PS This is all wrong

So there you are: your paper is written, and you’ve got it accepted in the world’s leading physics journal, and it has something really interesting to say. You’ve done the calculations and they just don’t match the observations. What this implies is dramatic: we’re missing a crucial part of the puzzle, some new physics, namely a fifth fundamental force of nature. Wow. OK, so that’s a tentative conclusion, but it’s what the numbers suggest, and you’ve been suitably circumspect in reporting it, and the referees have given the go-ahead.

Then, with the page proofs in hand, you decide to just go back and check the observations, which need a bit of number-crunching before the quantitative result drops out. And you find that the people who reported this originally haven’t been careful enough, and their number was wrong. When you recalculate, the match with conventional theory is pretty good: there’s no need to invoke any new physics after all.

So what do you do?

I’d suggest that what you don’t do is what an author has just done: add a cryptic ‘note in proof’ and publish anyway. Cryptic in that what it doesn’t say is ‘ignore all that had gone before: my main result, as described in the abstract, is simply invalid’. Cryptic in that it refers to the revision of the observed value, but says this is in good agreement ‘with the predictions above’ – by which you mean, not the paper’s main conclusions, but the ‘predictions’ using standard theory that the paper claims are way off beam. Cryptic in that this (possibly dense) science writer had to read it several times before sensing something was badly wrong.

In fact, I’d contend that you should ideally withdraw the paper. Who gains from publishing a paper that, if reported accurately, ends with a PS admitting it is wrong?

True, this is all a little complex. For one thing, it could be a postgrad’s thesis work at stake. But no one gets denied a PhD because perfectly good theoretical work turns out to be invalidated by someone else’s previous mistake. And what does a postgrad really gain by publishing a paper making bold claims in a prominent journal that ends by admitting it is wrong?

True, the work isn’t useless – as the researcher concerned argued when (having written the story and just needed to add some quotes) I contacted him, the discrepancy identified in the study is what prompted a re-analysis of the data that brought the previous error to light. But you have a preprint written that reports the new analysis; surely you can just add to that a comment alluding to this false trail and the impetus it provided. In fact, your current paper is itself already on the preprint server – you just need to cite that. The whole world no longer needs to know.

No, this is a rum affair. I’m not sure that the journal in question really knew what it was publishing – that the ‘note added in proof’ invalidated the key finding. If it did, I’m baffled by the decision. And while I’m miffed at my wasted time, the issue has much more to do with propriety. Null results are one thing, but this is just clutter. I realize it must be terribly galling to find that your prized paper has been rendered redundant on the eve of publication. But that’s science for you.