Research methods · Putting it to work
How to read a paper
Try this first
A study has four parts: abstract, methods, results, discussion. Before reading on, answer two things: in which part does the truth live, and in which part does the spin live?
You click a link a podcaster dropped: "new RCT shows this compound drops blood pressure." The abstract opens with that exact sentence, confident and clean. Most people read those two lines and stop — the abstract is right at the top, and it sounds like the answer. But the abstract is the one part of the paper the authors get to write about their own work. It is a pitch. Scroll past it to the methods table and the results, and you often find a different story: a tiny sample, a 12-week window, a 2 mmHg change that could be noise. Same paper, opposite verdict — depending on which part you trusted.
The one idea
Never take the abstract as the finding — it is the authors' spin. Read the methods first (what did they do, to whom, measuring what), then the results table (the real numbers), and treat the discussion as opinion. The single best check: does the abstract's claim match the results table?
A paper is four parts, and they are not equal
Every section has a job, and only two of them are evidence. The methods tell you whether the study was built to find truth or built to find a headline. The results table is the raw record — the actual numbers, before anyone narrated them. The abstract and the discussion are where the authors interpret, frame, and sell. Treat them as the marketing wrapped around the data, not the data itself.
Six questions for the methods section
The methods are where a study is built well or built to flatter itself. You do not need to understand the statistics; you need to ask six plain questions and notice if any answer is missing or weak.
| # | Ask | Weak answer that should worry you |
|---|---|---|
| 1 | Who, and how many? | 20 people; healthy 22-year-olds standing in for sick 60-year-olds |
| 2 | What design? | Survey or observational, dressed up as proof |
| 3 | Randomized & blinded? | Neither — subjects or assessors knew who got what |
| 4 | Outcome: surrogate or hard? | A blood marker moved, not an actual heart attack avoided |
| 5 | Effect size & CI? | Only a p-value or a percent; no absolute number, no interval |
| 6 | Funding / conflicts? | Paid for by the company that sells the product |
Work one, then finish one
Worked: A supplement brand cites "a clinical study" for its joint formula. Run the six. (1) Who: 38 adults, all recruited from the brand's own customer list. (2) Design: single-arm — everyone got the pill, no comparison group. (3) Randomized/blinded: no and no. (4) Outcome: a self-reported "stiffness score," not measured joint function. (5) Effect size: "felt better" with no number and no confidence interval. (6) Funding: the brand ran and paid for it. Five of six answers are weak. Before you ever reach the glowing abstract, the methods have already told you this study can support almost nothing.
Your turn: You open a paper and the abstract leads with a benefit that, when you check the methods, was a secondary outcome — not the primary one the trial was designed to measure. What do you suspect? (That the primary outcome probably failed, so the authors fished for any secondary result that looked good and led with it — classic outcome-switching and spin. Find the primary endpoint and see what it actually did.)
Why this matters
When a podcaster or a brand links "the study" behind a claim, you are not stuck choosing between trusting them and ignoring them. Sixty seconds in the methods and one look at the results table usually settles it. Say you are deciding whether to spend $40 a month on a longevity supplement because a confident host said a trial "proved" it works. You open the paper, skip the abstract, and check: who was studied, was there a control group, was the outcome real or a surrogate marker, and does the headline number actually appear in the results table. Often it doesn't — the abstract claims a win the table never delivered. That mismatch is your answer, and it just saved you $480 a year and the false belief that you are protected.
Recall check · no peeking
- Why should you not trust the abstract as the finding — what is it actually for?
- List the six questions to ask the methods section.
- What is the single best check on a paper, in one sentence?
Explain it back
In one plain sentence, tell a friend why you read the methods before the conclusion.