This is a preprint of an article published in Journal of the American Society for Information Science and Technology.

 

Evolution of research activities and intellectual influences in Information Science 1996-2005: Introducing author bibliographic coupling analysis

 

Dangzhi Zhao*

School of Library and Information Studies, University of Alberta, Edmonton, AB, Canada T6G 2J4, Email: dzhao@ualberta.ca.

 

Andreas Strotmann

School of Business, University of Alberta, Edmonton, AB, Canada T6G 2J4, Email: andreas.strotmann@ualberta.ca.

 

Abstract

Author co-citation analysis (ACA) has frequently been applied over the last two decades for mapping the intellectual structure of a research field as represented by its authors. However, what is mapped in ACA is actually the structure of intellectual influences on a research field as perceived by its active authors. In this exploratory paper, by contrast, we introduce author bibliographic coupling analysis (ABCA) as a method to map the research activities of active authors themselves for a more realistic picture of the current state of research in a field. We choose the Information Science (IS) field and study its intellectual structure both in terms of current research activities as seen from ABCA and in terms of intellectual influences on its research as shown from ACA. We examine how these two aspects of the intellectual structure of the IS field are related, and how they both developed during the “first decade of the Web”, 1996-2005. We find that these two citation-based author mapping methods complement each other, and that, in combination, they provide a more comprehensive view of the intellectual structure of the IS field than either of them can provide on its own.

1. Introduction

Recent years have seen great interest in the study of knowledge networks as manifested in recorded knowledge (e.g., published papers, books). Researchers analyze interrelationships between various facets of recorded knowledge (e.g., authors, publications, or institutions), often aided by visual network representations.

Analysis and visualization of knowledge networks can effectively assist in the discovery of new knowledge, and in the management and use of existing knowledge resources (Garfield, 1979; Swanson, 1986; Small, 1999; White, Buzydlowski, & Lin, 2000). This potential is now increasingly being studied and realized, as (a) recorded knowledge has become increasingly available in digital form, which provides large amounts and varieties of data for knowledge network studies; and as (b) sufficient computing power is now readily available to social scientists for analyzing and visualizing vast networks of information (Shiffrin & Börner, 2004; Börner, et al., 2004; Boyack, et al., 2007; Henzinger & Lawrence, 2004).

Information Science (IS) is one of the research fields whose knowledge networks have been studied frequently as the IS domain knowledge that researchers in this field have is important for understanding and interpreting the knowledge networks revealed, in particular when exploring new mapping methods. The most comprehensive study of this nature is White & McCain (1998), which studied the IS field during the decades before 1996 using mainly author co-citation analysis (ACA) methodology. It showed the scholars who significantly influenced IS, the structure of these intellectual influences, and the development of these influences and their structure over three 8-year periods.

More recent developments in IS have been studied using a number of methods. Astrom (2007) studied the field via a document co-citation analysis of documents that significantly influenced IS research over three 5-year periods between 1990 and 2004. Zhao & Strotmann (2008) employed an enriched ACA methodology for a comparison between the structure of intellectual influences on IS research during the first decade of the Web (1996-2005) and that of the time periods leading up to that decade as reported in White & McCain (1998).

The research activities in the IS field itself have not nearly been studied as frequently or thoroughly as have its intellectual influences, however. Using a co-word analysis approach, Jansens et al. (2006) examined the structure of IS research during 2002-2004 as represented in linkages of language use in research papers. Citation links have yet to be used for the study of research activities proper in the IS field.

The present study aims to fill this gap. To this end, we introduce author bibliographic coupling analysis (ABCA) as a method for mapping active authors in a research field for a more realistic picture of the current state of its research activities. We also use ACA to map the authors that have influenced these active authors in order to show the structure of internal and external intellectual influences on IS research. We examine how these two aspects of the intellectual structure (i.e., research activity and intellectual influence) are related and how they developed over two 5-year time periods during 1996-2005, the decade of the rise to prominence of the World Wide Web.

This study can help improve our ability to obtain thorough views of intellectual structures of research fields, and help enhance our understanding of, and confidence in, citation-based author network analysis.

2. Problem statement and research questions

As a similarity measure for use in clustering research papers, the bibliographic coupling (BC) concept was introduced a full decade earlier (Kessler, 1963) than co-citation (Small, 1973).

The BC strength (or frequency) between two documents is defined as the number of items they share in their reference lists. Knowledge network analyses that use BC frequencies to measure similarity therefore directly map recent publications (using their citing behaviour to judge relatedness between them) rather than the older publications that these publications reference (as co-citation analysis does). This supports the study of current research activities directly rather than via an interpretation of the influences on them (Weinberg, 1974; Egghe & Rousseau, 2002). A recent case study of BC patterns in nanotube-related patents found that BC is particularly suitable for use in forecasting, e.g., for anticipating technological breakthroughs (Kuusi & Meyer, 2007). It has also been found that research fronts may be detected based on weak signals in BC analysis as papers analyzed using BC analysis are not usually selected by the number of times they have been cited, unlike in co-citation analysis (Bassecoulard, et al., 2007; Glänzel & Czerwon, 1996).

The BC frequency between two documents is fixed once these two documents have been published, and therefore does not readily support the study of changes in research fields over time (Small, 1973). Author-aggregated BC analysis, however, circumvents this problem because the BC frequency between two authors (i.e., between their oeuvres) does continue to evolve as long as at least one of them continues to publish.

In addition, as current citation databases (e.g., ISI databases) tend to index all authors of source papers but only first authors of cited references, ABCA, which examines authors of source papers rather than authors of cited references, does not need to be limited to first authors in the study of author knowledge networks, and can therefore provide more thorough views of the structure, characteristics and development of knowledge networks.

By extending BC to author-aggregated (i.e., oeuvre) analysis, we therefore expect to be able to study the structure of the current state of research activities, as well as its evolution over time, taking into account authors’ contributions both as first authors and as non-first authors.

Despite its promises, BC as an indicator of relatedness between documents has rarely been applied to research evaluation or knowledge network analysis since its introduction in the 1960s, unlike co-citation analysis which has been dominating the mapping of knowledge networks and their evolution over time. This may in part be due to the difficulty of retrieving BC frequencies directly from the data source that citation analysis studies have heavily relied on, i.e., the databases provided by the Institute for Scientific Information (ISI). Recent years, however, have seen a resurgence in knowledge network analysis studies based on BC measures, both at the document (e.g., Jarneving, 2005; 2007) and at the journal level (e.g., Boyack, et al., 2007).

Here, we introduce author bibliographic coupling analysis (ABCA), i.e., we explore how to extend the BC concept from the document level to an author-aggregated (i.e., oeuvre) approach, and study whether this extension lives up to its promises. The present study is part of a continuing effort in this direction, and limits itself to one of several possible definitions of author BC frequency and one of several possible ways of selecting authors for analysis, in order to address the following research questions:

(1)              Is the analysis of author knowledge networks based on author-aggregated BC frequencies an effective approach to the study of intellectual structures of research fields?

(2)              What was the intellectual structure of the Information Science (IS) field in terms of its research activities and in terms of the intellectual influences on its research during 1996 - 2005? How are these two aspects related and how have they evolved over time?

Addressing these questions will contribute to a more thorough understanding of the intellectual structure of the IS field since ABCA adds a view of ongoing research activities of currently active researchers to that of influential authors of the field as provided by ACA, and will help enhance our confidence in citation-based author knowledge network analysis.

3. Methodology

Based on the discussions above, we make three assumptions in this study: (a) ACA is a well-established and effective approach to the study of intellectual structures of research fields; (b) the structure of current research activities is represented by the structure of the oeuvres of currently active authors in the IS field; (c) the intellectual influences on a field are indicated by authors’ oeuvres that have been highly cited in that field.

To address our research questions, we introduce ABCA, apply it to the IS research field, and compare its results with those obtained through ACA, for two consecutive five-year periods, 1996-2000 and 2001-2005. The IS field is a research field that we and readers of this journal are familiar with (Zhao & Strotmann, 2008). The IS domain knowledge we have allows us to better evaluate author groups and to perform meaningful comparisons of results between ABCA and ACA.

Data collection

We use the same dataset that we used for an earlier enriched ACA study of the IS research field (Zhao & Strotmann, 2008), namely, Web of Science full records for all articles published during the years 1996 to 2005 in 12 core IS journals as listed in Table 1. These journals were used to define the IS research field in the landmark article on ACA and information science by White & McCain (1998). The dataset includes 4,422 records of source papers that altogether have 110,785 references, i.e., about 25 references per source paper on average.

[insert here Table 1: Journals used to define information science]

We break this dataset into two subsets, each covering five years, i.e., 1996-2000 and 2001-2005, in order to gain a more detailed view of the development of the IS field during the decade that we explored before as a whole. We study the intellectual structure of the IS field and its evolution over time as manifested in these two sets of publications, both in terms of ongoing research activities using ABCA and in terms of intellectual influences on research in the field using ACA.

We developed software to parse these downloaded records, and to store the resulting data fields such as author names, publishing sources, and years of publication of both source papers and cited references in a data structure that was convenient for later data analysis, including the determination of author BC and co-citation frequencies.

ABCA – defining author BC frequency and selecting core authors

As in ACA, ABCA uses the author as its unit of analysis, and multivariate analysis methods such as factor analysis, cluster analysis or multidimensional scaling as statistical tools to reveal the underlying structure of the interrelationships between authors. The difference between the two lies in the way that relatedness between authors is measured: ABCA uses author BC frequency rather than co-citation counts as a measure of relatedness between authors, and uses these to map citing rather than cited authors.

There are two main choices that affect the resulting networks in ABCA: the precise choice of a definition of author-aggregated BC frequency, and the specific method for selecting a set of authors to represent the research field being studied.

·         Choice of author BC frequency definition

As the BC frequency between two documents is defined as the number of references that these two documents share, the BC frequency between two authors can be defined as the number of references these two authors’ oeuvres share.

On the one hand, this could be understood to mean the extent to which the individual publications in the two authors’ oeuvres share references, and the BC frequency between these two authors could thus be calculated by accumulating these document BC frequencies. On the other hand, we could more simply treat an author's complete oeuvre as if it were a single publication, and calculate the BC frequency between two authors as the overlap between the sets of references of their respective oeuvres.

We limit this first foray into ABCA to the latter definition of author BC frequencies, and also limit it to first-author-only BC analysis by defining an author’s oeuvre in the traditional manner as all publications with this author as the first author, even though information about all authors is readily available in our dataset.

More precisely, we determine a weighted reference set for each of the first authors in our dataset, defined as follows: If author A’s oeuvre contains X papers, author A’s weighted reference set will include all the different items in the reference lists of these X papers; if an item appears in N of the X reference lists in author A’s oeuvre, it will appear in author A’s weighted reference set with a weight of N. If the same item appears in author B’s reference set with a weight of M, then this item adds min(N, M) (i.e., the smaller of N and M) to the BC frequency between authors A and B. By weighing references in this manner when calculating BC frequencies between authors, we recognize that re-citation, i.e., the fact that an author frequently cites the same papers or authors multiple times over the years, largely defines an author’s citation identity (White, 2001).

·         Choice of author selection criterion for ABCA

For this exploratory study, we select the set of authors to examine via ABCA first by number of publications and then by average BC frequency, and leave the testing of different author selection methods to future studies. More specifically, with the goal of selecting 120 authors as in White & McCain (1998) and in Zhao & Strotmann (2008), we first determined the top 250 authors ranked by number of publications, and then selected the top 120 of these 250 authors as ranked by average BC frequency. An author’s average BC frequency is of course the sum of all this author’s BC frequencies with each of the other authors in the dataset, divided by the total number of authors in the dataset less one. A high average BC frequency thus may either indicate that this author is coupled bibliographically to a fairly significant degree to a large number of other authors, or it may derive from exceptionally strong BC links with a relatively small number of authors.

This method of selecting authors to include in an ABCA study considers both an author’s productivity and her connectivity with the field. Productivity-based author selection in ABCA is in direct analogy of citation-based author selection in ACA, and connectivity-based selection of “core authors” of a research field is consistent with findings from Glänzel & Czerwon (1996) that strongly and frequently coupled documents comprise “core documents” of a research field. Of course, unlike in co-citation analysis, “core” here in ABCA is a function of a level of integration into the field rather than impact of research.

Factor analysis

Using specially developed software, for each of the two time periods under scrutiny, a set of 120 authors was selected from our dataset using the method just discussed to represent the IS field, and a matrix of their BC frequencies was calculated. Similarly, a set of 120 most highly cited authors during each of the two time slices was selected, and a matrix of their co-citation counts was determined. The number of authors selected for this study was chosen to be comparable with the 120 authors used both in White & McCain (1998) and in our own recent ACA study of IS. This number of authors should suffice for the purposes of this study and is likely to cover major research areas in IS as citations received by these authors accounted for roughly 22% of all citations and publications by these authors accounted for roughly 25% of all publications in each time period. However, a variety of lower-tier topics may be missed as this number of authors only accounted for, after all, between 7% and 8% of citing authors, and between 0.6% and 0.8% of cited authors.

Each of these four matrices was analyzed using SPSS’ Factor Analysis routine to explore the underlying structure of the interrelationships between the selected authors. The diagonal values were treated as missing data and replaced by the mean in that routine. Factors were extracted by Principal Component Analysis (PCA), and the number of factors extracted was determined based on an examination of the Scree plot, total variance explained, communalities – how well a variable (i.e., an author’s oeuvre here) is explained by the factor model, and correlation residuals – the differences between observed correlations and correlations implied by the factor model (Hair, et al., 1998). This resulted in factor models that had good model fits, as shown in Table 2.

In the case of the matrix of author BC frequencies for 1996-2000, the resulting 11-factor model explained 68% of the total variance, and the differences between observed and implied correlations were smaller than 0.05 for the most part (87%). About 75% of the communalities are above 0.6, with the highest being 0.91. Other analyses produce comparable results (Table 2).

[insert here Table 2 – factor models and their model fits]

It is clearly seen here that the factor models from matrices of co-citation counts have significantly better model fits than those from BC frequencies, which means that the factorization is clearer in ACA than in ABCA. This is probably because co-citation counts are more concentrated and BC frequencies more dispersed as authors, especially well-established ones, often write on a range of topics, but are normally perceived as being influential in just one or two areas. This is consistent with findings from previous studies that research fronts based on weak signals may be detected in BC analysis for the price of higher levels of noise (Bassecoulard, et al., 2007; Glänzel & Czerwon, 1996).

We applied an oblique rotation (SPSS Oblimin) to the factor models, and labelled the resulting factors upon examining the articles written by authors in the corresponding factors. Oblique rotation resulted in a pattern and a structure matrix, both of which are visualized here as two-dimensional maps using the technique introduced in Zhao & Strotmann (2007; 2008) to aid interpretation. On these maps, authors are represented by square nodes and factors by circular nodes. The size of a factor node corresponds to the sum of the loadings on this factor by all authors who load sufficiently on it (i.e., with a value of 0.3 or higher in this case). The width of a line that connects an author with a factor is proportional to the loading of this author on this factor, as is its grey-scale value, with wider and darker lines signifying higher loadings. The color of an author node indicates the number of factors that this author loads on with a value of at least 0.3 each: yellow for authors who only load sufficiently on a single factor, green for those who co-load on two factors, red for three factors, and blue for four factors[1].

Comparisons

Results are compared between BC analysis and ACA in order to evaluate how the two aspects of the intellectual structure of the IS field, i.e., ongoing research activities and intellectual influences, are related to each other. Results are also compared between the two time periods in order to see how the intellectual structure of the IS field evolved during this decade. To this end, several relevant network features that they reveal were examined, such as which specialties (i.e., groups of authors) are identified, which specialties are most active, how these specialties are related to each other, and what the individual scholars’ memberships are in these specialties (White, 1990; White & McCain, 1998). Authors that appeared in all four matrices are highlighted in terms of how their citation identities and citation images are related to each other and how they evolved over time (White, 2001).

4. Results and discussion

We first examine the intellectual structure of the IS field during 1996-2000 in terms of research activities and intellectual influences, followed by an analysis of how this structure changed during 2001-2005. We then discuss the implications of these observations regarding ABCA and ACA as complementary approaches to the study of intellectual structures of research fields.

Information Science during 1996-2000

Figures 1 and 2 represent the pattern matrix results of a factor analysis with oblique rotation from author BC frequency and author co-citation count matrices, respectively.  Figures 3 and 4 visualize the corresponding structure matrix results.

As usual, we interpret large factors as research specialties, and an author’s loadings on a factor as indication of this author’s membership in that specialty. Figures 1 and 2 thus show the major specialties identified and the authors’ memberships in these specialties, as an author’s loading in the pattern matrix represents this author’s unique contribution to the corresponding factor, i.e., how well the research specialty represents the author’s work in IS, or influence on IS, respectively. Figures 3 and 4 show the interrelationships between specialties and thus the overall intellectual structure of the IS field, as loadings in the structure matrix represent a combination of author contributions to specialties and correlations between specialties (Hair, et al., 1998; Zhao & Strotmann, 2008).

[insert Figures 1 - 4 here]

Figure 1: ABCA results (Pattern matrix, 1996-2000)

Figure 2: ACA results (Pattern matrix, 1996-2000)

Figure 3: ABCA results (Structure matrix, 1996-2000)

Figure 4: ACA results (Structure matrix, 1996-2000)

 

Factors are labelled upon examining the articles written by authors in the corresponding factors, and are listed in Tables 3 and 4. In these tables, the size of a factor is the number of authors who primarily load on it, and thus indicates the activity level of each specialty in the IS field; the highest loading in each factor indicates how distinct the corresponding specialty is within the IS field. Factors that are smaller than 3 authors or do not have loadings higher than 0.6 are labelled as “undefined” as they do not provide sufficient information about whether they are specialties or what specialties they may be (Culnan, 1986, 1987; Hair, et al., 1998; McCain, 1990).

As authors, especially well-established ones, often write on a range of topics, we felt it more difficult to label factors from an ABCA than those derived via ACA where one can focus on the highly cited articles by each author. Given the meaning of BC, i.e., documents sharing references, we found that looking for common themes in articles by different authors in a factor appears to be a good way to go about labelling in ABCA. The picture we see from an ABCA therefore focuses on an author’s research areas that are in common with those of a good number of other authors, whereas ACA identifies areas of an author’ research that have had high impact on the field.

[insert here Table 3: Factors and their labels (BC frequencies, 1996-2000)]

[insert here Table 4: Factors and their labels (co-citation counts, 1996-2000)]

Table 3 shows the major areas in IS that attracted a good number of researchers in 1996-2000, and Table 4 the major specialties that influenced IS research significantly during this period. Based on information in these tables and in the maps in Figures 1-4, we observe the following:

(a) Most areas were quite consistent when comparing ongoing research with the influences on this research, but some had noticeably different foci. In the general area of user studies, for example, influential papers appeared to be from all aspects of user studies, especially user theories, but ongoing research appeared to be more about users’ interaction with information retrieval systems than about user information behaviour in general. In the Scientometrics area, influential papers were more about citation behaviour (e.g., motivations), while the actual research appeared to focus on citation analysis studies for research evaluation.

(b) Research on the interrelationship between scholarly communication and the Web appeared to be an emerging specialty during this period, as it was an active research area (Figure 1 & Table 3) but not an area of influence (Figure 2 and Table 4). As we will see below, this research area is in fact the predecessor of a Webometrics specialty that rapidly grew to become a distinct major research area in IS during the second time period we study here.

(c) Three specialties had a strong influence on IS research during this time period but were not among the active research areas in IS. These were: Assessment of networked library services, Historical and theoretical perspectives of IS, and Usability.

(d) Research on users’ interactions with IR systems was very active, and appeared to be growing: 33 of the 120 active authors we study primarily belong to this area, while only 26 of the 120 influential authors belong to the general user studies area. Research on system-oriented information retrieval appeared to be evolving in the opposite direction: there were fewer active authors than influential authors in this area, which suggests that it may have been decreasing in research activity. As we will see below, these observations are supported by data from the second 5-year period as well.

(e) From the percentage of the influential authors who were also active authors, we find that research on IR interaction appears to have been quite self-contained within IS, whereas research on IR systems drew heavily on research outside IS. 16 of the 26 influential authors in user studies (62%) were also active authors, compared to 6 out of 24 (or 25%) in the IR systems specialty. This suggests that research on IR interaction was not only active but also had developed its own influential theories, methods and experiences within IS. It had become one of the core research areas that characterised the IS field during this period.

(f) The two-camp structure of the IS field that has been observed frequently before (Harter, 1992, White & McCain, 1998, Zhao & Strotmann, 2008), with dense links within each camp (“literatures” and “retrieval”) and sparse connections between them, was quite clear both in the structure of active research (Figure 3) and in that of influential authors (Figure 4). Research in the IR camp focused on IR interaction and IR systems, with influences extending quite strongly to the OPAC area, as well as to an undefined area that appeared to have a strong component of historical and theoretical perspectives of IS (e.g., Rayward, Vickery, Hjorland, and Buckland). Research in the literatures camp was more evenly distributed among five specialties: collaboration, mapping of science, scientometrics, bibliometric models and distributions, and scholarly communication and the Web, but the influences did not include Web-related research.

Information Science during 2001-2005

Figures 5 and 6 represent the pattern matrix results of a factor analysis with oblique rotation of an author BC frequency and an author co-citation count matrix, respectively, covering the second 5-year period (i.e., 2001-2005). Figures 7 and 8 show the corresponding structure matrices. Factors are labelled upon examining the articles written by authors in the corresponding factors, and are listed in Tables 5 and 6.

 

[insert Figures 5-8 here]

Figure 5: ABCA results (Pattern matrix, 2001-2005)

Figure 6: ACA results (Pattern matrix, 2001-2005)

Figure 7: ABCA results (Structure matrix, 2001-2005)

Figure 8: ACA results (Structure matrix, 2001-2005)

 

[insert here Table 5: Factors and their labels (BC frequencies, 2001-2005)]

[insert here Table 6: Factors and their labels (co-citation counts, 2001-2005)]

 

Table 5 shows the major areas in IS that attracted a good number of researchers during 2001-2005, and Table 6 the major specialties that had a significant influence on these research areas during this period. Based on information in these tables and in the maps in Figures 5-8, we observe the following:

The factors representing current research and those showing its influences were again largely consistent, but with exceptions. Two specialties, Knowledge management and Patent analysis studies on innovation, influenced IS research during this time period but were not among the active research areas in IS, whereas the specialty Image retrieval appeared as one of the active research areas without having significant influence. Research on image retrieval thus appears to be an emerging specialty, and may keep growing in the future to produce its own influential works and authors just as Webometrics evolved from the first to the second 5-year time period we study.

Research on Webometrics was quite active, but already appears to be decreasing: while a full 21 of the 120 influential authors we study primarily belong to this area, only 15 of the 120 active authors conduct research in this area. Similarly, research on system-oriented IR appears to continue to decrease: 16 influential and 11 active authors primarily belong to this area.

The two-camp structure was still quite clear if not clearer in the structure of influences (Figure 8), but the two camps appeared to start to come closer in the structure of active research (Figure 7) with the IR systems specialty in the IR camp and the Webometrics specialty in the literatures camp together serving as a bridge.

Developments in Information Science from 1996-2000 to 2001-2005

Comparing the second 5-year time period with the previous one, we can see that the literatures camp has been very stable whereas the IR camp has undergone quite a bit of restructuring.

One of the five active research areas in the literatures camp from 1996-2000 (i.e., Collaboration) disappeared, and one (i.e., Scholarly communication and the Web) grew into a distinct specialty (i.e., Webometrics). A small but distinct research area has emerged (i.e., E-resources in scientific communication) which studies how scientists interact with electronic resources. Correspondingly, collaboration was no longer among the specialties that significantly influenced the research in this camp, whereas research on innovations through patent analysis was, indicating that this area of research was an important part of research in the Scientometrics specialty (e.g., Meyer). The rest of the camp essentially remained unchanged both in terms of size and of the nature of research and influences.

The IR camp, by contrast, developed from two dominating research areas to several more evenly distributed specialties, becoming more similar in overall structure to the literatures camp in this regard. The IR systems specialty became significantly smaller (shrinking from 20 to 11 authors). Research on OPACs was no longer an active research area, while Image retrieval emerged as a new active area of research. Two good-sized specialties split off of the large 1996-2000 IR interaction area: Information behaviour of various populations in various contexts, and Children’s Web searching behaviour. In terms of intellectual influences, the three specialties that had a strong influence on IS research but were not among the active research areas in IS during 1996-2000, i.e., Assessment of networked library services, Historical and theoretical perspectives of IS, and Usability, were no longer among the areas of influence during 2001-2005. On the other hand, Knowledge management and Patent analysis studies on innovation made the list of specialties that significantly influenced IS research during 2001-2005. Studies on users’ judgements of relevance stood out among the influences in the IR interaction area during this period.

We therefore observe an interesting contrast in the evolution of the two camps over the time period we study here.

The literatures camp, on the one hand, remained remarkably stable and unperturbed by the Web revolution, despite sprouting a completely new large specialty, Webometrics, and despite the fact that Web link and citation analysis have deep connections that one would expect to affect several areas in the literatures camp quite profoundly. We suspect that this impressive stability was due to the literatures camp’s strong and clear theoretical foundations (e.g., graph theory, citation theory, and bibliometric laws) and well-established methodology (e.g., multivariate analysis, and citation analysis), both of which continued to prove their value in the Web environment.

The IR camp, by contrast, displays evidence of major internal restructuring during this decade, constantly looking for new directions to accommodate its huge number of researchers in the face of an uncomfortably close vicinity to IR research in computer and information science and in management information systems. It appears that information scientists in the sense we define here as in White and McCain (1998) finally found a unique niche for their research within IR in this period, by focusing on the area of information users and use in an era where the Web has made end-user searching increasingly important and indeed ubiquitous. As a result, a huge number of researchers were attracted to this area, developing it in various directions.

IS researchers’ writing behavior and patterns of influences

Figures 1 & 5 show IS authors’ memberships in specialties as active researchers, i.e., the research areas in which they wrote or contributed, while Figures 2 & 6 depict areas of scholars’ influences on IS research or specialties in which their writings have been cited or recognized. We can therefore observe here some patterns both in the writing behaviour and the influences in the IS field.

Most IS researchers were highly focused and wrote in a single specialty (author nodes in yellow). About 30% of the authors did research in two specialties (green nodes), and only a few in three specialties (red nodes). The researchers who had broad research interests and activities include Qin, Van Raan, Vickery, ZhangY, and Zitt for 1996-2000, and Benoit, Kim, Liang, Ozmutlu, Yang, and Zitt for 2001-2005.

Scholars’ influences on IS research appear to have the same pattern in this regard: most were recognized in a single specialty, about one third in two specialties, only a few in three or more. Kling influenced four specialties, and Anderson, Brookes, ChenH, Price, Swanson, and Van Raan influenced three specialties for 1996-2000. Those who influenced three specialties for 2001-2005 include Allen, Belkin, ChenC, ChenH, Dillon, Hawking, Katz, Leydesdorff, Price, and Van Raan. Of course, some of the multiple memberships (e.g., KimK, ChenC, and ChenH) may have been the result of multiple authors under the same name (last name and first name initial – e.g., K and KS for the K in KimK, H, HC, HH, HS, HM, HL for the H in ChenH, and C, CM, CC for the C in ChenC), one of the drawbacks of author citation analysis.

Most of the multiple memberships occurred within each of the two camps, but there are a few notable exceptions. Qin wrote mostly in the literatures camp but also in the IR systems specialty in the IR camp during 1996-2000, and Benoit mostly in the IR camp but also in the Mapping of science specialty in the literatures camp during 2001-2005. In terms of influences during 1996-2000, Swanson’s work was recognized in both of the two major specialties in the IR camp, and in the Mapping of science area in the literature camp; Brookes influenced mostly the Bibliometric models and distributions specialty in the literatures camp, but also had some influence on user studies in the IR camp; and Anderson was recognized in the Scientometrics and Mapping of science areas as well as in the User studies specialty. During 2001-2005, Dillon and Howking had influences mostly within the IR camp but also influenced Webometrics in the literatures camp.

As was to be expected given the natures of ACA and author BC analyses, scholars from outside the IS field and historical figures only appear as influential authors in the ACA study but not as active researchers in the ABCA study. This includes many authors in the Scientometrics specialty (e.g., Merton, S. Cole, and A.J. Meadows), and a large group of computer scientists in the IR systems specialty (e.g., Croft, Harman, Lewis, Salton, van Rijsbergen, and Voorhees) or in the Webometrics specialty (e.g., Kleinberg, and Lawrence). This was probably one reason why each of these areas of active research shown from ABCA (Table 3) was smaller than the corresponding area of influence seen in ACA (Table 4). This suggests that studies in these three areas in the IS field have been drawing heavily on research outside of IS, and that ABCA appears to provide a more realistic picture of the state of research within the IS field whereas ACA reveals the structure of both internal and external influences on IS research.

Key IS scholars’ citation identities and citation images

Among the 120 authors studied, 45 authors were both active authors and influential scholars in the first period, and 35 in the second. The smaller overlap between active authors and influential authors during the second time period may be an indication of emerging researchers or of research areas that were importing ideas heavily from outside IS and have not produced their own influential works and scholars within IS. It may of course also indicate that influential authors were withdrawing from active research for reasons such as retirement and administrative responsibilities.

Fifteen authors were both active and influential in both time periods, and can thus be considered as major contributors to IS during this decade. These are Bates, ChenH, Cronin, Egghe, Ellis, Glanzel, Hjorland, Kling, Kostoff, Leydesdorff, Rousseau, Small, Spink, Van Raan, and Vinkler. Interestingly, most of these are literatures people, which is probably a major contributing factor to the comparative stability of the literatures camp during this decade that we observed above. Table 7, extracted from Figures 1, 2, 5, and 6,  presents these authors’ citation identities (defined by authors they cite in common as revealed in ABCA), and their citation images (defined by authors they are co-cited with as revealed in ACA) (White, 2001). As ChenH in fact represents multiple authors in cited references (though not in source papers), we exclude this author in the following discussion related to citation images or identities.

[insert here Table 7: Key authors’ citation identities and citation images]

During 1996-2000, about half of these authors wrote in a single specialty, and also half was influential in a single specialty. During 2001-2005, about two thirds of these authors wrote in a single specialty, and about one third influenced a single specialty. It appears that these authors became more focused in their research from the first to the second time period, but their influences spread more widely.

All authors’ citation images changed from the first to the second time period, and so did all but 4 authors’ citation identities: Egghe, Ellis, Small, and Spink. This dynamism of author citation identities reinforces our contention that, although a document’s “citation identity” is fixed, an author’s citation identity may change over time as authors remain active in publishing. ABCA can therefore be used to study evolution in the intellectual structures of research fields, just as co-citation analysis has been used for this purpose so far.

Authors whose research was consistently broad during the decade include Glanzel, Kling, Rousseau, and Van Raan. Consistently highly focused researchers, on the other hand, include Egghe on bibliometric models and distributions, Ellis on information behavior, and Small on mapping of science. Authors whose focus narrowed from 1996-2000 to 2001-2005 include: Bates, whose research on scholarly communication and the Web receded; Cronin, who moved to Webometrics from two other literatures specialties; Kostoff, who dropped out of research on IR systems; and Vinkler, who no longer appears in the mapping of science specialty. Authors whose research interests broadened include Leydesdorff who picked up Scientometrics to add research evaluation studies to his research on mapping science.

Authors who consistently influenced a single specialty during the decade include Cronin, Ellis, and Kostoff, and authors who consistently influenced more than one specialty include Hjorland, Leydesdorff, and Van Raan. Authors whose sphere of influence widened include Bates, Egghe, Glanzel, Rousseau, and Spink. Authors whose sphere of influence became more focused include Kling, Small, and Vinkler.

If we exclude ChenH as well as the 3 authors who fell into one of the undefined specialties, the primary citation identity and primary citation image are consistent for 9 of the 11 remaining authors during 1996-2000. For 2 authors, these two differ: Rousseau wrote primarily on scholarly communication and the Web, and secondarily on bibliometric models and distributions and on Scientometrics, but was recognized primarily for his secondary research activities. Vinkler wrote primarily on the mapping of science, but was primarily recognized for his work on collaboration. Because of citation delays and the cumulative nature of a citation image, the citation images of these two authors may well be based on, or heavily influenced by, their work published before 1996, which may well have been primarily on Bibliometric models and distributions, or on collaboration, respectively. The picture of the IS field revealed through co-citation analysis may therefore not be as current as that shown from bibliographic coupling analysis, as one would expect given that the former groups authors by the way they are cited, and the latter, by the way they cite.

Similar patterns were found for 2001-2005. Primary citation identities and images are consistent for 9 of 12 authors (again excluding authors in undefined areas and ChenH). For three authors, they differ: Rousseau continues to write primarily in one area and to be recognized in another – this time, writing primarily on bibliometric models and distributions, and influencing mainly Webometrics. Kostoff and Leydesdorff both wrote primarily on the mapping of science, but Kostoff’s main area of influence was Scientometrics, while Leydesdorff’s writings were impacting Webometrics. Rousseau’s citation image during this time period appears to have been heavily influenced by his writings in 1996-2000 (i.e., Webometrics), supporting our observation above about delayed recognition in studies based on citations.

Author bibliographic coupling analysis compared to ACA

As we have seen, ABCA adds a new and interesting perspective to citation-based author knowledge network analysis and visualization. Compared to ACA, it has both advantages and disadvantages, with respect to both theory and practice.

ACA elicits the intellectual structure of a research field as manifested in its literature by investigating the correlations between authors who are highly cited in that literature, based on a similarity measure that compares the ways in which authors have impacted the research. Studies such as White & McCain (1998) and Zhao & Strotmann (2008) have successfully deduced from the statistical properties of such a network of authors the intellectual structure of a research field, but it needs to be remembered that this network reflects the structure of significant influences of older works on current research and thus only indirectly represents the structure of current and ongoing research.

We can now see that this can be problematic in a field like IS that imports so much from other research fields across the social and natural sciences, and that is undergoing major changes, such as the restructuring of the IR camp in IS as a result of the significant impact of the Web. While it is true that the great sociologists like Merton or some areas of computer science have had a profound impact on this field, we find that the actual ongoing research within the field is quite distinct from those influences. This is particularly clear in the IR camp of IS, where computer scientists such as van Rijsbergen are highly cited and influential, but where the IS literature itself is much less concerned with the creation of the technology itself than it is with the human and cognitive aspects of information technology, i.e., with its users and use, and its implications for people and organisations. Delayed recognition, inherent in citations, further increases the distance between the current state of a field and that shown through ACA, as we saw in the case of author Rousseau who was recognized in the Webometrics specialty during 2001-2005 for his work done during 1996-2000 even though his research in fact had moved away from that area back to his major area of research: Bibliometric models and distributions.

By measuring the correlations between publishing authors themselves, ABCA thus provides an alternative and much more realistic view of the internal intellectual structure of a field and its current research activities by factoring out the external and older influences on the IS field that are included in the intellectual structure found through ACA.

Because influences are of considerable value for understanding the structure of a research field, ABCA cannot replace ACA as an author knowledge network analysis methodology, although it can substitute for it to some degree if an ACA is not possible. However, as we saw clearly above, ABCA can add considerable value to the results of an ACA by helping the researcher to identify those aspects of the structure that are both current and internal to the field.

Furthermore, the combination of these two views can provide additional unique insights into the intellectual structures of the IS field that neither one alone is able to provide. Here are some examples:

(a) The different sizes of a specialty in the two views may be an indicator of growth or decline – a specialty shows evidence of growth when it has relatively more active authors than influential authors (e.g., IR interaction during 1996-2000), and vice versa for symptoms of a decline (e.g., IR systems).

(b) When a specialty appears in the ABCA view but not in the ACA picture (e.g., Webometrics during 1996-2000, or Image retrieval in 2001-2005), this may suggest that this is an emerging major research area and has yet to produce its own influential works and authors.

(c) If a specialty only appears in the ACA picture but not in the ABCA map (e.g., Assessment of networked library services during 1996-2000), it is likely that this specialty is in the process of disappearing from the research focus.

(d) The percentage of active authors among influential authors in a specialty may reflect a degree of self-reliance of the specialty – self-contained for high percentages (e.g., IR interaction), or heavily importing when low (e.g., IR systems).

(e) An author’s citation identity and citation image together may provide interesting information about the style and / or the nature of his or her research. Some authors may be very focused in their own research, but may still impact research across the camp or even the field, whereas others may have wide research interests while being mostly recognized for just one area of their research. While most authors appear to primarily influence the same research area in which they primarily do research, a few scholars both influence and explore a range of research topics and areas.

ABCA thus complements ACA very well indeed, just as one would expect given the dual nature of co-citation and bibliographic coupling in a citation “lattice” (Egghe & Rousseau, 2002). Even though ABCA on its own can provide interesting insight into the intellectual structure of a research area, it is in conjunction with ACA that it provides an opportunity for an unprecedentedly thorough author citation network analysis that accounts for both the current internal structure and the influences, historic or external, that shape a field. In addition, we have found that these two analyses taken together appear to provide a glimpse of the evolutionary trajectory that the field is on.

Given then that ABCA complements ACA so well, we can ask what the price is that we need to pay for this added value. It must be considerable, one would think, if a measure that has been around for so long in principle is only now coming into its own in practice.

One of the great practical attractions of ACA methodology has been the relative ease with which existing citation databases (e.g., ISI databases along with DIALOG) could be queried directly for author co-citation counts, thus enabling researchers to do serious analyses without the need for internal access to those databases, or the need to build their own computational infrastructure for citation analysis.

The two main drawbacks of taking the counts produced this way were relatively minor compared to the immense value that the data brought: (a) counts were limited to first authors of cited references, and (b) they were measured across entire databases rather than within a field (White & McCain, 1998). The latter merely increases the distance between the internal structure of a field and the image produced via ACA, which, as we discuss and show in this paper, is already inherent in ACA. The former problem, on the other hand, is a serious issue in some fields (Zhao, 2006), but in others presents only a minor issue as the intellectual structure of a field revealed from first-author ACA is very close to that from an all-author ACA when the field's collaboration level is fairly low, as is the case in IS (Zhao & Strotmann, 2007).

We suspect that a major practical drawback of ABCA has always been that there is no comparably simple way to directly determine the BC frequencies in existing citation databases. Instead, it is necessary to follow our example and create a small citation database that represents the field under scrutiny, and to perform one's own BC frequency counting on this database. This method, of course, is also necessary if one wants to avoid the above-mentioned two problems with ACA.

On the other hand, putting together a small citation database for a field from existing citation databases such as Web of Science and Scopus is actually easier for ABCA than for ACA. This is because records downloaded from these databases contain all the data needed for deciding both first-author and all-author BC frequencies, but lack information needed for calculating all-author co-citation counts. That means that a large number of additional database queries may be required in order to build a database for ACA studies, especially when studying highly collaborative fields where counting non-first authors appears to be necessary (Zhao, 2006).

Calculating author BC frequencies is also easier than determining author co-citation counts. While authors of source papers are normally indexed thoroughly as separate fields in these databases, information on a cited reference is usually only provided as a single string in the downloaded record. This makes it necessary to write extra code for parsing the cited reference strings for author names in author co-citation counting, and the many different reference styles in these databases can make this process troublesome and unreliable. Author BC counting uses information about authors of source papers, and can treat an entire cited reference as a whole, obviating the need for parsing.

From a research methodology perspective, ACA has the considerable advantage over ABCA that, as shown earlier in this paper, factor analysis of a co-citation matrix produces factor models with considerably better model fits than that of an author BC frequency matrix does, with the practical advantage of making it easier to label the factors and to interpret the results.

On the other hand, the interpretation of results in ABCA can be much simpler than in ACA because information about author publications (e.g., titles and perhaps even abstracts) that is necessary for interpretation is contained in the dataset that the analysis was run on. In the case of ISI databases in particular, all information about source (or citing) papers is readily available in the downloaded records, while information about each cited reference is limited to its first author and the abbreviation of the name of the journal in which it was published, so that interpretation usually requires extra labour in the form of searching for the defining publications of a factor. In addition, the common problem of author citation analyses that multiple authors have the same last name plus first initial appears to be more critical in ACA than in ABCA, probably because a research field is always much narrower than the research it cites. In our study, for example, ChenH happened to be a single author in the source papers but represented several authors in the cited references. In large research fields, the latter can become a serious problem.

In summary, the price to be paid in order to run an ABCA is negligible compared to the significant additional insights one can gain from it. This is especially true when one studies a highly collaborative research field as one would need to put together a database from existing citation and other databases anyway in order to count non-first authors in ACA studies, and would also need to write computer programs for the counting rather than rely on the search facilities provided by existing citation databases.

5. Conclusion

In this paper, we introduced author bibliographic coupling analysis (ABCA), both as a stand-alone method for analysing the current research activities internal to a field, and to complement classical ACA. We applied this new methodology to the field of information science research during the first decade of the Web, 1996-2005, split into two five-year periods.

We found that ABCA is an effective method for providing a realistic picture of current active research within a research field, whereas ACA studies the external and internal as well as recent and historical intellectual influences on the field. When combining ACA with ABCA, it appears to be possible to gain a thorough view of the intellectual structure of a research field and to obtain an idea of the evolutionary trajectory that the field is currently on.

Most research areas in IS were quite consistent with respect to the ongoing research and the intellectual influences on the research. Exceptions did exist, however. These may suggest emerging specialties when the specialties appear in the ABCA results but not in the ACA picture (e.g., Webometrics, Image retrieval), or they may indicate the fading away of an area from IS research focus when the specialties are seen only in the ACA but not in the ABCA results (e.g., Assessment of networked library services).

The strong effect that the Web has had on the field (Zhao & Strotmann, 2007) manifested itself clearly here in both time periods of the decade we studied (i.e., 1996-2005), especially in the second half. However, the “literatures” camp remained remarkably stable in its internal structure under this impact, except for adding a large new specialty “Webometrics” which appears to have both emerged and already peaked during this period. The IR camp, on the other hand, underwent dramatic internal restructuring, but now appears to have finally found its own niche, namely, the study of users and use, within the large and diverse IR research community that extends well beyond IS as defined here.

For individual authors, a combined ABCA and ACA analysis provides the information needed for contrasting authors’ citation images and citation identities (White, 2001). This contrast may show the style and nature of an author’s research (e.g., Small’s focused and persistent compared to Kling’s wide-ranging and dynamic research interests and impact). It may also document the evolution of the research interests of an author (e.g., Rousseau) as the citation image may still reflect successful prior research (e.g., Webometrics) in a period of time when the author has already moved on to other or new areas of research (e.g., bibliometric models and distributions).

6. Outlook

While we have seen quite clearly that ABCA can provide significant insights into the intellectual structure and evolution of a field, especially when coupled with ACA, our research also raises a number of questions that invite further research.

In this exploratory foray into ABCA, we made a number of intuitive methodological choices that may or may not have been optimal. Further research is needed to weigh the pros and cons of first-author vs. all-author ABCA, which may differ from the pros and cons of all-author vs. first-author ACA; the pros and cons of different possible definitions of author BCF; the pros and cons of a host of possible selection criteria for ranking authors in order to determine “core authors” to consider in an analysis, especially for a combined ACA/ABCA study; the pros and cons of methods for smoothing raw BCF matrices prior to a factor analysis; and other methodological choices.

Beyond the core methodological issues, a number of other questions are raised concerning potential uses and usefulness of ABCA. While readers of this journal will agree that the picture painted by either ABCA on its own or by the combination of ABCA and ACA is intuitively compelling, it would be good to attempt a more thorough and stricter analysis of the validity of this method. Interviews with domain experts and card sorting are among the techniques that can be used for this purpose as they were used in validation studies of ACA.

As for potential uses of ABCA in conjunction with ACA, it shows obvious promise for those who are interested in tracing, or perhaps even forecasting, the intellectual evolution of a field, be it for research evaluation or research policy purposes. Less obvious is its potential as a visual browsing tool for digital bibliographic libraries, where we have begun looking, for example, at ways to utilize ABCA visualizations of the kind used in this paper. As more and more digital libraries and repositories have started to include reference lists of research papers, each cited reference often simply as a string, it may be fairly straightforward to provide an ABCA-based visual browsing interface for them, but it would be very difficult to use ACA for this purpose (Strotmann & Zhao, submitted).

Last but not least, we suspect that researchers will find more and more ways in which different analysis methods, applied to a single dataset, can complement each other in much the way that we saw ACA and ABCA do here. In the long run, we may thus be able to obtain more and more complete, detailed, and statistically valid analyses of the structure and evolution of research fields in particular, and of both the explicit and the implicit flow of knowledge in a network of  researchers, institutions, and other actors in general.

7. Acknowledgments

This study was funded in part by the Social Sciences and Humanities Research Council (SSHRC) of Canada, Genome Canada, and Genome Prairie. The authors would like to thank Paul Pype for his assistance with coding some of the programs used in this study. We are particularly grateful to the anonymous reviewers of this paper for their insightful comments and helpful suggestions.

References

Astrom, F. (2007). Changes in the LIS research front: Time-sliced cocitation analysis of LIS journal articles, 1990-2004. Journal of the American Society for Information Science and Technology, 58 (7), 947-957

Bassecoulard, E., Lelu, A., & Zitt, M. (2007). Mapping nanosciences by citation flows: A preliminary analysis. Scientometrics, 70(3), 859-880

Börner, K., Maru, J.T., & Goldstone, R.L. (2004). The simultaneous evolution of author and paper networks. Proceedings of the National Academy of Sciences of the United States of America, 101 (SUPPL. 1), 5266-5273

Boyack, K., Börner, K., & Klavans, R. (2007). Mapping the Structure and Evolution of Chemistry Research. Proceedings of the 11th International Conference of the International Society for Scientometrics and Informetrics, June 25-27, 2007, Madrid, Spain

Culnan, M.J. (1986). The intellectual development of management information systems, 1972-1982: A co-citation analysis. Management Science, 32(2), 156-172.

Culnan, M.J. (1987). Mapping the intellectual structure of MIS, 1980-1985: A cocitation analysis. MIS Quarterly, 11(3), 341-353.

Egghe, L., & Rousseau, R. (2002). Co-citation, bibliographic coupling and a characterization of lattice citation networks. Scientometrics, 55(3), 349-361

Garfield, E. (1979). Citation Indexing –– Its Theory and Application in Science, Technology, and Humanities. New York: John Wiley & Sons.

Garvey, W.D. (1979). Communication: The Essence of Science. New York: Pergamon.

Glänzel, W. & Czerwon, H. J. (1996). A new methodological approach to bibliographic coupling and its application to the national, regional and institutional level. Scientometrics, 37, 195–221.

Hair, J.F. Anderson, R.E., Tatham, R.L., & Black, W.C. (1998). Multivariate data analysis (5th edition). Upper Saddle River, NJ: Prentice Hall.

Harter, S.P. (1992). Psychological relevance and information science. Journal of the American Society for Information Science, 43, 602-615.

Henzinger, M., & Lawrence, S. (2004). Extracting knowledge from the World Wide Web. Proceedings of the National Academy of Sciences of the United States of America 101 (SUPPL. 1), 5186-5191

Jansens, F., Leta, J., Glanzel, W., & De Moor, B. (2006). Towards mapping library and information science. Information Processing and Management, 42, 1614-1642

Jarneving, B. (2005). A comparison of two bibliometric methods for mapping of the research front. Scientometrics, 65, 245-263

Jarneving, B. (2007). Bibliographic coupling and its application to research-front and other core documents. Journal of Informetrics, 1, 287-307

Kessler, M. M. (1963), Bibliographic coupling between scientific papers. American Documentation, 14, 10–25.

Kuusi, O., & Meyer, M. (2007). Anticipating technological breakthroughs: Using bibliographic coupling to explore the nanotubes paradigm. Scientometrics, 70(3), 759-777

McCain, K.W. (1990). Mapping authors in intellectual space: A technical overview. Journal of the American Society for Information Science, 41(6), 433-443

Schneider, J.W., Larsen, B., & Ingwersen, P. (2007). Comparative study between first and all-author cocitation analysis based on citation indexes generated from XML data. Proceedings of the 11th International Conference on Scientometrics and Informetrics, June 25-27, Madrid, Spain.

Shiffrin, R.M., & Börner, K. (2004). Mapping knowledge domains. Proceedings of the National Academy of Sciences of the United States of America 101 (Suppl 1): 5183-5185

Small, H. (1973). Co-citation in the scientific literature: A new measure of the relationship between two documents. Journal of the American Society for Information Science, 24, 265–269.

Strotmann, A., & Zhao, D. (submitted). Bibliometric maps for aggregated visual browsing in digital libraries. SIGIR Workshop on Aggregated Search, Singapore, 2008.

Swanson, D.R. (1986). Two medical literatures that are logically but not bibliographically connected. Journal of the American Society for Information Science, 38(4), 228 - 233

Weinberg, B.H. (1974). Bibliographic Coupling: A Review. Information Storage and Retrieval, 10(5/6), 189-96

White, H. D. (1990). Author co-citation analysis: Overview and defense. In C. L. Borgman (ed.), Scholarly communication and bibliometrics (pp. 84-106). Newbury Park, CA: Sage.

White, H. D. (2001). Authors as citers over  time. Journal of the American Society for Information Science and Technology, 52, 87-108

White, H. D., Buzydlowski, J. & Lin, X. (2000). Co-Cited Author Maps as Interfaces to Digital Libraries: Designing Pathfinder Networks in the Humanities. IEEE International Conference on Information Visualization. (London, July 18-22, 2000), pp 25-30.

White, H. D., & McCain, K.W. (1998). Visualizing a discipline: An author co-citation analysis of information science, 1972-1995. Journal of the American Society for Information Science, 49, 327-355.

Zhao, D. (2006). Towards all-author co-citation analysis. Information Processing & Management, 42: 1578-1591     

Zhao, D., & Strotmann, A. (2007). All-author vs. first-author co-citation analysis of the Information Science field using Scopus. Proceedings of the American Society for Information Science and Technology 2007 Annual Meeting, October 19-24, 2007, Milwaukee, Wisconsin, USA

Zhao, D., & Strotmann, A. (2008). Information science in the first decade of the Web: An enriched author cocitation analysis. Accepted for publication in Journal of the American Society for Information Science and Technology.



* Corresponding author

[1] Colors are only visible in the online version of this journal.