How to Read Science News and Evaluate Claims

How to read Science news is now part of basic scientific literacy. Students encounter claims about health, climate, space, artificial intelligence, nutrition, pollution, genetics, energy, medicine, wildlife, technology and the environment long before they read a scientific paper. Headlines can be accurate, oversimplified, preliminary, sensational, commercially framed or simply wrong. The student therefore needs a repeatable way to ask what was actually measured, who or what was studied, how strong the evidence is, and whether the headline says more than the research supports.

This Advanced Science Tutorials guide is written for parents and students in Sengkang, Punggol and across Singapore who search for how to read Science news, evaluate scientific claims, how to know if a study is reliable, correlation vs causation, peer review, sample size, scientific evidence, misinformation, research study and fact checking Science. It is a cross-level Science-literacy owner from Primary curiosity through PSLE and Secondary G1, G2 and G3.

This article does not ask students to distrust Science or distrust news. It teaches calibrated trust: stronger claims need stronger evidence, one study rarely settles a complex question, methods matter, sources matter, replication matters, and uncertainty is not the same as ignorance. The site’s broader evidence architecture remains at Scientific Method, Evidence & Measurement | How Science Knows, while the new Scientific Method for Students owner handles school investigations.

The 60-second Science-news check

What is the exact claim?
What evidence is offered?
Who or what was studied?
How many observations or participants were included?
Was the study experimental, observational, simulated or based on secondary data?
What was actually measured?
Does the result show correlation or stronger evidence of causation?
Was the result replicated or is it one early study?
Does the headline match the study’s scope?
Who produced the claim, and is there a conflict of interest?
What uncertainty or limitation is acknowledged?
What would count as evidence against the claim?

Scientific literacy in one sentence

Scientific literacy is the ability to understand scientific ideas well enough to evaluate evidence, interpret claims, recognise uncertainty and make informed decisions without pretending that every question has a simple or final answer.

Primary 1 and Primary 2: start with observation versus claim

Younger children can begin with a simple distinction: “I saw this” is not the same as “this always happens”. A single observation can be real without supporting a universal rule. Parents can ask: What did you actually observe? How many times? Could something else explain it?

This early habit builds the foundation for later evidence reasoning without turning childhood curiosity into formal research critique.

Primary 3 and Primary 4: evidence before explanation

Primary Science students can distinguish observations from inferences, compare fair and unfair tests, and learn that a conclusion should match the evidence. If a plant grew taller in one setup, the student should ask what else changed before claiming the tested factor caused the difference.

Use the site’s Primary 3 experiments, fair tests and evidence and Primary 4 experiments, data and conclusions routes for level-specific practice.

Primary 5 and Primary 6: claims, variables and data

Older Primary students can ask whether a comparison is fair, whether a graph supports a conclusion and whether a statement goes beyond the tested conditions. These are the same habits needed when reading public Science claims.

PSLE questions already train students to interpret data, evaluate experiments and explain relationships. Science-news literacy is therefore not a separate subject; it is real-world transfer of examination reasoning.

Secondary G1, G2 and G3: evidence quality becomes more explicit

Lower Secondary students can distinguish correlation from causation, think about sample size, measurement quality, confounding variables, replication and uncertainty. They can also learn the difference between a research paper, a preprint, a press release, a news story and a social-media post.

The official G1 and G2/G3 Lower Secondary Science syllabuses remain the curriculum references; this article extends those reasoning habits into public information.

Headline

Core idea. A headline is a compressed summary designed to attract attention, not a complete account of methods and uncertainty.

Common mistake. Students may treat the headline as if it were the scientific conclusion.

Reading move. Compare the headline with the article body and then with the study abstract if available.

Question to ask. Does the headline add certainty, universality or causation that the source did not claim?

A strong Science-news reader should be able to apply this idea without rejecting the entire article or accepting it wholesale. The goal is calibrated confidence: identify what the evidence supports, what remains uncertain, and what additional evidence would materially change the conclusion.

For parents, ask the child to locate the original evidence before debating the conclusion. For tutors, compare two headlines built from the same study and ask which wording is better calibrated. This turns media literacy into scientific reasoning rather than general scepticism.

Claim

Core idea. A claim is a statement that can be evaluated against evidence.

Common mistake. Students sometimes confuse opinion, prediction and observation with scientific claim.

Reading move. Rewrite the claim in the narrowest testable form before evaluating it.

Question to ask. What exact words would need to be true for the claim to hold?

Evidence

Core idea. Evidence is information used to support or challenge a claim.

Common mistake. Students may assume any number, graph or expert quote automatically counts as strong evidence.

Reading move. Ask where the data came from, how it was measured and whether it directly addresses the claim.

Question to ask. What evidence would be stronger than what is currently offered?

Source

Core idea. The source is the person, organisation or publication presenting the information.

Common mistake. Students may judge source quality only by visual professionalism or popularity.

Reading move. Identify whether the source is a journal, university, government agency, news outlet, company, advocacy group or individual account.

Question to ask. What incentives, expertise and editorial processes shape this source?

Primary source

Core idea. A primary scientific source reports original research or data.

Common mistake. Students may think a news story is the original study because it links to scientists.

Reading move. Look for the research paper, official dataset or direct study report when practical.

Question to ask. Can the article’s strongest claim be traced back to a primary source?

Secondary source

Core idea. A secondary source summarises, interprets or synthesises primary research.

Common mistake. Students may assume secondary means unreliable.

Reading move. Good reviews and reputable reporting can be highly useful, but they add a layer of interpretation.

Question to ask. Does the secondary source represent the primary evidence accurately?

Press release

Core idea. A press release communicates research or institutional news to media and public audiences.

Common mistake. Students may assume university branding guarantees neutral framing.

Reading move. Compare press-release wording with the paper’s abstract, limitations and effect size.

Question to ask. Did the press release simplify or strengthen the claim?

Preprint

Core idea. A preprint is a research manuscript shared before formal peer review.

Common mistake. Students may think preprint means false or, conversely, already fully validated.

Reading move. Treat preprints as accessible early evidence whose methods and conclusions have not yet passed journal peer review.

Question to ask. Has the work since been peer reviewed, corrected or withdrawn?

Peer review

Core idea. Peer review involves evaluation by other experts before or after publication, depending on system.

Common mistake. Students may think peer-reviewed means guaranteed true.

Reading move. Peer review can catch problems but cannot eliminate all error, bias or fraud.

Question to ask. Has later evidence replicated or challenged the finding?

Journal

Core idea. A scientific journal publishes research under specific editorial and review policies.

Common mistake. Students may think all journals have equal standards.

Reading move. Check the journal, publisher, indexing, correction policy and whether the paper type is research, review, commentary or editorial.

Question to ask. Is the article actually reporting original evidence?

Retraction

Core idea. A retraction signals that a published paper should no longer be relied on as part of the literature in the usual way.

Common mistake. Students may think retraction means Science as a whole failed.

Reading move. Retractions are part of scientific correction when serious problems are identified.

Question to ask. Was the paper corrected, retracted or replaced, and do newer sources still cite it appropriately?

Correction

Core idea. A correction updates an error without necessarily invalidating the whole study.

Common mistake. Students may interpret any correction as proof the research was worthless.

Reading move. Read what changed and whether the conclusion was affected.

Question to ask. Does the correction alter the effect size, methods or main claim?

Replication

Core idea. Replication tests whether findings can be observed again under similar or related conditions.

Common mistake. Students may think one famous study is enough.

Reading move. Repeated independent evidence increases confidence, especially when different teams and methods converge.

Question to ask. Has the result been reproduced beyond the original group?

Systematic review

Core idea. A systematic review uses a defined method to search, select and synthesise multiple studies.

Common mistake. Students may think a review simply collects papers the author likes.

Reading move. Strong reviews specify search strategies, inclusion criteria and appraisal methods.

Question to ask. Were studies selected systematically and were quality differences considered?

Meta-analysis

Core idea. A meta-analysis statistically combines results across multiple studies when appropriate.

Common mistake. Students may think more studies automatically mean a better answer.

Reading move. The result depends on study quality, heterogeneity, publication bias and model choices.

Question to ask. Are the combined studies sufficiently comparable?

Sample

Core idea. A sample is the subset actually observed or measured.

Common mistake. Students may treat the sample as identical to the whole population.

Reading move. Ask who or what was included and excluded.

Question to ask. Does the sample resemble the population the headline talks about?

Population

Core idea. The population is the wider group the study aims to understand.

Common mistake. Students may generalise from one age group, country or species to everyone.

Reading move. Check whether the target population matches the sample.

Question to ask. Is the claim broader than the population studied?

Sample size

Core idea. Sample size affects statistical precision and the ability to detect effects.

Common mistake. Students may think a large sample automatically removes bias.

Reading move. A huge biased sample can still answer the wrong population question.

Question to ask. Was the sample large enough and appropriately selected?

Sampling bias

Core idea. Sampling bias occurs when the selected sample systematically differs from the intended population.

Common mistake. Students may focus only on number of participants.

Reading move. Ask how participants, locations or observations were recruited or selected.

Question to ask. Who had no chance of appearing in the sample?

Selection bias

Core idea. Selection bias arises when inclusion into groups or analysis is related to variables that affect the outcome.

Common mistake. Students may treat all group comparisons as equivalent.

Reading move. Examine how groups formed and whether the process created systematic differences.

Question to ask. Could group membership be linked to the outcome independently of the claimed cause?

Random sampling

Core idea. Random sampling gives members of a population a defined chance of selection.

Common mistake. Students may confuse random sampling with random assignment.

Reading move. Sampling addresses representation; assignment addresses causal comparison.

Question to ask. Was randomness used to select participants, assign treatment, or both?

Random assignment

Core idea. Random assignment allocates participants to conditions by chance in experiments.

Common mistake. Students may think random assignment makes the sample representative of the whole world.

Reading move. It strengthens internal causal comparison but does not automatically solve generalisability.

Question to ask. Was assignment truly random and were groups comparable after assignment?

Control group

Core idea. A control group or condition provides a baseline for interpreting an intervention.

Common mistake. Students may think every study requires one identical kind of control group.

Reading move. The right comparator depends on the question: placebo, usual care, no treatment, alternative treatment or baseline.

Question to ask. Is the comparison fair and relevant?

Placebo

Core idea. A placebo resembles an intervention without its active component and can help separate treatment effects from expectations.

Common mistake. Students may think placebo means ‘fake’ and therefore scientifically useless.

Reading move. Placebo controls can strengthen blinded trials when ethically appropriate.

Question to ask. Were participants and assessors blinded to condition?

Blinding

Core idea. Blinding limits knowledge of group assignment to reduce expectation and measurement bias.

Common mistake. Students may think every experiment can be perfectly blinded.

Reading move. Some interventions are difficult or impossible to blind, so outcome measurement procedures become especially important.

Question to ask. Who knew the group assignments?

Double blind

Core idea. Double-blind commonly means participants and investigators or outcome assessors are unaware of assignment, but terminology varies.

Common mistake. Students may treat the label as a guarantee without checking who was actually blinded.

Reading move. Read the methods rather than relying on the phrase alone.

Question to ask. Which people were blinded, and could unblinding have occurred?

Observational study

Core idea. An observational study measures exposures and outcomes without assigning the exposure.

Common mistake. Students may treat observational associations as experimental proof.

Reading move. Observational studies can be powerful, especially when experiments would be unethical or impossible, but causal inference needs extra care.

Question to ask. What confounders could explain the association?

Randomised controlled trial

Core idea. An RCT randomly assigns interventions to strengthen causal inference.

Common mistake. Students may think every RCT gives a universal answer.

Reading move. Trial quality still depends on sample, adherence, blinding, outcome choice, analysis and generalisability.

Question to ask. Was the trial large enough, well conducted and relevant to the claimed population?

Cohort study

Core idea. A cohort study follows groups over time to compare exposures and outcomes.

Common mistake. Students may assume time order alone proves causation.

Reading move. Cohorts can establish temporal sequence but remain vulnerable to confounding.

Question to ask. How were exposure and outcome measured, and what differences existed between groups?

Case-control study

Core idea. A case-control study compares people with an outcome to controls and looks backward for exposures.

Common mistake. Students may think it directly measures disease risk in the same way as a cohort.

Reading move. The design is efficient for rare outcomes but requires careful control selection.

Question to ask. Could recall or selection bias distort exposure estimates?

Cross-sectional study

Core idea. A cross-sectional study measures variables at one time point.

Common mistake. Students may infer which variable came first.

Reading move. Cross-sectional data can describe prevalence and association but often cannot establish temporal sequence.

Question to ask. Can the design tell cause from consequence?

Case report

Core idea. A case report describes an individual or small number of cases.

Common mistake. Students may generalise dramatic cases to common outcomes.

Reading move. Case reports can identify unusual signals but cannot estimate population frequency reliably.

Question to ask. Is this an anecdotal signal or population-level evidence?

Animal study

Core idea. Animal studies can reveal mechanisms and safety signals under controlled conditions.

Common mistake. Students may assume effects in animals translate directly to humans.

Reading move. Species differences in metabolism, dose and biology limit direct translation.

Question to ask. Has the effect been studied in humans?

Cell study

Core idea. In-vitro cell studies isolate biological mechanisms in controlled conditions.

Common mistake. Students may assume an effect in cells means a food, drug or chemical will have the same effect in a whole person.

Reading move. Concentration, metabolism, tissue interactions and dosing can differ greatly in living organisms.

Question to ask. Is the tested concentration realistic in the real-world system?

Simulation study

Core idea. Simulations explore consequences of assumptions and models.

Common mistake. Students may treat model output as direct observation.

Reading move. Simulation results depend on equations, parameter choices and validation.

Question to ask. How well does the model reproduce real data?

Secondary-data study

Core idea. Researchers can analyse datasets collected for other purposes.

Common mistake. Students may assume large databases are automatically complete and unbiased.

Reading move. Definitions, missingness, measurement changes and selection still matter.

Question to ask. What was the dataset originally designed to measure?

Correlation

Core idea. Correlation means variables vary together statistically.

Common mistake. Students may treat correlation as proof one variable causes the other.

Reading move. Shared causes, reverse causation and coincidence can produce association.

Question to ask. What alternative causal structures fit the same pattern?

Causation

Core idea. Causation means changing one factor would change an outcome under defined conditions.

Common mistake. Students may demand an experiment for every causal claim.

Reading move. Causal inference can combine experiments, natural experiments, mechanistic evidence and strong observational designs.

Question to ask. What evidence rules out the strongest alternative explanations?

Confounding

Core idea. A confounder is related to both exposure and outcome and can distort the observed relationship.

Common mistake. Students may assume controlling for one obvious variable solves all confounding.

Reading move. Unmeasured and residual confounding can remain.

Question to ask. Which variables could plausibly influence both sides?

Reverse causation

Core idea. Reverse causation occurs when the supposed outcome may influence the supposed cause.

Common mistake. Students may infer direction from association alone.

Reading move. Time order and longitudinal data can help distinguish direction.

Question to ask. Could the relationship run the other way?

Effect size

Core idea. Effect size describes the magnitude of a difference or association.

Common mistake. Students focus only on whether a result is statistically significant.

Reading move. A tiny effect can be statistically significant in a large sample but practically unimportant.

Question to ask. How large is the effect in understandable units?

Absolute risk

Core idea. Absolute risk is the actual probability of an outcome in a group.

Common mistake. Students may be impressed by relative changes without knowing baseline risk.

Reading move. Going from 1 in 10,000 to 2 in 10,000 doubles relative risk but changes absolute risk by 1 in 10,000.

Question to ask. What are the event rates in both groups?

Relative risk

Core idea. Relative risk compares outcome probability between groups.

Common mistake. Students may interpret a large percentage increase as a large absolute change.

Reading move. Always pair relative risk with baseline or absolute risk when possible.

Question to ask. What does the percentage mean in counts per 100, 1,000 or 10,000?

Odds ratio

Core idea. An odds ratio compares odds rather than probabilities.

Common mistake. Students may read it as if it were always identical to relative risk.

Reading move. When outcomes are common, odds ratios can look more dramatic than risk ratios.

Question to ask. What measure did the study actually report?

Statistical significance

Core idea. Statistical significance evaluates compatibility of data with a statistical null model under assumptions.

Common mistake. Students may think significant means important or certainly true.

Reading move. Significance depends on sample size, variability, model assumptions and analysis choices.

Question to ask. What is the effect size and uncertainty interval?

p-value

Core idea. A p-value is the probability of data at least as extreme under a specified null model, not the probability the hypothesis is false.

Common mistake. Students often say p=0.03 means a 3% chance the result is wrong.

Reading move. Interpret p-values alongside design, effect size and uncertainty rather than as truth meters.

Question to ask. What was the analysis plan and how many tests were run?

Confidence interval

Core idea. A confidence interval gives a range derived from a procedure with defined long-run coverage properties.

Common mistake. Students treat the interval as an exact probability range for the true value in every interpretation.

Reading move. The interval reflects estimate precision under model assumptions.

Question to ask. Is the interval narrow enough to distinguish practically important effects?

Uncertainty interval

Core idea. Many fields use intervals to express uncertainty around estimates or model outputs.

Common mistake. Students may treat the centre estimate as exact.

Reading move. Read the full range and what sources of uncertainty are included.

Question to ask. Could plausible values support different practical conclusions?

Multiple comparisons

Core idea. Testing many hypotheses increases chances of apparently unusual results.

Common mistake. Students may focus on the one positive result among dozens of tests.

Reading move. Correction, preregistration and replication help manage multiplicity.

Question to ask. How many outcomes, subgroups or analyses were tested?

Subgroup analysis

Core idea. Subgroup results examine effects in subsets of participants.

Common mistake. Students may treat one striking subgroup as a separate established finding.

Reading move. Small subgroups have lower precision and more opportunities for chance patterns.

Question to ask. Was the subgroup specified in advance and replicated?

Preregistration

Core idea. Preregistration records hypotheses and analysis plans before results are known.

Common mistake. Students may think preregistration prevents all bias.

Reading move. It increases transparency but cannot fix poor design or measurement.

Question to ask. Were primary outcomes and analyses prespecified?

Registered report

Core idea. A registered report can receive in-principle acceptance based on question and methods before results are known.

Common mistake. Students may think publication then guarantees correct results.

Reading move. The format reduces publication bias toward positive findings but still requires good execution.

Question to ask. Were deviations from the planned method documented?

Publication bias

Core idea. Publication bias occurs when studies with positive or striking results are more likely to appear.

Common mistake. Students may assume the published literature contains every study performed.

Reading move. Funnel plots, registries and systematic searches can reveal missing negative evidence.

Question to ask. Could unpublished null results change the conclusion?

Researcher degrees of freedom

Core idea. Analysts often make many legitimate choices about exclusions, outcomes and models.

Common mistake. Students may think one reported analysis was the only possible analysis.

Reading move. Preregistration and robustness checks help reveal sensitivity to choices.

Question to ask. Would reasonable alternative analyses produce similar conclusions?

HARKing

Core idea. HARKing means presenting a hypothesis after results are known as if it had been predicted beforehand.

Common mistake. Students may think any plausible explanation after the fact counts as successful prediction.

Reading move. Post-hoc explanations can generate hypotheses but need new tests.

Question to ask. Was the hypothesis specified before seeing the data?

Replication crisis

Core idea. Some fields have found that influential results do not always replicate reliably.

Common mistake. Students may use this to conclude all Science is unreliable.

Reading move. The correct lesson is that methods, incentives and replication affect confidence, and Science can improve by detecting these problems.

Question to ask. What changes in practice have followed replication concerns?

Reproducibility

Core idea. Reproducibility can mean obtaining compatible results using the same data/code or repeating an analysis, depending on field.

Common mistake. Students use reproducibility and replication as exact synonyms.

Reading move. Check how the source defines each term.

Question to ask. Are data, code and methods available enough for verification?

Open data

Core idea. Open data allows others to inspect or reuse research data where privacy and ethics permit.

Common mistake. Students may assume open means automatically correct.

Reading move. Transparency improves scrutiny but does not guarantee quality.

Question to ask. Are variables and missing values documented?

Open code

Core idea. Open analysis code makes computational steps inspectable.

Common mistake. Students may assume code availability proves analysis validity.

Reading move. Code can still contain errors or encode questionable choices.

Question to ask. Has the analysis been independently checked?

Conflict of interest

Core idea. A conflict of interest is a relationship that could influence judgment or create perceived influence.

Common mistake. Students may assume industry funding automatically makes research false.

Reading move. Funding source is relevant context, but methods and independent replication still matter.

Question to ask. Were conflicts disclosed and are findings supported by independent evidence?

Funding source

Core idea. Funding identifies who paid for the work.

Common mistake. Students may ignore funding or overreact to it.

Reading move. Consider funding alongside study design, transparency and replication.

Question to ask. Would the conclusion remain credible if independently reproduced?

Expertise

Core idea. Expertise matters most when it matches the specific claim.

Common mistake. Students may accept a famous person speaking far outside their field.

Reading move. Ask whether the expert’s research or professional experience is relevant to the topic.

Question to ask. Is the claim supported by evidence beyond the person’s authority?

Consensus

Core idea. Scientific consensus reflects convergence of evidence and expert judgment in a field.

Common mistake. Students may think consensus means a vote that replaces evidence.

Reading move. Consensus is strongest when multiple independent evidence streams agree.

Question to ask. What reviews, professional bodies or datasets support the consensus?

Minority view

Core idea. A minority scientific view can occasionally become important, but being unpopular is not evidence of being correct.

Common mistake. Students may romanticise contrarian claims.

Reading move. Ask whether the minority claim has strong methods, predictions and independent evidence.

Question to ask. Does it explain more data with fewer unsupported assumptions?

Extraordinary claim

Core idea. An extraordinary claim is one that strongly conflicts with established evidence or requires a major revision of current models.

Common mistake. Students may think novelty itself makes a claim exciting and likely true.

Reading move. The evidential bar should rise with the size of the claimed revision.

Question to ask. Has the result been independently confirmed?

Mechanism

Core idea. A mechanism explains how a cause could produce an effect.

Common mistake. Students may think a plausible mechanism proves the effect occurs in the real world.

Reading move. Mechanistic plausibility supports a claim but still needs outcome evidence.

Question to ask. Does the mechanism operate at realistic doses and conditions?

Dose

Core idea. Dose is the amount of exposure to a substance, intervention or factor.

Common mistake. Students may treat presence of a chemical as equivalent to harmful dose.

Reading move. Toxicology depends on amount, route, duration and susceptibility.

Question to ask. What exposure level did the study actually test?

Exposure

Core idea. Exposure describes contact with an agent under defined conditions.

Common mistake. Students may infer exposure from environmental presence alone.

Reading move. Measure or estimate route, concentration, duration and frequency.

Question to ask. Does the exposure in the study resemble real-world conditions?

Hazard

Core idea. Hazard describes the potential to cause harm.

Common mistake. Students may treat hazard as identical to risk.

Reading move. A hazardous substance can present low risk under sufficiently controlled exposure.

Question to ask. What exposure is required for the hazard to matter?

Risk

Core idea. Risk combines probability and consequence under defined conditions.

Common mistake. Students may interpret a hazard label as certainty of harm.

Reading move. Risk assessment needs exposure, dose-response and uncertainty.

Question to ask. What is the actual probability under the relevant scenario?

Surrogate outcome

Core idea. A surrogate outcome is a proxy measured instead of the final outcome of interest.

Common mistake. Students may assume improving a biomarker always improves health.

Reading move. Some surrogates are validated; others may not capture all effects.

Question to ask. Does change in the surrogate reliably predict meaningful outcomes?

Self-reported outcome

Core idea. Self-reported measures depend on participants describing behaviours or symptoms.

Common mistake. Students may dismiss self-report entirely or accept it uncritically.

Reading move. Self-report can be useful but is vulnerable to recall and response biases.

Question to ask. Was the measure validated and was blinding possible?

Objective measure

Core idea. Objective measures use instruments or predefined criteria, but are not automatically unbiased.

Common mistake. Students may think instruments remove all judgment.

Reading move. Calibration, placement, thresholds and data processing still matter.

Question to ask. How was the measurement validated?

Composite outcome

Core idea. A composite combines several outcomes into one measure.

Common mistake. Students may assume all components matter equally.

Reading move. A positive composite can be driven by a less important component.

Question to ask. Which components changed, and by how much?

Relative change

Core idea. Relative change reports proportional difference from baseline.

Common mistake. Students may find large percentages persuasive without context.

Reading move. Pair relative change with absolute values.

Question to ask. What was the starting value?

Absolute change

Core idea. Absolute change reports the direct difference between values.

Common mistake. Students may miss meaningful proportional changes from small baselines.

Reading move. Both absolute and relative framing can be useful when reported transparently.

Question to ask. Which framing helps the audience understand practical magnitude?

Average

Core idea. An average summarises data but can hide variation and subgroups.

Common mistake. Students treat the mean as a description of every individual.

Reading move. Look at distribution, range and subgroup variation.

Question to ask. How spread out are the observations?

Median

Core idea. The median is the middle ordered value and can be more robust to extreme values.

Common mistake. Students may think mean is always the ‘real’ average.

Reading move. Compare mean and median when data are skewed.

Question to ask. Which summary better represents a typical case?

Outlier

Core idea. An outlier is an observation far from the rest of the data.

Common mistake. Students may assume outliers should always be removed.

Reading move. Outliers can be errors, rare true cases or signals of hidden structure.

Question to ask. Was there a pre-specified rule for exclusion?

Missing data

Core idea. Missing data occur when expected observations are unavailable.

Common mistake. Students may ignore missingness if the remaining sample is large.

Reading move. If missingness relates to outcome or exposure, estimates can be biased.

Question to ask. Who dropped out or lacked measurements?

Attrition

Core idea. Attrition is loss of participants over time.

Common mistake. Students may assume dropouts do not matter if many people remain.

Reading move. If dropouts differ systematically, group comparisons can be distorted.

Question to ask. Were attrition rates and reasons balanced between groups?

Measurement error

Core idea. Measurement error is difference between measured and true values.

Common mistake. Students may assume all error is random.

Reading move. Systematic error can bias results consistently.

Question to ask. Was the instrument calibrated and was the method validated?

Misclassification

Core idea. Misclassification assigns observations to the wrong category.

Common mistake. Students may treat categories as perfectly measured.

Reading move. Diagnostic tests and survey classifications have sensitivity and specificity.

Question to ask. Could classification error differ between groups?

Sensitivity

Core idea. Sensitivity is the ability of a test to correctly identify true positives.

Common mistake. Students may think a highly sensitive test is automatically accurate overall.

Reading move. Interpret sensitivity together with specificity and prevalence.

Question to ask. What happens when the condition is rare?

Specificity

Core idea. Specificity is the ability of a test to correctly identify true negatives.

Common mistake. Students may use specificity as a synonym for precision.

Reading move. Diagnostic performance depends on multiple measures.

Question to ask. What false-positive rate accompanies the test?

Base rate

Core idea. The base rate is how common an outcome or condition is before new evidence.

Common mistake. Students ignore prevalence when interpreting tests.

Reading move. Rare conditions can produce many false positives even with good tests.

Question to ask. What was the prior probability?

False positive

Core idea. A false positive is a positive test in someone or something without the target condition.

Common mistake. Students may think positive always means true.

Reading move. Test accuracy and base rate influence positive predictive value.

Question to ask. How often are positives confirmed by a stronger test?

False negative

Core idea. A false negative is a negative test despite the target condition being present.

Common mistake. Students may assume a negative result proves absence.

Reading move. Sensitivity, timing and detection limit matter.

Question to ask. Was the test capable of detecting the condition at that stage?

Prediction

Core idea. A scientific prediction states what should be observed if a model is accurate under defined conditions.

Common mistake. Students confuse prediction with confident guessing.

Reading move. Good predictions can be checked and potentially fail.

Question to ask. What observation would make the prediction wrong?

Forecast

Core idea. A forecast estimates a future state, often probabilistically.

Common mistake. Students may judge a forecast only as right or wrong after one event.

Reading move. Probabilistic forecasts should be evaluated across many predictions and calibration.

Question to ask. Did events assigned 70% probability occur about 70% of the time over many cases?

Model

Core idea. A model is a simplified representation used to explain or predict.

Common mistake. Students may treat model output as direct measurement.

Reading move. Models depend on assumptions, inputs and validation.

Question to ask. How well has the model performed on data not used to build it?

Model uncertainty

Core idea. Different plausible model structures can produce different results.

Common mistake. Students may assume one sophisticated model removes uncertainty.

Reading move. Compare ensembles, sensitivity analyses or alternative models where appropriate.

Question to ask. Which conclusions are robust across model choices?

Parameter uncertainty

Core idea. Model parameters are often estimated with uncertainty.

Common mistake. Students may focus on one best-fit number.

Reading move. Parameter ranges can change predictions.

Question to ask. Were uncertainty ranges propagated into final results?

Sensitivity analysis

Core idea. Sensitivity analysis tests how conclusions change when assumptions or parameters vary.

Common mistake. Students may see assumption changes as undermining the model.

Reading move. Robust conclusions should survive reasonable variation.

Question to ask. Which assumptions matter most?

External validity

Core idea. External validity concerns whether findings generalise to other populations, settings or times.

Common mistake. Students may generalise automatically from one study site.

Reading move. Compare context, demographics, environment and implementation.

Question to ask. Who is the evidence actually about?

Internal validity

Core idea. Internal validity concerns whether the study supports the claimed relationship within the studied sample.

Common mistake. Students may prioritise generalisability before checking whether the study itself is sound.

Reading move. Confounding, bias and measurement quality affect internal validity.

Question to ask. Is the observed difference plausibly attributable to the tested factor?

Ecological validity

Core idea. Ecological validity concerns how well study conditions reflect real-world contexts.

Common mistake. Students may assume laboratory control automatically makes findings directly applicable to daily life.

Reading move. Lab control can strengthen mechanism tests while reducing realism.

Question to ask. Do real-world conditions differ in ways that could change the effect?

Statistical power

Core idea. Power is the probability of detecting an effect of a given size under specified assumptions when it exists.

Common mistake. Students think a non-significant result proves no effect.

Reading move. Small studies can miss meaningful effects.

Question to ask. Was the study designed with enough observations for the expected effect?

Null result

Core idea. A null result means the analysis did not detect a specified effect under the design and assumptions.

Common mistake. Students may treat null as proof of no difference.

Reading move. Confidence intervals show what effect sizes remain compatible with data.

Question to ask. Was the study precise enough to rule out meaningful effects?

Negative result

Core idea. Negative findings can challenge hypotheses and reduce publication bias when reported.

Common mistake. Students think only positive findings are useful.

Reading move. Well-designed null results improve scientific knowledge.

Question to ask. Was the negative result published and integrated into reviews?

Replication failure

Core idea. A replication failure means a later study did not reproduce an earlier finding under its design.

Common mistake. Students may assume one failed replication proves the original fraudulent.

Reading move. Differences in methods, populations and statistical precision must be examined.

Question to ask. Do multiple replications converge?

Converging evidence

Core idea. Confidence increases when independent methods point to the same conclusion.

Common mistake. Students may focus on one definitive experiment.

Reading move. Different methods have different weaknesses, so convergence can be powerful.

Question to ask. Do observational, experimental, mechanistic and real-world data agree?

Triangulation

Core idea. Triangulation uses multiple approaches to study the same question.

Common mistake. Students think mixed methods weaken clarity.

Reading move. Independent biases are less likely to explain all converging results.

Question to ask. What does each method contribute that the others cannot?

Science consensus report

Core idea. Consensus reports synthesise large bodies of evidence through organised expert review.

Common mistake. Students may read only the executive headline.

Reading move. Read scope, confidence language, evidence basis and uncertainties.

Question to ask. Which claims are high confidence and which remain uncertain?

News embargo

Core idea. Embargoes give journalists advance access before publication under release timing rules.

Common mistake. Students may think simultaneous headlines prove independent confirmation.

Reading move. Many outlets may be reporting the same single study at once.

Question to ask. How many independent studies are actually behind the news wave?

Viral claim

Core idea. A viral claim spreads rapidly through social networks.

Common mistake. Students may use popularity as evidence.

Reading move. Trace the claim backwards to its earliest identifiable source.

Question to ask. Does the original source support the viral wording?

Screenshot evidence

Core idea. Screenshots can remove context, date, units or source links.

Common mistake. Students may trust a chart because it looks official.

Reading move. Reverse-search or locate the original chart and caption.

Question to ask. Are axes, units and data source visible?

Graph manipulation

Core idea. Graphs can mislead through truncated axes, selective ranges, dual axes or omitted uncertainty.

Common mistake. Students may assume graphs are neutral.

Reading move. Replot or mentally reconstruct the data with full scale and labels.

Question to ask. Does visual impression match numerical effect?

Cherry picking

Core idea. Cherry picking selects evidence that supports a desired conclusion while ignoring contrary evidence.

Common mistake. Students may think many supporting links prove a claim.

Reading move. Search for systematic reviews and evidence that could challenge the claim.

Question to ask. What happens when all relevant studies are considered?

Quote mining

Core idea. Quote mining extracts words from context to suggest a meaning the source did not intend.

Common mistake. Students may trust a dramatic quotation without surrounding text.

Reading move. Read paragraphs before and after and locate original source.

Question to ask. Does the full context change the apparent claim?

Appeal to authority

Core idea. Expert opinion can be informative but does not replace evidence.

Common mistake. Students may accept or reject claims solely by who says them.

Reading move. Use expertise as one signal and evidence quality as another.

Question to ask. What data or literature does the expert cite?

False balance

Core idea. False balance presents two positions as equally supported when the evidence base is very uneven.

Common mistake. Students may think fairness requires equal weight to every view.

Reading move. Compare evidential support, not speaker count.

Question to ask. Do professional reviews show genuine scientific uncertainty or fringe disagreement?

Misinformation

Core idea. Misinformation is false or misleading information shared without necessarily deceptive intent.

Common mistake. Students may assume every wrong claim is deliberate.

Reading move. Correct the claim and identify the reasoning failure without inventing motives.

Question to ask. What reliable source can verify the point?

Disinformation

Core idea. Disinformation involves intentionally deceptive information, but intent is often difficult to establish from content alone.

Common mistake. Students may label any disliked claim disinformation.

Reading move. Focus first on factual accuracy and evidence rather than guessing motive.

Question to ask. Is there credible evidence of coordinated deception?

Scientific uncertainty

Core idea. Uncertainty quantifies or acknowledges what is not precisely known.

Common mistake. Students may interpret uncertainty as ignorance or weakness.

Reading move. Every measurement and model has limits; responsible Science states them.

Question to ask. Does the source explain what is known with high confidence and what is less certain?

Provisional evidence

Core idea. Early evidence can be useful without being final.

Common mistake. Students may swing between ‘confirmed’ and ‘worthless’.

Reading move. Use calibrated language such as preliminary, suggestive, supported or well established.

Question to ask. How much independent evidence exists?

Consensus change

Core idea. Scientific consensus can shift when stronger evidence accumulates.

Common mistake. Students may think change means earlier Science was dishonest.

Reading move. Revision is expected when models fail new tests.

Question to ask. What new data drove the change?

Extraordinary certainty

Core idea. Words such as proves, miracle, breakthrough, cure, destroys, guarantees and completely safe deserve scrutiny.

Common mistake. Students may equate strong language with strong evidence.

Reading move. Compare rhetorical intensity with effect size, design and replication.

Question to ask. Would the scientific paper use the same word?

Health headline

Core idea. Health claims need attention to population, dose, comparator, outcome and absolute risk.

Common mistake. Students may generalise from mice, cells or one subgroup to everyone.

Reading move. Trace the evidence hierarchy and quantify absolute effects.

Question to ask. What is the realistic benefit or harm for the studied population?

Nutrition headline

Core idea. Nutrition studies often involve self-report, correlated behaviours and long-term outcomes.

Common mistake. Students may believe one food is proved good or bad by one study.

Reading move. Look for controlled feeding studies, cohorts and systematic reviews together.

Question to ask. How large is the effect and what dietary pattern surrounds it?

Environmental headline

Core idea. Environmental claims depend on location, exposure, timescale and ecological context.

Common mistake. Students may generalise from one site to the planet.

Reading move. Check sampling, units and spatial scale.

Question to ask. Is the result local, regional or global?

Climate headline

Core idea. Climate claims need long-term datasets, baselines, model context and distinction from individual weather events.

Common mistake. Students may use one storm or cold day to support or reject a trend.

Reading move. Look for multi-decadal records and attribution studies.

Question to ask. Is the claim about event occurrence, intensity, probability or long-term trend?

Space headline

Core idea. Astronomy claims often rely on indirect inference, imaging conventions and uncertainty.

Common mistake. Students may treat an artist’s impression as a photograph.

Reading move. Check whether the image is data, reconstruction, simulation or illustration.

Question to ask. What observation actually supports the discovery?

AI headline

Core idea. AI claims need clear benchmarks, comparison baselines, dataset definitions and real-world validation.

Common mistake. Students may accept ‘human-level’ without knowing the task.

Reading move. Check metric, test set, data leakage, deployment conditions and failure cases.

Question to ask. Does benchmark performance transfer to real use?

Technology headline

Core idea. Technology performance claims depend on test conditions and comparison products.

Common mistake. Students may assume laboratory peak performance equals everyday performance.

Reading move. Check duty cycle, environment, degradation, cost and system integration.

Question to ask. What conditions were required to produce the reported number?

Energy headline

Core idea. Energy claims need system boundaries, units, capacity factors, storage and lifecycle assumptions.

Common mistake. Students may compare unlike metrics such as peak power and annual energy.

Reading move. Put alternatives on a common metric and state assumptions.

Question to ask. What is included and excluded from the comparison?

Pollution headline

Core idea. Pollution claims require pollutant definition, concentration, exposure pathway and health or ecological endpoint.

Common mistake. Students may equate detection with dangerous exposure.

Reading move. Compare measured concentration with relevant exposure thresholds and evidence.

Question to ask. What dose or exposure level produced effects in the study?

Genetics headline

Core idea. Genetic associations rarely mean a single gene determines a complex trait.

Common mistake. Students may interpret genetic influence as destiny.

Reading move. Complex traits often involve many variants plus environment.

Question to ask. How much variance does the genetic factor explain?

Education headline

Core idea. Education studies depend on student population, teacher implementation, outcome measure and comparison condition.

Common mistake. Students may assume one intervention works everywhere.

Reading move. Check effect size, replication and context.

Question to ask. Was the intervention compared with normal teaching or a weak control?

Animal-behaviour headline

Core idea. Animal studies require careful operational definitions and avoidance of anthropomorphic overinterpretation.

Common mistake. Students may translate behaviour directly into human emotions or intentions.

Reading move. Check what was measured and whether alternative interpretations exist.

Question to ask. What behaviour supports the psychological label?

Citizen-science data

Core idea. Citizen-science projects can generate huge datasets but need quality-control procedures.

Common mistake. Students may dismiss non-professional observations or accept all equally.

Reading move. Training, validation, repeated observations and statistical methods can make citizen data useful.

Question to ask. How are observations verified?

Government data

Core idea. Government datasets often have strong collection systems but still require definitions and revisions.

Common mistake. Students may treat official numbers as timeless and context-free.

Reading move. Read methodology, date, revision notes and population definitions.

Question to ask. Did the definition change across years?

Company research

Core idea. Companies can produce valuable research but may have commercial incentives.

Common mistake. Students may reject it automatically or accept marketing claims uncritically.

Reading move. Separate the data and methods from promotional framing; look for independent replication.

Question to ask. Are raw methods and comparisons transparent?

NGO research

Core idea. Non-government organisations can produce useful evidence while also pursuing missions.

Common mistake. Students may assume mission-driven work is either automatically biased or automatically trustworthy.

Reading move. Evaluate methods, sources and transparency independently of alignment with values.

Question to ask. Can the findings be checked against independent data?

University research

Core idea. University affiliation signals research context but not infallibility.

Common mistake. Students may assume institutional prestige guarantees correctness.

Reading move. Check design, statistics, transparency and later evidence.

Question to ask. Has the finding been replicated outside the institution?

Predatory journal

Core idea. Predatory publishing can mimic scholarly appearance without rigorous editorial standards.

Common mistake. Students may trust any PDF with journal formatting.

Reading move. Check publisher reputation, indexing, editorial board and peer-review transparency.

Question to ask. Is the journal recognised by reputable databases or institutions?

Citation count

Core idea. Citation count measures how often a work is cited, not how often it was confirmed.

Common mistake. Students may interpret 1,000 citations as 1,000 replications.

Reading move. Papers can be cited to criticise, contextualise or reuse methods.

Question to ask. What do the citing papers actually say?

Impact factor

Core idea. Journal impact metrics describe citation patterns at journal level and do not directly validate individual papers.

Common mistake. Students may treat high-impact publication as proof.

Reading move. Judge the paper’s method and evidence directly.

Question to ask. Would the study still be convincing if published elsewhere?

Open access

Core idea. Open access means readers can access the paper without subscription under particular terms.

Common mistake. Students may confuse access model with research quality.

Reading move. Open and subscription journals can both publish strong or weak research.

Question to ask. What are the journal’s review and editorial standards?

Retrospective analysis

Core idea. Retrospective studies analyse past records or data.

Common mistake. Students may assume records were collected specifically for the current question.

Reading move. Missing variables and inconsistent measurement can limit inference.

Question to ask. Was the necessary information recorded reliably at the time?

Prospective study

Core idea. Prospective studies define measurements before outcomes occur.

Common mistake. Students may think prospective automatically means experimental.

Reading move. It can still be observational if exposures are not assigned.

Question to ask. Was the exposure assigned or merely observed?

Longitudinal study

Core idea. Longitudinal research follows participants or systems over time.

Common mistake. Students may confuse duration with causal strength.

Reading move. Longitudinal data help establish sequence but confounding can persist.

Question to ask. What changed over time besides the main variable?

Natural experiment

Core idea. Natural experiments use external events or policies that create comparison conditions without researcher assignment.

Common mistake. Students may think they are uncontrolled anecdotes.

Reading move. Strong designs can exploit quasi-random exposure or discontinuities.

Question to ask. Why is the comparison plausibly independent of other causes?

Instrumental variable

Core idea. Instrumental-variable methods use a variable that affects exposure but should affect outcome only through that exposure under assumptions.

Common mistake. Students may treat advanced statistics as magic causation.

Reading move. The method is only as good as its assumptions.

Question to ask. Are the instrument assumptions credible?

Difference-in-differences

Core idea. This method compares changes over time between treated and comparison groups.

Common mistake. Students may focus only on before-after changes in one group.

Reading move. The key assumption is that trends would have been parallel without treatment.

Question to ask. Were pre-intervention trends similar?

Regression discontinuity

Core idea. Regression-discontinuity designs exploit treatment cutoffs to compare cases near a threshold.

Common mistake. Students may generalise results far from the cutoff.

Reading move. Causal strength is strongest locally around the threshold.

Question to ask. Was the assignment rule manipulated?

Bayesian evidence

Core idea. Bayesian analysis updates prior beliefs with new evidence.

Common mistake. Students may think prior beliefs are arbitrary opinions.

Reading move. Priors can be informed, weakly informative or sensitivity-tested.

Question to ask. Do conclusions change strongly under reasonable prior choices?

Evidence hierarchy

Core idea. Different questions favour different study designs; no single hierarchy fits every scientific field.

Common mistake. Students may assume RCTs are always superior to every other evidence type.

Reading move. For mechanisms, harms, rare events, astronomy or Earth systems, other designs can be more appropriate.

Question to ask. What design best answers this specific question ethically and practically?

Scientific argument

Core idea. A strong scientific argument connects claim, evidence and reasoning.

Common mistake. Students may present facts without showing why they support the conclusion.

Reading move. Use claim → evidence → reasoning explicitly.

Question to ask. Can another explanation fit the same evidence?

A five-layer Science-news reading method

Layer 1 — Headline: What is being claimed?
Layer 2 — Article: What study or evidence is described?
Layer 3 — Source: Can you locate the original paper, dataset or official report?
Layer 4 — Method: Who or what was studied, how was it measured, and what comparison was made?
Layer 5 — Context: What do systematic reviews, replications or professional consensus say?

How to read a scientific abstract

An abstract usually compresses background, methods, results and conclusion. Students should look for sample, design, main outcome, effect size and limitation language. They should not read only the final conclusion sentence.

A useful exercise is to cover the conclusion, read methods and results, and write a cautious one-sentence conclusion before comparing it with the authors’ wording.

How to read graphs in Science news

Check axes, units, scale, baseline, uncertainty bars, sample size and whether the graph shows raw values, percentages, model estimates or cumulative totals. A dramatic graph can be numerically modest if the axis is truncated; a small-looking change can be important when baseline risk is high.

Use How to Read Science Diagrams, Graphs and Tables Without Guessing for the full representation protocol.

How to compare two conflicting Science stories

Check whether they are reporting the same study or different studies.
Check whether the populations differ.
Check whether the outcomes differ.
Check whether one is observational and the other experimental.
Compare effect sizes and uncertainty, not just direction.
Check whether one report is newer and incorporates more evidence.
Look for systematic reviews or consensus reports.
Accept that two findings can differ without either being fraudulent.

Science-news practice for Primary students

Use short age-appropriate stories. Ask: What was observed? How many things were tested? Was there a fair comparison? Does the headline say “all” when the study tested only a few? Keep the emphasis on evidence, not on accusing sources of dishonesty.

This develops the same habits used in fair-test and graph questions.

Science-news practice for Secondary students

Secondary learners can compare abstracts, press releases and news stories, identify study design, distinguish association from causation, interpret absolute and relative risk, and inspect limitations. They can also use public data to verify simple claims.

The best task is often to rewrite a headline so it matches the evidence more accurately.

A twelve-week Science-news literacy programme

Week 1: claim, evidence and source.
Week 2: primary versus secondary sources.
Week 3: samples, populations and bias.
Week 4: experiments versus observational studies.
Week 5: correlation, causation and confounding.
Week 6: effect size, absolute risk and relative risk.
Week 7: statistical significance and uncertainty.
Week 8: replication, reviews and publication bias.
Week 9: graphs, screenshots and data visualisation.
Week 10: health, environment, climate and technology headlines.
Week 11: misinformation, false balance and expert claims.
Week 12: full article-to-source audit and headline rewrite.

How parents can use Science news at home

Choose one short Science story a week.
Ask the child to state the claim in one sentence.
Locate the evidence source.
Identify sample, design and measured outcome.
Ask what the article does not tell us.
Rewrite the headline more cautiously if needed.
Return later if new evidence appears and update the conclusion.

How tutors can use three students

Give all three students the same headline but different supporting excerpts: one gets the press release, one gets the abstract, and one gets a news summary. Ask them to infer the claim and confidence independently, then compare what information changed their judgment.

This makes source layers visible and shows why evidence quality matters more than rhetorical confidence.

Frequently asked questions

How can I tell if a Science article is trustworthy?

Trace the claim to a primary source, inspect the study design, sample and outcome, look for effect size and uncertainty, and compare with broader evidence or replication.

Does peer review mean a study is true?

No. Peer review is a quality-control process, not a guarantee. Later replication, correction and synthesis still matter.

Does correlation mean causation?

No. Correlation is compatible with causation, reverse causation, confounding or coincidence. Stronger causal inference needs design and evidence that rule out alternatives.

Is a large sample always better?

A larger sample usually improves precision, but cannot by itself fix biased sampling, poor measurement or the wrong research question.

What is more important: p-value or effect size?

Both can matter, but effect size and uncertainty are essential for understanding practical magnitude. A small p-value does not tell you whether an effect is important.

Why do Science headlines change over time?

New studies, better measurements, replication and synthesis can refine earlier conclusions. Updating in response to evidence is expected scientific behaviour.

Should I distrust company-funded research?

Funding is relevant context, but it does not determine truth by itself. Examine methods, transparency, conflicts and independent replication.

What should a student do when two experts disagree?

Check whether they have relevant expertise, what evidence each cites, and whether the wider evidence base is genuinely divided or one position has much stronger support.

Final operating rule

Do not ask only, “Is this Science news true or false?” Ask a better sequence: What is the claim? What was measured? Who or what was studied? What design produced the evidence? How large is the effect? What uncertainty remains? Has it replicated? Does the headline stay inside the study’s scope? Scientific literacy is not permanent doubt. It is the ability to place confidence where the evidence deserves it—and to update that confidence when better evidence arrives.

Small Group Tutorials

Advanced Science Tutorials | How to Read Science News and Evaluate Scientific Claims