Research retractions are meant to correct the scientific record, but multiple studies suggest they often come too late to stop questionable findings from circulating widely online.
A 2022 analysis of retracted papers across major online platforms found that most papers have already “exhausted” public attention by the time a journal retracts them, limiting how much a retraction can reduce further sharing of the original findings.
Retractions often arrive after attention fades
In the 2022 study, researchers quantified attention to 3,851 retracted papers over time across 14 online platforms, including social media, news coverage, blogs, and knowledge repositories.
The study reported that by the time retractions occur, many papers are no longer being discussed much in the period immediately beforehand. It found that 80.2% of retracted papers received no mentions in the two months leading up to retraction, and 93.6% received no mentions in the last month before retraction.
Because online attention naturally decays over time, the researchers concluded that retraction has limited impact on reducing online attention to problematic papers, in part because the “window” where a retraction might curb sharing has often already closed.
Retracted papers can still get outsized attention
The same 2022 analysis found that retracted papers tend to receive more attention after publication than comparable nonretracted papers across multiple types of platforms. It also reported that the “popularity surplus” for retracted papers can be higher on curated platforms such as news outlets and knowledge repositories, not just on social media.
The researchers also examined whether attention was critical or uncritical, focusing closely on Twitter. They found that tweets about papers that were later retracted tended to express more criticism than tweets about control papers, suggesting social media discussion can contain early signals of concern about questionable work.
A large dataset shows how attention and citations behave
A separate 2021 PLOS ONE study examined retracted literature published from Jan. 1, 2010, through Dec. 31, 2015, using records maintained by the Retraction Watch Database and collecting attention data from Altmetric and citation counts from Crossref.
In that dataset of 3,008 retracted articles, the study reported that 56.1% received some Altmetric attention and 5.5% were considered “popular” (Altmetric Attention Score greater than 20). It also found that the median time from publication to retraction was 457 days.
The study identified differences in why papers get retracted, depending on how widely they spread. It reported that unreliable results were the most frequent reason for retraction among popular articles, while fake peer review was the most common reason among other articles in the dataset.
The 2021 analysis also found that popular original articles tended to receive substantially more attention than their retraction notices. In other words, even when journals issue a retraction notice, the corrective message may not travel as far as the original claim.
Retractions can fail to correct the public record
A related report summarizing this area of research argues that both journals and journalists can help perpetuate misinformation when retractions are slow or when the public is not informed that widely reported findings were later withdrawn.
That report describes how retraction-related communication can also have unexpected effects on social media. It highlights a case from the 2021 PLOS ONE analysis involving a retracted 2012 JAMA Pediatrics paper about using an Elmo sticker on an apple, where the retraction was followed by many tweets that promoted the invalid finding.
The same report notes that retractions are rare relative to the size of the scientific literature and cites an estimate that about 4 in 10,000 papers are retracted. It also says Retraction Watch has built a database of more than 26,000 retractions going back to 1756.
