The first one was last week and involves the so-called arsenic life paper. Published in Science in 2010 (that was online; the print version came out in July 2011, and that’s already an extraordinary fact about the story), the paper provides evidence that a bacterial species/strain (an extremophile from a toxic lake in California) could substitute arsenic for phosphorus in its macromolecules (DNA and RNA most notably). The authors’ conclusion, that this microbe could not only tolerate arsenic (this is not in dispute) but use it in DNA and RNA, was immediately questioned and widely doubted. I think it is now effectively universally known to be wrong; a paper that basically asserted this in its title (“GFAJ-1 Is an Arsenate-Resistant, Phosphate-Dependent Organism”) was one of two refutations published in Science a year after the “arsenic life” article appeared.
Here are links to writeups of the story from the last couple of weeks:
Blog post by the editor-in-chief and executive editor at Science
Excellent article at Ars Technica
News piece at Science
Official retraction notice in Science
Much of the discussion/debate in my feed is about what seems to be a change in the criteria for retraction. The retraction notice (written by the editor-in-chief) specifically asserts this:
Science did not retract the paper in 2012 because at that time, Retractions were reserved for the Editor-in-Chief to alert readers about data manipulation or for authors to provide information about postpublication issues. Our decision then was based on the editors’ view that there was no deliberate fraud or misconduct on the part of the authors. We maintain this view, but Science’s standards for retracting papers have expanded. If the editors determine that a paper’s reported experiments do not support its key conclusions, even if no fraud or manipulation occurred, a Retraction is considered appropriate.
In my opinion, this apparent shift in retraction policy is a pretty big deal. The argument against retraction tends to center on the fact that this standard (experiments do not support key conclusions) could be used to suggest retraction of tens of thousands of articles from the entire history of science. I think the standard can be applied carefully and thoughtfully, and in fact I think Science has done that in this case. But the decision is necessarily a lot more complicated than simply asking whether experiments support key experiments. I’m very interested in what others at PS think.
The second newsworthy retraction happened about the same time and involves a preposterous paper that was cited and discussed here on PS. The news of this retraction is nicely reported by The Transmitter:
This paper was transparently bad science, and while Neurochemical Research is not a high- or even moderate-impact journal, it is also not a paper mill or a borderline-predatory money-printing gambit. In our discussions here on the forum (I haven’t looked them up, we can post links in the replies if interested), we wondered why the paper had been published (or even reviewed) by a respectable journal.
Apparently lots of other people wondered too, and some of them examined the data and found numerous irregularities. The paper was retracted (by the journal) with the standard rationale of “The Editor-in-Chief therefore no longer has confidence in the presented data.”
In both cases, the retractions were undertaken by the editor(s) and were opposed by the authors. (There are many cases when retraction is initiated by authors, or is agreed to by all or some of the authors.) In both cases, there are significant concerns about whether the experiments can support the claims the authors made. But in the second case, there are additional and serious concerns about data integrity, indeed likely misconduct, that are universally agreed NOT to apply to the arsenic life paper. So in some ways, the bogus COVID vaccine autism paper is a standard retraction. But without the data integrity failures, we would have a difficult decision like the one faced by Science with the arsenic DNA: flawed experiments that don’t support any interesting conclusion.