Unnoticed Errors

Interesting article in the latest Journal of the Medical Library Association: "The effectiveness of the practice of correction and republication in the biomedical literature" (PDF here).

As you might suspect from the title, it's about a study done to see how effectively corrected versions of papers "take over" from the flawed original version, looking at how often each tends to be cited by later papers.

Discouragingly, it appears that corrections don't have any sort of universal power--the originals (which, by definition, contained some error of fact) tend to continue to be cited after the correction is published, with original and corrected versions seeing roughly similar citation rates.

So basically, a lot of researchers aren't aware of the fact that a correction has been made. Presumably anytime a researcher does know about a correction, she or he will use the corrected version (unless making a point specifically about the error, I suppose), but it's not as if we can retrospectively tag every article that has a correction with a scarlet C to call peoples' attention to the fact (although Amazon's brush with automatically updating the books on your Kindle makes one think).

And how many people, while busily searching for useful articles, will also busily double-check to make sure all of the articles they find are the very latest version? Especially since usually there won't be any correction to find; most of the time, if you do that double-checking, you've spent precious time and found nothing.

I imagine that even if the flawed version is cited later, it's probably not always a disaster--there could be an error in the transcription of some figures, say, that would not necessarily have any bearing on the conclusions of the paper (which would have been based on the correct figures). If some later researcher is only citing the conclusions, not the actual figures, they're basically fine.

Obviously, if that later researcher (or a researcher citing that later researcher!) tried to replicate the work in the original paper, or to in any way rely on the actual figures, then the error would be a problem, so I'm all for correcting things just in case, but it may not always come up right away.

The article suggests that a database of corrected papers would be helpful--a sort of TurnItIn for citations, so you could go to one place and check all the papers you wanted to use.

I could indeed see this being handy, especially if it worked with your bibliographic citation management software. You could just send it your bibliography (or even your entire collection of citations!) and it would check them against its list of officially published corrections, and tell you if there were any matches.

It would be especially handy if it could also pinpoint for you the precise correction that was made, so you could see whether or not your own conclusions would be affected. As I said, I assume you'd always want to cite the corrected version, but this sort of program could alert you as to whether you just needed to swap out a citation in your bibliography, or would need to actually rewrite part of your paper.

So we're going to need cooperation from publishers, a powerful database program that recognizes citation management software, development geniuses, time and money for building the thing, a massive publicity push to let everyone know that this new tool exists, and we'll be good to go!

.