Citing practices in ecology: can we believe our own words?

Peer-reviewed articles are the foundation of modernecological science. An essential component of mostecology papers is a clear, well-crafted argument thatbuilds upon the existing research base within the subjectarea in question and substantiates important assump-tions, technical information and opinions by accuratelyidentifying (i.e. citing) the source material.Despite the central role of citing references in thewriting of research papers, there have been no studies todate that assess the suitability and fidelity of citations inecology journals. Anecdotal evidence certainly suggeststhat citation malpractice is widespread. Many of us haveencountered instances where the support of an assertionby the cited reference proves to be ambiguous, non-existent, or even contradictory (often we only noticethis when our own work has been mis-cited!). A relatedpractice is the citing of ‘‘empty’’ references (Harzing2002), also known as ‘‘lazy author syndrome’’ (Gavras2002), where the citation actually attributes a finding oran opinion to a secondary source such as a review paper,editorial, etc. But how pervasive is citation malpracticeand how can it be controlled?We randomly selected three papers from each of thetwo most recent issues (prior to April 2006) of 51journals listed under ‘‘Ecology’’ in the Science CitationIndex (all with an impact factor 1) to examine theappropriateness and accuracy of citations used. Fromeach of the 306 papers (the ‘‘primary articles’’) we thenselected randomly one citation from the reference list(the ‘‘cited article’’) and identified the assertion it waspurportedly supporting. To avoid ambiguity we onlyused assertions supported by a single citation and, toreduce potential bias from especially errant authors,only one citation per primary article was selected. Thecited article was then obtained and examined carefully(independently by three of the authors: PT, DY andLD) and its appropriateness was categorised into one offour groups by a majority decision (Table 1). At allstages of the study, papers obtainable only through theNational University of Singapore digital library systemwere used. As just six articles per journal were selected,among-journal trends were not examined.We found that the original assertion was ‘‘clearlysupported’’ by the citation in 76.1% of the cases; thesupport was ‘‘ambiguous’’ in 11.1% of the cases; andthe citation did ‘‘not support’’ the original statement in7.2% of the cases. The remaining 5.6% of the caseswere classified as ‘‘empty’’. How do these mis-citationrates compare with other disciplines? One well-pub-licised example from physics and engineering (Ball2002, Muir 2002, Simkin and Roychowdhury 2003)analysed how misprinted citations are propagatedthrough the scientific literature and concluded that asmany as 80% of all cited papers are not read by theauthors citing them (Simkin and Roychowdhury 2003).This result, however, is most probably an overestima-tion (copying citations from other sources is notnecessarily associated with whether those papers haveactually been read). A number of biomedical studieshave used an approach similar to our own, althoughthey applied the analogous categories ‘‘major error’’ and‘‘minor error’’ rather than ‘‘no support’’ and ‘‘ambig-uous’’. Combined error rates found by Fenton et al.2000 (17%) and Lukic´ et al. 2004 (19%) are compar-able to our result of 18.3% for ‘‘no support’’ plus‘‘ambiguous’’, though other results for medical journalsrange from 12.3% (Gosling et al. 2004) to 35.2%(Goldberg et al. 1993). To our knowledge, emptycitation data are absent for all the sciences.Various proximate factors could be responsible formisappropriation of citations. Many instances will begenuine mistakes where the article has been misread ormisunderstood or, through the long process of writing a