How to read a research paper efficiently

Academic Writing, Reading Research and Presenting

“Researchers spend a great deal of time reading research papers. However, this skill is rarely taught, leading to much wasted effort.”

— S. Keshav, How to Read a Paper

Most students learn to read papers the same way they learn to read anything else in school: start at the top, read every word, finish at the bottom. For a dense, jargon-heavy research paper, this is slow, exhausting, and often the wrong strategy entirely — you'll spend forty minutes on the first two pages before discovering the paper isn't even relevant to what you needed.

The three-pass method

Computer science professor S. Keshav's widely used approach breaks reading into three increasingly deep passes, so you only invest as much time as a paper turns out to deserve.

Pass one (5–10 minutes): should I bother with this paper at all? Read the title, abstract, and introduction carefully. Read the section and sub-section headings, but nothing under them yet. Glance at any mathematical content just to get a sense of its theoretical basis. Read the conclusion. Skim the references, noting any you already recognise. By the end of this pass you should be able to answer: what problem is this solving, why does it matter, what's the main contribution, and is the writing clear enough to be worth your time? If the answer is "not relevant" or "not worth it," stop here — that's a successful first pass, not a failed one.

Pass two (about an hour): grasp the content, not the details. Read the paper with more care, but skip proofs and other heavy technical detail for now. Look closely at the figures, diagrams, and other illustrations in particular — they're often the fastest way into a paper's core idea, and poorly drawn figures can themselves be a signal about the paper's overall quality. Mark unfamiliar terms and any references you might need to chase down later. After this pass, you should be able to summarise the paper's argument in your own words to someone else, along with its main supporting evidence — even if you couldn't yet reproduce its proofs or rebuild its experiments from scratch.

Pass three (four or more hours): understand it in depth. The goal here is to virtually re-implement the paper — mentally reconstruct the entire structure of the work, question every assumption, and imagine how you'd have presented the same idea yourself. This pass is where you'd catch subtle flaws, spot implicit assumptions the authors didn't call out, and generate ideas for your own related work. Reserve this level of effort for papers that are directly central to your project or literature review — not every paper you touch needs it.

Why is this important?

Without a deliberate strategy, it's easy to either skim everything shallowly (and miss what actually matters in the papers that count) or read everything at full depth (and run out of time before covering enough ground). The three-pass method solves this by making "how much time should I spend on this paper" an explicit, early decision rather than something you only realise in hindsight, three hours in.

Using this for a literature review

Keshav also describes how the same method scales up: do a first pass on every paper that looks potentially relevant (via search terms and citation chasing) to build a rough map of the area, then a second pass on the ones that seem central, and a full third pass only on the small handful that are genuinely foundational to your own work. This naturally produces a shortlist, rather than forcing you to fully read dozens of papers before you know which ones actually matter.

Tips

Always know why you're reading a given paper before you start — for a specific project detail, for general background, for a literature review — since that shapes how deep a pass is actually worth doing.
Take notes as you go, even brief ones. See the Note-Taking Systems for Research resource for a system that makes this genuinely useful later, not just a pile of forgotten annotations.
Don't be afraid to abandon a paper after pass one. Recognising a paper isn't useful to you is time saved, not time wasted.
For a third pass, actively question the paper rather than just absorbing it — what would you have done differently, what's left unexplained, what would break the argument?

References

Keshav, S. (2007). How to Read a Paper. ACM SIGCOMM Computer Communication Review, 37(3). The original short paper this method comes from — worth reading in full, appropriately enough, using the three-pass method itself.

PDF: How to Read a Paper (Keshav)

HKU Researcher Connect — Reading Papers Efficiently with the Three-Pass Approach