How to read a research paper

Pass 1 · Five minutes · Is this worth my time?

• Read the title and the abstract. Write the main claim in one sentence, in your own words.
• Look at every figure and its caption. Figures are usually the result; the text is the argument around it.
• Read the last paragraph of the introduction. It usually says exactly what the paper sets out to do.
• Skim the conclusion. Does it match what the abstract promised?
• Decide: keep reading, save for later, or drop it. Dropping papers is a skill, not a failure.

Pass 2 · Thirty minutes · What did they actually do?

• Read the methods slowly. Who or what was studied, how many, and how was the key thing measured?
• For each figure, say out loud what the axes are before you look at the pattern.
• Underline every claim. Next to each, mark the evidence that supports it, or a question mark if there is none.
• Circle words you do not know and look up only the ones that block the main idea.
• Note one reference you might want to read next.

Pass 3 · As long as it takes · Would I believe this?

• Could anything else explain the result? A confounder, a selection effect, chance?
• Is the sample big enough, and like the people or things the conclusion talks about?
• Do the authors name their own limitations? Which one matters most?
• If you ran this study again, what would you change?

Where to find papers you can actually open

Google Scholar (look for the [PDF] link on the right), PubMed Central for health and biology, arXiv for physics, maths and computer science, and the authors' own websites. If a paper is paywalled, emailing the author for a copy is normal and usually works.

Your one-paragraph summary

1. The paper asks:
2. To answer it, the authors:
3. They found:
4. I am not fully convinced because:
5. The next question I would ask is: