The three-pass method

In a short and widely shared article, "How to Read a Paper" (2007), computer scientist S. Keshav describes reading a paper in up to three passes. Each pass goes deeper, and you can stop after any of them.

First pass: the bird's-eye view (5 to 10 minutes)

Read the title, the abstract and the introduction. Read the section and sub-section headings, but nothing else in the body. Read the conclusion. Glance at the references and tick off the ones you already know.

After the first pass, you should be able to answer Keshav's five Cs:

1. Category: what type of paper is it? (a new method, a measurement, an analysis of a system...)
2. Context: which other papers is it related to? What theory does it build on?
3. Correctness: do the assumptions look valid?
4. Contributions: what are its main contributions?
5. Clarity: is it well written?

For many papers, this is enough. You know what it is about and that it is not relevant, or not yet.

Second pass: the content (up to an hour)

Read with more care, but skip details such as proofs. Look closely at the figures, diagrams and tables. Are the axes labelled? Are there error bars? Mark references you have not read for later. Afterwards, you should be able to explain the paper's main point, with evidence, to someone else.

Third pass: the deep understanding (several hours)

To fully understand a paper, Keshav suggests trying to virtually re-implement it: make the same assumptions as the authors and re-create the work in your head, or in code. Compare your version with theirs. This shows you hidden assumptions, weak points and ideas for your own research. He estimates this pass takes a beginner four or five hours, and an experienced reader about an hour.

RememberMost papers you look at only need the first pass. Save the third pass for the few papers your project is built on.