ExplainerMedia
Media framing: how the same fact can tell two stories
A headline, a selection of facts and a few reservations can shape how the same event is read. Experiments and press council decisions help distinguish the effects of framing, misleading presentations and fabrications.
On November 7, 2017, the Canadian Press news agency published an article titled “Debate over a Conservative bill to ease gun-carrying restrictions.”1 According to the Quebec Press Council, this headline confused carrying and possessing firearms and was therefore not faithful to reality. In decision D2017-11-127, dated October 19, 2018, the organization upheld the inaccuracy, dismissed the complaint of sensationalism and granted absolution because of a prompt correction, as the headline had been corrected on November 7, 2017.1
This case concerns an error. But a difference in perception can also arise without an identified error, through the selection of information alone. According to researchers Amir Tohidi, Samar Haider and Duncan J. Watts, in an experiment published in the peer-reviewed scientific journal Scientific Reports in 2025, different selections of information about the same events changed participants’ opinions and emotions. However, their protocol combines selection and tone, and the texts did not undergo independent fact-checking. It therefore does not demonstrate that two accurate articles necessarily produce opposing perceptions. Researchers’ study.2
This article follows our investigation into what really happened in the Hugging Face incident. That investigation distinguished the evaluation protocol and the disabled safeguards, cross-checked against several sources, from the account of the incident itself, which was available only in publications by OpenAI or its ecosystem. The question here becomes more general: what does a public presentation retain about the conditions and limitations of the fact it describes?
What the evidence allows us to say
- Established
- ReportedAccording to Robert M. Entman, framing is based on selection and salience.4
This is a theoretical definition attributed to his 1993 article, not an independent experimental finding. The four functions he attributes to framing are not quoted here in their exact wording.
- EstablishedSeveral press ethics organizations have found that headlines created a misleading or unfair impression: terminological confusion, an accusation presented as an established fact, or a personal characteristic associated with events without justification.5167
The convergence concerns the existence of such shortcomings. Each individual decision is based on a single source and remains attributed to the organization that issued it.
- Established
- DisputedExaggerating a scientific press release increases its likelihood of receiving media coverage.111213
The British studies by Sumner and colleagues do not detect this advantage, whereas the Dutch study by Schat and colleagues, as reported by Bratton and his team, observes one. This Dutch result is reported only second-hand, by Bratton and his team. The reason for this divergence remains unknown.
- UnknownThe sequential order of information within an article has an effect of its own, independent of its selection and visual prominence.5
The cited decisions and studies mainly document editorial hierarchy. None isolates or quantifies this sequential effect.
Established Shown by verifiable documents or data.
Reported Stated by an identified source, not independently confirmed.
Disputed Contradicted by other documented evidence.
Unknown No available source can establish it.
Framing, creating salience, shaping standards of judgment
According to researcher Robert M. Entman, in his theoretical article published in 1993 in the scientific journal Journal of Communication, framing consists of selecting and making certain elements salient. This definition invites us to examine both what appears in the account and what is absent from it. Entman also attributes to framing the functions of defining the problem, diagnosing causes, making moral judgments and recommending solutions; their exact wording is not quoted here.4
Priming concerns more specifically the standards used for judgment. According to Shanto Iyengar, Mark D. Peters and Donald R. Kinder, in two experiments published in 1982 in the scientific journal American Political Science Review, the issues highlighted by television news received greater weight in evaluations of presidential performance. Their detailed statistical results are not reproduced here, so the magnitude of the effect is not quantified.3
Researchers Jessica T. Feezell, Rebecca A. Glazier and Amber E. Boydstun, in a study published online in 2019 in the peer-reviewed scientific journal Politics, Groups, and Identities, revisit a distinction drawn from Shanto Iyengar’s work: episodic framing presents a problem through a case or a person, while thematic framing places it within a broader set of collective conditions. According to their experiment involving 1,655 people, framing changes the attribution of responsibility in the case of poverty, but the expected effect does not appear for religious radicalism. In this study, the effect therefore varies according to the subject. The protocols, results and effect sizes of Iyengar’s original experiments on these two types of framing are not reproduced here. Researchers’ study.14
The experiment by Tohidi, Haider and Watts, for its part, involved 2,141 participants randomly assigned to preregistered conditions. According to the researchers, selecting positively, neutrally or negatively toned elements in synthetic articles shifts certain attitudes. It allows different presentations of the same event to be studied, but it does not fully separate the effect of choosing the facts from that of their tone, nor does it establish the consequences of prolonged media exposure.2
Negativity and repetition: distinct effects
According to researchers Sarah Bachleda, Fabian G. Neuner, Stuart Soroka, Lauren Guggenheim, Patrick Fournier and Elin Naurin, in a study published in 2020 in the peer-reviewed scientific journal Personality and Individual Differences, the preference for negative news varies between people and contexts. Their work uses surveys in the United States, Canada and Sweden. It does not justify assuming that every reader always prefers an alarming headline. Study notice.15
Repetition acts on another dimension: perceived truth. According to Aumyo Hassan and Sarah J. Barber, in two experiments published in 2021 in the peer-reviewed scientific journal Cognitive Research: Principles and Implications, repeated statements are judged to be truer. The largest gain appears at the second exposure, after which additional gains decrease. These experiments measure truthfulness judgments; they do not demonstrate that every repetition of a headline leads to lasting acceptance of its interpretation. Study notice.9
From scientific finding to press release, then to the news story
The transformation can begin before a newsroom becomes involved. According to researchers Petroc Sumner, Solveiga Vivian-Griffiths, Jacky Boivin and colleagues, in an observational study published in 2014 in the peer-reviewed scientific journal BMJ, exaggerations in news stories were strongly associated with those in university press releases. Their corpus included 462 press releases from 20 British universities and 668 news stories. They compared, among other things, the strength of advice, causal claims and extrapolations from animals to humans with the original scientific articles. 2014 study.11
According to Sumner, Vivian-Griffiths, Boivin and colleagues, in a second analysis published in 2016 in the peer-reviewed scientific journal PLOS ONE, the association appears in a corpus of 534 press releases from scientific journals and 582 news stories. Reservations present in the press releases also appear more often in the articles. These two publications concern distinct corpora but share several authors: they do not constitute two fully independent confirmations. Their observational method is also insufficient to establish that the press release, by itself, causes each exaggeration. 2016 study.1211
The results do not agree on every point. According to Luke Bratton, Rachel C. Adams, Aimée Challenger and colleagues, in a peer-reviewed replication published in 2019 in the scientific journal Wellcome Open Research, the association appears for causal claims and extrapolations to humans, whereas it does not appear for advice, contrary to the earlier results of Sumner’s team. Replication study.131112
Another divergence concerns the amount of coverage. According to the British studies by Sumner and colleagues, exaggeration is not associated with more coverage, whereas the Dutch study by Schat and colleagues, as reported by Bratton and his team, observes more frequent coverage of exaggerated press releases. This Dutch result is reported only second-hand, by Bratton and his team. The reason for this difference, whether it relates to countries, periods or other factors, has not been established.111213
The difference between a scientific finding and a media account therefore also lies in the scope of the statement. Presenting an association as a cause, turning an observation into advice or extending a result obtained in animals to humans strengthens the conclusion beyond the work being compared. These are precisely the transformations examined by Sumner and colleagues, and later by Bratton and his team.1113
Headlines, wording and omissions: ethics decisions
In a case involving an acquitted man, the Quebec Press Council reports an asymmetry in visibility: according to the organization, the alleged victim’s testimony was covered in an article on page 7 of La Voix de l’Est, while the acquittal was covered in a brief on page 18 and was not reported on lapresse.ca. According to the Council, one headline presented the man as guilty of acts of which he had been acquitted. The organization reprimanded La Voix de l’Est and lapresse.ca. These findings, reported solely by the Council, describe an editorial hierarchy and a headline contrary to the outcome of the proceedings; they do not measure their effect on readers. The summary published by the Council reproduces neither the full headline, nor the date, nor the exact text of the articles. Council summary.5
According to the Australian Press Council, an Australian press self-regulatory body, the headline and first paragraph of an article published by news.com.au about Lynette Daley on 10 May 2016 misleadingly and unfairly suggested that she had consented. The Council finds breaches of General Principles 1 and 3, while considering the editor’s note and the link added promptly as a corrective measure sufficient. The published decision indicates neither its number nor its date.6
According to the same organization, a Daily Mail Australia headline published on 31 October 2025 highlighted a person’s transgender status without any reported connection to the events. The Council finds breaches of General Principles 3 and 6. As described by the Council, this case concerns the relevance of a personal characteristic in the account, not a demonstration that the characteristic was false. The published adjudication indicates neither its date nor its number.7
The omission may also concern the context of a judicial conclusion. According to the Australian Press Council, three articles published by The Australian in March and May 2024 about Shane Drumgold presented Justice Kaye’s findings in a misleading, inaccurate or unfair manner. The Council notes, in particular, that communications underlying a conclusion of an apparent reasonable apprehension of bias were not given sufficient prominence. It finds breaches of Principles 1, 3 and 4, but not Principle 2 concerning correction.16
Regarding the funding of the Australian Press Council, The Guardian, in its Weekly Beast column, presents News Corp as the organization’s majority member and describes its contribution as « up to 70% of its annual $1.7m budget ».17 The media outlet also writes that News Corp « effectively controls the APC as a majority member ».17 This assessment and amount are those of The Guardian, the only publication to advance them; no independent detailed accounts are provided here to confirm them.
These decisions do not demonstrate that all framing is faulty. They identify specific problems: terminological confusion, suggested guilt, implicitly attributed consent, a personal characteristic whose relevance has not been established, or essential context presented insufficiently. They illustrate situations in which checking sentences separately is not enough to assess the overall presentation.156716
Media outlets found at fault for different acts
The Newbury Weekly News published, on 15 May 2025, an article titled « Thatcham man [name removed] denies sending obscene or menacing message ».18 According to the Independent Press Standards Organisation, a British press self-regulatory body, the article incorrectly described a release on bail as being subject to conditions. In decision 05702-25 of 16 April 2026, the organization finds a breach of the accuracy clause and requires a correction. This decision establishes an error in reporting judicial proceedings; it establishes no intention to mislead.18
The Jayson Blair case belongs to a different category. According to the internal investigation published in 2003 by The New York Times, the journalist fabricated comments and scenes and reused material from other media outlets. The documentary company EBSCO also reports these fabrications and this plagiarism in a summary based, in particular, on the newspaper’s investigation. This second publication therefore draws on the same primary evidence: it does not constitute an additional independent investigation. The case concerns the invention of content, not merely the choice of an angle.1920
How to read an article
The cases and studies examined suggest several concrete checks:
- Compare the headline with the body of the text. Does the headline retain the necessary reservations, attributions and distinctions? The case of carrying and possessing weapons, as judged by the Quebec Press Council, shows the importance of a single term.1
- Identify who is asserting what. In a report on judicial proceedings, an accusation, an established fact and the outcome of the proceedings must remain distinct.2118
- Look for the conditions underlying the scientific result. Does the article describe an association or a cause? An animal or human study? A result or advice added when it was communicated?11
- Examine what is given prominence. Do an accusation and the outcome of the proceedings receive comparable visibility? According to the Quebec Press Council, the coverage of a man’s acquittal by La Voix de l’Est and lapresse.ca provides an example of asymmetrical treatment.5
- Check the relevance of personal details and the presence of decisive context. According to the Australian Press Council’s decisions, these choices can make a presentation unfair or misleading.716
- Trace the information back to the initial source. A reservation may disappear between the study, the press release and the news report; its presence in the press release is associated with its inclusion in the article, according to Sumner and colleagues.12
Finally, to compare two accounts of the same event, one can identify the facts they share, the facts present in only one text, the words in the headline and the prominence given to reservations. This reading draws on the dimensions studied by Entman and by Tohidi, Haider and Watts. It helps examine the construction of the account without inferring intent from its framing alone.42