Zettelkasten, PARA, or Plain Folders? Choosing a Note Method
Match Zettelkasten, PARA, Cornell notes, a daily log, or search-first folders to your kind of work, then test one in two weeks with a timed find test.
Ask this guide
Answers come straight from this guide’s text. Nothing is generated, and nothing you type leaves your browser.
A note system is easy to abandon when it answers a question its owner never asks. A method built to turn reading into essays does little for someone who mostly needs last quarter's contract terms in ten seconds, and a tidy tree of project folders does little for a student with an exam in three weeks.
This guide works from the other end: sort your work into four kinds, see how five well-known methods fail, run a two-week trial with a timed find test, and move over without reorganizing every old note.
Start with what you will ask your notes: four kinds of work
A method is a bet on what your future self will want from a note.
| Kind of work | The question you ask your notes | Try first |
|---|---|---|
| Project-driven (client work, launches) | "What did we decide, and what is next?" | PARA, with a daily log |
| Idea-driven (writing, research, a thesis) | "What do I already think about this, and what connects to it?" | Zettelkasten |
| Study-driven (courses, exams) | "Can I recall and explain this on the date?" | Cornell notes |
| Reference-heavy (how-tos, settings, specs) | "Where was that setting or snippet?" | Plain folders plus search |
Work often has one dominant kind and a smaller second one. Give the dominant kind its method and let a daily log plus search cover the rest. The table is our synthesis, not a validated classification. To find the dominant kind, ask what a typical note is doing three months on:
- Re-read before a test: study-driven.
- Combined into something you write: idea-driven.
- Pulled into a project, then finished: project-driven.
- Looked up now and then: reference-heavy.
If notes mostly record what happened and why, start with a daily log. Reminders that expire within a day belong on a sticky note (see desktop sticky notes on a Mac).
Five methods, how each one fails, and what to do about it
The warning signs and fixes below are our rules of thumb, not tested ones.
Zettelkasten: one idea per note, linked, for long-running writing
Niklas Luhmann (1927–1998), a German sociologist at Bielefeld, kept a paper slip box that his university's archive puts at about 90,000 notes written between 1951 and 1996. The digitization team counts two boxes: about 24,000 notes from roughly 1952 to 1962 and 66,000 from 1963 to 1996.
The box was not folderless: the second box had 11 top-level sections, one of them "ad hoc notes", and the first had 108 sections. A new note only had to attach to the previous one, giving numbers like 1/1, 1/1a, and 1/1a1, and other connections were references written on the cards, about 30,000 in the second box. In the second box, a keyword register of about 3,200 entries (roughly one per 20 notes, our arithmetic) gave only a few entry points per keyword; the references led to the rest.
Sönke Ahrens's How to Take Smart Notes (2017; second edition 2022) adapted the practice for students and nonfiction writers. It separates fleeting notes you discard, literature notes about sources, and permanent notes, one idea each in your own words, which form the slip box. A made-up chain:
Fleeting: does re-reading actually work?
Literature (Roediger and Karpicke, 2006): tested students beat
re-readers after two days and a week; re-reading won at five minutes.
Permanent: Re-reading wins in five minutes and loses in a week, so for
material I need next month, test myself. Links: Cornell recite step.
Where it fits: research or writing that returns to the same questions for years; Luhmann developed his social theory over more than thirty years. How it fails: collecting without connecting. Warning sign: you can't name three notes that changed something you wrote. Fix: one linked permanent note per source before you collect the next.
PARA and progressive summarization: sort by what is actionable
Tiago Forte presents PARA in Building a Second Brain (2022) and The PARA Method (2023). Projects hold efforts with an end goal, Areas responsibilities that never finish, Resources topics you are following, and Archives whatever has gone inactive. Notes sit with the project they serve now, not under a subject. Progressive summarization is the companion: when you return to a note, bold its core, highlight the best of the bold, and for a few add a summary in your own words.
How it fails: sorting becomes the work, and Resources turns into a bin. Warning sign: choosing a folder takes longer than reading the note will. Fix: if Resources runs past one screen, move anything unopened for 90 days to Archives.
Cornell notes: a page layout that sets up recitation
Walter Pauk, a longtime Cornell education professor, devised the system in the 1950s as a teaching assistant; his book How to Study in College (1962) spread it. Cornell's Learning Strategies Center handout specifies a 6-inch notes column, a 2½-inch cue column, and a 2-inch summary strip, then five steps: Record in telegraphic sentences, write Questions in the cue column, Recite by covering the notes and answering aloud, Reflect, and Review for at least ten minutes every week.
How it fails: the cue column gets filled in and never answered, leaving a page layout, not a method. Warning sign: cues you can't remember answering. Fix: book the recite step in your calendar within a week of the lecture.
A daily log: capture in order, extract on a schedule
A daily log is one running page per day, in time order, with no filing. One variant is Tony Stubblebine's interstitial journaling (a 2017 Medium post): short entries between tasks with times, a few sentences on how the work went, and the first action for the next task. NBC News cites no research for it, so treat it as one person's practice. A made-up entry:
09:40-10:25 Drafted the Q4 report intro. Slow start; the numbers were
not final. Next: open the finance export and fill table 2.
On Friday, scan the week: move what you would act on to your task list and what you would reuse to a titled note.
How it fails: nothing is ever extracted. Warning sign: never having acted on an entry more than a week old. Fix: give the Friday sweep a fixed slot; our 20-minute weekly review has one.
Search-first plain folders: titles do the filing
The lowest-ceremony option is a handful of folders, descriptive titles, and search. The most relevant evidence we found is from email: in a CHI 2011 study of 345 users making more than 85,000 refinding actions, Whittaker and colleagues found that people with elaborate folders relied on them, but the preparation did not improve retrieval success, while search and threading helped. That suggests rather than shows that notes behave the same way.
How it fails: titles like "Meeting notes" return hundreds of near-identical hits. Warning sign: opening six results to find one. Fix: rename any note you had to dig for.
Tip: Title every note with a date, the project or person, and the noun you would search for, for example
2026-10-14 Acme contract termination clause(a made-up example).
Side by side: setup, upkeep, and the smallest version that works
Setup and upkeep are our estimates for one person; the smallest versions are our suggestions.
| Method | Setup | Weekly upkeep | Smallest version that still works |
|---|---|---|---|
| Plain folders plus search | 15 minutes | 0 to 5 minutes | Title rule only; no subfolders |
| Daily log | 5 minutes | 10-minute sweep | One page per day; extract on Friday |
| PARA | About an hour | 10 to 15 minutes | Four top-level folders; subfolders only for live projects |
| Cornell notes | None | 10-minute review (Cornell's minimum) plus reciting | Cues for the three main points, recited once within a week |
| Zettelkasten | Hours to learn; no known payoff threshold | 30 to 60 minutes per source | One permanent note per source, linked to an existing note |
What the research supports and what it does not
We found no controlled study comparing Zettelkasten, PARA, daily logs, and plain folders, and the small studies of Cornell-style notes we saw disagree, so any claim that one method is best outruns the evidence. What exists is evidence about mechanisms, consistent with the idea that what you do after capture matters more than how notes are filed, though no study here tested that.
Roediger and Karpicke (2006) had college students study prose passages, then restudy them or take free-recall tests without feedback. Restudying won at five minutes, but tested students retained substantially more after two days and a week, even though restudying left students more confident. That is the case for Cornell's recite step.
Karpicke and Blunt (2011, Science) found in two experiments with students that retrieving a science text from memory produced more learning than building concept maps from it, even when the final test asked for concept maps. Students tended to expect the maps to work as well or better, so connecting ideas is a poor substitute for practicing recall.
The evidence that changing how people take notes improves learning is weak. Kobayashi's 2006 meta-analysis of 17 studies of note-taking interventions, on learning from lectures or texts, found an overall mean effect near zero (d = 0.02) against no intervention, with some benefit at lower schooling levels and none at higher ones. That is no verdict on Cornell or Zettelkasten, but a fancier format is not, by itself, evidence you learn more.
The practical question is how often the step that makes a method work (reciting, linking, triaging, titling, or extracting) happens in your week. The trial measures that.
A two-week trial with a timed find test
Test one method at a time, for 14 days, on new notes only.
- Day 0, baseline (25 minutes). List five things you looked for in the past month (a decision, a number, a command). Time each in your current setup, marking any you can't find in three minutes. Decide the upkeep you will allow, such as 10 minutes a day.
- Days 1 to 14, use the method. Don't convert old notes. Each day, tick off the upkeep step: reciting, linking, triaging, or extracting.
- Day 14, find test and scorecard (20 minutes). Choose five items from your trial notes, at least two from week 1 that you have not opened since. Time each, then fill in the scorecard.
Method tried: ________ Dates: ________ to ________
Setup (min): ____ Upkeep (min): week 1 ____, week 2 ____
Days upkeep skipped: ____ / 14
Find test (5 items): baseline ____ s, ____ not found
day 14 ____ s, ____ not found
day 60 ____ s, ____ not found
One note that changed what I did or wrote: ____________________
Verdict: keep / adapt / drop
These decision rules are ours, not validated. Keep for now if you found things faster than baseline and did the upkeep on at least 10 of 14 days. Adapt if you found things faster but skipped upkeep on more than four days: shrink the step to its smallest version in the table above. Drop if nothing was faster; the method isn't paying for its upkeep.
One bias to discount: a 14-day pile is tiny next to years of old notes, so almost any method finds faster on day 14. Treat "faster" as necessary, not sufficient, and repeat the test at the 60-day review below.
For example, say you trialed plain folders plus search. At baseline, three items took 40, 95, and 180 seconds and two were not found. On day 14, five trial items took 20, 25, 45, 60, and 70 seconds, none were missed, and you did the title cleanup on 12 of 14 days. Keep for now and recheck at day 60. Five items is a rough check, not statistics: ignore small gaps.
Moving over without a big-bang reorganization
Re-filing every old note fails on arithmetic: 500 notes at two minutes each (an assumption; time yourself) is about 17 hours before the new system has produced anything. Use three rules instead:
- Start on a date. Notes created from then on follow the new method; old notes stay put and searchable.
- Touch it, move it. When you open an old note for real work, bring it across in the new method's form: a date and noun in the title, a project folder, or one idea in your own words with a link. Don't bulk-import highlights into a Zettelkasten; they are literature notes at best.
- Review at 60 days. List old notes you needed and could not find: that is your import list. Then repeat the find test with items you have not opened since.
Where to start this week
- Today (20 minutes): Sort your last ten notes with the three-months-on test, then run the baseline find test. If it is already fast and calm, you don't have a method problem.
- Monday to day 14: Run the method the first table suggests, then fill in the scorecard.
Frequently asked questions
Can I combine methods, or is that how systems get messy?
Yes, but give each method a different job. Pairings that keep jobs separate (ours, not tested): Cornell for lectures plus search-first folders for everything else; a daily log plus Zettelkasten for writers; PARA plus a daily log for project work. Past two methods, every note adds a filing decision (our rule of thumb).
How many notes before a Zettelkasten pays off?
We found no research that gives a threshold, and Luhmann's pace (roughly five notes a day for over thirty years, by our arithmetic) is no benchmark. A checkpoint of our own, not validated: after about 30 permanent notes, pick a topic and try to draft 300 words from linked notes alone. If you can't, they are probably collected, not connected.
Does the app, or an AI tool, matter?
Less than the habits do; Luhmann's box was paper. Our checklist for an app, not tested: stable titles or IDs, links that survive renaming, search, and export to plain text or Markdown, so switching later stays cheap. Our reading on AI: the effortful step in permanent notes and Cornell summaries is restating material in your own words, so let AI quiz you on your notes rather than write them. See learning with AI without fooling yourself.
Sources and further reading
- Niklas Luhmann-Archiv, Bielefeld University — Universität Bielefeld. Note count, years, and thirty-year theory; the 2018 digitization slides (in German) give per-box sections, references, and registers.
- The PARA Method — Tiago Forte, Forte Labs. The four categories; see also Progressive Summarization for the layers.
- How to Take Smart Notes — Sönke Ahrens, Open Library record. The 2017 and 2022 editions. The same catalogue dates Forte's Building a Second Brain to 2022 and The PARA Method to 2023.
- The Cornell Note-taking System — Cornell Learning Strategies Center handout, hosted by Hope University. Column sizes and steps; see also Cornell Alumni Magazine's profile of Pauk (2022) for the 1950s origin.
- How interstitial journaling can help you focus your day at work — NBC News, 2019. The parts of an entry; no research cited.
- Am I wasting my time organizing email? — Whittaker and colleagues, CHI 2011. The email refinding study.
- Test-enhanced learning — Roediger and Karpicke, Psychological Science, 2006, and Retrieval practice produces more learning than elaborative studying with concept mapping, Karpicke and Blunt, Science, 2011. Testing versus restudying, and retrieval versus concept maps.
- Conditional effects of interventions in note-taking procedures on learning: A meta-analysis — Kobayashi, Japanese Psychological Research, 2006. The 17 studies and d = 0.02, as reported in the abstract.