The new Retraction Watch Database is the largest and most comprehensive of its kind; the long-awaited database includes over 18,000 retractions dating back to the 1970s. What’s more, each retraction within the database has been assigned a reason for retraction based on a detailed taxonomy. Working alongside Retraction Watch, Science Magazine analysed over 10,000 retracted articles in the database, revealing some key themes and challenging a number of common perceptions:
- Despite a growing absolute number of annual retractions, retraction rates have plateaued; only about four in every 10,000 papers are now retracted.
- Although editorial practices and policing by journals have improved, more editors should step up, particularly those from journals with unusually low historical retraction rates.
- Relatively few authors (500 of >30,000 named in the database) are responsible for a disproportionate number of retractions (approximately 25% of the 10,500 retractions analysed), reported to be typically due to intentional misconduct rather than genuine error.
- Smaller scientific communities have a bigger problem, thought to be a result of under-developed policies and institutions inadequately equipped to enforce rules.
- Around half of retractions in the database involved fabrication, falsification or plagiarism, yet nearly 40% did not involve misconduct.
- Stigma associated with the term “retraction” could be exacerbating the problem; it may be more effective to reserve this “death penalty” for cases of intentional misconduct, and to distinguish those which had honest errors or problematic practices as “corrected”.
Past reports of retraction rate surges and scientific misconduct shone a spotlight on retraction, leading to increased efforts to eradicate “bad science”. This new Retraction Watch database is an important outcome of these efforts. The authors of Science Magazine’s feature — Jeffrey Brainard and Jia You — argue that perceived surges in retractions may be the consequence of “a community trying to police itself”, rather than “an epidemic of fraud”.