Reading a research summary

Why this matters

Sooner or later someone will send you a study, a white paper or a news story and ask, "Should we do this?" A new material is "40% stronger", a new tool "cuts errors in half", a supplement is "linked to better sleep". You do not need a PhD to read research usefully. You need a routine for pulling out three things: the claim, the method and the limits. With that routine you can decide whether a result applies to your situation, or whether it needs a closer look.

The parts of a research paper

Most scientific papers follow the same structure:

• Title and abstractThe short summary at the start of a paper giving the question, method and main result.: a short summary of the question, method and main result. Start here, but do not stop here, because abstracts are written to sound impressive.
• Introduction: why the question matters and what was already known.
• Methods: exactly what was done: who or what was studied, how many, what was measured and how.
• Results: the data, usually in tables and figures.
• Discussion and limitations: what the authors think the results mean, and what could be wrong with them.
• References: the earlier work it builds on.

A useful reading order is: abstract, figures and tables, methods, then discussion. Read the introduction if you need background.

Pull out the claim

Write the main claim in one sentence, in your own words, with numbers if there are any. For example: "In a test of 30 steel samples, a new heat treatment increased tensile strength by about 12% compared with the standard treatment."

Notice the difference between the headline ("40% stronger") and the actual result. Press releases and articles often round up, drop the conditions, or report the best case.

Check the method

Ask these questions:

• What kind of study is it? A controlled experiment (researchers change one thing on purpose) supports cause and effect. An observational studyA study where researchers record what happens without changing anything, so it can show a link but not cause. (researchers just record what happens) can only show a link, for the reasons you learned in the correlation lesson. A review or meta-analysisA study that combines the results of many earlier studies, often giving the strongest evidence. combines many studies, which is often the strongest evidence.
• How big was the sample? 12 people or 12,000? Small samples give noisy results.
• Was there a control or comparison group?
• Who or what was studied? Results on mice, on one factory's machines, or on university students may not apply to your situation.
• How was the outcome measured? Self-reported "felt better" is weaker than a measured value.

Find the limits

Good papers list their own limitations. Read that section closely. Also look for:

• Effect sizeHow big a result is in practice; a statistically significant effect can still be too small to matter., not just significance. A result can be "statistically significant" and still be too small to matter in practice. A 0.5% improvement that costs a fortune may not be worth it.
• Uncertainty. Look for ranges such as "12% (95% confidence intervalA range, such as 4% to 20%, that the true effect probably falls within; a wide range means more uncertainty. 4% to 20%)". A wide range means the true effect could be much smaller.
• Funding and conflicts of interest. Who paid for the study? Funded research is not automatically wrong, but it is a reason to check carefully.
• Peer review and publication venue. Was it published in a reputable peer-reviewed journal, or is it a preprintA paper shared publicly before peer review, best treated as early evidence. that has not been reviewed yet? Preprints are not bad, but treat them as early.
• Replication. Has anyone else found the same thing? One study is a starting point, not a settled fact.

Worked example

A short summary reads: "Workers who used standing desks reported 30% less back pain."

• Claim: standing desks reduce back pain by 30%.
• Method: you find it was a survey of 85 office workers who chose standing desks, compared with those who did not. Pain was self-reported.
• Limits: observational (people who choose standing desks may already be more active), small sample, self-reported outcome, no measurement of how long people actually stood.
• Your verdict: "An interesting link worth watching, but it does not show that standing desks cause less back pain. A controlled trial would be needed."

That three-line summary is exactly what a manager or teammate needs from you.

Common mistakes

• Reading only the headline or abstract. The details that matter are in the methods.
• Treating a link as a cause. Check whether it was an experiment.
• Ignoring who was studied. Results may not transfer to your population, material or system.
• Trusting a single study. Look for reviews or replications.
• Dismissing everything. Every study has limits; the goal is to weigh evidence, not to reject it all.

Quick recap

• Read abstract, figures, methods, then discussion.
• Summarise the claim in one sentence with numbers.
• Check study type, sample size, comparison group and how outcomes were measured.
• Look for effect size, uncertainty, conflicts of interest and replication.

Try it

Open the Do tab. You will pick a short, free research summary or article and write a one-paragraph claim / method / limits summary of it.