See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/242401176

Social Network Analysis: A Powerful Strategy, also for the Information

Sciences

Article  in  Journal of Information Science · December 2002

DOI: 10.1177/016555150202800601

CITATIONS READS

1,012 5,669

2 authors, including:

Ronald Rousseau
KU Leuven
485 PUBLICATIONS   10,114 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

NSFC Project: Measurements of Interdisciplinarity: Knowledge Sources, Information Linkages, and Collaboration activity (71573085) View project

All content following this page was uploaded by Ronald Rousseau on 19 May 2014.

The user has requested enhancement of the downloaded file.

 Journal of Information Science
http://jis.sagepub.com

Social network analysis: a powerful strategy, also for the information sciences
Evelien Otte and Ronald Rousseau
Journal of Information Science 2002; 28; 441
DOI: 10.1177/016555150202800601

The online version of this article can be found at:

http://jis.sagepub.com/cgi/content/abstract/28/6/441

Published by:

http://www.sagepublications.com

On behalf of:

Chartered Institute of Library and Information Professionals

Additional services and information for Journal of Information Science can be found at:

Email Alerts: http://jis.sagepub.com/cgi/alerts

Subscriptions: http://jis.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations (this article cites 23 articles hosted on the

SAGE Journals Online and HighWire Press platforms):
http://jis.sagepub.com/cgi/content/refs/28/6/441

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008

© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
 The effect of postings information on searching behaviour

1
2
3
Social network analysis: a powerful
4
5
6
strategy, also for the information
7
8
9
sciences
1110
1
2
3
4
5
6
7
8 Evelien Otte is one large network is not unrelated to this [2–4].
9 Social network theory directly influences the way
20 UIA, Rectoraat – Academische Planning, Wilrijk, Belgium
researchers nowadays think and formulate ideas on the
1 Ronald Rousseau Web and other network structures, such as those shown
2 in enterprise interactions [5]. Even within the field of
113 UIA, IBW and KHBO, Belgium
sociology, network studies are becoming increasingly
4 important.
5 Received 2 April 2002 In this article we will study social network analysis
6 Revised 28 August 2002 (SNA) and show how this topic may be linked to the
7 information sciences. Of course, Internet studies will
8 also be mentioned, as the World Wide Web represents
9 a social network of a scale unprecedented in history [5].
Abstract.
30 Interest in networks, and in particular in social
1 Social network analysis (SNA) is not a formal theory in soci- network analysis, has only recently bloomed in soci-
2 ology but rather a strategy for investigating social structures. ology [6, 7]. There are, however, many related disci-
3 As it is an idea that can be applied in many fields, we study, plines where networks play an important role.
4 in particular, its influence in the information sciences.
Examples are computer science and artificial intelli-
Information scientists study publication, citation and co-
5 gence (neural networks), recent theories concerning the
citation networks, collaboration structures and other forms
6 Web and free market economy [8], geography and trans-
of social interaction networks. Moreover, the Internet repre-
7 sents a social network of an unprecedented scale. In all these port networks [9, 10]. In informetrics, researchers study
8 studies social network analysis can successfully be applied. citation networks, co-citation networks, collaboration
9 SNA is further related to recent theories concerning the free structures and other forms of social interaction
40 market economy, geography and transport networks. The networks [11–19]. Underlying any concrete network
1 growth of SNA is documented and a co-author network of lies a graph, a structure studied by mathematicians
2 SNA is drawn. Centrality measures of the SNA network are since Euler solved the problem of the Königsberg
3 calculated. bridges.
4
5
6 1. Introduction 2. What is social network analysis?
7
8 Network studies is a topic that has gained increasing Social network analysis, sometimes also referred to as
9 importance in recent years [1]. The fact that the Internet ‘structural analysis’ [20], is not a formal theory, but
50 rather a broad strategy for investigating social struc-
1 Correspondence to: Professor R. Rousseau, Universiteitsplein 1, tures. The traditional individualistic social theory and
2 B- 2610 Wilrijk, Belgium. E-mail: [email protected] data analysis considers individual actors making

Journal of Information Science, 28 (6) 2002, pp. 441–453 441

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
Social network analysis

choices without taking the behaviour of others into to node j. Node i is called the initial node of link e, 111
consideration. This individualistic approach ignores i = init(e), and node j is called the final node of the link: 2
the social context of the actor [21]. One could say that j = fin(e). If the direction of a link is not important, or 3
properties of actors are the prime concern here. In SNA, equivalently, if existence of a link between nodes i and 4
however, the relationships between actors become the j necessarily implies the existence of a link from j to i, 5
first priority, and individual properties are only we say that this network is an undirected graph. A path 6
secondary. Relational data are the focus of the investi- from node i to node j is a sequence of distinct links 7
gations. It should be pointed out, however, that indi- (i, u1), (u1,u2), . . ., (uk,j). The length of this path is the 8
vidual characteristics as well as relational links are number of links (here k+1). In this article we only use 9
necessary in order to fully understand social undirected graphs. Consequently, the following defini- 10
phenomena [21]. Wetherell et al. [22, p. 645] describe tions are only formulated for that case. A co-authorship 1
SNA as follows: network is an example of an undirected graph: if author 2
A co-authored an article with author B, automatically 3
Most broadly, social network analysis (1) conceptualises
social structure as a network with ties connecting members author B co-authored an article with A. An undirected 4
and channelling resources, (2) focuses on the characteristics graph can be represented by a symmetrical matrix 5
of ties rather than on the characteristics of the individual M = (mij), where mij is equal to 1 if there is an edge 6
members, and (3) views communities as ‘personal communi- between nodes i and j, and mij is 0 if there is no direct 7
ties’, that is, as networks of individual relations that people link between nodes i and j. 8
foster, maintain, and use in the course of their daily lives. 9
20
Another important aspect of SNA is the study of how 3.2. Components
1
structural regularities influence actors’ behaviour. It is
A component of a graph is a subset with the character- 2
clear that ideas originating in SNA can offer added
istic that there is a path between any node and any 3
value to investigations in many disciplines, in partic-
other one of this subset. If the whole graph forms one 4
ular those mentioned in the Introduction.
component it is said to be totally connected. 5
One distinguishes two main forms of SNA: the ego
Next we define some indicators describing the struc- 6
network analysis, and the global network analysis. In
ture (cohesion) of networks and the role played by 7
‘ego’ studies the network of one person is analysed. An
particular nodes [9]. Many more are described in the 8
example in the information sciences is White’s descrip-
literature, but we will restrict ourselves to these 9
tion of the research network centred on Eugene Garfield
elementary ones. 30
[23]. In global network analyses one tries to find all rela-
1
tions between the participants in the network.
2
SNA, although considered here mainly within the 3.3. Definition: density
3
field of sociology, is an interdisciplinary technique
The density is an indicator for the general level of 4
developed under many influences, the most important
connectedness of the graph. If every node is directly 5
ones coming from mathematics and computer science.
connected to every other node, we have a complete 6
In sociology itself SNA can be described as originating
graph. The density of a graph is defined as the number 7
from sociometrics (important names here are Lewin and
of links divided by the number of vertices in a complete 8
Moreno), the Harvard School (with W. Lloyd Warner)
graph with the same number of nodes. For an undi- 9
and the Manchester anthropological school (with
rected graph G with N nodes, the density D is defined 40
Barnes, Mitchell and Bott).
as: 1
2
2 * (#L (G)) 3
3. Some notions from graph theory D = –––––––––
N (N  1) 4
5
3.1. Directed and undirected graphs
The density is sometimes called the gamma index [9]. 6
A directed graph G, a digraph, consists of a set of nodes, 7
denoted as N(G), and a set of links (also called arcs or 8
edges), denoted as L(G). In this text the words ‘network’ 3.4. Definition: centrality [24, 25] 9
and ‘graph’ are synonymous. In sociological research The most important centrality measures are: degree 50
nodes are often referred to as ‘actors’. A link e, is an centrality, closeness centrality and betweenness 1
ordered pair (i,j) representing a connection from node i centrality. 2

442 Journal of Information Science, 28 (6) 2002, pp. 441–453

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
 E. OTTE AND R. ROUSSEAU

111 Degree centrality of a node is defined as the number between j and k. Betweenness gauges the extent to
2 of ties this node has (in graph-theoretical terminology, which a node facilitates the flow in the network. It can
3 the number of edges adjacent to this node). In mathe- be shown that for an N-node network the maximum
4 matical terms degree centrality, d(i), of node i is defined value for b(i) is (N23N+2)/2. Hence the standardized
5 as: betweenness centrality is:
6
7 d(i) = m
j
ij 2b(i)
bs(i) = ––––––––––
8
N  3N + 2
2
9
10 where mij = 1 if there is a link between nodes i and j,
1 and mij = 0 if there is no such link. In a co-author Besides the Wiener index and the BRS compactness
2 graph the degree centrality of an actor is just the measure mentioned above, every centrality measure
3 number of authors in the graph with whom she has can be used to derive a centrality measure, C, for the
4 co-authored at least one article. The degree centrality in whole network. This is done as follows:
5 an N-node network can be standardized by dividing by
6 N–1: dS(i) = d(i)/(N(1).
7 Closeness centrality of a node is equal to the total
j

Cmax  Cj 
Cnetwork = ––––––––––––––––
8 distance (in the graph) of this node from all other nodes.
9 max value possible
As a mathematical formula closeness centrality, c(i), of
1120 node i can be written as:
This formula can be applied for determining degree,
d
1
2 closeness and betweenness centrality. The summation
c(i) = ij
3 i goes over all nodes of the network; Cmax is the largest
4 value obtained in the network under study, and ‘max
5 where dij is the number of links in a shortest path from value possible’ refers to the maximum value possible
6 node i to node j. Closeness is an inverse measure of for the numerator, given the total number of nodes. It
7 centrality in that a larger value indicates a less central can be shown that the total network C-measure is 1 for
8 actor while a smaller value indicates a more central a star (one central point and all other nodes connected
9 actor. For this reason the standardized closeness is only to this central node).
1130 defined as cS(i) = (N–1)/c(i), making it again a direct
1 measure of centrality. 3.5. Definition: cliques
2 Individual closeness measures can be combined to
3 define global measures, characterizing the cohesion of A clique in a graph is a subgraph in which any node is
4 the total network. The best-known ones are the Wiener directly connected to any other node of the subgraph.
5 index [26, 27] and the BRS compactness [28–30]. Figure 1 and Table 1 present a simple example of
6 Finally, betweenness centrality may be defined three networks and differences in their characteristics.
7 loosely as the number of times a node needs a given The density index, D, indicates that the networks a,
8 node to reach another node. Stated otherwise, it is the b and c (considered in this order) become increasingly
9 number of shortest paths that pass through a given dense. All centrality measures show that node u is the
1140 node. As a mathematical expression the betweenness centre, and that the other nodes become increasingly
1 centrality of node i, denoted as b(i) is obtained as: central (that is to have a larger centrality value) when
2 considering graphs a, b and c (in this order). The

 g–––
3 networks themselves, taken as a whole, show less and
jik
4 b(i) =
j,kg jk
5
6
7 where gjk is the number of shortest paths from node j to
8 node k (j, k ≠ i), and gjik is the number of shortest paths
9 from node j to node k passing through node i.
50 According to Borgatti [25], the purpose is to provide a
1 weighting system so that node i is given a full centrality
112 point only when it lies along the only shortest path Fig. 1.

Journal of Information Science, 28 (6) 2002, pp. 441–453 443

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
Social network analysis

Table 1 111
Density and centrality measures for the networks of Fig. 1 2
3
Network a Network b Network c 4
D (2*4)/(5*4) = 0.4 (2*6)/(5*4) = 0.6 (2.8)/(5.4) = 0.8 5
dS(s) = dS(t) = dS(v) = dS(w) 1/4 = 0.25 2/4 = 0.5 3/4 = 0.75 6
dS(u) 4/4 = 1 4/4 = 1 4/4 = 1 7
dnetw 12/12 = 1 8/12 = 0.667 4/12 = 0.333 8
cS(s) = cS(t) = cS(v) = cS(w) 4/7 = 0.571 4/6 = 0.667 4/5 = 0.8 9
cS(u) 4/4 = 1 4/4 = 1 4/4 = 1 10
cnetw 1 0.778 0.467 1
bS(s) = bS(t) 2
bS(v) = bS(w) 0 0 1/3 = 0.333
3
bS(u) 1 4/6 = 0.667 2/3 = 0.667
4
bnetw 1 16/24 = 0.667 (4/3)/24 = 0.056
5
6
7
less centrality. The relative values also illustrate the year 2000 were not yet complete. Hence these are also 8
difference between the three centrality measures not shown in the graph. 9
considered here. Figure 2 clearly shows the fast growth of the field in 20
For more information on graphs the reader is referred recent years. More specifically, the real growth began 1
to References [31–36]. around 1981, and there is no sign of decline. This is 2
most obvious in SA, but Medline Advanced also shows 3
a modest increase. This proves that other fields, besides 4
4. The development and growth of social sociology, have used the term and the techniques. 5
network analysis Next, quantification of this growth was attempted. A 6
linear regression analysis was performed on the SA 7
It is often stated that SNA has recently experienced data. This led to the equation 8
rapid growth, but this statement has rarely been proved. 9
In order to corroborate the latter, three databases were p(t) = 20.74 + 5.958t (1) 30
consulted: CSA Sociological Abstracts Database (SA), 1
Medline Advanced and PsycINFO. It is clear that the where time t = 1 in 1974 and p(t) denotes the number of 2
first one would have sufficed for the investigation, but published articles as a function of t. The correlation co- 3
it was interesting to find out whether related fields used efficient is 0.956, which is highly significant. An attempt 4
this technique and, if so, whether a similar growth to fit an exponential distribution yielded unsatisfactory 5
could be detected there. The yearly number of articles results. From this it was concluded that the field has 6
related to SNA was counted, as well as the number of experienced a linear growth over the last 25 years. 7
subjects within the field that were discussed. For the A similar fitting exercise on the Medline data 8
latter aspect the subject headings of Sociological resulted in the following regression line: 9
Abstracts were used (see Appendix, Table A1). 40
p(t) = 5.018 + 1.232t (2) 1
2
4.1. Growth in the number of published articles with a correlation coefficient of 0.927. 3
Searching in SA (for the period 1963–2000) 1601 arti- The cumulative number of articles published in the 4
cles were retrieved having ‘social network analysis’ in field (SA data only) was next investigated, from the year 5
the ‘Subject heading’ field. In Medline Advanced 308 1974 on. This is illustrated in Fig. 3. 6
articles were retrieved and in PsycINFO 105. The 1601 Applying non-linear regression leads to the equation: 7
articles found in SA have publication dates between 8
1969 and 2000. There are, however, only two articles P(t) = 34 + 0.96t 2.28
(3) 9
from the year 1969 and two from 1971. These are 50
omitted from the graph (Fig. 2). Similarly, when the where P(t) denotes the cumulative number of articles in 1
search was performed (beginning 2001) data for the the SA database, with t = 1 for the year 1974 (r2 = 0.998). 2

444 Journal of Information Science, 28 (6) 2002, pp. 441–453

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
 E. OTTE AND R. ROUSSEAU

111 160
2
3 140
4
5
120
6
7
8 100
Number of articles

9
10 80
1
2 60
3
4
40
5
6
7 20

8
9 0
1120 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
1 Year of publication
2
articles in Sociological Abstracts articles in Medline Advanced articles in PsycINFO
3
4 Fig. 2. Growth of social network analysis.
5
6
7 Note that a linear equation for year-by-year data mathe- the fields of medicine and psychology (the graphs split
8 matically leads to a square law for cumulative data. up only from 1984 on). This is no surprise as the
9 Statistical fitting yields an exponent of 2.28 instead of method was developed in sociology (at least under that
1130 the theoretically expected value of 2, a result that falls name), and only later adopted in other fields.
1 within our expectations. The three graphs (Figs 2–4) demonstrate the fact that
2 The sum of the results from the three databases (Fig. it was only in the early 1980s that SNA started its
3 4) shows that the development of SNA began later in career. The main reasons for this are the institutional-
4 ization of social network analysis since the late 1970s,
5 and the availability of basic textbooks and computer
6 software.
7 The institutionalization of the field began with the
8 foundation in 1978 by Barry Wellman of the
9 International Network for Social Network Analysis
1140 (INSNA). This is the professional association for
1 researchers interested in social network analysis. Its
2 principal functions are the publication of the informal
3 bulletin Connections, containing news, scholarly arti-
4 cles, technical columns, abstracts and book reviews;
5 sponsoring the annual International Social Networks
6 Conference (also known as Sunbelt) and maintaining
7 electronic, web-based services for its members. The
8 society also publishes, in association with Elsevier, the
9 peer-reviewed international quarterly Social Networks.
50 The earliest basic text that the authors know of
1 Fig. 3. Cumulative number of articles on social network dealing exclusively with social network analysis is
112 analysis. Knoke and Kuklinski’s Network Analysis, published in

Journal of Information Science, 28 (6) 2002, pp. 441–453 445

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
Social network analysis

200 111
2
180 3
4
160
5
140
6
7
120 8
9
Number of articles

100 10
1
80 2
3
60
4
40
5
6
20 7
8
0 9
74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
Year of publication of articles
20
1
sum of articles in Sociological Abstracts, Medline Advanced and PsycINFO 2
sum of articles in Sociological Abstracts and in Medline Advanced
3
4
articles in Sociological Abstracts
5
6
Fig. 4. Sum of SNA articles from the three databases.
7
8
1982. Other important books having influenced the ject classification scheme of Sociological Abstracts. 9
growth of the discipline are Wellman and Berkowitz’ This scheme divides the field of sociology into 33 30
Social Structures: a Network Approach (1988), Scott’s subjects, each subdivided into a variable number of 1
Social Network Analysis: a Handbook (1991), and subfields (see Appendix, Table A1). Social network 2
Wasserman and Faust’s Social Network Analysis: analysis is a subfield under the main heading ‘Complex 3
Methods and Applications (1994). Organization’. 4
The development of dedicated software also led to an Articles may be assigned to different subject head- 5
increase in interest in the field and its methods. The ings. Most articles found under the heading SNA are 6
best-known (and very user-friendly) program for the also classified under other headings. The fact that an 7
analysis of social networks is UCInet (a free evaluation article is classified with the code for social network 8
version can be downloaded from www.analytictech. analysis (0665) and one or more other codes indicates 9
com/ucinet_5_description.htm). UCInet can easily be that the author of that article has either discussed rela- 40
combined with Krackplot, a well-known program for tionships between SNA and that other subfield, has 1
drawing social maps. Other examples of computer pro- applied SNA together with techniques from that other 2
grams for social network analysis are Gradap, Multinet, subfield, or has applied SNA in that other subfield. 3
Negopy and Pajek. From 1984 on (for each year) the total number of 4
different subfield codes that were assigned together 5
with the code 0665 for SNA were counted. Figure 5 6
4.2. Articles dealing with social network analysis and
shows the linear best-fitting function of the number 7
relationships with other subjects
of additional (i.e. together with 0665) codes (Pearson 8
It is to be expected that growth in the number of studied correlation coefficient = 0.89). Further, a striking 9
subjects follows the growth in the number of articles change was observed in the subjects studied using 50
on social network analysis. In order to study this SNA, or related to it: in the early 1990s most articles 1
the following method was applied based on the sub- dealt with family and socialization, while at the end of 2

446 Journal of Information Science, 28 (6) 2002, pp. 441–453

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
 E. OTTE AND R. ROUSSEAU

111 70 5.1. Network analysis of the central cluster

number of different codes

2 60
Density. The density is an indicator for the level of
3 50 connectedness of a network. It is given as the number
4
40 of lines in a graph divided by the maximum number of
5
30 lines (the case where every author is connected to every
6
20 other one). Hence it is a relative measure with values
7
between 0 and 1. The density for the central network of
8 10
network analysts is 0.05, so this network is clearly not
9 0
dense at all, but very loose.
10 1980 1985 1990 1995 2000
Degree centrality. Degree centrality is equal to the
1 year of publication
number of connections that an actor (a node) has with
2 other actors. In this network being a central author
Fig. 5. Number of different subfield codes assigned together
3 means that this scientist has collaborated (in the sense
with the code for SNA.
4 of co-authored) with many colleagues. The author with
5 the highest degree centrality is Barry Wellman (Univer-
6 this period the SNA articles mostly dealt with the soci- sity of Toronto), who has a degree centrality of 9. The
7 ology of health and medicine. Indeed, social network degree centrality of the whole network is 11%, indi-
8 analysis is now often applied in AIDS and drug abuse cating that many authors are not connected.
9 studies. Closeness. Another way of studying centrality is
1120 using the closeness indicator. This indicator is more
1 general than the previous one, because it takes the
2 5. An SNA co-authorship network structural position of actors in the whole network into
3 account. A high closeness for an actor means that he or
4 In this section a network analysis was performed of she is related to all others through a small number of
5 authors in the field of social network analysis. The paths. The most central author in this sense is Patrick
6 central players are pointed out as are the underlying Doreian (University of Pittsburgh). The closeness of the
7 collaborative relationships between authors. Co- whole network is 14%.
8 authorship, a (strong) form of collaboration, is not the Betweenness. This measure is based on the number
9 only way to describe relationships between scientific of shortest paths passing through an actor. Actors with
1130 authors. Citation network, for instance, could reveal a high betweenness play the role of connecting different
1 other relationships, but these are not studied in this groups, as ‘middlemen’. Again Patrick Doreian has the
2 article. highest betweenness. The betweenness of the whole
3 In the 1601 articles dealing with SNA there were 133 network is 47%.
4 authors occurring three times or more. Forming an Cliques. UCInet found 16 cliques, meaning 16
5 undirected co-authorship graph (of these 133 authors) subgraphs consisting of three or more nodes. The
6 led to a big connected component of 57 authors, two largest one consists of six authors: Bernard, Johnsen,
7 components of four authors, two components of three Killworth, McCarthy, Shelley and Robinson. The
8 authors, seven small components consisting of two second largest one consists of the five authors: Erger,
9 authors and 48 singletons. The central cluster of Lovaglia, Markovsky, Skvoretz and Willer.
1140 57 authors will be concentrated on. Most important
1 scientists in the field belong to this cluster. There are,
2 however, exceptions, the most notable one being 6. A bibliometric analysis of the SNA
3 Ronald S. Burt (University of Chicago), who has 17 arti- database
4 cles in the Sociological Abstracts database. As these
5 articles are either written as a single author or with Barry Wellman is the most prolific author in the field of
6 authors who have only one article in the database, they social network analysis, based on the Sociological
7 were excluded. Presumably these collaborators were Abstracts database. He published 31 articles in the
8 students. investigated period (21 as first author). Table 2 shows
9 Network analysis was performed using UCInet while the most prolific authors (using total counts) in the field
50 the map was drawn with Pajek (Package for Large of SNA. More details can be found in Otte [7].
1 Network Analysis). Figure 6 shows the network of The author publication frequencies can easily be
112 network analysts. described by a Lotka distribution, which is a power

Journal of Information Science, 28 (6) 2002, pp. 441–453 447

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
Social network analysis

111
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
20
1
2
3
4
5
6
7
8
9
Fig. 6. The network of network analysts. 1, D.D. Brewer; 2. E.J. Bienenstock; 3. S.D. Berkowitz; 4, M. Gulia; 5, P. Bonacich; 6, 30
H.R. Bernard; 7, V. Batagelj; 8, K. Carley; 9, K.E. Campbell; 10, P. Doreian; 11, J.S. Erger; 12, L.C. Freeman; 13, K. Faust; 14, A.
1
Ferligoj; 15, N.E. Friedkin; 16, T.J. Fararo; 17, J. Galaskiewicz; 18, J.S. Hurlbert; 19, C. Haythornthwaite; 20, V.A. Haines; 21,
N.P. Hummon; 22, I. Jansson; 23, E.C. Johnsen; 24, D. Krackhardt; 25, P.D. Killworth; 26, M.J. Lovaglia; 27, B.A. Lee; 28, P.V.
2
Marsden; 29, B. Markovsky; 30, M.S. Mizruchi; 31, D.L. Morgan; 32, C. McCarthy; 33, M. Oliver; 34, S. Potter; 35, B. Potts; 36, 3
T. Patton; 37, D. Ruan; 38, J. Skvoretz; 39, J.W. Salaff; 40, T.A.B. Snijders; 41, J.J. Suitor; 42, F.N. Stokman; 43, G.A. Shelley; 4
44, M. Spreen; 45, J. Szmatka; 46, S.R. Thye; 47, M.A.J.Van Duijn; 48, G.G. Van de Bunt; 49, B. Wellman; 50, C. Webster; 51, S. 5
Wasserman; 52, D. Willer; 53, E.P.H. Zeggelink; 54, K.L. Woodard; 55, S.L. Wong; 56, N.S. Wortley; 57, S. Robinson. 6
7
8
law. Using the software program available in the where, now, f (y) denotes the relative number of authors 9
journal Cybermetrics [37], it was found that: with y first authorships, y >0. Again the fit is excellent 40
(the Kolmogorov–Smirnov Dmax = 0.004). For the role of 1
0.790 Lotka’s law in informetrics the reader is referred to 2
f (y) = ––––– (4)
y 2.727 Wilson’s [38] review article. 3
As it is the thesis of this article that network analysis 4
where f(y) denotes the relative number of authors with is a field equally important to sociology as to the infor- 5
y publications. According to the Kolmogorov–Smirnov mation sciences, the database LISA was consulted in 6
statistic the fit is excellent (Dmax = 0.009). Counting only order to find out whether the top scientists in SNA had 7
first authors (see Appendix, Table A2, for data) the also published in journals covered by this library and 8
following frequency distribution was obtained: information science database. It turned out that, of the 9
47 most prolific SNA authors (that is those who wrote 50
0.802 at least six articles), 12 had articles in LISA (not neces- 1
f (y) = ––––––– (5) sarily as first author). Articles were published in
y2.800 2

448 Journal of Information Science, 28 (6) 2002, pp. 441–453

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
 E. OTTE AND R. ROUSSEAU

111 Table 2
2 Most prolific authors in SNA
3
4 Author Number of Author Number of
published articles published articles
5
6 Wellman, Barry 31 Klovdahl, Alden S. 9
7 Skvoretz, John 24 Lovaglia, Michael J. 9
8 Bonacich, Phillip 20 Snijders, Tom A.B. 9
9 Everett, Martin G. 20 Stokman, Frans N. 9
10 Willer, David 19 Wasserman, Stanley 9
1 Burt, Ronald S. 17
2 Friedkin, Noah E. 16 9 authors 8
Borgatti, Stephen P. 14 8 authors 7
3
Johnsen, Eugene C. 14 11 authors 6
4
Faust, Katherine 13 13 authors 5
5 Markovsky, Barry 13 20 authors 4
6 Doreian, Patrick 12 57 authors 3
7 Marsden, Peter V. 12 223 authors 2
8 Mizruchi, Mark S. 10 1308 authors 1
9
1120
1 journals such as Scientometrics, JASIS(T), Journal of documents’ (authorities) and reviews (hubs) [1].
2 Classification, Information Processing and Manage- As early as 1972 Nance and co-workers [47] studied
3 ment (and its precursor, Information Storage and information networks as directed graphs. In their article
4 Retrieval) and Social Science Computer Review. Table the message transfer structure was the central notion.
5 3 gives the list of these authors. The fact that Patrick These authors defined measures of network structure
6 Doreian heads this list is perhaps not surprising in view such as the accessibility and the flexibility in message
7 of his prominent role in the SNA network. Note that transfer.
8 even this mini-list yields a perfect Lotka distribution Shaw [48] used the random graph hypothesis (lines
9 (exponent equal to 1.97). of a graph are randomly selected from the set of all
1130 possible ones) to study the validity of thresholded co-
1 citation graphs. Logan and Pao [49, 50] investigated the
2 7. Use of network analysis in the structure of co-author graphs and determined central
3 information sciences authors based on their position in the co-author graph.
4 The presence of these central authors created order and
5 In this section a short, and hence necessarily incom-
6 plete, overview of articles and authors in the informa-
7 tion sciences (or at least in information science and Table 3
8 multidisciplinary journals) that have used the network SNA authors and number of articles in LISA (1969–2001)
9 approach in their investigations is given.
1140 In information science studies publications, cita- Author Number of articles in LISA
1 tions, co-citations [39, 40] as well as collaborations give Patrick Doreian 8
2 rise to networks [9]. Recently other collaborations, such Nan Lin 5
3 as movie actor collaborations, have also inspired fellow Barry Wellman 4
4 scientists [41]. These authors and others link their Kathleen Carley 2
5 research to the so-called small-world phenomenon or H. Russell Bernard 2
6 ‘six degrees of separation’ phenomenon [42–45]. A Douglas R. White 2
7 small-world network is then characterized as a network Barry Markovsky 1
8 exhibiting a high degree of clustering and having at the Tom A.B. Snijders 1
9 same time a small average distance between nodes. Michael J. Lovaglia 1
50 Moreover, the ‘hubs’ and ‘authorities’ approach is Linton C. Freeman 1
Thomas J. Fararo 1
1 related to the Pinski–Narin influence weight citation
Karen E. Campbell 1
112 measure [46] and mimics the idea of ‘highly cited

Journal of Information Science, 28 (6) 2002, pp. 441–453 449

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
Social network analysis

structure in the graph. They were, moreover, especially tion scientists are becoming interested in the Internet 111
important for the information transfer and exchange (under the names webometrics, cybermetrics), it is clear 2
within the graph. These ‘middlemen’, however, were that social network analysis will find more and more of 3
very often not represented in first author relationships. a place in the information sciences. A perusal of Man- 4
In his thesis Pritchard [9] investigated the question fred Kochen’s book The Small World [44] clearly shows 5
whether it is possible to classify information transfer that more than 10 years ago he was already fully aware 6
networks on the basis of their topological structure. In of the importance of network theory for sociology as 7
his work he noted the close relationship between trans- well as for the information sciences. Finally, in conclu- 8
port geography and information transfer, and made sion, the relationship between networks, percolation 9
successful use of graph-theoretic measures. Appli- theory and the ‘informetric laws’ is pointed out [54]. 10
cations were given for five citation networks: four 1
patent networks and a comprehensive bibliography on 2
bibliometrics. It was found that the bibliometrics Acknowledgements 3
network and the patent networks had different patterns. 4
Martinsons et al. [51] have recently shown that the The first author thanks Peter Van Aelst (UIA) for guid- 5
field of strategic management has entered the main- ance during her master’s work. The authors thank M. 6
stream of social science. They studied the network of Dekeyser for editorial help. 7
journals in the field, using an asymmetric theory 8
(journal-to-journal citations are not symmetric), where 9
the notions of feeder and receiver journals are central. References 20
Network sociologists Haythornthwaite and Wellman 1
[52] used a social network approach in their JASIS [1] J. Kleinberg, Authoritative sources in a hyperlinked 2
environment, Journal of the ACM 46 (1999) 604–632. 3
article studying how work and friendship ties in a
[2] C. Borgman, From Gutenberg to the Global Information
university research group were related to the kind of 4
Infrastructure. Access to Information in the Networked
media used for information exchange. They found that World (MIT Press, Cambridge, MA, 2000).
5
face-to-face means of communication were preferred, [3] C. Lynch, The evolving Internet: applications and 6
supplemented primarily by e-mail. Further, those network service infrastructure, Journal of the American 7
persons who had the most frequent contacts used more Society for Information Science 49 (1998) 961–972. 8
different media. In a somewhat similar vein, [4] L. Björneborn and P. Ingwersen, Perspectives of webo- 9
Kretschmer studied the structure of interpersonal rela- metrics, Scientometrics 50 (2001) 65–82. 30
tionships based on co-authorships [53]. She found an [5] P. Raghavan, Social networks on the Web and in the 1
increasing social distance with declining similarity (as enterprise. In: N. Zhong, Y.Y. Yao, J. Liu and S. Ohsuga 2
measured by co-authorships) expressed succinctly as (eds), Web Intelligence: Research and Development. 3
Lecture Notes in Artificial Intelligence 2198 (Springer, 4
‘birds of a feather flock together’.
Berlin, 2001), pp. 58–60.
5
[6] L. Garton, C. Haythornthwaite and B. Wellman, Study-
ing online social networks, Journal of Computer-
6
8. Conclusion Mediated Communication 3(1) (1997). Available at: 7
www.ascusc.org/jcmc/vol3/issue1/garton.html 8
Social network analysis is a typical example of an idea [7] E. Otte, Social network analysis: the origin, the devel- 9
that can be applied in many fields. With mathematical opment and the structure of the field (in Dutch). Master’s 40
graph theory as its basis it has become a multidiscipli- thesis (UIA, Department of Political and Social Sciences, 1
nary approach with applications in sociology, the infor- Belgium, 2001). 2
mation sciences, computer sciences, geography etc. In [8] J. Markoff, On the Web, as elsewhere, popularity is self- 3
this article the growth of social network analysis within reinforcing, The New York Times 21 June 1999. 4
sociology is documented. Based on the study of a [9] A. Pritchard, On the structure of information transfer 5
networks. M.Phil. thesis (School of Librarianship, Poly-
collaboration network it was found that Barry Wellman 6
technic of North London, London, 1984).
and Patrick Doreian are the most central scientists in 7
[10] S. Démurger, Infrastructure development and economic
the field. These authors have also published several growth: an explanation for regional disparities in China? 8
articles in information science journals. It has been Journal of Comparative Economics 29 (2001) 95–117. 9
shown where social network analysis can be linked to [11] H.D. White and B. Griffith, Author cocitation: a literature 50
work in the information sciences. Now that not only measure of intellectual structure, Journal of the Ameri- 1
computer scientists but also more and more informa- can Society for Information Science 32 (1981) 163–171. 2

450 Journal of Information Science, 28 (6) 2002, pp. 441–453

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
 E. OTTE AND R. ROUSSEAU

111 [12] R.J.W. Tijssen, A quantitative assessment of interdisci- useful metrics, ACM Transactions on Information
2 plinary structures in science and technology: co-classi- Systems 10 (1992) 142–180.
3 fication analysis and energy research, Research Policy 21 [29] L. Egghe and R. Rousseau, BRS-compactness in
4 (1992) 27–44. networks: theoretical considerations related to cohesion
5 [13] R. Todorov, Displaying content of scientific journals: a in citation graphs, collaboration networks and the
co-heading analysis. Scientometrics 23 (1992) 319–334. Internet, Mathematical and Computer Modeling (in
6
[14] G. Melin, R. Danell and O. Persson, A bibliometric press).
7
mapping of the scientific landscape on Taiwan, Issues [30] L. Egghe and R. Rousseau, A measure for the cohesion of
8 and Studies 36 (2000) 61–82. weighted networks. Journal of the American Society for
9 [15] H. Kretschmer, Distribution of co-author couples in jour- Information Science and Technology 54 (2003) (in press).
10 nals: ‘continuation’ of Lotka’s law on the 3rd dimension. [31] W.K. Chen, Applied Graph Theory (North-Holland,
1 In: M. Davis and C. Wilson (eds), Proceedings of the Amsterdam, 1971).
2 8th International Conference on Scientometrics and [32] A. Gibbons, Algorithmic Graph Theory (Cambridge
3 Informetrics (BIRG, University of NSW, Sydney, 2001), University Press, Cambridge, 1985).
4 pp. 317–325. [33] F. Harary, Graph Theory (Addison-Wesley, Reading,
5 [16] M. Mymoon and K.S. Raghavan, Mapping biotechnology MA, 1969).
6 research using co-classification analysis. In: M. Davis [34] J. Scott, Social Network Analysis: a Handbook (Sage,
7 and C. Wilson (eds), Proceedings of the 8th International London,1991).
8 Conference on Scientometrics and Informetrics (BIRG, [35] N. Trinajstić, Chemical Graph Theory (CRC Press, Boca
University of NSW, Sydney, 2001), pp. 447–457. Raton, FL, 1992).
9
[17] P.S. Nagpaul, Visualizing the cooperation networks of [36] R.J. Wilson, Introduction to Graph Theory (Longman,
1120
elite institutions in India. In: M. Davis and C. Wilson London, 1972).
1 (eds), Proceedings of the 8th International Conference [37] B. Rousseau and R. Rousseau, LOTKA: a program to fit
2 on Scientometrics and Informetrics (BIRG, University of a power law distribution to observed frequency data,
3 NSW, Sydney, 2001), pp. 459–468. Cybermetrics 4 (2000) 4. Available at: www.cindoc.
4 [18] Y. Fang and R. Rousseau, Lattices in citation networks: csic.es/cybermetrics/articles/v4i1p4.html
5 an investigation into the structure of citation graphs, [38] C.S. Wilson, Informetrics. Annual Review of Information
6 Scientometrics 50 (2001) 273–287. Science and Technology 34 (1999) 107–247.
7 [19] L. Egghe and R. Rousseau, Co-citation, bibliographic [39] D.J. de Solla Price, Networks of scientific papers. Science
8 coupling and a characterization of lattice citation 149 (1965) 510–515.
9 networks, Scientometrics 55 (2002) 349–361. [40] R. Garner, A Computer Oriented, Graph Theoretic
1130 [20] B. Wellman and S.D. Berkowitz, Structural Analysis Analysis of Citation Index Structures (Drexel University
1 in the Social Sciences 2: Social Structures: a Network Press, Philadelphia, PA, 1967).
Approach (Cambridge University Press, Cambridge, [41] A.-L. Barabási and R. Albert, Emergence of scaling in
2
1988). random networks. Science 286 (1999) 509–512.
3
[21] D. Knoke and J.H. Kuklinski, Network Analysis. Sage [42] S. Milgram, The small world problem, Psychology Today
4 University Paper Series on Quantitative Applications in 2 (1967) 60–67.
5 the Social Sciences, no. 07–028. (Sage, Newbury Park, [43] M.E.J. Newman and D.J. Watts, Scaling and percolation
6 CA, 1982). in the small-world network model, Physical Review E 60
7 [22] C. Wetherell, A. Plakans and B. Wellman, Social net- (1999) 7332–7342.
8 works, kinship, and community in Eastern Europe, [44] M. Kochen (ed.), The Small World (Ablex, Norwood, NJ,
9 Journal of Interdisciplinary History 24 (1994) 639–663. 1989).
1140 [23] H.D. White, Toward ego-centered citation analysis. In: B. [45] L. Björneborn, Small-world linkage and co-linkage,
1 Cronin and H.B. Atkins (eds), The Web of Knowledge Proceedings of the 12th ACM Conference on Hypertext
2 (Information Today, Medford, NJ, 2000), pp. 475–496. (Aarhus, 2001).
3 [24] L.C. Freeman, Centrality in social networks: I. Con- [46] G. Pinski and F. Narin, Citation influences for journal
4 ceptual clarification, Social Networks 1 (1978) 215–239. aggregates of scientific publications: theory, with appli-
[25] S.P. Borgatti, Centrality and AIDS, Connections 18 (1995) cations to the literature of physics, Information
5
112–115. Processing and Management 12 (1976) 297–312.
6
[26] H. Wiener, Structural determination of paraffin boiling [47] R.E. Nance, R.R. Korfhage and U.N. Bhat, Information
7 points, Journal of the American Chemical Society 69 networks: definitions and message transfer models,
8 (1947) 17–20. Journal of the American Society for Information Science
9 [27] D.H. Rouvray, Predicting chemistry from topology, 23 (1972) 237–247.
50 Scientific American 255(3) (1986) 36–43. [48] W.M. Shaw Jr, Critical thresholds in co-citation graphs,
1 [28] R.A. Botafogo, E. Rivlin, and B. Shneiderman, Structural Journal of the American Society for Information Science
112 analysis of hypertexts: identifying hierarchies and 36 (1985) 38–43.

Journal of Information Science, 28 (6) 2002, pp. 441–453 451

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
Social network analysis

[49] E. Logan and M.L. Pao, Analytic and empirical measures organization, Journal of the American Society for 111
of key authors in schistosomiasis, Proceedings of the Information Science 49 (1998) 1101–1114. 2
American Society for Information Science 27 (1990) [53] H. Kretschmer, Patterns of behavior in coauthorship net- 3
213–219. works of invisible colleges. In: B.C. Peritz and L. Egghe 4
[50] E. Logan and M.L. Pao, Identification of key authors in a (eds), Proceedings of the Sixth Conference of the
5
collaborative network, Proceedings of the American International Society for Scientometrics and Infor-
6
Society for Information Science 28 (1991) 261–266. metrics (School of Library, Archive and Information
[51] M.G. Martinsons, J.E. Everett and K. Chan, Mapping the Studies, Jerusalem, 1997), pp. 197–208. 7
scholarly development of strategic management, Journal [54] J. Bogaert, R. Rousseau and P. Van Hecke, Percolation as 8
of Information Science 27 (2001) 101–110. a model for informetric distributions: fragment size 9
[52] C. Haythornthwaite and B. Wellman, Work, friendship, distribution characterised by Bradford curves. 10
and media use for information exchange in a networked Scientometrics 47 (2000) 195–206. 1
2
3
4
5
6
7
8
9
20
1
2
3
4
5
6
7
8
9
30
1
2
3
4
5
6
7
8
9
40
1
2
3
4
5
6
7
8
9
50
1
2

452 Journal of Information Science, 28 (6) 2002, pp. 441–453

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
 E. OTTE AND R. ROUSSEAU

111 Appendix Table A2

2 First authorships in the SNA database
3 Table A1
4 Sociological Abstracts Classification Scheme (sa Author Number Author Number
Classification codes), with details for the field of Complex of of
5
Organization articles articles
6
published published
7 as first as first
0100 Methodology and research technology
8 author author
0200 Sociology: history and theory
9
0300 Social psychology
10 Wellman, Barry 21 Borgatti, Stephen P. 8
0400 Group interactions
1 Burt, Ronald S. 17 Doreian, Patrick 8
0500 Culture and social structure
2 Skvoretz, John 15 Knoke, David 7
0600 Complex organization
Bonacich, Phillip 14
3 0621 jobs, work organization, workplaces and unions
Friedkin, Noah E. 13 8 authors 6
4 0623 military sociology
Everett, Martin G. 11 13 authors 5
5 0624 bureaucratic structure/organizational sociology
Mizruchi, Mark S. 10 19 authors 4
6 0665 social network analysis
Faust, Katherine 9 35 authors 3
7 0671 sociology of business and entrepreneurism
Markovsky, Barry 9 132 authors 2
8 0674 voluntary organizations/philanthropy
Marsden, Peter V. 9 870 authors 1
0700 Social change and economic development
9
0800 Mass phenomena There were, moreover, 578 authors who were never first author.
1120
0900 Political sociology/interactions
1 1000 Social differentiation
2 1100 Rural sociology and agriculture
3 1200 Urban sociology
4 1300 Sociology of language and the arts
5 1400 Sociology of education
6 1500 Sociology of religion
7 1600 Social control
8 1700 Sociology of science
9 1800 Demography and human biology
1900 The family and socialization
1130
2000 Sociology of wealth and medicine
1 2100 Social problems and social welfare
2 2200 Sociology of knowledge
3 2300 Community/regional development
4 2400 Policy, planning, forecasting
5 2500 Radical sociology
6 2600 Environmental interactions
7 2700 Studies in poverty
8 2800 Studies in violence
9 2900 Feminist/gender studies
1140 3000 Marxist sociology
3100 Clinical sociology
1
3200 Sociology of business
2
3300 Visual sociology
3
4
5
6
7
8
9
50
1
112

Journal of Information Science, 28 (6) 2002, pp. 441–453 453

Downloaded from http://jis.sagepub.com at PENNSYLVANIA STATE UNIV on February 5, 2008
© 2002 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.

View publication stats