Reading a research paper like a novel — start to finish, every word — is slow and demoralizing. Experienced researchers don't do it. They read in passes, deciding at each step whether the paper is worth more of their time. Here's the widely used three-pass method, adapted for anyone.
Pass 1 — the five-minute skim (should I read this at all?)
- Read the title, abstract, and section headings.
- Read the introduction's last paragraph and the conclusion.
- Glance at the figures and tables.
After five minutes you should be able to say what problem the paper tackles, what they did, and what they claim to have found. If it's not relevant, stop here — you've saved yourself an hour.
Pass 2 — the content read (what did they actually do?)
Now read the whole paper, but don't get stuck on heavy math or proofs. Focus on the method (what they did and why), the results (what they found), and the figures (which usually carry the real story). Note anything you don't understand or want to check later. This pass takes about an hour and is enough to summarize the paper to someone else.
Pass 3 — the deep read (could I reproduce it?)
Only for papers central to your work. Here you go line by line, question every assumption, check the equations, and mentally re-run the experiment. The goal is to understand it well enough to spot hidden flaws and reproduce the results. This is slow — budget several hours.
What each section really tells you
- Abstract — the whole paper in miniature; read it first, always.
- Introduction — the problem and why it matters; the last paragraph often lists the contributions.
- Methods — how they did it; the section that decides whether the results are trustworthy.
- Results — what they found, usually best understood through the figures.
- Discussion — what the authors think it means (their interpretation, not fact).
- Limitations — what the study can't tell you; never skip this.
Judging quality quickly
- Is the sample size reasonable, and is there a control or baseline?
- Do the conclusions actually follow from the results, or overreach?
- Is the data or code shared so others can check it?
- Is it peer-reviewed, or a preprint that hasn't been vetted yet?
- Do the authors state their limitations honestly?
AI summaries (like PaperFinder's) are a great Pass 1 accelerant — they surface the problem, method, results and limitations in seconds. But they can miss nuance or state things too confidently, so treat a summary as a map, not the territory: use it to decide what to read, then verify against the paper itself.