How science is organised · Organisation de la science
| English | Français |
|---|---|
| peer review/pɪə rɪˈvjuː/ | évaluation par les pairs |
| replication/ˌreplɪˈkeɪʃn/ | réplication |
| declaration of interests/ˌdekləˈreɪʃn ɒv ˈɪntrests/ | déclaration d'intérêts |
A published claim still needs checking
- An invented study says a cup sleeve keeps water warmer. Sleeved cups started at 80 °C and bare cups at 60 °C, so their final temperatures cannot isolate the sleeve's effect.
- Recalculating the same records may find an arithmetic mistake. It does not create a new experiment or remove the unequal starting conditions.
Understand what reviewers examine
- Peer review 同行评审 asks relevant specialists to scrutinise methods and reasoning; they may also examine data. Practices and access vary, so the label alone cannot certify every claim.
- A reviewer can flag the unequal starting temperatures and request a better comparison. Receiving that comment is evidence of scrutiny, not evidence that the revised experiment has happened.
What is the role of peer review?
Reviewers examine methods and reasoning and may examine data. Practices vary, and the review label does not guarantee correctness or supply a new independent experiment.
Collect new observations for replication
- Replication 可重复性 involves a new study addressing the finding with an appropriately comparable procedure. A second group should describe its cups, starting temperatures, measurement method and time interval.
- A planned second experiment supplies no new result yet. Once completed, interpret agreement or disagreement using the methods and uncertainty; neither automatically settles every claim.
Why can a new independent experiment help evaluate a striking initial finding?
New observations can test the finding with a comparable method. Agreement or disagreement needs interpretation of the design and uncertainty; replication does not guarantee a simple verdict.
Match the completed action to what it contributes.
Reading and recalculation use existing records; a performed new experiment supplies new observations.
Disclose interests and examine their effects
- A declaration of interests 利益声明 reports relevant relationships, such as donated sleeves or financial support. Examine possible effects on the question, comparison and reporting.
- A manufacturer funding the study does not by itself prove the claim false or true. Look for complete records, appropriate methods and checks that are independent of the claim's promotion.
A study funded by a company with an interest in the result should be dismissed.
Disclose the interest and examine methods, reporting and possible incentives. Funding alone establishes neither falsity nor superiority.
Let teams contribute checks, not authority alone
- Researchers, laboratories, universities and companies can contribute resources and expertise. Clear methods and accessible records help other people assess the work.
- Scientific consensus is broad expert judgement grounded in accumulated evidence. It differs from one senior person's statement, a funder's preference or a vote replacing the data.
Three events have different meanings: a reviewer identifies a design problem; the authors recalculate their old readings; another school performs and reports a new comparable experiment. Only the third supplies new experimental observations.
What is a scientific consensus?
Agreement should be supported by the accumulated evidence and relevant expertise. It remains open to justified revision.
Separate completed work from intentions
- Report which checks have actually happened. “An independent group plans to repeat the test” should not become “an independent group confirmed the result”.
- In the cup case, the original observations may still be reported with their design limitation. A better comparison starts the groups under comparable conditions and records the new measurements.
Identify the method, data, interests and completed checks behind a claim. Review and replication can improve the evidence without making any publication immune to revision.
In one sentence, say why "peer-reviewed" is not the same as "proved".
Example: “Peer review checks methods and reasoning, but the label does not guarantee correctness or remove the need for further evidence.”
An independent group planning an experiment has already confirmed the original result.
False: a plan is not a completed test or a new result.
Which change best addresses the stated unequal starting-temperature problem?
Use comparable starting conditions to avoid confusing their effect with the sleeve effect.