lnfonnalion Processlllg .l Management, Vol. 17, pp. 39-50. 1981
`Printed in Great Britain
`
`0306-4S7:l/81/010039-12$02.00/0
`Pergamon Pms Ltd.
`
`THE RELATIONSHIP OF INFORMATION SCIENCE
`TO THE SOCIAL SCIENCES:
`A CO-CITATION ANALYSISt
`
`HENRY SMALL
`Institute for Scientific Information, 3501 Market Stred, University City Science Center, Philadelphia,
`PA 19104. U.S.A.
`
`(Received 3 April 1980)
`
`Abstract-A co-citation cluster analysis of a three year (1975-77) cumulation of the Social
`Sciences Citation Index is described, and clusters of information science documents contained
`in this data-base are identified using a journal subset concentration measure. The internal
`structure of the information science clusters is analyzed in terms of co-citations among
`clusters, and external linkages to fields outside information science are explored. It is shown
`that clusters identified by the journal concentration method also cohere in a natural way
`through cluster co-citation. Conclusions are drawn regarding the relationship of informatiOI}
`science to the social sciences, and suggestions are made on how these data might be used in
`planning an agenda for re~earch in the field.
`
`INTRODUCTION
`One way of gaining insight into the state of a research field or discipline is to examine the
`publications produced by its practitioners. To the extent that practitioners in the field publish
`the results of their investigations, this mode for assessing the state of a field can reflect with
`great specificity the content and problem orientations of the group. Of the many ways that
`publications can be analyzed and counted, perhaps the most revealing kind of data are the
`references cited by the practitioner group in their publications. References to earlier literature
`tell us about the author making them as well as the items being cited. When references are
`cumulated over a significant volume of source literature such as in the Science Citation Index
`(SCn®and Social Sciences Citation Index (SSCJ)®, the collective patterns reveal the concerns
`of the field as symbolized by the documents and authors cited. This is how earlier cited
`literature can inform us on the current conceptual framework: the act of citing involves an
`association of a notion or idea expressed in the text with a cited document[l]. Hence,
`each reference is connected to a concept. The cumulative pattern of such contexts provides a
`representation of the cognitive structure of the research field.
`The objective of a study currently underway at the Institute for Scientific Information is to
`examine the structure and development of the field of information science using the published
`literature of information science as data, and the techniques of citation analysis. We know very
`little about how the field of information science has developed over the past several years. On
`the one hand the field might be viewed from a technological standpoint, the primary ac(cid:173)
`complishments of which are creation of machine readable data bases and retrieval systems.
`From another perspective, however, the field of information science can be seen as an
`investigation into the nature of information, the theoretical basis for retrieval, the evaluation of
`retrieval, and the way that human beings use and transmit information. We expect that the
`literature of information science will reflect both the conceptual and technical concerns of the
`field. By using the statistical techniques of citation analysis we hope to get a picture of how the
`field has developed, the main lines of research in the field, its principal foci of interest, and
`where the field appears to be going.
`Some studies have attempted to use published literature to arrive at insights into the
`structure of the field. Saracevic reviewed the first five volumes of the Annual Review of
`Information Science and Technology using bibliometric techniques and concluded that ARIST
`was biased toward the technology and practice of information science and against fundamental
`
`tResearch supported by NSF grant IST-78-16677.
`
`39
`
`EXHIBIT 2034
`Facebook, Inc. et al.
`v.
`Software Rights Archive, LLC
`CASE IPR2013-00479
`
`

`

`40
`
`H. SMALl.
`
`research in the area[2]. In a breakdown of most cited authors in ARIST Saracevic found about
`10% theoreticians, 40--50% experimenters, and the rest developers.
`Donohue has provided the most ambitious bibliometric analysis of the "information
`science" literature, although his choice of journals to include in the database was somewhat
`idiosyncratic (e.g. inclusion of the Journal of the Acoustical Society of America as an
`information science journal)[3]. In general, the main corpus used in Donohue's study represen(cid:173)
`ted information science in its more technical, mathematical and engineering sense, rather than
`information science related to documentation, retrieval and library science. Therefore, the
`structure of "information science" in Donohue's study cannot be taken as a guide to what we
`expect to find in our study. (For example, the list of clusters Donohue obtained using
`bibliographic coupling among documents included groups designated as acoustics, computing,
`cybernetics, engineering, logic, numerical mathematics and statistics).
`Salton has undertaken a citation analysis of individual researchers in information science
`using two comprehensive bibliographies and two points in time a decade apart[4]. His main
`conclusion is that the field has changed significantly over the decade and that this perhaps
`indicates not only intellectual ferment, but perhaps also a lack of focus on key problems. From
`Salton's study we would anticipate a lack of lasting theoretical orientations and a predominance
`of experimental work without the benefit of a theoretical framework.
`A recent survey of concerns in the field of information science by Pratt, based on a
`qualitative analysis of the ARIST series which does not use bibliometric techniques, finds the
`major topics of current research to be [5]:
`( 1) Library problems
`(2) Economics of information
`(3) The nature of information
`(4) Techniques of measurements.
`All of the studies mentioned provide some hints of what we could expect in our citation
`analysis, but certainly none stands out as a definitive benchmark against which to compare our
`results.
`Our study takes two different approaches to the analysis of information science as a field.
`First, we are using a special citation file extracted from the Social Sciences Citation Index
`database, consisting of a core set of journals in information science over a nine year period
`(1969-77), to explore the internal structure of information science and its development. The
`second approach is to show how information science links to the other fields and disciplines in
`the social sciences. To accomplish this we use existing cluster data files at ISI. It is this latter
`work which I report on in the present paper.
`
`CLUSTER ANALYSIS OF THE SSCI
`The starting point for this study was a cluster analysis of a special three year cumulation
`(1975-77) of the Social Sciences Citation Index (SSCI). Earlier we had performed a similar
`analysis of the SSCI for the period 1972-74[6, 7], and the new analysis of the 1975-77 file will
`allow us, eventually, to examine rates of specialty change within the social sciences. The results
`I will report here are only for the 1975-77 file and do not attempt to assess change over time.
`The clustering procedure was identical to that used in the original study. I will review only
`the essentials here. First, all documents in the file cited ten or more times during the three year
`period are selected. Table l presents some statistics on the cluster analysis, and indicates that
`of the over two million cited items in the three year file, about 25,000 were cited ten or more
`times. This is a fairly weak criterion for selection when compared with our usual threshold of
`15 citations per document per year for cluster analyses of annual SCI's[8]. Following selection
`of these highly cited items, all co-citations among the 25,000 were determined, that is, the
`number of times any pair of them is cited together in the three year period. As indicated in the
`Table there were 1.8 million unique pairs of co-cited items thus formed. The raw co-citation
`counts for each pair were normalized by dividing by the sum of citation frequencies for the two
`items minus the number of co-citations. This is essentially the fraction of citations to the two
`items that are co-citations, which is equivalent to the so-called Jaccard coefficient used in
`numerical taxonomy [9]. These coefficients were the basis for the cluster analysis. The cluster(cid:173)
`ing agorithm used (called single-link clustering) [I 0] requires only that we specify a threshold for
`
`

`

`The relationship of information science to the social sciences
`
`41
`
`Table I. Statistics on clusters from 1975-77 Cumulative Social Sciences Citation Index (SSCI)
`
`1. total citations ( 1975-1977)
`
`2. distinct cited items
`
`3,399,058
`
`2,196,127
`
`3. highly cited i terns (~ 10 citations)
`
`24,954 (1.14%)
`
`4. distinct co-cited pairs of cited items
`
`1,846,585
`
`5. distinct co-cited pairs at level 22~
`
`6. clusters at level 22% (~ 2 cited items)
`
`7. mean cited items per cluster
`e. mean citing items per cluster
`
`10,418
`
`2,095
`
`4.1
`
`39.9
`
`the normalized co-citation strength to generate a set of disjoint clusters containing the cited
`items. Setting this threshold at 0.22 generated about 2000 clusters each containing two or more
`cited items. The average cluster size was 4.1 cited items. This set of 2000 clusters in the social
`and behavioral sciences formed the universe from which we selected clusters on information
`science topics. Of course, information science is expected to represent only a small fraction of
`the clusters in this file, which is dominated by fields such as psychology (experimental and
`social), sociology, economics, psychiatry, and so on.
`
`THE SELECTION OF INFORMATION SCIENCE CLUSTERS
`The procedure used to select information science clusters from the 2000 1975-77 SSCI
`clusters was to define a set of information science journals which appear as source journals in
`the SSCI. Fifty journals were selected and are listed in Table 2. This list should not be regarded
`
`Table 2. Information science journal subset
`
`1. American Archivist
`
`2. Annual Review of Information Science and Technology
`
`3. Aslib Proceedings
`
`4. Bulletin of the Copyright Society of the U.S.A.
`
`5. Bulletin of the Medical Library Association
`
`6. Canadian Journal of Information Science
`
`7. Canadian Library Journal
`
`8. College and Research Libraries
`
`9. Drexel Library Quarterly
`
`10. Government Publications Review
`
`11.
`
`IEEE Transactions on Information Theory
`
`12.
`
`IEEE Transactions on Engineering Management
`
`13.
`
`IEEE Transactions on Professional communication
`
`14.
`
`Information and Control
`
`15.
`
`Information Processing & Management
`
`16.
`
`Information Sciences
`
`17.
`
`Information Scientist
`
`18.
`
`International Classification
`
`19.
`
`International Forum on Information and Documentation
`
`20.
`
`International Journal of Computer & Information Sciences
`
`21.
`
`Journal of the American Society for Information Science
`
`22.
`
`Journal of Chemical Information & Computer Sciences
`
`23.
`
`Journal of Documentation
`
`24.
`
`Journal of Education for Librarianship
`
`25.
`
`Journal of Librarianship
`
`

`

`42
`
`H. SMALL
`
`Table 2 (Contd).
`
`26.
`
`Journal of Library Autanation
`
`27.
`
`Journal of Library History Phiosophy ' Comparative Librarianship
`
`28.
`
`Journal of the Patent Office Society
`
`29. Law- Library Journal
`
`30~ Library & Information Science
`
`31. Library Resources & Technical Services
`
`32. Library Trends
`
`33. Library Quaterly
`
`34. Libri
`
`35. Methods of Information in Medicine
`
`36. Nachrichten FUr Dokumentation
`
`37. Nauchno-Teknischeskaya Informatsiya. Seriya l~ Organizatsiya I
`Metodika Informatsionnoi Raboty
`
`38. Nauchno-Teknicheskaya Informatsiya. Seriya 2.
`Protessy I Sistemy
`
`Infor.matsionnye
`
`39. On-line Review
`
`40. Proceedings of the American Society for Information Science
`
`41. Pattern Recognition
`
`42. Program-New of Computers in Libraries
`
`43. Review of Public Data Use
`
`44. Social Science Information
`
`45. Social Studies of Science
`
`46. Special Libraries
`
`47. Unesco Bulletin for Libraries
`
`48. Wilson Library Bulletin
`
`49. Zeitschrift f~r Bibliothekswesen Und Biblioqraphie
`
`50. Zentralhlatt f~r Bibliothkswesen
`
`as a definitive list of journals in the field, but rather as one way of defining the field, which is
`subject to an empirical test later in our analysis. It should be noted, for example, that computer
`science journals were intentionally not included in the list, and that we were slanting the list
`toward the library/information science direction. Some journals included were concerned with
`the mathematical study of communication (information), but the majority deal with the more
`traditional view of information science as an off-shoot of documentation.
`The procedure to select clusters, similar to that used in an earlier study[?], was to calculate
`the fraction of source (citing) papers for each cluster which fall in the specified journal set. A
`distribution of these fractions is obtained (see Fig. 1) which ranges from clusters having 100%
`of their citing papers in journals which are members of the set, to clusters which have none of
`their citing papers in these journals. For purposes of comparison a similar distribution for the
`field of chemistry is included on the same graph {from SCI not SSCI cluster data). The
`chemistry distribution shows clearly a group of disciplinary clusters centered on about 80%
`concentration in the journal sub-set for chemistry, and a smaller interdisciplinary group of
`clusters centered at 55%. The up-swing of the distribution to the left shows all the clusters
`which are not, or only marginally, in the field. The information science distribution shows
`similar disciplinary and interdisciplinary peaks, though on a much smaller scale: there are only
`eleven clusters of the 2000 which have 30% or more of their citing papers in information science
`journals, and only 22 with 10% or more. The latter group is listed in Table 3.
`
`CHARACTERISTICS OF THE IN FORMA TJON SCIENCE CLUSTERS
`Information science by our definition, therefore, comprises at most one percent of the
`clusters in the three year social and behavioral sciences database (about 20 of 2000
`
`

`

`The relationship of information science to the social scieences
`
`43
`
`1000
`
`Number of information science and chemistry clusters
`by percentage citations in disciplinary journal set
`
`~
`E
`::J z
`
`Information science
`
`0-10
`
`11-20
`
`21-30
`
`61-70
`
`Percentage of citations to cluster in disciplinary journal set (by decile)
`Fig. I. Number of information science and chemistry clusters by percentage citations in disciplinary journal
`set.
`
`clusters). Nevertheless, we can examine the clusters identified to see what they can tell us
`about the field. Referring to Table 3, each of the clusters with 10% or more of their citing
`papers in information science journals has been listed in descending order by percentage
`concentration. The order of clusters in Table 3 could be interpreted as the degree of disciplinary
`purity. This ranges from 100% for the "Precis" cluster to 14.3% for "copyright law" and
`"medical use of computers".
`The cluster number in the left column is an arbitrary identification number assigned by the
`computer to each cluster as it is generated. An approximate name was given to each cluster
`based on an examination of the titles of the cited and citing documents. In the last two columns
`the size of the cluster is given both in terms of the number of cited items and the number of
`items citing them.
`Table 4 lists all documents cited ten or more times which comprise the clusters having at
`least 40% of their citations from information science journals. The number of times each
`document was cited in the SSCI from 1975 to 1977 is indicated in the right-hand column. Of
`course, citations to an item can come from any source journal in the SSCI coverage, not just
`those journals listed in Table 2.
`Most of the clusters are very small (the smallest number of cited items a cluster can have is
`two). In addition, a number of clusters have been given the same name (e.g. there are "on-line
`retrieval and data bases" clusters "a" and "b"). This redundancy occurs because it was not
`possible to distinguish the content of these clusters based on the titles of the papers, and they
`were therefore given the same name. This suggests that these areas are being fragmented at this
`level of association (22% normalized co-citation) and we will show in a moment that this is
`indeed the case. Both the small size of the clusters and their fragmentation tell us that
`information science has a weaker structure than areas such as psychology, sociology or
`economics which emerge as larger and more coherent specialties at this level. If the co-citation
`threshold had been lowered, the fragmented information science areas would have congealed,
`but for other subject areas, composite macro-clusters would have formed, indicating too low a
`threshold in their case.
`
`

`

`44
`
`H. SMALL
`
`Table 3. Information science clusters (greater than 10% participation in journal set)
`* Cited
`
`Cluster
`Number
`
`Percent
`Participation
`
`Cluster Name
`
`Items
`
`*Citing
`
`l. 239
`
`2. 1140*
`
`3. 606*
`
`4. 311*
`s.
`
`563*
`
`6. 1999*
`
`7. 777*
`
`s. 1998*
`
`9. 824*
`
`10. 1299*
`
`100.0
`
`Precis
`
`91.7
`
`91.1
`
`88.9
`
`87.5
`
`82.6
`
`81.2
`
`80.0
`
`69.6
`
`47.3
`
`On-line retrieval & data bases (a)
`
`Bradford's law & bibliometrics (a)
`
`On-line retrieval & data bases (b)
`
`>r·heory of indexing & retrieval (a)
`
`Theory of indexing & retrieval (b)
`
`Bradford's law & bibliometrics (b)
`
`Theory of indexing & retrieval (c)
`
`Library serials planninq
`
`Citation analysis (a)
`
`2
`
`7
`
`2
`
`2
`
`2
`
`2
`
`31
`
`24
`
`45
`
`18
`
`16
`
`23
`
`16
`
`50
`
`23
`
`55
`
`17
`
`11. 1300•
`
`12. 101*
`
`l3. 2021*
`
`14. 1239*
`
`1!·. 406
`
`16. 548*
`
`17. 1096
`
`lB. 1844*
`
`19. 2013
`
`20. 896*
`
`21. 1354
`
`22. 1421
`
`41.2
`
`29.4
`
`24.1
`
`20.8
`
`20.3
`
`20.0
`
`20.0
`
`19.2
`
`18.3
`
`18.2
`
`14.3
`
`14.3
`
`Citation analysis (b)
`
`Sociology of Science (a)
`
`Zipf*s law
`
`Sociology of science (b)
`
`Fuzzy systems
`
`Sociology of science {c)
`
`Symbolic interactionisrn
`
`Sociology of science (d)
`
`Information theory
`
`Sociology of science (e)
`
`Copyright law
`
`Medical use of computers
`
`• on single-link network
`
`3
`
`12
`
`2
`
`2
`
`34
`
`29
`
`24
`
`64
`
`20
`
`15
`
`26
`
`GO
`
`55
`
`21
`
`21
`
`We also investigated how the clusters in information science compare with cluster samples
`in other fields in terms of size, age of cited material, and percentage of cited items that are
`books. The results are shown in Table 5 for the comparison fields of sociology, economics,
`psychology and particle physics. The last named field was included as an extreme case of a
`large and fast moving specialty in the physical sciences, obtained of course from the SCI. It is
`clear that information science clusters are very small compared with clusters in these other
`areas. Regarding the recency of literature cited, information science falls roughly between
`sociology and economics, when measured by either the mean publication date of cited items or
`Price's index[ll] (the percentage of items which fall within the last five years). Somewhat
`surprisingly, information science clusters contain relatively few books as cited items, about the
`same percentage as psychology. This indicates the tendency for important contributions in the
`field to appear in journal article form.
`
`THE STRUCTURE OF INFORMATION SCIENCE CLUSTERS
`The next question was how the selected clusters in information science relate to one
`another? To investigate this we make use of residual co-citation linkages-the linkages between
`documents at levels lower than the clustering level. In other words, when clusters are formed at
`level 22% we utilize only the strongest co-citation links among the highly cited documents. But
`there are many weak links below the clustering threshold among documents in different
`clusters, and summing these residual links enables us to measure the strength of inter-cluster
`relationship. When a matrix of these intercluster links is analyzed using the technique of
`multidimensional scaling[l2], a two-dimensional COJ;lfiguration of points (each representing a
`cluster) is obtained in which high co-citation between clusters corresponds to proximity in the
`
`

`

`The relationship of information science to the social sciences
`
`45
`
`Table 4. Highly cited documents in information science clusters (with greater than 40% of their citations in
`information science journals)
`
`Cluster Name
`Cluster
`Number
`Precis
`
`239
`
`Document
`
`times cited
`1975-1977
`
`Derek Austin., "The Development of Precis:
`A Theoretical and Technice.l History, n
`Journal of Documentation 30: 47-101, 1974.
`
`19
`
`19
`
`10
`
`On-line retrieval and data bases
`
`311
`
`A~ Stephanie Barber, Elizabeth D. Barraclough
`and W. Alexand~r Gray, "On-line Information
`Retrieval as a Scientists Tool,n Information
`Storage and Retrieval .2_: 429-1•40, 1973.
`
`Jeffrey Katzer, "The Cost-Performance of an On-line, 12
`Free-Text Bibliographic Retrieval System,"
`Information Storage and Retrieval .2_: 321-329,
`1973.
`
`1140
`
`S'tanley A. Elman, ncost Comparison of Manual and
`On-line Computerized Literature Searching,n
`Special Libraries 66: 12-17, 1975.
`
`Barbara Lawrence~ Ben H. Heil and Margaret H~
`Graham,
`•t f'.Iaking On-line Search Available in an
`Industrial Research Environment," Journal
`of the American Society for Information
`~.?.2_: 364-369, 1974.
`
`Bradford's Law and bibliometrics
`
`606
`
`S.C. Bradford, Documentation (Washington, D.C.:
`Public Affairs Press, 1948).
`
`B. C. Brookes, "The Derivation and Application of
`the Bradford-Zipf Distribution," Journal of
`Documentation 24: 247-259, 1968.
`
`B. C. Brookes, "Numerical Methods of Bibliographic
`Analysis,n Library Trends 22:18-43, 1973.
`
`B.C. Brookes, "Bradford's Law and the Bibliography
`of Science,"~~: 953-956, 1969.
`
`20
`
`10
`
`14
`
`19
`
`10
`
`15
`
`Ferdinand F. Leimkuhler, nThe Bradford Distribution," 17
`Journal of Documentation SJ: 197-207, 1967.
`
`B.C. Vickery, "Bradford's Law of Scattering,"
`Journal of Documentation i: 198-203, 1948.
`
`16
`
`Elizabeth ;\. Wilkinson,."The Ambiguity of Bradf'ord 1 s
`Lav," Journal of Documentation 28: 122-129, 1972.
`
`11
`
`777
`
`~~rk P. Carpenter and Francis Narin, nClustering
`of Scientific Journals/' Journal of the American 12
`Society for Information Science 24:425-436, 1973.
`
`Francis Harin, :!ark Carpenter and r;ancy 3erlt,
`"Interrelationships of Scientific Journals,."
`Journal of the American Society for Information
`~ gJ: 323-331, 1972.
`
`12
`
`Theory of indexing and retrieval
`
`563
`
`Abraham Bookstein and Don R. Swanson,. "Probabilistic 11
`~dels for Automatic Indexing," Journal of the
`American Society for Infonnation Science g_z_:
`312-318. 1974.
`
`M.E. ~ron and J.L. Kuhns, "On Relevance,
`Probabilistic Indexing and Information
`Retrieval," Journal of the Association for
`Computing Machinery 7: 216-242, 1960.
`
`1998
`
`Gerard Salton, Automatic Information Organization
`and Retrieval (New York: McGraw-Hill, 1968).
`
`Gerard Salton~ ed. The Smart Retrieval System:
`Ex eriments in Automatic Document Processin
`Englewood Cliffs, New Jersey: Prentice-Hall,
`1971).
`
`10
`
`41
`
`23
`
`

`

`46
`
`H. SMALL
`
`Table 4 (contd)
`
`Cluster Name
`Cluster
`Number
`
`1999
`
`Document
`
`Times Cited
`1975-1977
`
`Cerard Salton and M.E. Lesk, "Computer Evaluation
`of Indexing and Text Processing," Journal of
`the Association for Computing Mach~5-;-
`B-36, 1968.
`
`Gerard Salton and C.S. Yang, "On the Specification
`of Term Values in Automatic Indexing~" Journal
`of Documentation .s:!_: 351-372, 1973. - - -
`
`Karen ~~parck•Jones ,"A Statistical Interpretation
`of Term Specificity and its Application in
`Retrieval," Journal of Documentation 28:
`ll-21, 1972.
`-
`
`10
`
`12
`
`10
`
`Library serials planning
`
`824
`
`Chinr:-Chih Chen, "The Use Patterns of Physics
`Journals in a Large Academic Research
`Library," Journal of the American Society
`for Information Science 23: 254-264, 1972.
`
`Alexander Sandison, "Densities of Use, and Absence
`of Obsolescence, in Physics Journals at MIT,"
`Journal of the American Society for Information
`Science ~: 172-178, 1974.
`
`18
`
`Citation analysis
`
`1299
`
`E. Garfield, I.H. Sher and R.J. Torpie, The Use
`of Citation Data for Wri tin
`the Hist_o __
`of Science Philadelphia: Institute for Scientific
`Information, 1964).
`
`10
`
`B.C. Griffith, H.G. Small, J.A. Stonehill, and
`S. Dey, "The Structure of Scientific Literatures
`II: Toward a Macro- and Microstructure for
`Science," Science Studies~: 339-365, 1974.
`
`18
`
`H. Small, "Co-citation in the Scientific Literature: 18
`A New Measure of the Relationship between Two
`Documents," Journal of the American Society for
`Information Science 24: 265-269, 1973.
`
`H. Small and B. C. Griffith, "The Structure of
`Scientific Literatures I: Identifying and
`Graphing Specialties," Science Studies 4:
`17-40, 1974.
`-
`
`1300
`
`E. Garfield, "Citation Indexes for Science,"
`Science~: 108-lll, 1955.
`
`M. Weinstock, "Citation Indexes," in: Encyclopedia
`of Library and Information Science (New York:
`Marcel Dekker, 1971) pp.l6-4o.
`
`28
`
`10
`
`12
`
`configuration. Figure 2 shows the results of the scaling for the clusters having 10% or more
`journal concentration (clusters 1096, 1354 and 1421 had no links to other areas and were
`excluded from the analysis). Each point (cluster) has ben labeled by the percentage concen(cid:173)
`tration and all non-zero inter-cluster linkages have been drawn as lines connecting the points.
`This analysis clearly shows that the redundant cluster names are the result of a frag(cid:173)
`mentation of larger clusters which appear on this map as closely grouped points. The more
`"pure" information science clusters fall on the right and the lower middle portion of the map.
`These include "Precis", "theory of indexing", "on-line retrieval", "Bradford's law", and "serials
`planning". The less pure clusters are located toward the upper left and include a grouping of
`"sociology of science" clusters, "citation analysis", "Zipf's law", "information theory", and
`"fuzzy systems". The last two are the most mathematical of the clusters in the set. "Citation
`analysis" and "Bradford's law" appear to play mediating roles between the more pure
`information science clusters and the sociologically oriented areas.
`Another interesting question concerning these selected clusters is the extent of their
`linkages to areas outside the information science set. This should roughly parallel their per cent
`concentration in the journal set since the latter presumably measures the cluster's degree of
`inten.lisciplinarity. Table 6 shows the ratio of linkages a cluster or cluster group has to other
`clusters within the information science set or outside that set. For example, the group of three
`
`

`

`The relationship of information science to he social sciences
`
`47
`
`fuzzy systems
`20.3
`
`19. 2
`
`information
`theory
`18.3
`
`Precis
`100
`
`Fig. 2. Multidimensional scaling of information science clusters (clusters with more than 10% citations in
`information science journals).
`
`clusters on "theory of indexing and retrieval" has six of its eleven linkages to clusters within
`the information science set, and five linkages to clusters outside the set. Similarly, "on-line
`retrival" and "Bradford's law" have a majority of their links to clusters in the set. Hence, these
`are inward looking clusters, By contrast, "fuzzy systems", "information theory", and "citation
`analysis" are outward looking clusters having a majority of their links to clusters outside the set.
`As expected, these in/out ratios are closely correlated with the percentage concentration of
`articles in the journal set. These results strongly suggest that information science is not
`entirely separate from the fabric of the social sciences as a whole, and that it might be
`interesting to systematically explore the linkages which lead outside the set of selected
`information science clusters.
`
`Table 5. Comparison of information science clusters with clusters in other fields
`i
`
`Information
`Science
`
`Sociology
`
`Economics
`
`Particle
`Psychology I Physics
`
`Number of clusters
`
`Mean size (cited items)
`
`Mean size (citing items)
`
`Mean year of publications of cited
`items in clusters
`
`13
`
`2o8
`
`29o 3
`
`1967o9
`
`Percentage of cited items that are books
`
`l3o 9
`
`36
`
`5o5
`
`69ol
`
`1966 0 6
`
`39o 0
`
`Percentage of cited items published
`in last 5 years (Price's Index)
`
`47 o2
`
`46o5
`
`44
`
`50 3
`
`65o9
`
`277
`
`9ol
`
`96o8
`
`19
`
`l2o4
`
`153o8
`
`1968o4
`
`1968 0 8
`
`1972o4
`
`24o5
`
`52o8
`
`l4o 5
`
`Oo9
`
`57 o9
`
`72o9
`
`

`

`48
`
`H. SMALL
`
`THENETWORKOFCLUSTERS AROUND INFORMATION SCIENCE
`Up to this point we have examined the information science clusters identified by journal
`concentration in isolation from all other clusters in the social sciences. This isolation is of
`course not the case. To explore these external connections we use the full index of inter-cluster
`linkages, specify a starting cluster and an inter-cluster link threshold. The starting cluster was
`No. 563 "theory of indexing and retrieval," but it is important to point out that the final network
`obtained is independent of starting point. All linkages from the starting cluster to any other
`clusters in the file which had summed normalized co-citations of one or more were followed.
`These clusters were in turn treated as starting clusters, and their strong links followed. The
`resulting network is shown in Fig. 3, where clusters are represented as circles containing a
`cluster identification number and the lines connecting them represent summed co-citation links
`of one or more.
`The first finding was that we eventually reached a number of points where the number of
`linkages leading to other clusters became very large and, in effect, exploded into other larger
`networks. These are indicated on the Figure by short lines emanating from the cluster. Upon
`inspection of the document titles in these clusters, it was evident that one set of clusters was
`branching off into social psychology while another set was leading into sociology. Links from
`all the other clusters on the network were exhaustively followed and many of these clusters
`turned out to be identical to ones identified by the independent measure of journal concen(cid:173)
`tration. Jn fad 16 of the 22 clusters selected by having 10% of their citing articles in the
`
`on-line document retrieval
`and data bases
`
`bibliometrics 824 library serials
`planning
`
`1 s Laws
`
`of soci aJ.
`psychology
`
`PSYCHOLOGY
`
`+J organizations
`
`in academia
`
`SOCIJLOGY
`
`Fig. 3. Network of clusters around information science in the 1975-77 SSCI (network constructed at
`threshold one of the summed co-citation links between clusters).
`
`

`

`The relationship of information science to the social sciences
`
`49
`
`Table 6. In/out ratios for selected clusters and cluster groups
`
`Cluster or Cluster GrOUE;( cluster numbers)
`
`l. on-line document retrieval
`and data bases ( 311,1140)
`
`2. Bradford's lav and
`bibliometrics (606,777)
`
`3. theory of indexing and
`retrieval ( 563,1995,1999)
`
`4. citation analysis (1299,1300)
`
`5. information theory (2013)
`
`6. fuzzy systems (406)
`
`Ratio
`In
`
`Out
`
`5/6
`(. 83)
`
`l/6
`( .n)
`
`9/13
`(.69)
`
`4/13
`(. 31)
`
`6/11
`(.55)
`
`5/11
`(. 45)
`
`13/42
`(. 31)
`
`29/42
`(. 69)
`
`5/33
`( .15)
`
`28/33
`(. 85)
`
`2/22
`(. 09)
`
`20/22
`(. 91)
`
`information science journal set, appear on the network. (Their cluster numbers are starred on
`both Table 3 and Fig. 3). Furthermore, all clusters on the network which were selected by
`journal concentration are contiguous on the network, i.e. are not separated from one another by
`non-selected clusters. This result corroborates our selection of the journals for the information
`science set in the sense that the literature appearing in these journals behaves as a coherent
`whole.
`Information science, as represented by this network, appears poised somewhere between
`psychology and sociology, with a very strong link to sociology via the sociology of science, and
`a more tenuous link to psychology through a cluster called "creativity and achievement". At the
`same time, information science, at least in the context of the social and behavioral sciences,