ThreePassPaperReading
title: Three-Pass Paper Reading type: technique created: 2026-05-21 last_updated: 2026-05-21 related: ["radar/techniques/SeniorEngineerHandbook", "radar/techniques/ExploringGenerativeAI"] sources: ["http://ccr.sigcomm.org/online/files/p83-keshavA.pdf"] radar_quadrant: Techniques radar_ring: Assess radar_position: inner
A structured method for reading academic and technical papers efficiently, described by S. Keshav in the ACM SIGCOMM Computer Communication Review. The method prevents time waste on papers that don't merit full attention while ensuring thorough understanding of those that do.
First pass — 5-10 minutes. Read only: title, abstract, introduction, section headings, conclusions, and references (scan for familiar names). Goal: determine the paper's category (measurement, analysis, description of a system), context (which other papers it relates to), correctness (do the assumptions seem valid?), contribution (what does it claim to add?), and clarity (is it well written?). Decision: stop here, or continue to the second pass.
Second pass — up to 1 hour. Read the paper carefully, ignoring proofs and mathematical derivations. Look at figures and diagrams closely — they often carry the central result. Jot margin notes. Identify terms you don't understand and mark them. Goal: grasp the main thrust of the paper. Decision: understand the paper well enough to summarise it to someone else, or continue to the third pass.
Third pass — 4-5 hours for beginners, 1 hour for experienced readers. Re-implement the paper mentally: challenge every assumption, identify every implicit claim, consider alternative approaches. Goal: reconstruct the paper from memory — understand its innovations and its weaknesses. This pass is only worthwhile for papers that will directly influence your own work.
Without a structured approach, reading a paper means reading it linearly from start to finish — committing the full time regardless of whether the paper merits it. The three-pass method creates decision gates that allow early exit. Most papers only need the first pass. A fraction need the second. Very few need the third.
This compounds over a career: a researcher or practitioner who reads two papers per week but applies the three-pass method extracts signal from 100 papers per year rather than grinding through 20 in full.
The method applies to any long-form technical document: white papers, RFC specifications, architecture decision records, and long-form engineering blog posts. The first-pass heuristics (abstract, headings, conclusions) transfer directly.
Three-Pass Paper Reading sits in the Assess ring of the Techniques quadrant, at inner position. First studied via Keshav's SIGCOMM paper. The technique is widely cited in academic communities and requires no tooling — only the discipline of applying the three-gate structure. That citation record is third-party evidence, not first-person use: the method has not been adopted as the user's own reading practice, and the Adopt ring requires that, so it stays at Assess. Inner position reflects that any practitioner who regularly reads technical literature could apply this method as a default, at zero cost and with compounding return. The gate to Trial is first-person adoption of the three-pass structure on real papers.
