What is the bias that makes researchers prefer values already published?

Question

Some time ago I read that values (for a given process or associated to a certain physical phenomena) that have already been published, are more likely to be published in the future; compared to new values that differ substantially from them.

Is this true or is my memory failing me and I read something completely different? Does this "publication bias" have a formal name? Is there some meta-study that could be referenced?

Answer 1

I can't find a citation for it, but Charlie Poole has called this phenomena "Directional publication bias" - there's a publication bias that's toward a particular finding (in this case, what's already published) that's beyond just what you'd expect to see with the usual significant/non-significant result publication bias.

Answer 2

You may also be interested in some of the following which discuss various features of selective reporting and citation distortions.

@article{greenberg09,
   author = {Greenberg, S A},
   title = {How citation distortions create unfounded authority:
      analysis of a citation network},
   journal = {British Medical Journal},
   year = {2009},
   volume = {339},
   number = {b2680},
   keywords = {reporting of research, bibliometrics}
}
@ARTICLE{melander03,
  author = {Melander, H and Ahlqvist-Rastad, J and Meijer, G and Beermann, B},
  year = 2003,
  title = {Evidence b(i)ased medicine - selective reporting from studies
          sponsored by pharmaceutical industry: review of studies in new drug
          applications},
  journal = {British Medical Journal},
  volume = 326,
  pages = {1171--1173},
  keywords = {systematic overviews, integrity}
}
@article{cope10,
   author = {Cope, M B and Allison, D B},
   title = {White hat bias: examples of its presence in obesity research
      and a call for renewed commitment to faithfulness in research
      reporting},
   journal = {International Journal of Obesity},
   year = {2010},
   volume = {34},
   pages = {84--88},
   keywords = {reporting of research, integrity}
}

Hope those are interesting.