Medical Writing, Vol:8 / No:1, 2000
IN-DEPTH INTERVIEW :
Interview with Eugene Garfield, Ph.D.
By Chungji Kim
On April 19, 1999, the Asahi Newspaper reported in its science section ("Kagaku") that a paper of Prof. S. Nagai, of the Osaka University School of Medicine, was ranked number one in terms of citations in scientific papers published during the previous year. His article on apoptosis, published in Nature in January of 1998, was the first Japanese paper to top the annual list of the most often cited papers. The Asahi article added that the ranking was based on a citation analysis performed by the Institute for Scientific Information (ISI) and was reported in its publication, Science Watch.
In a district of Philadelphia called University City, the 4-story ISI building is nestled among those of the University of Pennsylvania, Drexel University, and other institutions. Its reception area and corridors are adorned by a variety of artwork, which seems to symbolize how science and art are intertwined, as is shown by the group of ISI databases: the Science Citation Index (SCI), the Social Sciences Citation Index (SSCI), and the Arts & Humanities Citation Index (A&HCI). Dr. Eugene Garfield, the founder of ISI and now its Chair Emeritus, met me at his office on the 3rd floor. During the interview he frequently turned to his computer to access his own website as well as others and retrieved documents relevant to our conversation. When he handed me the first printout, he remarked that this was turning out to be a new style of interview, the documented interview. Many of these documents are cited below in this article.
[Note: When the year in a citation has an asterisk (such as 1998*), the cited document can be found in "Eugene Garfieldís Bibliography" as full text on his website <http://www.garfield.library.upenn.edu>.]
History of Academic and Professional Career
Q: Will you share some of your memorable episodes in your career?
I was born in the Bronx, New York, in 1925. As I had developed a great interest in the American West as well as chemistry and mathematics, I went to the University of Colorado to study chemical engineering. I was then seventeen. During World War II, I served in the army, but I stayed in the U.S. When I came out of the army, I went back to school, this time further west, in Berkeley, California and became a pre-med major. Two years later, I decided to go back to New York. I was going to school under a special provision of the GI Bill for people who had received a medical discharge, but I was told that I did not have enough years left under this Bill to finish medical school. So I switched to biochemistry.
After earning a B.S. in chemistry at Columbia University, I started working there as a lab assistant to a professor of chemistry. After my second laboratory explosion, however, I decided that this wasnít the right career for me and looked for something else. As part of one job interview, I went to a meeting of the American Chemical Society held in New York. I wandered into a session chaired by Prof. James W. Perry, then of Massachusetts Institute of Technology. At this session I suddenly realized that one could earn a living by doing literature searching and documentation work on chemical information. I talked to Prof. Perry right after the session. A month later, he introduced me to Dr. Sanford Larkey, of the Johns Hopkins University School of Medicine, Baltimore, who was looking for a chemist to work on his research project, the Johns Hopkins Universityís Welch Medical Library Indexing Project (the Welch Project). The main purpose of this project was to investigate new methods to produce Current List of Medical Literature--the predecessor to the present Index Medicus. Until I joined this project in 1951, I had never done any indexing, but I was interested in the cataloging of information. I had been fascinated by books and libraries since my childhood.
The Welch Project was funded and sponsored by what is today called the National Library of Medicine (NLM). NLM originally was the Surgeon Generalís Library, then the Armed Forces Medical Library, and finally, NLM. Thatís how I learned all about indexing methods, including the use of controlled thesauri such as Medical Subject Headings (MeSH). The project aided the Current List of Medical Literature in revising MeSH. We also investigated other indexing methods like Chemical Abstracts (CA), Biological Abstracts, as well as new methods of machine indexing and compilation. I was not an indexer per se, but a researcher. In late 1952 I came up with the idea for a Ďtable of contentsí page service, covering library and documentation journals. I called this service "Contents in Advance." I also served as a volunteer CA abstractor of Spanish pharmacology papers. Through the Welch Project I came to know most of the pioneers of information science, in part, because this project was the only one in existence at that time.
In 1953 the project received a lot of media coverage when I organized a symposium on machine methods for scientific documentation. William C. Adair, a retired Vice President of the Shepard Citations, Inc., read newspaper accounts of our project. He wrote to me and said that when he was Vice President, he had been approached by several people asking whether or not the companyís citation system, Shepardís Citations, could be applied to engineering or science. But the company decided not to do anything about it because their main customers were lawyers; science was completely out of their field. And the size of the science engineering literature was vastly larger. As a result of this letter, I went to a public library in Baltimore to look at Shepardís Citations. When I saw it, I exclaimed, "Eureka!" because in my mind I had been looking for this kind of structure. I called him and said, "Mr. Adair, I want to write an article about a citation index for science, but I want to give you credit. As the Associate Editor of American Documentation, I invite you to write a short article first so that I can cite it." He did and it was eventually published in American Documentation in June of 1955. After the Welch Project ended the summer of 1953, I decided to begin work for a masterís degree in library science at Columbia University and completed a paper, "Citation Indexes for Science," which was published in Science in July of 1955*. So being a part of the Welch Project for two years was a key turning point of my life.
Q: Did you have to modify Shepardís Citations a lot for science?
Oh, absolutely! First of all, Shepardís Citations is called a citator system. It was never called a "citation index." It was confined to case law. The legal system in the USA, except Louisiana, operates under the doctrine of stare decisis, the Latin term which means that all courts must follow their own precedents. Any American lawyer involved in a trial should know which relevant cases should be cited. The lawyer must first find a relevant previous decision relating to his or her current case by consulting a digest, index, or encyclopedia. The Digest will provide the coded citation for the case in point. Then the lawyer looks up this reference in Shepardís Citations and finds the subsequent citing cases and whether the given decisions: (a) affirmed, (q) questioned, or (o) over-ruled. Although Shepardís Citations lists the new cases which refer to a particular starting case, it has nothing to do with searching the law journals.
While there are thousands of law cases recorded from all 50 states, they are still not as numerous as citations in the scientific literature. Already at that time, the order of magnitude of citations in scientific literature was 50 to 100 times greater than for law cases. It was only after we started SSCI in the 1960s that the Shepard organization realized the necessity of a citation index covering the law literature. They started receiving complaints from law librarians about their failure to cover the law journals. So, in 1968, they introduced Shepardís Law Review Citations.
Q: How did you establish your company, ISI?
Even before I had received a masterís degree in 1954, I was asked to work on machine methods for storing drug information by a pharmaceutical company in Philadelphia. Since I wished to do other work as well, I was retained as an independent, private documentation-information-consultant. In the following year I started publishing privately a contents-page service, called Managements DocuMation Preview. This service was adopted by another client, Bell Telephone Laboratories, which was interested in covering the business management literature as well, including social and behavioral science. We soon changed its name to Current Contents (CC) of Management and Social Sciences. In 1954, I entered a Ph.D. program in structural linguistics at the University of Pennsylvania. I produced another CC, CC of PharmacoMedical, Chemical & Life Sciences to distribute among drug firms. In 1958 I converted it to a subscription service under the name of CC of Life Sciences so that physicians and individual researchers also could subscribe. In 1960 I changed my companyís name from Eugene Garfield Associates to the Institute for Scientific Information, and in the following year I received a Ph.D. degree.
I still remember very vividly the moment in 1959 when Prof. Woodward of Harvard University called; he later (in 1965) won the Nobel Prize in chemistry. He saw an article listed in CC of Life Sciences that he needed very badly. One of his competitors in Germany had published something that he had to see right away. He said to me, "We donít have this article in our library, yet." In those days, it was not customary to receive journals by airmail. It was expensive and there was no air cargo, either. However, at ISI, we had journals sent to us by airmail. So I said to him, "I have the article here and Iíll send you a copy." He answered, "I have to know now!" Fortunately, the article was very short, just one or two pages. So I read it to him on the phone because I knew German.
What I learned from this is that CC could alert you to the existence of an article but would not bring you actual articles. Itís sort of like leading a horse to water but not letting it drink. So I decided to start a document delivery service. In 1967 we began a service called ASCA, the Automatic Subject Citation Alert, which is a personalized selective dissemination of information (SDI). This is the oldest and in fact the first commercial SDI service. Users produced their own profile using keywords, key authors, and key citations, reflecting their interests. In response we provided them with computer-generated ASCA reports. The users could change the profile whenever they felt it necessary. When users asked to receive copies of the actual articles simultaneously with the bibliographic report, we sent them; we called this service, "ASCAmatic." Many pharmaceutical companies found this service very useful. However, researchers in an academic position did not use this service very much. Instead, they requested a reprint directly from authors because ASCA and CC provided addresses for authors. Actually, reprint requesting became a major communication tool among scientists. If you want to know more about ASCA, you simply type in "asca" in "Full Text Search" on my website, <http://www.garfield.library.upenn.edu>.
Q: Your website is at the University of Pennsylvania. What kind of affiliation do you have?
I am on the Board of Overseers of the University Library. I taught information retrieval theory there for many years and remained an adjunct professor in computer and information science. One Ph.D. student eventually received a doctorís degree under my direction covering on-line searching. I still have a close connection to the University, and recently donated several murals for the library and the museum.
Q: Were there any opportunities for your company, ISI, to become a government-funded institute or become a National Institute for Scientific Information?
In 1958 I received a letter from a Nobel prizewinner in medicine, Prof. Joshua Lederberg in the Genetics Department of Stanford University. He said that my citation indexing system, described in the 1955 Science article, should be developed. He encouraged me to apply for a governmental grant and helped me with the application. In 1960, ISI received a 3-year grant from the National Institute of Health (NIH: now the National Institutes of Heath) to work on a Genetics Citation Index Project.
Unfortunately some other private company had misused grant money from the U.S. government so NIH decided not to give any grants to for-profit organizations. So to receive money from the government, you had to arrange to get a contract. So the original grant was converted to a contract and NIH transferred funds to the National Science Foundation (NSF) because NSF was used to dealing with contracts. Under a contract, you had to deliver a product that your customer orders; it is different from a grant. So we produced an experimental Genetics Citation Index, which covered three separate periods of the literature in genetics. Part I covered 15 years of one journal, Part II covered 35 genetics journals for five years, and Part III covered the entire science citation index literature for 1961. It was distributed free to 1,000 libraries and geneticists in this country. The most important section, Part III, selected material from 100,000 scientific citation index covered journal articles. I asked the NSF staff, "Why donít we continue and produce other indexes, the full science citation index, so other fields can benefit?" They wouldnít do it! So we had to decide whether or not we should invest the money not only to produce the 1961 Science Citation Index (SCI) but also to promote it. I took a calculated risk to publish it for $500 in 1963. Then we also announced a quarterly SCI service to begin in 1964 covering 613 interdisciplinary journals and 1.4 million citations. Back then we used keypunching and magnetic tapes for computer processing. After we launched SCI, ISI began to plunge into deep financial trouble [here Dr. Garfield laughed], which took me years to recover from.
If NSF had decided to continue to publish SCI, could it have been better? Itís difficult to say. All would depend on the people involved, but it could have been a success. I guess that SCI would probably have grown faster. Under the government, it might have been cheaper; maybe we could have covered more foreign journals right away. I would have preferred to sell SCI to individuals or to departments rather than just to big libraries and pharmaceutical companies. On the other hand, there might have been limitations. For example, I donít think that it would have been as valuable if NLM had decided to take over the SCI project because NLM was limited to medical journals. Would NLM have been willing to branch out into ancillary fields? One of the advantages of our SCI is that it is multidisciplinary. To cover the science of medicine, for example, you have to have biochemistry, molecular biology, and other journals in borderline fields; this is crucial. Only recently, the government has launched PubSci, a government website that will offer a multidisciplinary search outside of medicine, which is getting back to the idea of an almost unified index. I think that NIH did have a multidisciplinary approach to research on subjects of medical importance. But we never later went back to NIH to ask them if they wanted to sponsor the SCI project, since they produced Medline. But we, as a private company, were able to do the work as we had wanted and planned. The success of CC, in particular, helped us to go on.
Q: For 31 years you wrote a column in CC, "Essays of an Information Scientist." What made you start writing those essays?
Subscribers have a kind of love-hate relationship with CC. They know that CC provides valuable information, but itís hard work going through CC and staying on top of a given subject. So my idea was that we should give the subscribers something light to read, some interesting editorial materials to reduce this stress. In 1962, I started writing very, very short pieces, one or two pages, and had just two essays for that year. In 1970, the number of essays grew to 36, and I began a weekly column in 1972. Gradually over the years, it grew and grew, becoming a major editorial effort. In 1977, we started Citation Classics; we invited the authors, whose papers were highly cited, sometimes over 500 times, to write their story or make a comment. In smaller fields, we had lower citation thresholds. This column, which eventually appeared in the front of all seven discipline-specific CC editions every week, was equivalent to a journalís editorials. I got a lot of ego-gratification from writing those columns. There was also a public-relations effort, a way to explain the citation index to people, which is not an easy job. We received good feedback; our readers told us that they liked the columns. The columns helped people to understand the citation index, too. We had a captive audience. So we kept it up. At the peak I had an editorial staff of 15 people just for this.
The column was discontinued in 1993 when I sold ISI to the JPT Publishing in New York. They cut the column purely to increase short-term profit, which I think was a mistake. I personally feel sorry that I didnít do something about it, but I had already launched the biweekly newspaper, The Scientist, and my editorial efforts shifted toward that. Many times I regret that I didnít continue because by now there would have been an added seven-yearís worth of essays. We could have had more Citation Classics articles in CC, too. Anyway, all of the essays were compiled as 15 printed books and now you can browse all of them on my website. Only the first 2,000 Citation Classics were published in book form and I am still hoping to put all five thousand on my website.
Q: Why did you start The Scientist?
Good question. But it is a long story. The best thing is to look at the website <http://www.the-scientist.com>, where I explain the mission of this newspaper. Anybody can access the site; it is free of charge. Basically, it deals with the human side of science, or more precisely, of life sciences. This newspaper started in 1986 and used to be published by ISI, but it has not been connected to ISI for almost 10 years now. I am the President and the Editor-in-Chief of this newspaper. The office and staff are in a different building, just one block from here. But I spend most of my time here at ISI. After I sold ISI, I remained as Chair for some time, and Iím now Chair Emeritus. I am treated very much as a university professor emeritus. I have a limited consulting contract. All of my staff here at ISI are on the payroll of the newspaper, working for me in this ISI building. I just feel comfortable here, you know, like wearing an old shoe [here Dr. Garfield smiled]. And I believe the feeling is mutual. But I have no say in the day-to-day decisions of this company, which is now owned by the Canadian firm calling Thompson.
Citation Index versus Subject Index
Q: What are the advantages of a citation index over a subject index?
Medline, for example, is based on a subject index method used in Index Medicus; it uses MeSH to control vocabulary. Index terms have to be determined in advance, and professional indexers are involved. So a subject index is vulnerable to scientific obsolescence, subjective judgement, and considerable delay. SCI, on the other hand, is a hybrid system based on natural language generated not only from title words but author-provided terms as well as expanded keywords, called "KeyWords Plus." In other words, SCI has its own system of self-generated index terms, which is also well suited for man-machine indexing and derived from the papers which are cited in current articles. However, the piece-de-resistance of SCI is the Citation Index proper, the cited-reference index, which completely bypasses language all together.
You have to understand the whole philosophy of what citation index means. It is not easy to explain in just a few words, but the point is that any document can be a point from which indexing can begin. Letís say that you are writing a paper about Subject X. Even though you do not put the word X in the title of your paper, people understand that your paper addresses a very specific topic of Subject X. There could be hundreds of papers about the subject of cancer radiology, for example, but each paper is not just about cancer radiology; it is much, much more specific.
Using SCI CD-ROMs or one of the ISI websites, the Web of Science <http://www.webofscience.com>, you can find a particular paper of your immediate interest, letís say Paper A, by just typing any KeyWords, author name, address word, journal name, the year of publication dating almost back to 1945 now, or any combination of these in "General Search." This is how you conduct a traditional subject Search. But once you locate Paper A through one of these categories, you can find not only all the references that Paper A cites but also the papers that cite Paper A. In addition, you can do a specialized kind of citation search, called "Related Records," in which you can find all the other papers that are closely related to Paper A. At the top of the "Related Records" summary list is the paper that is most related to Paper A; this paper may or may not cite Paper A, but these two papers are closely related because they cite many of the same papers; they have references in common. Now, you can go down the list, or you also can go sideways to see which other papers have cited this most closely related paper of Paper A. You may find hundreds or thousands of related papers. It is up to you how far you want to go with this expansion, but you are able to judge at each step of the way from the title and its abstract whether or not you are getting closer to what you are looking for. We started this method years ago. Medline or its website, PubMed, now recently adopted the term "Related Articles"; this is similar but not the same and certainly is not based on citation relatedness but other semantic criteria.
In the printed versions of SCI, we could not do this kind of search very quickly. Long before the Internet began, we started, in 1988, to provide CD-ROMs. Using CD-ROMs or Web of Science, you can do a citation search in 10 minutes that used to take hours to do. If your institution has a site license for Web of Science, you can combine all three citation databases: SCI, SSCI, and A&HCI.
You could say that citation analysis is like fishing. You start with a line and hook because you want one fish. Then you want to know which school of fish that one fish is connected to. You decide what size of a net to use; you can expand your search with a big net, but you also can narrow it down using a small net.
Q: How do you compare medical literature searches in SCI and Medline?
I always tell researchers that they can combine SCI search with Medline. Every researcher should know this. You can use Medline, but as soon as you find an interesting paper, you go with that citation into SCI and then you do a citation search. Use Medline as a quick and dirty way to get a starting reference; a starting reference is the key point. If you do this process with your own paper, it will tell you where you are in your research field; in this sense, SCI is a navigational tool.
Citation Index was designed as, and is, a system of very precise information retrieval. Information retrieval serves several purposes. You come to know papers from the past, which I call information recovery. Then these papers open up new ideas to you, that is information discovery. You can also trace the historical development from early papers to current papers. If you have a certain paper in your field that is frequently cited, you know you can say that the paper has a high impact in your field. Using the ISIís citation databases, you can find out who are the most cited authors on a very specific subject. If you want to know who the leading figures are in your research field, that is the way. I myself could do this search without knowing anyone in your field.
We saw from the beginning that citation analysis could be used in a quantitative way to measure the impact of papers. You also can measure impacts by institution, country, or journal.
Impact of the Journal Impact Factor
Q: Is the Journal Impact Factor (JIF) your favorite subject to talk about?
Is it my favorite subject? [Here Dr. Garfield laughed.] There is always so much else to talk about but somehow people donít let me avoid it. I am asked to speak and write about impact factors more than any other topic.
Q: How did you create JIF?
I mentioned "impact factor" in my 1955 Science paper already, but very vaguely for a historical purpose. In the early 1960s, we created JIF to help select journals for SCI, and JIF has been calculated as follows since then:
Calculation for the Journal Impact Factor (JIF), an example:
Then I wrote a paper, "Citation Analysis as a Tool in Journal Evaluation" (1972*) mainly to help librarians choose journals for their library. In 1979, we decided to report these numbers in SCI Journal Citation Reports (JCR) each year. 1999 JIF = (number of citations in 1999 to articles published in 1997 and 1998) / (number of citable articles published in 1997 and 1998).
About 15 years ago, I did a study on internal and general medical journals ("Which Medical Journals Have the Greatest Impact?" 1987*). There I emphasized "article-by-article audit" of journals. What I mean by this is that one should compare journals for each type or category of papers. We first categorized the articles into research articles, editorials, letters, reviews, and so on, mainly using 1981 and 1982 SCI. Then we compared the citations and total impact factor for each category across the journals.
Itís an interesting story. I wrote this paper because Dr. Arnold Relman, who was the editor of the New England Journal of Medicine (NEJM), asked me. After I wrote it, he made me rewrite it two or three times and then he turned it down! He said, "The result makes NEJM look too good." He claimed it was unseemly for him to publish this paper because of this result. So I sent the paper, which was now two and a half years after I had first sent him the manuscript, to Dr. Edward Huth, the editor of Annals of Internal Medicine at that time. Ed published it immediately. I have never said this publicly, but itís all right. You can quote this; Arnold is retired [here Dr. Garfield smiled], and I must say that in spite of that experience, I have the greatest admiration for what he accomplished at NEJM.
Q: Did you anticipate then that JIF could affect the reputation of biomedical journals as much as JIF does now?
Yes, I did. I knew the influence; thatís why I wrote that paper! In my recent article ("From Citation Indexes to Informetrics," 1998*), I cited a paper on JIF of the journals in clinical physiology and nuclear medicine (Clinical Virology 17: 40-418, 1997). The Danish authors concluded that JIF in their field did a good job on ranking the journals in terms of cumulative citation frequency of papers.
Q: According to that paper of yours (1998*), 56% of all Ďcitedí papers between 1945 and 1988 had only 1 citation (Table 1). I often hear, however, that any prestigious journal should have JIF of "two" or higher. What do you think about this threshold number?
It is simply an arbitrary number; I canít justify it scientifically. In some fields, two would be the highest. Recently I wrote an article in Occupational Medicine ("Refining the Computation of Topic Based Impact Factor: Some Suggestions," 2000*). Even the top journalís impact factor in occupational medicine is low, not even two, but the highest one is the highest one. It depends on how much literature there is in a particular field and many other factors such as citation density. There are no absolutes. For example, "two" is nothing special in biochemistry because those journals have, in general, high JIF due to the high average number of references per paper in that area. However, in another field where researchers do not cite the current literature densely, JIF can be extremely low. So you have to be aware of the area or field for which you are making a comparison.
Q: Many biomedical journals have tried to increase their JIF. What do you think about this fierce competition?
Hopefully the competition will improve the quality of journals. What can increase JIF has often been discussed. In general, review papers may increase citations, but not always, depending on what topic, how good they are, and so on. There are a dozen or more types of papers. The denominator of JIF depends on what you include there. The question is whether or not ISI is consistent. There are some kinds of editorial materials that the journals are misclassifying; that needs to be pointed out. I have always said, however, if you want a good impact, you have to get good-quality papers; it is the best way.
I keep telling journal people that they should never even mention JIF beyond the first decimal place. I mean, to quote a JIF like "12.345" is ridiculous. Its JIF is "12.3"; why do you need these two extra digits? It gives a false idea of precision. Now you say that ISI does report these numbers. We only do this to give an easy way to separate journals; otherwise we would have many journals with 12.3, and these journals would have to be listed in an alphabetical order. In order to solve this problem, ISI reports the numbers as exactly as they come out. That practice probably should be abandoned.
Q: There are some critiques of JIF among medical journal editors. One example is that if a journal is lucky enough to have an occasional big-hit paper, its JIF is increased out of proportion to the journal's influence. What do you think would be different if ISI had used, for example, the median or the percentage of cited papers instead of the mean?
There are plenty of things to be done with JIF, but ISI can only respond very slowly. They are dealing with the database that is producing about 20 million citations a year. It is not easy to change things, but they should be more adaptive. They will have to make changes. I am sure that they will. If you are an editor for, letís say, the field of dermatology or anesthesia, you could find some way of evaluating impact within just that field in a somewhat different way than the standard. But ISI must cover all the journals and it is not easy to adopt methods that are different for every category.
You should know, however, that there are already many variations of impact number. Every scheme that you can imagine for modifying impact numbers has been more or less discussed inside ISI. The question is how practical it is. Besides JCR, ISI has a separate database called "Journal Performance Indicators (JPI)," which a lot of people donít know about. Using JPI as a separate ISI file, you can find the journalís impact number over each year of a 20-year period, arranged either alphabetically or ranked by impact within each category, such as "Cardiac & Cardiovascular Systems," or several categories combined.
Q: I noticed in your Bibliography on your website that you recently published articles on JIF in German for German journals. Why did you do this?
The reason I published these articles in German is because there was an editorial in a German-language journal which I found to be nonsense. So I wrote to the editor of that journal that I thought these authors had written nonsense. The editor, who turned out to speak beautiful English, called me and said, "OK, then write me a letter and I will translate it for you." He published it as a Letter to the Editor in German ("The Impact Factor and Using It Correctly,"1998*) and I posted the English and German versions on my website.
First of all, these German authors claimed that JIF has no validity because the German clinical journals have relatively low JIF. I wrote in this letter that all clinical journals in vernacular languages have a relatively low JIF; that is very natural. They are not in fact even primarily research journals; they do not publish a lot of research papers. Clinical-practice articles do not get cited as much as do research articles. Secondly, the authors had a complete misunderstanding of the nature of the ISI databases. So I had to inform them that most German research is not published in German; most of the German researchers publish in English-language journals. ISI does not discriminate against German journals. Local clinical-practice journals publish to a very narrow audience. So, for this group of German journal readers I just explained that they should not feel any sense of shame that German-language clinical journals are cited less than English journals; English today is the lingua franca of science. In the period from 1880 to 1940, German was the dominant language in science.
Now, what if they want to have their papers cited more? Should they publish in English or bilingual? Publishing in English does not necessarily improve JIF. Publishing in both the vernacular language and English will help improve the quality of the journal or maybe it will increase readership. But it will not necessarily be reflected in citations. Physicians read articles in the clinical-practice journals and use them, but they donít cite the papers. But the more the journal is read, the more its general prestige and awareness will be improved. Research physicians who donít know German would be encouraged to cite more when relevant.
Q: How should we use JIF properly?
As I said earlier, you should compare journals by type of articles, "article-by-article audit." Journal to journal comparison can be done in many ways, but a high JIF belongs to a highly prestigious journal; you canít avoid that. However, when I wrote the paper published in Annals of Internal Medicine, I had never thought that it would be taken up so heavily.
JIF is not created for all the purposes that some people want to use them for. A lot of people who are using JIF are doing it for advertising purposes. Why would some journal like Nature need to quote its JIF? Everybody knows that it is the top journal. Advertising people want to say, "You see, we have the highest impact journal: higher than Science " because these journals are competitors. They want every little extra evidence to prove something.
The real problem is when you start using JIF as the substitute for citation analysis of individuals. I have always said that you should not use JIF to evaluate a person or department. If I manage to get my paper published in Nature, does that guarantee that it is a great paper? Even if the paper is never cited?
If you want to take the citation records of a person, then you may want to compare the performance of his or her article to the average performance of that journal for the same year. One of the ISI publications, Personal Citation Report, lists "Expected Citation Rate," which is the average number of citations in that journal in a particular year. If you published a paper in a certain journal and year, you can compare your paperís citation number to this Expected Citation Rate for that year. There is, however, some problem with this comparison. Each year the number is changing. So, if your paper begins to be cited, for example, 5 years after publication, you have to allow the same number of years for other papers published that same year. Often times it is a matter of comparing papers which are most related.
Diversity and the Potential of Citation Analysis
Q: Dr. S.Yamazaki in Japan published the results of ranking Japanese research institutions based on citations received (Nature 372: 125-6, 1994). What other kinds of papers have been published by researchers outside ISI?
There are thousands of studies published using the ISIís citation databases. One of the things that I will stress is this; There have been enormous opportunities with these databases to do the history of science. You now have an easy method--I mean relatively easy method because it has never been easy--for tracing historical developments on very different topics by searching from early papers to current papers. If I spend a day with the information from Web of Science on your topic or research field, I can give you a brief history of it. Now, I say that is what every researcher should do; you should have a history of the subject that you are interested in. It may sometimes be difficult to do this time-wise, but you can certainly begin to construct it and see what is necessary to fill it out by using this citation network. It is crucial to know how citations are connected, how they are related. I think it should be a kind of routine thing for any researcher to figure out how their research is connected to the overall literature.
I have given hundreds of lectures around the world about the use of SCI. I have always said, "If you should get one thing out of my lecture or any lecture about citation indexing--I donít care about anything else--, I want you to know that every scientist should develop what I call citation consciousness." For any paper or book that you are citing you should be aware who has cited that paper. This search will tell you where you are and what you have to know. Thatís all! You can expand your search as much as you want, but you should not use a reference unless you know what has happened to it; just like a lawyer who goes into court wants to know what happened to a cited case. You should ask yourself, "Is this paper that I am citing still valid?" Somebody might have published a paper last week saying that the paper you want to cite is just a bunch of junk. Will you then still cite this paper to support your data?
In some cases it is not as easy; the paper you want to cite may have been cited a thousand times. Nonetheless, not too long ago somebody challenged the Watson and Crick original hypothesis of the double helix of DNA, published in 1953 Nature, because there is some discrepant evidence in crystallography. So if I want to cite the Watson and Crick paper, I will do a search to see who has cited it. Now, there is another side of it. This Watson and Crick paper does not get cited that much any more. All classic papers eventually suffer from, what was described in one of my essays in CC, "Obliteration by Incorporation." It is not ordinarily necessary to cite the Watson and Crick paper anymore. The greatest compliment that you can receive is that people donít think itís necessary to cite you; nobody cites the citation index any more, and not many bother to cite me anymore [Dr. Garfield smiled] even though their papers are about a very related subject. However, that is not the same thing as being ignorant of the literature.
Every week something new appears in the literature on citation; somebody finds a new way to use the citation data that I wouldíve never thought of. In my article ("From Citation Indexes to Informetrics," 1998*:), I cited a paper on gender discrimination in research grants (Nature 387: 341-343, 1997). The authors of this paper, two Swedish female researchers, are not the first to try to use the citation index to prove gender-biased research funding. But they did their study in a very interesting way. Although they donít say very much about citation analysis, that is the basis of their study. I am sure that there are many more studies that are going on but have not been published.
Q: What kinds of papers do you expect to see more of from citation analysis?
I would like to see more papers on the history of science. Particularly in Japan, reviewing the Japanese literature would be a great thing to get across the culture. Researchers should be given good academic credit for writing good reviews: not just bibliographical reviewing but critical reviewing. The Citation Index makes it easier for you to write a review. When you do it, however, you should make it sure that you have covered most of the subject.
Q: Have you ever thought that you might have accelerated the evolution of science and scientists by selecting the Ďfittestí through your creation of citation analysis?
When you said "the fittest," I think that most of the best scientists receive recognition, but there are a lot of people, for a variety of reasons, who donít. Through citation analysis over many years we found hundreds and hundreds scientists who had never been publicly recognized. That is why I started Citation Classics in CC and I think that I have helped a lot of those scientists getting recognition. I am occasionally told that we should have called some of these articles "Citation Laureates" because quite a few of those authors in Citation Classics were later awarded the Nobel Prize.
English versus Non-English in Science
Q: Do you think that being born or raised in an English-speaking country is an advantage in the career of a scientist on a global scale?
To me, for bright scientists anywhere in the world it wonít make any difference. You have brilliant scientists in Japan. When you are talking about the language barrier, I have to say that in some cases it probably did hinder their global career to develop. Thatís unfortunate, but even then it does not prevent them from doing good work. It just prevents their work from being known better, which is a shame. One example is that in 1996 when the Royal Swedish Academy of Science awarded the Nobel Prize to Drs. Curl, Kroto, and Smalley in chemistry for the discovery of fullerenes, which was called "carbon 60" at the beginning, the Academy did not include the Japanese scientist, Dr. E. Osawa, who published on that earlier. I am afraid that the information from this Japanese author was not disseminated adequately. It was a rather complicated matter, but I felt bad about this.
Q: You have visited Japan many times. What do you think about the teaching of English in Japan?
I have often said to my friends in Japan that it is unfortunate that people are not taught to speak English. When you teach a foreign language, you should begin with children speaking it: the younger, the easier. In Japan, the emphasis is on reading from the beginning. One flaw in the Japanese system is that spoken English is not emphasized. Children can learn the spoken language easily. It is much easier when they are young. Later on when they read English-language texts they will be able to speak and think about those texts in English and speak and write with colleagues and in general be able to communicate. There is also a Japanese culture that does not encourage the young to speak up.
Science is not just publishing papers; it is convincing your colleagues and peers of your ideas. It is like selling a product; you are marketing your ideas. Thatís what Japanese and everybody else have to do. Nonetheless, in spite of their tight cultural background, Japanese scientists can and do break out of their cultural straightjacket and become great scientists. It is ironic that the Japanese have not applied the basic tenet of their own famous Suzuki method in music to learning languages.
Q: There are some complaints among Japanese researchers that Americans do not make much effort to gather information from the Japanese-language literature. How do you respond to this?
Yes, I know. English dominance has been interpreted as being kind of chauvinistic or imperialistic. It is true that ISI has always been biased against non-Roman alphabet journals. Although we have from the beginning processed Japanese-language journals, we could not afford to transliterate all the Japanese characters and letters. I have never known a scientist who deliberately avoided a Japanese publication. But the cost of translation could be an impediment. I think this is a job for Japanese people to do. I recently heard that they might start a Japanese SCI. The Chinese already have their SCI.
I hope the Internet will provide a solution to the space problem. On the Internet, you could have multi-lingual versions for each paper. Of course I am aware that bringing this to reality could be very difficult for Japanese people, but if they provide such English information to ISI, I hope ISI will take it.
We are at the stage in developed science where English has become mandatory. Who knows, maybe someday it will be Chinese! Today, however, Japanese scientists who want to play in the international arena have to find a way to publish in English and they do! In November of last year, I gave a talk in Japan and showed that the same thing that has happened in Germany has happened in Japan; Japanese scientists are publishing primarily in English (Figure 1: Percentage of Non-English Papers Published Worldwide; Figure 2: Article Output by Country.) [Note: Both figures are from the talk in Japan in 1999, "A Citation Analystís Perspective on Japanese Science"]. Even Japanese-language journals now have English articles; this happens in the Republic of Korea, too.
Future in Science Communication and Information Science
Q: What do you think about currently available search engines and their index systems on the Web? Is there any room for improvements? What kind?
When I use the Web, I sometimes find it very difficult to search. Search engines can be very frustrating. The people who are running these engines have ignored 40 years of machine-indexing research. Of course their files are huge, but there would be a lot of value to their doing more research. I wrote a short article, "SDI and ĎPush Pullí Technology" (1997*). In this article I referred to our ASCA system because what they call "crawlers" on the Web give users information only one way, "push." You have to have feedback from users, "pull." I think the providers of search engines could learn some from our experience on ASCA to better meet their usersí needs.
Q: Given the information capacity of the Internet, do you think in the near future that electronic versions of scientific journals may mandate that the authors provide not only hyper-links to references but also highlight the actual quotations in each citing paper?
That is a very interesting idea. Yes, we should be able to highlight actual quotations. There already exists a method for doing context analysis. Future citation indexes will have the citing sentence, paragraph, or page. It may be called "annotated citation index." Let me tell you one episode. Many, many years ago, a professor of the State University of New York said to me, "Citation index needs to have an annotation like in Shepardís Citations." I said to him, "You donít seem to understand that we cannot provide these annotations because itís impossibly expensive. Each paper contains a list of references. What is the annotation that you are going to put there for each citation?" Shepardís Citations includes annotations but they are dealing with a much smaller volume of literature. We are dealing with not only a million articles a year, but each article contains on average about 20 references. So the only thing you could do is to automatically connect to the full text and extract the sentences that give you some idea of it.
Now, there are problems with that. Most citations are not just quotations. A citation index is not equal to a quotation index; they are not the same thing. For example, somebody who cites my paper in Science in 1955 says, "Garfield started the citation index in 1955 -- period!" This person cites my paper, but gives no quotation. You might say, "Oh! Iíd like to know if the citation is positive or negative." How can you tell if it is positive or negative? The general conclusions of the paper may be given in another part of the paper than the annotated paragraph or not even clearly stated. Most citations in science are neutral.
Q: Will the ISI databases start including on-line only journals?
They already have a few electronic journals. I think the problem is that they have not begun to index the address of websites, or URLs, being cited. That will have to happen. However, when an on-line journal changes its URL, the documents are moved to a new URL. You can eventually find it, but linking them all correctly is a problem now.
Q: Will we soon be able to hyperlink from Web of Science to NLMís websites, PubMed and PubMed Central?
ISI and NLM are discussing the possibility. The National Research Library Alliance and ISI have already been working to link each otherís website so that researchers can access full-text articles on the Web. ISI has convenient access to the full text through its document delivery service called "ISI Document Solution." The Derwent Innovations Index, a patent index system, can already be combined with the ISI citation databases on Web of Science. Researchers donít have to go to their library any more; they gather and analyze information at their desktop computer.
I am currently the President of the American Society for Information Science (ASIS). I and all the pioneering information scientists before me ("On the Shoulders of Giants," 1998*) have spent our lives working toward this kind of network. We have tried to create complete access to the world of literature so that researchers become able to see what has been written about anything that they are interested in and how these things are related. It is indeed a pleasure to see that individual scientists are now having a desktop digital library for real.