words pass from ear to fingers without stopping anywhere useful in between. Handwriting is too slow for that, so handwriters had to summarize, and summarizing is comprehension work.
Two follow-ups made the paper famous. In the second experiment, the researchers explicitly told laptop users not to take verbatim notes. It didn't help. The instruction failed to reduce how much people transcribed, which suggested the pull toward stenography is close to automatic once your hands are on a keyboard. In the third, half the students got ten minutes to study their notes before a test a week later. There was no main effect of medium on its own, but longhand plus study beat every other combination, and among students who could review, verbatim overlap again predicted worse conceptual scores.
That's a tidy story with a clear villain, and it landed on a movement already underway. Individual professors had been banning laptops since the early 2000s, and Mueller and Oppenheimer's own opening paragraph cites a 2007 law-review article titled "Banning laptops in the classroom." What the 2014 paper supplied was the citation that movement had been missing.
The 2019 Replication That Found Almost Nothing
Five years later, Kayla Morehead, John Dunlosky, and Katherine Rawson ran a direct replication and published it in Educational Psychology Review. Dunlosky is not a critic on the fringe of this field; he's the lead author of the 2013 review that sorted ten study techniques into high, moderate, and low utility, and one of the most careful people working on learning science.
They kept the design and added two groups the original didn't have. One group took notes on an eWriter, a digital tablet you write on with a stylus, which separates "writing by hand" from "writing on paper." The other group took no notes at all.
The results went sideways, though not in the direction you'd expect. In the first experiment, longhand note-takers came out ahead on factual questions, at d = 0.45. That's an advantage the original study never found, because Mueller and Oppenheimer's factual scores were level. On conceptual questions, the ones where the original found its headline effect, there was no reliable difference between conditions. On a test two days later, note-taking method had no effect on performance at all. Pooling the direct replications produced small effects favoring longhand that didn't reach significance.
And then there's the no-notes group. Those participants didn't perform worse than any of the note-taking groups. Sit with that for a second, because it reframes the whole argument. If people who wrote nothing scored like people who wrote something, the experiment can't tell you much about how to write. A comparison this coarse isn't sensitive enough to rank two note-taking media.
What did predict performance was mundane and much more useful: the overlap between what a student's notes contained and what the test asked. Notes that happened to cover the tested material produced better scores. The content of the notes mattered. The instrument that produced them barely registered.
What 24 Studies Say About Handwriting vs Typing Notes
A single replication doesn't settle a literature either. In July 2024, Abraham Flanigan and colleagues published a meta-analysis in Educational Psychology Review covering 24 studies drawn from 21 articles on typed versus handwritten lecture notes, with a pooled sample of about 3,000 college students.
Two numbers came out of it, and the contrast between them is the most informative thing in this entire debate.
Handwriting won on achievement, at Hedges' g = 0.248 overall (p < .001, 95% CI 0.181 to 0.315). By convention that's a small effect. It's real and it's statistically solid across two dozen studies, but it is not the dramatic gap the 2014 headline implies. One moderator moved it: students given a formal review period showed g = 0.421, against g = 0.208 for students who weren't. Review roughly doubles the benefit, and hold onto that, because it matters more than the medium does.
Typing won on volume, at g = 0.919 (p < .001) across the 15 effect sizes that measured it. That's a large effect. Typists record far more material than handwriters do, and it isn't close.
| Source | Year | What it measured | Result |
|---|---|---|---|
| Mueller & Oppenheimer | 2014 | Conceptual and factual recall, 3 experiments | Longhand ahead on conceptual items; verbatim overlap predicted worse scores |
| Morehead, Dunlosky & Rawson | 2019 | Direct replication plus eWriter and no-notes groups | No reliable conceptual difference; longhand ahead on factual, which is not the original pattern; no-notes group performed comparably |
| Flanigan et al. | 2024 | Meta-analysis, 24 studies, achievement | Handwriting ahead, g = 0.248 (small); g = 0.421 with review |
| Flanigan et al. | 2024 | Meta-analysis, 15 effect sizes, note volume | Typing ahead, g = 0.919 (large) |
| Van der Weel & Van der Meer | 2024 | High-density EEG during writing, 36 analysed | Condition difference in theta-alpha connectivity; no learning outcome measured |
Read the table as one sentence and it says this: the medium has a huge effect on how much you write down, and a small effect on what you learn. The small effect is almost certainly a downstream consequence of the huge one.
The Brain Scan That Went Viral, and the Commentary Under It
In January 2024, Ruud van der Weel and Audrey van der Meer published an EEG study in Frontiers in Psychology that has since been picked up by more than 290 news outlets. Forty university students wrote words with a digital pen and typed the same words on a keyboard while researchers recorded brain activity through a 256-channel sensor array; 36 datasets survived artifact rejection.
Handwriting produced widespread theta-alpha coherence across left, midline, and right parietal hubs and central cortex. Typing showed markedly less. The paper reported 32 significant cluster differences representing 16 distinct neural connections. The coverage wrote itself: science proves writing by hand rewires your brain for learning.
Then, in January 2025, Pinet and Longcamp published a commentary in the same journal raising five problems with that reading.
First, the study measured brain activity during writing. It never measured learning or memory retrieval, so it can't tell you whether any of that connectivity produced a durable trace. Second, the typing condition had participants use a single right index finger, which is not how anyone types and which turns a skilled bimanual task into an unfamiliar pointing task. Third, visual feedback differed systematically between the two conditions. Fourth, typing skill was neither assessed nor controlled. Fifth, and worst for the headline, the statistics were run only on the difference between the two conditions, never on either condition alone, which the commentary says undercuts the paper's own title.
None of this makes the EEG result fake. Handwriting and typing genuinely are different perceptual and motor acts, and it's unsurprising that they light up differently. What the study can't support is the inference everyone drew from it. A difference in neural coherence while you form letters is not evidence that you'll remember the content better next Tuesday. For that you need a memory test, and the studies that ran memory tests found g = 0.248.
This is the same pattern that shows up in the illusion of competence: a vivid signal gets treated as proof of learning, when learning is a separate thing that has to be measured separately.
What Actually Separates Handwriting From Typing: Selection
Strip out the noise and one finding appears in every study, in the same direction, at a size nobody argues about. Typing produces more words. Mueller and Oppenheimer found it. Flanigan's meta-analysis put a large effect size on it. It reproduces because it's mechanical: your fingers are faster than your hand, so the bottleneck moves.
And the bottleneck was doing the work.
When you write by hand at a lecture's pace, you physically cannot capture the whole thing. Every few seconds you make a decision: is this worth the ink? That decision requires you to hold the idea, compare it against what came before, judge its importance, and compress it into fewer words. Four cognitive operations, forced on you by the slowness of your wrist. Psychologists call this class of constraint a desirable difficulty, an obstacle that makes performance worse right now and learning better later.
The evidence for selection as the mechanism, rather than the pen as the mechanism, is scattered across all three papers. Mueller and Oppenheimer found verbatim overlap predicted worse conceptual performance even with note-taking medium held constant in the model. Morehead and colleagues found the overlap between note content and test content predicted scores. Flanigan's meta-analysis shows typing's advantage is volume, precisely the thing that correlates with doing worse.
Which means the useful question isn't "pen or keyboard." It's "did anything get left out?" If the answer is no, you were a transcription device, and it doesn't matter what you were holding.
Highlighting sits in the lowest tier of Dunlosky's rankings, and the narrow reason it isn't worthless is this same one. A highlighter can't record anything. It can only mark what's already there, so the only operation it supports is choosing. When you highlight three sentences in a 2,000-word essay, the value isn't the three sentences, it's the 1,900 words you decided not to mark. That's the same work a pen imposes on a note-taker, applied to reading.
The Variable Nobody Puts in the Headline: Review
Look again at the one moderator that moved Flanigan's achievement result. Students who got a formal review period showed g = 0.421. Students who didn't showed g = 0.208. Review roughly doubles the benefit, and it was the moderator that mattered: question type and assessment timing both came back null.
That matters because note-taking does two different jobs and most people only think about one. There's the encoding function, which is the learning that happens while you write. And there's the external storage function, which is the learning that happens later when you come back. Handwriting has a small edge on the first and, going by Flanigan's review subgroup, a slightly larger one on the second. What paper loses is neither of those. It's access: a notebook isn't searchable, doesn't sync, and can't be pulled up in the six minutes you have before an exam.
The hard truth is that most notes of either kind are never reopened. A notebook that closes at the end of term and a document that gets buried in a downloads folder fail for the same reason. The encoding benefit is real but modest, and it's the only benefit you collect if review never happens.
Retrieval is what converts notes into knowledge, and it's the best-evidenced finding in the whole field. Active recall beats rereading at about g = 0.61 in Adesope and colleagues' 2017 meta-analysis, roughly two and a half times the handwriting effect. So if you're optimizing, the order is obvious: get review working first, then argue about the pen.
A digital workflow stops being a compromise here and starts being the better option. Glasp's web highlighter keeps every passage you marked with its source URL attached, so a highlight from March is one search away instead of somewhere in a stack of notebooks. Your Kindle highlights land in the same place. You can ask Glasp's AI chat a question and have it answer from what you actually marked, which is a review session that starts in two seconds instead of twenty minutes.
How to Type Notes That Don't Fail You
If you type, you're not doomed. You're just unprotected. The pen came with a built-in constraint and a keyboard doesn't, so you have to supply one. These are the constraints worth imposing, in rough order of how much they help.
Set a budget and hold it. Decide before the lecture that you'll write no more than a screen. A hard cap does the same job slowness used to do: it forces triage. One screen per hour is a reasonable starting cap. Whatever number you settle on, the enforcing is where the learning happens, not the number.
Ban the speaker's exact words. Make it a rule that nothing goes in the document in the phrasing you heard it. Paraphrasing is the entire mechanism, and it's the one thing Mueller and Oppenheimer couldn't get their participants to do on request. You'll do better than their subjects because you're choosing it rather than being told.
Leave the left margin empty for questions. The Cornell note-taking method works just as well in a text editor as on paper. The cue column is a retrieval prompt you're writing for your future self, and having somewhere to put questions changes what you notice while you listen.
Close every other tab. A pen can't open Slack. A large share of what gets blamed on the medium in classroom studies is plain multitasking, and that part you really can fix with a keyboard shortcut.
Write one sentence at the end with the source closed. Not a summary of the notes, a summary from memory. It takes 30 seconds and it's a retrieval attempt, which is worth more than the previous hour of typing.
Mark, then write. When the source is a document or a video rather than a live lecture, highlight first and compose second. Highlighting is selection without the temptation to transcribe, because there's nothing to transcribe into. Then turn the marked passages into a few sentences of your own. For video, YouTube Summary gives you a transcript you can highlight the same way you'd highlight an article, so the same selection discipline transfers to a lecture recording.
Handwriting or Typing: Choose by Situation, Not by Principle
The medium question has a different answer depending on what's in front of you. Treating it as a principle is what produces laptop bans in statistics courses where students need to type LaTeX.
| Situation | Better choice | Why |
|---|---|---|
| Fast lecture, conceptual material | Handwriting | The speed limit forces selection when you have no time to decide deliberately |
| Equations, diagrams, chemical structures | Handwriting or stylus | Spatial content is far slower to type than to draw |
| Technical talk with code or exact terminology | Typing | Precision matters more than compression, and you'll need to search it later |
| Reading articles and papers | Highlight, then write | Selection without a transcription option, and the source stays attached |
| Long-term reference you'll search in a year | Typing or digital highlights | Handwriting loses on retrieval, and retrieval is most of the value |
| Video lectures and recorded talks | Transcript plus highlighting | You control the pace, so slowness is no longer doing useful work for you |
| Books you read on a device | Digital highlights, exported | The alternative is highlights that stay locked in the reading app |
One pattern runs through the right-hand column. Handwriting helps in exactly the conditions where you're forced to keep up in real time and can't pause. The moment you can pause, rewind, or reread, the pen's advantage evaporates, because slowness was only ever a proxy for deliberate choice. When you can be deliberate directly, be deliberate directly.
There's a social dimension too, and it's the one thing neither medium has historically offered. Notes in a notebook are read by one person. When highlights are public, as they are in Glasp's community, you can see which passages other readers marked in the same article and find out whether you selected well. That feedback loop doesn't exist for private notes of either kind. Every note-taking method is optimizing for something, and the choice gets easier once you know what.
Frequently Asked Questions
Is it better to take notes by hand or on a laptop?
By hand, slightly, and by less than you've been told. The largest meta-analysis of college students, published in 2024, found handwriting ahead on achievement at Hedges' g = 0.248, a small effect, rising to 0.421 for students who got to review their notes first. The famous 2014 study that reported a much bigger gap did not survive direct replication in 2019. If you type but you paraphrase and review, you're in better shape than someone who handwrites verbatim and never looks back.
Was the "pen is mightier than the keyboard" study ever replicated?
No. Morehead, Dunlosky, and Rawson ran a direct replication in 2019 and found the longhand advantage had moved to the factual questions, with no reliable difference on the conceptual ones where the original effect lived. Two days later, note-taking method made no difference at all. A control group that took no notes performed comparably to all three note-taking groups, which suggests the paradigm isn't sensitive enough to rank media.
Does writing by hand improve memory?
The 2024 EEG study by Van der Weel and Van der Meer found broader brain connectivity during handwriting than during typing, and it was widely reported as proof that handwriting improves memory. It doesn't show that, because the study measured brain activity while participants wrote and never tested recall. A 2025 commentary by Pinet and Longcamp also flagged that participants typed with a single index finger, which is not a fair comparison condition. The studies that did test memory found small effects.
Is taking notes on an iPad as good as writing on paper?
Probably, and that's a useful clue. The 2019 replication included an eWriter group specifically to separate handwriting from paper, and it didn't behave differently from longhand on paper in any meaningful way. That supports the idea that the effect comes from the speed constraint of forming letters rather than from anything about paper itself.
How many notes should I take during a lecture?
Fewer than feels comfortable. There's no magic number, but the mechanism says your notes should be a small fraction of the source. If a transcript would run 6,000 words and your notes run 2,000, you were transcribing. If they run 200 and each line is in your own phrasing, you were thinking. Reducing volume is the goal, not a side effect.
Does highlighting count as note-taking?
It counts as the selection half, and it skips the paraphrasing half. Highlighting alone is one of the weaker study techniques: Dunlosky's 2013 review rated it low utility, and it was the only one of the ten techniques to earn an outright negative rating on criterion tasks. Highlighting followed by writing a few sentences in your own words is a different activity, and it's close to what handwriting forces on lecture note-takers. The marks give you the selection, and the sentences give you the encoding.
The Pen Was Never the Point
A single 2014 experiment with 65 students in its first study became the standard citation for a classroom policy that was already spreading. The direct replication came back nearly empty, the meta-analysis came back small, and the neuroscience everyone shared didn't measure learning. That's not a scandal, it's ordinary science working slowly in public, but it does mean the confident version of this advice was never warranted.
What's left after the correction is more useful than the myth was. Two things predict whether notes teach you anything, and neither is the instrument. You have to leave things out, and you have to come back. Handwriting helps with the first by accident, because your hand is slow. It hurts the second, because paper can't be searched.
Pick the tool that fits the situation, then add the constraint the tool is missing. If you type, cap your volume and paraphrase. If you write by hand, build a review habit, because your encoding advantage expires the moment the notebook closes.
If most of your reading happens on a screen, the two things to fix are making selection deliberate and making review cheap. Glasp's web highlighter keeps what you marked with its source attached, across articles, PDFs, and Kindle books, so the coming back part stops depending on discipline you may not have on a Thursday night. Mark less than you want to. Then go look at it again.
