How to read a research paper without getting lost.

The first paper I was told to read took me four hours and I could not have told you what it argued. I had highlighted about half of it. That is the normal failure, and it is not a reading-speed problem. It is a sequencing problem: a paper is not written to be understood in the order it is printed.

What follows is six passes over the same document. Each pass has one question. You are allowed to be bored by everything that does not answer that question.

01 / Get the claim, before anything else

Read the title, the abstract, then jump straight to the last two paragraphs of the discussion. Nothing in between. Then write one sentence in your own words, starting with the words "The authors argue that".

If you cannot finish that sentence, you have already learned something useful: either the paper is badly framed, or it is outside your background and you need a review article first. Both are fine outcomes for eight minutes of work.

The authors argue that students who received written feedback within 48 hours revised more substantially than students who received the same feedback after two weeks.

That is the whole paper, compressed. Every later pass either supports that sentence or complicates it.

02 / Find the question the paper is answering

Go to the last third of the introduction. That is almost always where the question lives, usually right after a sentence beginning "However" or "Yet little is known about". Authors put the gap there because reviewers look for it there.

Copy the question out. Then check it against your one-sentence claim from pass 01. A well-built paper has a claim that answers its own stated question exactly. When they drift apart / the question is about learning, the claim is about satisfaction scores / you have found the paper's weakest seam, and you found it in ten minutes.

03 / Trace the evidence, not the prose

Now skip the text entirely and read only the figures and tables, captions included. For each one, answer two things: what was measured, and which direction did it go.

Do this before reading the results section. The results section is the authors' interpretation of the figures, and if you read it first you inherit their reading. Looking at the table cold, you notice things like a control group of eleven people, or an effect that exists in one of four measures and is being reported as if it existed in all four.

A habit worth building

Write the numbers down by hand, in a line: n = 214, treatment 68% / control 61%, p = .04. Once it is in your handwriting you tend to ask whether seven percentage points matters. On the page it just looks like a finding.

04 / Interrogate the method

Methods sections are dull on purpose and this is where the paper is actually won or lost. Three questions cover most of it:

• Who or what was studied, and who was left out? Undergraduates at one university in one term is a real finding about a narrow world.
• How was the thing measured? "Engagement" measured by clicks is a different claim from "engagement" measured by an interview.
• What was compared to what? No comparison, no causal claim, however confident the abstract sounds.

This approach builds on a well known predecessor: S. Keshav's three-pass reading method ("How to Read a Paper," 2007), widely circulated in computer science, makes a similar point in fewer words: on the second pass you should be able to explain the paper to someone else, and on the third you should be able to spot how you would have done it differently. Method is where that third pass lives.

05 / Look for restraint

Read the limitations. Slowly. This is the fastest way to sort careful work from confident work.

A serious limitations paragraph names a specific threat and tells you what it would take to rule it out. A decorative one says the sample was small and the topic deserves further study. If the limitations section is decorative, treat every claim in the abstract as one size stronger than the evidence supports, and read the conclusion again with that discount applied.

06 / Write the note you will actually reuse

Five lines, in a single file you keep for the whole project:

1. The question, in the authors' terms.
2. The design, in one clause: quasi-experiment, two cohorts, self-reported outcome.
3. The strongest single piece of evidence.
4. The limitation that genuinely worries you.
5. One question of your own that the paper made possible.

Line five is the one that matters. It is where a literature review stops being a summary of other people's work and starts becoming your argument. Nine papers, nine line-fives, and you can usually see the shape of your own question in them.

---

A last practical note. You do not owe every paper six passes. Passes 01 and 02 take about ten minutes and are enough to decide whether the paper belongs in your project at all. Most do not. Spending an hour finding that out is the actual time sink, not the reading itself.

---

Further reading: Keshav, S. (2007). How to Read a Paper. The original one page guide.