populations that scarcely need vaccina- National Foundation for Infantile Paralysis. 17. D. M. Horstmann et al., J. Exptl. Med. 86, 7. J. F. Enders, T. H. Weller, F. C. Robbins, 309 (1947). tion, because of a low incidence of pa- 109, 84, (1949). 18. D. Bodian, Am. J. Hyg. 60, 358 (1954). ralysis, already have an abundance of 8. J. S. Youngner, Proc. Soc. Exptl. Biol. anid 19. H. K. Faber and L. Dong, J. Exptl. Med. Med. 85, 202 (1954). 101, 383 (1955). poliomyelitis strains in the community 9. J. E. Salk, J. S. Youngner, E. N. Ward, 20. D. Bodian, Federation Proc. 13, 685 (1954). that are actively immunizing children Am. J. Hyg 60, 214 (1954). 21. Pediatrics, 15, 107 (1955). soon after are born. 10. D. Bodian, ibid. 60, 339 (1954). 22. Am. J. Hyg. 56, 78 (1952). they 11. G. Shwartzman et al., Conference on Biology 23. W. McD. Hammon, L. L. Coriell, P. F. of Poliomyelitis, N.Y. Acad. Sci. 20 Jan. 1955. Wehrle, J. Am. Med. Assoc. 151, 1272 (1953). References and Notes 12. D. Bodian and H. A. Howe, J. Exptl. Med. 24. Summary Report: Evaluation of 1954 Field 1. D. Bodian, Am. J. Hyg. 55, 414 (1952). 85, 231 (1947). Trial of Poliomyelitis Vaccine. Vaccine Evalu- 2. D. M. Horstmann, Proc. Soc. Exptl. BRiol. 13. A. B. Sabin and R. Ward, J. Exptl. Med. ation Center, Univ. of Mich., Ann Arbor, Med. 79, 417 (1952). 73, 771 (1941). 12 Apr. 1955. 3. , R. W. McCollum, A. D. Mascola, J. 14. J. F. Kessel et al., J. Exptl. Med. 74, 601 25. D. Bodian, The Dynamics of Virus and Exptl. Med. 99, 355 (1954). (1941). Rickettsial Infections, (Blakiston, New York, 4. D. Bedian and R. S. Paffenbarger, Jr., Amiz. 15. H. A. Wenner and E. F. Rabe, Am. J. Med. 1954). J. Hyg. 60, 83 (1954). Sci. 222, 292 (1951). 26. H. A. Howe, D. Bodian, I. M. Morgan, Am. 5. D. Bodian, Ibid. 57, 81 (1953). 16. A. B. Sabin and R. Ward, cited in J. Mt. J. Hyg. 51, 85, (1950). 6. This work was aided by a grant from the Sinai Hosp. 11, 185 (1944). 27. H. A. Howe, Am. J. Hyg. 60, 371 (1954).

case. Classified indexes are also depend- ent upon a subject analysis of individual articles and, at best, offer us better con- sistency of indexing rather than greater specificity or multiplicity in the subject Citation Indexes for Science approach. Similarly, terminology is im- portant, but even an ideal standardiza- A New Dimension in Documentation tion of terminology and nomenclature will not solve the problem of subject through Association of Ideas analysis. What seems to be needed, then, in ad- dition to better and more comprehensive Eugene Garfield indexes, alphabetical and classified, are new types of bibliographic tools that can help to span the gap between the subject on October 2, 2007 approach of those who create documents "The uncritical citation of disputed approach to subject control of the litera- -that is, authors-and the subject ap- data by a writer, whether it be deliberate ture of science. By virtue of its different proach of the scientist who seeks infor- or not, is a serious matter. Of course, construction, it tends to bring together mation. knowingly propagandizing unsubstanti- material that would never be collated by Since 1873 the legal profession has ated claims is particularly abhorrent, the usual subject indexing. It is best de- been provided with an invaluable re- but just as many naive students may be scribed as an association-of-ideas index, search tool known as Shepard's Citations, swayed by unfounded assertions pre- and it gives the reader as much leeway published by Shepard's Citations, Inc.,

sented by a writer who is unaware of the as he requires. Suggestiveness through Colorado Springs, Colo. (2). A citation www.sciencemag.org criticisms. Buried in scholarly journals, association-of-ideas is offered by conven- index is published for court cases in the critical notes are increasingly likely to tional subject indexes but only within the 48 states as well as for cases in Federal be overlooked with the passage of time, limits of a particular subject heading. courts. Briefly, the Shepard citation sys- while the studies to which they pertain, If one considers the book as the macro tem is a listing of individual American having been reported more widely, are unit of thought and the periodical article court cases, each case being followed by apt to be rediscovered." (I) the micro unit of thought, then the cita- a complete history, written in a simple tion index in some respects deals in the code. Under each case is given a record

In this paper I propose a bibliographic submicro or molecular unit of thought. of the publications that have referred to Downloaded from system for science literature that can It is here that most indexes are inade- the case, the other court decisions that eliminate the uncritical citation of frau- quate, because the scientist is quite often have affected the case, and any other dulent, incomplete, or obsolete data by concerned with a particular idea rather references that may be of value to the making it possible for the conscientious than with a complete concept. "Thought" lawyer. This type of listing is particu- scholar to be aware of criticisms of indexes can be extremely useful if they larly important to the lawyer, because, earlier papers. It is too much to expect a are properly conceived and developed. in law, much is based on precedent. research worker to spend an inordinate In the literature-searching process, in- Citation indexes depend on a simple amount of time searching for the biblio- dexes play only a small, although signifi- system of coding entries, one that re- graphic descendants of antecedent pa- cant, part. Thosewho seek comprehensive quires minimum space and facilitates the pers. It would not be excessive to de- indexes to the literature of science fail to gathering together of a great volume of mand that the thorough scholar check point out that such indexes; although they material. However, a code is not abso- all papers that have cited or criticized may be desirable, will provide only a lutely necessary if one chooses to compile such papers, if they could be located better starting point than the one pro- a systematic listing of individual cases or quickly. The makes this vided in the selective indexes at present reports, with a complete bibliographic check practicable. Even if there were no available. One of the basic difficulties is history of each of them. Thus, it would other use for a citation index than that to build subject indexes that can antici- be possible to list all pertinent references of minimizing the citation of poor data, pate the infinite number of possible ap- under each case with sufficient com- the index would be well worth the effort proaches the scientist may require. Pro- required to compile it. ponents of classified indexes may suggest is to this Mr. Garfield is a documentation consultant with This paper considers the possible util- that classification the solution offices at 1530 Spring Garden St., ity of a citation index that offers a new problem, but this is by no means the 1, Pa. 108 SCIENCE, VOL. 122 pleteness to give the index more of the would be listed other code numbers rep- therefore potentially from 50 to 100 appearance of a bibliography. However, resenting articles that had referred to the times as great as it is for Shepard's Cita- this would result in an extremely bulky article in question, together with an in- tions. volume. dication of whether the citing source However, not all of these 50,000 pub- There are analogies in bibliographic was an original article, review, abstract, lications are being covered in our present operations. For example, in cataloging review article, patent, or translation, and indexing activities, and yet this has not books for booksellers' or library catalogs, so forth. In effect, the system would pro- prevented us from continuing indexes of an attempt is made to find references to vide a complete listing, for the publica- standard type or from starting new ones. each book in one or more authoritative tions covered, of all the original articles Lack of complete coverage is not neces- bibliographic sources, such as the cata- that had referred to the article in ques- sarily an argument against a citation in- logs of the British Museum (BM), Biblio- tion. This would clearly be particularly dex. It is in fact an argument in its favor. theque Nationale (BN), or the Library useful in historical research, when one is Coverage could perhaps be limited to the of Congress (LC). The "authority" card trying to evaluate the significance of a list of periodicals covered by one of the used in cataloging sometimes looks like particular work and its impact on the leading indexing services. This approach a Shepard entry. literature and thinking of the period. would, of course, have an immediate dis- Another example is a book-review di- Such an "" may be much advantage. Such a subject selection gest, in which one finds for each book more indicative than an absolute count would mean that less directly related sub- title a series of references and selections of the number of a scientist's publica- jects of interest would be excluded, and from published reviews, critical and tions, which was used by Lehman (3) these are the publications that the indi- otherwise. Certain indexing publications and Dennis (4). The "impact factor" is vidual is least likely to cover in his own periorm a similar function. similar to the quantitative measure ob- research. It would be necessary to con- Some time ago I became concerned tained by Gross (5) in evaluating the sider all the pros and cons in a selective with the problem of developing a cita- relative importance of scientific journals, approach and then to determine the pos- tion code for science. This was necessary a method later criticized by Brodman sible utility of such a tool. For example, for the efficient manipulation by mechan- (6) but used again by Fussler (7). would a citation index to the 1500 peri- ical devices of entries to scientific indexes. Other advantages would also obtain. odicals covered by the Current List of In the course of this research I developed In a way such listings would provide each Medical Literature be of real value, or, a very simple system for identifying an scientist with an individual clipping serv- similarly, a citation index to the 5000 individual scientific article that had ap- ice. By referring to the listings for his periodicals covered by Chemical Ab- peared in the periodical press. The re- article, an author could readily deter- stracts? The Current List would, in fact, sulting numerical code consisted of two mine which other scientists were making offer a good starting point, since it al- parts. The first part was a serial number, reference to his work, thus increasing ready provides a unique code for the on October 2, 2007 used instead of an abbreviation, to iden- communication possibilities between sci- 100,000 items indexed by it each year. tify each periodical; it was similar to the entists. It is also possible that the indi- Presumably these are the most significant serial numbers employed in the World vidual scientist thus might become aware contributions in the covered fields for the List of Scientific Periodicals, by no means of implications in his studies that he was year. If 10 is the number of references in a new idea. For example, Die Biblio- not aware of before. the average article, then about 1 million graphie der fremdsprachigen Zeitschrif- Most authors like to see how their citations would be involved. The prepa- ten Literatur has for many years used works are received. Bringing together all ration of that number annually is not un- such a system to save space. book reviews and abstracts is very im- reasonable. Shepard's has already used in its The second part of the code number portant, for it is not possible for an well over 50 million citations pub- www.sciencemag.org was also a serial number, assigned to each author to keep up with the thousands of lishing activities. article in a particular publication, start- publications in which his contribution The ultimate success of a citation in- ing with 1 and continuing throughout all might be reviewed. This applies equally dex would depend on many factors. For volumes. The code thus gives no indica- to publishers. It would not be impossible example, if each periodical would assign tion of year or volume number, a serious to include books in the citation index. In- unique code numbers to the articles pub- shortcoming. The article number is also deed, as a first suggestion, the use of lished, it would be possible for authors not unique, having been used by the Pro- Library of Congress card numbers as the to list these numbers in their bibliogra- seem save the work of cod- ceedings of the Society for Experimental identifying code for books would phies and, thus, to Downloaded from Biology and Medicine since its inception. appropriate. ing on the part of the citation index staff. These two serial numbers taken together, It is necessary next to discuss some It is unlikely that such a development it can be seen, can identify any published realistic questions concerned with the could take place in less than 5 or 10 periodical article. It soon became appar- realization of such an index. Bitner (8) years, but it is comparable to the problem ent, after such codes had been utilized on has estimated that 30,000 cases are cov- of getting publishers to include Library an experimental basis, that the use of the ered by Shepard's Citations in 1 year, of Congress card numbers in their pub- codes would facilitate the compilation of the cases and articles appearing in not lications. a citation index. (Other coding systems more than a few hundred publications. When such a large volume of data is would be equally applicable.) In 1953 about 1 million citations were to be handled, mechanical devices of A citation index to science would have added-close to 40 citations per case. high speed and versatility could be used the following main characteristics. First What is the prospect in scientific litera- to great advantage and would probably there would be a complete alphabetic ture? The last published edition of the determine success or failure. Once the listing of all periodicals covered, in addi- World List of Scientific Periodicals con- coding is done, compilation itself is quite tion to the code number for each periodi- tained more than 50,000 titles in science mechanical. This could be done by means cal. This list would be similar to the and technology. It is variously estimated of conventional filing slips; the Shepard World List, but without the library hold- that between 1 and 3 million new scien- organization itself has used them success- ings information. The main portion of tific articles are published each year. The fully for 80 years. However, it would be the citation index would list in straight Journal of the American Chemical So- facilitated by a mechanical approach numerical order the code numbers for ciety alone publishes more than 3000 per using punched cards. all the articles covered. Under each code year, including approximately 2000 origi- The utility of a citation index in any number, for example, 3001-6789, there nal articles. The order of magnitude is field must also be considered from the 15 JULY 1955 109 point of view of the transmission of In another instance the RCA article adaptation syndrome (14). A systematic. ideas. A thorough scientist cannot be was unexpectedly cited in the joumal search was then made of all papers that satisfied merely with searching the litera- Electronic Engineering in an article on were published in the Journal of Clinical ture through indexes and bibliographies information theory (12) that I was read- Erdocrinology subsequent to Selye's if he is going to establish the history of ing because of an entirely different in- paper up to 1951-a period of 5 years, an idea. He must obviously do a great terest. No subject indexer could have an- including well over 500 articles. Every deal of organized, as well as eclectic, ticipated this crossbreeding of interests. bibliography in each of the 500 articles reading. The latter is necessary because Perhaps there are many other articles was checked for a reference to Selye's it is impossible for any one person (the and books unknown to me that have article. Twenty-three articles were found indexer) to anticipate all the thought made similar references to this device. to make such reference; each of them processes of a user. Conventional subject How can they be located when the main was then checked for the character of indexes are thereby limited in their at- subject matter of the article is, on the the information provided. tempt to provide an ideal key to the lit- surface, so unrelated in ? Examination of the citation list (Table erature. The same may be said of classi- One might say that it would be pos- 1 ) shows the great variety of subject mat- fication schemes. In tracking down the sible to index articles more thoroughly ter included. One thing became quite origins of an idea, the citation index can to achieve the same results. For example, clear, even to the uninitiated-that is, the be of real help. This is well illustrated by the article on information theory, if thor- influence of Selye's article has been quite an example from my own experience. oughly indexed, might have included an pronounced. Such evidence is extremely Many years ago the Radio Corporation entry under reading devices for the blind. valuable to the historian. of America developed a reading-aid for Yet if this were done, our periodical in- It is interesting to note that, although the blind (9). This device had an elec- dexing services would clearly become all the articles cited were indexed in tronic system for converting printed let- hopelessly overloaded with material that Quarterly Cumulative Index Medicus, ters into recognizable sound patterns. is not necessary to lead us to the micro not one is to be found there under the Using the device, a blind man could scan unit-the entire article or one of its heading "Adaptation." In fact, it is sur- a printed page; in a set of headphones major sections. Although it might be said prising not to find any articles from this he could hear a series of sound patterns, that no scientist interested in the greater journal under this subject heading. each letter having its own recognizable comprehensiveness to be found in a cita- It also becomes quite obvious that sound pattern. In effect, the words were tion index would object to having such a many references to Selye's paper were spelled out, letter by letter, in code. I great mass of references in a subject in- general and contribute little or nothing was particularly interested in this device dex, this is impracticable. It would re- because I had been independently work- quire an army of indexers to read the Table 1. Index sample based on article on October 2, 2007 ing on a device that would copy print, articles and identify the exact subject by Hans Selye, "General adaptation syn- letter by letter, and reproduce it for matter of every paragraph or sentence. drome" [J. Clin. Endocrinol. 6, 117 bibliographic and other purposes. The Yet this would be necessary. To illustrate, (1946) ]. The code number for this journal two devices had something in common in it is only in the very last paragraph of in the World List is 11,123a; the article that they both employed scanning de- the article on information theory that one number is arbitrarily taken as 687; and vices. I then wanted to learn whether would find a reference to reading devices the code number for the article is 11123a- anyone had ever suggested that the RCA for the blind. 687. The 23 articles that cited Selye's reading-aid could be used for this pur- Were an army of indexers article are listed, followed by A hypo- available, it thetical citation index entry for Selye's pose. It will be apparent that if anyone is still doubtful that the proper subject -ttticle. R, review article; A,.-abstract; 0, had known of the RCA device and had indexing could be made. Over the years original article. www.sciencemag.org thought of adapting it for copying pur- changes in terminology take place, that poses, a reference to the article might vitiate the usefulness of a standard sub- 1. Williams, R. H.: Thyroid & Adrenal have been made. This reference could ject index. To a certain extent, this is Interrelations, 7: 52-57 (1947). easily have been included in an article overcome through the citation approach, 2. Venning, E. H.: Glycogenic Corti- or patent that was not at all related to for the author who has made reference coids, 7: 79-101 (1947). the problem of reading devices. A cita- to a paper 40 or 50 years old has inter- 3. Forbes, et al.: 17-Ketosteroids in tion index would have given me just the Trauma and Disease, 7: 264-288 preted terminology for us. By using (1947). what I was after. Nothing could substi- Downloaded from authors' references in compiling the cita- 4. Talbot, et al.: Excretion of 11-Oxy- tute for extensive reading, but a great tion index, we are in reality utilizing an corticosteroids, 7: 331-350 (1947). deal of time could have been saved bv army of indexers, for every time an 5. Castillo, E. B. del, et al.: Syndrome bringing the appropriate works to my at- author makes a reference he is in effect of Rudimentary Ovaries, 7: 385-422 tention. indexing that work from his point of (1947). In the course of my reading I did find view. This is especially true of review 6. Forsham, P. H., et al.: Pituitary a few references to this device, one in a articles where each statement, with the Adrenocorticotropin, 8: 15-66 book (10), and several others in peri- following reference, resembles an index (1948). odical articles, one of which was a Ger- entry, is 7. Pincus, G., et al.: Rhythm in Biped superimposed upon which the Excretion, 8: 221-226 (1948). man article on the mechanization of function of critical appraisal and inter- 8. LeCompte, P. M.: Width of Adrenal philological analyses and concordance pretation. To the indexer this has its ad- Cortex in Lymphatic Leukemia, 9: building. The latter article (11) did not vantages as well as its disadvantages 158-162 (1949). discuss my own special interest in copy- (13). 9. Wolfson, W. Q.: 17-Ketosteroids in ing devices, but it did show the similarity To determine in a practical way what Gout, 9: 497-513 (1949). between the author's and my own think- the citation index could offer, it was de- 10. Stein, H. J., et al.: Hormonal Re- ing from the point of view of letter-rec- cided to track down the citations made sponse to Heat and Cold, 9: 529-547 ognition devices, which is what the RCA in one journal to a ar- (1949). single significant 11. Davis, M. E.: Eosinophils in Preg- device attempts to be. In other words, ticle, in order to compile a sample entry nancy and Labor, 9: 714-724 both of us were interested in this device for the citation index. At the suggestion (1949). as a letter-recognition device for the of Erich Meyerhoff, I selected Hans 12. Conn, J. W.: Na and Cl of Sweat as analysis of text. Selye's famous article on the general Cortical Index, 10: 12-23 (1950). 110 SCIENCE, VOL. 122 13. Recant, L., et al.: Effect of Epine- sented indicates that the citation index the citing article and the code for the phrine on Eosinophils, 10: 187-229 offers interesting possibilities for another book (its LC card number). (1950). approach to bibliographic control. After all the articles had been coded, 14. McArthur, J. W.. et al.: Urinary The -next step in compiling the index it would next be necessary to sort the Excretion of Corticosteroids in Dia- betic Acidosis, 10: 307-312 (1950). for the Selye article would be to seek out cards by the code numbers for the items 15. Bors, E.: Fertility in Paraplegic additional references to it in more periph- cited. This would yield a group of cards Males, 10: 381-398 (1950). eral journals, but obviously the farther for each cited article. These would then 16. Grossman, S., et al.: Idiopathic Lac- away you get from the immediate subject be sorted by code numbers for the citing tation following thoracoplasty, 10: area of the main article, the fewer the articles. This completes the coding and 729-734 (1950). references to it you will locate. Yet these sorting. The next step would be prepara- 17. Cooper. J. S., et al.: Metabolic Con- may well be the most useful references of tion for the printer. sequences of Spinal Cord Injury, 10: all, for the cross-fertilization of subject From this description it will be appar- 858-870 (1950). most prob- volume of ma- 18. Hioco. D.: Adrenal Metabolites in fields is one of our important ent that, although a great Bronchial Asthma, 10: 1570-1578 lems in science literature. terial is to be covered, relatively unskilled (1950). It will be well to close with a brief persons can perform the necessary coding 19. Jailer, J. W.: Pituitary-Adrenal Sys- description of how the citation index and filing. Professional supervision would tem in Infants, 11: 186-192 (1951). might be compiled. The first step would still be required, because certain decisions 20. Deane, H. W.: The Adrenals in Ex- be the selection of the particular group require skilled judgment, for example, perimental Hypertension, 11: 193- of periodicals to be covered; next, the when ibid. or loc. cit. must be carefully 208 (1951). period to be covered, say, only that since interpreted. Footnotes tend to ma-ke cod- 21. Hioco, D., et al.: Epinephrine and 1900. ing somewhat cumbersome. The code I ACTH in Bronchial Asthma, 11: 395-407 (1951). The problem actually has two facets: have described is merely an example used 22. Schaffenberg, C. A., et al.: p-Hy- the selection of periodicals to be covered to illustrate the method in principle. If droxypropiophenone (PHP) and in order to obtain citations, and the selec- the system were adopted, then in the other so-called pituitary inhibitors, tion of those articles for which we want future every author ought to be required 11: 1215-1223 (1951). a citation record. For example, all ar- to include the serial number of each item 23. Talbot, N. B., et al.: Urinary Water- ticles in journals in the Current List of he referred to, so as to facilitate not only Soluble Corticosteroids, 11: 1223- Medical Literature that have remained the compilation of citation indexes but 1236 (1951). in continuous publication since 1900 also other operations such as requests for might be coded, in which case the Jour- reprints (15, 16). Citation Index Entry nal of Clinical Endocrinology would not In a certain sense a citation index is 11123s-687 be included. However, we might include not very different from a compendium on October 2, 2007 464-9789(R) as citation sources all journals covered like Beilstein, which gives a rather com- 869-3366 (R) by the Current List. Thus, the bibliogra- plete record of a compound, compiled 1105-9876(A) phies appearing in articles in the Journal by a similar method. A citation index 1123-4432(R) of Clinical Endocrinology would supply for the literature of chemistry would un- al l,123-0752(0) references to the basic group of articles. doubtedly make the preparation of such -0779(0) Each coder would be assigned a group works as Beilstein much easier than it is -7264(0) -7331(0) of articles in a particular journal. The at present. The new bibliographic tool, -7385 (0) first step would be to number each article like others that already exist, is just a

-0866(0) in the journal in ascending order, by starting point in literature research. It www.sciencemag.org -8221(0) utilizing a complete table of contents of will help in many ways, but one should -9158(0) that journal from its inception. not expect it to solve all our problems. -9497 (0) Once a code number has been assigned and Notes -9529(0) to each article, the proper codes may References to each This 1. P. Thomasson and J. C. Stanley, Science 121, then be assigned periodical. 610 (1955). Thomasson and Stanley were com- might be the number given in the World menting on C. Zirkle's discussion of the use to the readers' enlightenment, since exact List, with new numbers for any periodi- of fraudulent data [Science 120, 189 (1954)]. 2. W. C. Adair, Am. Documentation 6, 31 page references are not provided. In sev- cals not to be found there. (1955). Downloaded from eral cases the Selye article is even cited Actual coding starts with the first ar- 3. H. C. Lehman, Sci. Monthly 78, 321 (1954). 4. W. Dennis, ibid. 79, 180 (1954). but not referred to in the text. Selye's in- ticle in a particular periodical. The coder 5. P. L. K. Gross and E. M. Gross, Science 66, fluence on all of these authors is quite prepares a 3- by 5-in. card for each cita- 385 (1927). apparent. In particular instances the ci- made the article. Each card 6. E. Brodman, Med. Library Assoc. Bull. 32, tion in 479 (1944). tations are of value in locating confirma- should give (i) the code number for the 7. H. H. Fussler, Library Quart. 19, 19 (1949). tory evidence of some of Selye's claims. citing article, (ii) the code number for 8. H. Bitner, personal communication, April 1954. Thus, in the case of a highly significant the article cited, and (iii) a classification 9. V. K. Zworykin and L. E. Flory, Proc. Am. article, the citation index has a quanti- of the citing article as an original con- Phil. Soc. 91, 139 (1947). tative value, for it may help the historian 10. R. R. Shaw, Machines and the Bibliographi- tribution, review article, abstract, and so cal Problems of the Twentieth Century (Univ. to measure the influence of the article- forth. of Illinois Press, Urbana, 1951) p. 19. (Re- that is, its "impact factor." With regard Many references will be excluded by printed from Bibliography in an Age of Science (Univ. of Illinois Press, Urbana, 1951). to a less significant work, one would sus- the limits of coverage set up. Thus all 11. R. Busa, Nachr. Dokumentation 3, 14 (1952). pect that the bibliographic advantages references to articles not in the prescribed 12. A. M. Andrew, Electronic Eng. 25, 471 (1953). 13. E. Garfield, "Review literature as a source of might be increased, because the scientist list of journals would be excluded. critical entries for scientific indexes, unpub- or librarian would be provided with ref- All books would be excluded unless lished paper, December 1952. erences not to be found in conventional in which case the 14. H. Selye, J. Clin. Endocrinol. 6, 117 (1946). otherwise specified, 15. E. Garfield, Science 120, 1039 (1954). indexes. The preliminary evidence pre- reference card would carry the code for 16. J. A. Behnke, Science 120, 1055 (1954).

15 JULY 1955 III