Let's Be Positive About This
Reproducibility has always been a struggle in experimental papers; we chemists moan about it, but it looks as though medicine may have exposed the underside of the ice berg.
According to Researchers at Bayer's labs, two thirds of papers are just not reproducible. And this is papers which are accepted by the scientific community as valid and quality works. In attempting to replicate 67 of them, Bayer only succeeded fully in replicating 14, and partially replicating a further 8. Naturally, this is not all due to sloppy academia: sometimes these problems are due to the complexity of biological systems, unrecognised flaws or systematic errors - very sensitive changes which may not have been observed or reported. In my experience, lack of experimental details is more often the real problem with reproducibility, but more and more often scientists are being put under pressure to publish one-time positive results because of the need to validate funding and secure promotions.
Some have mooted the idea that it should be the responsibility of peer reviewers to replicate results in journal submissions to check for reproducibility, but where they will get the funding for the equipment, machine time and chemicals for – let alone how they will create the time for this – was not fully explained in the proposal. They probably won't.
I'm a little dubious however, that reproducibility every time is a must-have (though obviously ideal). Wasn't the Rutherford gold foil experiment famously conducted by picking out only certain data – those alpha particles which came back? Or charging a droplet to measure energy quanta?
Source
Comments