Gene's worlds: Comparing Eugene Garfield's co-citation and tri-citation image PFNets over time



The structure of Eugene Garfield's Citation Image over two time periods (1978–87 and 19982007) is explored using both “standard” Author Co-citation Analysis and Tri-citation Analysis- co-citation with Garfield included as a third cited author. A comparison of PFNets for each decade and mapping approach shows that tri-citation can bring out more specific and context-relevant author relationships than co-citation that is based on all articles citing any pair of Citation Image authors.


This poster explores the use of tri-citation (co-citation with the focal author as context for data gathering) over different time periods to obtain a contextual view of the focal author's contributions and relationships that is lost in standard ACA mapping of Citation Images (White, 2000) and a finer-grained portrayal of an author's subject position over time vis a vis the other Image authors. The author of interest is Eugene Garfield. Garfield's career spans (and, in large part, created) the rise of interest and application of citation analysis and other bibliometric methods for studying scholars and scholarship. This project captures “Garfield's Worlds” near the beginning (1978–1987) and in the most recent decade (1998–2007).


I obtained a list of authors frequently co-cited with Garfield in each time period from Social SciSearch (Dialog File 7) and used the LIMITALL and RANK commands to constrain the search to a replicable time frame and to sort the set of co-cited authors retrieved in frequency-rank order. I retained all authors in the top 50 who had at least 30 citations with Garfield. Table 1 lists the authors and their co-citation counts in each time period. Authors with bolded names are listed in both the early and late Citation Images.

Table 1. Citation Image Authors
AuthorsCounts 1978–1987AuthorsCounts 1998–2007
Price DD212Moed HF181
Cole JR131Small H175
Small H127Seglen PO169
Narin F113Cronin B162
Merton RK103Price DD158
Zuckerman H81Merton RK126
Cole S75Glanzel W125
Line MB73Egghe L120
Moravcsik MJ73van Raan AFJ119
Meadows AJ59White HD114
Crane D58Narin F110
Griffith BC58MacRoberts MH109
Cawkell AE52Rousseau R109
Vlachy J52Leydesdorff L96
Chubin DE48Cole S81
Hagstrom WO44Braun T80
Gilbert GN44Ingwersen P77
Frame JD42McCain KW76
Kuhn TS42Schubert A74
Garvey WD41Cole JR71
Brookes BC40Harter SP71
Endler NS39Vinkler P69
Inhaber H39Lawrence S68
Bradford SC38Zuckerman H58
Edge D37Kostoff RN55
Kessler MM35Borgman CL54
Wade N35Crane D54
White MJ35Van Leeuwen TN54
Martyn J35Martin BR51
Braun T34Kuhn TS50
Lindsey D34Moravcsik MJ50
Nalimov VV34Callon M49
Pinski G34Salton G48
Lancaster FW33Thelwall M47
Singleton A32Bradford SC46
Bayer AE32Brin S46
Carpenter MP32Gilbert GN46
Cronin B32Nederhof AJ46
Lawani SM32Oppenheim C45
Sullivan D32Cozzens SE43
Myers CR31Latour B43
Oromaner M31Lotka AJ41
Kaplan N31Lewison G40
Rushton JP30Bordons M39
  Meadows AJ39
  Swanson DR39
  Smith R39

Co-citation and tri-citation counts for the authors with Garfield as participant and as context were gathered using standard Dialog search protocols. Following White (2003), a PFNet was extracted from each matrix of raw co-occurrence counts and the structure re-visualized in Pajek with node sizes reflecting the degree centrality (number of node connections) of each author in the network.

Results and Discussion

The co-citation and tri-citation analyses are presented as a series of labeled PFNets. The co-citation PFNets (Figures 1 and 2) are based on co-citation counts for all authors (Garfield and the authors in Table 1) across SSCI. The tri-citation PFNets (Figures 3 and 4) are based on co-citation counts of the authors in Table 1 with Garfield retained as a third author in each search. The regions of the PFNets are labeled based on the cited authors' highly cited areas of research and publication. Space considerations preclude an extended discussion of each PFNet and a fuller bibliography for cited authors; only the highlights are provided below and presented in the actual poster. The results of additional, more detailed analyses are available from the author upon request.

The Co-citation Structure of Garfield's Citation Image

Figure 1.

1978–1987 Co-citation PFNet

Figure 2.

1998–2007 Co-citation PFNet

The co-citation PFNets show links among the Citation Image authors based on co-citations in all of SSCI. In Figure 1, there are 3 main authors. Garfield attracts the most other authors. His monograph on citation analysis (Garfield 1979) and early articles from Science and Nature (Garfield 1955, 1972, 1976) are among his most cited works. Thus his subnetwork essentially represents all of citation analysis and its application across the social sciences, including using bibliometrics for evaluation of literatures, fields, and scholars. Price anchors a set of authors who are known either for bibliometric models or for studies of scientific communication and literatures - two of the topics for which Price was best known (see, e.g. Price 1961, 1963, 1965). Merton is the central author in the subnetwork representing North American sociology of science.

In the later decade, Garfield is clearly the most central figure with 19 authors directly connected to his node. The label “Bibliometrics & Scientometrics” is an attempt to be inclusive of the authors who are directly connected. Merton (sociology of science) and Latour (science studies) offer contrasting views of how the practice of science should be studied and understood. On the left side of the network, we find authors such as Ingwersen, Borgman, Salton, and Harter whose work deals with a subset of information retrieval, human information behavior, and bibliometric studies. These authors connect to Garfield via Salton→Small and Ingwersen→ Rousseau and Egghe, two mathematical bibliometrician.

The Tri-citation Structure of Garfield's Citation Image

Figure 3.

1978–1987 Tri-citation PFNet

Figure 4.

1998–2007 Tri-citation PFNet

The tri-citation PFNets restrict co-citation counts between authors to those coming from papers that also cite Garfield. Thus Garfield's oeuvre becomes the context in which all other authors are co-cited. The universe has changed from all of social science to bibliometrics and scientometrics-arguably the most important aspects of Garfield's many-faceted career.

Both networks show the increased detail and topic specialization that accompanies the introduction of context. Authors who may be more frequently recognized for non-bibliometric content are likely to have their links re-organized and repositioned. In the earlier decade, Price moves from an intermediate position to center stage. Authors are linked directly to Price based on their work on Pricean areas such as communication in science and quantitative measures of scientific literatures and their use. On the lower right, Narin brings together authors concerned with quantitative evaluations of national and international science and technology (most are his colleagues/coauthors). Above, Small's introduction of co-citation mapping (Small, 1973) connects authors concerned both with the structures arising from citation linkage and the varied concepts that citation can represent. The left hand side of the network brings together the sociologists of science for whom citation data are indicators of social stratification in science (e.g. Cole 1973) and citation analyses in the social sciences.

In the most recent decade, there are several almost equally central authors in the PFNet, based on number of direct links to others. Cronin's early monograph on the Citation Process (Cronin 1984) and more recent discussions of citation linkage and the internet (Cronin et al., 1998; Cronin, 2001) put him in a mediating position between the two “traditional” areas of bibliometric research-studies of the structure of fields and literatures and evaluation of the performance of the authors and institutions. Latour's position and prominence has changed noticeably with the addition of Garfield's oeuvre as context. Within the social sciences literature as a whole, Latour was the center of a subnetwork focusing on his contributions to science studies (including social construction of scientific knowledge and actor-network theory) whereas in the contextual network he serves merely as one of two bridges (Price is the other) between Small (the central figure in structural citation studies) and Merton (sociology of science). On the left side, Moed's node brings together early and contemporary scientoetricians who both provide critical evaluations of bibliometric measures and use bibliometric and other quantitative methods to evaluate individual, institutional and national R&D performance


The Citation Image authors are those who are highly cited with the focal author (ego in White's terminology). A “traditional” co-citation mapping puts ego in the broad context of the citation database (science, social science and/or humanities) and can bring to prominence major topics with which he or she may only be tangentially associated. Ego may not even be the central figure in the network-Garfield hold that position here but other experiments (McCain 2008, McCain & Rozaklis 2008) show that ego may serve primarily as a “hook” to retrieve authors with much higher visibility and connectedness overall (as was the case with Garfield's Citation Image in the Arts & Humanities Citation Index (White, 2000). Adding ego as the context in which the co-citation counts are obtained appears to provide a more specific, nuanced view of the associations made by citing authors. The positions and centrality of the Citation Image authors are a closer representation of the contributions of ego. One drawback of tri-citation is that it requires that ego have a substantial citation history. Authors who might be interesting from a research perspective, but who are not well-cited as first authors, will not generate a sufficient number of citing papers to yield usefully high tri-citation counts.