If you're trying to read a research paper front-to-back like a book, you're doing it wrong. Twenty pages of dense methods, statistical notation, and field-specific jargon are not designed for linear reading. They're designed to be interrogated. Researchers who read 10 papers a week and students who read one a month usually use a different strategy, not a different IQ — and the most famous version of that strategy is S. Keshav's three-pass method from his short paper at the University of Waterloo.
PaperDraft is a writing assistant, not a paper generator — the draft is your starting point, not your submission. You are responsible for editing, verifying sources, and following your school's academic integrity policy.
Below, how to read a research paper in three targeted passes, what each pass is for, and how to decide when to stop.
The premise: most papers don't deserve a deep read
The first uncomfortable truth is that most papers you skim will not warrant a full read. That's not laziness — it's triage. You're looking for a handful of papers that will actually shape your thinking. The three-pass method is about efficiently deciding which papers those are.
Each pass has a specific purpose and a specific time budget. You don't move to the next pass until you've decided the paper is worth it.
Pass one: the 5–10 minute skim
The first pass is reconnaissance. Your goal is to decide whether the paper is worth reading at all. You read:
- Title, abstract, and introduction. This tells you what the paper claims.
- Section and subsection headings. This tells you how the argument is organized.
- The conclusion. This tells you what the paper says it found.
- Glance at the references. If you recognize names, you're in familiar territory.
Do not read the methods or results sections yet. Do not try to understand every sentence. You're building a map, not reading the terrain.
At the end of pass one, you should be able to answer five questions Keshav calls the five Cs:
- Category — What type of paper is this? (survey, empirical study, theory, system description)
- Context — What other work is it related to?
- Correctness — Do the assumptions appear valid?
- Contributions — What does it claim to contribute?
- Clarity — Is it well written?
If the answers are clear and the paper isn't relevant to your work, stop. You've saved yourself hours. If it's relevant but poorly written, note it and move on. If it's relevant and worth more, move to pass two.
Pass two: the 1-hour structured read
The second pass is about understanding what the paper actually says, but not yet how it proves it. Spend up to an hour and:
- Read the paper carefully but skip the proofs and heavy math. The goal is to grasp the argument, not reproduce it.
- Examine the figures and tables. In most empirical papers, the figures carry the real argument. Read every axis label. Ask whether the charts actually support the claims in the text.
- Mark unclear terms and passages. Don't stop the flow to look them up — note them for later.
- Note the references you want to follow up on. These become your citation chain.
At the end of pass two, you should be able to summarize the paper's main thrust to someone else in a few sentences — what question it asks, what it claims, and the shape of the evidence. If you can't, you either need another pass-two reading or the paper isn't well-argued enough to invest more time in.
This is where most students stop when reading for a class. For a paper you'll actually cite or build on, you may need pass three.
Pass three: the virtual re-implementation
The third pass is for papers central to your own work — the three to five sources you'll actually rely on in your argument. It takes several hours.
The idea is to mentally re-implement the paper. For a theoretical paper, that means working through the proofs. For an empirical paper, it means imagining what you'd do to replicate the study with the same data. For a system paper, it means reconstructing the architecture on a whiteboard.
At this level of reading, you're looking for:
- Unstated assumptions. What does the paper take for granted that you might question?
- Missing citations. What related work isn't cited, and why?
- Alternative methods. Could the same question have been answered differently?
- Errors. Real ones — in math, in methodology, in interpretation of prior work.
After pass three, you should know the paper well enough to identify both its strengths and the specific places where a thoughtful critic would push back. This is where a paper goes from "a source I cite" to "a paper I understand."
Four papers deep and still no draft on the page? PaperDraft gives you a structured first draft — thesis stub, outline, cited opening sections — so you can spend your time on deep reads instead of staring at a blank document. It's a drafting assistant, not a submission. Try PaperDraft — free
When to use each pass
- Every paper you consider: pass one, always. 5 to 10 minutes, no exceptions.
- Papers relevant to your argument: pass two. Up to an hour.
- The 3–5 papers central to your thesis: pass three. Several hours each.
If you're writing a literature review over a week, you might pass-one 30 papers, pass-two 10, and pass-three 3. That's normal. It's also how you avoid drowning in PDFs you partially read and can't quite cite.
Common mistakes
Trying to read every paper deeply. You can't. A 20-source paper means 20 pass-threes at 3 hours each is 60 hours of reading. That's a full work week just for reading. Pick your deep reads intentionally.
Reading linearly. Start with the abstract, not the introduction. Read the figures before the results text. Read the discussion before the methods if you want to know what the paper actually claims.
Not taking notes as you read. Pass one and two reading without notes is reading you'll have to redo. See our note-taking guide for a system.
Confusing pass one with the whole read. A skim is not a read. Don't cite a paper you've only skimmed — either invest in pass two or drop the citation.
Skipping the references. The references list tells you the paper's intellectual neighborhood. It's where your citation chaining finds its next targets.
How a drafting assistant fits
Once you've done your three-pass reading and identified the key claims you'll cite, PaperDraft can take your source list and scaffold a structured first draft — a thesis stub, an outline, and opening sections with citation stubs pointing to the papers you've read. You still do the reading and the interpretation — the tool cannot read a paper for you — but it handles the scaffolding so your reading time translates into actual writing. The assistant doesn't replace the pass-three work; it just frees up the hours so you have time to do it.
FAQ
Who came up with the 3-pass method?
S. Keshav, a professor at the University of Waterloo (later Cambridge), wrote the short paper "How to Read a Paper" around 2007. It's become one of the most widely circulated papers in research methodology, partly because it's exactly the kind of focused, useful writing the method itself trains you to appreciate.
How long does pass one actually take?
With practice, five minutes. Early on, 10–15 as you get used to reading out of order. The key is strict time-boxing — if you're still in pass one after 20 minutes, you've already committed to pass two without deciding to.
Do I need to use all three passes for every paper?
No. That's the whole point. Most papers stop at pass one. A few go to pass two. Only the papers that matter to your argument get pass three.
Can I skip the methods section in pass two?
You can skim it. You cannot skip it entirely — methods determine whether the results mean what the paper claims they mean. A flashy result from a bad method is not evidence.
How do I take notes while using the 3-pass method?
Use different note structures per pass. Pass one: one sentence on whether to continue. Pass two: bullet points on claims and evidence. Pass three: a structured summary. See our note-taking for research papers guide for templates, and check the synthesis matrix template if you're organizing multiple papers. For the full research-to-writing pipeline, see our guide to writing a research paper.