How to read research papers effectively

Most graduate students were taught to read papers the way they were taught to read textbooks — linearly, beginning to end, highlighter in hand. That works for textbooks. It is the wrong tool for research papers. This guide covers a reading method, a curation rule, and a note system that compound across years.

The three-pass method#

Adapted from S. Keshav's classic note, the three-pass method treats reading as an escalating commitment. You only invest more time once the paper has earned it.

First pass — orientation (5 minutes)#

Read the title, abstract, introduction, section headings, and conclusion. Skim the figures. At the end you should be able to answer: What problem does the paper claim to solve? What does it contribute? Is this paper relevant to me right now?

Most papers stop here. That's by design — you only have so much attention, and most papers in your field aren't worth more than five minutes of it.

Second pass — argument (60 minutes)#

If the paper passes orientation, read it for the argument. Skip equations on first read. Look closely at every figure caption. Mark the central claim. The goal is to summarise the paper in three sentences without looking back.

Third pass — replication (4+ hours)#

Reserve the third pass for papers you want to build on. Walk through the methods as if you were going to replicate the study. Re-derive the equations. Critique every design choice. By the end you should know the paper's weaknesses as well as the authors do.

How to choose what to read#

Most graduate students under-curate. A rough rule that has held across our cohorts: for every paper you read deeply, you should have first-passed ten.

• Start at the source. Find the canonical paper in your area — the one others cite as background. Read it deeply.
• Work the citation graph forward. Use Google Scholar's "cited by" to find papers that built on it.
• Work the citation graph backward. Mine the bibliography for foundational work you haven't read.
• Subscribe to two journals, not ten. Read every issue of two; ignore the rest until they cross your radar.

Notes that compound#

The single highest-return habit in graduate school is a consistent note system. The format matters less than the consistency, but a structure we've seen work across cohorts:

1. One file per paper, named `Author Year — short title.md`
2. Three sections: Claim, Method, Why I should remember this
3. One sentence each for the first two; one paragraph for the third
4. Link forward: every note ends with two or three links to other notes

After two years of consistent notes, you'll have a personal knowledge graph. After five, you'll find yourself citing your own past notes more than the live literature.

Building a reading practice#

Pick a fixed slot in your week. Two hours, undistracted, every Friday morning — whatever survives in your calendar. The students who read consistently outpace the students who read intensively. Compounding always wins over heroics.

Wrapping up#

Reading research papers is a skill, not a chore. The three-pass method gives you permission to abandon papers early; the curation rule gives you a way to choose what to read; the note system makes everything you read available to your future self. Together, they turn reading into one of the most reliable engines of research progress.