Study skills

How to tell whether a viral science story is actually true

A practical guide to tracing viral biology claims back to the evidence, using the microplastics-in-the-brain story to expose weak methods, missing controls and headline hype.

By Business Bliss · 1 September 2026 · 21 min read

Laboratory glassware and measuring cylinders — Image by PublicDomainPictures from Pixabay

A video appears in your feed with a claim designed to stop the scroll: scientists have found a spoonful of plastic in the human brain. The presenter may be wearing a laboratory coat, the caption may name a prestigious journal and the comments may already be full of people promising to throw away every plastic container they own. Somewhere beneath the dramatic music, there may even be a real study.

That does not mean the viral version is true. It does not mean it is false either. Most misleading science stories are not invented from nothing; they begin with genuine research and then stretch it. A finding in a small group becomes a fact about everybody. An association becomes a cause. A result from cells or mice becomes advice for humans. A cautious estimate becomes a household object that people can picture.

This guide uses the widely reported microplastics-in-the-brain story as a worked example. It is an unusually good case because the original research was important, peer reviewed and published in a respected journal, yet the headline version still ran well ahead of what the evidence could securely establish.

A viral science story is rarely one claim. It is usually a stack of claims, and each layer needs its own evidence.

Start by breaking the headline into separate claims

Before searching for the paper, rewrite the story in plain, testable language. The phrase "there is a spoonful of plastic in your brain" may contain at least six different claims:

Detection: material identified as plastic was found in human brain samples.
Measurement: researchers could estimate how much was present with reasonable accuracy.
Generalisation: the samples were representative enough to say something about a typical living person.
Extrapolation: a measurement from one part of the brain could be applied to the whole organ.
Trend: the amount in human brains has risen over time.
Health effect: the plastic causes dementia, cancer or some other disease.

Those claims do not stand or fall together. A study might provide convincing evidence that polymer material is present while leaving the exact quantity uncertain. It might find a difference between people with and without a condition without showing which came first. It might justify concern and further research without justifying a diagnosis, a detox regime or a shopping list.

This is the first habit worth learning: do not ask whether the whole headline is true until you have separated the parts that can be tested independently.

Find the original study, not somebody's description of it

A TikTok video, newspaper article, university press release or podcast can point you towards a study, but none of them is the study. Search for a distinctive phrase from the claim together with words such as study, journal, authors or DOI. When you find a paper, check the title, journal, publication date and DOI rather than assuming the first PDF is the final version.

For the brain-plastic story, the main source was a paper called Bioaccumulation of microplastics in decedent human brains, published in Nature Medicine on 3 February 2025. That immediately tells us more than most headlines did. The samples came from deceased people, not from brain scans or biopsies of a random sample of the living population, and the paper was a brief research communication rather than a clinical trial.

Also identify the publication status. A preprint can be valuable and may contain excellent work, but it has not yet completed journal peer review. The bioRxiv guidance, for example, explicitly warns readers that preprints have not been certified by peer review and may contain errors. Conversely, "peer reviewed" does not mean "finally proved", as the history of the brain-plastic paper shows.

Check what kind of paper you have found

Not every item on a journal website reports a new experiment. It may be a review, editorial, commentary, letter, case report, conference abstract or correction. Each serves a different purpose. A systematic review may summarise many studies; a narrative review may select evidence less formally; a commentary may argue about another paper without collecting new data.

For a viral claim, record four details before reading further:

Study type: cells, animals, human observational research, clinical trial, review or something else.
Population: what or who was actually studied.
Exposure or intervention: what was measured, administered or compared.
Outcome: what the researchers directly observed rather than what a headline inferred.

What the microplastics study actually did

The researchers analysed post-mortem samples of frontal cortex, liver and kidney. Their principal comparison used specimens collected in New Mexico in 2016 and 2024, with between 20 and 28 separate participants for each organ and time point. They also obtained older brain specimens from tissue banks, covering deaths between 1997 and 2013, and examined a separate set of 12 brains from people with recorded dementia diagnoses.

The tissue was chemically digested using potassium hydroxide. The remaining solid material was concentrated and analysed mainly by pyrolysis gas chromatography–mass spectrometry, usually shortened to Py-GC/MS. In simplified terms, this technique heats the prepared material and looks for chemical breakdown products associated with particular polymers. The researchers also used other methods, including infrared spectroscopy and electron microscopy, to provide supporting chemical and visual evidence.

This distinction matters because the study did not simply place an intact brain under a microscope and count recognisable pieces of packaging. It used several analytical steps to isolate material and infer polymer identity and mass. Every one of those steps creates questions about recovery, contamination, calibration and interference from the biological tissue itself.

What the researchers reported

The median estimated concentration in frontal-cortex samples was 3,345 micrograms per gram in the 2016 group and 4,917 micrograms per gram in the 2024 group. The brain measurements were substantially higher than those in the liver and kidney, and polyethylene made up most of the estimated polymer mass. The authors described an increase of about 50 per cent between the two groups.

They also found much higher estimated concentrations in the 12 dementia samples. This became an obvious route to alarming headlines, but the paper itself was careful. The authors noted that brain atrophy, damage to the blood–brain barrier and poorer clearance could allow material to accumulate in dementia. They expressly stated that no causality should be assumed.

That sentence should govern any honest report of the dementia result. The study did not show that microplastics caused dementia. It could not establish whether plastic accumulation preceded disease, followed disease-related changes or was influenced by another factor associated with both.

Where the "spoonful" came from

The spoon was not found in a brain. It was a media-friendly extrapolation. Commentators took the reported 2024 concentration, applied it across the approximate mass of an adult brain and arrived at roughly seven grams, similar to the weight of a small disposable plastic spoon.

That comparison is memorable, but it carries assumptions that the headline tends to hide:

the frontal-cortex sample must represent the concentration throughout the brain;
the analytical method must have estimated polymer mass accurately;
the sampled people must be representative of the wider population;
an average or median estimate must be treated as though it describes every individual;
the weight of a spoon must not quietly turn into a literal spoonful by volume.

A useful comparison can therefore become a false mental picture. The careful version is not "scientists removed a spoonful of visible plastic from every brain". It is closer to this: researchers estimated polymer-associated mass in small frontal-cortex samples, and extrapolating the median concentration across a whole adult brain produced a mass comparable to a small plastic spoon.

Check the sample size, but do not treat it as a magic number

"The sample was small" is one of the most common criticisms of research, and often one of the least informative. A sample of 25 can be woefully inadequate for estimating a subtle difference across the whole population, yet useful for demonstrating that a measurable phenomenon exists. Human brain tissue is difficult to obtain, so a modest study may still make an important first contribution.

The better question is: was the sample large and representative enough for the claim being made?

In the microplastics paper, the main organ-and-year groups contained 20 to 28 participants. The dementia comparison contained only 12 cases, split across Alzheimer's disease, vascular dementia and other diagnoses. Infrared confirmation was carried out on five brain samples, while the published imaging came from a subset of ten brains without dementia and three with dementia. Those numbers do not erase the findings, but they set limits on how confidently they can be generalised.

Count independent biological samples, not every reading

A paper may report hundreds of microscope images, thousands of cells or repeated machine readings from a handful of people. Those are not automatically hundreds or thousands of independent participants. Repeating a measurement can improve precision, but it does not create new biological subjects.

Ask what the true experimental unit is. In a human tissue study, it may be the donor. In an animal study, it may be the individual animal, the litter or sometimes the cage. In cell culture, wells prepared from the same culture may be technical replicates rather than independent biological replicates. Treating technical repeats as if they were independent samples can make a result look much more certain than it is.

Look at who is missing

Representativeness is not just about n. The principal samples in this study were post-mortem tissues collected through one medical-investigation system in New Mexico. Each participant contributed one small sample from each organ, and the brain material came from the frontal cortex. The researchers acknowledged that variation between grey and white matter, vascular regions and other anatomical areas remained unknown.

Think about how those facts affect the claim. The study can tell us about the analysed tissues under its collection and measurement conditions. It cannot, by itself, tell us the exact brain burden of every age group, country, occupation or living person.

Read the methodology like a detective

The methods section is often the least inviting part of a paper, but it is where a viral claim either acquires a foundation or starts to wobble. You do not need to understand every instrument setting to ask useful questions.

How were samples selected? Randomly, consecutively, retrospectively or because they were convenient and available?
How were they collected and stored? Could the container, preservative, laboratory air or handling process affect the result?
What did the method measure directly? A molecule, a chemical signature, an image, a questionnaire score or a proxy?
Has the method been validated for this material? A method that works in water may behave differently in fatty human tissue.
What are its detection limits and error rates? Can it distinguish a small signal from background noise?
Were samples processed in the same way? Differences between batches can masquerade as biological differences.
Were analysts blinded? Knowing which samples belong to which group can influence judgement where interpretation is subjective.
Were exclusions decided in advance? Removing inconvenient data after seeing the results can bias the analysis.

For the brain study, one central issue was the use of Py-GC/MS in a complex, lipid-rich tissue. The original authors recognised that residual biological material could interfere with the mass spectra and that lipids might be mistaken for signals attributed to polyethylene. They argued that their digestion and separation process reduced the problem and that complementary techniques supported the finding.

Later researchers challenged whether the contamination controls and validation were sufficient for the reported concentrations. This does not automatically mean that no plastic was present. It means the exact measurement, particularly the amount attributed to polyethylene, deserves more caution than a viral post usually allows.

For cell and animal stories, check dose and route

The same method applies to other biological headlines. If a substance "kills cancer cells", ask whether it also kills healthy cells and whether the concentration could ever be reached safely in a human body. If mice developed a problem after exposure, check whether they swallowed, inhaled or were injected with the substance, how the dose compares with ordinary human exposure and whether the animals were genetically unusual.

A result can be sound within the experiment and still have little immediate relevance to real life. The route from a Petri dish to a clinical recommendation is long, and social media often edits out almost all of it.

Look for controls that rule out the boring explanation

Controls are not ceremonial extras included to make a paper look scientific. They test whether the result could have appeared even when the proposed biological explanation was absent.

Negative controls show what happens without the exposure, treatment or target being studied.
Positive controls show that the method can detect a known effect when it should.
Procedural blanks go through the preparation and measurement process without the biological sample, exposing contamination from reagents, equipment or the laboratory.
Recovery or spiking tests add a known amount of material and check how much the method finds after processing.
Comparison groups help distinguish the feature of interest from age, disease, location or another variable.
Randomisation and blinding reduce systematic differences and human expectations where the study design permits them.

Contamination controls are especially important when the substance being measured is everywhere. In the microplastics study, the researchers analysed blank samples of potassium hydroxide and formalin, checked the polymer composition of tubes and pipette tips, and had five 2016 brain samples analysed independently at another university with consistent results. Those are genuine strengths.

There were still limitations. The specimens had not originally been collected over several decades with microplastic contamination in mind, and later critics argued that more extensive blank controls and validation were needed. A sensible assessment therefore records both points: the study did not ignore contamination, but expert disagreement remained about whether its safeguards were enough for precise quantification.

Ask what else could produce the same pattern

Suppose a study reports more of substance X in people with disease Y. There are several possible explanations:

X contributes to Y;
Y changes the body in a way that increases X;
a third factor influences both X and Y;
the groups differ in another relevant way;
the measurement behaves differently in diseased tissue;
selection, storage or laboratory processing creates the apparent difference;
chance produces a pattern in a small dataset.

The dementia finding illustrates reverse causation particularly well. A damaged blood–brain barrier or impaired clearance might permit particles to accumulate after disease develops. Without evidence about the order of events, "higher levels in dementia cases" cannot be rewritten as "plastic causes dementia".

The time comparison needs similar care. The 2016 and 2024 tissues came from different people; this was not a study that measured the same individuals twice. The older tissue-bank samples also came from different regions. A rising pattern across collection years is interesting, but it can reflect changes in exposure, population, collection, storage or measurement conditions. Statistical adjustment can address measured differences, not every unmeasured one.

Do not let a p-value do the thinking for you

Viral reports often use "statistically significant" as a synonym for "true and important". It is neither. A p-value is calculated under a statistical model and says something about how compatible the observed data are with a specified null hypothesis. It does not give the probability that the scientific claim is true, and it does not tell you whether an effect is large enough to matter biologically or clinically.

The American Statistical Association's statement on p-values is worth reading because it addresses exactly these misunderstandings. When checking a result, look beyond the threshold of p < 0.05 and ask:

How large was the difference?
How wide were the confidence intervals?
Can you see the individual data points, or only a summary bar?
Was the distribution highly skewed?
How many outcomes, subgroups and statistical tests were examined?
Were the main hypotheses specified before the data were analysed?
Would the conclusion survive a different reasonable analysis?

For health-risk stories, compare absolute as well as relative risk. A "doubling" can mean an increase from one case in 10,000 to two, which is a very different proposition from an increase from one person in five to two in five. For laboratory studies, distinguish statistical significance from biological relevance: a tiny change can be measured precisely without being meaningful in an organism.

Peer review is a filter, not a truth certificate

The original brain-plastic study was peer reviewed. It was also corrected on 31 March 2025. The author correction dealt with duplicated supplementary image panels, incorrect scale bars, clarification of a particle table and a sample-preparation step involving ethanol that had been omitted from the written method.

Those corrections did not amount to a retraction, and the journal did not say that the central conclusions had collapsed. They do, however, show why students should check the current version rather than downloading a paper once and treating it as frozen forever.

In November 2025, a separate group published a formal methodological challenge in the same journal. They questioned the contamination controls and validation and argued that biological material, including lipids, could affect the reported polymer concentrations. The original researchers then published a reply, acknowledging uncertainty in current nanoplastic methods while defending their digestion process, cross-laboratory results and use of several complementary techniques.

This is peer review working in public. Review before publication may catch errors and force improvements, but specialists often scrutinise a method much more intensely after a striking paper appears. Publication is an important checkpoint, not the end of scientific argument.

Look on the journal page for notices labelled correction, expression of concern, retraction, Matters Arising, comment or reply. The Crossmark service can also show whether participating publishers have registered corrections, retractions or other updates.

Find out what happened after publication

Search the exact paper title together with words such as replication, critique, commentary, correction and retraction. Check which later papers cite it and why. A citation may support a finding, dispute it, reuse its method or merely mention it in passing, so the number of citations alone is not a quality score.

In April 2026, a separate study of microplastics and nanoplastics in brain tumours and healthy human brain tissue reported polymer material in almost all of the samples it examined. Its dataset included 156 samples from 113 brain-tumour patients and 35 healthy brain samples, although those healthy samples came from only five post-mortem donors.

That later work strengthens the case that microplastics or nanoplastics can be detected in human brain tissue. It does not independently prove the original seven-gram estimate, establish a 50 per cent population-wide rise, or show that the particles cause dementia or tumours. Replication is always claim-specific. A new paper may reproduce the presence of a phenomenon without reproducing its quantity, trend or medical consequences.

Systematic reviews can be helpful once enough comparable studies exist, but a pooled result is not automatically reliable when laboratories use incompatible collection, digestion and detection methods. In an emerging field, disagreement over measurement may be the main scientific story.

Compare the headline with the strongest sentence the data justify

Once you understand the paper, perform a simple translation exercise. Replace the viral wording with the most specific sentence that the study design can carry.

Headline: "Everybody has a spoonful of plastic in their brain." Defensible version: A study estimated polymer-associated mass in post-mortem frontal-cortex samples, and extrapolating the 2024 median across a whole adult brain produced a weight of roughly seven grams.
Headline: "Plastic in the brain has increased by 50 per cent in eight years." Defensible version: Different groups of post-mortem samples collected in 2016 and 2024 had median concentrations that differed by about 50 per cent under the study's analytical method.
Headline: "Microplastics cause dementia." Defensible version: Twelve brains from people with recorded dementia had higher estimated concentrations, but the study could not establish causation or the direction of the relationship.
Headline: "Scientists proved that plastic is poisoning our brains." Defensible version: Researchers reported evidence of plastic-associated material in brain tissue; the health consequences and exact burden remain under investigation.

Pay particular attention to verbs. Causes, proves, cures, reverses, prevents and destroys all demand stronger evidence than was associated with, was detected in or was consistent with.

The distortion does not always begin with a journalist. A study of health-related science news and academic press releases found that exaggerated advice, causal language and human inference in news reports were strongly associated with similar exaggeration in the press releases supplied to journalists. Read the paper, the press release and the coverage as three separate documents.

A five-minute check for a TikTok science claim

Save the exact wording. Do not investigate a softened version of what was actually claimed.
Find the identifiable study. Look for the paper title, authors, journal, year and DOI. "Scientists say" is not a citation.
Identify the evidence level. Was the work carried out in cells, animals or humans? Was it observational or experimental?
Find the real sample size. Count independent people, animals or cultures, not images or repeated machine readings.
Read the abstract conclusion and the limitations. Then search within the paper for causal, limitation, confound, uncertain and further research.
Inspect at least one key figure and its caption. Check units, axes, spread, comparison groups and whether the result comes from a small subset.
Check controls and alternative explanations. Ask what else could create the same result.
Check the publication record. Look for corrections, criticism, retractions and later studies.
Write a one-sentence verdict. State what is supported, what is uncertain and what the study did not test.

This will not turn five minutes into a full critical appraisal, but it is usually enough to distinguish a paper from the story that has been built around it.

How to use the same method in a biology assignment

At university level, describing a paper is not the same as analysing it. A paragraph that repeats the abstract may be accurate but still weak. Good biological writing connects each conclusion to the study design, explains what the controls rule out and identifies the point at which interpretation becomes inference.

Students working through difficult biological research, laboratory evidence and scientific data may find specialist biology assignment help useful, particularly where a topic combines experimental methods, statistics and competing interpretations. The most useful support should still leave a visible chain from the original evidence to every conclusion in the finished work.

A practical critical-appraisal paragraph can follow this structure:

Claim: State exactly what the authors concluded.
Evidence: Identify the experiment, comparison and result that support it.
Strength: Note a feature that increases confidence, such as an appropriate control, independent validation or converging methods.
Limitation: Explain a realistic source of uncertainty and how it affects the interpretation.
Verdict: Say whether the evidence strongly supports, tentatively supports or does not establish the claim.

For example: The study supports the presence of polymer-associated material in post-mortem brain samples because the authors used chemical analysis alongside spectroscopy and microscopy, included reagent blanks and reproduced measurements from five samples in another laboratory. Confidence in the exact reported mass is lower because brain digestion left more residual biological material than liver or kidney digestion, and specialists later disputed whether lipids and contamination had been excluded sufficiently. The findings therefore justify further investigation of brain accumulation but do not establish a universal whole-brain burden or a causal role in dementia.

Use reporting guidelines as maps, not quality stamps

The EQUATOR Network collects reporting guidelines for different research designs. STROBE is relevant to observational studies, CONSORT to randomised trials and ARRIVE to animal research. These checklists help you see what information should have been reported, but compliance is not proof that the design was good. A beautifully reported weak study remains weak; a checklist helps you identify the weakness more accurately.

Give each part of the claim a confidence label

A binary verdict of "true" or "false" is often too crude. A more honest approach is to label each part of the story:

Well supported: several sound lines of evidence point in the same direction.
Supported with qualifications: the evidence is meaningful but depends on the studied population, method or assumptions.
Plausible but unresolved: there is a signal or mechanism worth investigating, but the present data cannot settle it.
Not established: the study did not use a design capable of answering the claim.
Contradicted: good evidence directly points the other way.

As of September 2026, a fair reading of the brain-plastic story would look something like this:

Well supported: plastic-associated particles or polymer signals can be detected in human brain tissue using current analytical techniques.
Supported with qualifications: the 2025 study measured higher concentrations in its brain samples than in its liver and kidney samples, and higher median concentrations in its 2024 group than its 2016 group.
Plausible but unresolved: human brain burdens may be rising as environmental exposure increases, but the size and universality of that trend remain uncertain.
Uncertain: the typical human brain contains exactly the mass represented by a plastic spoon.
Not established: microplastics cause dementia, brain tumours or other neurological disease.

Notice that not established does not mean impossible. It means the evidence being discussed cannot bear that conclusion. This wording protects you from both credulity and overconfident debunking.

Why scientific scepticism should not become cynicism

It is tempting to react to an exaggerated headline by declaring the entire study nonsense. That can be as careless as believing the headline. The original paper may have identified a genuine and important problem even if its most memorable numerical interpretation is uncertain.

The publication history of the microplastics study is a useful picture of science as a process: an initial result, a correction, specialist criticism, an author response and further research using different samples and methods. None of those stages offers permanent certainty. Together, they allow the claim to become narrower, clearer and more reliable.

The most useful question is therefore not simply, "Is this viral story true?" Ask instead:

Which exact part of it is supported?
What was directly measured?
Which assumptions connect the measurement to the headline?
How confident should we be, and why?
What evidence would change that judgement?

That is the difference between consuming science as content and reading it as evidence. The aim is not to become the person who automatically disbelieves every surprising result. It is to make every noun, number and verb in the story carry only as much weight as the research underneath it.

Sources and further reading

← Back to the blog