Does "Publication Bias" Lead to Biased Science?

• Published in: Oikos 1997-05, Vol.79 (1), p.167-168
• Main Authors: Silvertown, Jonathan, McConway, Kevin J.
• Format: Article
• Language: English
• Subjects: Animal, plant and microbial ecology; Biological and medical sciences; Conservation biology; Ecological competition; Ecology; Fundamental and applied biological sciences. Psychology; General aspects. Techniques; Lung neoplasms; Meta analysis; Methods and techniques (sampling, tagging, trapping, modelling...); Opinions
• ISSN: 0030-1299; 1600-0706
• DOI: 10.2307/3546101

In a survey of a wide range of biological journals, including Journal of Ecology of which one of us is Editor, Csada et al. (1996) show that only a small fraction of published papers report that the main hypothesis is not supported by the data. It is highly improbable that a similarly small fraction of the studies actually conducted fails to reject a null hypothesis and so Csada et al. (1996), like many before them (reviewed by Sterling et al. 1995), conclude that there is a bias in the selection of results for publication. It is argued that this bias could originate with self-censorship by authors themselves, or with Editors or some unspoken collusion between the two. While we would not quarrel with the evidence that there is a widespread bias against the publication of negative results, we wish to argue that the consequences of this for science, and for ecology in particular, are not as detrimental as is suggested. Indeed to view this phenomenon as a "bias" at all one must adopt a particular (and in our view wrong) model of the scientific process. The "bias" does not reflect some imperfection in the way scientific results are reported (in which case there would be more reason to worry) but rather the fundamental nature of scientific enquiry itself. The statistical "bias" need not lead to fundamentally biased science.