Every day we encounter claims about health, science, history, money and public life. Some are supported by strong evidence; others rely on anecdotes, selective statistics or confident repetition. Learning to evaluate evidence is a practical form of intellectual self-defence.
What counts as evidence?
Evidence is information used to support or challenge a claim. Its usefulness depends on the question. A historical document may be valuable evidence about what a person wrote, but not necessarily proof that the account is accurate. A laboratory experiment can test a mechanism under controlled conditions, but may not show how it works in everyday life.
The first step is to identify the exact claim. Broad statements often hide several smaller claims that require different kinds of evidence.
Match the evidence to the question
Different questions call for different methods. Randomised controlled trials can help estimate the effects of interventions. Observational studies can reveal patterns in real-world populations. Historical research evaluates documents, material evidence and context. Qualitative research can illuminate experience, meaning and social processes.
No single research design answers every question. A good evaluation asks whether the method is suitable for the claim.
Correlation is not causation
If two things occur together, one does not automatically cause the other. A third factor may influence both, the direction of influence may be reversed, or the association may be coincidental.
Causal claims require stronger reasoning, such as a plausible mechanism, appropriate comparisons, control of confounding factors and evidence that alternative explanations are less likely.
Look beyond a headline
When reading a study or report, ask:
- Who produced it, and what expertise do they have?
- What question was actually investigated?
- How many people, cases or observations were included?
- How were participants or sources selected?
- What limitations did the authors acknowledge?
- Does the conclusion match the results?
- Can independent researchers reproduce or corroborate the finding?
A press release or headline may simplify a study. Whenever possible, inspect the original research and independent assessments.
Bias, uncertainty and replication
Bias can enter through study design, measurement, participant selection, analysis or publication decisions. Researchers use methods such as preregistration, blinding, transparent data practices and replication to reduce some risks.
A single study rarely settles a complex question. Confidence grows when multiple independent lines of evidence point in the same direction. Scientific conclusions can change when better data or methods become available; that is a strength of a self-correcting process, not proof that all claims are equally uncertain.
Common misunderstandings
- A source can be reputable and still make an error.
- Peer review is a quality-control process, not a guarantee of truth.
- A large sample does not automatically remove bias.
- An expert opinion is useful, but the evidence and reasoning still matter.
- Uncertainty does not mean we know nothing; evidence can support conclusions with different degrees of confidence.
Frequently asked questions
What is the most reliable source of information?
There is no universal source. Prefer relevant primary evidence, transparent methods, qualified expertise and independent corroboration.
Is a peer-reviewed paper always trustworthy?
Peer review improves scrutiny but cannot guarantee that a study is correct, unbiased or reproducible.
How can I spot misinformation?
Check the original source, verify the date and context, compare independent reporting, inspect the evidence and be cautious of sensational claims that offer certainty without showing their reasoning.
Sources and further reading
- National Academies: Reproducibility and Replicability in Science
- Stanford Encyclopedia of Philosophy: Scientific Method
- Cochrane: Understanding evidence
Part of the Knowledge Wiki by Melvin Varghese.
