How to Read Research Papers Fast: The Three-Pass Method

The fastest way to read a research paper is not to read it front to back. Read it in passes, and decide after each one whether it deserves more of your time. A first pass of five to ten minutes (title, abstract, introduction, headings, conclusions) tells you what the paper claims and whether it matters to you. A second pass of up to an hour, focused on the figures and the argument, lets you explain it to someone else. Only the few papers you must review, replicate or build on need a full third pass. Most of the time you save comes from reading less of most papers, not from moving your eyes faster.

The three-pass method for reading a scientific paper

The best-known version of this approach comes from S. Keshav, a computer scientist at the University of Waterloo, in a two-page article titled "How to Read a Paper", published in ACM SIGCOMM Computer Communication Review in 2007. It was written for graduate students, but the method works in any field.

First pass: decide whether to read it at all

Read the title, abstract and introduction carefully, then only the headings, then the conclusions, and skim the reference list for work you already know. At the end, Keshav suggests you should be able to answer five questions:

• Category: what type of paper is it (an experiment, a measurement, a review, a new method)?
• Context: what other work and theory does it build on?
• Correctness: do the assumptions look valid?
• Contributions: what does it add?
• Clarity: is it well written?

Many papers stop here, and that is the point. A paper outside your area, or one whose claims you only need to know about, has done its job after ten minutes.

Second pass: understand the argument

Read the paper properly but leave aside proofs and derivations. Spend real time on the figures: check that axes are labelled and that results show their uncertainty. Note the key points in the margin and mark references you have not read. If the paper still makes no sense after this pass, you can set it aside, read some background first, or commit to the third pass.

Third pass: re-create the work

Read as if you had to reproduce the study, making the same assumptions as the authors and challenging each one. This is slow by design, and it is the only pass that lets you judge a paper rather than describe it. Reserve it for papers you are reviewing, replicating or relying on.

Ask after every pass: do I need more from this paper? The answer is usually no, and that is where the time goes back into your week.

In what order should you read the sections?

Papers are written in a fixed order (abstract, introduction, methods, results, discussion), but they do not have to be read that way. A 2017 survey by Hubbard and Dunbar of undergraduates, PhD students, postdoctoral researchers and academics at a UK university found that readers at different stages value different sections. Undergraduates ranked the abstract as the most important section; postgraduate researchers ranked figures and tables highest; academics placed the methods third on average. Fewer than a third of undergraduates found the methods easy to read, against roughly four in five PhD students and academics. It was one university and a self-selected sample, but the pattern matches common advice: experienced readers go to the evidence.

A sensible order depends on why you are reading:

• Keeping up with your field: abstract, figures and their captions, conclusions. Stop there unless something surprises you.
• Deciding whether to cite a claim: abstract, the result that supports the claim, the methods behind that result, and the limitations paragraph of the discussion.
• Learning a new area: introduction first, because it summarises the field and points to the key earlier papers, then the discussion.
• Replicating or reviewing: everything, methods most carefully of all.

How to skim the abstract and figures

Skimming a paper is a skill with a structure, not a blur. Most abstracts follow the same sequence: background, aim, method, main result, conclusion. Find the sentence that states the main result and the one that says how it was obtained; those two carry most of the information.

Then go through the figures in order, reading each caption in full. Ask what is being compared, what the axes measure, how large the effect is and how uncertain it is. In many papers the figures tell the whole story, and the text explains how the authors got there. The general technique is covered in skimming vs scanning.

When not to speed read a paper

The research review by Rayner and colleagues (2016) concludes that reading much faster than your normal rate costs comprehension, and most papers are exactly the kind of dense, unfamiliar text where understanding slips first. Slow down, and read at your natural pace, for:

• Methods. Whether a result can be trusted depends on details such as the sample, the controls and the analysis. Skimming them means taking the conclusion on trust.
• Key results you will rely on. If you will cite it, teach it or build on it, read the passage that supports it in full.
• Anything you do not understand on first reading. Going back over a difficult sentence is part of understanding it; see regression in reading.

Where an RSVP reader helps, and where it does not

An RSVP reader shows text one word or phrase at a time at a set speed. For papers it fits a narrow job:

• It does not work for equations, tables, figures, code, reference lists or multi-column layouts. These depend on their position on the page and turn into noise when flattened into a stream of words.
• It can help with long stretches of prose: introductions, discussions, review articles and book chapters, where the task is to follow an argument and you tend to drift.

In practice, do your first pass on the PDF itself. Then, for a long introduction or discussion you want to get through, copy that section into a paced reader as its own text, or import the PDF and read the prose parts. A full-text mode with a moving highlight, 200 to 275 words per minute and three or four words per chunk is a reasonable start for dense material. The details of importing and what survives are in how to speed read a PDF.

How to take notes that make the next pass faster

1. After the first pass:write one sentence stating the paper's main claim, plus your answers to the five questions. This is often all you will ever need.
2. During the second pass: note the key evidence for each claim, with the figure or table number, and list the references you want to follow.
3. After the second pass: write a three-to-five sentence summary in your own words and one question the paper leaves open. If you cannot write the summary, you have not finished the pass.
4. Keep the notes with the reference in whatever reference manager or document you use, so you never re-read a paper to remember what it said.

Reading many papers for a literature review

Keshav applies the same method to surveying a new field:

1. Use an academic search engine to find three to five recent papers, and give each a first pass.
2. Read their related-work sections. If you find a recent survey paper, read that and you are largely done.
3. Otherwise, look for citations and author names that keep recurring: those are the key papers and researchers. Collect them, and see where those researchers have published recently.
4. Give the collected papers two passes each, and repeat if they all cite something new.

How to practise reading papers faster

For one week, do five first passes a day on papers from your field, with a ten-minute timer for each. Write the main claim and your five answers before opening the next one. In the second week, add one second pass a day of up to an hour, ending with a written summary. After two weeks you will triage a reading list noticeably faster, because you will have practised deciding, not just reading. For students juggling several courses, see speed reading for students.