Reading a claim about your notes
95 min
Two hosts talk the lesson through. The voices are synthetic; the script was written from this lesson and checked against it, and asserts nothing the lesson does not.
- Sort a claim about note-taking into measurable, unmeasurable as stated, or false, and say what would settle it
- Identify what a note-taking claim was measured on, and say what a college-lecture sample licenses
- Decide what to do about your own note-taking, with what each option costs
Five meta-analyses will become six. Somebody will run the EEG study again with two hands on the keyboard, and somebody else will publish the commentary on that. Within a few years the numbers in lessons 3 to 6 will be the wrong numbers.
None of that touches what this lesson is for. The seven lessons behind it were also, quietly, a training in one operation: taking a sentence somebody says about notes and working out what would have had to happen for it to be true. That operation outlives its examples, and this lesson is where it gets written down.
The sort you already have
The institute's standing three-way sort is used here rather than taught again. Digital Literacy lesson 8 introduced it, Using AI Effectively lesson 11 sharpened it, and Time Management lesson 8, Habits and Self-Discipline lesson 7, Focus and Deep Work lesson 7 and Memory lesson 7 have each put it to a different subject.
Measurable: a study could check it, and the sentence says enough for you to imagine the study. Unmeasurable as stated: it turns on a word nobody has defined, so no result could disagree with it. False: somebody checked and it didn't hold.
The four claims worked below were written for this lesson rather than quoted from anywhere, and the sorting of each is this course's reasoning rather than a finding.[7]
Work one. "Handwriting beats typing for lecture notes." Measurable, plainly, and measured more often than almost anything else in this subject. Lesson 5's table has five separate pooled answers to it running from +0.008 to +0.250, three of which cleared zero.5 So the sort is easy and it doesn't finish the job. A claim can be measurable, measured five times over, and still unsettled, and the sort's only promise is that the disagreement is about evidence rather than about words.
Work two. "A good note-taking system makes you think more clearly." Unmeasurable as stated, and worth being precise about why. It isn't that the claim is silly. It's that nothing in the sentence says what would count as thinking more clearly, or which systems are good ones, so there's no result that could come back and contradict it. A sentence nothing could falsify isn't a finding, whatever else it is.
Notice what would fix it. Name the system, name an outcome somebody could score, and name the comparison, and you have a study. The repair is usually available and usually not made, because the vaguer sentence is the one that sells.
Work three. "Typing is worse because you write too much." False, in the specific sense that the reason given has been checked and points the other way. Higher word count was associated with better quiz performance in the 2014 study and again in its direct replication.4 Typists do write more, which reproduces reliably. The bit welded onto it, that writing more is what hurts them, is the bit the data contradicts.
Work four, and this is the shape you'll meet most. "Don't type your notes, because handwriting activates more of the brain." Two claims stuck together, and they want separate verdicts.
The second half is lesson 6's EEG study, which found more elaborate connectivity patterns while writing single words by hand than while typing them, on 36 university students.[6] That half is a measurement.
The first half is advice about your working life, and nothing in the study is about your working life or about learning at all, since nobody in it learned anything that was then tested. A reader who accepts both halves together, or rejects both together, has done neither.
Before you read on. Take a claim you have actually been told about note-taking, in whatever words the person used. Sort it, and then say what would settle it.
Show the answer
The second half is the part worth the effort, and it's the part people skip.
Sorting a sentence takes a few seconds and feels like progress. Saying what would settle it forces you to name an outcome, a comparison and a population, and most claims about note-taking fall apart at the third one. "Would handwriting beat typing?" is answerable. "For whom, doing what, tested when?" is where the five meta-analyses started disagreeing with each other.
If the claim you picked turned out to be unmeasurable as stated, try the repair from work two before you move on. It's a better habit than the sort itself.
The question this course adds: where was it measured
Each course on this path adds one question of its own to the sort. Memory lesson 7 added the sample: how many people, and who. Focus and Deep Work lesson 7 added the instrument. This course's addition is the setting, and it's here because of something Stage 1 turned up about this subject in particular.[7]
Every measurement of note-taking behind these eight lessons happened in a lecture hall, to a college student, with the outcome taken inside the hour or inside the week. The one exception is the EEG study, which is a laboratory with no lecture and no outcome at all, and that is a different gap: lesson 6's rather than this one's.
The lecture is not a coincidence and it isn't a criticism of the literature either. A lecture is a controllable event with a natural comparison group sitting in it, which is exactly what an experiment wants.
Now hold that against the claims you actually meet. They're about somebody in a meeting, somebody reading a paper at their desk, somebody keeping a notebook for eleven years. None of those is a lecture, and none of them is tested on Friday.
So the setting question is three questions, and they take about ten seconds.
Where did the measurement happen? A lecture hall, a laboratory, a workplace, somebody's own week.
How long after? Within the hour, within the week, or long enough to matter to you.
And what was the task? Listening to a lecture is not reading, and neither is sitting in a meeting where you are also expected to speak.
A claim that survives all three reaches your case. A claim that fails one of them may still be true of you; what it isn't is evidenced for you, and the difference between those two is this course's whole subject.
Before the table. Six sources. Guess which one licenses the widest claim, and which one has been cited hardest for a claim it does not license.
Show the answer
They aren't the same source, and the second one is easier to guess than the first.
The widest licence belongs to the 2024 meta-analysis, because pooling 24 studies across 21 articles[5] is what buys you a statement about college students in general rather than about the participants in any one room. Even then the statement it buys is "among college students", which the abstract says three times and scopes a fourth.
The most heavily worked source is the EEG study, which measured 36 people writing single words and was widely reported as evidence about children and classrooms, its own title inviting that.[6] Its own last sentence is a recommendation about children, which lesson 6 gives in the authors' words, and no child was in it.
That pairing is the lesson. The widest licence and the widest circulation attach to different papers, and nothing about a result makes the second follow from the first.
What each source licenses
Lesson 1's table said how far into each source this course got. This one says what each one lets you claim, which is the other half of the same question. The last two columns are this course's reading of the sources rather than anything the sources say about themselves, and they are the part to argue with.[7]
| The source | Measured on | What it licenses | What it does not |
|---|---|---|---|
| The 2024 meta-analysis[5] | 24 studies, 21 articles, college students in lectures | A pooled achievement advantage for handwriting, g = +0.248, and a larger volume advantage for typing | Anybody who is not a college student in a lecture |
| The 2022 meta-analysis[3] | 77 effect sizes from 39 samples, from a search summary | That a pooled analysis of 73 effect sizes returned +0.008, interval −0.16 to 0.18 | Any account of how it was done, since this course read only the figure in another paper's table |
| The direct replication[4] | 74 laptop and 68 longhand participants, one lecture, a quiz after a brief distraction | That the note-taking difference reproduced and the quiz difference did not, on those 142 people | Anything about a delayed test, or about what reviewing the notes would have done |
| The replication and extension[2] | Students, two experiments, including a group taking no notes | That some trends suggested longhand superiority but performance did not consistently differ between groups, and that its authors call the question premature | What the eWriter condition found, which the abstract does not say |
| The famous 2014 study[1] | Students, three studies, conceptual questions | That three studies found a longhand advantage on conceptual questions, with verbatim transcription proposed as the mechanism | Any sample size or effect size, neither of which is in the abstract this course read |
| The EEG study and its commentary[6] | 36 university students writing single words, one finger on the keyboard | That two motor acts are accompanied by different connectivity patterns | Anything at all about learning, which nobody in that study measured |
Read the last column on its own. In four of the six rows, what the source cannot support is close to what a reader will have met it saying , which is this course's reading rather than a count anybody has made.[7]
And read the first row against the third. The 2024 meta-analysis and the 2021 replication are looking at the same literature from two distances: alongside their 142 participants, Urry and colleagues also pooled eight similar studies, and that pooled figure is one of the five rows in lesson 5's table, sitting beside the 2024 paper's own.5 A disagreement between two sources is not always two camps. Sometimes it's one body of work being cut two ways, which is what lesson 5 found and what makes this subject's disagreement an unusually honest one.
Four books this course did not open
The route most readers take into this subject doesn't go through a meta-analysis. It goes through Sönke Ahrens's How to Take Smart Notes, Tiago Forte's Building a Second Brain, Niklas Luhmann's own account of his card index, or Tony Buzan on mind mapping. That this is the usual route is this course's expectation about its readers rather than something anybody counted.[7]
None of the four was opened for this course.[7] The research file names all four as unread and the build takes that seriously: a lesson citing one of them for a claim fails, and this lesson is exempted from that only so it can say their titles. Nothing in these eight lessons rests on any of them, and they are named here because you will be handed one and ought to know what the course can and cannot say about it.
Which is less than you might hope and more than nothing. Stage 1 searched for peer-reviewed work measuring whether any personal knowledge management system improves anything, and found none.[7] That is reported as a finding rather than as a gap in the searching, because the searching was the ordinary kind and this is a large and commercially successful literature.
An absence of tests is not a failed test, which is the distinction lesson 5's two non-significant rows exist to protect and the one this section would defend in either direction. What you have for these four books is the sort, the setting question, and one specific thing from lesson 7: the best known of the systems, the Zettelkasten, rests its public case on the output of a single scholar nobody compared with anybody.
Eight lessons in. Which of the things this course has told you should you hold most firmly, and which most loosely?
Show the answer
Worth answering before you read on, and the order isn't the one the lessons were in.
Sort them the way the lesson has been sorting everything else, by what would have to happen for each to stop being true.
The two questions survive anything. The setting question and lesson 2's denominator question are not findings, so no result can overturn them; a meta-analysis published tomorrow would be read with both and would need both. The table above is the argument for keeping them, since four of its six rows carry a licence narrower than what a reader will have met.
The field's summary of where the benefit sits would take a lot to move. Taking and reviewing notes beats taking and not reviewing them. That reaches you as source 5's summary of its own literature rather than as a result this course checked,5 and it is the one sentence here that would change what you do tomorrow.
Every figure could move by next year, and some of them will. The five pooled effects depend on which studies go into the pool, and lesson 5 showed five honest poolings producing five answers. The 30 percent rests on three cited papers. The 74 and the 68 are one room.
And the mechanism is the piece to bet against. Verbatim transcription as the reason handwriting helps was a proposal in the 2014 abstract; it reproduces as a description of what typists do, and its link to worse performance was not robust in the replication.4 If a version of the effect survives, the story about why is the part likeliest to be different.
The one thing this course will not do
Nothing in these eight lessons will tell you whether your own notes are any good. Not last week's meeting page, not the system you have been building for three years, not the notebook you keep meaning to reread. There is no test in here you can run on one note.
That sounds like modesty and it is closer to arithmetic. Every result this course has is an average over a room of people doing a task somebody chose, and an average over a room says nothing about any one person in it. The thing that can say something about your own practice is the project, and the reason it asks you to log a week rather than read a table is that a table cannot do that job.
So the two overreaches to watch for are mirror images, and both are ordinary.
One reader finishes lesson 4, reads that a no-notes group did no worse, and stops taking notes: that result was measured on students in a lecture, tested soon afterwards.2 Another finishes lesson 5, takes the largest of the five pooled effects, and rebuilds a working life around a pen. Each has taken a number measured on somebody else and spent it on themselves.
Three things people get wrong at this stage
"There must be research on this." On handwriting against typing there is, five poolings of it. On whether a personal knowledge system does anything, none that this course could find.
"The science says handwriting." It says +0.008 to +0.250 depending on which studies you pool, among college students, with all five pointing the same way and two of the five intervals including zero.5
"If it were nonsense somebody would have said so." Publishing an objection takes a journal, a reviewer and a reason; not objecting takes nothing. Silence around a method is the cheapest thing in this subject.
Practice
Take twenty minutes and one claim. The point is the trail rather than the verdict.
Take a specific claim about note-taking that you have seen repeated with a number in it. A percentage, a multiplier, a number of days. Write it down in the words you met it in.
Then work backwards. Who said it? What do they cite? Follow that, and follow what that cites. Stop after twenty minutes whatever happens.
There are six endings and every one of them is a result. Which of the first five you reach more often than the others is not something anybody has counted, so nothing below says.[7]
- You reach a study and it says what the claim says. Rare, and when it happens, check the setting before you celebrate.
- You reach a study and it says something narrower. The usual shape when there is a real paper at the bottom of the trail. Write down what was dropped between the paper and the sentence.
- You reach a study that says something else. Work three above is this shape.
- You reach a person who gives no source of their own. An author, a speaker, a post that cites nothing onward. Not nothing: it tells you the claim's authority is one person's judgement, which is a thing you can weigh.
- The trail closes. It circles back to something you have already followed, or it ends in "studies show". That is the most informative of the six, because it means the number travelled a long way without anybody checking it.
- Twenty minutes ran out. Record where you had got to and how many steps it took. How deep a trail goes before it runs out is itself a fact about the claim.
Write one line saying which ending you got and how many steps it took. That line is the whole exercise.
Allow twenty minutes. This is the last section of the course project, and it can be done on its own if you have not been keeping the log.
Pick one thing about your own note-taking and reach one of three endings. Not a resolution to take better notes, which is unmeasurable as stated and you now have a name for that.
Ending one: change something. Name the change specifically enough that somebody could tell in a month whether you had done it. Then name what it costs: minutes a week, something you'll stop doing, or an idea you'll lose because you're writing instead of listening.
Ending two: keep something deliberately. You type your notes, you've read all eight lessons, and you're going to carry on typing. That's a defensible decision and it's a decision, as long as you write down why and what evidence you're leaning on. All five meta-analyses point the same way and three of them clear zero, which is an argument for switching that you may judge too small to act on, and is not one you can wave off.
Ending three: stop deciding about it. Some things aren't worth a policy. If you take four notes a month and go back to none of them, the review routine from lesson 7 is a solution to a problem you don't have, and saying so is worth more than a habit you'll drop in a fortnight. This ending is the exercise's rather than the project's: the project's section 4 asks for one of the first two, because it marks the cost you named, so if you are writing this as your project section, pick one of those and keep this one for yourself.
Then, for whichever ending you chose, one line on what would tell you it was the wrong call. That line is the difference between a decision and a preference, and it is the thing almost nobody writes.
Connections
Back. This lesson uses all seven before it. Lesson 1 is the two functions and the scope table; lesson 2 is the denominator; lessons 3 to 5 are the medium question and what happened when people checked it; lesson 6 is the step count for an imaging claim; lesson 7 is the review routine and what maintaining a system costs.
Across. Digital Literacy lesson 8 is where the three-way sort was introduced and Using AI Effectively lesson 11 is where it was sharpened. Memory lesson 7 added the sample question and Focus and Deep Work lesson 7 added the instrument question; this lesson adds the setting. Memory lesson 4 and this course's own lesson 3 are where tracing a claim back to its source was worked as a method, and How to Learn Anything is where the review routine's evidence actually lives.
Out. Nothing on the path depends on this course, and what it hands forward is the habit rather than the figures.
Go deeper
- Typed Versus Handwritten Lecture Notes and College Student Achievement: A Meta-Analysis, Flanigan and colleagues, 2024. This course read the abstract, the introduction, two sections, Table 6 and one limitations subsection, from an open copy of the PDF, and Table 6 is the object worth your time: five meta-analyses of one question, side by side, with their intervals.
- Don't Ditch the Laptop Just Yet, Urry and 87 colleagues, 2021. This course read the abstract only. Worth reading for how a replication reports a split result, where the mechanism reproduced and the outcome did not.
- The four books named above. Read one if you want to know what the advice literature says, and read it knowing that no lesson here has checked any of it.
Sources
- Pam A. Mueller and Daniel M. Oppenheimer, "The pen is mightier than the keyboard: advantages of longhand over laptop note taking", Psychological Science 25(6), 2014, pp. 1159 to 1168. Read at abstract level only; the paper was not opened, which lesson 3 says at the point of use. Supports: three studies, conceptual questions, and verbatim transcription as the proposed mechanism.
- Kayla Morehead, John Dunlosky and Katherine A. Rawson, "How Much Mightier Is the Pen than the Keyboard for Note-Taking?", Educational Psychology Review 31, 2019, pp. 753 to 780. Read at abstract level only. Supports: that performance did not consistently differ between groups including a no-notes group in experiment 2, and the authors' own word "premature".
- Daniel Voyer, Scott T. Ronis and Narissa Byers, "The effect of notetaking method on academic performance", Contemporary Educational Psychology 68, 2022, article 102025. Not read. Its pooled figure reaches this course through Table 6 of source 5, and its counts through a search summary, both said in the table above.
- Heather L. Urry and 87 colleagues, "Don't Ditch the Laptop Just
Yet", Psychological Science 32(3), 2021, pp. 326 to
- Read at abstract level only. Supports: 74 laptop and 68 longhand participants, the quiz after a brief distraction with no study, the word-count association running the other way, and the verbatim finding not being robust in the replication.
- Abraham E. Flanigan, Jordan Wheeler, Tiphaine Colliot, Junrong Lu and Kenneth A. Kiewra, "Typed Versus Handwritten Lecture Notes and College Student Achievement", Educational Psychology Review 36, 2024, article 78. Read in part: the abstract, the introduction, "Trends in Existing Literature", Table 6, the discussion of note-taking quantity, and one limitations subsection. Supports: g = +0.248 and the volume figure; Table 6's five pooled effects and intervals; the sentence naming three of the five as statistically significant; and the field's own summary that the benefit emerges from review.
- F. R. Van der Weel and Audrey L. H. Van der Meer, "Handwriting but not typewriting leads to widespread brain connectivity", Frontiers in Psychology 14, 2024, article 1219945, with Adrien Pinet and Marieke Longcamp's commentary, Frontiers in Psychology 15, 2025, article 1517235. Both read in part, at the levels lesson 6 states. Supports: 36 university students, single words, the one-finger typing condition, and the existence of three published objections.
- The course's own constructions, each labelled where it appears in the body. The four worked claims are sentences invented for this lesson rather than anything quoted, and how each is sorted is this course's reasoning. The setting question is this course's own addition to the sort, and the observation that every measurement of note-taking behind these lessons happened in a lecture is a statement about the six sources in this file rather than about the literature at large, and the body names the EEG study as the exception to it. The last two columns of the licence table are this course's reading of the sources, not anything they say about themselves. And the absence of peer-reviewed evidence for personal knowledge management systems is the result of this course's own Stage 1 search, which is a weaker thing than a systematic review and is reported as such.
Check your understanding
This lesson has a 6-question quiz. Pass it and the questions come back on a schedule in Review, so what you learned stays learned. Your progress is saved in your browser; no account needed.