We have written repeatedly (LINK, LINK) about the border between the fields of Good Research Practice and Research Integrity being more nuanced than many of us have previously assumed. The European Code of Conduct for Research Integrity (ALLEA) provides several examples of unacceptable research practices that have nothing to do with fabrication, falsification or plagiarism, forms of research misconduct that most scientists are aware of.
Different forms of research misbehavior listed in the Code are likely to be more common than outright misconduct, but there is little data on the frequency of these practices. One reason is that unacceptable research practices are listed in the Code without definitions and examples, which makes it unclear how these behaviors should be identified and handled. It may therefore be useful to identify and discuss examples that may or may not fall into the categories of unacceptable research practices.
In this commentary, we focus on a specific practice presented in the Code as “Citing selectively to either enhance own findings, or to please editors, reviewers or colleagues” (p. 8).
What is the evidence that selective citation does take place?
One source of information comes from analytic efforts that aim to characterize self-citation rates for large sets of authors assuming the majority does not engage in reference list manipulations. For example, in a recent preprint, a subset of active, highly published authors (n=20,803) was examined in PubMed. It was found that the frequency of citations to any of a single author’s entire body of published papers within one published paper is approximately power-law distributed. Based upon this distribution, it was estimated that approximately 3,284 (16%) of all authors may have engaged in reference list manipulations to some degree, possibly in an opportunistic fashion.
Another source of information is based on the number of citations or rather non-citation of work that contradicts the hypothesis or the conclusions of a previously published study.
Our readers may have followed a scientific debate on whether or not the (+) enantiomers of the well-known opioid antagonists naloxone and naltrexone are TLR4 antagonists (LINK and LINK).
In short, results reported by one laboratory could not be confirmed in follow-up studies by other scientists. This is a common situation and there may be several reasons why follow-up studies were not successful. We intentionally omit this discussion without comment on why confirmatory efforts may have failed in this particular case.
This debate is, however, very important when interpreting the results of subsequent studies where (+)naloxone and (+)naltrexone are used as research tools probing the role of TLR4 receptors:
In this context, our attention was drawn to a recent publication, where 14 out of a 58 citations refer to the work published by the senior author. In contrast, there were no citations of reports (e.g. LINK, LINK or LINK) that challenge the hypothesis that (+)naloxone and (+)naltrexone are TLR 4 antagonists with the potential to be used for treatment of substance use disorders.
Scientists do of course have the right to disagree with each other and many journals allow them the option to exclude certain colleagues from being invited as reviewers. However, the impact of such selective citation can be dramatic, especially if the discordant datasets are published in journals with different visibility or if the “publication productivity” of the parties involved in the debate is significantly different with one side publishing much less.
This recent publication is an interesting illustration. It appeared in 2019 and has already been cited (last accessed on September 28, 2020) 10 times, including 3 self-citations. Importantly, the citation bias in the 2019 publication is fully transmitted into the papers that cite this publication. This is the main negative consequence of selective citation – it distorts the picture and this distortion may be long-lasting.