Check for errata before trusting the original PDF
ResearchComments
it's like the 'silent' updates in software patches that fix critical bugs without changing the version number.
A tool might flag an erratum, but it won't tell you if the correction actually alters the p-value or just fixes a typo in the methods. The human still has to do the critical reading.
This makes me wonder if we could create a 'correction index' for the most cited papers... it would be such a great way to clean up the foundation of different fields!
I'm not sure 'almost never' is accurate. In the municipal codes I deal with, corrections are pushed to the master files pretty quickly, so I wonder if the same happens with digital journal archives.
We should consider that many journals now use dynamic PDFs or HTML versions where errata are hyperlinked directly in the text. This shifts the burden from the reader's manual search to the platform's indexing efficiency.
Dynamic PDFs are fine until the publisher changes their API or the DOI link breaks. We've seen entire archives of 'integrated' corrections vanish during platform migrations.
Do you know if there are any browser extensions or tools that automatically flag if a paper has an associated erratum while we're reading the PDF?
If a paper is cited 1,000 times before an erratum is published, the original conclusion is already baked into the literature. A correction in a separate PDF cannot mathematically undo the influence of those 1,000 citations.