December 7, 2021

I, Science

The science magazine of Imperial College

The number of retraction articles have increase 10-fold in the last decade as more ways of detecting scientific misconduct have been found ...

Retraction blog

The process of getting a science paper published isn’t as straightforward as one would imagine. Although getting science to the stage of publication seems strenuous enough for scientists, for some this is not the end. Some articles have post publication corrections published for them. Post publication corrections can be in the form of correction articles, where the corrections are published in a separate shorter article. However, the ultimate post publication punishment is in the form of retraction articles.

Retraction articles can be a nightmare for the authors, if it is made out that an article has been plagiarised or some sort of scientific misconduct has happened. In 2011, an article published in Nature highlighted that the number of retraction notices published had increased by 10-fold in the last decade, even though the number of articles published had only increased by 44%. Partly, this is due to development of technology that allows us to be able to detect plagiarism. So does a lot of scientific misconduct take place? In recent surveys, 1-2% of scientists admitted to falsifying or fabricating data at some point in their careers. Reasons may include getting funding, gaining respect in their field or simply to keep up with their competitors. Retraction notices are also a pain for journal editors, who in far too many cases try to avoid them, as they affect the prestige of their journal. Each retraction notice is unique and editors often find it difficult to declare genuine reasons for why an article was retracted. Thus, retraction articles are often published with vague notices such as “The editor noticed problems with the paper and hence, the paper was retracted”. This does not help the readers, as it does not stress the issues with the paper .

The question then arises – do retractions actually work? Their idea is to warn the readers of the content of an article, so they are not misled by it. The Nature article mentioned earlier also highlights a study of 235 retracted papers published between 1966-1999, which have been cited more than 2000 times without the retraction notices being cited. A similar study carried out in between 1997-2000 shows that only 4% of retraction articles were cited where the original articles were cited. These statistics are quite worrying make it clear that a new way of handling post publication correction needs to be developed or the current system needs to be improved. This is especially because of the obsession of scientists with PDFs; a survey carried out by NatureJobs last month revealed that 82% of scientists preferred the PDF, either on their desktop or printed to the HTML version of an article. This is somewhat surprising and in contrast to the digital movement of science. Perhaps, this trend is due to PDFs being more accessible to make notes on when they are on the desktop or printed. Programs such as readcube have also played a role in bringing PDFs to life by making them more interactive. Working for a science publisher, I have been told many times “most mathematicians don’t care about the HTML”. In the case of post publication corrections, this is a disturbing movement, as post publication corrections made to articles would not update the PDFs that readers have already saved on their desktop of printed. To address this issue, CrossRef, the organisation that provides DOIs to articles, created a service called ‘CrossMark’. Readers simply click on a ‘CrossMark’ logo present on PDF or HTML versions of an article and a pop-up notification immediately informs them whether the version of the article that they are reading is the most up-to-date version of the work or if further post publication corrections have been published.

But are all retractions scientific misconduct? No – some are honest mistakes. The rise in the number of retraction notices can be seen in a positive way, as perhaps, more scientists are admitting their mistakes to benefit and clear up scientific literature. Unfortunately, the post publication procedure followed varies between different publishers and journals. An initiative to keep track of retractions as well as to keep an eye on publishers is Retraction Watch, which has not only drawn attention to the large number of retractions happening, but has also led to publishers changing their policies. SCIGen is a computer program that was created by researchers at MIT in 2005, which randomly combines strings of words to create fake science papers. Although, Cyril Labbé, a French computer scientist, has made a program that detects articles created from SCIGen; two weeks ago, he informed two publishers of 120 fake articles that had been published in their journals. But, why would researchers want to create completely fake articles? Labbé previously showed that publishing fake articles is an easy way to boost a researcher’s h-index, a measure of the scientific productivity and impact of a researcher. Another problem is the amount of time that it takes for retractions and post publication corrections to appear correctly in PubMed Central. Harold Garner, the creator of the eTBLAST plagiarism detection software, said that a search in the Medline database for similar citations resulted in 56 retractions in the following months.

As the science gets more complex, so do the ways in which it can be manipulated. The battle against scientific misconduct will continue with new and innovative software with which it can be detected. However, retraction notices should not be discouraged. Journal editors should not hesitate to retract when necessary, as it is in interest to flush out bogus content from scientific literature. Publishers need to also collaborate more with another as well as follow a more similar post publication protocol so a clear norm is achieved in years to come.