When Scientists Press the Red Button – What Paper Retractions Reveal About Scientific Practice

In January 2020, Nobel Prize–winning chemist Frances Arnold stunned the scientific community when she announced the retraction of a paper she had published just a few months earlier in the journal Science. The reason: the study’s findings could not be reproduced in the laboratory.

Every year, dozens of scientific papers are retracted, either by their authors or by the editors of the journals that published them. Retraction is a serious step that signals major concerns about a paper and its findings, but it does not remove the paper from the scientific record. Anyone can still access and examine it, although citing a retracted paper as reliable evidence is strongly discouraged. What happens when science presses its most painful red button?

Scientific Safeguards

The ability to reproduce published findings is a cornerstone of the modern scientific method. Research on a particular subject may continue for decades or sometimes even centuries, involving many scientists, each examining a different part of the larger picture. Every researcher makes only a temporary contribution, standing on the shoulders of giants from the past—and sometimes on those of the scientist down the hall. Even a small flaw in the reliability of findings shared with the scientific community – once through libraries and today online – can therefore leave other researchers standing on unstable ground. Ignoring such flaws is rightly considered unethical and unprofessional.

When a scientific paper is submitted for publication, it undergoes peer review: researchers in the same field carefully examine the manuscript, comment on it and recommend corrections. The process is intended to identify and address possible flaws in the research or in the way it is presented. But peer review is never foolproof, especially when authors conceal or misrepresent details, whether deliberately or inadvertently. The system generally assumes that researchers submit their work in good faith. Even when they do, however, reviewers may struggle to identify spurious findings caused by an error or by a one-off combination of circumstances. Science seeks explanations that apply as broadly as possible, so findings that cannot be reproduced undermine that goal.

For researchers, the immediate questions are how irreproducible findings come to be published and what can be done to prevent this. After all, no reputable scientist wants to damage their professional reputation. Yet modern science faces several structural problems that make the issue difficult to address. Academic advancement today depends heavily on publishing a large number of high-quality studies. This places intense pressure on young researchers to publish at almost any cost in order to secure their professional future—and may tempt some to look for shortcuts.

Senior, experienced researchers are not immune to mistakes either. For scientists who lead large research groups, problems may arise from inadequate oversight. Researchers of Arnold’s stature often supervise extensive teams of lab managers, staff scientists, postdoctoral researchers, graduate and undergraduate students, technicians and programmers. In practice, they operate much like the chief executives of small or medium-sized companies.

Keeping such a complex operation running smoothly is a considerable challenge, particularly for people with no formal management training. As a result, a lead researcher may fail to notice important details. Even so, Arnold’s decision to retract the paper and publicly acknowledge the problem was welcomed and ultimately reinforced her credibility. It would, of course, have been better if the problem had never arisen. But once it did, Arnold acted in keeping with her conscience and her commitment to scientific integrity by notifying the journal.

Uncovering the Failures

Unfortunately, not all scientists share Arnold’s commitment to scientific integrity. Papers are often retracted because of serious ethical or professional failings, ranging from fabrication and negligence to honest errors and undisclosed conflicts of interest. Such conflicts may arise, for example, when researchers hold positions at several institutions at once or have ties to a company that stands to benefit from the paper’s publication, yet fail to disclose those connections when submitting the manuscript.

When researchers fail to uphold the highest standards of scientific integrity, it often falls to others to raise the alarm and ensure that the scientific record is corrected. In recent years, researcher Elisabeth Bik has uncovered numerous blatant manipulations of Western blots – a common laboratory technique used to detect proteins in a sample – in papers published in the life sciences and medicine. Through painstaking work, Bik has contributed to the retraction of more than a thousand papers. It is an extraordinary number of publications now formally discredited because of duplicated or fabricated images – practices that can undermine public trust in science at its very foundation.

Another case that received extensive media coverage illustrates how difficult it can be to expose scientific misconduct or fraud when a paper’s authors refuse to cooperate. The researchers did not retract the paper themselves. Instead, the journal eventually announced its retraction – but only 15 years after its publication.

The paper, published in Science in 2010, claimed that scientists had discovered a form of life that incorporated arsenic into its biological processes, in place of the phosphorus used by all other known organisms. The findings raised doubts from the outset, including among scientists such as biochemist Dan Tawfik of the Weizmann Institute of Science. But the lead author, NASA researcher Felisa Wolfe-Simon, firmly maintained that the findings were valid and emphasized that the paper had passed a legitimate peer-review process without its central claims being rejected. The debate continued for many years, until the journal ultimately accepted the critics’ arguments and, after further review, announced the paper’s retraction.

In a recent and unusual case, a paper was retracted following an internal review at Rutgers University in the United States. One of its authors was geneticist George Church, a highly prolific researcher whose name appears on more than 700 publications. The paper described a new treatment intended to slow the aging process. Several of its other authors were researchers at the biotechnology company BioViva, where Church serves as a scientific adviser. The journal retracted the paper at the university’s request and with Church’s agreement, citing “data discrepancies” in two of its figures.

BioViva reacted angrily to the retraction and even threatened legal action against Rutgers University. Church, by contrast, acknowledged that the paper contained errors, although he described them as minor and said they did not invalidate its scientific conclusions. Meanwhile, Elisabeth Bik, who has developed a remarkable ability to identify irregularities in scientific papers, found several additional problems beyond those cited when the paper was retracted.

It is important to remember that retracting a paper is the doomsday option of scientific publishing—a dramatic last resort step that people generally prefer to avoid. In this respect, it makes little difference whether the retraction is initiated by the researchers themselves or imposed by the journal’s editors. When isolated errors are discovered or legitimate scientific disagreements arise, journals often prefer a less drastic response. Science, like any human institution, is not free from mistakes. Because retraction is intended to correct and improve the scientific record, rather than to punish or assign blame, it should be used with care.

Less drastic remedies are available, depending on the circumstances. When authors discover a relatively minor technical error that does not affect the paper’s main conclusions or require a new round of peer review, the journal will usually publish a correction notice. If the error was introduced by the journal, the notice is traditionally called an erratum, from the Latin for “error.” If the mistake originated with the paper’s authors, it may be called a corrigendum.

Another fairly common practice is the publication of formal responses. In the natural sciences, these are often prompted by disagreements over a particular finding or research method. Such exchanges allow scientific debate to take place openly and transparently, enabling the wider research community to assess the evidence for itself.

The internet has also given rise to open platforms for these discussions—a kind of virtual town square where researchers can post comments and critiques of papers, either under their own names or anonymously. In many cases, other scientists can use these platforms to identify problems as early as the preprint stage, before a paper has undergone formal peer review at a journal. One prominent example is PubPeer.

Readers interested in the subject may also wish to follow the blog Retraction Watch.