| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
,1
* Alfa Institute of Biomedical Sciences (AIBS), Athens, Greece; and
Department of Medicine, Tufts University School of Medicine, Boston, Massachusetts, USA
1Correspondence: Alfa Institute of Biomedical Sciences (AIBS), 9 Neapoleos St., Marousi 151 23, Greece. E-mail: m.falagas{at}aibs.gr
The number of citations to previous work of an individual scientist has been considered, in part, to be an important indicator of academic performance in research. Some of these citations, of course, are to the author’s own work; the Germans call this "Eigenlob," or self-praise. We analyzed data obtained from 340 papers (294 original articles and 46 review papers) in a sample of leading biomedical science journals published during the first two months of 2005. Seventeen and 20% of references of papers published in clinical and basic science journals, respectively, referred to previous work of the authors (self-citations). Nineteen and 11% of references of original articles and review papers, respectively, were self-citations. The proportion of references of the examined papers that were self-citations was higher for original articles than review papers (Odds Ratio (OR)=1.39, 95% Confidence Interval (CI) 1.21–1.61, P<0.001) and for papers published in nonclinical journals than in clinical journals (OR=1.36, 95% CI 1.23–1.51, P<0.001). In addition, the number of authors was also associated with the proportion of references that were self-citations (OR=1.03 per author, 95% CI 1.02–1.04, P<0.001). Self-citation was approximately equally divided between the first and the last author in papers of clinical journals while it mainly referred to the last author in papers of nonclinical journals.
INTRODUCTION
The number of citations to publications of an individual author has been considered by the academic community as one of the most important indices of performance in research. This is because the number of citations to previous publications of a specific author is thought to represent recognition of his or her work and to be an estimate of the total impact of scientific work that in some respects incorporates both the quality and the quantity of the produced research. This index together with the total number and average impact factor of publications of a scientist are frequently taken under serious consideration (among other factors) in the selection process of faculty recruitment for various academic positions (1
, 2)
.
Some members of the academic community have questioned the value of the total number of citations in estimating the performance of an individual scientist in research (3)
. A practice related to the total number of citations that is sometimes criticized by members of the scientific community is that of self-citation of previous work of authors of papers published in scientific journals (1
, 3)
. It may be considered reasonable for an author to refer to his/her previous work, especially if the publications are the result of a series of consecutive efforts in a specific research field. This is particularly relevant in the case of referring to previous publications that include a reported methodology also used in a new study. However, there are concerns among members of the scientific community that the practice of self-citation is sometimes overused, some say abused, by a proportion of authors. Although there has been a continuing controversy among scientists regarding author self-citation, there have been few analyses examining this issue (4)
.
REVIEWED JOURNALS
Based on impact factor calculations, we reviewed the papers published in the first two months of 2005 in six leading journals: JAMA, the New England Journal of Medicine, the Lancet, Science, Nature, and Cell. Only original articles and review papers were included in the analysis. Publications such as editorials, commentaries, letters to the editor, and news were excluded. We also excluded articles published in the reviewed multidisciplinary journals (Science and Nature) that were not related to biomedical sciences. We included those articles published in Science that were classified by the editors of the journal in the following categories: medicine, microbiology, virology, parasitology, biochemistry, genetics, neuroscience, immunology, physiology, developmental biology, structural biology, cell biology, cell signaling, molecular biology, cell cycle, evolution, and anthropology. Articles published in Nature that were determined by us to be related to the above categories were also included in our analysis.
Papers that had more than 30 authors as well as papers that had the name of a group of scientists without indicating names of individual authors on the title page were excluded from the analysis. Although we planned to analyze data regarding the original articles and review papers of the specified journals published in the first two months of 2005, we included in our analysis the first March 2005 issue of Cell, instead of the second February 2005 issue of the journal because this was a special issue that included reviews only.
Data extracted from each reviewed paper included: journal name, title of paper, names and number of authors, number of references, number of self-citations, and number of self-citations of the first and the last author. We analyzed the distribution of the various variables to obtain estimates of self-citation. In addition, we sought determinants of self-citation among the available variables by performing a backward, stepwise, multivariable logistic regression model.
FREQUENCY OF SELF-CITATION IN THE REVIEWED JOURNALS
We reviewed 340 papers published in the first two months of 2005 in the selected journals. From these reviewed papers, 158 and 182 were published in the clinical (JAMA, New England Journal of Medicine, and Lancet) and nonclinical journals (Science, Nature, and Cell), respectively. The numbers of authors and references were higher in papers published in clinical than in nonclinical journals (mean numbers of authors and references 7.9 and 45.9, respectively, compared to 7.1 and 32.6 for the clinical and nonclinical journals). From the analyzed papers, 294 were original articles and 46 were review papers. The number of authors was higher and the number of references lower for original articles compared to review papers (mean numbers of authors 8.2 vs. 2.7; mean number of references 32.6 vs. 78.3). In Table 1
, we present descriptive data of the distribution of the number of authors, number of references, number of self-citations of any author, and number of self-citation of the first and the last author for analyzed papers.
|
Three papers that had more than 30 authors were excluded from the analysis of variables possibly associated with self-citation. These were a paper with 54 authors published in Nature (02/24/2005), one with 54 authors published in Science (02/25/2005), and one with 42 authors published in JAMA (01/12/2005). Two papers published in the Lancet (02/26/2005) were also excluded from this analysis because they had the name of a group of authors but no individual authors on the title page.
SELF-CITATION OF THE FIRST AND THE LAST AUTHOR
Self-citation in papers of nonclinical journals mainly referred to the last author while self-citation in papers of clinical journals was about equally divided between the first and the last author of the study (Table 2
). The number of authors of a paper was inversely associated with the proportion of self-citations that specifically cited previous publications of the first author (OR=0.96, 95% CI 0.94–0.98 per author, P<0.001) or the last author (OR=0.94, 95% CI 0.92–0.95 per author, P<0.001).
|
SELF-CITATION IN CLINICAL AND NONCLINICAL JOURNALS
The type of paper (original article vs. review), the type of journal (nonclinical vs. clinical), and the number of authors were found to be associated with self-citation in both the univariable and the multivariable analyses using a backward, stepwise, multivariable logistic regression model. Specifically, the proportion of references of the studied papers that were self-citations of one or more authors was higher for original articles than review papers (OR=1.39, 95% CI 1.21–1.61, P<0.001) and for papers published in nonclinical journals than in clinical journals (OR=1.36, 95% CI 1.23–1.51, P<0.001). In addition, the number of authors of a publication was associated with the proportion of references that were self-citations (OR=1.03 per author, 95% CI 1.02–1.04, P<0.001).
COMMENTARY
There have been several previous publications on self-citation of specific journals (i.e. citation to articles published in the same journal) (5
6
7
8)
. Several citation analyses have examined various aspects of scientific publications including the impact of blind peer review, the hierarchy of scientific evidence, the relative language of self-citation, and the temporal differences in self-citation of scientific journals (2
, 9
10
11
12
13)
. Only a few investigators have systematically studied the issue of author self-citation. Previous analyses of author self-citation have focused on specific research fields (4)
. To our knowledge, this is the first analysis that examined author self-citation in leading journals. We found that about 17% of references in papers published in leading clinical journals as well as 20% of references in papers related to biomedical sciences published in leading basic science journals were self-citations. We also found that approximately 19 and 11% of references of the examined original articles and review papers, respectively, were self-citations.
It is interesting to note that self-citation is more common in original articles than in review papers. Although the latter are usually written by experts with considerable previous research work in the reviewed field, the authors apparently avoid extensive self-citation. It should be emphasized that the association between the type of paper and self-citation cannot be attributed only to the fact that the number of authors is higher in original articles than in review papers because the association is persistent after adjustment for this confounding factor in the multivariable analysis.
As expected, our analysis found that the number of authors of a paper is associated with a higher proportion of self-citations. However, the analysis of the proportion of self-citations to previous work of the first and the last author of the reviewed papers provided us with interesting findings. Self-citation in papers of nonclinical journals mainly referred to the last author while self-citation in papers of clinical journals was approximately equally divided between the first and the last author of the study. This finding suggests some noteworthy trends in decisions regarding the order of authorship in papers published in clinical and basic science journals.
Our analysis had several limitations. First, we did not try to determine whether the use of self-citations in the reviewed papers was justified (i.e., whether there were alternative references that would better support or refute the statements made) since this was a task with apparent major methodological difficulties. Second, we analyzed data from only six journals (based on impact factor calculations) of about 6,000 journals that are included in the PubMed database. Third, we focused only on the major types of papers of the reviewed journals, namely original papers and review papers, without examining the issue of self-citation in publications of other types that would also be of interest. Fourth, we did not include in our multivariable analysis other important cofactors, such as the country of origin of the reviewed papers and number of previous publications of authors, which could also be among the factors determining the frequency of self-citation.
In conclusion, we found that self-citation was not rare in the reviewed journals, an observation that is likely true for the majority, if not all, of scientific journals. In addition, it seems that there is an association among the type of the paper (original article vs. review paper), the type of journal (nonclinical vs. clinical), and the number of authors with a higher proportion of self-citations in the multivariable analysis.
|
ACKNOWLEDGMENTS
M.E.F. takes responsibility for the integrity of the data reported in this essay and the accuracy of the data analysis.
FOOTNOTES
The opinions expressed in editorials, essays, letters to the editor, and other articles comprising the Up Front section are those of the authors and do not necessarily reflect the opinions of FASEB or its constituent societies. The FASEB Journal welcomes all points of view and many voices. We look forward to hearing these in the form of op-ed pieces and/or letters from its readers addressed to journals{at}faseb.org.
REFERENCES
Related Articles
This article has been cited by other articles:
|
|
 |
|
  M. E. Falagas, V. D. Kouranos, R. Arencibia-Jorge, and D. E. Karageorgopoulos Comparison of SCImago journal rank indicator with journal impact factor FASEB J, August 1, 2008; 22(8): 2623 - 2628. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Glick You are what you cite: The role of references in scientific publishing J Am Dent Assoc, January 1, 2007; 138(1): 12 - 14. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
