II EUGENE GARFIELD

I INSTITUTE FOR SCIENTIFIC 4NF0RMAT10N@3501 MARKET ST PHILADELPHIA PA 19104

I How 1S1 Se4ects Journals for Coverage:Quantitative and Qualitative Considerations

INumber 22 May 28, 1990

ISI” takes into account many factors when makingjournal coverage decisions. Three types ofinformationare discussedand illustrated:citationdata, journat standards,and expertjudgment.Theessay is basedon a lecturepresentedat the Symposiumon ScienceJournalEvacuation,the NationalScienceCouncilof Taiwan, Scienceand TechnologyInformationCenter, Taipei, on March21 and22, 1990.

I recently had the pleasure of visiting Tai-pei, Taiwan, as a guest of the National Sci-ence Council. Our very gracious host, Tao-Hsing Ma, director of the Science and Tech-nology Information Center, invited me toaddress two groups of journal editors, in-formation specialists, and librarians and ar-ranged au informative tour of the nation’sleading research institutions. At the conclu-sion of our tour, we were charmed by Pat-zen Wu, ex=utive secretary of the Coun-cil’s Science and Technology advisorygroup.

The subject of my talk was one that fre-quently comes up at ISI@—jourrudselection.We receive a steady stream of calls and let-ters asking how 1S1decides what journalsare covered in various editions of CurrentContents” (C@’), the Science CitationIndex” (SCF ), and other services. Editorsin particular are the most inquisitive-per-haps because they believe that coverage by1S1and other international secondary infor-mation services attracts greater notice totheir journals.

1S1’sjournal selection considerations haveoften bem dkcussed in CC. 1.2The Taiwanpresentation is the most recent update on thisperennially popular topic, and I’ve adaptedit in this essay for the benefit of interestedreaders. This overview will, to some extent,generalize and simplify what is actually apainstaking process of research, analysis,

and consultation by a professional staff ofsubject specialists.

Three Broad Criteria

Basically, three types of information aretaken into account when evaluating journalsfor coverage, ranging from the quantitativeto the qurditative: citation data, journal sEm-dards, and expert judgment. 1S1is uniqueamong the world’s leading information ser-vices for indexing all cited references as weJlas complete bibliographic information on allitems published in a journal-not just articlesand reviews but letters, editorials, errata andretractions, book reviews, and other items.1S1’s consolidated database for tbe SCI,social Sciences Citation Indexe, and Arts& Humanities Citation Indexm now includesabout 18,CKX),000source items publishedfrom 1945 until today in thousands of jour-nrds, and more than 217,000,000 citedreferences.

These citation data are a source of quan-titative indicators that an be used to evaluateexisting journals with established track rec-ords, as is discussed later. But selection ofnew journals often relies on other, morequalitative considerations. Journal standardsare an example. A journal’s ability to meetits declared schedule and frequency is per-haps the most basic expectation. Standardscan also include editorial requirements for

185

abstracts, titles, and references set by pro-fessional associations of publishers and txh-tors. 3,4 Peer review of submissions, edito-rial board membership, and the reputationof the publisher or sponsoring society areother indicators of journal quality.

Finally, jourrud selection also relies on thesubjective judgment of experts in a partictdarfield-subscribers, editors and publishers,and 1S1’s many editorial advisory boardmembers and staff specialists.

Why Be Sdeetive at All?

1S1’sobjective as a secondary informationservice is to provide comprehensive cover-age of the world’s most important journalsfor our subscribers’ current awareness andinformation retrieval needs. But comprehens-ive does not necessarily mean all-inclusive.The success of CC is partly due to the factthat most of its readers already find morethan they need in it. In fact, many readersfeel 1S1should do whatever possible to con-tain the proliferation of journals-not to en-courage it.

Of course, this does not mean we urgepublishers to limit the number of journalsin the marketplace. As I’ve stated before,there is nothing we can or should do to pre-vent the legitimate’ ‘proliferation” or’ ‘twig-ging” of joumrds. 5.6 Curtis G. Benjamin,former president, McGraw-Hill Book Com-pany, New York, coined the word “twig-ging” to describe the relentless specializa-tion of scientific knowledge and, as a result,the burgeoning number of and markets forscientific publications.7 But, as many criticshave pointed out, the problem is that pub-lishers sometimes launch journals prema-turely before an adequate market develops.

The fact is, no matter how many joumrdsare in the market, only a small fraction ac-count for most of the articles that are pub-lished and cited in a given year. This is il-lustrated in Figure 1, which shows the dis-tribution ofjournal articles and citations inthe 1988 SCl database.

In a recent New Scientist ‘‘Forurn” fea-ture on scientific journals, British researchchemist Susan Aldridge discussed the“80/20 rule” for time management-that is,

that 80 percent of results come from doing20 percent of the tasks.g While the rule istypically intended as a useful reminder tobusiness executives who face mounds of pa-perwork, Aldridge suggested it applies aswelI to researchers trying to gain controlover the flood of literature. She speculatedthat the SC1database would provide “a tineexample of the 80/20 rule. ”

The graph shows that Aldridge’s hunchis indeed true for citations but not for arti-cles. The solid line indicates that 900 (21percent) of the 4,400 journals indexed in the1988 SC] rweived 83 percent of the8,000,000 citations processed for the Jour-nul Ci~”on Reportsm (JCR@) that year. Butthe dotted line shows that it took 2,0Wjour-nais (46 percent) to publish 86 percent ofthe 435,000 original research or review ar-ticles and technical notes included in the1988 JCR.

The graph, of course, is merely anotherillustration of the well-known Bradford andZipf distributions and various other statis-tical patterns. The Literature on these sub-jects is too voluminous to cite here but wasreviewed previously. g Recently, a ‘‘Biblio-metrics Toolbox” software package wascreated that can generate such graphs fromthe “SCIand other data. 10

These data illustrate the fact that 1S1couldconceivably limit itself to the top 500 jour-nrds and still provide comprehensive cov-erage of the most important publications.Thus, 1S1’scoverage of over 7,000 journalsgoes well beyond what the average subscrber needs or perhaps even wants to knowabout the research literature. In the past Ihave referred to this as Gafileld’s law ofconcentration. ]1

Citation Data for Journal Evaluation

Several types of journal citation data cOv-ering a particular year or period of time canbe derived from the 1S1database. Librarians,information researchers, editors, and otherswho regularly use 1S1’sJCR for evaluationand analysis rdready are farnihar with mostof these: total citations, impact, what a par-ticular journal cites most frequently, whatjournals cite it, and so on. 12

186

Figure 1: Distribution of publlahed item and citations among science Jonrnak, 1988 SCF

It should always be stressed that citationdata must be carefully interpreted-and theirlimitations clearly underskmd-when theyare used for evrthtating anything. This sub-ject has been extensively and repeatedly dis-cussed in CC and needs only brief mentionhere. IS-ISFor example, the number of au-thors and journals varies greatly betweenand withii disciplines, as do their citationlevels and rates. Smaller fields like botanyor mathematics do not generate as many ar-ticles or citations as, say, biotechnology orgenetics. Also, in certain fields it may take10 or more years for an article to attract ameaningful number of citations, while inother research areas citations can typicallypeak afier only a few years. These pointsshould be kept in mind as we consider thefollowing examples.

Journal Rankingsby Total Citations

The list of the 25 most-cited journals inthe 1988 SC1 (Table 1) probably agreesclosely with most readers’ mental list of themost important scientific journals. Hardlyanyone would dispute the inclusion of theJournal of Biological Chemistry, Ntaure, theProceedings of the National Academy of Sci-ences of the USA, the Journal of the Amen”-can Chemical Society, Science, or any

others. The same basic group of journalstends to be most cited year after year. A fewmay gradually decline or be replaced by suc-cessful newcomers like Cell as editors andaudiences change. But most successftd jour-nals sutvive and prosper for decades.

Table 1: The 25 mnat-cited JmmrmJa, 1988SCFJCJF

Jourmd 1988 Citations

J. Biol. Chem. 172,726Nature 167,897Pmc, Nat. Acad. Sci. USA 167,464J. Amer. Chem. Sot. 122,492Science 106,393J. Chem. Phys. 84,098Phys. Rev. Lett. 73,497N, EngJ. J. Med. 69,103Phys. Rev. B—Condensed Matter 68,179Lancet 67,723Biochim. Biophys, Acts 65,593J. Jmrmmol. 64,358Cell 63,776Biochemistry-USA 52,123Amer. J. Physiol. 50,108Brain Res. 49,167J. Clin. Jnvest. 46,243Astrophys. J. 44,733Biochem. J. 43,589J. Cell BioL 41,906Cancer Res. 41,731Biochem. Biophys. Rcs. Comrrmm 41,353J. J?xP. Med. 37,079J. Org. Chem. 35,9443J. Phys. Chem. 34,856_.. ..........-. -...

187

Not surprisingly, the list is dominated bylarger jou~s ad the big life-sciences spe--cialties. Fourteen also ranked among the top25 by number of articles published. To com-pensate for this putative size advantage, ajournal’s impact-the average number of ci-tations per article-can be considered ratherthan its total citations.

Journal Rankings by Impact

In Table 2 impact is calculated as follows:the number of articles published by a jour-nal in 1986 and 1987 is divided into thenumber of citations they received in 1988.For example, the ,4rmud Rm”ew of Bio-chemistry published 67 articles in 1986 and1987. They received a total of 3,237 cita-tions from IS1-covered journals in 1988.Thus, its impact factor is 48.3.

The list is obviously dominated by reviewjournals, which tend to publish fewer con-tributions than original research journals,but these are cited much more frequently.Table 3 presents artother impact ranking,showing only journals that published at leastlCJOarticles, which effedvely excludes mostreview journals.

Eighteen life-sciences journals are listed,compared to two each for chemistry andphysics, Again, while impact compensatessomewhat for the size of a journal or litera-ture, it tends to favor research areas thatmore heavily cite recent research publishedin the last two years. As we found severrdyears ago, the average number of referencescited per article is perhaps the most signifi-cant contributing factor. This may or maynot be a reflection of the field size. lb

Calculating the Impact of JournalsNot Covered by 1S1

There seems to be a widespread miscon-ception that the JCR gives citation data on-ly for those journals indexed in the 1S1da-tabases. In fact, any publication that is citedby ISI-covered journals will be included inthe JCR’S ranking of journals by total cita-tions received and in the cited journrd list-ing. While data on the number and type of

Tabk 2: The 25 highsQt hnpsectjourds, 1988SCPJCW

Joti

Amm. Rev.B&hem,

Pharmacol, Rev.Amu. Rev,

IlmmmOl.Amu. Rev. Cell

BioLAdvan. Chem.

Phys.CellN, Engl. J, Med.Adwm. ~ChC

Nucl. Pmt.Amu. Rev.

Neurosci.Adwm. Prot.

Chem.ScienceAdvan. Immunol.Microbiol. Rev.NatureAmIu. Rev.

Genet.Rev, Mod, Phys.LancetAmm. Rev. Pkmt

Physiol.Elwtrnemd.

Chem.Physiol, Rev.Advm. NUCL

Phys.J. Exp. Med.EMBO J,hmnunol. TmlayEndmrine Rev,

1%361%37

48.3

29.425.4

24.2

24.0

23.921.118.2

17.0

16.5

16.516.416.315.815.1

15.114.513.4

12.3

12.211.9

11.810.910.710.6

Amcles

67

1749

33

11

a63716

6

36

4

1,6162049

2,37543

47955

39

4

4-I10

623I,055

16454

citation?

3,237

5001,245

799

264

20,63713,142

109

6i 1

66

26,5%328797

37,425650

71113,S28

521

49

575119

7,36911,5381,747

570

source items published is not provided fora nonsource journal, the number of timesit was cited is shown.

This makes it possible to calculate impactfactors of journals not included in 1S1’sda-hbases. A simple and effed.ive method fordoing so was described in a recent letter intheJournal of Docunretuation, by B.IL Sert,A. Karanjai, and U .M. Munshi, B~bliomet-rics Section, INSDOC, New Delhi, In-iia. i7 They manually counted the numberof source items published by the IndianIow-nal of Malariology over a two-year pe-riod. They also recorded the number ofbrnes the journrd cited itself and added theseto the number of citations from XXcOvered

188

Table 3: The 25 highest Jmpact jourrtats publishingat least 100 WtiCk.S, 1988 SCF JCR@

CellN, Engl. J. Med.ScietiNatureIancetJ. Exp. Med.EMBO J.Jmmunol. TudayChem. Rev.Pruc, Nat. Acad.

Sci. USAJ. Cell Biol,Trends Neurosci.Ann. Jntem. Med.Phys. Rep.—Rev.

Sea Phys. Lat.Phys. Rev. Len,Trends Biuchem.

*i.Mol. CeU. Biol.J. Clin. Invest,J. Immurml.Arch. Gen.

PsychiatBluudAccuunt. Chem.

Res.CirmdationJ. Mol. BIO1.J, Biol. Chem.

Impact

23.921.116.515.814.511,810.910.710.410.0

9.79.28.58.2

8.27.9

7.77.66.96.8

6.86.7

6.76.66.5

1986-1987 1988Articka citations

863 20,637716 15,142

1,616 26,5962,375 37,425

955 13,828623 7,369

1,055 11,538164 1,747100 1,040

3,%8 39,805

983 9,582186 1,702538 4,555145 1,194

3,022 24,821252 1,984

1,187 9,1701,013 7,6902,666 18,409

251 1,708

1,087 7,445124 826

994 6,636665 4,359

5,335 34,632

journals as listed in the JCR. By dividingself- and SCLcitations by source items, theywere able to calculate impact. Also, S. Mar-i;ii, Association of ~lerttific Unions of Yu-goslavia, Zagreb, and colleagues have U.SEAthe same method for some 10 years. 18

Sen et al. suggest that this method can beused by editors in Third World and Com-munist block countries to evahutte and, ifnecessary, improve the “quality” of theirjournals as well ag to bring to ISI’s atten-tion journals that may be suitable for cov-erage. Many Third World editors have al-ready asked us why their publication is notcovered by 1S1when its impact, althoughlow, is comparable to other journals we doindex. This reflects another common mis-conception-that impact factors are the soleor single most important criterion for cov-erage. In fact, journal impact is only one ofseveral quantitative and qualitative factorsdescribed in this essay that we take into ac-count.

Table 4 The top25juurmdsby five-yeaI impsct thatpublished at least 10Uarticles, 1984 SCJ@. Citationdata were taken from the 1984-1988S(X.

Journal

CellJ. Exp, Med.Pruc. Nat. Acad.

Sci, USAEMBO J.J. Clin. Invest.Arch. Gerr.

Psychiat.NUCLAcid. Res,J. BioL Chem.J, Neurosci.Circ. Res.Mol. Cell. Biol,J. hIllnUKd.Mol. Pbamracol,NaturePhys. Rev. lat.BluudJ. Mol. BioLScienceEur. J. Immunol.NeuruscienwGeneJ. Clin. Enducrinol

Metab.J. Comp. Neurol,Nuc1. Phys. BJ. Virol.

W84Jmpact Mlcles

77.444.238.5

33.731.127.2

26.725.726.725.424,624.322.221.120,920.620,419,719,319,218.618.4

18.418.217.7

427294

1,681

495487197

7232,424

293172395

1,076147

3,5611,445

444416

2,W1211272247420

377576563

1984-198$citations

33,03012,989@,777

16,68815,1305,364

19,33362,250

7,8154,3759,725

26,1763,262

75,07430,2019,1568,497

41,1074,0745,2274,5867,743

6,93910,5049,976

-s by Five-Year Impact

Recently, 1S1created a database that in-cludesjive-yertr impact factors, and the top25journals ranked by this measure are listedin Table 4. In these examples, impact is cal-culated by dividing the number of articiespublished by a joumrd in 1984 into the num-ber of citations they received from 1984through 1988. For example, Cd’s 4271984articles were cited 33,030 times from 1984to 1988, giving an impact factor of 77.4.

Basically the same set of journals comesout on top in Tables 3 and 4 whether im-pact is calculatd over two or five years.Fourteen journals appear on both lists, andboth are dominated by life-sciencesjournals.

[tem-by-ltern Impact

1S1’s database can also derive separatexmventional two-year impact factors for themrious types of’ ‘source items” published[na journal-articles, research reviews, let-

189

ters, notes, editorials, and errata and retrac-tions. For example, in 1986 we studied 40leading medical journals and compared theseso-called “item-by-item” impacts. 1g Re-view articles had the highest impact in fourof the top five journals-the Annals of In-ternal Medicine, JAMA-Jounud of theAmerican Medical Association, Lancet, andthe New England Journal of Medicine(NEW). Original research articles were theimpact leader in the Bn”tishMedical Jour-nal (BMJ) and were second in Lancet andNEYM. Interestingly, either discussions orproceedings ranked semnd in impact for theAnnals, BIUJ, and JAMA.

This information demonstrates the dNer-sity—in kind, number, and impact-of thetypes of items typically published in jour-nals. The impact of letters, editorials, pro-ceedings, and discussions argues in favor ofcomprehensive coverage of all items, as 1S1provides, and against limiting coverage toarticles and reviews only.

Tracking Journal CiMtion (Yassics”

The highest impact items ever publishedby a particular journal can also be identifiedin the 1S1database. An example is shownin Table 5, which reproduces the first pageofa241 -page printout on Cell that includesover 2,500 articles cited at least 50 timesfrom 1945 through 1988. An abbreviated

reference is given to keep the ftle withinmanageable limits.

The report ranks articles by total citationsand details year-by-year citations to showhow rapidly each item was cited, and wheth-er its impact is rising, steady, or declining.These annual citation trends cars also un-cover possible cases of delayed recognition,which was illustrated in the first essay in ourCC series on the most-cited papers ever.zoEditors have used these fdes to prepareprogress reports to boards of directors, sup-port budget and funding proposals, compareperformance against peers and competitors,and plan commemorative issues and for var-ious other reasons, including simple curi-osity and professional pride.

Internationality

The geographic representation of a jour-nal is another consideration. Unless a jour-nal of interest to only a small region of theworld is exceptional in some way, we areless likely to cover it.

1S1’sdata can be used to indicate a jour-nal’s “internationality, ” in two senses: thenationality of items it publishes and the na-tionality of the articles that cite it. Table 6shows the nationality of 1984 articles fromthe International Journal of Gzncer, definedby the address of the jirst author. That is,if a US institution wa8 listed, the article was

Table5 The4X Cell artfcfes moat cftedin the 1945-1988 SCP, first page of a 241-page printout on Urisjournalfrom the ISI” Journat Classics fde.

JmIRIat TotalJOratAsstbor Year vol. L? murk citations

BERK A J77=1 78=2386=260 87=267

O’FARJU3LLP Z77=0 78= 1686=250 87=259

REINHERZ E L80=12 81=97

MANJATIS T78=1 79= 3087=125 88=118

KELLY K83=0 84=92

GILMANA G84=17 85=136

79=5488= 254

79= 6588=216

82=182

80=120

85=214

86=262

1977 1280=99 81=122 82=156

1977 1280=124 81=170 82=181

1980 1983=252 84=222 85=203

1978 1581=174 82=153 83=188

1983 3586=267 87=314 88=260

1984 3687=364 88=272

72183=215 84=233

113383=226 84=201

82186=184 87=146

68784=174 85=163

603

577

1 1,95185=267

2 1,94085=232

3 1,42988=131

4 1,37486=128

5 1,147

6 1,051

190

.......................... ...............

credited to the US even though the authormay be a visiting researcher from the UK,France, or India, or the coauthors may bebased in other countries. The number ofsuch international collaborations is small,however, typically accounting for less than5 percent of all SCI-indexed source itemsper year.

Of the251 articles published by this jour-nal in 1984, 60, or 24 percent, were fromthe US. The UK is second with 29 articles(12 percent), followed by Japan with 22,France (20), and Italy (19). In total, 32 na-tions were represented by the first authorsof 1984 articles in the International Jour-nal of Cancer, and the top five countriesidentified above accounted for 60 percent.

Table 7 shows the nationrdity of the1984-1988 articles that cited this journal’s1984 publications at least 20 times. Again,US articles top the list with 1,091 citations,or31 percent of all citations received by theInternational Journal of Gancer. Japan issecond with 399 citations (11 percent), fol-lowed by the UK (360, 10 percent), Italy(226, 6 percent), and France (223, 6 per-cent).

All of these examples show that citationdata can provide a wide range of informa-tion on journals. As stated at the outset, cita-tion data are one category of information 1S1

Table 6 The rmtionafity of first authors of articlespubtished in the hr/ernarionat Jourmd of Cancer,ranked by total 19S4 articles snd showing total1984-1988 SCl@ citations and five-year impactfactors. Onfy tfmae nations with at least five ardclesarc listed.

1984 19S4-1988 Ffve-YearCouotry Items citations Impact

us w 707 11.8UK 29 398 13.7Japan 22 474 21.6France 20 260 13.0Italy 19 324 17.1Sweden 13 160 12.3Israel 11 110 10.0Canada 10 78 7.8FRG 9 120 13.3Australia 6 38 6.3The Netherlands 6 16s 28.0Norway 5 94 18,8Switzerland 5 126 25.2

All Glfrera (19) 36 453 12,6

TOTAL 251 3,510 14.0

considers in evaluating journals for cover-age. Another key consideration is a set ofbasic journal standards.

k18iC Jollimd Standards:Timeliness Above AU

One of the most basic obligations a jour-nal owes its subscribers is timeliness or reg-ularity. It is unethical and unacceptable forpublishers to allow journals to appear chron-ically late, weeks or months *r their coverdate. Of course, tempormy production prob-lems or other factors may sometimes causejournals to be delayed. But if the journal can-not maintain or manage an adequate backlogof manuscripts, the publisher should mergeit or throw in the towel. And if a publisheris unresponsive, then subscribers should re-fuse to accept false publication dates and oth-er rip-offs.z I

Whether or not a journal follows intema-tionfd editorial conventions may also influ-ence 1S1’sdecision to cover it. 1S1has ad-vocated a variety of editorial practices andstandards that apply equslly to establishedand new journals. More informative jour-

TaMeZ W rmiomlfty of ffratmrthoraof 1984-1988SCP papers that cited 19S4items in the InternationalJournal of Cancer. Gnfy those nations with at least

20 citationsare listed.

citing 1984198scountry Citatforla

us 1,091Japan 399UK 360Italy 226Frarrce 223FRG 182Sweden 132The Netherlands 128Csnada 116Austrafia 74SwiIzerkmd 55Norway 54Israel 52Relgium 49Finfand 49Denmark 47Austria 46USSR 39SoUthAfrica 30Czechoslovakia 20

All Others(25) 138

TOTAL 3,510

191

nal titles, fully descriptive article titles andabstracts, complete bibliographic informa-tion for all cited references, full addressesincluding telephone and fax numbers forevery author of all published items, and con-tents page formats are several exam-p]es.422.23

Editorial policy on language is anotherconsideration. We do cover a large numberof foreign-language journals, but the pres-ence of informative abstracts or summariesin English is essential. If editors truly wantwider notice of their journals by the inter-national research community, they ought topublish article titles, abstracts, and cited ref-erences in English. If its English title andabstract seem especially interesting, morescientists might lx encouraged to go to thetrouble of reading the foreign-languagearticle.

Other indicators of quality are also impor-tant-such as whether a journal relies onpeer review to assess the relevance of a sub-mitted manuscript, the reliability of its meth-ods, the originality of ftiings, the com-pleteness of references, and other factors.24

We also look at the recent track record,that is, the previously published works ofthe editorial board members and contribut-ing authors. For example, we examine howotlen they have published; in which journalstheir articles appear; and if their works havebeen cited. We also check their referencelists to make sure the authow are citing abroad range of important journals. This en-sures that the authors have consuke$ the lit-erature and are members of a broad scien-tific community.

The reputation of a particular publisheror professional society is some help in evrd-uating new journals. Of course, a good trackrecord on old journals does not necessarilyguarantee the performance of new journals.Publishers have been known to launch jour-nals prematurely, often at the prodding ofspecial-interest groups. Also, some gover-nment-sponsoredor professional dues-subsi-dized journals may occasionally be adverse-ly affected by annual fluctuation in budgetsand membership renewals.

By considering a combination of citationdata and these basic editorial standards, itis possible to make an informed journal-cov-erage decision in most cases. However, 1S1does rely also on the subjective judgment ofexperts in the field.

Expert Judgment

Since its beginningin the early l%f_ts,1S1has benefited from the expert consultationsprovided by the editorial adviso~ boards forour various products and services. Themembers are distinguished researcherswhose collective multidisciplinary expertisegives 1S1vrduable input about important newand existing publications.

For many new journals, we ako solicitcritiques from people working in the disci-plines covered by the journals. Of course,this has its drawbacks. If a new journalcovers a specialized topic, the experta in thatfield are probably the most likely to put thehighest value on greater coverage of thattopic. This self-interest is weighed when weevaluate critiques and recommendations so-licited from outside experts.

Of course, 1S1’sstaff of subject specialistsalso considers the invited and uninvited ad-vice of subscribers, editors, publishers, andothers about journals we ought to cover ornot drop. 1S1welcomes all informed opin-ions about the value of a particular new orestablished journal. In fact, we are planninga new large-scale survey of 1S1subscriberslater this year that will give us essentitd feed-back on journrd coverage.

How to Recommend Journals

As was stated at the outset, we get manyletters and calls suggesting various journalsas candidates for coverage by 1S1.As a helplid reminder, we request that all nominationsbe accompanied by one or more recentcopies of the journal. It is essential to pro-vide us with the editor’s full name and com-plete address, including telephone and faxnumbers. We also request that you specify

192

the publisher, laqguage, publication frequemcy, and other basic information.

In conclusion, I’d like to reiterate that

journal selection is a painstaking process at1S1performed by staff specialists. It involvesa combination of citation analysis, research

on timeliness, reliability, and other stan-

dards, and the judgment of outside expertsin the field. Our goal is to provide compre-

hensive coverage of the most important jour-

nrds for our subscribers. By basing our se-lection decisions on a thorough and bakmced

analysis of several factors, 1S1will continueto meet its subscribers’ expectations for highquality current awareness and informationretrieval services.

*****

My thanksto Alfred Welljams-Dorof andJudith E. Schaefferfor their help in the preparation of this essay.

019WLU

REFERENCES

1. GarfieldE. Jnurnal selection for Current Contents: editorial merit vs. political pressure. Emuys of aninforrrmdon scientist.’ ghostwriting and other essays. Philadelphm: 1S1press, 1986. Vol. 8. p, %-104.

2. —------- How do we acleet journrds for Current Contents? fbid., 1981. Vol. 4. p. 309-12.3. Jnternatforsal Cnmrnittee of FvfedfcalJourrmf Edftora. Uniform requirements for manuscripts submitted

te biomedical joucnafa. Arm Intern. Med. 108:258-65, 1988.4. GarfieldE. Idiosyncrasiesand errors, or the terrible things journals de to us. Op. cit., 19S4.

Vol. 6. p. 6-12,5. —------ l@ter to edirnr. (More on scientific joumsfs. ) N. Engf. J. .4feri.307:506, 1982.6. —. Tne many journals? Nonsense! 3?ICscientist 2(5):11, 7 March 1988.7. Ber@nirt C G. Snaring prices and sinking sales of scienct monographs. science 183:282-4, 1974.8. Afdridge S. How to cut a swathe through the literature. New W. 125(1707):69, 10 March 1990.9. Garffefd E. Bradford’s law and related statistical patterns. Essays of an informan”on m“entis~,

PMfadelphia: 1S1press, 1981. Vol. 4, p. 47683,10. McLafn J P. Review of ‘“BiblionretricsToolbex” by T.A. Breoks. .f, Amer. Sec. Injorm. SCi.

41(1):70-1, January 1990.11, GarPleM E. The mystery of the transposed joumrd lists-wherein Bradford’s law of scatterin8 is

generalized according te Wield’s law of concentration. Op. cit., 1977. Vol. 1. p. 222-3.12. ——. How re use Journal Citm”on Reporrs, including a special safute to the Johm Hopkins

Meo%crrlJournal. Bid., 1984. Vol. 6. p. 131-S.13, —. The impact of citation counts. Times High. Educ. Suppl. 15 July 1988. p. 12. (Reprinted in:

Current Cbntents (37):3-5, 12 September 1988.)14, -—--—. Uses and misuses of citation frequency. Essays of on in~m”on scientist: ghostwriting and

other essays. Fhiladelphiw 1S1press, 1986. Vol. 8. p. 403-9.15. ------- How te use citation analysia for faculty evaluations, and when is it relevant? Parts 1 & 2.

fbid., 19S4. Vol. 6. p, 354-72.16. —----- The number of biochemical articles is growing, but why also the number of references per

article? J6id., 1981. Vol. 4. p. 414-25.17. Ear B K, Kamo,jal A & IWmafd U M. Letter te editer. (A methcd for determining the impact fatter

of a non-SCI joumai.) J. Dac. 45:139-41, 1989.18. MnrfrW S. Letter tn editer. (hnpact factors of peripheral jeumals,) J. Dac. 45:330-1, 1989.19. Gorffe[dE. Which medical journafs have the greatest impact? Arm Intern. Med. 105:313-20, 1986.

(RepMted in: Essays of an inforrnarion scientist: peer review, refereeing, fraud, and other essays.Phiidelphiz 1S1press, 1989. Vol. 10. p. 7-13.)

20. —. The most-cited papeta of all time, SCI 1945-1988. Parts 1A & IB. Cwmmr Contents(7):3-14, 12 Februmy 199@ (8):3-13, 19 February 1990.

21. -—-. Fatse publication datea and other rip-nffs. Essays of an information scientist. Plriladelphirc1S1PFSSS, 198i3,vol. 3. p. 488-91.

22. —. What a difference an “A” makes. Ibid., 1981. Vol. 4. p, 208-15.23. To indent or net te indent? How tn improve journal contents page formats. fbid., 19S0.

Vol. 3. “p. 267-70.24. RudolphJ & Brackstme D. Trro many scholars ignore the basic rules of documentation. C6ron. High.

E&c. 11 Aprd 1990. p. A56.

193