Impact Factors Can Mislead Was 'Valid' Or Not
Total Page:16
File Type:pdf, Size:1020Kb
going to determine whether one's research Impact factors can mislead was 'valid' or not. Research productivity is now evaluated by a National Committee SIR - Impact factors (IFs) for scientific purchased from ISL In each category we selected by the Spanish Science and journals, developed by the Institute for compared the ranking of journals by IF as Education Ministry, which simply rates Scientific Information (ISi) and published printed in the JCR to the one based on our one's research according to the number of in the section "Journals per category, correct IF, by calculating the number of SCI publications accumulated over a six ranked by Impact Factor" of the Journal journals moving at least 1, 3, 5 or 10 posi year period. Citation Reports (JCR), are frequently used tions. The table shows the five categories This may work in some fields, but it fails to evaluate the status of scientific journals affected most severely, measured through in botany, geology and zoology, for exam or even the publication output of scientists. the percentage of journals moving at least ple, because these disciplines do not gener The IF of a journal in year T is defined as one position in the ranking. The categories ate as many articles or citations as, say, the number of citations in year T to docu- listed relate not only to the medical sciences biotechnology or genetics. The combination ments published in that journal in years but also to chemistry and engineering of less frequent citations and a publishing T - 1 and T- 2 , divided by the number of sciences. The percentage of journals drop- market now saturated with new journals FIVE CATEGORIES SHOWING u\RGEST DIFFERENCES IN RANKINGS ping by at leaSt 10 posi- renders the SCI estimated impact factor of BY IF PRINTED IN ISl'S JCR AND BY CORRECT IMPACT FACTOR tions exceeds 3.5 in all these publications close to zero. Many titles categories except one. are never indexed by the SCI. Category No. % Journals with rank difference Among these, there is a I am a graduate zoologist, doctor and journals 2: 1 2:3 2:5 2:10 substantial number of university professor who was failed by a 'top' journals. National Committee which, it seems, takes Chemistry 130 89.2 40.8 6.9 4.6 Our analyses suggest account only of the SCI ratings. It is ironic Medicine, general that journal editors or sci never to have failed an examination as an and internal 152 80.3 32.2 11.2 3.9 entific publishers could, undergraduate, to have been awarded a Pharmacology in principle, artificially doctorate in zoology and to have obtained a and pharmacy 171 78.4 25.7 8.2 3.5 Medicine, research raise the IFs of their jour professorship and then to have been failed and experimental 86 76.7 14.0 7.0 2.3 nals. To put it bluntly, by a National Committee. Despite a Engineering, electrical if a scientific publisher research track record judged by editorial and electronic 114 70.2 18.4 6.1 3.5 succeeds in publishing boards, granting agencies and international important review articles colleagues to have been active and produc citeable documents published in that jour as an editorial, or including a lively corre tive, I was graded 4 on a scale of 10. nal in years T- 1 and T- 2 . But the spondence section, the IF of his or her jour There is considerable room for improve concept of citeable document is not defined nal may go up substantially. Moreover, ment in the way the system operates. accurately by ISL scientists whose publication output is Jose I. Saiz-Salinas Eugene Garfield has pointed out that, weighted by the IFs printed in the JCR may Dpto Biologfa Animal y Genetica, for 40 leading medical periodicals, journals see their scores descend when correct IFs Universidad def Pafs Vasco, differ with respect to the numbers and types are used. 48.080 Bilbao, Spain of documents they publish, and variations Evaluators of scientific output or journal exist in impact for different types'. We performance should be cautious in using obtained evidence that the IFs of many the impact factors printed in ISI's Journal Slow, slow... journals included in the Science Citation Citation Reports. Index (SCI) are inaccurate because of an H.F. Moed SIR - You perpetuate the myth that HTTP inappropriate definition of citeable docu Th. N. van Leeuwen is inherently slower than FTP for single file ments2. Centre for Science transfers (Nature 380, 380; 1996). This is ISi classifies documents into types. In and Technology Studies, patently untrue; both protocols relegate the calculating the nominator of the IF, ISi Leiden University, responsibility for file transfer to a lower counts citations to all types of documents, 2300 RB Leiden, level protocol, TCP (transmission control whereas as citeable documents in the The Netherlands protocol), identical in both cases. In fact, denominator ISi includes as a standard only HTTP is marginally more efficiently 1. Garfield. E. Current Contents. 12 January 1987. normal articles, notes and reviews. How 2. Moed, H. F. & van Leeuwen, Th. N. J. Am. Soc. Inf. Sci. designed than FTP as it transfers the file ever, editorials, letters and several other 46, 461--467 (1995). down the same connection as the request, types are cited rather frequently in a num whereas FTP opens an entirely new connec ber of journals. When they are cited, these tion for the file transfer. types do contribute to the citation counts in Failed professor So, if it can be experimentally verified the IF's numerator, but are not included in that an HTTP is indeed slower than FTP ( a the denominator. In a sense, the citations to SIR - Your journal has shown itself sensi fact often claimed but rarely supported by these documents are 'for free'. For instance, tive about Spanish scientific policy and any evidence) one must conclude that the taking into account only citations to normal assessment of and awards to university difference is caused by poor implementa articles, notes and reviews in The Lancet, professors and researchers. tion of the protocol, either by the client or the 'correct' impact factor of this journal in The present system of marking and the server, or both. 1992 would be 43 per cent lower than the IF assessment, based only on the Science Of course, HTTP is poorly designed in listed in the JCR. Citation Index (SCI) of the Journal Citation other ways, particularly when used to trans We calculated for each SCI journal a Reports, is unfair to those who have con fer modem Web pages made up from many 'correct' IF for the year 1994 by taking into tributed to international journals but appar files, as the article correctly points out. account in the IF's numerator only ently not the right ones. Ben Laurie citations to articles, notes and reviews, and Nobody doing research in Spain at the A. L. Digital Ltd, including in the denominator the same beginning of the 1980s could foresee that a 5 Fairlawn Grove, three types. We used a special datafile list of journals compiled by the SCI was London W4 5EL, UK 186 NATURE · VOL 381 · 16 MAY 1996 .