In a recent article published in the Royal Society Open Science journal, Paul E. Smaldino and Richard McElreath describe the destructive and self-perpetuating cycle of “bad science”. The authors compare the subject of poor research to Darwin’s Theory of Evolution by Natural Selection and claim that current incentives encourage researchers to “propagate poor research methods and abuse of statistical procedures”. They argue that it “arises from the positive selection of methods and habits that lead to publication” resulting in the “natural selection of bad science”. The authors say that the primary reason for this continued behaviour is the fact that career progression is still intrinsically linked to publication record. Therefore, methods and analyses associated with perceived publication success will be copied and spread or “naturally selected”. The authors go on to present evidence and modelling to demonstrate that:
- Adequate statistical power, that has repeatedly been singled out as a problem, has not improved with time
- Methods and research originating from laboratories with higher numbers of publications are more often repeated
- Result reliability is only weakly selected for
The apparent success that accompanies publication quantity will select for processes that generate the most publishable data rather than the most scientifically sound methods. The authors conclude that fundamental changes at the institutional level are needed to break the cycle of “bad science”.