Read a study
Read for the chain of evidence. A memorable result is easier to assess when you can reconstruct who was studied, what was compared and what was measured.
Start with the claim
Write the research question in one sentence. Identify the population, proposed relationship and outcome. Is the paper describing a pattern, predicting an outcome, interpreting experience or estimating the effect of an intervention? Compare the authors’ conclusion with the question their design can answer.
Operationalization makes abstract terms inspectable. Preregistration can help distinguish planned tests from decisions made after seeing data.
Reconstruct the comparison
Find the recruitment method, eligibility rules, group allocation and timing. Random sampling concerns who enters a sample; random assignment concerns who receives which condition. Neither word substitutes for the other. For observational comparisons, identify plausible alternative explanations and how the researchers addressed them.
- Who was excluded, declined or lost to follow-up?
- Were the same people measured repeatedly? Could order, practice or carryover matter?
- What differs between groups besides the proposed cause?
Randomized experiments · Quasi-experiments · Longitudinal studies
Inspect the measures
Look beyond the construct’s name to the actual task, items, scoring and observation rules. Ask what evidence supports the intended interpretation in this population and setting. Consistent scores do not automatically establish that the measure captures the intended construct.
Read the uncertainty
Find the estimated effect and its units, not just whether a threshold was crossed. Check the interval, sample size, missing-data handling and number of analyses. A small p value does not give the probability that the hypothesis is true, and a large one does not establish no effect. Ask how much the estimate could change under reasonable analytical choices.
Effect sizes · Confidence intervals · Hypothesis testing · Statistical power
Check ethics and openness
Look for consent arrangements, ethics review, exclusions and any deception or debriefing. Check whether a protocol, materials, code or appropriately protected data are available. Openness makes some decisions inspectable; it does not certify the study’s design or conclusions. Sharing must respect consent and privacy.
Informed consent · Research ethics review · Open data and materials · Replication
Practice with a constructed abstract
Teaching scenario: Forty volunteers choose either a new memory app or their usual study routine. One week later, the app group remembers more words on a researcher-created list. The report gives p = .04 but no effect estimate, interval, allocation procedure beyond self-selection, or validation evidence for the list. These details are invented for this reading exercise.
Compare your reading with the method notes
- The comparison is self-selected, so motivation and prior memory ability are plausible alternative explanations. The abstract does not isolate a causal effect.
- The outcome is recall on this particular list. Generalizing to everyday memory requires additional measurement and transfer evidence.
- The p value alone does not reveal the size or importance of the difference. Request the estimate, interval, group sizes and missing-data information.
- A short abstract may omit material available in the full paper. Treat omitted information as a question to check, rather than proof the procedure never occurred.
Try a claim-choice exercise or return to the method directory.