MEDOK (2) p.7-11, 1975
MINI-COMPUTER OR ON-LINE ACCESS?
Eugene Garfield, Ph.D.
E-mail: garfield@codex.cis.upenn.edu
Website: http://eugenegarfield.org
Dr. Garfield’s paper was originally presented at the meeting of "Te-klubben" (an informal association of medical librarians mainly in the Oslo region) May 26, 1975, at the Biomedical library of the University library in Oslo.
Thanks to Mrs. Elisabeth Buntz, chief librarian at the Aker sykehus, MEDOK has the possibility to publish Dr Garfield's paper. Mrs. Buntz has secured Dr Garfield’s consent to publish, and written a copy of his manuscript.
The editorial staff wishes to point out that not all the opinions presented by Dr Garfield correspond to their own beliefs.
________________________
Some of the librarians in this audience were present in Amsterdam in May 1969 where I presented a paper at the Third International Congress of Medical Librarianship. Tonight I would like to give you a short progress report on some of my earlier speculations. In short, where have we been, where are we now, and where are we going.
MEDLINE is probably the most widely used on—line bibliographic system available today. Possibly Chemical Abstracts is second. However, in the United States and elsewhere the heavy use of MEDLINE has nothing to do with its basic superiority to other systems. If a government body in the United States subsidized other systems they too would be heavily used. The use will be very heavy whenever the subsidized price competes with the unsubsidized cost of using printed indexed. Indeed, we know that many users are now prepared to pay a little more for an on—line search because they may have an aversion for a printed index. Using a terminal seems less tedious in spite of the delays in getting connected and the frequent communications problems, especially in Europe.
In Amsterdam I pointed out that there are a variety of technologies available for rapid display of information. A properly designed microfilm store of the Index Medicus might out-perform MEDLINE. But we are not yet in the habit of measuring cost—effectiveness in government subsidized systems. It is interesting to observe that in the one government subsidized system where all databases compete on a comparable level the SCI system is the one selected by most users. I refer here to Canadian SDI.
Most people in Europe have not yet had an opportunity to use the on-line SCISEARCH. Many of you have had experience of the SCI. You can readily imagine how convenient it can be to simply key in the authors name and have immediately displayed the most current articles that have cited the particular publication you started with. Indeed, the advent of SCI and SCISEARCH has accelerated and made more efficient the use of NEDLINE and other on-line systems. This can be done with either the Permuterm or Citation sections of SCI. PSI can suggest appropriate search terms. Or else the Citation Index can be used to find a needingly relevant article. Then you can call out from MEDLINE the MeSH terms used to index that tide. This enables you to use a controlled vocabulary more efficiently.
Some people believe that I am emotionally biased against the MEDLARS system because it is a competitor to ISI’s system. It is not generally known how closely I was associated with the creation and development of the MeSH system. Dr. Sanford Larkey, now deceased, former director of the Johns Hopkins Univ. Welch Medical Library, and I, spent two years (1951-53), refining MeSH, term by term. The category system still used in MeSH was developed by Helen Field, Williamina Himwich, Dr. Larkey and I. So, if I am at times critical of the MeSH system, I think I earned the right to be so. I assure you that at ISI we are well aware that at times the more generic type of indexing done at NLM has its advantages. But no one has ever done a proper study of cost—effectiveness in MEDLINE or SCI. It is significant that NLM finally recognized the need to use natural language when it added the title word search capability to the system. Perhaps in Europe, if not America, a less emotional approach will recognize that most logically MEDLINE and SCISEARCH ought to be combined.
My association with MEDLINE was not limited to MeSH analysis. After I left Johns Hopkins I went to Columbia University School of Library Service because I knew my credentials would never be complete without a library degree. About a year later I was retained as a consultant to Seymour Taine, editor of Index Medicus. Frank Rogers was then NLM director. He had graduated from library school the year before I had. As a result of this consultancy the Index Mechanization report appeared and it formed the basis of the ultimate MEDLARS system.
Some of you are aware that I am here in Oslo to conduct a seminar at the Norwegian School of Library Science on the relationship between citation indexing and classification. I will also be discussing ISI's research and development of an on-line system that goes beyond those on-line systems available. None of the existing on-line systems is satisfactory because they do not fully utilize the special capabilities of citation linkages. Nevertheless, some of these capabilities are apparent when you use SCISEARCH with the SDC Orbit system. Since the bibliography or footnotes of each paper is contained in the basic ISI tape record you can display the bibliography of any article title retrieved by SCI, and most articles retrieved by HEDLINE. With this information one can do what we call cycling. However, this is only the beginning.
In the new ISI software system we store, for each document, not only its bibliography, but also its own citation index. Perhaps it will help you visualize the topology of this special kind of connectivity table if you imagine a bicycle wheel. Half of the spokes go up to about 15 cited papers. The lower half are the links to the papers which subsequently cite that paper. By studying the simple statistical properties of citations eventually we know that the average paper will be cited something less than the average numbers of references in the average paper for that field.
Let me return once again to my futuristic Amsterdam talk, which was entitled Citation indexing and historio-bibliography and sociology of science. I predicted at that time that we would enter an era when historiography and bibliography would merge. The future historian will go to a medical library or more likely use his own terminal to call out a name, a paper or a word corresponding to a particular event in the history of medicine or science. The ISI system will display on a television console a historiography, which shows the macro or microstructure of the scientific subject involved. Such a system requires a very large read-only memory. Therefore, the imminent advent of significantly cheaper computer memories will accelerate the availability of such systems. The SCI is a vastly larger file than most other existing bibliographic systems. To fully utilize its potential requires much larger memory. The reduction in the cost of memory will have a very significant impact on the cost of accessing citation files. This will in part be true for systems like MEDLINE but the basic software design will not necessarily alter their greater demand for central processing unit as distinct from memory. However, if the cost of memories is going to drop for large computers it is also going to drop for so-called mini—computers. Imagine now that you have at your disposal your own mini—computer with a memory sufficiently large to store an entire year of the SCI or the Index Medicus or both. How important will the MEDLINE network be under that set of circumstances? Why would you dial up MEDLINE or SCISEARCH if you could have instantaneous access to your own not-so-mini computer? The ability to update such files is a non-trivial problem. But now technology will enable us to find more efficient ways of distributing and reproducing tapes or discs for this purpose. In the future, instead of receiving your monthly Index Medicus you might receive an updated magnetic tape disc.
Minicomputers could solve the significant queuing problem in on—line systems. -As traffic increases queuing problems -increase too, and more central computer facilities are needed. Indeed it may be necessary to duplicate the central computer facilities for each batch of 50 to 100 major users.
Mini-computers controlled locally and immediately accessible may, in the future, be used in combination with on—line files containing only the most current information. Thus your own minicomputer might search a ten—year SCI file, but you would also augment a search via an on-line terminal to cover the last few months. This is equivalent to having Current Contents on-line if you are not inclined to search through the weekly indexes. It is easy to speculate where all this is leading, but printed indexes, like scientific journals, have had an amazing stability. Many disenchanted on-line users depend even more on printed indexes. Furthermore, we believe there is a constantly expanding community of users. However, if librarians should decide to discontinue subscriptions to those printed indexes in favor of more expensive on-line facilities this will inevitably raise the price of the computerized services even more. Someone has to pay the cost of inputting the data in the first place. The ultimate extension of the on-line systems would be the elimination of printed indexes and the total input cost would have to be borne by the computer facility. This is why tape subscribers and on-line users of SCI get an amazing bargain. For $ 20.000 per year, and even less for retrospective files a rational system tape user is paying less than 1% of costs. Printed indexes and abstracts still maintain a significant degree of popularity, partly because we do not assign a cost to the time of students and other heavy users of these tools. But we should not forget that printed indexed provide a display capacity that the existing on—line systems do not have. When our new system employs large video screens we may be able to display entire pages of data — just as one might do with a sophisticated microform system.
As my concluding contribution I will remind you that in my Amsterdam paper I first published the list of the 50 most cited authors of 1967. To gain admission to that select group an author required more than 600 citations in that year. Five years later, only about half of these names would have appeared on a list of most cited authors for 1972. It is interesting to note that the 1967 list was published in Nature 1970 and within a few years more than 6 of the names listed received a Nobel Prize and of course a large number had already received Nobel Prizes. In a current report in Science, the same list was republished so I thought I would tell you something more current. I have here a list of 150 scientists ranked by their total citations from 1961-1972. The remarkable stability of citations to such eminent scientists is illustrated by the fact that about 10 of the original 1967 names do not appear on the list for the 12-year total.
The following names appear in the top 50 that were not in the original list.
1. H. Selye, 2. E.J. Corey, 3. C.H. Fiske, 4. S. Moore,Given enough time I could show you that there is an interesting correlation between the time when a Nobel prize is received and the peak of citation activity for a particular scientist. E.g. a recent Nobelist C. deDuve peaked in 1968 but it will be interesting to see what effect if any the award has on his future record. If previous experience means anything it will probably be slight. The work of G. Natta peaked in 1965. Quite different from this pattern is the work of R.B. Woodward whose work has not yet peaked. This is also true for E.J. Corey whose work was cited over 1,000 times in 1972.5. 0. Warburg, 6. R. Kuhn, 7. E.A. Kabat, 8. S. Siegel
9. S. Chandraeekhar, 10. H.A. Krebs, 11. B.B. Brodie,
12. G. Gomori.
Amongst the next group worth noting are A.L. Hodgkin, A. Carisson, W.C. Schneider, G.A. Bray, L.F. Chen, J.O.Hirschfelder, and D.D. Sabatini.
This exercise has reminded me that a new list should be published soon taking into account data up to and including 1974. Maybe you will invite me back in another five years to see how well we have forecasted this time.
References
Garfield, E., Citation indexing, historio-bibliography, and the sociology of science. In: Proc. of the 3rd Internat. congress of medical librarianship, Amsterdam 5-9 May 1969. Ed by K.E. Davis & W.D. Sweeney, pp 187—204. Excerpta Medica, Amsterdam 1970. (Reprints available).Garfield, E., Citation indexing for studying science. Nature 227:669-71, 1970. (Reprints available).
The National library of medicine index mechanization project. Bull. Med. Libr. Assoc. 49: Nr 1, P.2, 1961, 96 pp.