correspondence Journals: redundant publications are bad news Publishing the same work twice is unethical and casts doubt on the integrity of research.

Sir — We have developed an electronic publications remain the central systematic search tool to estimate the score> 0.9 requirement for academic advancement, amount of duplicate publications in the 0.9 > score> 0.8 a reasonable solution seems unlikely. 70 ophthalmological journals listed by Nevertheless, it is imperative that the 0.8 > score> 0.7 Medline. Our results show that there problem of redundant publications be 0.7 > score> 0.6 is a considerable number of duplicate addressed, for it is the responsibility of all publications. If this holds true for other 0 50 100 150 those who care about objective research disciplines, it is bad news for research. and evidence-based medicine. For our survey, we matched the title Figure 1 Estimated number of redundant Stefania M. Mojon-Azzi*, Xiaoyi Jiang†‡, and author(s) of each of the 22,433 articles publications for matching scores of 0.6 or more, Ulrich Wagner*, Daniel S. Mojon‡§ published in the 70 journals between 1997 where 1ǃtotal overlap. *Research Institute for Management in Health and 2000 using a duplicate-detection Services at the University of Applied Sciences, algorithm1, and found that 13,967 pairs careful peer-reviewers or editors, they St Gallen, Switzerland of articles give a matching score of 0.6 or cannot provide complete protection. †Department of Electrical Engineering and more. Of these, we manually reviewed a Scientific journals can combat redundant Computer Science, Technical University of Berlin, random sample of 2,210. We found 60 publication in various ways2, but in Germany genuinely ‘duplicate’ publications and practice the penalties for duplicate ‡Department of Ophthalmology, Kantonsspital, estimate that 1.39% of the analysed articles publication are minimal3. 9007 St Gallen, Switzerland are redundant. Because of the very Proper deterrents are needed: for §Scientific Secretary, Swiss Society of restrictive selection process and the example, better education on publication Ophthalmology, Kantonsspital, St Gallen, impracticality of detecting all duplicate guidelines, the introduction of registers Switzerland publications, and because the estimated for planned and ongoing clinical trials, 1. Jiang, X. & Mojon, D. S. in Proc. 1st Int. Workshop New amount of duplicates increases with lower and a change in assessment criteria from Developments in Digital Libraries 79–88 (ICEIS, Setúbal, Portugal, 2001). matching scores (Fig. 1), we regard this quantity to quality when papers are 2. Cho, B. K. et al. Ann. Thorac. Surg. 69, 663 (2000). estimate to be the tip of an iceberg. submitted for posts or grants. As long as 3. Franken, E. A. Acad. Radiol. 5, 407–408 (1998). Of the 70 journals, 32 were victim to duplicate publication — 27 journals published the first paper and 26 the world except for all the others, members of duplicate, on average 6.4 months later Journals: how to decide the research community should cooperate (standard deviation 4.7, range 0–21.3 what’s worth publishing to answer several questions. months). We found no statistically We need to know whether peer review significant difference between the average Sir — Your News Feature (Nature 419, (in whatever form) is more effective than journal impact factor of the first (1.13) and 772–776; 2002) raises important questions alternatives. Does it identify submissions the second journal in which the duplicate about the reliability of peer review, but of higher quality than do other selection article was published (1.42) (Wilcoxon- falls back on the justification often used methods, or chance, or no selection? Does signed ranks test P>0.1). The analysed by editors to shield themselves from peer review significantly improve the publications were by 210 authors, widespread dissatisfaction with the system clarity, transparency, accuracy and suggesting by extrapolation that a total of as currently practised: “If it ain’t broke, usefulness of published papers compared 1,092 authors could have been involved in don’t try to fix it”. with the submitted versions? redundant publication during the time We believe it may never have been If peer review in its current, descriptive period that we analysed. The scientific working in the first place. form is ineffective or less than effective, conclusions of the original and of the Perhaps peer review, in its current we should experiment with more duplicate(s) were identical in 88.3% of form, cannot be expected to detect fraud. analytical forms of assessment. For cases; we found slight changes in 6.7%; and But can we even rely on it to improve the example, the quality of a new study could major changes (different results despite chances that what is published is the best be assessed in the context of a pre-existing identical samples, or omission of patients) science, communicated as accurately as systematic review of studies on the topic. in 5% of cases. possible, and that what remains Such a population approach may make it Duplicate publications are unethical. unpublished is dispensable? easier to assess the contribution of an They waste the time of unpaid, busy peer Various studies, mostly in biomedical individual new study. At the same time, reviewers and of editors; inflate further the journals, have reported only modest assessment should be standardized already over-extensive scientific literature; author satisfaction (at best) with the and specific for different experimental waste valuable production resources and review process, irrespective of the quality designs, and peer reviewers should be journal pages; lead to flawed meta-analysis; of the review. Papers that eventually trained to use a single, structured- exaggerate the significance of a particular became very highly cited were often assessment instrument. set of findings; distort the academic reward rejected by the journal of first choice. Peer Ultimately, it is the larger population system and copyright laws; and bring into review is costly, biased, can be inefficient, of readers (rather than a possibly biased question the integrity of medical research. does not always identify important work, sample of referees) who should decide Republication of data yields no benefit and can allow publication of articles with whether the changes made during review other than to the authors. serious deficiencies or omissions. substantially improve the document as a It is important that journal editors can Rather than falling back on the record of a peer’s contribution to science. trust their authors. Although many churchillian cliché quoted in your feature New systems should be tried that involve duplicate publications are discovered by that peer review is the worst system in the readers in the review process either after

NATURE | VOL 421 | 16 JANUARY 2003 | www.nature.com/nature © 2003 Nature Publishing Group 209 correspondence

‘traditional’ review (as in post-publication with submissions from authors hopeful of demanding than to reproduce received commentary, rapid replies and the like) publishing in a journal that will give them wisdom. I hope that the presence of active or by developing a ‘definitive’ text by bench-credibility in a world of instant research groups in high schools may help consensus before publication. Language judgements, these editors must screen to correct this misconception. experts have been investigating readers’ submitted papers to see if they meet the Peter Hughes reactions to texts for many years; it is time journal’s needs before sending them Westminster School, 17 Dean’s Yard, for editors and publishers in the ‘harder’ out for peer review. Therefore, most London SW1P 3PB, UK sciences to use their methods to extract submissions are rejected for reasons other useful experimental data from these than flawed scientific reasoning. reactions. I have no criticism of this approach: it Tom Jefferson makes sense in the commercial world of Animal research needs Health Reviews Ltd, Via Adige 28a, Anguillara journal production. The problem arises organized defence Sabazia, Rome, Italy when scientists and administrators of Karen Shashok science use the placement of papers to Sir — Your Opinion article “Promoting Comp. Ruiz Aznar 12, 2-A, 18008 Granada, Spain judge the worth of researchers, the worth animal research” (Nature 420, 447; 2002) of institutions, the best places to award delivers a much-needed message. grant money and the best places to fund Ten years ago, I volunteered to join a fellowships. The more we couple National Institutes of Health programme Journals: impact factors allocation of resources to publication in to educate young people about the need are too highly valued ‘top’ journals, the more we are effectively for intact animals in biomedical research. handing over the direction of research to a Local high schools and colleges were Sir — Linda Butler in Correspondence small group of professional editors, who sufficiently receptive to encourage me (Nature 419, 877; 2002) shows that never sought this responsibility and who to continue. researchers in Australia are publishing (excellent at their intended jobs though In the past four or five years, however, more papers since the number of they may be) are unlikely to be the best my approach to the education authorities publications was introduced as a people to bear it. has fallen on deaf ears, although all of the performance indicator for research. Butler Most of us are, at least sometimes, the teachers, and many of the students, voiced points out there is now “little incentive to judges as well as the judged. If we do not praise for the programme early on. One strive for placement in a prestigious consistently take the trouble to judge teacher told me that they did not want to journal. Whether a publication is a papers by their content rather than by their run into problems with animal activists groundbreaking piece in Nature or a location, the direction of science will come by allowing me to speak. Most of the pedestrian piece in a low-impact journal, to be determined, however uninten- population is disappointingly uneducated the rewards are identical”. tionally, by an editorial élite. We shall have about science and a significant percentage The point is well-made, but her phrase only ourselves to blame. is anti-science. highlights another growing problem in Jamie Davies Next month, I will send approximately measuring performance which, if Edinburgh University College of Medicine, 30 letters to local colleges and high schools unchecked, threatens to have a major Teviot Place, Edinburgh EH8 9AG, Scotland in an attempt to rekindle the interest. I impact on science policy and progress. believe that major biomedical and medical The problem is an over-reliance on societies, and journals, should constantly journal impact factors to judge the urge an educational campaign to deliver worth of scientists. Bright students enjoy our message to the public. It is increasingly common to hear correcting the textbooks People like myself are very willing to scientists making snap judgements about volunteer to speak, design handouts, and the quality of others’ work simply by Sir — Students aged 16–17 have been so on, but a central focus group is needed. perusing the names of the journals in doing chemistry research at Westminster Charles G. Smith which they publish, with no actual attempt School for the past five years (see the News Address supplied to read their papers. This is a dangerous feature “Put your lab in a different class”, habit, for quite brilliant work can appear Nature 420, 12–14; 2002). Our projects in a ‘lesser’ journal, either because its all have their origins in the normal subject is not currently fashionable or curriculum, as many points of quite DNA discrepancy because its author has special reasons for elementary chemistry have not been Sir — We should be able to trust any preferring a specialist forum. The habit is investigated for half a century or more. author, whether or not a scientist, to also dangerous because it erodes the With modern techniques we can amplify deliver an accurate description of the past. capacity of the research community to (and often correct) what is written in the Indeed, your final editorial of 2002 exhorts determine its own direction. standard textbooks. scientists to work to retain the public trust An ex-colleague of mine, for example, Our first paper, on the addition of (Nature 420, 719; 2002). Thus, it is even liked to publish his excellent work on nerve hydrogen halides to alkenes, has now been more unfortunate that Naturejobs states in regeneration, which could have been published (J. Chem. Soc., Perkin Trans. 2, the same issue (Naturejobs 3; 19/26 published anywhere, in a very specialist 810–813; 2002), and other work is nearing December 2001) that Rosalind Franklin surgical journal because that is where he completion. Students gain by having to and Maurice Wilkins worked on DNA thought it would be most likely to inspire think about a problem for a year or more, structure at the University of Cambridge: immediate clinical use. and experiencing the disappointments they were famously, of course, at King’s The professional editorial staff of very as well as the satisfaction inherent in College London. high-impact journals such as Nature have original work. Alex May a primary responsibility to the success of Too often, bright pupils are put off Division of Mathematical Biology, National their journal: circulation, advertising, from studying science because they think Institute for Medical Research, The Ridgeway, impact statistics and reputation. Deluged they will be asked nothing more Mill Hill, London NW7 1AA, UK

210 © 2003 Nature Publishing Group NATURE | VOL 421 | 16 JANUARY 2003 | www.nature.com/nature