Lesson 29 of 36 · 14 min
How to read a scientific paper
A paper is an argument supported by methods and data. The abstract tells you what the authors want you to notice; the methods and results tell you what they actually did and found.
The reading order
Do not read a paper from top to bottom like a story.
1. Research question
What exact question was the study designed to answer?
2. Methods
Who was studied, what was compared, how was treatment assigned and how long did follow-up last?
3. Primary endpoint
Which outcome was designated as the main test before the results were known?
4. Results
What was the effect estimate, uncertainty and adverse-event pattern?
5. Limitations
What design features restrict the strength or generalizability of the conclusion?
6. Conclusion
Only now ask whether the authors' interpretation matches the data you just read.
Tables beat adjectives
Look for numbers before words such as impressive, meaningful or promising.
A strong paper lets you identify the numerator, denominator, comparator, effect estimate and uncertainty. Those details make the result inspectable.
If the abstract says an intervention 'significantly improved' an outcome, the next question is how large the improvement was and whether the confidence interval leaves room for a trivial or very large effect.
Primary vs secondary
Not every positive result carries the same evidentiary weight.
Primary endpoints are usually the outcomes around which the study was designed and powered. Secondary and exploratory analyses can be informative, but the more outcomes tested, the easier it becomes to find an apparently positive result by chance.
A headline built around a secondary subgroup result should therefore be read differently from a prespecified primary endpoint.
Paper-reading rule
Methods first. Results second. Interpretation last.
If you read the conclusion before understanding the design, you are more likely to inherit the authors' framing instead of evaluating it.
Key takeaways
- ✓Start with the research question and methods, not the conclusion paragraph.
- ✓Identify the primary endpoint before interpreting positive secondary findings.
- ✓Read effect estimates and uncertainty rather than relying on descriptive adjectives.
- ✓Check population, comparator, follow-up and missing data before generalizing the result.
- ✓A well-written conclusion can still be broader than the underlying design supports.
Further reading
Lesson finished
Ready for the checkpoint?
The quiz opens on its own page. Get every answer correct to mark this lesson complete and continue.
Take lesson quiz →Previous
How to research a peptide ←
Next lesson
Endpoints, effect size & uncertainty →
Continue when you are ready, or come back to this lesson whenever you need a refresher.