CHAPTER — 3

INFORMETRIC ANALYSIS

Qenesis & Development

Evolution of the concept informetrics can be tracedback to the introduction of the term ‘Statisticalbibliography’ [l] by Hulme in 1923 to denote theapplication of quantitative methods to library andbibliographic works. After Hulme, Gosnell used the term inhis thesis to ‘emphasise on the quantitative aspect thanqualitative’ [2]. Raisig also claimed that there waspotential utility of statistical bibliography as a methodof analysing information needs [3].

Statistical bibliography originated from two terms,statistics and bibliography. ‘Statistics' is derived fromthe word ‘status’ meaning, state, position, standing.Websters dictionary defines it as "facts or data of anumerical kind, assembled, classified and tabulated so as

to present information about a subject" [4]. Bibliographyis derived from 'biblion' meaning ‘book' and 'graphos'meaning ‘to write‘. Webster's dictionary defines it as "alist of authors writings or the literature dealing with acertain subject or author" [4]. Hulme defines bibliographyas the "science of organization of knowledge" [l].Statistics and bibliography combined to form statistical

56

bibliography, which generally denotes quantitative methodsin Library Science.

In 1948 S.R. Ranganathan suggested the use of’librametry' to refer to ‘quantitative research inlibraries’ [5]. He also recommended the application ofstatistical and mathematical techniques in Indian librariesto improve their day to day activities and services. In1969, Pritchard used the term 'bibliometrics' instead ofstatistical bibliography to denote the "application ofmathematical and statistical methods to books and othermedia of communication" [6]. He considered statisticalbibliography as a misnomer as it may be misinterpreted asbibliography of statistics. Fairthorne describedbibliometrics as the "quantitative treatment of propertiesof recorded discourse and behaviour appertaining to it"[7]. The term bibliometrics was accepted widely instead ofstatistical bibliography after the publication ofPritchard's paper in 1970 [8]. Bibliometrics consists oftwo terms 'biblio' meaning book and ‘metrics’ meaningscience of meter ie. measurement. Bibliometric studies arelimited to recorded information. Pritchard defines it asfollows:

"The definition and purpose of bibliometrics is to shedlight on the process of written communication and of thenature and course of a discipline (in so far as this isdisplayed through written communication) by means of

57

counting and analysing the various facets of writtencommunication".

The British standard glossary of documentation termsexplained bibliometrics "as the study of the use ofdocuments and patterns of publication in which themathematical and statistical methods have been applied"[9]. Potter defines it as "the study and measurement of thepublication patterns of all forms of written communicationand their authors" [10]. Sengupta describes it as"organization, classification and quantitative evaluationof publication patterns of all macro and microcommunication along with their authorships by mathematicaland statistical calculus" [ll].

According to Hertzel, "Bibliometrics the science ofrecorded discourse - which uses specific methodologies,mathematical and scientific, in its research — is acontrolled study of communication. It is the body of aliterature, a bibliography quantitatively or numerically orstatistically analysed - a statistical bibliography; abibliography in which measurements are used to document and

explain the regularity of communication phenomenon" [12].

Bibliometric studies fall in two broad groups —descriptive (productivity count) and evaluative (literatureusage count). Descriptive studies may concern geographic,time or discipline. Evaluative cover references orcitations. According to Nicholas and Ritchie "Bibliometrics

58

provide information about the structure of knowledge andhow it is communicated" [13]. They divided the bibliometric

studies into two groups those describing thecharacteristics or features of the literature (descriptive)and those examining relationships formed between componentof a literature (behavioral). Evaluative studies use thereferences to literature used by research workers in afield. The scope of bibliometrics include studying therelationship within a literature or describing aliterature, focusing on consistent patterns involvingauthors, monographs, journals or subject.

Bibliometrics has grown into a vast subject involvingresearchers from different branches of human knowledge.Thus synonymous terms like 'Scientometrics' coined byRussians to application of statistical methods to scienceemerged. FID developed the term ‘informetrics’ to meanmeasure and quantify information. It also established acommittee on informetrics. Brookes, at the internationalconference on bibliometrics and informetrics held atOntario in l989, suggested that 'Informetrics‘ is theappropriate term to cover all the "quantitative studiesrelated to information science" [l4]. Thus it is seen thatinformetrics evolved from earlier terms _like statisticalbibliography, librametry, bibliometrics and scientometrics.Informetrics is "the study of quantitative aspect ofinformation in any form, not just records or

59

bibliographies, and in any social group, not justscientists. Thus it looks at the quantitative aspects ofinformal or spoken communication, as well as recorded, andof information needs and uses of the disadvantaged, notjust the intellectual elite. It can incorporate, utiliseand extend the many studies of measurement of informationthat lie outside the boundaries of both bibliometrics andscientometrics" [15].

As the term bibliometrics is expanded intoinformetrics, the definitions, descriptions andregularities also applies to informetrics. Informetrics isdefined as "use and development of a variety of measures tostudy and analyse several properties of information ingeneral and documents in particular" [16]. According toLancaster, "Informetrics covers all quantitative analysesof information transfer, whether or not they involve thepublished literature“ [17]. Scientometrics, the term usedby Russians to quantitative studies of science isconsidered a term related to informetrics. "Scientometricsis the study of quantitative aspects of science as adiscipline or economic activity. It is part of thesociology of science and has application to science policy­making. It involves quantitative studies of scientificactivities, including among others, publication, and so

60

overlaps bibliometrics to some extent” [15]. From thedefinition the scope of scientometrics is limited tostudies of science, whereas the informetrics studies arespread over all fields of knowledge. Because of the widerscope of informetrics a series of international conferencesin the topic which discuss the emerging methods andapplications for research in this field are organised.According to Lancaster, the subject has grown from simpledata analysis to well defined subject involving appliedstatistics, modelling, simulation, cluster analysis, studyof citation network etc.

Lit@r§tv£eRevi¢y

One of the first papers based on significantstatistical data was by Cole and Bales in 1917 [18].Statistical analysis of literature on comparative anatomyis done using publication counts and graphic illustrationsin the paper. This paper came out before the subjectbibliometrics was formed as a seperate discipline or evenbefore formation of the concept statistical bibliography.The study by Hulme in 1923 is considered as the firstanalytical account on growth of literature. In this, theproductivity of scientific literature was criticallyevaluated for a period of time and came to the conclusionthat the decline and rise of scientific literature areinfluenced by population change, political and economicmovements. Another importance of the paper was that for the

61

first time the term statistical bibliography was used inscientific literature in 1923.

Later in 1926, scientific productivity formula whichlater became known as Lotka's law was published [19]. Gross

and Gross wrote a paper based on the studies of referencesin "Journal of American Chemical Society" [20]. From theirstudy it was decided to make use of a number of specificperiodicals needed for a college library. This was thebeginning of the first recorded study of citation analysisand later remained as a model for further studies.

Bradford, in 1934, published a paper on scatteringwhich later came to be known as Bradford's law ofscattering, the first important law to appear ininformetrics research [21]. Bradford, while investigatingon 300 abstracting and indexing journals stumbled upon thefact that "out of 750,000 articles, only 250,000 are dealtwith and the remaining 500,000 are missed". This led tofurther investigation by using a bibliography on appliedGeophysics (1928—1931) and Lubrication (1931-1932) compiled

by Lancaster Jones. From this, Bradford established ahypothesis that "to a considerable extent, the referencesare scattered throughout all periodicals with a frequency,approximately related inversely to the scope" [22],Bradford's law is discussed in detail in his book"Documentation" published in 1948.

62

The next contribution to bibliometrics was by Zipf,G.K. when he put forward the relation between rank andfrequency of word usage in his book published in 1935 [23].This later came to be known as Zipf's law of wordfrequency. By using M.L. Hanley‘s "Index of words for JamesJoyce's Ulysses", Zipf studied the frequency and that leadto the formation of law.

Fussler, in a classic paper in 2 parts discussed “theimportance of literature of various subject fields toresearch in Chemistry and Physics" and the temporal span ofthe literature, the principal forms of the literature, thenational origins of the literature used in the UnitedStates, and some attention is devoted to the more importantserial titles" [24].

Garfield, in 1955 presented the concept "Impact factor"[25]. The influence of a journal or an article during aspecific period was termed impact factor. Garfield furtherstated that journals can be ranked by frequency and byimpact of citations for science policy studies. "Impact isa measure of relationship between citations and articlespublished" [26]. Another method to measure journals orarticles cited is by ’Immediacy index‘ which indicates howquickly they are cited. These values are published inJournal Citation Reports. Narin used measure of ‘influence’

of a journal [27] by calculating citation ratio betweentwo journals ie. the extent to which one cites the other.

63

Kessler stated ‘bibliographic coupling‘ [28, 29] as anew method for grouping technical and scientific papers. Apaper reviewing bibliographic coupling was published byWeinberg in l974 [30] which referred to Kessler’s paper as‘classic paper in bibliographic coupling‘.

A paper on epidemic theory, a technique explaining thetransmission of ideas by means of literature was publishedby Goffman and Newill in 1964 [3l].In this, an ‘infectiousdisease epidemic‘ is compared to ‘intellectual epidemic‘.The purpose of this was to help to describe the publicationactivity within a given discipline to determine thenecessity for an information retrieval system. This modeldepends on the rate of change over time of 3 groups ­infectives, susceptibles and removals. Suscelptibles arethose who have not yet contributed paper to literature butmay do so in future. Infectives are those who havecontributed earlier and are no longer contributingarticles. Removals are those who are neither infective nor

susceptibles. This can be applied to clusters of papers —an original paper being infective and papers which citeoriginal — susceptible.

The next important paper in chronological order is byPritchard in 1969 in which the term ‘bibliometrics' wasfirst suggested instead of statistical bibliography [6].Small, describes relation between two documents by a newkind of coupling — “co-citation" which is "the frequency

64

with which two items of earlier literature are citedtogether by later literature" [32].

After 1970, a proliferation of papers appeared onvarious methods and applications. Earlier papers weremainly on theoretical and empirical aspects. Theoreticalpapers accepted mathematical formula rather than empiricaldata to study various distributions. Vickery examinedBradford's analysis and found there is difference in verbaland graphical expressions of the law [33]. Kendall tookinto account the references in operational research andcame across a relation between Bradford and Zipfdistribution [34]. Leimkuhler analysed Bradford's law andsuggested that the ‘law of scattering’ predicts the numberof references for a given portion of journals while the‘distribution’ gives emphasis on the number of journalsrequired to obtain a given portion of reference [35].Brookes criticised the above work and opined that it istheoretically possible but practically difficult to apply[36]. Yablon¢ky examined in detail the bibliometric andscientometric core—scatter distributions and concluded that

there exist a close similarity [37] between these twoparameters. Accordingtx>Haitun [38] Zipf's distribution isbasic to all the other distributions. Bookstein suggeststhat the three informetric distributions as unified models

[39, 40, 41]. Egghe gives more importance to the studies onthe classical bibliometric laws and the similarity anddifference between them [42] while Rousseau discuss the

65

relation_between informetric laws and how they differ [43]in certain vital aspects.

Literature on empirical studies are based on two typesof studies — probability (prediction) and evaluation. Someimportant papers of similar nature are by Pratt [44] Drott[45] Lawani [46] Goffman and Morris [47] Brookes [48] and

Egghe [49].

A third group of papers which describe mathematicalmodels are of importance in the field. Some examples areBrooke's logarithmic model, [50] Price's CumulativeAdvantage Distribution, [51] Bookstein's model linkingBradford, Lotka and Zipfs laws, [52] Coile's model ofScientific productivity, [53] Hubert”s model for journalproductivity, [54] Sengupta's model providing an offsettingweightage formula, [55] Naranon s power law model, [56]O'Neill's model pointing to limitations of Bradforddistribution, [57] Schorr, [58] Voos [59] and Subramanyam[60] on Lotka's law, Wilkinson's model pointing outambiguity of Bradford s law, [61] Worthen s contagionmodel, [62] Vlachy’s work on lotka's law, [63] RavichandraRao [64] and Tague’s [65] negative binomial model. Earlier

models developed by Bradford, Lotka and Zipf have notapplied any Goodness~of-fit tests to check the validity ofdata. Later studies mentioned above applied tests like Chi~square and Kolmogorov—Smirnov tests to accept the models.

Some other work applied regression models to test the data.

66

A new method ie. explicativity developed by Good I.J. isexplained by Tague [66] in her paper as a new method totest models to data. At present models suggested by Brookesand Bookestein, Ravichandra Rao and Tague have gained wider

applications in information studies than any other models.

After the starting of journal 'Scientometrics’ in 1977edited by Tibor Braun, many papers related to informetricsare published in this journal. A number of review Papers onInformetrics which give an overview of the subject andexplain the key concepts form a major part. Important amongthem are those of Broadus, [67] Hertzel [12] Bensman, [68]Lawani [69], Rao [70], Ikpaahindi [71] Narin [27] Zunde

[72] White and Mccain [73]. Bibliographies on informetricswere complied such as by Pritchard and Wittig [74] Hjerppe[75] Schubert [76] and Sellen [77].

By the advancement of technology, informetric studiesare conducted using data collected by online search. Papersby Oppenheim [78], Persson [79] Hibbs [80] Lancaster andLee [81], Stefaniak [82] deal with online studies. Thismethod might get more practical applications in futurethough it has some limitations at present, such as coverageof database, data retrieval method etc.

lnformetric Laws

In many fields researchers find some regularities whilecounting events or tabulating size of things. These are

67

often known as models, measures or laws and are derivedfrom empirical observations or statistical inferences.Informetrics as such deal with the study of library andinformation dissemination processes by using quantitativetreatment of the properties and behaviour of knowledge. Insuch studies some theoretical formulations and generallyvalid laws are discovered. These laws may loose theirvalidity or change when there is an alteration in theexisting environment. In informetrics this type of largescale changes are unlikely to take place. Therefore theselaws remain valid eventhough some new changes are suggestedby later researchers. The three classic laws ininformetrics are Lotka's law of Scientific productivity(based on Author productivity in terms of paperspublished), Bradford's law of scattering (based ondistribution of articles over various journals) and law ofword occurrence (ranking of word frequency in a text).These laws and several new models based on these arediscussed below.

L9tka'§llaw

Lotka in 1926 formulated the law of frequencydistribution of authors in relation to number of articlesproduced. This became known as the inverse square law ofscientific productivity. It states that the relationbetween frequency distribution of ‘y’ persons making 'x‘contributions in xny I c. The value of constant when n = 2

68

was then calculated. Lotka explained the phenomenon asfollows:­

"In the cases examined, it is found that the number ofpersons making 2 contributions is about one fourth ofmaking one, the number making 3 contributions is about one­

ninth etc the number making n contribution is about l2 ofthose making one, and the proportion of all contributions,that make a single contribution is about 60 per cent" [19].

Lotka_ studied counts of chemists, and theirpublications based on data from Chemical Abstracts andAuerbach's bibliography on Physics. The law was termedLotka's law by 1949 and the applicability to otherdisciplines was tested only by 1973. Murphy studied theapplicability in humanities and found it was fit to thisfield [83], Voos, in 1974 measured productivity of authorsin the field of information science and found that "the

relationship in this field is 1 instead of Lotka’s plz"n 3'5 n[59]. In 1974 Schorr, applied the law to library scienceand showed that Lotka’s law does not apply to the field oflibrary science. In 1975 Schorr studied map librarianshipand concluded that the law fits in this field. But Coileshowed that Schorr‘s calculation was wrong as the law doesnot fit to map librarianship [84].

Pao, states that several studies have assumed theinverse square relation as the basis for testing, and

69

others derived the value of constant 'c' from thepercentage of single paper contributors which cannot betraced back to Lotka's assumptions. "Therefore, a uniformmethod should be agreed upon by those attempting a test.Comparison and generalisation on author productivity may bepossible only if compatible data are available and resultsare significant" [85].

There have been many analytical approaches differentfrom Lotka’s law for scientific productivity. Narin,concluded that "scientific talent is highly concentrated ina limited number of individuals“ [27]. Dennis, found aclose correlation between quantity of scientificpublication and achievement of eminence. Price, concludedthat the number of elite in science is small compared tototal number of scientists. According to Price an elitemean an eminent scientist producing scholarly writing.Bookstein, suggested a theoretical model which is ageneralised version of Lotka's law [39], f(x) = _;5__ where

Xzfl

k and.¢C are constants. According to this, the number ofauthors with x papers is proportional to l/X5.

Some recent studies on author productivity by Pao,Nicholls and Griffith used the version of Lotka‘s law byBookstein, in their studies and estimated the values of nrather than using n = 2. They counted authors and suggesteda goodness of—fit test for the model. Nicholls showed that

70

the generalised version is "surprisingly well fitting andstable" [86] and Pao concluded "overwhelming conformity"[87] to this model.

Price suggested that "Half of the scientific papers arecontributed by the square root of the total number ofscientific authors" [88]. This became known as Price'ssquare root law and was proved to be invalid boththeoretically and empirically [89] by further studiesconducted by Nicholls.

while applying Lotka‘s law, the problem of creditingauthorship to multiauthored paper occurs. Originally Lotkacounted only first author in multiauthored paper. Booksteindiscusses this in a paper and concluded that "if Lotka‘slaw holds for one accounting method, it will hold for anyother one in which the change in the typical amount ofcredit given to authors per paper may vary from author toauthor but does not depend strongly on how much the authors

published. If this is true, the investigator can give anyreasonable system of assigning credit to authors whilestudying author productivity [90].

In author productivity studies it is seen that thenumber of single paper producers are more. This aspect hasbeen given a thorough scrutiny by many others. It is shownthat authors who are more productive are having morecollaborative studies than single paper producers. Because

71

of the transdisciplinary nature of research topics, thereis more scope for multiauthored paper than single authoredpaper. Jointly authored papers are cited more times than asingle authored paper. Lawani has shown that "citation rateand quality of paper (as judged by a form of peer review)both correlate positively with number of authors per paper“[91]. In addition to collaboration, individual productivityis affected by working environment, motivation, rewardsystem etc.

B1'§df_9F_d 'aI~_<'=!:s

This law is the most prominent informetric law as it isapplied to the control of literature. Bradford discoveredthe scattering phenomenon while studying the extent towhich literature in a discipline is scattered over a rangeof journals. This was published in his classic article in1934 [21]. Earlier, he was working on the necessity forstandard- bibliographical methods to avoid duplication andwaste of time, money and to give better informationservice. While studying 300 abstracting and indexingjournals duplication and omission of articles were found.This finding led to further investigation of twobibliographies on Applied Geophysics and Lubrication. Forboth subjects, tables giving number of journals producingcorresponding number of articles were listed and thenarranged them in the decreasing order of productivity. Thelist was found to have three zones each having the same

72

number of references. The number of journals in one zonewhen divided by number of journals in the preceeding zonewas found to be a constant which is known as Bradfordmultiplier. In the data studied it was approximately 5.Using the data Bradford constructed two graphs. Oneplotting the logarithm of cumulated number of journalsagainst cumulated number of articles. He observed that thelater part of the curve was close to a straight line andobserved that "the aggregate of references in a givensubject, apart from those produced by the first group oflarge producers, is proportionalof sources concerned when these

of productivity" [21]. Based ongraph was constructed to developthe straight portion of curve

to the logarithm of numberare arranged in the orderthis observation anotheran algebraic relation forFrom this the law was

originated which states that:

"If scientific journals are arranged in the order ofdecreasing productivity of articles on a subject, they maybe divided into a nucleus of periodicals more particularlydevoted to the subject and several groups of zonescontaining the same number of articles as the nucleus, whenthe number of periodicals in the nucleus and succeedingzones will be as l:n:n2 [21].

Bradford's law received wide attention after thepublication of his book in l948 in which the law isdiscussed in detail. In the book he also used the same data

73

as in the first paper in which the law was presented. Sincethen the law had been used in various studies andmathematically is expressed as:

F(x)=a+blogx where F (x) is the cumulated number ofreferences contained in first x most productive journals,'a' and ‘b' are constants.

Bibliographic studies in a number of fields confirmthat the dispersion of articles in a set of journalsconform to the statistical distribution of the typedescribed by Bradford. But arriving at_ a mathematicalexpression for this distribution led to many articles onthe topic.

In 1948 Vickery published a paper based on analyses of1600 periodical references. He found that Bradford’s lawwas not in total agreement with his algebraic expression.Vickery concluded that the relation which Bradforddiscovered only "fits the upper portion of the observedcurve" [33]. Cole, also took a new look at referencescattering in which he plotted the cumulative fraction ofreferences against the lQgarithm11fthe cumulative fractionof titles and he named the slope obtained as "reference­scattering coefficient" [92].

Leimkuhler was the first to give a theoreticalexpression to the scatter of journals, which was found to

74

be the inverse function of the Bradford distribution.Leimkuhler's Bradford distribution is given as

In (I + Bx)F (x) = ——————————— —— where x denotes the fraction ofIn (I — Bx)

documents in a collection which are most productive, F (x)denotes the proportion of total productivity contained in

the fraction x. The parameter P is related to the subjectand completeness of collection [35].

Brookes gave a simplified form of this formulation asthis required much tedious computation for practical use.The new form is B (r) = a + k log r, where B (r) representtotal number of articles published in journals upto rank r,a and k are constants. This became the widely usedformulation of Bradford's law [93].

Wilkinson in 1972 suggested that the formula providedby Brookes and Leimkuhler did not describe the samephenomenon. The error was due to Bradford s graphicalrepresentation of his Applied Geophysics data and theverbal expression of the law. Wilkinson observed thatLeimkuhler derived his distribution from Bradford's verbal

representation and Brookes from the graphicalrepresentation [61]. In comparative tests Brookesformulation conformed more closely to empirical data.

Naranan modelled the dynamic process of accumulation of

articles in journals in terms of a power law distribution

75

assuming that number of articles and journals growexponentially [94].

This assumption was not found valid as Bradford's lawfits even when articles and journals do not growexponentially. Brookes suggested that Naranan's analysis ismore valid for Lotka’s law than for Bradford's law.

Lawani, in a study of literature of agriculture foundthe distribution conformed to Bradford's law. The curveobtained was linear with a Groos droop [95] for journals oflowest rank. This was a point of controversy. Brookesviewed that this was due to the incomplete bibliography.But O'Neill concluded that the Groos droop is not due tothe incomplete bibliography but is a part of Bradforddistribution [96].

Goffman and Morris applied Bradford's law to a study onlibrary acquisitions. They stated that "Bradford's lawshould apply to the use of periodicals in a library as wellas to the dispersion of articles among journals" [47]. Theyshowed that the distribution of both circulatingperiodicals and their users obey the law. Hence thesmallest core of journals which must belong to thelibrary's collection can be defined. This core shouldconsist of a minimal nucleii of periodicals circulating inthe library plus the minimal nucleii of journals devoted tothe subjects of most interest to the library's nucleus of

76

users. As the funds allow successive zones of periodicalcorresponding to circulation and user interest can beadded. As a result the library collection can be maintainedin an orderly fashion and viable state thus providing itspatrons with the most potentially usable material for thefunds at disposal.

Fasler, in a comment on the above wrote that it was the

most promising method but with a warning that "before it ispossible to discontinue a journal subscription, it isnecessary to make sure that such action will not causegreat inconvenience [97]. In 1975 pope claimed that "thearea in which Bradford's distribution has the greatestpotential is in collection development" [98].

In 1977 Brookes did a complete reflevaluation of the lawand stated that "the analysis of Bradford's law hashitherto been applied to theoretical models which are toostatic, too deterministic and too physical. All Bradforddata are derived by observing the activities of a set ofsources over some appropriate period of time and by notingthese activities, as measured in terms of the numbers ofitems each source accounts for in that time. Thus theBradford's law is concerned with:

l) a finite set of active sources whose activities aremade to manifest with the generation or consumption ofa specified type of item.

77

2) observation of those activities over a specificsampling period.

3) items of some homogenous kind which are discrete andcountable [50].

Statistical distributions depend on relationshipsbetween the number of active sources, the range andintensity of their activities, and the period ofobservation which provides the sample data. All Bradforddistributions are samples of some ongoing activity but alltoo often, the sample data have been regarded asconstituting a total population. Brookes was a majorwriter on Bradford's law. In one paper he states that anew statistical theory based on both Bradford and Zipf'slaws ‘together "provide the most convenient analyticalinstrument for the exploration of social science data"[48].

A general formulation of the law is given by Asai [99]in which five types of laws formulated by earlier papersare combined and made into one. This paper provides a non»

linear regression technique for estimating the slope,intercept and "shift in a straight line to log rank" of theBradford curve. Chen and Leimkuhler [100] noted thatAsai's formulation treats rank as a continuous variablewithout gaps which lessens the realism of the model. Theysuggest the ‘index approach‘ to rectify this. They alsoidentify two parameters that affect the shape of Bradford

78

curve (non—linear) in least productive and most productivezones. Drott and Griffith have shown that the linear slopeof Bradford curve is related to the number of articlesplotted and the intercepts are related to the number ofjournals [l0l].

Karmeshu et al derives a rationale for Bradford's lawbased on random subdivision of papers over the field ofjournals and on individual scientific productivity [lO2].Bradford‘s law and other statistical laws are explained byPrice as cumulative Advantage Distribution of the success­breeds—success type [51]. He obtains a relation of theform F(n) = (m+l) B (m+2) when n is number of successes m

is a constant and B a Beta function.

A few examples are:

a) A journal which has been used is more likely to be usedagain.

b) An Article in a journal cited many times is morelikely to be cited again.

c) An author of many papers is more likely to publishagain than one who is less prolific.

Ravichandra Rao also supports this but shows that thenegative binominal distribution describes this phenomenonbetter than cumulative advantage distribution [l03]. Instatistics this is generally described by a hyperbolic

79

distribution function known as Yule distribution. This isshown to vary with sample size. Yule characteristic

l+(sample variance — mean/mean)2

K l “““ "'§;§.£;;_;E"EI;;;;; _____ "

Garfield advances a measure of concentration ofarticles in journals [l04]. This Bradford phenomenon showsa gradual concentration of articles to a few core journalsin a subject. This is due to the explosion of literatureand number of journals. Bradfords law enables to do aliterature search among numerous journals and saves theusers from ‘documentary chaos‘.

Maia and Maia show that there is no ambiguity inBradford’s theory of distribution [l05]. It is alsoverified that in the Bradford's experiment the verbal andgraphical expressions are in agreement showing that thereis no ambiguity. Earlier, Wilkinson also studied theambiguity of Bradford's law and concluded that adocumentation researcher has no means to decide which

formulation (verbal or graphical) should be used. O‘Connoor and Voos also examine the relation between theoryand empirical laws [lO6]. Another paper by Drott alsostudies the theoretical and empirical development ofBradford's law [l07]. But Bookstein in a recent paperdiscussed the various informetric laws as the differentversions of a single regularity and explored the

80

consequences of demanding that informetric laws areresilent to ambiguity [90].

Burrell studied the Bradford phenomenon and opines astochastic process termed ‘Waring process‘ which conformsto the general features of Bradford's law [108]. Inanother work he examined the dynamic nature of bibliometric

processes by analysing a bibliography compiled over anextending period of time using Bradford and Leimkuhlercurves. The theoretical and practical importance ofstochastic processes to model these systems are also given[l09].

Recently, Aparna Basu attempted to find a theoreticalfoundation for Bradford's law and suggests a modified log­linear two parameter model to explain journal productivity[ll0].

3.3.3 Zipf[s paw

Zipf, explained this law in 1935 based on the frequencyof occurence of words in a text. It states that "if thenumber of words occurring once in a given sample is takenas x, the number of different words occurring twice, threetimes, four times, 'n' times in the sample is respectivelyl l 1 1iii? ilii "a __qn-Innli

22 I000 O O O O I O O O O O QOI

of x". From this zipf developed a formula ab2=kwhere a is the number of words occurring b times.

81

Zipf explained the law as a consequence of a general"principle of least effort" in his book published in 1949[lll]. Zipf applied his principle to study Hanley's Indexof words for James Joyce‘s Ulysses‘ and found a clear-cutcorrelation between rank (r) and frequency (f) of word,which is represented as rf = C. when the law is appliedexactly, 'a hyperbola is obtained while plotting thefrequencies against rank.

Many studies were conducted on Zipf's law. Wyllys,expressed it as "one of the most puzzling phenomenon inbibliometrics" [ll2], Simon, in a paper observed thefitness of the Yule distribution to a number of empiricaldata including word frequency [ll3]. Hill, wrote a groupof papers on Zipf’s law. Mandelbrot, published severalobservations on the generalizations of Zipf's law. In l975Wyllys presented a paper at the 38th Annual meeting ofASIS, which gives that "inclined towards mysticism, Zipfnot only leaped to the conclusion that the true slope ofrank frequency curve was -l, but also claimed that thisregular slope resulted from some fundamental force ofnature" [ll4].

Hubert, reviewed the law and its development in laterwork. Fedorowicz applied the distribution of indexingterms in inverted files of bibliographic databases usingZipf and other models [ll5, ll6]. In another paper byTague and Nicholls, attempts in relating the law to file

82

design and general bibliometrics is done [ll7]. The simpleZipf size-frequency distribution for tokens is presented as

g$ = a/xb where gx is the number of type with x tokens, ais the number of types with single token and b dispersionof token over types.

Zipf‘s law, like other informetric laws can be relatedto the forms of description traditional in statistics likecumulative distribution function. It is also observed thatPareto's law in economics which relates the income andnumber of people are a variant of Zipf‘s law.

InformetricydistributionspLotkaeBradfordeZipf_relation

Number of studies have been made regarding thetheoretical and practical implications of the informetriclaws. These led to the study of relationship between thedifferent laws. Earlier relation between Bradford andZipf's law were explained by researchers like Kendall, thatthese two are almost equivalent. He supported the ideathat Bradford distribution is really Zipfian and provided amore refined statistical explanation for the straight lineobserved by Bradford. Leimkuhler also agreed with Kendallstheory that the two laws were essentially two differentangles of looking at the same thing. Brookes studied theBradford~Zipf phenomenon and says that "the near identifyof Zipf’s and Bradfords laws are not immediately obviousbecause, in practice, the most marked deviation ofempirical data from the mathematical expectations of

83

Bradford's law are likely to occur among most productivejournals of the nucleaus" [36]. Journals also lackstatistical unity as they are not issued at regularintervals and number of articles per issues also vary.

Garfield writes that "while each ‘law’ applies to adifferent specific phenomenon, they all tend to demonstrateone thing that a few (journals, scientists etc.) accountfor the many (articles, citations etc)" [l04]. He relatesthe Pratta index which measure the degree of concentrationof papers in a subject to Bradford's law. Many theoriesunifying the various statistical distributions are putforward. Garfield proposes a theory of ‘concentration’which points out that for any field of science, articlesare concentrated essentially within the same highly citedor multidisciplinary journals. Price proposed a unifyingconceptual model for Bradford, Zipf, Lotka and otherstatistical laws by the theory of "cumulative advantageprocesses" which is based on success~breeds~successphenomenon [51]. Ravichandra Rao also supports thisphenomenon in the case of use of documents.

All the informetric distributions are claimed to beessentially one by authors like Yablonsky [37], Bookstein[39, 40, 41]. Haitun [38] considered Zipf‘s as fundamentaldistribution. Brookes [48], argues that a new statisticaltheory based on Zipfian distribution is needed forapplications to social sciences. He proposes the use of

84

Bradford and Zipf law together for analytical explorationof social science data. In another paper he says that bothBradford's and Zipf's are rank—frequency distributions. Healso opines that there remains a theoretical gap yet to bebridged as long as there remains some aspects inbibliometrics beyond the reach of techniques dependent onanalysis of frequency distribution.

A common functional relationship between the three laws

are derived by Chen and Leimkuhler using the data fromKendall's study of bibliography of operational research[ll9]. They also propose a more useful formulation for thethree laws.

Egghe gives an inventory of bibliometric laws anddevelops the ‘criteria by which they are the same ordifferent (eg. asymptotically)‘ [42]. The informetricdistributions are described as ‘variants of a singledistribution‘ by Bookstein [40]. These provide rich anddiverse issues and approaches for research in the field.

Research in the field using the laws have shown closesimilarity or even identity between the models underlyingthe empirical distributions. However since Bradford's lawdeals with journal studies and Lotka's law with authorstudies it is easy to see the literatures as differentforms of the same mathematical phenomenon. The advantage

of informetric distribution is their simplicity and

85

familiarity. But their use will depend on how well theyassist in making decisions.

¢i£ati°n-Ana1y§is

Informetric studies depend mainly on document units orwritten records of communication ie. books, journals,articles, reports etc which form the objects of studies.As the physical documents are difficult to handle in largequantities, the "document representations" are consideredfor informetric studies. Currently bibliographies and databases or data from users are also used in the studies.These can be considered as secondary sources in informetric

studies. The "document representations" are usually knownas reference or citations. References are theacknowledgement one document gives to another whilecitations are acknowledgement one document receives from

another. According to Weinstock "when a scientist ortechnologist publishes an article, he should refer toearlier articles which identify earlier researchers whoseconcepts, methods etc. were used by the author to develophis own article" [l20].

Citation analysis is the most common technique which is

considered an authentic tool for journal evaluation. Itis based on the concept that articles citing an earlierwork in scientific paper have much of their content incommon. This has led to many studies including citationcounts, impact factor, bibliographic coupling, co~citation

86

and obsolescence. Citation studies are valid for assessingthe quality of research produced by individuals andinstitutions. It also correlates to the use of librariesby researchers. According to Garfield, Citation analysis"provides a number of interesting and useful insight intothe network of journals that function as primary, formalcommunication medium of science" [l2l].

"Citation analysis means the analysis of the citationsor references which form part of the articles in journals"[70]. Citations in journal indicate the connection betweentwo documents — one which cites and the other which is

cited, whereas citations in secondary periodicals does notimply any connection between documents. The main objective

of citation analysis is to evaluate and interpret citationsreceived by articles, authors, institutions and otheraggregates of scientific activities. It is also useful inmeasuring communication links in sociology of science.

Barker in his thesis noted that there are two types ofcitation studies — (a) studies based on productivity in allor part of scientific literature and (b) those reflectingthe use of all or part of the literature [l22]. In thefirst type the source may be major abstracting or indexingjournals, review journals, bibliographies or data bases.The second type of citation study is based on literatureused by an author. Use of Bibliographies or abstracts arenot considered true citation studies as they are compiled

87

for the purpose of use and do not represent actual use made

by the users. True citation analysis deals with works citedas having actually been used.

Many arguments for and against the validity ofcitations by scientists are often given. Some such as toomuch self citation, negative citation, citing paperswithout seeing the original, citing to get favour of mightyand comfort lowly etc., are put forward against thevalidity of citation studies as actual use studies. Butthe honesty and integrity of the majority of the scientistswho may not involve in such practices is to be givenweight. In short, citations are the major tool forevaluating quality of research journals.

Many studies have found numerous reasons for citationsto papers by the authors. Reasons for citations receivedby some highly cited papers given by Oppenheim and Renn[123] are:

l. Historical background2. Description of other relevant work3. Supplying information or data for comparison4. Use of theoretical equation5. Use of methodology

6. Theory or method not applicable or not the best one

Selection of source for citation studies are usuallydone by sampling as the collection of documents are usually

88

too large. One main source of data for citation studiesare the science citation Index (SCI) published by Instituteof Scientific Information (ISI) from 1963 onwards. Thesecitation indexes give a list of documents that have beengiven as source journals covered by the index. The citeddocuments are ordered alphabetically by the author. Thuscitation index forms a structured list of all citations ina given collection of documents. From 1973 ISI startedpublishing Journal Citation Reports (JCR) which are alsouseful in citation studies. This gives three lists—first alist of most frequently cited Journals for the periodcovered by JCR; second a list of journals in which they arecited (source journal listing); a third part shows the listof citing journals to the frequently cited journals(reference journal listing). The use of SCI for citationstudies has its own merits and demerits,§ome problems aredue to self—citation, multiple authorship, homonym, errorsin bibliography etc.

Initially the collection characteristics such as form(monograph, journal etc), type (articles, notes, news etc),subject (specific subject), set of authors, institutionsetc are to be determined for citation studies. Then theobjects of the study such as references, citations etc areto be selected. Other parameters like sources detail,level of aggregation, variables to be studied, period ofstudy, methods, approach and purpose are to be determined

89

prior to the study. There is a large number of literatureavailable on citation analysis. Some of the earlier workare those of Garfield [124] Tagliacozzo [125] and Hjerppe[l26].

The first record of citation analysis is an article byGross and Gross in 1927 in which the citation count isused to measure adequacy of a college library. Followingthis, a number of papers defining the importance anddispersion of the various segments of scientific literaturecame out. The publication of paper by Bradford analysingthe importance of a small number of core journals for aspecific subject became the basis for research afterwards.More attention was made on managing the rapidly growingscientific literature. Price made use of a number ofliterature counts to devise his outline of scientificenterprise. Citation counting began to attract moreattention as a means of structuring scientific literature.Some of the important work which belong to this type are byGarfield [26], Mccain and Bobick [128] and Brown [l29].

Citation counts are often taken as a measure of the use

of journals though the validity of this assumption is oftencriticised. Scales [130] in a study of the most usedjournals in National Lending Library compared it withhighly cited journals listed in JCR. She conlcuded thatthere is no relation between actual use and citations.Brown suggested that ranking of periodicals in a discipline

4

90

also was made by selecting related secondary periodicals assource journals. Frequency of abstracts was taken toreflect the frequency of citation in a primary journal.This method also have limitations as the objective of theabstracting journal may be comprehensive to the subjectcoverage. Sengupta, in various studies suggestmodifications like weightage formula [55], bibliometricparameters [131] and using annual review to collect sourcedata [132] to eliminate the limitations in the use of rawcitations. Recently informetric studies using data fromon—line data bases are also suggested and attempted [l33,134]. The scope of studies based on citation analysis arevast. Some include ranking and evaluation, growth anddecay, aging or obsolescence, clustering, bibliographiccoupling, trend analysis etc. The most important andcommonly used parameters among them are discussed below.

Ranking_&Evaluation

Citation analysis is used commonly for preparing ranklist of journals and evaluate journal in specific subjects.Earlier, most libraries relied on the judgement by theselection committee for deciding the journals forsubscription. This usually depended on the utility in alocal library. But for a library specialising in adiscipline? utility alone may yield misleading data as theusers turn out to be biased and the existing titles may beused repeatedly. Moreover if the funds are limited and the

91

demand for titles are more, then the librarian may have toadopt some criteria to find the most relevant titles. Themost journal ranking studies use more than one criteria toevaluate the journals related to specific disciplines.Ranking journals using citation analysing combined with‘impact factor‘ study are very useful in the presentcontext.

This type of citation studies help to identify coreperiodicals and formulate a need based acquisition policyin a scientific library. It also enables to assess thequality of research journals. Evaluative studies may havesome limitations as this depends on the sample selectedwhich may be from geographically and linguistically limitedarea. To overcome this, the number of sources can beincreased covering almost all areas. The method ofstreamlining acquisition of journals in research librariesby finding core journals is economical and practical.Bradford's law or the other models derived from this areusually applied for journal ranking studies. Some studiesare also based on data collected from users in one or agroup of similar libraries.

Journal citation studies show that journals arestrongly connected through the references in the articlesthey contain. Thus the core titles may be the highlyconnected journals. Garfield has done a number of journalcitation studies on various disciplines [l35]. The most

92

informetric studies are on journal citations or their use.This is because the journals are a major communicationchannel for research in science and other fields.

Citation analysis also enables to study the number ofpapers produced by an author (author productivity orscientific productivity) by an institution (institutionalproductivity) by a country, (national productivity), thenumber of citations received by an author and, major areasof a subject where active research is going on etc.Measuring science by this method is helpful in determiningthe science policy of industrialised countries as well asless developed countries of the world.

Author productivity studies are done by using Lotka'slaw which is based on the assumption that frequency ofauthors producing papers is constant. Some of the studiesby Pao [l36], Nicholls (89) and Gupta [137] tests theapplicability of lotka's law and its generalized form.Tests such as chi—square and K-S statistical tests wereused to test the applicability of the law. Otherinvestigations on scientific productivity led to thePrices theory of cumulative advantage distributionaccording to which if an individual (scientist, journal)becomes successful (publish yields an article) at the firstattempt, the possibility of success at subsequent attemptsincreases. However, a failure does not diminish theprobability of success on the next attempt. According to

4

93

Bookstein [138] Lotka's law is invariant under the impactof society on the pattern of scientific productivity.

Prices success—breeds—success phenomenon can also be

explained by the 80/20 rule. According to this 80% ofitems (articles, _number of publications, number ofcitations received etc.) are accounted for by 20% ofsources (authors, journals etc.) In other words 20% ofjournals receive 80% of citations or 20% authors contributeto 80% of literature. Egghe [139] shows that the 80/20rule is much stronger if the underlying distribution isLotka type.

ImpactjFactory(I§)

Citation frequency of a journal is not only a functionof significance of the material it published but also thequantity of material it publishes. Thus the most cited

journal may be the most productive journal also. In viewof this relation between size and citation frequency, thesize can be taken to assess the citation frequency. Thusthe method of calculating "impact factor" or the influenceof a journal article over a specific time was adopted injournal ranking studies. This method was first done byGarfield in the Journal Citation Report (JCR), a by productof Science Citation Index.

Journal Impact factor reflects the average citationrate per published article. It is calculated by dividing

94

the number of times a journal has been cited by the numberof articles it has published during some specific period oftime.

Number of citations received in year 3 by articlespublished in year l and year 2.

IF c "'1;;;.;;;";;";;;;;1;;';;;11;;;;';;;;;;;;;'1"gig?"

Ranking of journals based on impact factor tend to bedifferent from ranking based on absolute citation. Impactfactor shows the quality of the journals. Higher theimpact factor, greater is its quality. Journals rankedaccording to their impact factor can also be used forjournal selection in a library.

Some limitations of using impact factor may occur suchas fixing impact of journals not covered by sciencecitation Index. A method to overcome this is originated byINSDOC, which is used by Sen etal [140] in a paper. In thismethod

IF - _§i-f;¥2­Yl+Y2

where Y1 and Y2 are the number of papers published in ajournal during two consecutive years, X1 is the number ofcitations these papers receive by citation Index journals

and X2 is the number of citations these papers receive fromthe journal itself during consecutive year.

A new method of finding Discipline Impact Factor (DIF)formulated by Hirst [141] in which the citations made by a

95

few core journals are taken for computing impact factor.This gives an indication of the impact to the discipline.In this method, first a candidate journal of a specificdiscipline is selected from the most cited list of titles.The discipline influence score is computed by

No. of times J1 cited Journal ADISA = ---------------------------------------- -­Total No. of times J1 cited all Journals

where DISA is the discipline influence score of journal Ain the candidate journal set. J1 a member of DisciplineJournal set and 'n' is total number of journals inDiscipline journal set.

A study by He and Pao demonstrated this disciplinespecific journal selection and compared the results withthat of expert evaluation, impact factor and total citationranking [l42]. User's preference was found more close toDiscipline influence score than the other two methods.

Using the method of measuring impact factor, the outputof a scientist,scientific organization or scientific agencycan be analysed.Individual impact is a measure of impact ofan individual's published output [143]_Institutional impacthelp to evaluate the output of research organisations in acountry, various nations [144] or of a specific agency[l4R]. Opinions are divided on the evaluation of impactfactor. Some people consider it as a misleading indicatorof true impact [l46]. Still some think that the ranking and

96

impact are relevant only in specific disciplines. In somesubject, impact factor of journal varies from country tocountry [l47]. Use of a normalised impact factor (NIF) issuggested by Sen and Kumar [148] to overcome differencesamong fields. This is given by adopting a 10-pointweightage scale. The_top ranking journal in each subject is

given a weight of 10 and others are scaled downaccordingly. This ‘method is useful to study NIF atinstitutional or individual level.

Similar to impact factor, the ‘immediacy index’ and'influence‘ are also suggested as measures to journalcitation studies. Earlier, Garfield has shown that thenewly published journals in a subject show high impact[26}. Immediacy index is a measure of how quickly thepapers in journals are cited. It is calculated as fol1ows:—

Number of citations received in year Xby items published in year x

—III$ii-Iiiii130-QjiiI|1ijQ|Qiic|_~QQ-‘iii-Q'—ii-mnqnuixicmunw-“iii

Number of items published in Year x

Tomer has shown that journals with high impact have highimmediacy index [l46].

Another measure of impact is the influence of ajournal. This method is used by Narin [27] which involves

complicated calculations which can be done by computeronly. This involves finding citation ratio between twojournals. From this data the influence weight can becalculated. The influence weight depends on the citation

97

relationship of a journals with all the other journals in adatabase. Influence weight multiplied by average number ofreferences give the influence per publication, the value isthen multiplied by number of papers to get the totalinfluence. This calculation is more refined than impactfactor as it takes into account the interrelationship ofjournal citations.

Impact factor and Immediacy index are ratios. Impactfactor is the ratio of number of citations which a journalreceives in the course of a given year to the number ofcitable items published by that journal within the twopreceeding calendar year. Immediacy Index is the ratio ofnumber citations a journal receives in its most recentcomplete year of publication to the number of source itemspublished during the same interval.

Ranking of journals based on informetric methods arefound mostly to agree with the expert opinion. Someresearchers are adopting new methods combined with citationstudies to formulate list of journals more specific todifferent subjects. Sengupta has adopted three parametersto rerank periodicals in the field of biochemistry [l3l].The parameters were (a) scientific interest of a paperbased on a number of papers published (b) compactness ofthe information content in a scientific periodical (c) thescientific value of a paper in relation to compactness ofpresentation.

4

98

Aqinseorobseleacsnse

Obsolescence is the process whereby the materialsbecome no more useful or reliable. Aging or Obsolescence isalso a measure of quality of a journal. It helps to assessthe decline in use of a set of documents over time. Studies

on aging are conducted in collection management aspects inspecific libraries or in the journal references as theindicator of previously published work. The reasonsattributed for aging of journals may be varied likechanging scenario in scientific technique or extension ofscientific knowledge etc. Rate of aging varies withdiscipline and is found as a characteristic of scientificand technical literature.

Aging studies are of two types-synchronous studieswhich use data of citations from a journal or a subjectduring a period to measure how old are the cited journalsor compute ‘median citation age‘. In second type»diachronous studies, citation history of a document orsubject from the beginning to the end of a period arestudied and ‘half-life‘, is calculated. It is seen thatboth types give different results.

According to Brookes the rate of obsolescence isrelated to the rate of growth of scientific literature andthe number of contributions in the field [l49]. If thesemeasures are constant, then the rate of obsolescence alsoremains constant. Ravichandra Rao and Meera studied the

99

relation between growth and obsolescence and remarked that

"faster the growth of literature then quicker theobsolescence as well as the half~life" [l50]. Wallacehypothesized that "those journals that were mostproductive would have short active lives and as thejournal productivity decreased, the average active lives ofthe articles would increase". According to Griffith "halfthe total citedness of a volume of an average scientificjournal would be exhausted within five years after thepublication and more than 90% after 20 years" [l5l].

Obsolescence has been the concern of librarians forquite some time because of the overabundance of materials,lack of housing facility etc. By studying the aging, the‘half _life' or the time during which one half of theliterature currently active originated can be calculated.The half life and median citation age shows how far back intime one must go to account for the age of one half of thebibliographic references published in a journal in aparticular year. Price studied the percentage of referencesto works published in the most recent five year periodknown as Price's Index [l52]. He attributed the citation ofrecent papers to "Immediacy effect" due to the citation ofephemeral papers at the research front.

Aging pattern shows the characteristic of sourceselected. In social sciences highly cited journals aged theliterature more quickly than a randomly selected sample of

100

journals. Highly productive journals showed low mediancitation age and low productive showed high median citationage.

Line and Sandison [153] have criticized the study ofaging using citation data or use data. They state that thegrowth of literature must be allowed for by calculating thedensity of the use for each title considered. Linecalculated the density of use (use per metre of shelf) andfound the evidence of increasing density of use withincrease of age [l54]. There was no aging found but the lowcitation density was attributed to the inaccessibility ofthe old volumes which were kept in the basement of thelibrary.

In‘ a study Christdvao came to the conclusion thatarticles from developed countries aged the literaturefaster than articles from developing countries [l55]. Thisindicates that the difference in aging is related to theinternational or local relevance of research topic. Aversain a diachronous study using Journal Citation Reports datafrom 1972 to 1980, concluded that for one group ofcitations showed peak in the third year and declinedthereafter while that for the highly cited group showedpeak in the sixth year and then dropped off slowly [l56].Two types of aging pattern is related to the role of papersin subsequent research. Technical and experimental paperspeaked later and aged less rapidly than papers with

101

specific research findings. Obsolescence studies enablecollection management in libraries, help in informationgeneration, study of growth of literature etc.

Qibliographicmcoupling and clustering

In 1962 Kessler reported on bibliographic couplingbetween scientific papers. According to him a number ofpapers have a meaningful relation to each other when theyhave one or more references in common. When the unit of

coupling is a Single item of references in common then itis said to have similarity in content between the twopapers. This study helps to identify related papers in aspecific subject.

Clustering occurs when one item of paper is referencedby a number of journals. All these journals ikwnua cluster.Carpenter and Narin studied clustering of journals in thedisciplines of Physics, Chemistry and Molecular Biology[l57]. This aids in classifying related journals moreprecisely based on their relationship. This is based on theassumption that journals which deal with the same subjectarea may have similar referencing pattern and the journalswhich deal with the same subject may refer to each other.Cluster studies are indicators of birth, growth and deathof science and their social structure. Cluster studies were

done by using bibliographic coupling but now this isreplaced by co—citation, a measure developed by Small. Co~

4

102

citation changes with time but bibliographic coupling isstatic.

Coecitation

Co-citation studies help in mapping of literatures.When one paper cited two other papers together it isimplied that the two papers are related. If the papers arecited together by more authors, then their relation isclearly indicated. Earlier studies by Small, shows linksbetween documents which are obtained by Comcitation counts

[32]. Such documents are arranged in a cluster with thehighly co~cited document in the centre of cluster. As newpapers are published, the co-citation links also change.Co~citation analysis can be done by using data collectedfrom databases in computer. It helps to form clusters ofdocuments, authors etc. Another type of study co~citationcontext analysis enables to produce research reviews andSwanson gives the importance of such studies in informationScience [l58].

Clusters of cited documents are groups made evident byco—citation strengths. The strength of relationship isproportional to the number of times the papers are citedtogether. This technique is used in citation indexing inISI databases.

Citation analysis have been applied in variousdisciplines successfully mainly for journal evaluation.

103

Though there are some limitations such as self-citation,multiple authorship, synonyms, merging or splitting ofjournal titles etc to such studies, the advantages aremore. It helps to classify journals according to scientificimportance based on their impact on research, measurescientific influence" and productivity, enables propermanagement of collection in libraries, informationretrieval in computerised services etc. Citation studiesare becoming more easier and relevant by using onlinedatabases

Current Trends

Informetric techniques are used in the identification

of trends in a subject and also in the study of scientificcommunication. Informetrics is accepted as a research areathat gives additional understanding of the form andstructure of scientific communication. Some of the uses are

for studying

l) growth of a subject and its literature2) evaluate quality of research of an individual, an

organisation or a country

3) productivity of an individual scientist4) study history of science and science policy5) obsolescence or aging of scientific literature6) evaluating individual journals or a group of journals

especially for acquisition decisions.

104

7) identify users of different subjects8) authorship trend of scientific literature9) measure usefulness of SDI services

l0) develop experimental models correlating or bypassingexisting ones.

Of the informetric distributions, the most used isBradford's law and its various refinements. It is used toassist in designing information systems, rationalizinglibrary services and in making more economic and fruitfuluse of periodicals. Brookes gave the use of Bradford's lawin ­l) computerised bibliographic search systems (Medlars)2) management of special libraries3) discarding of ‘aged’ periodicals4) planning special library systems5) subject bibliography compilation without difficulty.

The two characteristics of Bradford s law are itsuniversality and stability. The universality states thatthe law holds for all subjects at all times. The stabilitymeans violation of conditions for the validity of the lawsuch as narrow time span of bibliography or narrow subjectdefinition is not affecting the law significantly.

Research on informetric techniques applied to variousdisciplines have been done which has led to an increase innumber of papers on the topic. Most of the studies wereconducted to evaluate the scientific publications,

105

scientists, research organizations etc. Many reports oninformetric laws and its applications are also found.Citations are observed to be valid measures of quality inmost of the studies though some show citations are not truemeasure of quality. In such cases combination of othermeasures such as expert opinion, can also be taken into

­

account for forming conclusions.

Recently informetric measures of research productivityand quality are applied by decision makers to enable properutilization of available resources. Quantitative methods inscience and technology assist in determining fundingpriorities for research in the present context of economicrecessions. Quantitative studies using informetric methodsare also taken as science policy indicators in manycountries of Europe. Such studies are published in journalslike ‘Research policy‘ ‘Scientometrics‘ and otherpublications [l59]. The use of informetric methods inscience policy decision—making has generated the same range

and level of criticism as in the use of citation analysis.But these studies can be considered as one of the manysources of information for science policy indicators.

Applications of informetric methods has beennecessitated by the information explosion. By thedevelopment of computer databases most of the studies arenow done by gathering data online. Stefaniak gives review

106

on works done in U.S. and Europe [82]. A variety ofmeasures for downloading data are available on "TheBibliometrics Toolbox" created by Brooks [l60].

As applied to library and information science,informetric analyses have also been used for book andperiodical acquisitions, library use studies, weeding ofobsolete ‘materials etc. Now these methods are applied toevaluate the structure of literature in variousdisciplines. Thus the quantitative methods which weretermed ‘Bibliometrics' to denote ‘written communication‘

have progressed to information science which makes the term

‘informetrics' the most appropriate one in the presentcontext. Informetrics has thus become very significant andeffective tool in research for quantitative and qualitativemeasurement of human knowledge.

References

Hulme,E.W. 1923. gtatistioal bibliography in_ relation _tothe_ qrowth pof modern giviligation. Lectures, May 1922.Grafton, London, p.9.

-. Gosnell,C.F. l943. The rate pf Wobsolescence in_ Qollegglibraryh book_ collection. Ph.D dissertation, New YorkUniversity. p.16.

Raisig,L.M. 1962. Statistical bibliography in healthsciences. Bulletin Medical Libraries Association 5Q: 450.

107

Websters gNew WorldDictionary of American_ Languagg 1964.

The World Publishing Company, New York, p.144.

Ranganathan,S.R. 1969. Librametry and its scope. DRTCSeminar 7, Paper DA.

Pritchard,A. 1969. Statistical bibliography or biblio~metrics” Qourgal ofinpocumentation, Z2 (4) p.349.

p

Fairthorne,R.A. 1969. Empirical hyperbolic distributions.(Bradford—zipf—Mandelbrot) for Bibliometric description and

prediction. dournal oi Documentation, Q6: 319.

Pritchard,A 1970. Computers, bibliometrics and abstracting

services. Research in Librarianship. lg: 94"99.

British Standard Institution, 1916 gritishp gStandardglossary? gof: Qocumentationg terms. Prepared under thedirection of Documentation standards committee. p.7.

Potter,W.G. 1981. Introduction gibrary Trends, Q9: p.5.

Sengupta,I.N. 1985. Bibliometrics, a bird‘s eyeview. IASLIC

Bulletin, QQ: 167-174.

Hertzel, Dorothy,H. 1987. Bibliometrics, history of thedevelopment of ideas in. in: Enczglopaedia of Qibrarvw andInformation Science Q2: l44~2l9.

Nicholas,D and Ritchie,M. 1978. Literature wand biblio;metrics. Clive Bingley, London. p.9—ll.

108

Brookes,B.C. 1990. Biblio-, Sciento—, Infor—metrics°?°

What are we talking about. lg: Egghe, L and Rousseau, R,Ed. Informetrics_1989/90, Elsevier, Amsterdam. p. 31-43.

Tague,J. 1992. An introduction to informetrics. InformationRrqcessingtandyflanagement. Z5 (1): 1-3.

Ravichandra Rao,I.K. and Neelameghan, A. 1992. FromLibrametry to informetrics. An overview and Ranganathan‘s

contributions. Lipri, 42: 242-257.

Lancaster,F.W. 1991. Bibliometric_ method_ in’, assessingproductivity, and, impact Tof researcn. Sarada RanganathanEndowment for Library Science, Bangalore. 52 p.

Cole,E.J. and Eales,N.B. 1917. History of comparativeanatomy. Pt. I. Statistical analysis of the literature.yScience Progress. p. 578-596.

Lotka,A.J. 1926. Frequency distribution of scientificproductivity. qournal of_Washington Academy of Science, 16:317-325.

Gross,P.L.K. and Gross,E.M. 1927. College libraries andchemical education. Science, 66: 385-389.

Bradford,S.C. 1934 Sources of information on specificsubjects. Engineering, 137: 85-86.

Bradford,S.C. 1948. DocumentatiQn. Crosby Lockwood, London.

p. 85.

109

Zipf,G.K. 1935. Ehe épsypho-biologyof ylanguagez anintroduction todynamic philology. Cambridge, MIT Press,Mass.

Fuss1er,H.H. 1949. Characteristics of the researchliterature used by Chemists and Physicists in the UnitedStates. LibraryiQuarterly. pt I: 19-35 and pt II: ll9—143.

Garfield,E. 1955. Citation indexes for science. Science,122: 108-111.

Garfield,E. 1976. Significant journals of science. Nature,264: 609-615.

Narin,F. 1976. Evaluative bibliometrics. Computer Horizons.Inc. Cherry Hill, N.J. p. 41-44.

Kessler,M.M. 1963. An experimental study of bibliographiccoupling between technical papers. IEEEg?Transactions, p.49.

Kess1er,M.M. 1963a. Bibliographic coupling betweenscientific papers. gmerican Documentation, 14: 10-25.

»

Weinberg,B.H. 1974. Bibliographic coupling~a review.information Storage Retriegal, 10: 190.

Goffman,W and Newi11,V.A. 1964. A generalization of thetheory of epidemics—an application to the transmission ofideas. Nature, 204: 225-228.

110

Sma1l,H. 1973. Co-citation in the scientific literature: Anew measure of the relationship between two documents.' ieJourna1_;of American §og_g§y for Information Science, 24:256-269.

Vickery,B.C. 1948. Bradford's law of scattering. Journalojépocumentation. 4: 198-203.

Kenda1l,M.G. 1960. The bibliography of operationalresearch. Qperational Research Quarterly. 11: 31-36.

Leimkuhler,F.F. 1967. The Bradford distribution. Journal

of Documentation, 23: 197-207.

Brookes,B.C. 1968. The derivation and 'application ofBradford — Zipf distribution. Journal of Documentation. 24:247-265.

Yablonsky,A.I. 1980. On the fundamental regularities of thedistribution of scientific productivity. Scientometrics 2­__ _,_ __ 4. 7 '3-34.

Haitun,S.D. 1982. Stationery scientometric distributions.Pt. I-III Scientometrigsgiz 5-25, 89-104, 181-194.

Bookstein,A. 1976. Bibliometric distributions. LibraryQuarterly, 46: 416-423.

Bookstein,A. 1990. Informetric distributions pt. I. Unifiedoverview. Journal of g§mericanyM§ociety pfor_ Information

gcience, 41: 368-75.

lllBookstein,A. 1991. Theoretical properties of informatricdistributions. lg; Ravichandra Rao, Ed. Informetrics 91 p.l7-35.

Egghe,L. 1988. On the classification of classicalbibliometric laws. Journal of Documentation. Q5: 53-62.

Rousseau,R. 1990. Relation between continuous versions of

bibliometrics laws. Jpurnal of “American _Society _forInformation Science, 51: 68~75.

Pratt,A.D. 1977. A measure of class concentration inbibliometrics. Jqurnal_of American_Societ¥for Information§cience,gg8: 285"292.

Drott,M.C. Mancall,J.C. and Griffith,B.C. 1979. Bradfordslaw and libraries. Present applications — Potentialpromises. §§LIB Proceedings Q1: 296-304.

Lawani,S.M. 1973. Bradford's law and the literature ‘ofAgriculture. lnternational Library Reyiey, 5. 34l~50.

Goffman,W and Morris,T.G. 1970. Bradford‘s law and library

acquisitions. Nature, 221: l205—l207.

Brookes,B.C. 1989. Towards informetrics: Haitun, La place,

Zipf, Bradford and the alvey programme. Qournal ofDocumentation, 40: 120-143.

Egghe.L. 1986. Methodological aspects of bibliometrics.Library Science, 25: 179-191.

112

Brookes,B.C. 1977. Theory of Bradford's law. Qournal _ofQocumentation, 33: 180-209.

Price,D.J.S. 1976. A general theory of bibliometrics andother Cumulative Advantage Process. Journal of AmericanSqeiety for Information Scienge, 21: 292-306.

Bookstein,A. 1979. Explanation of bibliometric laws.Collegtion Management, 3: 151-162.

Coile,R.C. 1977. Lotka‘s frequency distribution ofScientific productivity. Journal ofiAmerican Society torInformatiQnScienqe, 28: 366.

Hubert,J.J. 1977. Bibliometric models- for journalproductivity. Social Indicators Research. 4: 441-473.

Sengupta,I.N. 1984. Weightage formula to rerank biochemical

periodicals. International Library Review. 16: 447-458.

Naranan,S. 1971. Power law relations in sciencebibliography - a self consistant interpretation. JournalQiQocumentation. 21: 83-97.

O'Neill,E.T. 1973. Limitations of the Bradford‘sdistribution system ~ their benefits. Proceedin9s_ Ame leanr1_

Sogietyrformlnjprmation science. 2Q: 177-178.

Schorr,A.E. 1974. Lotka‘s law and library science. ResearchiQuarterly. 12: 32-33.

113

Voos,H 1974 Lotka and information science, Journa1 TofAmericanSQciety forlnformation Science. 22: 270-272.

Subramanyam,K. 1981. Lotka's law and library literature.

Library Research. 8: 167.

Wi1kinson,E.A. 1972. Ambiguity of Bradford's law.Journa1 of

pocumentation. 28: 122-130.

Worthen,D.B. 1973. Epidemic process and the contagionmodel. Journal of American Societyfor Information Science,

24: 343-346.

V1achy,J. 1978. Frequency distribution 'of scientificperformance: bibliography of Lotka's law and relatedphenomena,Scientometrics 1: 109-130.

Ravichandra Rao,I.K. 1980. Distribution of Scientificproductivity and social change. Journalof American Societyfor Information Science 31: 111-121.

Tague,J. 1981. Success breeds success phenomena andbibliometric processes. §ourna1MofAmerican Society? forInformation Science, 32: 280-86.

Tague,J. 1991. How good is a model? Explicativity as aGoodness~of—fit measure. lg; Ravichandra Rao, I.K. Ed:

Informatrics 91: p. 484.

114

Broadus,R.N. 1987. Early approaches to Bibliometrics.Qournal jof American §ociety for Information _Science, Q8:127-129.

Bensman,S.J. 1982. Bibliometrics laws and library usage as

a social phenomena. gibrary Research, 4: 279“312.

Lawani,S.M. 1981. Bibliometrics: its theoreticalfoundations, methods and applications. Libri, Q: 294-315.

I

Ravichandra Rao,I.K. 1983. Quantitative methods for libraryand information science. Wiley Eastern, New Delhi, 271 p.

Ikpaahindi,L. 1985. An overview of bibliometrics itsmeasurements, laws and their applications. Libri, Q5: 163—177.

Zunde,F. 1984. Empirical laws and theories of information

and software sciences. Information Processing & Management,

22; 5-18.

White,H.D. and Mccain.K.W. 1989. Bibliometrics. Annual

ReviewofyInformationScience & Technology. Z2: 119-186.

Pritchard,A and Wittig,G.R. 1981. §ibliometriqs+abibliography and_inde§. Vol 1: 1874-1959.

Hjerppe,R. 1980. A__pjb1io9raphy poff bibliometrics__yan§

citation Z_indexingp and, analysis. Royal institute ofTechnology Library, Stockholm (TRITA—Lib»2013).

115

Schubert,A. 1983. Quantitative studies of science: Acurrent bibliography. (Pnf;ars serially every year from1983 onwards) Scientometrics.

Sellen, Mary,K. 1993. Bibliometrics — an annotatedbibliography 1970-1990. G.K. Hall, New York. 169 p.

Oppenheim,C. 1985. Use of online databases in bibliometrics

studies. £3: Qroceedings f9th _[nternationa1yon1ineinformation meeting. Learned Information Ltd. Oxford. 355­404.

Persson,O. 1986. Online bibliometrics - a research tool forO

everyman. §cientometrics 19: 69-75.

Hibbs,J.E., Bobner,R.R., Newman,I., Dye,C.M., and Benz,C.R.

1984. How to use online databases to perform trend analysisin research. Online, 8: 59-64.

Lancaster,F.W. and Lee,J.L. 1985. Bibliometric techniques

applied to issues management, a case study. Qournaly ofgmericangociety for lnformationyScience,36: 389-397.

Stefaniak,B. 1987. Use of bibliographic databases for

scientific studies. gcientometricsylgz 149-161.

Murphy, Larry,J. 1973. Lotka‘s law in the humanities.

Journal Qfgmerican Societyforglnformationficience) 24:461-62.

116

Coi1e,R.C. 1976. Bibliometric studies of scientificproductivity. Progeeding;yU of”? Amerian _ Society N forInformation Science, 13: 20.

Pao,M.L. 1982. Lotka's test. Collection Management 5: 112.

Nicho1ls,P.T. 1987. TnepLot§ayQypothesis and _bib1iometricmetnqdology. Pn.D. Dissertation, University of WesternOntario‘, Canada 175 p.

Pao,M.L. 1986. An empirical examination of Lotka's law.

Journaly of Ame{icanTSociety_for Information yficience. 21:26-33.

Price,D.J.S 1963. gittle Science, big? science, ColumbiaUniversity Press, New York 118 p.

Nicho11s,P.T. 1988. Price's square root law: empiricalvalidity and relation to Lotka‘s law. informationprocessing andJManagement, 2Q: 469~77.

Bookstein,A. 1990. Informetric distributions pt. II.Resilence to ambiguity. Qournal of American_ Societyy forInformation Science 41: 376-86.

Lawani,S.M. 1980. Quality, Collaboration andycitationsy incanceri;researcn. Ph.D. thesis, Florida State UniversityPress.

Cole,P.F. 1962. A new look at reference scattering. Journal

of Documentation, 18: 58.

117

Brookes,B.C. 1981. A critical commentary on Leimkuhler‘s

exact formulation of Bradford's law, Qgurnal gofQocumentation. 87: 77.

Naranan,S. 1970. Bradford's law of bibliography of science,

an interpretation. Qature, 227: 1377.

Groos,V. 1967. Bradford’s law and the Keenan—Atherton data.American Documentation 18: 46.

O‘ Neill,E.T. 1970. Journal usage? patterns_ and gtheirimplications gin the planning of glibraryp systems. Ph.D.Dissertation. Purdue University.

Fasler,A. 1970. Exceptions to Bradford's law(Correspondence) §aturg, Q27: 101.

Pope,A. 1975. Bradford's law and the periodical literature

of information science. Qournal of American_ Society _forIn§ormation Science, 2§: 212.

Asai,I. 1981. A general formulation of Bradford‘sdistribution the graph oriented approach. Journal gofAmerican $ociety_j9r Infiormation Science, Q2: 113-119.

Chen,Y.S. and Leimkuhler,F.F. 1987. Bradford s law, an

index approach, Scientometrics, ll: 183-198.

Drott,M.C. and Griffith,B.C. 1978. An empirical examination

of Bradford s law and the scattering scientific literature.

102

103.

4

105

106

107.

108

109.

118

Journal yof American Society for Injormagion Science, 29:238—246.

Karmeshu,N”, Lind.C and Cano,V 1984. Rationale forBradford's law. §cientometrics, Q: 233—24l.

Ravichandra Rao I.K. 1981 Documents and user distribution_ I I __p__ _T__,_ _,___ __ __ ,____ _-_pin transaction records of Canadian University Libraries.Ph.D. Thesis. University of Western Ontario, Canada.

Garfield,E 1980. Bradford's law and related statisticalpatterns, §ur£ent,§omments, May 12. Reprinted in: §ssay§ ofanflnfgrmation ficientist, i: 476-483.

Maia,M.J.F. and Maia, M.D. 1984. On the unity of Bradford's

law. Journal of Documentation, 40: 206"216.

O’ Connor,D.O. and Voos,H. 1981. Empirical laws, theory

construction and bibliometrics. Library Trends, Q0: 9-20.

Drott,M.C. 1981. Bradford's law: theory empiricism and the

gaps between, Library Trends, Q0: 41-52.

Burrell,Q.L. 1988. Modelling the Bradford phenomenonQournalojzgocumentation, §§: 1-18.

Burrell,Q.L. 1992. The dynamic nature of bibliometricprocesses: a case study, Ln: Ravichandra Rao,I.K. Ed:lnformetrics — 91. p. 97-129.

110

111

112

113.

114

115

116

117

119

Basu, Aparna. 1992. on the theoretical foundations ofBradford's law. Ln: Ravichandra Rao,I.K. Ed. Informetrics :91 p. 51-67.

Zipf,G.K. 1949. Human behayiour and the principle of leas_ __ 1 . _ . --- ._~-_ .-__effort. An introduction to Human ecology, Addision - Wesley

Press, Reading, Mass p. 24.

Wyl1ys,R.E. 1981. Empirical and theoretical basis ofZipf‘s law, Library Trends, QQ: 53.

Simon,H.A. 1955. On a class of skew distribution functions.

Biometrica, Q2: 425.

Wyllys,R.E. 1975. Measuring scientific prose with rankfrequency ('Zipf') curves, A new use for an old phenomenon.

Information Revolution, 12: 30.

Fedorovicz,J. 1982. A Zipfian model of an AutomaticBibliographic system. An application of Medline. Journal of

Americangociety for Information Science, 98: 223-232.

Federovicz,J. 1982a. The theoretical foundations of Zipf‘slaw and its applications to the bibliographic databaseenviornment. Journalpof American Societ¥_pfor%_1nformation

Science, Q3: 285-293.

Tague,J and Nichol1s,P. 1987. The maximal value of a Zipfsize variable. Sampling properties and relationship toother parameters. Information Processing;& Management, 21:155-170.

118

11

120.

121.

22.

23.

24.

25.

9.

120

Brookes,B.C. 1988. Comments on the scope of bibliometrics.

IQ: Egghe & Rousseau Ed: In§ormetrics;§7[§8, Elsiver p. 29­41.

Chen,Y.S. and Leimkuhler,F.F. 1985. A relationship betweenLotka's law, Bradford's law and Zipf's law. Journal_ ofAmsriqanéocistxliorIn€¢rmati9n-9¢i¢n9s- 3Z= 397-314­

Weinstock)M. 1971. Citation Indexes. lg: Encyclopaedia ofLibrary and Information Science. V. 5, Marcel Dekker,New York p. 15-40.

@arfie1d-E- 1979- Qitayiqniilaqsrinq - iE§_i;he9§r__§9Q2221192§i92§._ini§§isnseJ_;eshn9lQgy and h9m§nitis§- Wiley

New York, 274 p.

Barker-D-L- 1966- Qnarasssrisiis§_9£__r@§@§r2h_rlitsrstyns

used; by_chepists. Ph.D dissertation, University ofIllinois.

Oppenheim,C and Renn, Susan,P. 1978. Highly cited oldpapers and the reasons why they continue to be cited.l92rln§li ,<>.€_ 51112.1: isssi I9-..<>r_<=.2‘-s 12>: _€92_I.r2§Qr.m@§i_9rr 9iren2e.- .2.9.=

226.

Garfield,E. 1972. Citation analysis as a tool in journaleva1uati@n- 2222222» lZ§= 471 79 Rs2ri9§ed_i2= Eeeezsisfan

In§2rm@ii99i§2i2nP§§E- l= 527-544­

Tagliacozzo,R. 1967. Citations and Citation indexes — Areview. Methods Information Science 6: 136-142__:_________ ______,_ __ Lil I _ _ _ _.__? _ _ _ _,____._ I ___ V

121

Hjerppe,R. 1978. An outline of bibliometrics and citationanalysis. Report? TRITA 1 LI3- 6815. Royal Institute ofTechnology Library, Sweden, 82 p.

Price,D.J.S. 1965. Networks of scientific papers. Science,

132: 510-515.

Mc Cain,K.W. and Bobick,J.E. 1981. Pattern of journal use

in a departmental Library Qournal of American Society £95lnfiormation_ScienQe, Q2; 257-267.

Brown,C.H. 1956. Scientific serials charactristics andlists of most cited publications in Mathematics, Physics,Chemistry, Geology, Physiology, Botany, Zoology andEntomology. Asso§'ation of College andyReference Libraries,lg .. :,*_ V_ V_,. _

Chicago, Monograph No. 16 p. 1-15.

Scales,P.A. 1976. Citation analysis as indicators of theuse of serials, a comparison of ranked title lists producedby citation counting and from use data, Journal yyofDocumentation, 32: l7-25.

Sengupta,I.N. 1986. Three New Parameters in bibliometricresearch and their applications to rerank periodicals inbiochemistry.Scientometrics, 19 235-242.

Sengupta,I.N. 1990. Bibliometrics and identification ofcore periodicals. geraldyofLibrarywscienge, 29: 226-247.

133

134

135

136

137

138

139

140

122

Moed,H.F. 1988. The use of on line databases forbibliometric analysis. lg: Egghe & Rousseau, Ed Infometrics1987/88. Elsevier p. 133-146.

Whitney, Gretchen 1992. Asian authorship in majorinternational bibliographic databases. lg: Ravichandra Rao,I.K. Ed: Informetrics-91. Sarada Ranganathan Endowment for

Library Science, Bangalore p. 501-526.

Garfield,E. 1977. Essaysgof an pinformation scientist. ISIpress, Philadelphia. 4 V.

Pao,M.L. 1985. Lotka's law: a testing procedure:In£ormationgProces§ingManagement, 21: 305-320.

Gupta,D.K. 1989. Lotkas law and its application to author

productivity distribution in psychological literature ofAfrica 1966-1975 Pt II— §erald_of Library §cience, 28: 318­326.

Bookstein,A 1977. Pattern of scientific productivity andsocial change a discussion of Lotka‘s law and bibliometricsymmetry. Journal gofg;American $ociety_Wforg Information

§cience, 28: 418-427.

Egghe,L. 1986. On the 80/20 rule. scientometrics, 10: 55­68.

Sen,B.K. Karanjai,A and Munshi,U.M. 1989. A method ofdetermining the impact factor of a non-SCI journal. Journal

of Documentation, £5: 139-141.

123

Hirst,G. 1978. Disciplinary impact factors a method fordetermining core journals lists. Journal _ofmfl_AmeriganSociety_for_lnfqrmation §cience, 29: 171-172.

He,C and Pao,M.L. 1986. Discipline Specific journalselection algorithm. lnfgrmation_Processing_andgManaqemnet,

£2: 405-416.

Abdu1la,S.B. and Lancaster,F.W. 1991. The contribution ofscientists to the popular literature, their role as expertwitnesses and their influence on their peers : a case studyin the field of acidinin. Scientometrics 20: 55-64.1. 1 1 _ 1 ___'

Vinkler,P. 1990. Bibliometric analysis of publicationactivity of a Scientific Research Institute. Ln: Egghe &Rousseau, Ed: Informetrics89/90. p. 309—33.

Arora, Kamalesh and Sen,B.K. 1991. Use of impact factor as

a valid measure for evaluating the performance ofscientific papers. lg: Ravichandra Rao, I.K. Ed:lnformetrics—91 p. 36-50.

Tomer,C. 1986. A statistical assessment of two measures of

citation ~ the impact factor and the immediacy index.lnformation Processing § Management, 22: 251-258.

Arunachalam,S and Manorama,K 1989. Are citation—based

quantitative techniques adequate for measuring science onthe periphery? Scientometrics, 15: 393~408.

148

149

150

151

152

153

154

124

Sen,B.K. and Kumar,S. 1990. Evaluation of recent scientific

research output by bibliometric method. Paper presented _atthe_ international: conference on _$cience mlndicators_Wfordeveloping countries, Paris, Unesco.

Brookes,B.C. 1970. The growth, utility and obsolescence of

scientific periodical literature. Journal of Documentationg§= 283-284.

Ravichandra Rao,I.K. and Meera,B.M. 1992. Growth and

obsolescence of literature an empirical study. 12:Ravichandra Rao, Ed: Infometrics - 91, p. 377-394.

Griffith,B.C” Servi.D.N. Anker,A.L.,Drott,M.C. 1979. Theaging of scientific literature, a citation analysis.Qournal of Documentation, 35: 179-196.

Price,D.J.S. 1970. Citation measures of hard science, soft

science, technology and non science. lg: Nelson C &Pollock, D. Ed. Qommunication among gscientistsg andengineers. Lexington MA. Heath P. 3~22.

Line,M.B. and Sandison,A. 1974 ‘Obsolescence’ and changes

in the use of literature with time. Qournalp ofDocumentation, 30: 283-350.

Line,M.B. 1987. Relation between productivity and‘Obsolescence’. Journal of American Societyfor InformationScience, 38: 307-308.

125

Christavao,H.T. 1985. The aging of literature of biomedicalsciences in developed and developing countries.Scientometrics, Z: 411-430.

Aversa,E.S. 1985. Citation patterns of highly cited papersand their relationship to literature aging: a study of theworking literature, gcientgmetrigs, Z: 383-389.

Carpenter,M.P. and Narin,F. 1973. Clustering of Scientific

journals. Journal {pf gAmerican:SocietZp for} informationScience, 2Q: 425-36.

Swanson,D.R. 1988. Historical note: Information retrieval

and the future of an illusion.§ournal of American gSocietXI

for Information Science, 39: 92-98. iKing,J. 1987. A review of bibliometric and other scienceindicators and their role in research evaluation. Journal

of Information Science, 13: 261-276.

Brooks, T.A. 1987. The bibliometrics tool box version 2.5Software and documentation available from North CityBibliometrics, Seattle.