`of the Relationship Between Two Documents
`
`A new form of document coupling called co-citation is
`defined as the frequency with which two documents are
`cited together. The co-citation frequency of two sci(cid:173)
`entific papers can be determined by comparing lists of
`citing documents in the Science Citation Index and
`counting identical entries. Networks of co-cited papers
`can be generated for specific scientific specialties, and
`an example is drawn from the literature of particle
`
`physics. Co-citation patterns are found to differ sig·
`nificantly from bibliographic coupling patterns, but to
`agree generally with patterns of direct citation. Clusters
`of co-cited papers provide a new way to study the spe(cid:173)
`cialty structure of science. They may provide a new
`approach to indexing and to the creation of SDI
`profiles.
`
`HENRY SJMLL
`
`Institute for Scientific In/ormation
`Philadelphia, Pennsylvania 19106
`
`• Introduction
`
`Bibliographic citations in scientific papers have been
`used by a variety of researchers to establish relationships
`among documents. Both direct citation-the citing of
`an earlier document by a new document-and biblio(cid:173)
`graphic coupling-the sharing of one or more references
`by two document-s-have received considerable attention
`(1, 2, S, 4). A related measure, which has been over(cid:173)
`looked in earlier discussions, is co-citation. Unlike biblio(cid:173)
`graphic coupling which links source documents, co-cita(cid:173)
`tion links cited documents and is, therefore, analogous to
`a measure of descriptor or word association (5). The
`purpose of this paper is to define this new kind of
`"coupling" and to distinguish it from bibliographic
`coupling, using a.n actual example from the literature of
`physics. Co-citation patterns are found to differ signifi(cid:173)
`cantly from bibliographic coupling patterns, but to agree
`generally with patterns of direct citation.
`
`• Measuring Co-citation
`
`The strength of co-citation between two cited papers
`or books can be easily determined from a citation index
`such as the Science Citation Index (SCI). Each of the
`
`two papers is located in the Citation Inde:~: section of
`the SCI, and their lists of citing papers a.re scanned.
`The number of identical citing items defines the strength
`of co-citation between the two cited papers. An identical
`citing item is simply a. new document which has cited
`both earlier papers; therefore, co-citation is the fre(cid:173)
`quency with which two items of earlier literature are
`cited together by the later literature (6).
`To be strongly co-cited, a large number of authors
`must cite the two earlier works. Therefore, co-citation is
`a relationship which is established by the citing authors.
`In measuring co-citation strength, we measure the degree
`of relationship or association between papers as perceived
`by the population of citing authors. Furthermore, be(cid:173)
`cause of this dependence on the citing authors, these
`patterns can change over time, just as vocabulary co(cid:173)
`occurrences can change as subject fields evolve. Biblio(cid:173)
`graphic coupling, on the other hand, is a fixed and per(cid:173)
`manent relationship because it depends on references
`contained in the coupled documents. Co-citation pat(cid:173)
`terns change as the interests and intellectual patterns of
`the field change.
`When two papers are frequently co-cited, they are
`also necessarily frequently cited individually as well.
`If it can be assumed that frequently cited papers repre(cid:173)
`sent the key concepts, methods, or experiments in a
`EXHIBIT 2024
`Facebook, Inc. et al.
`v.
`Software Rights Archive, LLC
`CASE IPR2013-00479
`
`Journal of the American Society for Information E
`
`
`
`field (7), then co-citation patterns can be used to map
`out in great detail the relationships between these key
`ideas. This may lead to a more objective way of
`modelling the intellectual structure of scientific special(cid:173)
`ties. Changes in the eo-citation patterns, when viewed
`over a period of years, may provide clues to under(cid:173)
`standing the mechanism of specialty development.
`
`• Example of Co-eitation
`
`An illustration of a co-citation pattern which exists in
`the scientific literature is shown in Figure 1. Each box
`in the network diagram represents a cited paper in a
`specialty of particle physics which might be roughly
`described as: "Theories of broken chiral symmetry and
`
`current algebras." (The full bibliographic citation for
`each of the articles in the network is given in Table 1.)
`The number of undirected lines which connect the papers
`reflects, according to the code shown, the strength of
`co-citation coupling. The two strongest linkages are
`between Lovelace '68 and Veneziano '68, and between
`Gell-Mann '68 and Glashow '68. The network was drawn
`from data given in Table 2, omitting co-citation linkages
`which fall below the threshold of seven.
`The procedure known as "cycling" was used in identi(cid:173)
`fying the papers (8). The highly cited paper by Gell(cid:173)
`Mann (1968) was used as a starting point. All the papers
`which cited Gell-Mann '68 in the first quarter 1971 SCI
`were examined. All citations for the papers in the bibliog(cid:173)
`raphies of the citing papers were then located in the
`quarterly SCI. If any of the items was cited ten or more
`
`Fiouu 1
`
`Co-citation Network for Frequently Cited Papers in Particle Physics
`(Data from the 1971 SCI)
`
`Veneziano G
`Lovelace C ~~~~~~~~~~~~~~
`Phys Letters B 1=
`Nuovo Cimento
`28B 264 '68
`57 190 '68
`
`Wilson K
`Phys Rev
`179 1499 '69
`
`Gasiorowicz S
`Rev Mod Phys
`41 531 '69
`
`Key: author
`joumal
`vol pg yr
`
`Scale: Co-citation
`1-6
`7-12
`13-24
`25-48
`49and up
`
`Lines
`not shown
`1
`2
`3
`4
`
`266
`
`Joumal of the American Society for Information Science-July-August 1973
`
`
`
`TABLE 1. Papers in the network
`
`BJORKEN, J. D. "Applications of the Chiral U(6)®U(6)
`Algebra of Current Densities," Phys Rev 148: 1467 '66
`GAsiosowicz S. and D. A. GEFFEN "Effective Lagrangians
`and Field Algebras with Chiral Symmetry," Rev Mod
`Phys 41: 531 '69
`GELL-MANN, M., "Symmetries of Baryons and Mesons,"
`Phys Rev 125: 1067 '62
`GELL-MANN, M., "The Symmetry Group of Vector and
`Axial Vector Currents," Physics 1: 63 '64
`GELL-MANN, M., R. J. OAKEI! and B. RENNER, "Behavior of
`Current Divergences under SUaXSUa, Phys Rev 175:
`2195 '68
`GLAsuow, S. L. and S. WEINBERG "Breaking Chiral Sym(cid:173)
`metry," Phys Rev Lett 20: 224 '68
`LoVELACE, C., "A Novel Application of Regge Trajectories,"
`Phys Letters B 28B: 264 '68
`VENEZIANO, G., "Construction of a Crossing-Simmetric,
`Regge-Behaved Amplitude for Linearly Rising Tra(cid:173)
`jectories," Nuovo Cimento 57: 190 '68
`WEINBERG, S., "Pion Scattering Lengths," Phys Rev Lett
`17: 616 '66
`WILBON, K. G., "Non-Lagrangian Models of Current Alge(cid:173)
`bra," Phys Rev 179: 1499 '69
`
`times it became part of the cluster of key papers to be
`studied. Co-citation relationships among all items in the
`cluster were determined in the manner described earlier.
`The resulting network of ten papers is shown in Figure 1.
`It should be emphasized that while these ten papers may
`not represent all the frequently cited papers in this
`specialty, they probably constitute the "core" of that
`literature.
`How can we explain the strength of these co-citation
`linkages? In some cases, we are dealing with relation(cid:173)
`ships which are strongly recognized by physicists in this
`specialty. For example, 50 authors of papers in 1971
`cited both Gell-Mann '68 and Glashow '68. When we
`examine the Gell-Mann '68 paper, we find that Cell(cid:173)
`Mann not only directly cites the Glashow paper, he also
`discusses its high similarity to his own work: "Many
`authors have taken steps in the same direction, and in
`particular we should mention the work of Glashow and
`Weinberg as being the most closely related to ours" (9)
`(Author's italics). Furthermore, these papers display a
`strong bibliographic coupling, citing five common previ(cid:173)
`ous papers. Based on this example, one might be tempted
`to hypothesize that co-citation, like bibliographic cou(cid:173)
`pling, measures subject similarity. However, the same
`agreement between bibliographic coupling and co-cita(cid:173)
`tion strength was not observed in other cases.
`Table 2 displays the absence of any clear relationship
`between bibliographic coupling strength and co-citation
`
`frequency. This is especially evident in the case of Love(cid:173)
`lace '68 and Veneziano '68, the strongest co-citation
`linkages on the diagram. These papers bad a biblio(cid:173)
`graphic coupling strength of only one. Once again, a
`reading of the Lovelace paper reveals why it came to be
`so closely associated with the Veneziano paper, and,
`further, why it was linked to the Gell-Mann work. In
`the first paragraph of his paper Lovelace stated: "Re(cid:173)
`cently Veneziano gave a simple formula which exhibits
`the Regge pole=resonance duality. We show here how
`this new development in Regge theory can be applied in
`a rather unexpected direction, . . . There is. also reason,
`from at least seven parallel predictions, to believe [the
`Veneziano formula] connected with chiral symmetry."
`(10) Lovelace has applied the Veneziano formula in a
`new way which succeeds in establishing a link with the
`Gell-Mann work on chiral symmetry. Thus, two papers
`which were strongly linked by co-citation were only
`weakly tied by bibliographic coupling, although they
`were clearly related in content.
`Of the 36 possible couplings among the 9 papers in
`Table 2, there are 8 instances of high co-citation between
`papers (six or more times) in which there was no biblio(cid:173)
`graphic coupling at all. In 11 cases of very strong
`co-citation (15 or more), eight had bibliographic coupling
`strengths of either one or zero (four had couplings of one,
`and four had zero). These results suggest that biblio(cid:173)
`graphic coupling is a less reliable indication of subject
`similarity than co-citation, although co-citation may, in
`addition, reflect "semantic" relations among cited papers,
`analogous to those observed in patterns of co-occur(cid:173)
`rence. Bibliographic coupling has been used to assemble
`groups of papers on particular subjects (11). Co-citation
`could also be used to form these groups after the papers
`have been cited, but the present data suggest that the
`results of these two grouping procedures would differ
`quite significantly.
`A much better predictor of strong co-citation linkages
`between papers is provided by the direct citation pat(cid:173)
`terns, that is, the citing of one paper by another. Figure
`1 shows that the strongest eo-citation links are between
`papers connected by direct citation. Of the 11 cases of
`strong co-citation (15 or more), eight are direct citation
`connections. In fact, all cases of direct citation are
`represented by co-citation strengths of at least 5, and
`many are much higher. The average co-citation fre(cid:173)
`quency for papers connected by direct citation was 17.8,
`while the average for papers not connected by direct
`citation was 4.9. 0 n the other hand, there are nine cases
`of co-citation strength of 6 or more which are not con(cid:173)
`nected by direct citation. Thus, direct citation does not
`predict all strong co-citation linkages between papers in
`a field, although it more closely parallels such linkages
`than does bibliographic coupling.
`
`Journal of the American Society for Information Science-July-August 1973
`
`267
`
`
`
`TABLE 2. Co-citation, bibliographic coupling, and direct citation between pairs of papers
`
`~
`II:
`0
`
`~
`
`5
`1
`no
`10
`8
`ves
`18
`1
`no
`15
`0
`ves
`50
`5
`ves
`
`.. <:.)
`~
`:;
`
`0
`1
`6
`.no
`
`cq
`SP
`
`~
`::11
`~
`
`9
`0
`yes
`6
`9
`yes
`
`;
`~
`::11
`:::!:
`~
`7
`1
`yes
`9
`2
`yes
`19
`4
`yes
`
`I
`
`~
`~
`~
`~
`13
`3
`no
`11
`5
`ves
`32
`0
`ves
`20
`0
`ves
`
`Bjorken '66
`
`Gasiorowicz '69
`
`Gell-Mann '62
`
`Gell-Mann '64
`
`Gell-Mann '68
`
`Ulashow '68
`
`Lovelace '68
`
`Veneziano '68
`
`Weinberg '66
`
`G.>
`
`G.>
`~
`
`~
`l
`.s
`2
`0
`no
`1
`3
`no
`3
`0
`no
`4
`0
`no
`7
`0
`no
`1
`1
`no
`
`co-citation
`bibl. coupling
`direct citation
`
`G.>
`
`~
`0
`..
`-~
`~
`2
`0
`no
`1
`0
`no
`0
`0
`no
`0
`0
`no
`7
`0
`no
`0
`0
`no
`68
`1
`ves
`
`J
`2!1
`G.>
`.&J
`-~
`til::
`2
`2
`no
`9
`6
`yes
`6
`3
`no
`9
`1
`ves
`19
`1
`no
`10
`2
`no
`21
`1
`ves
`17
`0
`110
`
`Cl>
`SP
`
`=
`~
`
`11
`4
`yes
`1
`17
`no
`7
`1
`yes
`6
`1
`no
`18
`4
`ves
`7
`0
`ves
`1
`4
`.no
`1
`0
`no
`0
`1
`no
`
`Each paper in the table is indicated by the first author's last Dllme &Dd the ye&r the paper was published. There ere three entries for e&ch pair
`of papers. The top entry in each box is the co-citation frequency (determined from the 11'11 SCI), the second entry is the bibliographic coupling
`strength, and the third is a "yes" if the later paper cited the earlier one, and "no" if it did not.
`
`• Conclusions
`
`It apepars that an interpretation of the significance of
`strong co-citation links must rely both on the notion of
`subject similarity and on the association or co-occurrence
`of ideas. We have established, at any rate, that this
`measure in many cases reflects the existence of direct
`citation links and corresponds to significant intellectual
`connections within the field. Two applications in infor(cid:173)
`mation retrieval are immediately apparent. A secondary
`index based on highly co-cited papers would allow
`sequential searches through a citation index, retrieving
`a list of new documents at each co-cited entry point.
`Second, co-citation could be used to established a cluster
`or "core" of earlier literature for a particular specialty.
`This 'core" could serve as a "profile" for that specialty
`and, thus, the basis of an SDI system.
`Another area for the application of co-citation is in
`the study of the specialty structure of science (1S).
`The pattern of linkages among key papers establishes a
`
`structure or map for the specialty which may then be
`observed to change through time. Through the study of
`these changing structures, co-citation provides a tool for
`monitoring the development of scientific fields, and
`for assessing
`the degree of
`interrelationship among
`specialties.
`
`• Acknowledgement
`
`I would like to thank Professor Belver Griffith of
`Drexel University and Drs. Eugene Garfield and Morton
`Malin, Institute for Scientific Information, for discussions
`and valuable suggestions.
`
`References
`
`1. KEssLER, M. M., "Bibliographic Coupling Between Sci(cid:173)
`entific Papers," American Documentation, 14 (No.1):
`10-25 (1963).
`
`268
`
`Journal of the American Society for Information Science-July-August 1973
`
`
`
`2. GARFIELD, E., "Citation Indexes for Science," Science,
`122: 108-111 (1955).
`3. SALTON, G., "Associative Document Retrieval Tech(cid:173)
`niques using Bibliographic Information," Journal of
`the ACM, 10 (No.4): 440-457 (1963).
`4. PRICE, N. and S. ScHIMINOVICH, "A Clustering Experi(cid:173)
`ment: First Step Towards a Computer-Generated
`Classification Scheme," Information Storage and Re(cid:173)
`trieval, 4: 271-280 (1968).
`5. STEVENS, M. E., V. E. GIULIANO and L. B. HEILPRIN,
`Eds., Statistical Association Methods for Mechanized
`Documentation (National Bureau of Standards Mis(cid:173)
`cellaneous Publications 269, 1965).
`6. We can also give a more formal definition of co-cita(cid:173)
`tion in terms of set theory notation. If A is the set
`of papers which cites document a and B is the 'set
`which cites b, then AnB is the set which cites both
`a and b. The number of elements in AnB, that is
`n(AnB), is the co.-citation frequency. The relative
`co.-citation frequency could be defined as n(AnB) -7-
`n(AUB).
`7. GARFIELD, E., "Citation Indexing, Historio-Bibliography,
`and the Sociology of Science," Proceedings of the
`
`Third International Congress of Medical Librarian(cid:173)
`ship, Amsterdam, 5--9 May 1969, (Excerpta Medica
`1970) pp. 187-204.
`8. CAWKELL, A. E., "Search Strategies using the Science
`Citation Index," in Computer Based Information Re(cid:173)
`trieval Systems, B. Hougton, Ed. (London, 1968) pp.
`27-44.
`9. GELL-MANN, M., R. J. OAKES and B. RENNER, "Be(cid:173)
`havior of Current Divergencies under SUa X SUa,"
`The Physical Review, 175 (No.5): 2195--2199 (1968).
`10. LoVELACE, C., "A Novel Application of Regge Trajec(cid:173)
`tories," Physics Letters, 28B (No.4): 264-268 (1968).
`11. ScHIMINOVICH, S., "Automatic Classification and Re(cid:173)
`trieval of Documents by Means of a. Bibliographic
`Pattern Recognition Discovery Algorithm," Informa(cid:173)
`tion Storage and Retrieval, 6: 417-435 (1971).
`12. Programs are currently being developed for auto(cid:173)
`matically determining co.-citation and using these data
`as input to a cluster generating program. The results
`of some preliminary experiments using this system to
`examine scientific specialties will be reported in a
`forthcoming paper.
`
`Journal of the American Society for Information Science-July-August 1973
`
`269
`
`