Killing the 'zombie papers': Can we rid science of retracted articles?

Why does research whose results are generally contested continue to be disseminated? How could we reduce the time between the identification of a "fake" and its removal from documentary databases?

In the scientific field, the retraction of a publication is supposed to mark the end of its influence. Yet, it is common to find that these articles, across all disciplines, continue to be cited long after their official retraction, keeping now-discredited ideas alive. We call these articles "zombie papers."

Their persistence raises major challenges: not only do they contribute to the spread of misinformation, but they also risk diverting future research, skewing the results of meta-analyses, or misleading public policies or clinical practices. Why do these zombies persist for so long, and how can we accelerate their demise?

To understand the extent of this phenomenon, we recently conducted a study analyzing 25,480 retracted articles published between 1923 and 2023, covering a very large part of the disciplinary fields, using the Retraction Watch database . Note that the retraction can come from the publishers or editors-in-chief of a journal, the authors of the article or even their institution of affiliation and occurs when a significant problem (errors in the data, fraud, plagiarism, etc.) is detected after the publication of the article.

We highlight that the average time between publication and retraction is approximately 1,045 days, or nearly three years. This time can be much longer in some extreme cases , sometimes exceeding several decades.

Serious misconduct such as falsification or fabrication of data, which are often difficult to verify, also significantly lengthens retraction periods. Geographical differences also appear: Western European and North American countries tend to withdraw articles more quickly, while other regions such as Asia, South America, and Eastern Europe have longer retraction periods, reflecting diverse editorial and institutional practices.

Our results also indicate that paid (subscription) academic journals generally retract problematic articles more quickly than open access journals, probably thanks to more rigorous and better structured editorial processes and "quality control" mechanisms. As for so-called "predatory" journals (illegitimate or fraudulent journals publishing articles without regard for their quality), they sometimes carry out rapid retractions, not out of genuine scientific concern, but under pressure from oversight bodies or the academic community that denounce their practices, or in an attempt to preserve their already fragile credibility.

To better understand the dynamics of the disappearance of "zombie papers," we propose a theoretical model called Zombie Population Decay Dynamics (ZPDD). Inspired by ecological models, this framework simulates the temporal evolution of the population of retracted articles and assesses which editorial strategies could make them disappear more quickly. The model includes a crucial parameter: "residual capacity," which corresponds to a fraction (estimated at around 5%) of zombie articles that persist despite attempts at elimination. These publications remain cited, whether they have been clearly identified as retracted or whether they slip through the cracks of editorial mechanisms.

Three editorial strategies are studied: data transparency (mandatory submission of raw datasets), reproducibility (pre-registration of protocols, highlighting replication studies), and the improvement of plagiarism detection tools. The results show that data transparency—through researchers submitting their databases and codes upon submission of their articles—is the most effective measure. It allows for faster detection of errors or fraud, thus accelerating retractions. The other interventions also contribute to improving the overall rigor of the editorial process, even if their direct impact is slightly less when these practices are isolated.

The effects of editorial policies appear to vary greatly across disciplines. In applied and quantitative sciences, where research relies on accessible empirical data, implementing transparency policies produces rapid results: we observe that it is possible to eliminate half of "zombie papers" in approximately 1,200 days. In the humanities and social sciences, which generally employ more qualitative approaches, the timeframes are longer, since the same reduction often requires more than 2,000 days. This difference is due to the nature of scientific methods: empirical results are more easily verifiable than theoretical reasoning or literary analyses, which makes error detection more complex in certain fields.

The significant regional variations we observe – with faster retractions in Western Europe and North America than in South America or Asia due to higher standards of good practice – illustrate the need for editorial policies adapted to local contexts, taking into account available resources and infrastructure.

Our work shows that no single intervention is sufficient to solve the problem of "zombie papers." Combating this phenomenon requires an integrated approach combining several levers: data transparency, reproducibility, and the continuous improvement of fraud and plagiarism detection tools. In this regard, the establishment of open-source platforms dedicated to sharing scientific data should be encouraged to ensure free and sustainable access to the information needed to validate research.

Similarly, in terms of reproducibility, it would be desirable (i) to require authors to pre-register their experimental protocols before starting experiments in order to prevent a posteriori manipulation of results and (ii) to encourage and promote the publication of replication studies to identify possible errors or methodological biases and thus avoid the spread of erroneous results. Finally, in addition to plagiarism detection tools, the creation of specialized committees within scientific journals would facilitate the investigation of cases of fraud, by guaranteeing rigorous and impartial monitoring of dubious academic practices.

Furthermore, the visibility of retractions must be optimized. It is therefore essential that each retracted article be clearly indicated on all scientific dissemination platforms in order to limit its influence, particularly through citations.

Although killing "zombie papers" is a major challenge, the coordinated mobilization of publishers, researchers, and institutions is an essential lever for action to guarantee the integrity of the scientific corpus and ensure the robustness of the knowledge produced.

This article is republished from The Conversation under a Creative Commons license. Read the original article.