So the science Twitterz and bloggysphere is being roiled by the LaCour scandal, which involves a whole lot of scientific fraud along with ridiculously shoddy oversight. But what I keep reading in every story–and then which is dropped like a hot potato–is something along the lines of ‘scientists are under tremendous pressure in an environment where grant funding is getting harder to attain’ (though it’s probably written better than that…).
If we don’t address the funding crisis, this certainly won’t be the last time something like this happens. Granted (SEE WHAT I DID THERE?), LaCour’s dishonesty is breathtaking in its scope, its poor tracks covering, and its compulsiveness. But there are probably more cases which aren’t so obvious, and were committed by someone who isn’t as flashy and extroverted as LaCour seems to be. Yet those studies are ‘out there’, and they probably will never be exposed, never mind withdrawn. While there will always be the ‘extravagant’ cheaters, we can reduce the incentives for smaller-scale, harder to catch cheaters–which can damage the literature just as much as the high profile stuff.
Unrelated to the elephant in the room of funding, there have been suggestions that this wouldn’t have happened if there were post-publication peer review. Leaving aside the problem that, as this story notes, there are huge incentives not to accuse someone of fraud (or even incompetence), what we really need are two basic ‘reforms.’
First, we need clear and transparent methods sections–methods need to be placed in the actual paper, not a supplemental section that few people read and reviewers are less likely to thoroughly interrogate. Fuck the financial prerogatives of the Glamour Magz. Related to this, reviewers need to have much higher standards for methodology–did the authors describe the methods such that one could replicate the experiment (i.e., you shouldn’t have to call or email the author to find out what he actually did)?
Second, data need to be publicly released after publication. There are many different publicly-accessible repositories; use them.
These will not be a panacea–nothing is–but other people need to be able to look at your data and replicate your work.