The most efficient way to read a research paper is the three-pass method described by S. Keshav in his short paper "How to Read a Paper" (2007). First pass: skim for 5 to 10 minutes to decide whether the paper matters. Second pass: read carefully for about an hour, ignoring proofs and fine detail, to grasp the content. Third pass: re-create the work in your head, which takes a few hours and is only for papers you must master or review.
Why reading papers front to back fails
Papers are not written like novels. They are written to be defended, so the claims, methods and caveats are spread across sections. Reading from the abstract to the references line by line can take hours and leave you unable to say what the paper actually showed. A structured approach lets you stop early when a paper is not useful, which is most of the time.
Pass one: the bird's-eye view (5 to 10 minutes)
- Read the title, abstract and introduction.
- Read the section and subsection headings, ignoring the content.
- Read the conclusions.
- Glance at the references and tick the ones you already know.
Then answer Keshav's five Cs. Category: what type of paper is this (measurement, new method, review, theory)? Context: which other papers is it related to and what base does it use? Correctness: do the assumptions look valid? Contributions: what are the main claims? Clarity: is it well written? After this pass you can decide whether to read on, file it or drop it.
Pass two: the content (about an hour)
Now read with more care, but skip the proofs and the dense technical detail. Look hard at the figures, tables and graphs. Check whether axes are labelled, whether error bars or confidence intervals are shown and whether results are statistically sound. Mark references you want to follow up. Afterwards you should be able to summarise the main thesis, with supporting evidence, to someone else.
If you reach the end of this pass and still do not understand it, you have three options: set it aside and come back after you read the background, read the background now, or persist to the third pass.
Pass three: re-create it (several hours)
This pass is for papers you are reviewing or that your own work depends on. You try to re-create the paper: make the same assumptions and work out how you would have done the study. Compare with what the authors did. This is where you spot hidden assumptions, weak controls and ideas for future work.
Questions to ask any paper
- What is the research question, and is it clearly stated?
- What is the sample or dataset, and how large is it?
- What is the method, and could someone replicate it from the description?
- What are the main results, with numbers?
- What do the authors say are the limits? What do you think they are?
- Do the conclusions follow from the results or go beyond them?
How to take notes that you can use
Write notes in your own words, not copied sentences. A simple template for each paper keeps them comparable:
| Field | What to write |
|---|---|
| Reference | Authors, year, title, venue |
| Question | One sentence |
| Method | Design, data, key steps |
| Findings | Two or three results with numbers |
| Limits | Yours and the authors' |
| Why it matters to me | How it connects to your project |
Remember what you read
Reading ten papers and remembering none of them is common. Retrieval practice helps: after each paper, close it and write down the question, method and main result from memory, then check. Review those notes after 1, 3 and 7 days. If your reading list includes methods or terminology you need to master, such as statistical tests, turn them into flashcards or short Q&A so they come back on a schedule.
Reading many papers
When you read a field for the first time, begin with a recent review article to get the vocabulary and the map of the area. Then follow the most cited papers and the newest ones. Search tools such as Google Scholar, OpenAlex and Semantic Scholar help you see who cites whom. Keep a spreadsheet of what you have read with the key findings, because it will become your literature review.
Using Type to Study with papers
If you upload a paper or a set of papers as PDFs (up to 5 files at a time), Type to Study can build a course from them, with lessons that cite the pages they come from, flashcards for key terms and an AI tutor that answers from those lessons. It does not replace reading the paper critically, but it helps with the retention step. If you only have a topic, it can also build a course from research papers listed in OpenAlex, with every source shown.
The aim is not to read everything. It is to decide quickly what deserves your attention and then to understand that fraction properly.