On 26 June 2015, Science magazine published an article in its section “Policy Forum” entitled “Promoting an Open Research Culture” (B. A. Nosek et al. Science, Vol. 348, pp. 1422-1425, DOI: 10.1126/science.aab2374). The article and two related pieces (“Self-correction in science at work”, and “Solving reproducibility“) published in the same issue seem to have been inspired by the perception that there is an irreproducibility crisis affecting science. This delution has been installed in the media by some journalists and other outsiders to science.
Prof. Ariel Fernandez opined on the subject. His pronouncement is featured in Science. The contribution is reproduced below in accord with Terms and Conditions on User Submissions to Science:
Some journalists and some science outsiders have installed the belief that science is in the midst of a reproducibility crisis. These people are being listened to, at least by some editors, while they feverishly advocate for higher standards of transparency in regards to the way in which scientists conduct and report their findings. The underlying misconception that led to this delusional thinking may well end up sliding into hysteria if scientists keep taking advice from outsiders on how to conduct their business. The misconception sprouts from the odd notion that scientific publications are meant to report monolithic truths that must withstand the acid test of time. Nothing further from the truth, and while scientists comply and try to raise the bar on transparency and accountability, they better take steps to debunk the myth that research papers distill anything other than provisional assertions subject to endless revision.
Much of the science reported is a-priori likely and expected to be faulty merely on statistical grounds. John Ioannidis, a professor of medicine at Stanford University, wrote in 2005 a paper in the journal PLoS Medicine entitled “Why most published research findings are false” (http://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.00…) where, using statistical arguments, he estimated that the likelihood that a scientific paper contains false results is nearly 50%. His analysis reveals that under a great diversity of conditions, most scientific findings are likely to simply represent “measures of prevailing biases”. This statistical study was conducted with the utmost rigor and prompts us scientists to regard research reports with lower expectations, more in the context of a progression of provisional attempts at attaining an independent pristine truth. And please, please, let us focus on running our business ourselves, or we will have no one else to blame for the current crisis, be it real or delusional.
Dr. Ariel Fernandez Stigliano is the former Karl F. Hasselmann Professor of Bioengineering at Rice University.
阿列尔·费尔南德斯 (Ariel Fernandez Baidu biosketch)