How to read a research paper
A practical order of operations for judging a study without taking its abstract at face value.
Abstracts are written to be persuasive, so reading one and stopping is how most misreading happens. A more reliable order is: methods first, then results, then the abstract — checking whether the conclusion the authors drew is the one their data support.
A short checklist covers most of it. What kind of study is it — in vitro, animal, observational, randomised — which fixes the ceiling on any causal claim, per the evidence hierarchy. How many subjects, and was the primary endpoint pre-specified or chosen afterward? Was the endpoint a real outcome or a surrogate? Is there a control, and is it a fair one? For animal work, does the model resemble the human situation in the ways that matter? And who funded it — not disqualifying, but relevant.
Then ask what would have to be true for the conclusion to hold, and whether the paper establishes it. One study is a data point; synthesis and replication are what turn data points into knowledge.
Related articles
- MethodsPeer review and the evidence hierarchyWhat peer review does, what it does not do, and how evidence is ranked.
- MethodsBiomarkers and surrogate endpointsMeasuring something easy in place of something that matters — useful, and quietly dangerous.
- MethodsSystematic reviews and meta-analysisSynthesising a whole literature rather than trusting a single study.
- MethodsReproducibility in biomedical researchWhy a substantial share of published findings do not replicate — and what reduces the risk.
- MethodsPreclinical modelsWhat animal models can and cannot tell you about human regeneration.
- MethodsBench to bedsideThe pathway from laboratory observation to clinical use, and where it usually breaks.