9781849668200

What is Social Network Analysis?
‘What is?’ Research Methods series

Edited by Graham Crow, University of Southampton
ISSN: 2048–6812
The ‘What is?’ series provides authoritative introductions to a range of
research methods which are at the forefront of developments in the social
sciences. Each volume sets out the key elements of the particular method and
features examples of its application, drawing on a consistent structure across
the whole series. Written in an accessible style by leading experts in the field,
this series is an innovative pedagogical and research resource.
What is Online Research? Forthcoming books:

Tristram Hooley, John Marriott and
What is Qualitative Interviewing?
Jane Wellens
Rosalind Edwards and Janet Holland
What is Social Network Analysis?
What is Narrative Research?
John Scott
Molly Andrews, Mark Davis, Cigdem Esin,
What is Qualitative Research? Lar-Christer Hyden, Margareta Hyden,
Martyn Hammersley Corinne Squire and Barbara Harrison
What is Discourse Analysis? What is Inclusive Research?
Stephanie Taylor Melanie Nind
What are Qualitative Research Ethics?
Rose Wiles
What are Community Studies?
Graham Crow
What is
social network
analysis?
John Scott
First published in 2012 by
Bloomsbury Academic
an imprint of Bloomsbury Publishing Plc

50 Bedford Square, London WC1B 3DP, UK
and
175 Fifth Avenue, New York, NY 10010, USA
Copyright © John Scott 2012
This work is published open access subject to a Creative Commons

Attribution-NonCommercial 3.0 licence (CC BY-NC 3.0, https://
creativecommons.org/licenses/by-nc/3.0/). You may re-use, distribute,
reproduce, and adapt this work in any medium for non-commercial
purposes, provided you give attribution to the copyright holder and the
publisher and provide a link to the Creative Commons licence.
CIP records for this book are available from the British Library and the
Library of Congress
ISBN 978-1-78093-848-6 (hardback)

ISBN 978-1-84966-817-0 (paperback)
ISBN 978-1-84966-819-4 (ebook)
ISBN 978-1-84966-820-0 (ebook PDF)
This book is produced using paper that is made from wood grown in managed,
sustainable forests. It is natural, renewable and recyclable. The logging and
manufacturing processes conform to the environmental regulations of
the country of origin.
Printed and bound in Great Britain by MPG Books Group, Bodmin, Cornwall.
Cover design: Burge Agency
www.bloomsburyacademic.com
Contents
List of Figures vii

Series foreword ix
1 Introduction 1
2 History of social network analysis 7
3 Key concepts and measures 31
4 Applications of network analysis 57
5 Criticisms and frequently asked questions 85
6 Software for social network analysis 103
Bibliography 111
Index 123
v
This page intentionally left blank
List of Figures
1 An acquaintanceship sociogram 10
2 Sociometric structures: star, Y, chain, circle 13
3 Strong and weak ties 15
4 Clique structure 18
5 Overlapping cliques in a social hierarchy 19
6 Blockmodel role structure 21
7 Affective distance in friendship 23
8 Multidimensional scaling of a network 24
vii
Series foreword
The idea behind this series is a simple one: to provide concise

and accessible overviews of a range of frequently-used research
methods and of current issues in research methodology. Books
in the series have been written by experts in their fields with a
brief to write about their subject for a broad audience who are
assumed to be interested but not necessarily to have any prior
knowledge. The series is a natural development of presentations
made in the ‘What is?’ strand at Economic and Social Research
Council Research Methods Festivals which have proved popular
both at the Festivals themselves and subsequently as a resource
on the website of the ESRC National Centre for Research
Methods.
Methodological innovation is the order of the day, and the
‘What is?’ format allows researchers who are new to a field to
gain an insight into its key features, while also providing a useful
update on recent developments for people who have had some
prior acquaintance with it. All readers should find it helpful to
be taken through the discussion of key terms, the history of
how the method or methodological issue has developed, and
ix
x What is social network analysis?
the assessment of the strengths and possible weaknesses of the

approach through analysis of illustrative examples.
The examples considered here relating to the process of
innovation, the nature of relations within a community, and the
operation of power are only a fraction of the social phenomena
whose character can be illuminated by social network analysis,
but they are enough to establish that networks deserve to be
taken seriously by anyone seeking to understand how the social
world works. The book also demonstrates how the growing
methodological rigour of social network analysis has not been
at the expense of comprehensibility; the underlying ideas that
people are connected through networks but that these networks
may be hidden to outsiders remain powerful and persuasive
ones.
The books cannot provide information about their subject
matter down to a fine level of detail, but they will equip readers
with a powerful sense of reasons why it deserves to be taken seri-
ously and, it is hoped, with the enthusiasm to put that knowl-
edge into practice.
Graham Crow
Series editor
1 Introduction
Social network analysis conceptualises individuals or groups as

‘points’ and their relations to each other as ‘lines’. It is concerned
with the patterns formed by the points and lines and involves
exploring these patterns, mathematically or visually, in order to
assess their effects on the individuals and organisations that are
the members of the ‘networks’ formed by the intersecting lines
that connect them. It therefore takes the metaphorical idea of
interaction as forming a network of connections and gives this
idea a more formal representation in order to model structures
of social relations. Treating a social structure as a network is the
cornerstone of social network analysis.
Social network analysis has changed a lot in the last forty years
and the network analyst of today works very differently from
the one of the early 1970s. In this book I will trace these changes
from the early origins of the approach in simple metaphorical
ideas to the contemporary rigour and formalism that character-
ises it. My own involvement in network analysis over this period
has meant that I have myself experienced the changes that have
led to the highly technical and sophisticated measures that now
1
2 What is social network analysis?
face the newcomer. The book can usefully begin, therefore, with
a contrast between how it was done before and how it is done
now.
My personal interest in social network analysis began because
I had an interest in economic power and a fascination with the
work of those who claimed to have identified the key financial
groups in contemporary economies and to have shown the
webs of connection that tied them together through class-based
links of schooling, club membership, and kinship. It seemed to
me that this could best be studied through an investigation of
interlocking directorships and through the use of systematic
network methods. Small studies by Richard Whitley (1973) and
by Phil Stanworth and Tony Giddens (1975) were beginning to
appear at this time, and these authors had drawn diagrams of
board connections among small numbers of companies and
had reported the measured ‘density’ of their networks. These
studies seemed to offer an advantage over the more popular
diagrammatic representations produced by radical journalists
in which the nature of the connections and the implications of
the patterns were not considered. However, I felt that there had
to be a way of extending these systematic methods to the larger
selections of companies that I wanted to study.
The early 1970s were a time at which British sociology was
beginning to make advances in the use of mathematical tech-
niques to study individuals and their relationships. The British
Sociological Association Quantitative Sociology Group had
been formed as a focus for this work and I wrote a short piece
for its newsletter in which I requested help in my search for
useable techniques of social network analysis. This brought me
Introduction 3
into contact with leading figures in North America, Britain, and

continental Europe: Barry Wellman, Joel Levine, Mike Schwartz,
Frans Stokman, Rob Mokken, Clyde Mitchell, and Tony Coxon
all made the ‘first contact’ between 1975 and 1976 that allowed
me to boldly go across this new frontier of sociological methods.
From these people I learned ideas, acquired programs, and
received much needed intellectual support. I learned of possible
measures and of computer programs that could, with some diffi-
culty, help me in my task.
Before I could undertake any social network analysis, however,
I had to collect some data. My data collection had begun
some years previously when I began a project on ownership
and control in the Scottish economy and the involvement
of Scottish companies in the development of the oil and gas
resources of the North Sea (published in Scott and Hughes 1976;
1980). Arriving at a list of companies to study was easy, and it
was then fairly straightforward to compile lists of the names
of their directors. Two sets of record cards were produced: (1)
a set of company cards, one for each company, recording the
names of all the directors in each company; (2) a set of direc-
tor cards, one for each director, recording the names of all the
companies on which the director sat. The second set of cards
was difficult to produce, as it involved searching through all the
company cards to identify any person who appeared on more
than one company card. This involved a manual, error-prone
process of spotting similar names and then searching a variety
of sources in order to discover whether J. Smith, John Smith, H.
John Smith, and Brigadier General H. J. Smith were all, in fact, the
same person or needed to be treated separately. An additional
problem was that company information had been reported in

the directories used at various dates during the year and so John
Smith may have become Sir John Smith, Lord Smith, or even Lord
Tottenham, part of the way through the year. Not for some years
was it possible to enter the company and director names into
a computer and to sort them automatically into alphabetical
order so as to bring similar names closer together in the list.
When a reasonably clean data set had been produced, some
quantitative analysis could begin. This involved manually sorting
the cards into piles on the floor in order to calculate frequencies:
for example, the numbers of directors with 2, 3, 4, or more direc-
torships in the companies studied, and the numbers of compa-
nies with boards of particular sizes or numbers of ‘multiple
directors’. This task, too, was soon helped by my investment in
a pocket calculator. However, many network measures that are
now standard could not be produced in this way and I had to
resort to drawing network diagrams by hand—using long rolls
of wallpaper—in order to map out the principal connections.
These diagrams served an interpretative purpose for me as a
researcher, allowing me to identify the well-connected and the
less well-connected companies and to recognise areas of intense
connection but they were impossible to use as means of rigorous
data analysis or for presentation to others. I explored various
simplified diagrams that gave ‘artist’s impressions’ of the data
in the hope that I could clarify the broad structure of the large
network, but none of these proved satisfactory.
Once I had access to computer programs from the colleagues
who had contacted me, things became a lot easier. For a study
of British companies (Scott and Griff 1984) I transferred the
Introduction 5
data onto punched cards which could then be read into the
university’s mainframe Control Data Corporation (CDC) Cyber
computer. The British project came to form a part of an inter-
national project headed by Frans Stokman, who headed the
team that had produced a network program called GRAph
Definition and Analysis Package (GRADAP), also stored on a
larger CDC computer at Gröningen University. My punched
cards were transferred to a large data tape, and this was posted
to Gröningen for analysis there—no other computer was large
enough to handle the program and its data. A few weeks later
I would begin to receive the output in the form of reams of
computer printout on concertinaed paper.
The output received was, however, a revelation. Most of
the now-standard network concepts could be generated by
GRADAP, and I could begin to report the basic structural features
of my network—a mere three years or so after I had begun to
compile the data. All the researchers in the ten participating
countries worked in the same way and the Gröningen team
produced equivalent output for each participant, so ensuring
strict comparability in the analysis of our various data sets.
Some measures were not, however, included in GRADAP. For
these I relied on a program written by Clyde Mitchell. Clyde
provided me with a copy of the program—CONCOR—on
punched cards and the university computer staff wrote a small
program that could convert my GRADAP files into a form suit-
able for use with CONCOR. The computer demands for CONCOR
were so great that it could run only because my university rented
time on the Manchester super-computer: which was almost as
powerful as one of today’s mobile phones! Each time I used the
program I consumed a half of the university’s weekly budget,

and each time I made a mistake in punching the cards—which
was often—I had to rerun the program, sometimes a week later,
when more budget was available. The physicists who regarded all
computers as their sole property were none too pleased about
this.
Since personal computers became more easily available from
the late 1980s, the research process has changed dramatically.
Pioneers of contemporary social network analysis have produced
cheap software that can be run on any desktop. The initial steps
still require a great deal of manual cleaning of the data, even
when these have been harvested directly from an online data
source. Standard computer software also allows easy conversion
and transfer of data. Analyses that would have required an over-
night run on the Manchester super-computer in 1984 can now
be analysed on a laptop in less than a second. Some networking
procedures are now routinely implemented on mobile phones
on a daily basis by anyone who uses Facebook. Newcomers to
social network analysis are able to start their work more quickly
and more easily than ever before.
These changes are what make this book timely. I have tried
to produce a simple primer that introduces and illustrates the
basic ideas of social network analysis and that will allow users
to quickly use the software themselves and to move on to more
advanced introductions that will give a deeper understanding of
the concepts and measures that are being used.
2 History of social network
analysis
The roots of social network thinking are to be found in relational

or structural approaches to social analysis that developed in
classical sociology. While some approaches to sociology and
anthropology used the ideas of culture and cultural formation
to explain social patterns of feeling, thought, and behaviour, and
others stressed the material environment and the physicality of
the body as crucial determinants, a particularly important strand
of social thought focused its attention on the actual patterns
of interaction and interconnection through which individuals
and social groups are related to each other. In some cases this
involved a conception of the ‘social organism’, or social system,
as a structure of institutions that constrain the subjectivity and
actions of those who occupy positions within those institutions.
For other theorists, however, greater attention was given to the
immediate face-to-face encounters through which individuals
relate to each other and that are constantly refigured through
the actions of these individuals.
It was among the latter theorists that the metaphors of the
social ‘network’ and its equivalents—such as the social ‘web’ or
7
the social ‘fabric’—first emerged. German social theorists such as

Ferdinand Tönnies and Georg Simmel took up this idea in their
‘formal sociology’, seen as a sociology of the ‘forms’ of interac-
tion that carry and contain the diverse subjectively meaningful
contents that motivate the actions of individuals. The translation
of this work and the publication of much of it in the American
Journal of Sociology encouraged many US sociologists of the first
decade of the twentieth century to pursue an ‘interactionist’
approach to social life. Charles Cooley, Albion Small and George
Mead were, perhaps, the leading figures in this movement of
thought. In Germany itself, the sociologies of Alfred Vierkandt
and Leopold von Wiese explored the interweaving of actions
into large-scale social forms such as the market and the state. A
number of these theorists explicitly adopted a terminology of
‘points’ and ‘lines’ to depict the networks of connections that
tie individuals into the ‘knots’ and ‘webs’ of social structure.
Wiese (1931) was, perhaps, among the very first to use these
quasi-mathematical ideas explicitly in a theoretical monograph,
labelling points with letters and referring to the directionality
and circularity of interweaving lines of connection.
Sociometry, small groups, and communities
The earliest empirical work on small groups and communities

occurred in the United States, where researchers with a back-
ground in psychology and psychoanalysis undertook a series of
investigations into the friendship choices made in educational
contexts by school children and college students. Friendship
History of social network analysis 9
choices among classmates were seen as a way of exploring the

cohesion of school class groups and the relative popularity of
particular pupils. Growing out of a long tradition of child study
that had peaked with Stanley Hall’s (1904) developmental study
of adolescence, the earliest published report on friendship
networks was that of Helen Bott (1928), who studied play activi-
ties among nursery school children.
The leading influence in the development of this work,
however, was an Austrian psychoanalyst who had migrated to the
United States. Influenced by the way in which Alfred Vierkandt
had combined a relational focus with a phenomenological
concern for the meanings and emotional significance of relation-
ships, Jacob Moreno devised systematic formal methods for
charting social relations among children. Moreno’s aim was both
to measure and to draw social relations, referring to his work as
‘sociometry’ and to his drawings as ‘sociograms’ (Moreno 1934).
Moreno observed children’s interaction and counted the
numbers of friendship choices made and received by different
class members, combining these into sociograms that depicted
each child as a point and their friendship choices as lines with
arrow heads. These arrow heads showed the direction in which
a choice was made: distinguishing ‘outgoing’ choices directed
at others from ‘incoming’ choices received from others. This
method allowed Moreno to identify the most popular ‘stars’
of attraction and the relative ‘isolates’ who received few or no
friendship choices. An example of one of his sociograms is shown
in Figure 1. Moreno was also able to see whether certain children
attempted to make friends with others but were not able to
secure reciprocal choices from those they sought out. Through
13 14
15 16
7 19
8 18
10 3
23
17
6
21 11
5
24
2
22
4
1 12 20
Figure 1 An acquaintanceship sociogram

(Source Moreno 1934: 145)
compiling the various friendship choices made by class members

into a single sociogram, Moreno aimed to use a sociometric
investigation to model the overall connectedness and emotional
climate of the group (see Jennings 1948).
One influence on Moreno’s work was the tradition of Gestalt
psychology. Kurt Lewin, also a German émigré to the United
States, was even more firmly embedded in this tradition of analy-

sis and pioneered a more general psychology of small groups.
Gestalt psychology involves a focus on the mental structures
that allow people to organise and make sense of their experi-
ences, basing its ideas on observations of non-human primates
(see Köhler 1917). Lewin aimed to translate this basic idea to
the social level, aiming to show that the social structures of
groups are the means through which their actions are organised
and constrained. His investigations into social groups were
concerned, then, with how such structures are produced and
the effects they have on the communication and actions of their
members.
Lewin’s starting point was to see groups as ‘fields’ of interac-
tion—hence his adoption of the term ‘field theory’ to describe his
approach. A group field is the life space within which people act,
and their friendship choices and other social relations are to be
understood as creating forces of attraction and repulsion within
the field that constrain the flow of ideas through the group. A
particular individual, for example, is able to communicate ideas
only through his or her direct contacts or through intermediar-
ies who are able to pass them on. The diffusion of ideas, then,
depends upon the structure of group relations within which the
communicating individuals are located.
The work of Lewin inspired a series of experimental studies
that led to the establishment of ‘group dynamics’ as a specialism
within social psychology (Cartwright and Zander 1953; Harary
and Norman 1953). It was in this specialism that researchers
began to use more systematic mathematical arguments to model
group structure. Using the mathematical approach called graph
theory, which investigates the formal properties of networks, or

‘graphs’, of all kinds, they began to operationalize ideas of the
‘density’ of sociograms and the ‘centrality’ of individuals. In graph
theory, the formal properties of points and lines in a network
become the objects of a mathematical analysis that discloses the
constraints that shape network form.
Researchers in group dynamics constructed formal models of
group structure, such as the star, the ‘Y’, the chain, and the circle
(see Figure 2), and held that these structures had very different
implications for effective communication because some indi-
viduals are in critical ‘central’ positions. If a group is structured
into a long chain of connections in which information is commu-
nicated by passing it through a series of intermediaries, then it is
likely that meanings will become slightly distorted and altered
with each step in the flow of communication. Much as happens
in the children’s game of Chinese Whispers, the message received
at the end of the chain may be quite different from that sent at
the beginning. In a group in which there are many direct connec-
tions and alternative channels of communication, on the other
hand, meanings are less likely to alter as they flow through the
group because the multiple channels introduce ‘corrections’
and so greater conformity in thought and behaviour is to be
expected.
Intermediaries in social groups, especially those at the centres
of ‘stars’, have been seen as the potentially more powerful
members of their groups: they are the influential opinion lead-
ers because of their central locations within the group. Research
in group dynamics has explored the ways in which relations of
dependence within groups can enhance or diminish power and
Figure 2 Sociometric structures: star, Y, chain, circle
foster particular structures of leadership (French and Raven

1959). Other investigators have focused rather more on the
extent to which social relations are reciprocated and, therefore,
the patterns of ‘balance’ and imbalance that characterise differ-
ent groups (Davis 1963). The research of Festinger (1957) linked
this with ideas of subjective mental balance in attitudes and
ideas so as to explore particular patterns of group response
(Festinger et al. 1956).
In the 1930s, George Lundberg (1936; Lundberg and Steele
1938) had extended basic sociometric techniques to the study
of village communities, but it was not until the 1950s that
sociometric techniques really began to move beyond small
and experimental groups to larger groups in real settings. The
Canadian social psychologist Elizabeth Bott—daughter of the
pioneering sociometric researcher Helen Bott—worked at the
Tavistock Institute, where sociometry and group dynamics had
its British base. Here she carried out a comparative study of
working-class and middle-class couples in London. Bott (1957)
showed that members of each class were embedded in different,
class-specific, structures of kinship and friendship and that these

networks influenced their ‘conjugal’ domestic relations within
their households. She examined, in particular, gender differences
in social networks and conjugal roles.
Bott worked closely with social anthropologists who were
already beginning to explore the implications of Alfred Radcliffe-
Brown’s (1940) view that the social relations of tribal societies
could be investigated through constructing models of the
‘structural forms’ of these relations. John Barnes (1954) brought
these ideas together in his report on the communal structure of
the fishing village of Bremnes in western Norway and his work
proved particularly influential in the work in Central Africa that
Max Marwick was establishing as an offshoot of the Department
of Sociology and Social Anthropology at Manchester University.
Clyde Mitchell, Scarlett Epstein, Bruce Kapferer, and others
worked severally on a series of studies of community and
kinship and their effects on gossip and strikes. These studies
were brought together in an influential collection that aimed to
show the power of graph theory as a model for social relations in
complex societies (Mitchell 1969).
From the late 1960s, researchers in the United States took
advantage of the advances being made in the use of computers
for data analysis and began to apply more systematic and rigor-
ous ideas in their studies of larger community and economic
networks. Granovetter (1974) looked at friendship relations as
sources of information about job opportunities and developed
the influential idea that people acquire the most useful infor-
mation from their more distant ties. When a job opportunity
becomes available locally, information flows quickly and rapidly
through the dense and well-connected local networks and

everyone tends to acquire the same information. More distant
job opportunities, however, come to be known about only by
those with generally looser connections beyond their immedi-
ate locality. Those with such connections may, therefore, have a
distinct advantage in their job search activities as they will have
more opportunities than those with only locally dense connec-
tions (see Figure 3). Granovetter (1973) described this as the
thesis of the strength of weak ties.
Figure 3 Strong and weak ties
Wellman (1979; Wellman and Hogan 2006; see also Fischer

1977; 1982) used social survey methods to collect information on
friendship and kinship relations in Toronto, Canada. His aim was
to explore whether individuals relied exclusively on local connec-
tions or were able to keep in touch with those who had moved
away to other parts of the city or country. He was able to exam-
ine people’s range of immediate social contacts, the frequency
and perceived intensity of these ties, and the opportunities
provided by the telephone and the car to maintain contacts over
extended distances.
Bearden and his colleagues (1975; and see Mizruchi
1982; Mintz and Schwartz 1985) investigated corporate
board-level connections in top US companies, while Helmers

and his colleagues (1975) undertook a similar study in the
Netherlands and rapidly extended this into an international
comparison (Stokman et al. 1985). This research highlighted the
relative ‘centrality’ of banks and financial companies in corporate
networks and the changing relationships between financial and
industrial companies in major economies. They documented
structures of coordination and communication among large
business enterprises and pointed to their effects on economic
performance and class cohesion.
Cliques, roles, and matrices
The work of the Manchester anthropologists studying African

societies was not the first attempt by social anthropologists
to investigate social networks. Lloyd Warner worked in the
Durkheimian tradition of Radcliffe-Brown and had undertaken
a conventional study of kinship among native Australians before
moving to the United States and joining with the psychologist
Elton Mayo to undertake an anthropological study of a Chicago
factory and its workforce. This pioneering use of anthropological
techniques to study ‘advanced’ societies proved important in
generating an alternative to purely sociometric studies.
Industrial psychologists working at the Hawthorne electrical
works in Chicago had been undertaking experimental studies
on the effects of physical conditions on work satisfaction and
output. They had discovered that improving the heating and
lighting conditions, allowing rest periods, and other physical
alterations to the workplace improved worker morale and led to

greater productivity. They were confused to find, however, that
similar changes occurred when physical conditions were restored
to their original state or even allowed to deteriorate. Uncertain
how to interpret these findings, they called on Mayo and Warner
to advise them. It was soon concluded that the workers were
responding to the experiment itself rather than to the changes in
physical conditions. The managers had specially selected a group
to study, had located them in an area of the factory separate
from other workers, and had, for the first time, seemed to show
an interest in their welfare. This phenomenon became known as
the ‘Hawthorne effect’ in experimental studies.
In arriving at these conclusions, Mayo and Warner pursued
their own observational and experimental studies in the factory.
Of particular importance was their observational study of a
wiring room where they observed friendly and hostile interac-
tions, cooperation, and offers of help (Roethlisberger and Dickson
1939). Some of their findings were reported as sociograms,
though they seem to have been inspired by the electrical wiring
diagrams that abounded in the factory rather than by published
sociometric work. Their most important work, however, used
tabular or matrix representations to depict the formation of
‘cliques’. Presenting observed interactions in a table in which
the rows represent individuals and the columns represent the
occasions in which they participate, allowed the researchers to
identify those individuals who interact frequently and the occa-
sions or circumstances in which they interact. They identified a
clique in the wiring room that comprised those who tended to
help each other and a set of isolates who found it difficult to get
I1 I3
W1 W2 W3 W4 W5 W6 W7 W8 W9
S1 S2 S4
–
Figure 4 Clique structure
Source: Roethlisberger and Dickson (1939: 507)
help when it was needed and drew this as a diagram (see Figure
4). Cliques were seen as formal representations of the commonly
expressed idea that people may feel they are part of an ‘in-crowd’
or that they are outsiders to it.
Warner decided to pursue similar issues through a community
study in the New England town of Newburyport. Regarding this
as a typical American town with its roots in the early colonial
period, he referred to it as ‘Yankee City’ in his published studies
(Warner and Lunt 1941; 1942; Warner and Srole 1945; Warner and
Low 1947; Warner 1963). Later in the 1930s he supervised a similar
study in the equally old southern town of Natchez, referring
to this as ‘Old City’ (Davis et al. 1941). In these studies, Warner
and his teams explored the formation of ‘cliques’, understood as
informal communal groupings based on feelings of intimacy and
solidarity and that existed alongside the formal associations of
church, business, leisure, and politics. This relationship between
communal and associational patterns, informal and formal ties,
can be seen in relation to Tönnies (1887) earlier suggestions on

the relationship between Gemeinschaft and Gesellschaft.
The Warner research showed that people tend to be members
of numerous overlapping cliques and that it is through the
intersection of cliques that overarching structures of community
solidarity and cohesion are produced. Warner and his colleagues
argued that examination of the pattern of rows and columns in
a data table can show the existence of cliques and the relation-
ships among them. Instead of using sociograms, they reported
the cliques in Venn diagrams in which circles and ellipses repre-
sented sets of interacting individuals, as they had done in the
Hawthorne studies (see Figure 5). At an aggregate level for the
whole community they allocated individuals to a social class and
then represented each social class as a row in a matrix of connec-
tions among the various cliques. This procedure allowed them to
identify the macro-structural positions found in the communi-
ties. Their research highlighted social divisions of both class and
ethnicity, showing the existence of a rigid ‘colour line’ separating
black and white communities.
Figure 5 Overlapping cliques in a social hierarchy

George Homans (1950) undertook a systematic review of small-

group studies, aiming at a theoretical synthesis of ideas. At the heart
of this synthesis was his use of sociometric ideas of the frequency
and direction of social relations, but he also explored the matrix
methods used by Warner for clique identification. Revisiting the
analyses of informal interaction undertaken in Natchez by Warner’s
team, he began to develop a systematic method of matrix analysis
that has since become a central part of social network analysis.
Homans looked at the meetings of 18 women at 14 social events,
as represented in an 18 × 14 matrix, and claimed that a simultane-
ous manual rearrangement of rows and columns could bring out
the internal clique structure. A matrix is typically arranged in an
arbitrary order, listing individuals and events alphabetically or, at
best, chronologically. Shuffling the order of rows and columns until
a strong pattern appears in the cells discloses the structure that is
hidden by these arbitrary arrangements.
This kind of trial and error rearrangement is slow and cumber-
some, even for a relatively small group, and Homans’s method
was not pursued until much later. It did, however, point the way
towards systematic investigations of structural positions within
networks of the kind that were being suggested by the social
anthropologist Nadel (1957). Aiming at a formal, mathematical
approach to anthropology, Nadel showed that algebraic meth-
ods for set analysis could be used to model the ‘roles’ within a
social structure. Where sociometric uses of graph theory focused
on the actual interactions of particular individuals, algebraic set
theories focused on the positions and roles that these individuals
occupied and at role relations at the level of the social structure
as a whole.
Algebraic and matrix methods were developed by Harrison

White and others (1963; Lorrain and White 1971) and they were
able to take advantage of advances in computing to undertake
matrix rearrangements for large-scale social networks. In an
approach that they called ‘blockmodelling’, they saw blocks of
cells in a matrix as representing the structural positions that
Nadel had sought (White et al. 1976; Boorman and White 1976).
The individuals who occupy each position are ‘structurally equiv-
alent’ to each other: they are, for network purposes, interchange-
able, and their individual characteristics and connections can be
disregarded. Thus, all fathers may be expected to relate in similar
ways towards their sons and daughters, while all teachers may
be expected to relate in similar ways towards pupils. The char-
acteristics of roles are social facts that are independent of the
particular attitudes and outlooks of their individual occupants.
Fathers Mothers
Children Teachers
Figure 6 Blockmodel role structure
Space and distance
The sociometric uses of graph theory measured the ‘distance’

from one individual to another by the number of links that
must be traversed to connect the two. This is a useful measure
of closeness, but it does not correspond to the everyday idea

of distance as something measured across a physical space. In a
sociogram, the physical arrangement of points is arbitrary, limited
only by the aesthetic attempt to minimise overlaps among the
lines. A measure of physical distance, however, requires a non-
arbitrary representation of the data. The distance between two
towns in miles, for example, can be measured ‘as the crow flies’
rather than by traversing a network of roads (of varying length)
and intersections. A number of network researchers have, there-
fore, attempted to construct models of ‘social space’ in which
straight line, ‘crow flies’ distances can be measured (Bogardus
1925; 1959). Embedding a network of connections in such a space
allows both the pattern of connections and relative distances to
be studied.
In some applications, a simple idea of perceived social distance
has been used. Individuals may be presented with a simple
friendship chart (see Figure 6) and asked to plot the position
of those they know in terms of their subjective or emotional
distance: close friends, people who you are acquainted with on
a day-to-day basis, or more distant friends. The resulting chart
gives a visual representation of a person’s social world in terms of
the degree of intimacy they have with various numbers of other
people (Wallman 1984: 61, 66–7; Spencer and Pahl 2006). These
charts have provided a useful approach to affective personal
networks.
A more formal idea of social space was inherent in the early
work of Lewin, but it really developed as a formal method in
psychology. Psychometric studies of attitudes had used methods
of ‘scaling’ to show the relative strength of attitudes and this led
self
people you
know well
people you see

in the street
distant friends
Figure 7 Affective distance in friendship
to attempts to measure two or more attitudes through the inter-

section of their scales in a cognitive space that could be taken to
represent a structural part of the mind. Such approaches were
called smallest space analysis (SSA) because they involved an
attempt to define the smallest number of dimensions (scales)
that would represent a particular cluster of attitudes. This
approach was generalised as multidimensional scaling (Kruskal
and Wish 1978) and began to be applied to social phenomena
as a way of reporting structural features of social space (Coxon

1982; and see applications in Hope 1972).
Of particular importance in plotting networks in social space
have been the methods of factor analysis and principal compo-
nents analysis, both of which attempt the construction of a
smallest-dimension social space in which clusters of individuals
or positions can be represented. The most recent development
of this approach has been multiple correspondence analysis
(Rouanet and Le Roux 2009), a version of which was used in the
stratification studies undertaken by Bourdieu (1979). Computer
programs now produce multidimensional scaling images that
show the actual social distance between points, as shown in
Figure 8.
Multidimensional scaling was applied in a study of communi-
ties by Edward Laumann (1966; 1973; and see Laumann and Pappi
1976). Focusing on positions rather than individuals, Laumann
Figure 8 Multidimensional scaling of a network

traced the friendship patterns of those in particular occupa-

tional categories. The frequency of friendship ties among pairs
of positions was taken as a measure of the distance between the
positions and computerised techniques were used to generate
an overall space in which the classes could be mapped according
to their friendship distance. This was, therefore, an attempt to
measure differential association among social classes. Laumann’s
study generated a three-dimensional model of community
structure in which the same social classes could be represented
as clouds or clusters of points.
One of the most influential studies that used multidimensional
scaling was that of Joel Levine (1972) on interlocking corporate
directorships. Using data on large US banks in 1966, Levine
constructed measures of similarities in patterns of connection
around each of the principal banks. He subsequently used this
method to produce a comprehensive ‘atlas’ of corporate connec-
tions (Levine 1984).
Advances in computer software have made it very easy to
carry out multidimensional scaling. One of the leading software
programs (PAJEK) uses a spring-embedding technique to posi-
tion a network in a multidimensional space and allows the
network to be rotated for the visual inspection of its structure.
Dynamics and social change
The majority of the early studies of social networks have been

static and descriptive. They have reported on the features of
social networks as they exist at a particular time, but they have
not generally attempted to explore the internal dynamics that

lead a network to change from one state to another. Where
a concern for time and change has been apparent, this has
involved the construction of a series of cross-sectional studies
with processes of change imputed but not directly studied (see
Scott and Griff 1984; Scott and Hughes 1980). A move towards
dynamic models is a relatively recent phenomenon and has
come about through work by physicists who have been unaware
of previous research by social psychologists, social anthropolo-
gists, and sociologists.
Motivated by an apparent decline in fundamental theoretical
problems in physics itself, a number of physicists have explored
the possible extensions of mathematical models from physics
to other intellectual fields. Barabási (2002) has been the leading
advocate of the application of physical models to social and
economic phenomena, seeing himself as a pioneer in virgin terri-
tory (Scott 2011b). Highlighting the importance of a paper by
Watts and Strogatz (1998), Barabási has produced an approach
that, for all its numerous limitations, does provide some novel
ideas that have helped to bring about a greater awareness of the
importance of dynamic analysis.
Work in this area has used Stanley Milgram’s (1967; Travers and
Milgram 1969) studies of ‘small worlds’ to explore the limits to
certain kinds of variation in network structure. Milgram was inter-
ested in the fact that strangers are often able to identify mutual
acquaintances or connections and will exclaim ‘what a small
world!’ To explore this phenomenon he carried out an experi-
ment in which he asked volunteers to pass a message to a named,
but unknown, individual in another country. The volunteers were
instructed that they must pass the message only to a person

known to them and that this second person must also pass the
message to a known individual. The message must, therefore, be
passed forward in the manner of a chain letter. Each person who
receives the message is instructed, however, to pass it on to an
acquaintance who they feel is likely to be able to get the message,
directly or indirectly, to the target individual. Milgram found that
messages could typically get from source to target through an
average of six connections, or five intermediate individuals. This is
the now-famous idea of ‘six degrees of connection’.
Watts and Strogatz began to explore the mathematical
properties of the networks in which these experimental
conclusions hold. They showed that only certain kinds of
networks have these small-world properties and that many
of the measures used in social network analysis depended on
their presence in the networks studied. Their focus of interest
was, therefore, on variations in network structure and shifts
of state from a small-world network to either more or less
densely connected networks. Duncan Watts (1999; 2003) has
shown that small-world properties exist in networks that are
clustered into zones of relatively high density and by a differ-
entiation between strong and weak ties. In such a network, the
overlapping of connections is so great that the line distances
between points are optimally low. A small-world graph
contains many ‘redundant’ links such that points tend to be
connected through multiple alternative paths. Watts showed
that small changes in the connectivity of such networks can
significantly alter their properties if these changes occur close
to the threshold levels that define the small-world conditions.
Radical structural changes can therefore follow from minor

local-level changes.
Structural change, then, is seen as resulting from the local-level
making and breaking of connections and occurs as an unintended
consequence of those actions. These kinds of change have been
modelled in agent-based computational models that aim to
simulate agent decision making and so to trace change over time.
This work developed independently of work by physicists (see
Snijders et al. 2010) but has rapidly been recognised as providing
an essential element in the dynamic models proposed by the
physicists. Tom Snijders’s model depicts individuals as rule follow-
ers who make or break their social relations according to particu-
lar decision rules. Individuals act ‘myopically’, without awareness
of the larger consequences of their actions (which are, typically,
unknown and unpredictable to them). Individuals acting in this
way produce incremental linear changes in the overall structure
of the network. When their actions reduce the number of redun-
dant links beyond a certain point, however, change may be radi-
cal and non-linear. At critical threshold points there is what Watts
has called a ‘phase transition’ that disrupts the capacity of the
network to continue developing as before. The diffusion of inno-
vations and the flow of capital, for example, may be completely
disrupted by such a transition. When local-level actions increase
the number of redundant links beyond an upper threshold point,
on the other hand, the network becomes so highly connected
that ideas and resources can rapidly spread through the network
at such a rate that all positional advantages are lost.
I have reviewed the history of social network analysis through
tracing developments in relation to a number of specific
methodological approaches. The earliest of these was graph

theory and the associated techniques of sociometry. This
provides an intuitive modelling of social networks and allows
a number of fundamental and advanced measures of network
organisation to be calculated. The second mathematical
approach I considered was the algebraic use of sets and matrices
to uncover the structure of positions and roles within a network.
While this approach is perfectly compatible with graph theory,
it highlights quite distinct sets of issues. Next I looked at spatial
models that emulate geographical mapping techniques to
produce spatial configurations of points. These techniques allow
a move away from the arbitrary configurations of sociometric
sociograms and towards more visually meaningful arrange-
ments. Finally, I considered some novel approaches to network
dynamics that make it possible to build on the static models
of graph theory, matrix algebra, and multidimensional scaling
and to construct accounts of structural change and network
development. In the chapter that follows, I will introduce the key
concepts employed in these approaches, and in Chapter 4 I will
look in more detail at some of the applications of those concepts
in substantive studies.
Further reading
Scott, John 2012. Social Network Analysis. Third Edition. London:

Sage. Chapter 2 gives a fuller account of the history of social
network analysis.
Prell, Christina 2012. Social Network Analysis: History, Theory and

Methodology. London: Sage. Chapter 3 provides an alternative
account of this same history.
Freeman, Linton C. 2004. The Development of Social Network
Analysis: A Study in the Sociology of Science. Vancouver:
Empirical Press. The definitive and standard full-length history
of social network analysis.
Bott, Elizabeth 1957. Family and Social Network. London:
Tavistock Publications. A good example of a classic early study
using network ideas.
Fischer, Claude S. 1982. To Dwell Among Friends: Personal
Networks in Town and City. Chicago: University of Chicago
Press. A more advanced middle-period study of personal
networks.
3 Key concepts and measures
Chapter 2 traced the history of social network analysis through

the three broad mathematical approaches that have domi-
nated the field: graph theory, algebraic approaches and spatial
approaches. In this chapter I will first consider the principal
methods of data collection for social network analysis, and I
will then outline and define, in sequence, the key concepts and
measures associated with each mathematical approach consid-
ered in Chapter 2. I will suggest that graph theory provides the
formal framework common to all these approaches. I will not
present highly technical definitions, as these are more appropri-
ate to the various handbooks of social network analysis (Scott
2012; Degenne and Forsé 1994; Prell 2012). Having done this, I will
set out some of the statistical procedures used in assessing the
validity of network measures in actual situations.
31
Collecting network data
Relational data for social network analysis can be collected

through a variety of methods. These include asking questions
about the choice of friends, observing patterns of interaction,
and compiling information on organisational memberships from
printed directories. All such methods require that an appropriate
selection of cases be made from the total range of possible data
sources. In many studies this will involve a complete, or near-
complete, enumeration of a whole population. For example, an
investigation into the interlocking directorships within the ‘top
200’ companies in an economy will typically involve the collec-
tion of data on all 200 companies. Similarly, an investigation into
the friendship patterns of the children in a school class will wish
to collect information on all the children in the chosen class.
This emphasis on complete enumeration reflects, in part, the
small size of the groups involved and the relative ease with which
complete data can be collected. More importantly, however,
it reflects the difficulty of using sampling techniques with
relational data. Many concepts and measures in social network
analysis, as will be seen, require the use of complete data on all
network linkages, and only a relatively few measures relating
to individuals can be assessed from sampled data. In general it
can be said that individual-level or ego-centric measures can be
generated from sample data, but that macro-level, structural
measures can be produced only with considerable difficulty
and uncertainty, or perhaps not at all. While this difficulty is not
absolute, it is a major limitation and much research is currently
Key concepts and measures 33
being undertaken to find ways of making more reliable infer-

ences from sample data on social networks.
Once relational data have been acquired, they can be prepared
for use in social network analysis. In the case of very small
networks, this will generally involve purely manual methods as
the analyst can easily keep track of the various individuals and
organisations involved. Even where the number of cases reaches
20 or 30, manual analysis of the data may still be possible. In the
case of large networks, however, analysis by computer will gener-
ally be necessary and the data will need to be stored within the
formatted files of the particular software that is to be used.
Data must be sorted on a case-by-case basis, as each case will
be represented by a ‘point’ in the sociograms and the network
analyses. Typically, each point and its connections are listed as
a numbered row in a data file. For example, each individual may
be listed with the organisations or meetings in which he or she
is involved. Each such point can be given an identifying refer-
ence label, perhaps the name of the individual or an abbreviated
form of this, and the label can be stored as part of the row or
in a linked file. The various organisations or meetings in which
the individuals are involved will also need to be given a label so
that this can be married up with the individual data. In some
analyses, an investigator will be concerned with the connections
among the cases themselves: for example, with the connections
among individuals who are involved with the same organisa-
tions. In these situations, the rows in the file will need to list the
other individuals to which each individual is connected, and this
kind of file can easily be generated from the original list by the
standard software programs.
Such a data file might show, for example, eighteen women and
their participation in a number of social events. Each woman
would be represented by a row in the file, and the row would
show the particular events that she has attended. From this
initial file it is possible to produce a file listing the women and the
others who they meet at the various events, and also a file listing
the various events that are similar because the same women
attend them. Similarly, a file for the analysis of interlocking
directorships would show each company as a row, with the rows
listing the directors of the company. From this can be produced
listings of individuals and their meetings with each other and of
companies and their board-level connections to each other. A
file showing the friendship choices of schoolchildren would typi-
cally record the connections among the children directly, each
row showing the other children that are chosen as friends.
Standard social network analysis packages can read these
so-called linked lists and can convert them into the various other
files that are required for specialist analyses. This normally occurs
behind the scenes with there being little or no need for the
investigator to directly manipulate the files. It is to the particular
concepts and measures available in these packages that I now
turn.
Graph theory and egocentric measures
True to its origins in sociometry, uses of graph theory tend to

employ the language of points and lines to describe the patterns
of connection in a social structure. Its various concepts and
measures aim to convey the properties of the visual image of

a sociogram. Visualisation is difficult for large and complex
networks, but the visual imagery behind the basic concepts of
graph theory can help us to imagine the more complex struc-
tures of which they are the building blocks. Sociograms are arbi-
trary aesthetic representations that maintain only the patterns
of connections and not any information about physical distance
and relative location. The mathematical language of graph
theory is, similarly, an attempt to grasp and summarize merely
the pattern of connections.
In the simplest cases, any particular point can be connected
to any or all of the other points by single lines. The total number
of lines connecting an individual is termed the ‘degree’ of the
corresponding point. Where the lines represent, say, relations
that result from the presence of two people at the same event or
their shared membership of an organisation, the lines are said to
be undirected. However, not all social relations are of this simple
type. Where friendship choices are being analysed, for example,
choices may not be reciprocated and so the ‘direction’ of the line
needs to be considered. Thus, choosing a friend can be seen as a
line directed outwardly from the individual making the choice to
the person chosen. Similarly, the receiving of a friendship choice
from another can be seen as an inwardly directed line. A graph
containing directed lines is technically termed a digraph.
Of course, social relations may not all be equally salient, and it
is often necessary to have some measure of the intensity or value
of the lines. Thus, identifying someone as a ‘close friend’ involves
greater subjective feelings of emotional intimacy than is the
case for a ‘distant friend’ or an ‘acquaintance’. It is conventional
to assign a numerical value to a line in order to represent the

intensity of the relationship, producing a valued graph. Any
such numerical value is likely to be arbitrary, being based on the
analyst’s best guess at the strength of the relationship. Typically,
intensity has been measured on an arbitrary 1 to 9 scale, with
a value of 9 representing the most intense relationship. One
particularly important measure of value is the ‘multiplicity’
of the line. This is a measure that recognises that two people
who are each members of three organisations may need to be
represented by a line with multiplicity 3, meaning that they have
a closer or more intense relationship than two people with only
one organisational membership in common.
The discussion so far has assumed that social relations are
positive: that people like each other or find their common
organisational memberships significant. However, many social
relations are negative or involve hostile attitudes: people dislike
each other, reject friendship choices, and so on. Where it is
important to distinguish positive and negative relations in a
network, a mathematical sign (+ or –) may be attached to the
line and the resulting graph is called a signed graph.
These basic concepts of direction, value, and sign allow a
mapping of the myriad ways in which individuals and organi-
sations relate to each other. They grasp the basic qualities of
interpersonal relations and the more abstract features of macro-
structural relations. The most fundamental measures are the
so-called egocentric measures that are based on a particular
focal individual or organisation, referred to as ‘ego’. An indi-
vidual’s immediate friendship choices can be represented as a
number of outgoing lines, the size of a person’s friendship set
being measured by the number of lines—technically referred

to as the ‘outdegree’ of the point. In the same way, friendship
choices received are represented and measured by the ‘indegree’.
Relations of hostility can be measured as the negative outdegree,
and all these relations can be regarded as varying in the strength
or intensity of the measured lines. Thus, an individual’s imme-
diate field of interaction—his or her ‘neighbourhood’—can be
fully charted with the basic concepts of graph theory.
Not all of the relationships of an individual or organisation
are immediate or direct. Some social relations are indirect,
involving intermediary points. For example, an individual may
not know a particular other, but may ‘know a man that does’.
Thus two individuals may be indirectly connected through
one or more common neighbours. In the words of an old song,
a young debutante may sing that she has danced with a man,
who danced with a girl, who danced with the Prince of Wales.
Such relationships are more distant than are direct relations,
and graph theory uses measures of ‘distance’ to represent this.
As I have already noted, this measure departs from the assump-
tions of spatial distance and is calculated by the number of
lines needed to connect two points. Ego is connected to others
at varying distances according to the number of lines that run
between them. A friend of a friend is connected by two lines
and so is said to be at distance 2 from ego. Thus, the singer of
the song was connected to the Prince of Wales at distance 3.
Distances greater than 2 typically lose their significance for the
focal individual very rapidly, and such egocentric measures need
to be used with care. Nevertheless, it can be important to know,
for example, the size of a person’s neighbourhood at distance 2,
as this will include a number of potential intermediaries who can

connect them to new contacts or sources of information. This
was the basis of Granovetter’s (1973) important argument about
the strength or importance of relations through relatively distant
acquaintances.
It is important, of course, to take account of the direction, sign,
and value of lines in measuring distance. If the lines connecting
two individuals vary in direction, it may make no sense at all to
measure the overall distance by the number of lines. Similarly, if
the sign of a second line is the opposite of the sign of the first,
then the intermediate point may be a barrier to communica-
tion rather than a facilitator of it. Finally, lines of high value may,
other things being equal, bring points ‘closer’ together than lines
of low value. These remarks highlight the fact that the proper-
ties of a social network can never be inferred simply from the
mathematical measures that it is possible to make. It is always
necessary to ensure that the measure can be given a substantive
sociological significance. A connection at distance 2 can easily be
given such an interpretation, but a connection at, say, distance 10
may not be at all significant for the individuals or organisations
concerned. Such judgements of significance will vary from one
type of relationship to another. Distance in a network of friend-
ship may be more significant than, say, distance in a network of
credit relations, though this may not always be the case. Thus,
a distant and apparently trivial link held in the computer of a
credit-rating agency may affect an individual’s credit rating quite
significantly. The interpretation of any distance measure depends
on the expert judgement of the analyst and the purposes of the
research.
The last of the egocentric measures to be reviewed here is the

neighbourhood density. This is a measure of the extent to which
the immediate contacts of a point are mutually connected
with each other. Density is what Bott (1957) referred to as the
‘connectedness’ of a person’s network and is high whenever,
for example, many of a person’s friends are also friends of each
other. In an egocentric network with lower density, a person’s
various friendships are more segmented. For example, if four
people are all directly connected to each other, rather than being
connected only through the focal ego, then ego’s neighbourhood
has a density of 100 per cent. This measure of density is more
typically represented on a scale of 0 to 1, with complete 100 per
cent density being represented as a value of 1. Many of the quali-
fications entered about line distance apply to the calculations of
neighbourhood density, especially where neighbourhoods are
defined at line distances of two or greater. Even in the simpler
case, however, it is important to take account of the direction,
sign, and value of lines in measuring and interpreting density.
Other issues in the measurement of density will be discussed
later in this chapter.
Graph theory and global measures
While egocentric measures are always focused on a particular

point—ego—the overall, or global properties of a network
can also be assessed. The connections of particular individuals
concatenate to produce a more or less cohesive structure of
relations that is unlikely to have been intended by any of the
participants. Graph theory provides a set of concepts for analys-

ing these macro-level properties of social networks.
In order to measure global properties, however, it is not gener-
ally possible to rely on sample data. The egocentric network
properties of a large number of individuals can be measured
from a sample survey so long as all the normal issues of represen-
tativeness are considered. In the case of most global properties,
however, the very act of sampling will result in a loss of the infor-
mation that is required. Global properties are not mere aggre-
gates or averages, as is usually the case in sample data, but are
themselves structural features that are apparent only if a whole
network is studied and not simply a sample drawn from it. This
is not true for all global measures, and work is currently going
on to find a solution to this problem. However, this is an issue
that always needs to be borne in mind when seeking to measure
global properties.
The most fundamental global properties of graphs are degree-
based measures. The idea of line distance, for example, is an
important global measure and may be more useful at the global
level than it is in egocentric networks. The various lines connect-
ing two points form a ‘path’ whose length can be defined by
the number of constituent lines that chain together to form it.
Movement through a network depends importantly on the path
distances involved. Thus, the communication of ideas or innova-
tions may occur through a whole chain of intermediaries with
little or no reliance on the participants all having direct personal
knowledge of each other. This was the key to Milgram’s discussion
of the ‘small world’. Some studies of the diffusion of information
have suggested that information degrades with distance simply
because each intermediary has less contextual knowledge about

the original act of communication and so may misinterpret or
distort the information. This is the problem alluded to in the
game of Chinese whispers.
Measures of path distance must, of course, take account of
the directionality, sign, and value of the various lines. The latter
is particularly important as it has been seen as a crucial deter-
minant of the efficiency of the network flow and movement
and so as reinforcing any degradation or attenuation effects. A
network in which the strength of lines is constant or increasing
across paths might be expected, other things being equal, to be
an efficient network for communicating ideas or transmitting
financial resources. A network in which path distances involve a
decline in value at successive steps might be expected to be far
less efficient and to result in a significant attenuation of mean-
ings and ideas with distance.
A whole collection of global measures cluster around the idea
of the centrality of points within their graphs. The degree of a
point—its total of incoming and outgoing lines—is the most
basic measure and has been termed local centrality. Calculating
the degrees of all points in a network and ranking them from
highest to lowest gives a rank order of local centrality. This
centrality is ‘local’ because it highlights points that are well-
connected in their immediate neighbourhoods. Such points
may not, however, be central in the more global sense that a
circle or sphere has a unique centre that can be understood in
quasi-spatial terms. Locally central points are well-connected
within particular parts of the network but may not be at all well-
connected in a global sense.
It is for this reason that a distance-based measure of ‘closeness’

has generally been preferred as a measure of global centrality.
The most commonly used measure of closeness involves calcu-
lating the aggregate distances from each point to all others. The
score for any particular point is a measure of how close or distant
the point is likely to be to any randomly chosen point. The point
with the lowest aggregate distance is the most central point in
the network. It will be apparent that all the usual qualifications
apply and that closeness must be calculated with respect to
appropriate indicators of direction, sign, and value.
The converse of centrality is peripherality and it can be useful
to know those points that are least close to other members of
their networks. Such points are not isolated but are poorly
integrated into their network. They are likely to have little influ-
ence and to be uninvolved in significant communication flows
through the network.
A final measure of centrality is one that measures the extent to
which a point is able to act as an intermediary in a large number
of network flows. This measure has been called ‘betweenness’
and refers to the extent to which a particular point is able to
serve as an intermediate point of contact between any two other
points. The person represented by such a point may be able
to control access or the flow of information to others because
of the ‘structural hole’ (Burt 1992) that exists between the two
others that he or she connects. A point has a high betweenness
centrality when it fills a large number of structural holes in the
network. A point with high betweenness centrality may not be
at all central in terms of its local degree or global closeness, yet it
may play an important part in the structure of the network.
A further set of global measures involve a move away from

the properties of points to the properties of the network as a
whole. These measures comprise the inclusiveness, density, and
centralisation of the network.
Inclusiveness simply measures the number or proportion of
the whole set of points that are actually connected into one or
more parts of the graph. Some points may be isolates, having no
ties to other points, while others will be connected, to a greater
or lesser extent, into larger structures. The inclusiveness of a
graph is simply the total number of non-isolated points, gener-
ally expressed as a percentage of the total number of points.
Inclusiveness is a rough and ready approximation to cohesion,
but it is usually more informative to measure the actual density
of the graph. The idea of neighbourhood density has already
been introduced and so the global density of a network should
be easy to understand. Global density is the actual number of
lines connecting the points of a graph (directly related to inclu-
siveness) expressed as a proportion of the total number of lines
that could possibly exist among the points. Thus, the existence of
a large number of isolates reduces the density of a graph. Density
can be calculated for valued and directed lines as well as for lines
in an undirected graph.
The density of a graph is a very useful and direct measure of
its cohesion, but it has one major limitation as a comparative
measure of social structure. In real situations, density varies with
the size of a network and this limits the possibilities of using
the measure to compare different types of network. It is highly
unlikely that agents are able to sustain more than a certain
number of relationships: our ability to be ‘friends’ with people,
for example, has its limits. Other things being equal, then, an
increase in the size of a network is likely to mean a reduction
in its density. This is exacerbated by the fact that the ability to
sustain relationships varies with the type of relationship. We
may, for example, be able to ‘recognise’ far more people than we
would acknowledge as ‘friends’, and we are likely to ‘love’ even
fewer. A comparison of the density of the recognition, friendship,
and love networks, even if they are of a similar size, will therefore
be difficult. Nevertheless, for networks of a similar type, the
global density can be a useful comparative measure if reported
alongside measures of the size and inclusiveness of the graph.
A further measure of global cohesion is the centralisation of
a network. Where centrality relates to the position of particular
points, centralisation relates to the overall structure of a network.
Centralisation measures the extent to which the cohesion of a
network is organised around a specific point or set of connected
points. The spokes on a bicycle wheel, for example, form a highly
centralised network around its hub. Measures of centralisation
can be based on the degree, distance, or betweenness of points,
and extensions of these concepts have involved the idea that it
is possible to identify the sets of points that comprise the centre,
margin, and periphery of the network as a whole.
Graph theory and network differentiation
The global measures considered so far are attempts to measure

different aspects of the cohesion of a graph understood as its
compactness. An alternative strategy is to focus directly on the
differentiation or fragmentation of a graph in order to uncover

the plurality and diversity of its structure. Such a strategy recog-
nises that connected points in a graph may not be connected
as a single whole but may instead be connected into a number
of distinct structural parts of the whole network. Two broad
approaches to the differentiation of networks into ‘subgraphs’
can be identified: the measurement of ‘components’ and the
measurement of ‘cliques’. Each of these types of subgraph has
distinct properties.
A component is a subgraph in which all points are directly
or indirectly connected to each other and there are no connec-
tions to points outside the subgraph. Information or resources
can, therefore, flow along a path through all the members of the
component but cannot reach any other points in the graph. A
graph may comprise one or more components of varying size,
and the number and size distribution of components is a funda-
mental measure of network differentiation and of the existence
of boundaries to the flow of information and resources.
The identification of components must go beyond the simple
components of undirected graphs and must take account of any
direction attached to lines as this will affect the ability of infor-
mation or resources to flow from one point to another. In some
cases it may be legitimate to disregard direction—for example, in
order to measure the ‘awareness’ of agents choosing friendship
partners—and this is what are referred to as ‘weak components’
in contrast to the ‘strong components’ in which all points are
connected through a path of uniformly directed lines.
A further refinement of the simple component idea is that
of the cyclic component built from intersecting cycles of
connection. A cycle is a directed path that returns to its starting

point. The overlapping of such cycles produces a cyclic compo-
nent in which all points are connected by one or more cycles
and no points have cyclic connections outside the component.
A cyclic component is a structural element within a strong
component and may be connected to other members of the
strong component through ‘bridges’ that do not lie on the cycle
itself.
The various types of component can be further analysed by
taking account of degree-based measures such as the intensity or
value of the constituent lines. Such an analysis shows the arrange-
ment of components of varying tightness, one inside another.
For example, if a measure of value has been attached to the lines,
it is possible to slice through the network at various levels of
intensity of connection. Such an analysis discloses the ‘contour’
structure of the network Components of the kind described so
far have taken account of all lines, regardless of intensity and
so are connected at the lowest level of intensity. If lines below
a chosen level of intensity are disregarded, then the higher level
of connection is specified and the analysis moves from the
‘valley’ of the network to its lower slopes. Progressively raising
the chosen level of intensity allows the analysis to approach the
‘peaks’ of the structure, with intermediate levels marking the
gradient of the network. Where points are connected at varying
levels of multiplicity, this multiplicity can be used to generate a
picture of the ‘nested’ structure. Nested components, then, stack
inside each other like Russian dolls.
The other approach to differentiation that I referred to involves
the identification of cliques. A clique differs from a component
in that it is a set of points in which each point is directly and

reciprocally connected to all other points. It represents, for
example, a tight friendship group or gang. The definition of a
clique is, therefore, far more restrictive than that of a compo-
nent, and cliques tend to be smaller subsets within components.
By taking account of the direction of the lines in a clique, it will
be possible to identify strong and weak cliques, and if the values
of the lines are taken into account it may even be possible to
identify a nested structure of cliques.
This idea of the clique is clear and readily interpretable, but it is
relatively unusual to find such a structure in real social networks,
except in those with exceptionally high density. For this reason,
social network analysts have generalised the idea into that of the
‘n-clique’ in which less restrictive conditions apply. This ‘n’ is the
criterion of connectedness in the clique and so altering the value
of ‘n’ allows the identification of looser cliques. In a 2-clique, for
example, all members are connected at distance 2 or less: each
point is either directly connected to all others or is connected
through an intermediary point. Such a concept can be seen as
representing social groups such as those in which individuals are
either friends or friends of friends. Although values of ‘n’ greater
than 2 can be used in clique analysis, they can less readily be
given a substantive sociological interpretation and so may be less
useful.
Various extensions of the clique idea have been suggested,
though they have, so far, been less widely used in sociological
work. The most important of these extensions are the ‘clan’, in
which all points are connected by a path of a specified maxi-
mum length, and the k-plex, in which all points are adjacent to a
specified number of other points. All such measures are available

within the principal network analysis programs, but great care
must be taken to show that they designate meaningful sociologi-
cal concepts before they are used.
A final and more useful extension of the clique idea is that of
the social circle, seen as a set of overlapping and intersecting
cliques (Alba and Kadushin 1976). This concept points to the
importance of chains of indirect connections in tying individuals
into larger structural elements. The idea of the social circle comes
close to the informal idea of the clique originally identified by
Warner and his colleagues in their community studies.
Algebraic measures of network structure
Algebraic methods represent the points of a graph in sets that

correspond to social ‘positions’ and that can be handled in equa-
tions or analysed in matrices. I have traced the origins of this
approach to social network analysis in the works of Warner and
Homans. The most important recent developments in algebraic
network analysis have been those inspired by Harrison White
and his work on ‘structural equivalence’.
Structurally equivalent points are those that can be regarded
as substitutable for each other: they occupy similar structural
positions in relation to others, though they may be connected to
different others. Culturally defined and institutionalised roles—
such as father, mother, child, teacher, and so on—are the clearest
examples, but the idea also involves purely relational categories
such as classes or political factions that generate structural
uniformities of experience and action. The approach taken by

Lorrain and White (1971) was to represent agents and their affili-
ations or participations in a matrix and then to simultaneously
rearrange both the rows and the columns until blocks of simi-
larly connected points appear. This ‘blockmodelling’ reduces the
structure of the whole graph to a structure of social positions,
such as a role set or a class structure.
The rearrangement of a matrix involves one or another form
of cluster analysis (Everitt 1974). Methods of cluster analysis
combine or separate points according to their similarity, and in
White’s approach the similarity between two points is measured
by the correlation found in their patterns of connection within a
graph (White et al. 1976; Boorman and White 1976). The method
used by White and his colleagues to identify structurally equiva-
lent positions was termed CONCOR and is available within
the standard social network analysis software programs. The
boundaries of clusters can sometimes be identified visually—as
was the case in Homans’s reanalysis of the Old City data—but
in larger and more complex networks they must be identified
from the patterns of high and low density found in the various
blocks. Once a repeated cluster analysis has produced a simpli-
fied pattern of zero and non-zero blocks of density, a reduced
network can be drawn in which non-zero blocks are represented
as super-points and zero blocks are represented by the absence
of connections. This is referred to as an ‘image graph’ and has
been seen as a delineation of the positional structure of a social
network.
An alternative matrix method is that of Doug White and
Karl Reitz (1983), which they call regular equivalence. Structural
equivalence takes account of direct connections, but regular

equivalence takes account of the wider patterns of connection.
It looks at similarity of connections at varying specified path
lengths. The resulting matrix of similarities can be analysed
through similar cluster analytic methods to yield image graphs.
This looser measure of substitutability is, arguably, more useful
for much social network analysis.
Algebraic methods offer powerful measures for the analysis
of social structure (see Pattison 1993), but they are difficult to
apply in highly fragmented and differentiated networks. They
may be usefully applied within components or fully connected
graphs, but are impossible to apply between components where
no connections exist.
Spatial and cartographical approaches
Spatial approaches such as smallest space analysis and multidi-

mensional scaling are attempts to plot points in a spatial field
through geometrical methods that allow the construction of
sociograms as true ‘maps’ showing a configuration of points
that depicts their relative physical distance and spatial direction
as well as retaining information on the familiar path distances
and line directions. As with blockmodelling, spatial approaches
start out from measures of similarity (or dissimilarity) but use
cartographical techniques to produce a best-fit configuration in
which the relative similarities among points correspond directly
to the relative physical distances that can be shown on paper or
a computer screen.
Central to these methods is the search for an appropriate

number of dimensions in which to embed the configuration.
A maximum of two dimensions can be directly represented
on paper, as is the case with an atlas of physical maps showing
places in relation to the north/south and east/west dimensions.
Computer software, however, allows the depiction of three-
dimensional configurations on screen and allows a configuration
to be rotated to show the shape and solidity of the networks
from various angles. Such representations give clear visual
impressions of the density, centralisation, and fragmentation of
the networks. It can be difficult to represent more than three
dimensions in physical form, and it may be necessary to resort
to simplifications such as, on paper, presenting the successive
two-dimensional views of a three-dimensional structure.
However many dimensions are used, there is always the ques-
tion of providing a sociological interpretation of each dimen-
sion. It is rarely the case that a configuration will take a directly
geographical form and place points in a north/south and east/
west two-dimensional space. In social space, dimensions will tend
to show scales of economic or political inequality, civic participa-
tion, religious affiliation and so on. The key task in spatial analysis
of any given configuration is to rotate it in its space in order to
yield the most meaningful interpretation of the arrangement of
points. Such an interpretation is never given by the method used
but must be made and supplied by the analyst.
The configurations produced by these geometrical methods
retain the ideas of relative distance and spatial direction that
were highlighted in Lewin’s work, and they avoid the arbitrari-
ness that is involved in the classic attempts to draw sociograms.
It is also possible to show the lines that connect points within a

spatial map, with the length of each line reflecting the ‘as-the-
crow-flies’ distances between points. Path distances, which in
basic graph theory simply represent the number of links, become
measures of actual spatial distance along the paths that connect
the points. Thus, spatial models combine spatial distance with
accurate representations of the ‘routes’ that must be taken in the
communication of information or resources through a network.
Notions of centrality may also be clear from the patterns of
lines, though globally central points may not appear as spatially
central to the whole configuration.
It is also possible to represent components, nesting, and
structurally equivalent positions within a spatial configuration.
Component boundaries can be drawn around its connected
points and measures of the value of the constituent lines can be
used to draw the nested boundaries within this. Such a diagram
corresponds closely to the depictions of hills and valleys on a
map by the drawing of contour lines. Such an approach gives a
ready depiction of the topography of a social space and can give
a useful way of interpreting its structure. Positional ideas may be
more difficult to represent with such clarity, as the members of
the various blocks may not be spatially contiguous. Nevertheless,
the use of different colours to represent the various blocks will
bring out the equivalence patterns and so help to identify the
relationships between block membership and the membership
of cliques and components.
These uses of spatial analysis have been made possible by
advances in computing. These have allowed the handling of large
data sets and the easy and quick production of sociograms, but
they have also allowed the use of powerful on-screen methods

for visualising the social networks generated. Standard software
allows the inspection of three-dimensional configurations
and allows the points to be coloured to indicate their proper-
ties. More specialised programs allow moving pictures to be
constructed so as to depict patterns of change over time.
Statistical inference
Much work in social network analysis has been both static and
descriptive. Recent work, however, has shown how it is possible
to move towards dynamic understandings of social change and
to construct explanations of the patterns described. This has led
to much greater attention being given to statistical methods for
assessing the significance of results and the validity of explana-
tions. This work follows the general and well-established prin-
ciples of statistical inference and hypothesis testing, adapting
these to the specific requirements of using relational data.
A hypothesis to explain observations must be produced by
expert use of the sociological imagination, but any hypothesis
proffered must be tested before it can be accepted with any
confidence. Statistical methods of hypothesis testing involve
comparing the observed results from particular patterns of
change to the results that might be expected to occur as a result
of random variations alone. That is, the methods attempt to
show the likelihood that any result could have occurred simply
by chance. If this probability is low, then some confidence can be
taken in the mechanisms that have been hypothesised.
A number of approaches to the significance testing of rela-

tional data have now been proposed, the most influential and
important being the so-called exponential random graph models
(ERGM) of Wasserman and his colleagues (see, for example,
Wasserman and Pattison 1996; Wasserman 1980). This approach
uses regression techniques to compare an observed pattern with
a range of possible graphs produced through random simula-
tions from the same data. Confidence in the significance of the
observed results is greatest when it corresponds with a randomly
generated pattern that has a very low probability of occurrence
by chance.
Further discussion of statistical inference is beyond the
scope of this book. The method will, however, be of increasing
importance in social network analysis, which has constantly
been bedevilled by the critical comment: ‘So what?’. Critics have
pointed to the fact that simply describing an observed configu-
ration of points and lines means nothing unless a plausible
sociological interpretation can be given to it. Statistical methods
will be an essential element in avoiding this critical response in
future.
Further reading
An appropriate working strategy for building on the ideas

outlined in this chapter would be to work through the following
sources in sequence:

Sage. The principal chapters in this book cover all the
concepts and measures in greater detail and are designed to
be read by those with little or no technical background in
social network analysis.
Scott, John and Carrington, Peter C. (eds.). 2011. The Sage
Handbook of Social Network Analysis. London: Sage. A
comprehensive collection of introductory material that
elaborates on basic concepts with greater background.
Section 3 covers key methodological ideas.
Wasserman, Stan and Faust, Katie 1994. Social Network Analysis:
Methods and Applications. New York: Cambridge University
Press. The best and most advanced text that gives a thorough
and comprehensive discussion of techniques of social network
analysis.
Carrington, Peter J., Scott, John and Wasserman, Stanley
(eds.). 2005. Models and Methods in Social Network Analysis.
Cambridge: Cambridge University Press. A collection of
specialised papers reviewing recent advances and leading-
edge developments in network methodology.
Background sources on all key areas of network analysis have
been brought together in various compilations and source
books:
Scott, John (ed.). 2002. Social Networks, Four Volumes. London:
Routledge. Volumes 1 and 2 include key sources on concepts
and techniques.
Holland, P. and Leinhardt, S. (eds.). 1979. Perspectives on Social
Networks. New York: Academic Press. A shorter collection
with a number of important source papers.
4 Applications of network
analysis
My outline of the key concepts and measures has, inevitably,

been rather abstract and formal. I want now to make these ideas
more concrete by illustrating their use in empirical studies that
have applied them in relation to substantive sociological issues.
I will do this through a consideration of three of the areas that
have made the most extensive use of social network analysis.
First, I will look at work on the diffusion of ideas and practices
through social networks, showing how the structure of a network
can shape the flow of information and resources. Secondly, I look
at studies of the scholarly networks created through the citation
of scientific papers, training at graduate research centres, and
participation in scientific conferences and workshops. I will show
that these studies have produced mappings of the intellectual
space within which scientific production takes place. Thirdly, I
will examine studies of corporate power through investigations
of interlocking directorships. I will explore issues of centrality
and its relation to corporate power and the existence of bank-
centred cliques and clusters.
57
Diffusion and the flow of information and resources
Diffusion studies are concerned with how the flow of informa-

tion and attitudes about new practices and techniques are
shaped by the structures of the networks in which intercom-
municating individuals are involved. The earliest writer to
highlight the need to investigate diffusion in this way was the
French lawyer and criminologist Gabriel Tarde, who saw imita-
tion as the basic psychological mechanism responsible for this
diffusion. In his Laws of Imitation (Tarde 1890), he set out an
account of the factors that constrain imitation and explored the
consequences of these for the spread of new ideas and technical
innovations. Interaction, he argued, is grounded in the natural
tendency for individuals to imitate the behaviour of those who
are psychologically close to them and with whom they identify.
Thus, interaction is necessarily a process in which individuals
behave, intentionally or unintentionally, in ways that those they
encounter may either take up or ignore. Innovations made by
one individual are, therefore, subject to selective retention and
replication, much as genetic variations are selectively replicated
in Darwinian theory.
Tarde saw chains of such processes of imitation as the basis
on which ‘rays’ or ‘waves’ of innovation spread from focal
innovators to permeate a whole network. As more and more
individuals adopt the innovation, it is able to spread in multiple
and intersecting waves that are constrained in their movement
by the paths and blockages inherent in a given pattern of social
connections. Cultural transmission is not, therefore, a simple and
unproblematic process but is complex in both its causation and
Applications of network analysis 59
its consequences. One of Tarde’s principal conclusions was that

the rate at which an innovation spreads can be described in an
‘S’ curve in which the number of those who adopt an innovation
increases slowly at first but then takes off exponentially until
it again slows down at the point where few potential adopters
remain. Tarde highlighted the importance of those who are
regarded as exemplars or role models. These are the people that
individuals look up to and respect and who are especially likely
to be imitated. These individuals are the influential ‘opinion lead-
ers’ in the process of diffusion.
Tarde’s suggestions had some influence in studies of political
communication and opinion formation, but he had little wider
impact until the 1940s. It was then that the full range of his ideas
began to be appreciated in a small number of studies in rural
sociology. The most important of these studies was that of Ryan
and Gross (1943), who studied the release of a strain of hybrid
corn to farmers in Iowa. Tracking its adoption from its initial
release in 1928, they showed how it eventually revolutionised
farming techniques in the State. The hybrid seed was developed
by research scientists at Iowa State University and was promoted
through the advertising and sales campaigns of the seed compa-
nies. Take-up of the corn, however, depended on the massive job
of persuading farmers to purchase new seed each year, rather
than the traditional practice of relying on a saving of seed corn
from the current year’s crop. The new seed would be successful
only if farmers could be persuaded to make a substantial and
continuing investment in the purchase of seed. Successful adop-
tion of the innovation therefore required a significant change in
behaviour.
Ryan and Gross showed that the rate of adoption over the
period of the study followed the ‘S’ curve suggested by Tarde.
Although sales publicity could make most farmers aware of the
new products, there were initially only a small number of early
adopters. Most of those who eventually used the new corn
did so only after discussion with neighbouring farmers who
could persuade them that using the new corn was worthwhile.
Persuasion through discussion, rather than simple imitation,
was the key to adoption, and people were especially likely to be
persuaded by the early adopting opinion leaders whose views
they valued. When a number of neighbours had adopted it
and advocated its use, other farmers were very likely to take it
up. Thus, the initial spread was slow until there were sufficient
adoptees for many farmers to have at least one adopter in
their neighbourhoods. Take-up then accelerated rapidly as the
number of new adopters increased across the State. Eventually,
however, a point was reached at which few non-adopting farm-
ers remained and the rate of adoption slowed down.
Only in the 1960s was there any advance on this work. James
Coleman and his associates (1966) undertook a study of the
introduction of the new antibiotic Tetracycline (referred to
in the study as ‘Gammanym’) and of its adoption by general
practitioners. Though unaware of the earlier work of Ryan and
Gross, they also discovered both that the pattern of adoption
followed an ‘S’ curve and that the opinion leaders were of key
importance in this process. Coleman also found that those with
the highest neighbourhood degrees, as measured by hospital
affiliation, attendance at staff meetings, and sharing an office
with other doctors, were more likely to be early innovators and
were most likely to be named as sources of information or as

friends by other doctors. These were the people who became
early innovators through a ‘chain reaction’ or contagion effect
that ran through their well-connected neighbourhoods. They
subsequently became influential contacts working through what
Granovetter was to term the ‘weak ties’ and so were responsible
for the spread of the innovation to other parts of the network
where local opinion leaders sponsored its take up and influenced
others in their area.
A few years earlier than the Coleman study, Everett Rogers
(1962) had undertaken a systematic review of research on inno-
vation and had drawn this together into a systematic review of
the area. Initially publishing this work in 1963, it took its definitive
form in the Third Edition of 1983 and appeared in its Fifth Edition
in 2003. Rogers showed that the overall inclusiveness, fragmenta-
tion, and density of interpersonal networks determine the extent
of exposure people have to new ideas and ways of behaving.
The relative location of individuals within this network—their
peripherality, the size of their neighbourhood, their involvement
in cliques and clusters—determine the extent and salience of
their exposure. He highlighted a social process in which there
were distinct phases of knowledge, persuasion, and decision:
individuals who adopt an innovation must become aware of it,
form an attitude towards it, and decide to adopt it.
Rogers argued that individuals are likely to become aware of
new ideas that meet their needs or interests as a result of their
exposure to the information that flows through a social struc-
ture. While people may sometimes actively search out innova-
tions, they are more typically dependent on formal and informal
messages that come to them through their regular channels of

communication. This is how they may even be made aware of
new products and services that they did not previously realise
they ‘needed’. Those who acquire knowledge early—the ‘early
knowers’— are those who have a large number of social contacts
and a wide sphere of interaction through which they can reach
them. Mass media channels are, however, of greatest importance
in the knowledge phase of the diffusion process.
Rogers rejected the dubious psychological assumption of a
natural propensity to imitate. An idea is taken up only if individu-
als are persuaded to act on their knowledge: they must develop
a favourable attitude towards it through assessing its likely pros
and cons. This is most likely to occur when a person is aware that
others who he or she have cause to trust are considering it or
have already adopted it. The more of a person’s contacts that
are in this state, the more likely is he or she to form a favourable
attitude. It is only at this point that an explicit decision to adopt
or not to adopt is made, with collective pressure being especially
important in bringing about conformity with the evolving deci-
sions within a person’s sphere of contacts. The more people an
individual is aware of who have made a decision to adopt, the
more likely is he or she to follow suit.
Within this process a key role is played by the opinion leaders
that occupy central positions within networks. They are the key
determinants of the rate of adoption because their position in
their social network makes them critical to the flow of informa-
tion through it—making them influential for a large number of
people—and because their respected status means that their
opinions carry a great deal of weight (which can be represented
by the intensity of the lines connecting them to others). Thus,

diffusion must be studied in relation to the structures of the
neighbourhoods in which people are embedded. The three-fold
process of knowledge, persuasion, and decision proceeds itera-
tively and cyclically, such that the critical mass of adopters builds
up and larger and larger numbers of individuals are exposed to
the pressures that encourage them to adopt an innovation. This
explains how the observed exponential shape of the ‘S’ curve is a
direct consequence of this ongoing process.
This approach to innovation has been applied in a number of
recent studies. In an investigation of the spread of Christianity,
Rodney Stark (1996) asks how it was that a tiny and obscure
Messianic movement on the edge of the Roman Empire with a
maximum of around 1000 adherents in 40AD was able to dislodge
pagan beliefs and grow to more than 33 million adherents by
350AD, the year that the Emperor Constantine converted to
Christianity and allowed the religion to become the dominant faith
of the western world. The annual growth rate for contemporary
religious movements is around 40 per cent and Stark assumed this
to be the case for early Christianity. Using this figure, he showed
that a constant and realistic annual growth rate does, indeed,
produce an exponential absolute growth on the scale suggested
by the evidence. The growth of Christianity followed an ‘S’ curve,
with the crucial upturn occurring between 250AD and 300AD.
In order to explain this pattern of growth, Stark drew on
the idea of differential association. Conversion to a new
religious group, he argued, is more likely when people have
stronger attachments to existing members than they do to
non-members. It is the balance of attachments rather than
doctrinal commitment, that explains the propensity to listen,

join, and then become committed. Thus, the basis for the growth
of a religious movement is to be found in the social networks
of direct and intimate interpersonal attachments. Individuals
are more likely to respond positively to those with whom they
are close and share attitudes and outlook. When recruited, they
are especially likely to bring in or ease the recruitment of their
immediate family members and intimate associates. Successful
religious movements are those that avoid becoming ‘closed’ and
are able to reach outsiders, especially through their weak ties.
Stark showed that early recruitment to Christianity occurred
through the networks of the Hellenised Jews, who were already
somewhat detached from Palestinian Judaism but were also
marginal to Greek society itself. It was through their networks of
mutual support and social solidarity that Christianity was able
to grow through the Synagogue communities. Thus, conversion
in the earliest years was most marked in such cities as Caesarea,
Damascus, Antioch, Corinth, Ephesus, and Rome. These were the
larger cities of the Graeco-Roman world, many being the dias-
pora cities of Asia Minor, where Jews constituted a large enough
group to form a ‘critical mass’ for conversion. These cities were,
however, relatively disorganised, as they were newly re-estab-
lished and with colonial and migrant populations. They were
also relatively ‘open cities’ in which many people would have had
weak ties to other such cities. It was here that Christianity had
its greatest appeal as providing an answer to the prevalence of
deprivation and suffering.
Oliver and her colleagues (Oliver et al. 1985; 1988b; Oliver et
al. 1988a; Marwell and Oliver 1993) have developed an approach
that extends the standard model of diffusion to the intermittent

and cyclical processes that allow the formation or strengthen-
ing of social movements and the organisations that carry their
ideas forward in concerted joint action and collective protest.
Following Roger Gould (1993), they argue that ‘closed’ organisa-
tions—those that are relatively self-contained clusters of related
individuals formed into overlapping cliques—can develop and
sustain the solidarity required for joint action far better than can
‘open’ organisations. Thus, a cell structure may be an appropriate
organisational form for a radical political group. They show also
that the diffusion processes in social movements can generate
periodic ‘spikes’ of activity and a cyclic rise and fall in protest
activity, especially where the extent of diffusion is strongly
influenced by news media coverage (Oliver and Myers 2003).
Gould himself (1995) applied similar ideas in his own study of the
formation of popular protest movements in nineteenth century
Paris.
These diffusion ideas were brought into the mainstream of
formal social network analysis by Ronald Burt (1987; 2005), who
connected social influence with analyses of social capital. Tom
Valente (1995) has usefully summarised these arguments in a
formal extension of Rogers’s argument.
Citation studies and the sociology of science
A long tradition of research has outlined the use of publication

patterns and, especially, citation patterns, to map forms of social
organisation in science. Building on suggestive research into the
importance of ‘scientific communities’ and groups of scientists

in the formulation and growth of scientific knowledge (Kuhn
1962; Price 1963), Diana Crane (1972) was one of the first to use
ideas from diffusion research to explore scientific innovation, the
formation of scientific specialisms, and their basis in processes
of recruitment, promotion, and co-authorship. While she used
sociometric measures, derived from Coleman’s software, she
concentrated on diffusion and scientific growth and did not
report her findings in what is now standard sociometric format.
Nevertheless, she produced a pioneering study that set the base-
line for later investigations into scientific networks.
Crane showed that the growth of knowledge within a scien-
tific specialism exhibits the characteristic ‘S’ curve of diffusion
and asked what it is about scientific communities that can
explain this. She pursued this question through a study of two
specialisms: the rural sociologists concerned with agricultural
innovation and mathematical work on finite groups. Authors
of published papers in each of these areas were the basis for a
mapping of networks of the scientific relations resulting from
involvement in informal discussions, published collaborations
(co-authorship), relations with teachers, and the influence of
colleges on the selection of research problems and methods.
Crane found a close correspondence between the structures
found in each of these forms of relationship: between two
thirds and three quarters of all authors were connected into a
large, low density component, and up to a half of the authors
reported that they thought these relationships were important
in their research. They formed a ‘circle’, as defined by Kadushin,
in that they were loosely bounded structures of communication
and awareness whose members are geographically dispersed and

so are largely engaged in correspondence, co-publication, and
intermittent and infrequent face-to-face discussion. A majority
of the members of each specialism were, directly or, more typi-
cally, indirectly in contact with a majority of the other members.
Groups of associates were identified on the basis of those who
were named as collaborators in scientific projects. These were
seen as the ‘invisible colleges’, the virtual laboratories and work
groups through which research is organised. The number of
publications produced by an individual was found to be the basis
of the longevity of their career as a member of the specialism
and of their ability, therefore, to influence norms and practices
within it. Isolates and members of small cliques were less likely
to be regarded as influential within the specialism. Successful
specialisms were those with dominant—central—individuals
who acted as the ‘leaders’ who trained or collaborated with a
large proportion of other members. Crane concluded that the
peripheral and isolated members of scientific communities must
rely on formal mechanisms of communication rather than the
informal networks of communication and so are less able to
keep up-to-date and are less likely to make significant contribu-
tions to their field.
Shortly after Crane’s study an exploration into the structure
of theoretical debates in American sociology was undertaken
by Nicholas Mullins (1973), using somewhat more systematic
network methods. Mullins saw theoretical ‘schools’ of thought
as invisible colleges formed from relations of communica-
tion, co-authorship, apprenticeship, and colleagueship. A new
theoretical orientation, he argued, develops from an initial
loose stage in which work is carried out by isolated theorists at

a number of different universities, through a ‘network stage’ in
which key publications become the focus of work by people in
frequent communication with each other, and a ‘cluster stage’ in
which groups of students and colleagues form around the key
specialists in a small number of universities. Finally, a theoretical
area may reach the ‘speciality stage’ in which students who have
begun independent careers carry the work to a larger number
of institutions and build up their own links to new students and
junior co-authors.
Theoretical groups were identified from a reading of the
literature and discussion with colleagues, and Mullins under-
took separate network analyses for each of the eight areas that
he chose to study. Data on training and careers were collected
from interviews and from information included in directories,
footnotes, and other published sources. Data on co-authorship
came from the indices of the American Journal of Sociology and
the American Sociological Review, together with the actual books
and articles produced.
Mullins’s first report was on what he called ‘Standard American
Sociology’, perhaps the most successful of all theoretical special-
isms. This grew from the initial suggestions of Parsons (1937) to
become the mainstream of American sociology through the
1950s and 1960s. This entered its network stage in 1935, with a
particular focus on Harvard and Columbia, and entered its cluster
stage a decade later. Under the intellectual leadership of Parsons,
a cluster of 13–15 students, teachers, and researchers forged
the central elements of the structural-functional approach.
From 1951 it expanded its influence at other universities and
established itself as the dominant theoretical orientation in the

United States.
Turning to the secondary tradition of symbolic interaction-
ism that emerged at Chicago and reached its specialty stage in
1952, Mullins shows this to be an extensive but somewhat looser
network of researchers. The network of co-authorship included
five components, the largest including thirty-seven members.
Central participants were Ernest Burgess, in the earliest stages,
and then Anselm Strauss, Howard Becker, and Everett Hughes.
This core group of symbolic interactionists, like two of the smaller
components were connected to authors from within the struc-
tural functionalist group, and Mullins took this and the fragmen-
tation of the network as indicators of a relatively loose structure.
This structure can be contrasted with ethnomethodology,
which emerged around Harold Garfinkel’s seminars at UCLA in
the mid-1950s, expanded through Berkeley and Santa Barbara,
and began a diffusion of influence as a specialism in 1971. The
group was much smaller than the structural-functional group
and the network of the invisible college was far more frag-
mented. The co-authorship network, for example, comprised
seven components, three of which were simply pairs. The largest
component included nine authors and had its focus in Aaron
Cicourel, who co-authored with all of the other eight members.
Indeed, without Cicourel, the large component would have
shrunk to a triad.
A particularly influential use of this approach to intellectual
development was Randall Collins’s study in The Sociology of
Philosophers (2000), in which he provided a theoretical rationale
for network studies. Collins started out from the assumption
that individual thought in philosophy and science can be under-

stood as the internalisation and coalescence of ideas circulating
through the networks in which scientists and philosophers are
involved. An intellectual may work physically alone, as a lone
scholar, but always within a social context that means that he
or she is never mentally alone. Thus, ‘Thinking is a conversation
with imaginary audiences’ (ibid., 52). Collins therefore argued for
the need to study intellectual activity through what have been
called the collaborative circles (Farrell 2001) and other networks
of association through which cooperation, influence, and argu-
ment take place.
This provides the rationale for selecting the particular rela-
tions that have been studied in the investigations of the invis-
ible colleges. Ideas are communicated face-to-face in meetings,
congresses, lectures, and workshops, through systematic and
informal training, and in written texts such as articles, mono-
graphs, textbooks, working papers, and so on. Changes in
communications technologies supplement these relations with
new forms of e-mail and on-line communication. These various
networks of linkage are embedded within ‘organisational bases’
that make communication possible: systems of higher educa-
tion, publishing, retailing, and professional organisation that
sustain their reproduction as an intellectual way of life. Studying
such processes over a period of more than two thousand years,
from Classical Greece to the present day, involves considerable
problems of data consistency and reliability. Collins uses much
qualitative and, it must be admitted, impressionistic data, aiming
to compile as much source material as possible on intellectual
apprenticeship and teaching, publication, recruitment and
promotion, and debate, and attempting to document negative,

conflict relations as well as the more positive ones. Collins’s
approach depends upon the ad hoc discovery of personal links
and on his ability to infer intellectual lineages from his knowl-
edge of the philosophical positions.
Collins’s work can best be illustrated by taking his analysis of
the period from 1765 to 1935: the period since Kant and in which
the universities underwent a radical transition from a patron-
age system to the contemporary system of research universities
(Collins 2000: Chapters 12 and 13). This was the period in which
the influence of idealist philosophers who had broken with theo-
logical ideas spread with the German model of the university and
gave birth to a whole swathe of new philosophies. Collins traces
the massively influential links from Kant to his pupil Herder and
to his close friend Goethe and then from these to Fichte, who
initiated a fully idealist system and was a central figure in influ-
encing Schelling, Schopenhauer, and Hegel. Conflict between
Hegel and Schopenhauer followed a split in the network that
presaged a decline in the influence of Schopenhauer at the same
time as a massive growth in the influence of Hegel. Hegel’s influ-
ence grew with the prominence of Berlin in the German univer-
sity system and was the basis for expansion abroad.
In Britain, Collins argues that university reform was marked
by the growth of idealism into a philosophical culture hitherto
dominated by the utilitarianism of non-professional philoso-
phers. Developing through Jowett, Green, and Bradley at Oxford,
it became the dominant approach to philosophy through the
interlinked careers of their students Caird, Seth, and Bosanquet.
Critical attacks from Bradley on other philosophers fragmented
the expanding system into a number of distinct clusters of ideas

and individuals: the evolutionary and positivistic thought of
Huxley, Lewes, and Eliot, and, through the central position of
Russell, the soon-to-be-dominant school of analytical philoso-
phy. The latter included Russell’s teacher Whitehead, Moore,
Keynes, and members of the tightly knit friendship clique of the
Bloomsbury Group, and it influenced the independent line taken
by Russell’s student Wittgenstein.
Collins’s work, then, provides a clear theoretical basis for the
use of citation and other data on scientific linkages to map intel-
lectual networks. His own work, inevitably, faced data problems
that meant he needed to take a more impressionistic approach
to data analysis than had been the case in the studies by Crane
and even by Mullins. However, recent work by Howard White
has brought such work into the mainstream of social network
analysis and established a clear basis for both theoretical and
methodological advance.
White and his colleagues (2004) studied the ‘Globenet’
research group, a pseudonym for a small international and inter-
disciplinary research group. The researchers aimed to investi-
gate both the social and intellectual structures of the Globenet
group and so looked at citation practices among members,
together with such social ties as friendship, research contacts,
and advice-seeking, and they also undertook some interviews.
The core of their analysis, however, concerned citation patterns.
Taking four time periods from pre-1989 up to 2000, they looked
at the changing patterns of both ‘intercitation’ (the citation
of each other) and of co-citation (joint citation of the same
source).
The study found that although three quarters of members

did not cite each other, there was a growth in the overall level
of intercitation over the whole period. Similarly, the density of
the network increased substantially. The distribution of intercita-
tions was found to be scale free, a small number of researchers
accounting for a large proportion of all citations. Those who
engaged in mutual citation tended to rate each other as ‘friends’,
were involved in direct collaboration, worked in the same disci-
pline, and knew each other from before Globenet was set up.
The editors of a collectively produced book were central to the
network and also played a leadership role within it. It was found
that they had especially high levels of out-citation, showing that
they took seriously the need to solidify the group by affirming its
collective identity in their own publications.
Co-citation patterns tended to reflect disciplinary differences,
though some individuals appeared as ‘interdisciplinary linchpins’
(White et al. 2004: 120). The extent of cross-disciplinary cita-
tion increased as the project developed. Intercitation increased
with the total amount of scholarly communication and with
communication outside of meetings. It correlated with written
(and e-mail) communication, but not with telephone use, and it
correlated negatively with face-to-face scholarly communication
through attendance at the same conferences and workshops.
White and his colleagues concluded that citation patterns
reflect both the social structure of the research group and its
intellectual consolidation. Over and above the effects of group
leadership, people tended to cite each other’s work when they
were better acquainted with each other. The development of
the Globenet project increased the tendency of its members to
cite the work of other members. It was also found, however, that
there was a tendency for citation to reflect intellectual affinities
of theory or method.
The possibilities of using systematic network methods are
vividly illustrated in work by Moody and Light (2006), who used
citation data to investigate the overall structure of sociology as a
discipline within the social sciences. They look at the co-citation
of social science journals in order to discover how similar any
two journals are in their contents in terms of citations to them
by papers in other journals. The raw data on similarities among
1,657 journals were mapped into a multidimensional space and
the authors devised a way of drawing contour lines that surround
journals with specified levels of similarity. Peaks in the contour
map, they argue, can be seen as ‘discipline’-specific clusters of
journals.
The mapping produced by Moody and Light (2006: 71, Fig 1)
shows strong and closely linked clusters for law and economics,
closely connected to similar clusters for political science. On the
immediately opposite side of the map is the strong cluster of
psychology journals. Spread between law and psychology are a
series of lower peaks corresponding to management and organi-
sational behaviour. In the opposite direction, the ‘foothills’ from
psychology trail through education and social work. Sociology
appeared as a moderately high peak in the dead-centre of the
map, while geography and anthropology appeared as low foot-
hills between sociology and political science. The height of the
peaks in the Moody and Light map correspond to the sharp-
ness of the boundaries defining a journal’s authors, and hence
the boundaries of a discipline. Sociology, they found, is not so
strongly inwardly influenced as psychology or economics, tend-

ing to draw on other disciplines and to contribute to them. By
contrast with economics, sociology has no clear structure of
central or core journals.
From this mapping of social science, they turn to a detailed
mapping of sociology itself, tracing changes in journal connec-
tions over the period 1970 to 1990. This time they looked at the
topics discussed in articles in order to investigate the state of
sociological discourse. Their analysis is, therefore, a conceptual
network of journal content based on a frequency count of
words. They show that in the 1970s, there were strong clusters
for community, education, race, and culture, with each of
these surrounded by looser groupings of topics. By the 1990s
this had completely changed, with strong areas being health,
family, gender, and science-teaching. Sex as a topic of study in
sociology journals had grown considerably and was linked to
a new large peak of articles on AIDS-HIV. All other areas were
much looser than before. Finally, by the late 1990s, the AIDS
peak had shrunk back somewhat and was matched by peaks on
health care, welfare, language, and stratification, together with
interlinked peaks on science, technology, and the sociology of
sociology.
Sociology, then, Moody and Light found to be in constant
interchange with neighbouring ‘stronger’ disciplines, and with
its specialisms restructuring in relation to trends in the outside
world. There is a constant shift in topics of study, although the
body of sociologists producing these may be more constant and
cohesive.
Interlocking directorships and corporate power
Studies on the role of company directors have a long history in

social science, though its immediate origins were in journalistic
and political investigations of the concentration and abuse of
economic power. Directorships have long been seen as sources
of power. A directorship in a company, or corporation, is a posi-
tion at the top of a company that confers legal authority over its
assets and employees on its occupant. Holding directorships in
two or more companies proportionately increases the power of
the individual concerned. Thus, tabulations of the number and
distribution of directorships has been seen as a way of charting
the degree to which corporate power is concentrated in the
hands of a small number of individuals or families.
The earliest studies of this power were those of Otto Jeidels
and John Hobson. Jeidels, a member of a Frankfurt banking
family who became a leading investment banker and himself the
holder of multiple directorships, published a study of board-level
relationships between banks and heavy industry in Germany
(Jeidels 1905). He tabulated the total numbers of directorships
held by directors of each of the big six banks and traced whether
these directorships were held by bankers or industrialists and
so could be assigned a direction: from banks to industry or vice
versa. His report tabulated what he described as a ‘community
of interests’ among the directors at the heart of the German
economy. Hobson, a labour activist and journalist, drew on his
experiences reporting the Boer War in South Africa to add a
section on corporate power to the Second Edition of his study
of The Evolution of Modern Capitalism (Hobson 1906). Taking
up popular comment on the ‘Randlords’, the leading gold and

diamond producers of South Africa, he tabulated the director-
ships held and their formation into an ‘inner ring’ of finance.
The arguments of Jeidels and Hobson were important influences
on Marxist work by Rudolf Hilferding (1910) and Vladimir Lenin
(1917), who documented the emergence of a dominant group
of ‘finance capitalists’ who had become the leading elements in
the various financial groups that pursued strategies of imperial
expansion across the globe.
Hobson, however, had gone beyond numerical tabulation
and had produced a diagrammatic representation of the inner
ring. Most probably inspired by radical metaphors of the
‘webs’ of influence and the ‘tentacles’ of large business groups,
Hobson drew a schematic and simplified diagram in which
circles representing financial groups were connected through
a mesh of criss-crossing lines, to numerous industrial ventures.
Hobson’s suggestive innovation was taken up by a US congres-
sional subcommittee enquiry into the monopolistic power of the
so-called Money Trust (US Congress 1913). The report not only
tabulated the distribution of directorships but also produced
large charts in which the connections among the large compa-
nies and corporate groups were laid out as maps of corporate
power.
The congressional report introduced, or at least popularised,
the term ‘interlocking’ directorships. When one individual
holds a board-level position in two companies, the two boards,
or directorates, are connected. This ‘interlock’ is a social rela-
tion between the two companies and was seen as a channel
of communication and potential influence between business
entities that the law and economic theory regard as separate

sovereign bodies. Thus, numerical tabulations of directorships
can be seen as reports of the quantitative significance of this
structure of power. This assumption became the basis on which
many later investigations by academics, governmental bodies,
and others aimed to investigate the monopolistic powers of
top bankers and industrialists regarded as the elite of financial
controllers.
A key theme in this research was the attempt to discover
and document the particular financial groups that lay at the
heart of modern economies. Variously described as ‘interest
groups’ or ‘empires’ of finance, these groups were not, however,
always defined in clear or strict sociometric terms as cliques or
components. They were seen as groups of companies subject to
common control, but no precise measure of control and hence
of the boundaries of this control were specified. The resulting
indeterminacy is apparent in the fact that studies of the same
economy at similar times have reported different numbers
of financial groups. Thus, Marxist economist Paul Sweezy
(1939) undertook an investigation for the National Resources
Committee of the US Congress in which he documented eight
such groups associated with various of the leading investment
banks. By the 1950s and 1960s, estimates of group structure
varied widely: Perlo reported eight groups but Dooley (1969)
reported 15 and Menshikov (1969) reported 22.
The first study of interlocks to introduce proper techniques of
social network analysis was that of Warner and Unwalla (1967).
Drawing on ideas that had been developing since Warner had
himself introduced the idea of the ‘clique’, Warner and Unwalla
described the structure of the US economy as organised around

a ‘hub’ of finance companies with ‘spokes’ that radiate out into
the wider economy. While this remained a metaphorical descrip-
tion, they did report its structure through tabulations of the
‘direction’ that could be attached to an interlock. When a direc-
tor holds a full-time executive position within a company, which
can thus be regarded as his or her principal business interest,
the interlocks created can be seen as an outgoing relation from
the base company. From the standpoint of other companies it is
an incoming relation from the base company. These two types
of interlock, Warner and Unwalla argued, can have a different
significance for the companies involved. Later research has
distinguished these ‘primary interlocks’ from the ‘induced inter-
locks’ that result among the companies on which the executive
sits as a non-executive director and the ‘secondary interlocks’
that are created by directors who are completely without execu-
tive positions (Stokman et al. 1985).
Over the following years, a number of studies have used
more rigorous sociometric ideas. Levine (1972) and Bearden
et al. (1975) used the new ideas being produced by Harrison
White and his colleagues and students to measure centrality
in corporate networks and the function of the cliques and
clusters formed around central companies. At the same time,
Mokken and Stokman oversaw an investigation of the Dutch
economy (Helmers and others 1975) and put together a large
international research group to study interlock patterns in ten
countries (Stokman et al. 1985). Bearden and his associates used
a measure of centrality based on the idea of closeness and that
took account of direction. They documented the existence of
an extensive national network of predominantly ‘weak’ ties

(secondary and induced interlocks) within which could be found
more intensely-linked clusters based on the ‘strong’ ties (primary
interlocks) created by executives. The clusters were organised
around focal banks and these bank-centred groups were loosely
integrated into an extensive national network.
The group led by Mokken and Stokman used the concept
of the component to identify any clearly bounded corporate
groups that there may be in the Dutch economy. An analysis of
undirected interlocks found a single large component contain-
ing 84 companies. A breakdown into subgroups on the basis of
their densities disclosed a smaller ‘core’ group and a surrounding
periphery. While the core of 17 central companies had a density
of 0.76, the density of the whole network was just 0.19. Thus, they
argued, the 17 central companies formed a densely connected
group of financial companies with influence over the other 67
companies in the large component.
Scott and Griff (1984) used component analysis to show that
the largest 250 companies in Britain in 1976 were formed into
one large component of 185 and just two smaller components.
An analysis based on the value of the lines—the number of
directors in common between two companies—showed the
existence of many more components, the largest containing just
17 companies. A major part of their analysis, however, concerns
cliques identifiable in the network of interlocks carried by
executives. The 156 companies in the large component of the
network of such primary interlocks contained eight 2-cliques
that varied in size from 10 to 15. The central points within each
clique were largely taken by banks or insurance companies.
The British network of 1976, therefore, comprised a structure

of overlapping bank-centred spheres of influence. It was within
and through these spheres that the bank and industry executives
who dominated their boards were able to exercise a degree of
control and influence over the companies and it was through the
overlapping relations of the larger structure of the weak ties that
they could ensure a degree of coordination across the economy
as a whole.
Levine (1972) recognised the problems involved in drawing
sociograms for large networks and so used multidimensional
scaling to chart structures of centrality and influence among
70 industrial companies linked to the 14 largest banks of 1966.
A three-dimensional representation showed that these inter-
locks had a regional pattern, that banks were the most central
companies, and that the third dimension sharply separated the
banks from the industrials. Levine explored this third dimension
further using cartographic techniques. Banks, he argued, stood
at the centres of clusters of influence and these clusters could be
mapped as discrete spaces on the surface of a sphere.
This remarkable burst of research between 1972 and 1985
contained all the key themes to emerge in the social network
analysis of corporate interlocks. Subsequent studies have applied
similar ideas to different economies, or have taken longitudinal
approaches to changes in corporate interlocks over time. One
area of strong development, however, has been the attention
accorded to transnational networks. Meindert Fennema (1982)
had undertaken an analysis of international links and had showed
the persistence of national and regional—generally language-
based—clusterings. The growing globalisation of economic
relationships, however, has led many to investigate whether such

nation-state-centred structures have disappeared. David Smith
and Doug White (Smith and White 1992) pointed the way with
a study of networks of international trading patterns that used
blockmodelling to document the existence of trading blocks
and their relationship to Wallerstein’s world-system categories of
core, semi-periphery, and periphery. This kind of approach was
taken up in interlock studies by Fennema and Carroll (2002) to
investigate the formation and structure of a transnational busi-
ness community of people with attenuated links to particular
national economies.
Further reading
Rogers, Everett 1962. Diffusion of Innovations, 5th Edition. New

York: Free Press, 2003. The definitive overview of diffusion
studies and approaches.
Coleman, James S., Katz, E. and Menzel, H. 1966. Medical
Innovation: A Diffusion Study. New York: Bobbs-Merrill.
Remains a classic study in the application of diffusion
methods.
White, Howard D. 2011. ‘Scientific and scholarly networks. In
The Sage Handbook of Social Network Analysis, John Scott and
Peter J. Carrington, eds. London: Sage Publications. A clear
and definitive account of citation and bibliometric methods.
Mullins, Nicholas C. 1973. Theories and Theory Groups in

American Sociology. New York: Harper and Row. An early
example of work on scientific networks that is especially
accessible for students of sociology.
Pennings, Johannes 1980. Interlocking Directorates. San Francisco:
Jossey-Bass. Chapters 1 and 2 provide one of the few general
accounts of this methodology for studying corporate power.
Carroll, William K. 2004. Corporate Power in a Globalizing World.
A Study in Elite Social Organization. Ontario: Oxford University
Press. A powerful recent example of work in this area.
Applications of social network analysis in other areas and special-
isms include:
Wellman, Barry 1979. ‘The community question: the intimate
networks of East Yorkers’. American Journal of Sociology 84:
1201–31. An important and influential report on community
structure and personal networks.
Werbner, Pnina 1990. The Migration Process: Capital, Gifts and
Offerings among British Pakistanis. Oxford: Berg Publishers.
Uses blockmodelling techniques to investigate migration
patterns.
Gould, Roger V. 1995. Insurgent Identities: Class, Community and
Protest in Paris from 1848 to the Commune. Chicago: University
of Chicago Press. Uses network techniques to investigate class
formation and solidarity.
Bearman, Peter S. 1993. Relations into Rhetorics: Local Elite Social
Structure in Norfolk, England: 1540–1640. Rutgers: Rutgers
University Press. An historical study using legal records to
investigate aristocratic class relations.
A range of specialised papers can be found in:

Scott, John and Carrington, Peter, eds 2011. The Sage Handbook
of Social Network Analysis, London: Sage Publications. Section
2 includes Chapters reviewing work on social support, online
communities, policy networks, terrorist networks, criminal
networks, cultural networks, and many other topics.
Scott, John ed. 2002 Social Networks: Critical Concept, Four
Volumes. London: Routledge. A collection of classic sources.
Volumes 2 and 3 cover family and community; Volume 3
covers corporate power and economic structure, and Volume
4 covers politics, protest, and policy networks.
5 Criticisms and frequently
asked questions
Social network analysis is a collection of concepts, measures,

and techniques for relational analysis. It is an approach that is
specifically designed to grasp the most important features of
social structures and it is unrivalled in this task. It can be used
to explore social relations themselves and also the cultural
structures of norms and ideas that help to organise those rela-
tions in conjunction with material circumstances. The various
applications reviewed in Chapter 4 have demonstrated the ways
in which relational and cultural structures can be investigated
with a few simple network concepts. Theories of social structure
inform and sustain the methods of social network analysis.
The social world, however, consists of more than just social
structures (Scott 2011a), and social network analysis itself cannot
reach beyond the structural concerns for which it was designed.
Many of the issues that arise in the study of social action require
a totally different set of theoretical ideas: ideas concerning
motives, intentions, rationality, emotionality, and subjectivity
together with the consequences, intended and unintended, to
which different types of action lead. Action theories therefore
85
require different methods and techniques—methods of frame

analysis and interpretation, for example—and a crucial ques-
tion in sociology concerns the complementarity or opposition
between theories of structure and action. My own view is that
they are not contradictory or alternative to each other but are
complementary. Indeed, I have shown in Chapter 2 that dynamic
models of network development rest upon a combination of
structural ideas and a theory of action and its consequences.
It is equally the case that cultural analysis involves more than
just the analysis of cultural structure. The study of culture rests
upon ideas of discourse, ideology, representation, and socialisa-
tion that are specific to cultural analysis and require specific
methods of their own: content analysis, narrative analysis, ideol-
ogy critique, and so on. These ideas, too, are complementary to
the structural ideas behind social network analysis. A number
of works in network analysis have now begun to combine these
frameworks together (see, for example, Mische 2007).
Thus, social network analysis is closely linked to structural
theories and, as such, comprises one element in a more compre-
hensive framework of sociological analysis. Action theories and
cultural theories, with their associated methods, are complemen-
tary within this framework. Social network analysis is limited to
its structural concerns, but it is an essential complement to other,
equally limited, approaches. The use of social network analysis in
particular research projects is likely, therefore, to require the use
of a variety of other methods of data collection and analysis. Even
in its own territory, social network analysis is likely to involve
a multi-method research design, and it is rarely the case that
even a predominantly structural study will use social network
Criticisms and frequently asked questions 87
techniques alone. This is equally the case with respect to data

collection. The collection of data for social network analysis
involves methods of data collection that are not specific to it:
most particularly, survey methods, ethnographic methods, and
documentary methods. An understanding of social structure
in the context of the actions of individuals and groups and the
cultural contexts in which they live and construct their actions
will involve placing social network analysis alongside the other
analytical methods appropriate to these areas. These multiple
methods of data collection and analysis—both quantitative
and qualitative—must be combined in an appropriately devised
research design.
These limitations to social network analysis are fundamental,
though they are no different from the limitations that apply to
any specialised approach. However, some potentially serious
limitations have been suggested by critics of network analysis,
and these must be considered and answered.
Criticisms and responses
Network analysis, like all social research methods, has been the
subject of much criticism by those whose preferences and intel-
lectual concerns lie in different directions. Indeed, social network
analysis has, perhaps, received more than its fair share of criti-
cism, and perhaps some of this criticism has been provoked by
the overstatements—or even the naivety—of some of its
advocates and users. Much criticism, however, is based upon
misunderstanding and misrepresentation. In this section I will

address the major criticisms and try to show how much credence
should be accorded to them. I will look at six commonly voiced
criticisms:
1 Isn’t it all frightfully new and exciting?
2 Isn’t it all rather trivial?
3 Isn’t it all unnecessary?
4 Isn’t it just pretty pictures?
5 Isn’t it simply too formal?
6 Isn’t it very static?
1 Isn’t it all frightfully new and exciting?

This statement is not so much a criticism as a complete misrep-
resentation of the history of social network analysis. During the
late 1990s, a number of physicists put forward mathematical
models of networks that, they held, could be applied widely in
social science applications and could fill a huge intellectual gap
in sociological understanding. Social scientists, they argued, had
simply ignored the fact that people are connected into struc-
tures of social relations that can be analysed in terms of their
distinctive emergent properties. Thus, this comment was a criti-
cism of the poor intellectual skills of social scientists, who had
failed to understand how important were the social networks in
which people are involved and had to wait until the physicists
came along and showed them how to do relational sociology.
This view was enthusiastically taken up by non-social scientists
and by journalists and other commentators. Many books and
articles appeared and enthused over the ‘new science’ of social
networks. We live in an increasingly ‘connected’ world, they

argued, and so we require these new techniques if we are to
study this new phenomenon.
I have shown in Chapter 2 that social network analysis has,
in fact, a long history in sociology, anthropology, and social
psychology: a history reaching back to the early days of these
disciplines. Even if attention is limited to formal and systematic
mathematical approaches, this history stretches back for over
half a century. So what, exactly, is going on? Why have physicists
and recent commentators not been aware of this history?
It has to be recognised that, to a degree, social network analysts
have themselves been culpable. Rather than actively proselytis-
ing beyond their disciplinary boundaries, they have simply got
on with the job and have carried out their specialised studies.
Unless people have read the sociological literature they will not
have been aware of the work in social network analysis. This is
why those who advocate this ‘new’ method are so completely
wrong. In claiming the critical need for their own approach to
sociological questions, they forgot to check whether sociolo-
gists had got there first: they simply did not bother to look at
any sociology (or, for that matter, any anthropology or social
psychology) before decrying the absence of relational concerns
in those disciplines (Scott 2011b).
Social network analysis, as practiced by social scientists, may
not be new but it is certainly exciting. The critics of social science
have got it right on this point at least. I hope that I have shown
in Chapter 4 and elsewhere that social network analysis provides
powerful and exciting insights into crucially important questions
and that the techniques outlined in this book can help us to

explore these.
2 Isn’t it all rather trivial?

Those who point to the triviality of social network analysis do
recognise that this particular specialism exists, but they fail
to recognise what it is actually about. The claim is often made
these days that sociologists who study social networks ought
really to be studying something more important. The study of
social networks is seen as concerned with the ‘social networking’
promoted by management consultants and the use of internet-
based ‘social network’ sites such as Facebook, Twitter, My Space,
and LinkedIn. This, many people assume, is what social network
analysis is all about.
The first point to make is that these phenomena are, in fact,
rather important social phenomena. An increasing amount
of communication and interaction is now internet-based and
people engage with each other through e-mail, live messaging,
and dedicated websites to an ever greater extent. The cyber
world is an increasingly important part of all our social worlds.
However, social network analysis does not only concern itself
with these web-mediated forms of interaction. It is concerned
with the whole range of face-to-face and spatially distanciated
social relations that have always been the principal objects of
sociological analysis.
Some critics, even within the discipline, have recognised
this point but persist with their criticism of the triviality—or
perhaps marginality—of social network analysis. This approach,
they argue, is concerned only with the interpersonal relations of
everyday life and cannot be used to investigate more important

matters of economic and political structure or the many macro-
social features of human life. This criticism is misguided on two
grounds. First, the interpersonal and the everyday are every bit
as important as the economic and the political. There are no
grounds for asserting the triviality or importance of one domain
of enquiry over another. Second, social network analysis is able
to address economic, political, and macro-social issues. I have
illustrated this in Chapter 4, where I showed the long history of
research into economic structures of relations that have investi-
gated national economic structures and the structures of global
economic systems.
Social network analysis, far from being trivial or concerned
with marginal issues, has addressed central social issues across
the whole spectrum of the discipline. It is an approach of major
significance.
3 Isn’t it all unnecessary?

This criticism may not see social network analysis as limited to
triviality, but it does hold that the findings of social network
analysis are lacking in significance. This is the oft-cited ‘so what?’
criticism of social network analysis. Such critics note the immense
time and effort spent on documenting the number and type of
ties that link people together into more or less dense chains of
connection but go on to ask that the analyst show that these
make some difference to how people act or to the outcomes
of their actions. Critics claim that everybody knows that such
links exist and that anyone, with the appropriate technical skills,
can map them out, but unless it can be shown that people take
some account of them or that things are different because of
them, then the research findings are merely documents of the
obvious and the irrelevant. The most radical of these critics
go on to assert that even if the researcher did try to assess the
importance of network connections, he or she would inevitably
find that they have no importance. All this research is, therefore,
completely unnecessary.
It has to be said that much social network analysis has,
in fact, remained at the purely descriptive level and has not
gone on to assess the significance of the connections for the
individuals involved. Some early computer-based research did
report measures simply because the available software made
them easy to compute: ‘if it can be measured then it must be
significant’ was their assumption. This kind of naïve empiricism
also marked early uses of SPSS for survey analysis, where the
option ‘Statistics all’ spewed out an array of statistics that were
assumed to be important simply because they were a part of
the package. However, this assumption is far from typical. I
have argued in Chapter 3 that it is essential that the structural
analyst know and understand the concepts that are being used
and that he or she can justify them as valid and appropriate.
When there is a prima facie reason to expect the social relations
to have an effect, then a descriptive mapping plays an essential
part in the research process. This is typically the case when a
project is firmly grounded in an on-going tradition of research
in which this significance has been explored or inferred. I have
shown in Chapter 3 that there have been good technical reasons
why conventional statistical significance tests cannot be used in
social network analysis and that it is only now that appropriate

methods of statistical analysis have become available.
The ‘so what?’ criticism has always carried less weight than
its users believe. It is a criticism that could be raised against any
approach to sociology where there is an absence of accepted
theoretical justification or established empirical grounding.
Using survey analysis to document the pattern of attitudes held
among the members of a group, for example, could be subjected
to the same criticism unless the researcher produced a theoreti-
cal rationale for the significance of the distribution of attitudes or
rehearsed the results of empirical studies. Much survey research
is just as descriptive as much social network research, and the
fact that the ‘so what?’ question is raised against one rather than
the other smacks more of prejudice than of reasoned argument.
Happily, such criticisms are becoming less frequent as social
network analysis becomes more established and its research
applications can allude to a larger theoretical and empirical
framework to legitimate its use.
4 Isn’t it just pretty pictures?

This is really a variant of the ‘so what?’ criticism and so can be
dealt with briefly. The criticism is particularly applied to studies
that concentrate on the display of sociograms rather than the
calculation of specific measures. The argument is that drawing a
sociogram, however colourful and striking, must still be shown
to be a ‘real’ social factor with distinct causal effects on actions
and their outcomes. While the validity of this argument can be
recognised, exactly the same considerations are relevant as have
been discussed in relation to the larger ‘so what?’ criticism.
5 Isn’t it simply too formal?

This criticism implies a dualism between ‘structure’ and ‘action’,
arguing that the formal clarity of network structure ignores
the meanings and definitions through which people construct
their social relations. Sociological explanations, they argue, must
take account of individual definitions of the situation and must
recognise that the particular formal connections discovered by
an outside observer are important factors in sociological analysis
only if and in so far as they are taken account of by individuals.
It is the individuals who interpret social relations that give them
their significance. The further conclusion is drawn that particular
connections—a directorship, a citation, a friendship—will mean
different things to different people and so it is meaningless to
treat them all as equivalent and to count them up or use them in
formal methodologies.
This is an important criticism if not taken to its extreme. Social
relations do vary in their significance for individual actors and
it is not at all straightforward to count numbers of friends or
to rate one friendship as twice or three times as important as
another. However, the difficulty in doing this does not warrant
the conclusion that it is impossible. It is always necessary for
social researchers to make an informed, expert judgement about
the meaning and significance of social relations to individuals,
but this must be done on the basis of evidence that warrants the
judgement as plausible. Counting, ranking, and valuing relation-
ships for use in social network analysis do not involve making
an essential and absolute statement about what the relation-
ship ‘really’ means to individuals. Use of these procedures is an
attempt to find a plausible and defensible way of transcending

individual subjectivity and drawing some general conclusions. It
is simply one particular way of undertaking a relational, struc-
tural analysis and only if the very idea of structural analysis is
rejected can the criticism be upheld in full. I have tried to argue
above that it cannot be rejected and that structural analysis has
a part to play alongside other approaches to sociology.
6 Isn’t it very static?

This final criticism has been thoroughly covered already. Critics
argue, however, that social network analysis has been static and
descriptive because this is all it ever can be. Social network analy-
sis, they argue, simply charts relationships and structures as they
exist at a particular time. Even if the critics do not claim that
this work is unnecessary, they do claim that it is fundamentally
limited and can tell only a small part of the whole story. Social
relations are changing all the time and a snapshot picture taken
at a particular point might be an extremely poor representation
of the social structure at a later time.
This is an important criticism, though it involves a consider-
able overstatement. It is true that a single cross-section, or even a
series of them, is a poor representation of a continually changing
social process. It is similarly the case that screen shots give a poor
impression of a movie. However, when used with care, descriptive
studies can serve as the first step towards more comprehensive
accounts. They are the starting point—not the end point—of
on-going research. As I showed in Chapter 3, methods better
able to grasp dynamic processes and that allow proper longitu-
dinal research to be undertaken are now becoming available and
are being more widely used. We all have to walk before we can
run, and the new mathematical concepts and software programs
allow the social network analyst to take this further step.
Frequently asked questions
In the first part of this chapter I have looked at some of the

principal questions raised by critics of social network analysis,
and I have tried to provide some responses to these. In this part
of the chapter I turn to the practical questions raised by those
who think they might find social network analysis useful but
are concerned about their own abilities to use or to understand
the techniques required. In most cases, the answers should be
apparent from the earlier, detailed discussions, but I have tried to
highlight the key issues here. Seven questions recur in practical
discussions of how to do a social network study:
1 How can I decide who to include as members of my
network?
2 When do I stop tracing connections among the members of
my group?
3 How can I decide the relative importance of different kinds
of network linkage?
4 How can I determine the strength of intensity of a social
relation?
5 Networks include positive as well as negative relations: does
this pose problems for social network analysis?
6 Do I need to know a lot of mathematics to use social

network analysis?
7 What are the ethics of social network analysis; isn’t it just a
form of snooping and spying?
1 How can I decide who to include as members of my

network?
This is the issue of the so-called ‘boundary problem’ in social
network analysis. In some cases, the boundaries of a network will
be clear and straightforward: the members of a school class or
year group, the members of an organisation, and so on. However,
groups that may appear to be clearly bounded may actually be
indeterminate. It might at first seem obvious who to include as
members of a family, for example, but does the ‘family’ include
aunts, uncles, and cousins, what degrees of cousinhood are
relevant, and what are the implications of divorce, remarriage,
or cohabitation? Can the same definition be used in all cases:
does each potential member of the network mean the same
thing by ‘family’? Boundaries must usually be determined by the
informed expert judgement of the researcher and can rarely be
decided on the basis of network measures themselves. A sociolo-
gist of the family, for example, must decide what, in the light of
existing knowledge, makes sense as a definition of the family and
so how it is to be bounded. Similarly, an economic sociologist
must decide, on the basis of prior research, whether a network
of ‘top’ companies should include 100, 200, or 300 companies.
There are, in fact, no easy answers to these questions, but that is
true of any sociological question worth asking.
2 When do I stop tracing connections among the members of

my group?
It is very easy to get carried away and to record long chains of
connection: to note down friends of the friends of a worker
who are committee members of political parties in which other
members are friends of the friends of an employer. This does not
necessarily mean that the worker and his or her employer have
any significant social relationship beyond the employment rela-
tionship itself. As with the boundary question, the investigator
must decide on the basis of sociological evidence and concep-
tualisation when to cut-off the search for relationships. Equally,
the investigator him or herself must decide whether relations
involving, say, line distances of five, six, or more constitute real
social relationships or can be regarded as having a sociological
significance. This decision is likely to vary from one type of rela-
tionship to another.
3 How can I decide the relative importance of different kinds

of network linkage?
This is not something that social network analysis itself can
resolve for you. Most relationships are complex and involve
many different emotions, purposes, and interests: social network
analysts describe them as ‘multiplex’. Someone may, for example,
be simultaneously a friend, workmate, and political collaborator
of another, and any separation of ‘friendship’, ‘work relation’, and
‘political affiliation’ is likely to be arbitrary. Nevertheless, this is
what sociologists must do in constructing ideal types of relations
from concrete patterns of connection. This analytical approach
to social relations is a prelude to deciding which is to be seen
as the most important for the practical purpose at hand. When

a number of potentially important relations have been identi-
fied, however, social network analysis does provide a way of
calculating which of these seems to have the greatest salience or
significance for the particular question being investigated: this is
simply the standard form of causal analysis in social science.
4 How can I determine the strength of intensity of a social

relation?
At the risk of repetition, this depends on your expert sociologi-
cal judgement. The strength of a relationship in social network
analysis is simply the number that is attached to the line, and this
number is something that has to be decided during the research
process. The researcher must use some kind of sociologically
grounded criterion of scaling in order to assign numerical values
to a line. This is relatively straightforward when deciding on
the absence or presence of a relation and so assigning a value
of 0 or 1, but even here it may not always be obvious when a
relationship has ceased to exist or come into being. It is far more
difficult to assign a value of 2 to one line and 4 to another, and it
may be highly problematic to conclude that the latter is twice as
strong as the former. In social network analysis this is technically
described as the problem of linearity in ratio scales. All that can
be said is that, as in all sociological analyses worth undertaking,
a plausible and justifiable estimate of strength must be made
and that this tentative and arbitrary estimation must be remem-
bered when it comes to analysing your results. Don’t forget that
you assigned the numbers in the first place and so the results will
reflect the plausibility of that initial judgement.
5 Networks include positive as well as negative relations:

does this pose problems for social network analysis?
Yes, it does, but only because it poses problems for the people
involved! Negative relations of dislike, withdrawal, exclusion, or
conflict may cause discomfort for the people involved, but for
the sociologist they pose no additional questions. A + or – sign
can be assigned to indicate their character and they can then
be treated in exactly the same way as any other relationship. A
negative relationship can be assigned a numerical value in the
same way as a positive one and, of course, all the same limita-
tions will apply.
6 Do I need to know a lot of mathematics to use social

network analysis?
The simple answer is ‘no’. The more complex answer is ‘maybe’.
The basic concepts and measures used in social network analysis
can all be generated very rapidly by the standard, and easy to
use, software packages described in the following chapter. To do
this requires absolutely no mathematical expertise and depends
simply on your skills with a keyboard and mouse to negotiate the
menu structure of the software. However, you do need to have
some understanding of what the various procedures are trying to
do and whether it makes sense to use them for your data. This is
the kind of understanding that I have tried to provide in Chapter
3 and, in more detail, in my handbook of social network analysis
(Scott 2012). In order to follow the discussions in some of the
more advanced texts, however, you will need to have some famil-
iarity with mathematical notation and procedures, though not
generally of a very high level. More important is the willingness
to grapple with the numbers and symbols and to expect some

difficulty in comprehension. As with any difficult enterprise in
social science, the benefits outweigh the heartaches.
7 What are the ethics of social network analysis; isn’t it just

a form of snooping and spying?
All sociology is a form of snooping: that’s what makes it such fun!
Studying networks is no more unethical that studying any other
aspect of a person’s life. It is important, of course, to observe the
normal considerations of confidentiality and anonymity where
these are requested and are appropriate, but the application
and use of social network analysis raises no additional problems
to those found with any other social research method. There
have been concerns over the use of social network analysis by
the police and the security services to investigate criminal and
terrorist networks and networks of political activists. There are
very genuine issues about whether academic social scientists
should cooperate in such research, but the validity of undertak-
ing this research is not a question of professional ethics but is
a much larger political question about the legitimate role and
scope of state activity. Social network analysis is not ‘value free’,
but neither is it an especially unethical form of social science.
Further reading
Scott, John and Carrington, Peter 2011. The Sage Handbook of

Social Network Analysis. London: Sage Publications. Section 1
includes reviews of developments in social network analysis in
various disciplines and considers the theoretical issues raised

in these. The Chapters by John Scott and by Bettina Holstein
are especially relevant to the issues discussed here.
Cresswell, John W. and Plano Clark, V. L. 2007. Designing and
conducting mixed methods research. Thousand Oaks, CA: Sage
Publications. A general souerce that does not discuss social
network analysis but gives a clear statement and rationale for
combined or multiple methods.
6 Software for social network
analysis
As I outlined in Chapter 1, the time was that many of the

processes involved in social network analysis had to be under-
taken manually. Even when computers came into wide use, the
possibilities for rapid data analysis were limited by the need to
refine, or even to provide from scratch, software programs that
would automate some of the basic tasks. All this has changed.
A number of software programs are now both easily and
cheaply available, and the newcomer to social network analysis
can rapidly begin to produce sophisticated analyses of her or
his data. This situation has its dangers, of course. The tempta-
tion to generate output is such that a researcher may produce
that output first and only then consider whether it is useful or,
indeed, valid. However, having got this far in this book, I hope
that my readers will be very aware of this problem and will not
fall into that trap.
In this chapter I want to provide a general guide to the use
of some of the most important packages available. In order to
use these effectively you may want to work through some of
the more technical readings given in Chapter 3, but this guide
103
should provide you with enough information to allow you to

begin producing some simple, but quite powerful, measures of
network structure.
Two programs dominate the field: UCINET and PAJEK.
UCINET is a commercial programme distributed by Analytical
Technologies (http://www.analytictech.com/). PAJEK is a freely
distributed programme available from a WIKI site (http://
pajek.imfm.si/doku.php?id=pajek). A word of warning: this
webpage includes a large photograph of a spider and the squea-
mish may prefer to go directly to (http://pajek.imfm.si/doku.
php?id=download)! Each program has its particular advantages
and limitations, and a choice may be a matter of personal prefer-
ence. This choice may be quite easy to make as UCINET includes
PAJEK as an add-in for its own program code. In addition to
these general programs there is a widely used statistical program,
R, that includes a specialist add-on for the statistical analysis of
social network data. This program is more complex to use, but
is extremely powerful and makes it possible to transfer data
to other statistical and graphical routines. These are the three
principal programs on which I will concentrate, but I will also
make some brief reference to a number of specialist visualisa-
tion programs that are associated with them and, in some cases,
distributed with them as add-ons.
UCINET: The pioneer program
UCINET was developed at the Irvine campus of the University of

California (UC, Irvine—hence the acronym). It was designed and
Software for social network analysis 105
produced by some of the leading social network researchers—

most notably, Lin Freeman, Steve Borgatti, and Martin Everett—
as a way of helping others to undertake the kinds of analyses that
they themselves were using in their empirical studies. It is now
available in Version 6 and is established as a tried and tested, and
very stable, platform for social network analysis.
UCINET appears on screen with a very simple menu structure,
from which data entry and information windows appear. Seven
top-level menu headings define the structure of the program:
• FILE
• DATA
• TRANSFORM
• TOOLS
• NETWORK
• VISUALIZE
• OPTIONS
As might be expected, the FILE menu provides access to the basic
file management tasks and access to its current add-ins. The core
of the program, however, is to be found under the DATA menu,
together with the TRANSFORM menu. DATA is the gateway to
data entry and the import and export functions. From here it is
possible to compile a data file in native UCINET format—typi-
cally simply the network as a linked list of data points—or to
import the raw data from text files and spread sheets. The DATA
menu also provides the routines needed to refine and prepare
data sets, using extraction and deletion procedures to determine
the inclusion or exclusion of particular points and subgroups.
Finally, the DATA menu allows for the sorting, permutation, and
transposing of the basic data and for converting it into different

formats. For example, the conversion of 2-mode into 1-mode
data is achieved in this way. TRANSFORM provides further
options for dichotomising or slicing a network. All these tasks
are handled very intuitively, with the option given at each stage
to save any newly generated files. There is also a useful provision
of error messages that can be supplemented by the comprehen-
sive ‘Help’ screens.
The TOOLS menu in UCINET leads to procedures for under-
taking some basic, auxiliary tasks. This is the way in which it is
possible to use cluster analysis—and to construct dendrograms
from clustered data—and to undertake metric and non-metric
multidimensional scaling, factor analysis, and correspondence
analysis. The key network concepts and measures are to be
found, unsurprisingly, under the NETWORK menu. Here are to
be found measures of distance, density, homophily, and central-
ity, as well as methods for identifying components, cliques, struc-
turally equivalent blocks, and other subgroups. It is also through
the NETWORK menu that a number of egocentric measures can
be produced, including analyses of betweenness and brokerage.
Particularly useful features of UCINET are the routines found
under its VISUALIZE menu. It is here that the NETDRAW
network-drawing package, the MAGE modelling program, and
the drawing facilities provided in PAJEK can be accessed. These
are all discussed below.
PAJEK
PAJEK—Slovenian for ‘spider’—was developed at the University

of Ljubljana by Vladimir Batagelj and his colleagues. It had its
origins as a specialist program for handling very large data sets
and includes a number of advanced measures. The program has
a more extensive menu structure than UCINET and does not
cover quite the same wide range of measures, but a number of
key measures are provided in addition to its ability to analyse
large data sets very rapidly. A manual for the program has been
published as an accessible guide to social network analysis (De
Nooy et al. 2005).
The basic elements in the program are to be found under the
FILE and NET menus. Here can be found the routines for input-
ting and editing network data and for exporting data in various
formats, such as that of UCINET. The NET menu provides vari-
ous routines for transforming networks, much as are provided
in UCINET’s DATA and TRANSFORM menus, though sorting
is accessed under the OPERATIONS menu. NET is also the
means through which components can be identified, while the
OPERATIONS menu allows the undertaking of blockmodelling.
The heart of the PAJEK program for many of its users is its
DRAW menu, which leads to a screen from which sub-menus
allow the drawing and displaying of networks and their
subgroups. Simple circle diagrams can be drawn, but the most
interesting option is, perhaps, the ability to generate two-
dimensional and three-dimensional visualisations based on
forms of multidimensional scaling. Visual representations can be
produced easily, and three-dimensional representations can be
rotated in various ways to display and inspect a network. Simple

but powerful techniques are available to include or exclude
labels for data points and to indicate the value or intensity of a
line. The representations produced can be exported in various
graphics formats for printing or for display independently of the
PAJEK program.
R and other programs
R (available at http://www.r-project.org/) is a program produced

as part of a general project for statistical computing. It comprises
a basic framework that serves as a platform for a number of
supplied modules and add-ons that all operate within its basic
environment. The two most important add-ons for social
network analysis are STATNET and SIENA. STATNET covers the
important techniques for hypothesis testing associated with
ERGM methods. SIENA was originally developed by Tom Snijders
as part of his StOCNET program, now supplanted by R. SIENA
is especially geared to the statistical analysis of longitudinal and
cross-sectional data, so allowing the assessment of the signifi-
cance of change over time.
Other, free-standing, programs that readers may wish to
investigate are NETDRAW, MAGE, and SONIA. NETDRAW, also
provided as a module within UCINET, is a powerful program
for drawing and editing basic network sociograms. It interfaces
rapidly with UCINET and works rapidly on its files. MAGE was
devised by chemists as a way of using stick and ball methods
to represent chemical bonding formulae (see http://kinemage.
biochem.duke.edu/kinemage/magepage.php). Because of its

data structure it is adapted easily for use in social network analy-
sis and provides both two-dimensional and three-dimensional
representations of data generated from UCINET or PAJEK and
allows the ‘sticks’ and ‘balls’ of a network (the lines and points)
to be coloured and labelled in various ways. SONIA, devised by
Snijders and supported by Dan McFarland and Skye Bender de
Moll, is a visualisation package for longitudinal data (see http://
www.stanford.edu/group/sonia/). Unlike the programs discussed
so far, SONIA works under JAVA and requires the installation
of at least the basic JAVA package. The environment gives the
program the ability to produce striking and easily transportable
visual representations of social network data as moving images.
As all the best marketing presentations say: other network
programs are available. A guide to these can be found in Huisman
and van Duijn (2011). The programs discussed in this chapter,
however, are the most easily accessible and useable programs for
the beginner and for more advanced users. If I can be permitted a
recommendation, I will suggest using UCINET, which gives direct
and integrated access to PAJEK, NETDRAW, and MAGE. Used
as an adjunct to your reading of more advanced texts on social
network analysis, the program and its add-ons will allow you to
do almost anything you can imagine—and many things that you
are unlikely to have imagined. As your knowledge improves, you
will find that you need to use R and SONIA, and you should, by
then, have the technical ability to handle them and to use them
sensibly.
Bibliography
Alba, Richard D. and Kadushin, Charles 1976. ‘The intersection of

social circles: a new measure of social proximity in networks’.
Sociological Methods and Research 5.
Barabási, Albert-László 2002. Linked: The New Science of
Networks. Cambridge, Mass.: Perseus.
Barnes, John A. 1954. ‘Class and committee in a Norwegian
island parish’. Human Relations 7.
Bearden, James and others 1975. ‘The nature and extent of
bank centrality in corporate networks’ in Scott, J. (ed.) Social
Networks, Volume 3. London: Sage, 2002.
Bogardus, Emory S. 1925. ‘Measuring social distance’. Journal of
Applied Sociology 9.
Bogardus, Emory S. 1959. Social Distance. Los Angeles: Antioch
Press.
Boorman, Scott A. and White, Harrison C. 1976. ‘Social structure
from multiple networks: II’. American Journal of Sociology 81.
Bott, Elizabeth 1957. Family and Social Network. London:
Tavistock Publications.
111
Bott, Helen 1928. ‘Observation of play activities in a nursery

school’. Genetic Psychology Monographs 4: 44–8.
Bourdieu, Pierre 1979. Distinction: A Social Critique of the
Judgment of Taste. London: Routledge, 1984.
Burt, Ronald S. 1987. ‘Social contagion and innovation: cohesion
versus structural equivalence’. American Journal of Sociology
92, 6: 1287–1335.
Burt, Ronald S. 1992. Structural Holes. New York: Cambridge
University Press.
Burt, Ronald S. 2005. Brokerage and Closure: An Introduction to
Social Capital. New York: Oxford University Press.
Cartwright, D. and Zander, A. (eds.) 1953. Group Dynamics.
London: Tavistock.
Coleman, James S., Katz, E. and Menzel, H. 1966. Medical
Innovation: A Diffusion Study. New York: Bobbs-Merrill.
Collins, Randall 2000. The Sociology of Philosophers. Cambridge,
Mass.: Harvard University Press.
Coxon, Anthony P. M. 1982. The Users’ Guide to Multidimensional
Scaling. London: Heinemann.
Crane, Diana 1972. Invisible Colleges. The Diffusion of Knowledge
in Scientific Communities. Chicago: University of Chicago Press.
Davis, Allisson B., Gardner, B. B. and Gardner, M. R. 1941. Deep
South. Chicago: University of Chicago Press.
Davis, James A 1963. ‘Structural balance, mechanical solidarity
and interpersonal relations’. American Journal of Sociology 68:
444–462.
De Nooy, Wouter, Mrvar, Andrej and Batagelj, Vladimir 2005.
Exploratory Social Network Analysis with Pajek. New York:
Cambridge University Press.
Bibliography 113
Degenne, Alain and Forsé , Michel 1994. Introducing Social

Networks. English translation, Beverley Hills: Sage Publications,
1999.
Dooley, Peter C. 1969. ‘The Interlocking Directorate’ in Scott, J.
(ed.) The Sociology of Elites, Volume 3. Aldershot: Edward Elgar
Publishing, 1990.
Everitt, Brian 1974. Cluster Analysis. London: Heinemann.
Farrell, Michael P. 2001. Collaborative Circles. Friendship Dynamics
and Creative Work. Chicago: University of Chicago Press.
Fennema, Meindert 1982. International Networks of Banks and
Industry. Hague: Martinus Nijhof.
Fennema, Meindert and Carroll, William K. 2002. ‘Is there a
transnational business community’. International Sociology 17,
3: 393–419.
Festinger, Leon 1957. A Theory of Cognitive Dissonance. Stanford:
Stanford University Press.
Festinger, Leon, Riecken, H. W. and Schachter, S. 1956. When
Prophecy Fails. New York: Harper and Row.
Fischer, Claude S. 1977. Networks and Places: Social Relations in
the Urban Setting. New York: Free Press.
Fischer, Claude S. 1982. To Dwell Among Friends: Personal
Networks in Town and City. Chicago: University of Chicago
Press.
French, John R. P. and Raven, Bertram 1959. ‘The bases of social
power’ in Cartwright, D. (ed.) Studies in Social Power. Ann
Arbor: University of Michigan Press, 1959.
Gould, Roger V. 1993. ‘Collective action and network structure’.
American Sociological Review 58: 182–96.
Gould, Roger V. 1995. Insurgent Identities. Class, Community, and

Protest in Paris from 1848 to the Commune. Chicago: Chicago
University Press.
Granovetter, Mark 1973. ‘The strength of weak ties’. American
Journal of Sociology 78, 6: 1360–1380.
Granovetter, Mark 1974. Getting A Job. Cambridge, Mass.:
Harvard University Press.
Hall, G. Stanley 1904. Adolescence. Its Psychology and Its Relation
to Psychology, Anthropology, Sociology, Sex, Crime, Religion.
Two Volumes. New York: D. Appleton.
Harary, F. and Norman, R. Z. 1953. Graph Theory as a
Mathematical Model in Social Science. Ann Arbor: Institute for
Social Research.
Helmers, H. M. and others 1975. Graven Naar Macht.
Amsterdam: Van Gennep.
Hilferding, Rudolf 1910. Finance Capital. London: Routledge and
Kegan Paul, 1981.
Hobson, John Atkinson 1906. The Evolution of Modern
Capitalism, Revised Edition. London: George Allen and Unwin.
Homans, George 1950. The Human Group. London: Routledge
and Kegan Paul, 1951.
Hope, Keith (ed.). 1972. The Analysis of Social Mobility. Methods
and Approaches. Oxford: Clarendon Press.
Huisman, Mark and van Duijn, Marijtje A. J. 2011. ‘A Reader’s
Guide to SNA Software’ in Scott, J. and Carrington, P.C. (eds.)
The Sage Handbook of Social Network Analysis. London: Sage,
2011.
Bibliography 115
Jeidels, Otto 1905. Das Verhältnis der deutschen Grossbanken zur

Industrie mit besonderer Berücksichtigung der Eisenindustrie.
Leipzig: Duncker & Humblot.
Jennings, Helen Hall 1948. Sociometry in Group Relations.
Washington, DC: American Council on Education.
Köhler, Wolfgang 1917. The Mentality of Apes. London: Routledge
and Kegan Paul, 1924.
Kruskal, Joseph B and Wish, Myron 1978. Multidimensional
Scaling. Beverly Hills, CA: Sage.
Kuhn, Thomas S. 1962. The Structure of Scientific Revolutions
(Second Edition, 1970). Chicago: University of Chicago Press.
Laumann, E. O and Pappi, F. U. 1976. Networks of Collective
Action: A Perspective on Community Influence Systems. New
York: Academic Press.
Laumann, Edward O 1973. Bonds of Pluralism. New York: John
Wiley.
Laumann, Edward O. 1966. Prestige and Association in an Urban
Community. Indianapolis: Bobbs-Merrill.
Lenin, Vladimir Ilyich 1917. Imperialism: The Highest Stage of
Capitalism. Moscow: Progress Publishers, 1966.
Levine, Joel H. 1972. ‘The sphere of influence’. American
Sociological Review 37.
Levine, Joel H. 1984. Levine’s Atlas of Corporate Interlocks.
Hanover, NH: Worldnet.
Lorrain, F. and White, Harrison C. 1971. ‘Structural equivalence
of individuals in social networks’. Journal of Mathematical
Sociology 1.
Lundberg, George 1936. ‘The sociography of some community
relations’. American Sociological Review 1, 5.
Lundberg, George and Steele, Mary 1938. ‘Social attraction-

patterns in a village’. Sociometry. 1: 375–419.
Marwell, Gerald and Oliver, Pamela 1993. The Critical Mass in
Collective Action. Cambridge: Cambridge University Press.
Menshikov, Sergei 1969. Millionaires and Managers. Moscow:
Progress Publishers.
Milgram, Stanley 1967. ‘The small world problem’. Psychology
Today 2: 60–7.
Mintz, Beth and Schwartz, Michael 1985. The Power Structure of
American Business. Chicago: Chicago University Press.
Mische, Ann 2007. Partisan Publics: Communication and
Contention across Brazilian Youth Activist Networks. Princeton,
NJ: Princeton University Press.
Mitchell, J. Clyde (ed.). 1969. Social Networks in Urban Situations.
Manchester: Manchester University Press.
Mizruchi, Mark S. 1982. The American Corporate Network, 1900 –
1974. London: Sage.
Moody, James and Light, Ryan 2006. ‘A view from above:
The evolving sociological landscape’. American Sociologist
Summer: 67–85.
Moreno, Jacob L. 1934. Who Shall Survive? New York: Beacon
Press.
Mullins, Nicholas C. 1973. Theories and Theory Groups in
American Sociology. New York: Harper and Row.
Nadel, Siegfried Frederick 1957. The Theory of Social Structure.
Glencoe: Free Press.
Oliver, Pamela, Marwell, Gerald and Prahl, Ralph 1988a. ‘Social
networks and collective action: a theory of critical mass, III’.
American Journal of Sociology 94: 502–34.
Bibliography 117
Oliver, Pamela, Marwell, Gerald and Teixeira, Ruy 1985. ‘A theory

of the critical mass, I’. American Journal of Sociology 91: 522–56.
Oliver, Pamela, Marwell, Gerald and Teixeira, Ruy 1988b. ‘The
paradox of group size in collective action: a theory of the
critical mass, II’. American Sociological review 53: 1–8.
Oliver, Pamela and Myers, Daniel J. 2003. ‘Networks, Diffusion,
and Cycles of Collective Action’ in Diani, M. and McAdam, D.
(eds.) Social Movements and Networks: Relational Approaches
to Collective Action. Oxford: Oxford University Press, 2003.
Parsons, Talcott 1937. The Structure of Social Action. New York:
McGraw-Hill.
Pattison, Philippa 1993. Algebraic Models for Social Networks.
Cambridge: Cambridge University Press.
Prell, Christina 2012. Social Network Analysis: History, Theory and
Methodology. London: Sage.
Price, Derek de Solla 1963. Little Science, Big Science. New York:
Columbia University Press.
Radcliffe-Brown, Alfred Reginald 1940. ‘On social structure’ in
Radcliffe-Brown, A.R. (ed.) Structure and Function in Primitive
Society. London: Cohen and West, 1952.
Roethlisberger, F. J. and Dickson, William J. 1939. Management
and the Worker. Cambridge, MA: Harvard University Press.
Rogers, Everett 1962. Diffusion of Innovations, 5th Edition. New
York: Free Press, 2003.
Rouanet, Henri and Le Roux, Brigitte 2009. Multiple
Correspondence Analysis London: Sage.
Ryan, Bryce and Gross, Neil C. 1943. ‘The diffusion of hybrid seed
corn in two Iowa communities’. Rural Sociology 8: 15–24.
Scott, John 2011a. Conceptualising the Social World. Principles of

Sociological Analysis. Cambridge: Cambridge University Press.
Scott, John 2011b. ‘The new social physics’ in Scott, J. and
Carrington, P.J. (eds.) The Sage Handbook of Social Network
Analysis. London: Sage.
Sage.
Scott, John and Griff, Catherine 1984. Directors of Industry.
Cambridge: Polity Press.
Scott, John and Hughes, Michael 1976. ‘Ownership and control
in a satellite economy: a discusssion from Scottish data’.
Sociology 10, 1.
Scott, John and Hughes, Michael 1980. The Anatomy of Scottish
Capital. London: Croom Helm.
Smith, David and White, Douglas 1992. ‘Structure and dynamics
of the global economy. Network analysis of international
trade, 1965–1980’. Social Forces 70: 857–893.
Snijders, Thomas A. B., Steglich, C.E.G. and van de Bunt, G.G.
2010. ‘Introduction to actor-based models for network
dynamics’. Social Networks 32: 44–60.
Spencer, Liz and Pahl, Ray 2006. Rethinking Friendship. Hidden
Solidarities Today. Princeton, NJ: Princeton University Press.
Stanworth, Philip and Giddens, Anthony 1975. ‘The modern
corporate economy’ in Scott, J. (ed.) The Sociology of Elites,
Volume 1. Aldershot: Edward Elgar, 1990.
Stark, Rodney 1996. The Rise of Christianity: A Sociologist
Reconsiders History. Princeton, NJ: Princeton University Press.
Stokman, Frans, Ziegler, Rolf and Scott, John (eds.). 1985.
Networks of Corporate Power. Cambridge: Polity Press.
Bibliography 119
Sweezy, Paul M. 1939. ‘Interest Groups in the American

Economy’ in Sweezy, P.M. (ed.) The Present as History. New
York: Monthly Review Press, 1953.
Tarde, Gabriel 1890. The Laws of Imitation. New York: H. Holt
and Co., 1903.
Tönnies, Ferdinand 1887. Community and Association. London:
Routledge and Kegan Paul, 1955 (based on the 1912 edition).
Travers, Jeffrey and Milgram, Stanley 1969. ‘An experimental
study of the small world problem’. Sociometry 32, 4: 425–43.
US Congress 1913. House Committee on Banking and Currency.
Money trust investigation. [The Pujo Report.] Washington:
Government Printing Office.
Valente, Tom W. 1995. Network Models of the Diffusion of
Innovations. Cresskill, NJ: Hampton Press.
Wallman, Sandra 1984. Eight London Households. London:
Tavistiock Publications.
Warner, W. Lloyd 1963. Yankee City. New Haven: Yale University
Press.
Warner, W. Lloyd and Low, J. O. 1947. The Social System of A
Modern Factory. New Haven: Yale University Press.
Warner, W. Lloyd and Lunt, P. S. 1941. The Social Life of a Modern
Community. New Haven: Yale University Press.
Warner, W. Lloyd and Lunt, P. S. 1942. The Status System of a
Modern Community. New Haven: Yale University Press.
Warner, W. Lloyd and Unwalla, Darab B. 1967. ‘The system of
interlocking directorates’ in Warner, W.L., Unwalla, D.B. and
Trimm, J.H. (eds.) The Emergent American Society: Volume 1.
Large-Scale Organizations. New Haven: Yale University Press,
1967.
Warner, W. LLoyd and Srole, Leo 1945. The Social Systems of

American Ethnic Groups. New Haven: Yale University Press.
Wasserman, Stanley 1980. ‘Analyzing social networks as
stochastic processes’. Journal of the American Statistical
Association 75: 280–294.
Wasserman, Stanley and Pattison, Philippa 1996. ‘Logit models
and logistic regressions for social networks: I. An introduction
to Markov random graphs and p*’. Psychometrika 60: 401–26.
Watts, Duncan 1999. Small Worlds: The Dynamics of Networks
Between Order and Randomness. Princeton: Princeton
University Press.
Watts, Duncan 2003. Six Degrees. The Science of a Connected Age.
New York: W. W. Norton.
Watts, Duncan J. and Strogatz, Steven H. 1998. ‘Collective
Dynamics of “Small-World” Networks’. Nature 393: 440–2.
Wellman, Barry 1979. ‘The community question: the intimate
networks of East Yorkers’. American Journal of Sociology 84:
1201–31.
Wellman, Barry and Hogan, Bernie 2006. ‘Connected lives: the
project’ in Purcell, J. (ed.) Networked Neighbourhoods. London:
Springer-Verlag, 2006.
White, Douglas and Reitz, K. P. 1983. ‘Group and semi-group
homomorphisms on networks of relations’. Social Networks 5.
White, Harrison C. 1963. An Anatomy of Kinship. Englewood
Cliffs: Prentice-Hall.
White, Harrison C., Boorman, Scott A. and Breiger, Ronald L.
1976. ‘Social structure from multiple networks: I’. American
Journal of Sociology 81.
Bibliography 121
White, Howard D., Wellman, Barry and Nazer, Nancy 2004. ‘Does
citation reflect social structure? Longitudinal evidence from
the “Globenet” interdisciplinary reserach group’. Journal of the
American Society for Information Science and Technology 55, 2:
111–126.
Whitley, Richard D. 1973. ‘Commonalities and connections
among directors of large financial institutions’ in Scott, J. (ed.)
The Sociology of Elites, Volume 1. Aldershot: Edward Elgar, 1990.
Wiese, Leopold von 1931. ‘Outlines of the “Theory of Social
Relations”’ in von Wiese, L. (ed.) Sociology. New York: Oskar
Piest, 1941.
Index
agent-based computational clique 17–18, 19, 45, 46–48, 65,

models 28 80–81
algebraic methods 48 ff. closeness 42
attraction 11 cluster analysis 49
co-authorship 66 ff.
balance 13 Coleman, James 60, 66
Barabási, Albert-László 26 collecting network data 32 ff., 97
Barnes, John 14 Collins, Randall 69–72
Bearden, James 15–16, 79 communication patterns 12, 40,
Bender de Moll, Skye 109 52
betweenness 42 community relations 18–19, 25
blocks and blockmodelling 21, complete enumeration 32
49, 52 component 45–7, 52, 80
Bott, Elizabeth 13–14, 39 CONCOR 5, 49
Bott, Helen 9, 13 conjugal roles 14
Bourdieu, Pierre 24 connectedness See density
Burt, Ronald 65 Cooley, Charles 8
Coxon, Anthony 3
Carroll, William 82 Crane, Diana 66
centralisation 44 cyclic component 45–46
centrality 11, 41
citation studies 65 ff. data files, structure of 33–34
clan 47
123
degree 35, 60 See also indegree, Granovetter, Mark 14, 38

outdegree graph theory 11–12, 29, 34 ff.
density 11, 39, 43 Griff, Catherine 80
descriptive methods 92 Gross, Neil 59
differential association 64 group dynamics 11, 12
diffusion 11, 28, 58 ff.
digraph 35 Hall, G. Stanley 9
dimensions in space 51 Hawthorne electrical works
directed relations 9, 20, 35, 38, 16–17, 19
47, 79 Helmers, H. M. 16
distance 21–24, 25, 38, 40, 51–52, Hilferding, Rudolf 77
98 Hobson, John 76, 77
Dooley, Peter 78 Homans, George 20, 48
dynamic processes in networks Huisman, Mark 109
26, 95–96 hypothesis testing 53–54
ego-centric measures 32, 34 ff. imitation 58, 62

Epstein, Scarlett 14 inclusiveness 43
ethical issues in network analysis indegree 37
101 innovation 58 ff.
exponential random graph intensity of relations 35–36, 38, 41,
models 54 46, 63, 99
interlocking directorships 2, 3–4,
Fennema, Meindert 81, 82 15–16, 24, 32, 34, 76 ff.
Festinger, Leon 13 isolates 9, 43
field theory 11
fields of relations 11 Jeidels, Otto 76
flows in networks 11, 12, 28, 62 job opportunities 14–15
formal sociology 8, 14
friendship relations 8–9, 11, 14, 15, Kadushin, Charles 66
22, 24, 25, 32, 35 Kapferer, Bruce 14
kinship 14, 15, 16, 97
Giddens, Anthony 2
global network measures 39 ff. Laumann, Edward 24
Gould, Roger 65 Lenin, Vladimir 77
GRADAP 5 Levine, Joel 3, 25, 79, 81
Index 125
Lewin, Kurt 10, 22, 51 origins of social network analysis

Light, Ryan 74–75 7 ff., 88–9
lines in a graph 35 outdegree 37
Lorrain, François 49
Lundberg, George 13 PAJEK 25, 104, 107–8
path 40, 52
MAGE 108–9 peripherality 42
Manchester University 14, 16 Perlo, Victor 78
matrix analysis 20, 29 phase transition 28
Mayo, Elton 16–17 physics 2, 88
McFarland, Daniel 109 plex 47–48
Mead, George 8 points in a graph 33, 34 ff.
Menshikov, Serge 78 power 76
Milgram, Stanley 26–27, 40
Mitchell, J. Clyde 3, 5, 14 Quantitative Sociology Group 2
Mokken, Robert 3, 79, 80
Moody, James 74–75 R program 108
Moreno, Jacob 9–10 Radcliffe-Brown, Alfred 14, 16
Mullins, Nicholas 67–68 Regular equivalence 49–50
multidimensional scaling 23–24, Reitz, Karl 49
29, 50, 74, 81 repulsion 11
multiple correspondence analysis Rogers, Everett 61–62
24 Ryan, Bryce 59
multiplexity 98
multiplicity 36 ‘S’ curve of diffusion 59, 60, 63, 66
sampling 32–3, 40
Nadel, Siegfried 20 school interaction 9, 32
negative relations 36, 38, 100 Schwartz, Michael 3
nested components 46, 52 Scott, John 80
NETDRAW 108 SIENA 108
non-linear change 28 signed graph 36, 100
significance testing 54, 92–93
Old City study 18, 49 Simmel, Georg 8
Oliver, Pamela 64 six degrees of connection 27
online communication 90 Small, Albion 8
opinion leaders 59, 60, 62 small groups 10
small world properties of Tarde, Gabriel 58–59

networks 26–27, 40 Tavistock Institute 13
smallest space analysis 23, 50 time in networks 25 ff.
Smith, David 82 Tönnies, Ferdinand 8, 19
Snijders, Tom 28, 108, 109
social circle 48 UCINET 104–6, 109
social networking distinguished Unwalla, Darab 78
from social network analysis
90 Valente, Tom 65
sociogram 9, 17, 33, 35, 51, 93 value of line 35–36
sociometry 20, 28 See also graph Van Duijn, Marijtje 109
theory Venn diagrams 19
software for social network Vierkandt, Alfred 8, 9
analysis 103 ff.
SONIA 109 Warner, W. Lloyd 16–17, 18, 48, 78
space 22, 29, 50 ff. See also fields Watts, Duncan 26, 27
of relations weak ties 15, 27, 61, 64
Stanworth, Philip 2 web metaphor 7, 8, 77
Stark, Rodney 63–4 Wellman, Barry 3, 15
stars of attraction 9, 12 White, Douglas 49, 82
statistical inference 53 White, Harrison 21, 48–49, 79
STATNET 108 White, Howard 72–73
Stokman, Frans 3, 5, 79, 80 Whitley, Richard 2
Strogatz, Steven 26, 27 Wiese, Leopold von 8
structural equivalence 21, 48–49
See also regular equivalence Yankee City study 18
structural hole 42
subgraphs 45
Sweezy, Paul 78

9781849668200

Uploaded by

Copyright:

Available Formats

9781849668200

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

9781849668200

Uploaded by

Copyright:

Available Formats

What is Social Network Analysis?

‘What is?’ Research Methods series

What is Online Research? Forthcoming books:

an imprint of Bloomsbury Publishing Plc

This work is published open access subject to a Creative Commons

ISBN 978-1-78093-848-6 (hardback)

Cover design: Burge Agency

List of Figures vii

The idea behind this series is a simple one: to provide concise

the assessment of the strengths and possible weaknesses of the

Social network analysis conceptualises individuals or groups as

into contact with leading figures in North America, Britain, and

problem was that company information had been reported in

program I consumed a half of the university’s weekly budget,

The roots of social network thinking are to be found in relational

the social ‘fabric’—first emerged. German social theorists such as

Sociometry, small groups, and communities

The earliest empirical work on small groups and communities

choices among classmates were seen as a way of exploring the

Figure 1 An acquaintanceship sociogram

compiling the various friendship choices made by class members

States, was even more firmly embedded in this tradition of analy-

theory, which investigates the formal properties of networks, or

Figure 2 Sociometric structures: star, Y, chain, circle

foster particular structures of leadership (French and Raven

class-specific, structures of kinship and friendship and that these

through the dense and well-connected local networks and

Figure 3 Strong and weak ties

Wellman (1979; Wellman and Hogan 2006; see also Fischer

board-level connections in top US companies, while Helmers

Cliques, roles, and matrices

The work of the Manchester anthropologists studying African

alterations to the workplace improved worker morale and led to

can be seen in relation to Tönnies (1887) earlier suggestions on

Figure 5 Overlapping cliques in a social hierarchy

George Homans (1950) undertook a systematic review of small-

Algebraic and matrix methods were developed by Harrison

Figure 6 Blockmodel role structure

Space and distance

The sociometric uses of graph theory measured the ‘distance’

of closeness, but it does not correspond to the everyday idea

people you see

Figure 7 Affective distance in friendship

to attempts to measure two or more attitudes through the inter-

as a way of reporting structural features of social space (Coxon

Figure 8 Multidimensional scaling of a network

traced the friendship patterns of those in particular occupa-

Dynamics and social change

The majority of the early studies of social networks have been

not generally attempted to explore the internal dynamics that

instructed that they must pass the message only to a person

Radical structural changes can therefore follow from minor

methodological approaches. The earliest of these was graph

Scott, John 2012. Social Network Analysis. Third Edition. London:

Prell, Christina 2012. Social Network Analysis: History, Theory and

Chapter 2 traced the history of social network analysis through

Collecting network data

Relational data for social network analysis can be collected

being undertaken to find ways of making more reliable infer-

Graph theory and egocentric measures