How to read a research paper: the three-pass method

By Koray Özcan · Updated November 7, 2026

Don't read a paper from start to finish. Read it in up to three passes, each deeper than the last: a five-to-ten-minute skim to decide whether it matters, an hour to understand it, and a slow, critical read only for the few papers you need to master. The method comes from computer scientist S. Keshav's short guide "How to Read a Paper" (ACM SIGCOMM Computer Communication Review, 2007), and it works in any field.

Pass 1: five to ten minutes, to decide

Read only these, in this order:

The title, the abstract and the introduction.
The section and subsection headings, nothing under them.
The conclusion.
A glance at the references, ticking off the ones you already know.

Then answer five questions, which Keshav calls the five Cs:

Category: what kind of paper is it? A measurement, a new method, a review, a trial?
Context: which other work is it built on?
Correctness: do the assumptions look reasonable?
Contributions: what does it claim is new?
Clarity: is it well written?

Most papers stop here, and that is the point. If the abstract does not hold up, or the paper is outside what you need, you have lost ten minutes, not an afternoon.

Pass 2: about an hour, to understand

Read the whole paper now, but skip the proofs and the dense derivations. Look closely at the figures, tables and graphs: are the axes labelled, are there error bars, do the results actually show what the text says they show? Write short notes in the margin, and mark the references you have not read but need to.

After this pass you should be able to explain the paper's main point and its evidence to someone else in a few sentences. If you can't, the paper may be badly written, or it may need background you don't have yet. Either way, decide whether to stop, read the background first, or go on.

Pass 3: hours, to master

The third pass is for papers you will build on, review or teach. Read every line and question every assumption. Keshav's advice is to rebuild the work in your head: make the same assumptions the authors made and work it out yourself, then compare. Where your version differs is where you find the paper's hidden assumptions and weak points. This takes a beginner four or five hours, and about an hour once you're used to it.

Abstract first, always

Every pass starts with the abstract, because the abstract is the authors' own claim about what they did. Read it before any summary, AI-written or otherwise. A summary can be a good way to decide whether to start, but it can drop the condition that makes a result true, such as "in mice", "in one hospital" or "under these settings". The abstract keeps those details.

Make it a habit

The method works best when you use it often, not once before an exam. A few first passes a week, across the fields you follow, keeps you current far better than a pile of unread PDFs. Nerdi is made for exactly this: three papers a day from the fields you choose, each with its original abstract and a free PDF when one exists, so every day starts with a first pass.