# How to read a research paper faster

You do not need to read a research paper the way you read a novel. Most papers will not earn a line-by-line pass. The reliable way to go faster is three passes: decide if it belongs, build the skeleton from figures and headings, then RSVP only the sections that carry the claim.

## How do you read a research paper faster?

Use three passes, and stop when the job of that pass is done. Pass 1 is a five-minute filter. Pass 2 is a ten-minute skeleton. Pass 3 is a focused RSVP read of methods, results, or argument — not the whole PDF by default. Treating "I opened the PDF" as "I read it" is the usual failure.

## What is pass 1 of reading a paper?

Pass 1 answers one question: does this paper belong in this review, or are you done? Read the title, abstract, and conclusion. Skim section headings. Spend about five minutes.

- What claim are they making?
- What kind of evidence (experiment, proof, survey, argument)?
- Is this the paper, or a citation you actually need instead?

Do this at a high RSVP speed (500–700 WPM) or just with your eyes. Do not mark yet. If the abstract already tells you it is off-topic, close it. A literature review dies when every PDF gets a full reading because it was in the folder.

## What is pass 2 of reading a paper?

Pass 2 builds the skeleton: method, result, caveat. Look at every figure and table caption, then the first and last paragraph of each section. Budget about ten minutes.

You should be able to say, out loud, "They did X on Y and report Z, with the usual limitations around W." If you cannot, the paper is denser than it looked — budget a real pass. If you can, you may not need pass 3 at all.

## When do you RSVP a research paper?

RSVP the paper only in pass 3, and only on the sections that carry the claim: methods if you need to trust the result, discussion if you need the argument, related work if you are mapping the field.

Run RSVP at a speed you can follow — 350–450 WPM for dense methods, faster for related work. Mark:

- The actual claim (not the hype in the intro)
- The dataset, setup, or assumption that would break the result
- A number you might cite
- A limitation you might need to mention

Do not write a summary while the words are moving. Mark the spot, finish the section, then review.

## Why is reviewing marks the actual reading?

Review is the reading because the artifact you need is a list of claims, methods, and limitations — not a feeling that you "got through" the PDF. After the pass, flip through your marks and add a one-line note: "claim", "method", "limitation", "cite this." Export that list.

S. Keshav's three-pass approach is the ancestor of this. The change is mechanical: RSVP makes pass 3 cheap enough that you will actually do it on the papers that survived the first two filters.

## How should you queue papers for a review?

Put the week's papers in a Readlist in the order you want to attack them, not in a folder named "to read." When one is done, the next is waiting. The system is: filter, skeleton, RSVP, review, next.