How science is organised
| English | Chinese | Pinyin |
|---|---|---|
| peer review | 同行评审 | tóng háng píng shěn |
| replication | 可重复性 | kě chóng fù xìng |
| declaration of interests | 利益声明 | lì yì shēng míng |
Science is done by institutions, not by lone geniuses
- Research happens in universities, companies and government laboratories, with funders and deadlines.
- Peer review 同行评审 is other specialists checking a paper before publication. It filters; it does not guarantee.
- Replication 可重复性 is how a result becomes trusted: someone else does it and gets the same thing.
What peer review does and does not do
- It checks that the method is sound, the analysis fits the data, and the claims match the evidence.
- It does not repeat the experiment, and it rarely sees the raw data.
- So "peer-reviewed" means "a specialist saw no obvious problem", which is worth a lot and is not proof.
What does peer review actually check?
It rarely sees the raw data and never repeats the work. It filters; it does not guarantee.
Funding
- Asking who paid is normal practice, not cynicism. Funders influence which questions get asked, even without influencing answers.
- Reputable journals require a declaration of interests for exactly this reason.
- ⚠ A study funded by an interested party is not automatically wrong. It is a reason to check the method more carefully, not to dismiss it.
A study funded by a company with an interest in the result should be dismissed.
It is a reason to check the method more carefully, not to discard it. Funding shapes questions more than answers.
Why one study is never enough.
A single result can come from a real effect, from chance, from an unnoticed flaw, or from a sample that happened to be unusual.
Replication separates them: an effect that is real shows up again in someone else's hands, and the other three usually do not.
This is why a news story about "a new study" is the weakest form of scientific evidence, even when the study is good.
A single striking result appears. Why wait for replication?
A real effect shows up again in someone else's hands; the other explanations usually do not.
Consensus is not a vote. It is what remains after many groups have tried to find a problem and failed. That is a different and much stronger thing than most people agreeing.
What is a scientific consensus?
A different and much stronger thing than most people agreeing.
Do not treat a preprint or a conference abstract as equivalent to a peer-reviewed paper. Both are legitimate parts of how science works, and neither has been through the check that the third one has — say which you are citing.
In one sentence, say why "peer-reviewed" is not the same as "proved".
Example: "Reviewers check that the method and reasoning hold up, but they do not redo the experiment, so only replication can confirm the result."