How to read a scientific paper quickly and well

Nobody reads a research paper the way they read a novel, from the first page to the last. Experienced researchers skim, jump to the figures, read the methods closely only when they need to, and decide early whether a paper deserves more time. This guide describes a method you can use on any paper, and what to look for in each section.

Start with why you are reading

Before you open the PDF, know what you want from it. Are you checking whether it is relevant? Looking for a specific result to cite? Learning a method you plan to use? Evaluating it as a reviewer? The answer decides how deeply you read. Most papers you encounter only need a quick look; a few deserve hours.

The three-pass method

A widely used approach, described by the computer scientist S. Keshav in a short 2007 essay called "How to Read a Paper", reads a paper in up to three passes, each deeper than the last. You stop as soon as you have what you need.

First pass: the bird's-eye view

Spend five to ten minutes. Read the title, the abstract and the introduction. Read the section headings, but not the text beneath them. Read the conclusion, and glance over the references to see which work the authors build on.

At the end of the first pass you should be able to say:

• what kind of paper it is: an experiment, an analysis of existing data, a review, a new method or a theory;
• what question it asks and what it claims to have found;
• whether its assumptions look reasonable;
• whether it is relevant enough to read further.

Second pass: the content

Spend up to an hour, reading with more care but skipping the finest details such as proofs or derivations. Look closely at the figures and tables: are the axes labeled, are the error bars or confidence intervals shown, and does what you see support what the text says? Mark references you have not read and might need.

After the second pass you should be able to summarize the paper's main argument and evidence to a colleague. For many papers, especially ones you will cite as background, this is enough.

Third pass: the deep read

For the papers central to your own work, read as if you had to reproduce the study. Question every assumption, check each step of the analysis, and think about how you would have designed it. This can take several hours, especially in an unfamiliar field. At the end, you should know the paper's strengths, its weaknesses and its hidden assumptions well enough to explain them.

What to look for in each section

• Abstract: the claim in miniature. Note the main finding and the design, and treat strong wording with caution until you have seen the evidence.
• Introduction: the gap. The last paragraph or two usually states exactly what the authors set out to do and why previous work fell short.
• Methods: the foundation of everything else. Who or what was studied, how many, how variables were measured, and how the data were analyzed.
• Results: the evidence. Read the figures before the text, and look at effect sizes and confidence intervals, not only p-values.
• Discussion: the interpretation. Separate what was found from what the authors think it means, and read the limitations paragraph closely.
• References: the context. They show which conversation the paper is part of and point you to the papers you should read next.

Questions to ask of any paper

• Does the design actually answer the question? A correlational study cannot establish cause, however it is worded.
• Is the sample large enough and representative of the population the conclusions are about?
• Were the measures valid and reliable, and were they chosen before the data were collected?
• Are the effects large enough to matter in practice, not only statistically significant?
• What would the authors' strongest critic say, and do they answer it?
• Do other studies agree, and if not, why might they differ?

Take notes you will use later

Notes taken in the margins of a PDF tend to vanish into a folder. Instead, write a short structured note for every paper you read past the first pass: the question, the design, the sample, the key result with its numbers, the main limitation and one line on why it matters to you. Write them in your own words, so you are not tempted to copy phrasing later.

Kept consistently, these notes become a synthesis matrix for a literature review or the annotations of an annotated bibliography.

Common traps

• Reading every paper deeply. Most papers only need a first pass, and deciding that quickly is a skill.
• Stalling on unfamiliar statistics. Note the method, read on, and look it up later if the paper turns out to matter.
• Trusting the abstract alone. Abstracts are written to sell the paper and often leave out the caveats.
• Losing the paper afterward. Save it with its DOI and your note the moment you decide it is relevant; How to organize research papers explains a simple system.

Reading in Mosaic

Mosaic's reader shows a paper's PDF next to its details. You can highlight passages, keep notes and tag the paper as you read, and everything stays with the paper in your project. When you want to compare a paper with the rest of your library, the assistant answers with citations you can open and check.