← Blog

The Interstitial Journal blog · Built on evidence, not hype

What to Keep and What to Delete

There is a particular kind of guilt that comes with a notes app: hundreds of saved articles, clipped highlights and half-written entries, none of it re-read, all of it apparently your fault. The guilt assumes that a bigger archive is a better one and that you have failed to maintain it. Both halves of that assumption are worth examining, because the research points somewhere less tidy — and considerably more forgiving.

By comun.ai · 3 August 2026 · ~6 min read

The "collector's fallacy" is a good phrase, not a finding

The term comes from personal-knowledge-management writing, where it names a real enough experience: the small hit of satisfaction from saving something, mistaken for the work of understanding it. It is a useful label. It is not a research construct — nobody has operationalised "collector's fallacy" and tested it, and you should treat any article citing it as though it were an established effect with suspicion.

What does exist is a body of work on the gap the phrase is gesturing at: the difference between material passing in front of you and material you did something with.

Producing beats filing

Two of the sturdier findings in memory research both say a version of the same thing.

• The generation effect. Norman Slamecka and Peter Graf showed that words participants generated themselves were remembered better than the same words simply read.¹ Strong Producing the material changes what happens to it.
• The testing effect. Henry Roediger and Jeffrey Karpicke found that retrieving material — being tested on it — produced substantially better long-term retention than re-studying it for the same amount of time.² Strong And notably, participants tended to predict the opposite, believing re-reading had served them better.

Clipping a paragraph into an app is neither generating nor retrieving. It is the highlighting-and-re-reading end of that comparison, with an extra step that feels like progress. Writing one line in your own words is the other end.

The note-taking study everyone quotes — and what happened next

You have probably met the claim that handwriting notes beats typing them. It comes from a genuinely famous experiment: Pam Mueller and Daniel Oppenheimer reported that students taking notes by hand outperformed laptop note-takers on conceptual questions, and proposed the mechanism that laptop users transcribed verbatim while longhand users were forced to summarise.³

It is a compelling story and it is not secure. Kayla Morehead, John Dunlosky and Katherine Rawson ran a direct replication with extensions and did not reproduce the key advantage.⁴ Contested We are flagging this rather than quietly citing the original, because the original is quoted constantly by people selling notebooks and note-taking systems, and the replication almost never is.

The safer reading is the one that survives both papers: what matters is probably not the implement but whether you processed the material or transcribed it. That is the generation effect again, and it does not care whether you used a pen.

Why the useful note is the effortful one

Robert Bjork's framework of desirable difficulties makes sense of the pattern: conditions that make learning feel harder in the moment often improve retention later, while conditions that make it feel fluent often don't.⁵ Moderate Saving is maximally fluent. It feels like custody of an idea, at almost no cost — which is precisely the profile of something that won't stick.

There is a related and much-cited idea worth grading carefully. Betsy Sparrow, Jenny Liu and Daniel Wegner reported that when people expect information to remain available, they remember where to find it better than they remember the thing itself.⁶ Early It is an appealing account of what an archive does to you, and it is one influential study whose interpretation has been debated since. Interesting; not load-bearing.

So what should you actually keep?

A rule that follows from the evidence rather than from tidiness: keep what you produced, not what you collected.

• Keep the line you wrote in your own words. It's the generated version, and it's the only part with a mechanism behind it.
• Keep what you couldn't reconstruct. A decision, who was in the room, what you promised, what you were worried about at the time. Memory reconstructs the gist and quietly rebuilds the rest — the details are the part worth storing.
• Delete the clipped paragraph, without ceremony. If it mattered, you'll meet it again. If it didn't, it was costing you attention every time you scrolled past it.
• Stop maintaining an archive you don't query. An unread archive isn't a debt. It's a signal that the saving step was the wrong step.

What this means for a journal of seams

An interstitial journal is a slightly different object from a knowledge base, and the distinction matters here. Its entries aren't reference material to be curated — they're timestamped marks at the boundaries of a day, written to get something out of your head and to leave the day legible later. We went through the retrieval side of that in What to Write When Nothing Happened, and the boundary itself in The Ten Seconds Between Meetings.

Which produces a permission most notes advice won't give you: you do not have to re-read it. A line that discharged an open loop when you wrote it has already done its job — the plan-making effect happens at the point of writing, not at the point of review.⁷ Moderate Anything you do go back for is a bonus, not the justification.

Where the evidence ends

To be precise about which parts are load-bearing. The generation and testing effects are strong. Desirable difficulties are moderate. The handwriting-beats-typing claim is contested and should not be repeated as settled. The transactive-memory result is early.

What is not established is the step we could easily imply and won't: that deleting most of your notes will make you think more clearly, work better, or feel lighter. Early No study has tested that, and we're not going to attach a percentage to it — the absence of invented numbers on this site is deliberate and permanent. What the research supports is narrower: producing beats filing, retrieval beats re-reading, and fluency is a poor guide to what you'll retain. This is reflection, not treatment. More on our science page, and on how your notes stay private on our privacy page.