On Network Theory - Org Sci

Published on March 2017 | Categories: Documents | Downloads: 25 | Comments: 0 | Views: 157
of 14
Download PDF   Embed   Report

Comments

Content

Organization Science
Vol. 22, No. 5, September–October 2011, pp. 1168–1181
issn 1047-7039 — eissn 1526-5455 — 11 — 2205 — 1168

http://dx.doi.org/10.1287/orsc.1100.0641
© 2011 INFORMS

On Network Theory
INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

Stephen P. Borgatti, Daniel S. Halgin
LINKS Center for Social Network Analysis, Gatton College of Business and Economics, University of Kentucky,
Lexington, Kentucky 40508 {[email protected], [email protected]}

esearch on social networks has grown considerably in the last decade. However, there is a certain amount of confusion
about network theory—for example, what it is, what is distinctive about it, and how to generate new theory. This paper
attempts to remedy the situation by clarifying the fundamental concepts of the field (such as the network) and characterizing
how network reasoning works. We start by considering the definition of network, noting some confusion caused by two
different perspectives, which we refer to as realist and nominalist. We then analyze two well-known network theories,
Granovetter’s strength of weak ties theory [Granovetter, M. S. 1973. The strength of weak ties. Amer. J. Sociol. 78(6)
1360–1380] and Burt’s structural holes theory [Burt, R. S. 1992. Structural Holes: The Social Structure of Competition.
Havard University Press, Cambridge, MA], to identify characteristic elements of network theorizing. We argue that both
theories share an underlying theoretical model, which we label the network flow model, from which we derive additional
implications. We also discuss network phenomena that do not appear to fit the flow model and discuss the possibility of a
second fundamental model, which we call the bond model. We close with a discussion of the merits of model-based network
theorizing for facilitating the generation of new theory, as well as a discussion of endogeneity in network theorizing.

R

Key words: theory; social network; flow model; bond model; endogeneity; structure
History: Published online in Articles in Advance April 11, 2011.

Introduction

characteristic elements of social network theorizing. We
have a particular interest in explicating the mechanisms
used in network theory to facilitate the generation of new
theory. In characterizing network theory, it is important
to emphasize that our objective is not to define what
should and should not be network theory. We do elaborate a view of what constitutes the heart of network
theorizing, but it is worth remembering that the network
analysis research program (in the sense of Lakatos 1980)
is a social enterprise that includes all kinds of different
researchers with different aims and backgrounds. There
is a great deal of work that is part of the broader SNA
research program that does not include the canonical elements we describe or that includes additional elements
that are not distinctive to the field.
It should also be noted that SNA theorizing encompasses two (analytically) distinct domains, which we
refer to as “network theory” proper and “theory of networks.” Network theory refers to the mechanisms and
processes that interact with network structures to yield
certain outcomes for individuals and groups. In the terminology of Brass (2002), network theory is about the
consequences of network variables, such as having many
ties or being centrally located. In contrast, theory of networks refers to the processes that determine why networks have the structures they do—the antecedents of
network properties, in Brass’s terms. This includes models of who forms what kind of tie with whom, who
becomes central, and what characteristics (e.g., centralization or small-worldness) the network as a whole will
have. In this paper, we focus on network theory proper,

Social network analysis (SNA) is currently popular.
As shown in Figure 1, publications referencing “social
networks” have been increasing exponentially over
time. The interest in networks spans all of the social
sciences and is rising even faster in physics, epidemiology, and biology. In management research, social
networks have been used to understand job performance (Sparrowe et al. 2001), turnover (Kilduff and
Krackhardt 1994; Krackhardt and Porter 1985, 1986),
promotion (Burt 1992), innovation (Obstfeld 2005),
creativity (Burt 2004), and unethical behavior (Brass
et al. 1998). Moreover, in management consulting, network analyses are fast becoming standard diagnostic
and prescriptive tools (e.g., Anklam 2007, Baker 2000,
Bonabeau and Krebs 2002, Cross et al. 2000).
Despite this popularity (and, perhaps, in part because
of it), there exists considerable confusion about network
theorizing. Even though certain network theories are
extremely well known—Granovetter’s (1973) strength
of weak ties theory has been cited more than 14,000
times1 —it is not unusual to read that network analysis
contains no theory of its own (Salancik 1995). In this
view, SNA is “just” a methodology, and what theory
there is “belongs to” other fields, such as social psychology. Moreover, as the term “social network” gains
caché, it is increasingly applied to everything from a
trade association to a listserv to a social media website
such as Facebook.
Our objectives in this paper are to clarify the concept of social network and to begin to identify the
1168

Borgatti and Halgin: On Network Theory

1169

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

Share of All Articles Indexed in Google Scholar with
“Social Network” in the Title, by Year

Share of all articles

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

Figure 1

1970

1975

1980

1985

1990

1995

2000

2005

2010

although we do find it useful to make a few comments
about theory of networks as well. In addition, we devote
a section of this paper to assessing whether considering
network theory without simultaneously treating theory
of networks harms the understanding of either.

What Is a Network?
A network consists of a set of actors or nodes along
with a set of ties of a specified type (such as friendship)
that link them. The ties interconnect through shared end
points to form paths that indirectly link nodes that are
not directly tied. The pattern of ties in a network yields
a particular structure, and nodes occupy positions within
this structure. Much of the theoretical wealth of network
analysis consists of characterizing network structures
(e.g., small-worldness) and node positions (e.g., centrality) and relating these to group and node outcomes.
It is important to realize that it is the researcher—
by choosing a set of nodes and a type of tie—that
defines a network. To appreciate the point, consider the
boundary specification problem (Laumann et al. 1983),
which refers to the question of how to select which
nodes to study. The naïve concern is that we may select
nodes “incorrectly,” accidentally excluding nodes that
should have been there and possibly including nodes that
should not have been. In reality, however, the choice of
nodes should not generally be regarded as an empirical
question. Rather, it should be dictated by the research
question and one’s explanatory theory. For example, we
may be interested in how centrality in an organizational
communication network is related to work performance.
Therefore, we study all communication ties among all
members of the organization. In making this choice, no
claim is made that only ties with other members of the
organization exist or matter, but rather that position in
the network defined by this kind of tie among this set
of actors has a measurable effect on performance. A different researcher might be interested in how a person’s
communications outside the organization interact with
the intraorganizational communication network to affect
performance. Yet another researcher, perhaps a psychologist, might ignore the influence of others altogether
(whether inside or outside the organization) and focus

on how personality or life experiences affect a person’s
performance.
In our view, part of the angst involved in the boundary
specification problem is due to confusing networks with
“groups.” A fundamental part of the concept of group is
the existence of boundaries. Even though we recognize
that boundaries may be fuzzy or uncertain (e.g., there
are part-time members, wannabees, conflicting views of
what the group is, etc.), the distinction between insiders
and outsiders is an important part of the group concept.
Therefore, when studying groups, we are justifiably concerned with establishing the boundaries of the group. For
example, if we are studying gangs in Los Angeles, we
would not want to approach the boundary specification
problem in a wholly etic2 way, such as defining gang
members as all young males living in a given area.
In contrast to groups, networks do not have “natural” boundaries (although, of course, we are free to
study natural groups, in which case the group boundaries determine our nodes).3 Networks also do not have
to be connected. A disconnected network is one in
which some nodes cannot reach certain others by any
path, meaning that the network is divided into fragments
known as components (see Figure 2). For those confusing networks with groups, this may seem an odd conceptualization. The advantage, however, is that it facilitates
the study of network evolution. For example, suppose
we study the freshman class at a university, focusing on
friendships. Initially, it may be that none of the freshmen is friends with any other, defining a maximally disconnected network with as many components as nodes.
Over time, friendships begin to develop, and the number of components may reduce rapidly. Eventually, it
is possible that all the actors are connected in a single
component in which every node can be reached from
every other by at least one path (even if very long).
Thus, by allowing the network to be disconnected, we
can trace the evolution of connectivity within it. In this
perspective, we do not ask “under what circumstances
will networks emerge” (Kahler 2009, p. 104), as if they
were groups. Rather, we ask how specific properties of
Figure 2

Network with Three Components

Borgatti and Halgin: On Network Theory

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

1170

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

the network, such as the level of fragmentation or characteristic path length, change over time.
A closely related issue is what “counts” as a tie.
A common beginner’s question is, which network questions should I ask in order to get at the network? Implicit
in the question is the idea—labeled the realist position
by Laumann et al. (1983)—that there is a “true” network
of relationships out there, and our job as researchers is to
discover it. Given that assumption, it is reasonable to ask
which social network questions have proven effective
at eliciting this network. However, a more sophisticated
view of social networks—labeled the nominalist position by Laumann et al. (1983)—holds that every network
question (such as “Who are you friends with?” or “Who
do you seek advice from?”) generates its own network,
and which to use is determined by the research question.
Thus, a given research question may lead us to examine the advice and friendship ties within an organization,
whereas another research question may lead us to examine “who-likes-whom” ties. No matter what kind of tie
we are interested in, measuring that kind of tie among all
pairs of nodes in the sample defines a network, and each
network will have its own structure and its own implications for the nodes involved. For example, being central
in a gossip network might be entertaining and beneficial,
whereas being central in a who-dislikes-whom network
might be painful and deleterious.
In practice, the kinds of ties that network theorists
tend to focus on can be categorized into two basic types:
states and events (see Table 1). States have continuity over time. This is not to say they are permanent,
but rather that they have an open-ended persistence.
Examples of state-type ties include kinship ties (e.g.,
parent of), other role-based relations (e.g., friend of or
boss of), cognitive/perceptual relations (e.g., recognizes
or knows the skills of), and affective relations (e.g., likes
or hates). State-type ties can be dimensionalized in terms
of strength, intensity, and duration.
In contrast, an event-type tie has a discrete and transitory nature and can be counted over periods of time.
Examples of event-type ties include e-mail exchanges,
phone conversations, and transactions such as sales or
treaties signed. Cumulated over time, event-type ties can
be dimensionalized in terms of frequency of occurrence
(e.g., the number of e-mails exchanged). It is these kinds
Table 1

Types of Social Ties

State-type ties

Event-type ties

• Kinship ties
• Interactions (e.g., giving advice to;
(e.g., brother of)
sending e-mail)
• Other role-based ties
(e.g., boss of or friend of)
• Cognitive (e.g., knows)
• Transactions (e.g., signing
• Affective
treaty with; making a sale)
(e.g., likes or dislikes)

of ties that researchers have in mind when they define
networks as a recurring pattern of ties (e.g., Dubini and
Aldrich 1991, Ebers 1997).
Both state-type ties and event-type ties can be seen
as roads or pipes that enable (and constrain) some kind
of flow between nodes.4 Flows are what actually pass
between nodes as they interact, such as ideas or goods.
Hence two friends (state-type social relation) may talk
(event-type interaction) and, in so doing, exchange some
news (flow). As we discuss in the next section, one large
swath of network theory is about how position in a backcloth network determines the timing or quantity of flows
to the actor occupying that position.
We might also note that, in empirical studies, researchers often make use of relational states and events
that are not, properly speaking, social ties. For example, a frequent proxy for social ties is group comembership, such as being on the same board of directors
or belonging to the same club. Similarly, coparticipation in events such as parties is used as a proxy for
unobserved social relationships. Other dyadic variables
of this type include geographic proximity (Allen 1977)
and similarity of traits such as behavior, beliefs, and attitudes (McPherson and Smith-Lovin 1987, McPherson
et al. 2001). From a theoretical point of view, comemberships, coparticipations, geographic proximities, and
trait similarities can all be seen either as dyadic factors
contributing to the formation of ties (e.g., meeting the
other members of your club) or as the visible outcomes
of social ties (as when close friends join the same groups
or spouses come to hold similar views).

Network Theorizing
To illustrate the nature and distinctive flavor of network
theorizing, we start by describing in detail two wellknown network theories, and we then analyze them for
their key characteristics. We begin with Granovetter’s
(1973) strength of weak ties (SWT) theory and then
move to Burt’s (1992) structural holes (SH) theory.
The SWT theory is organized as a set of explicit
premises and conclusions. The first premise of the theory is that the stronger5 the tie between two people, the
more likely their social worlds will overlap—that they
will have ties with the same third parties. As a result,
if A and B have a strong tie, and B and C have a
strong tie, the claim is that A and C have an increased
chance of having at least a weak tie (e.g., A and C are
acquaintances). This is a kind of transitivity—one that
some authors have called g-transitivity (Freeman 1979).
The reason for this transitivity, Granovetter argues, is
that the underlying causes of tie formation have this
kind of transitivity built into them. For example, people
tend to be homophilous, meaning that they tend to have
stronger ties with people who are similar to themselves
(Lazarsfeld and Merton 1954, McPherson et al. 2001).

Borgatti and Halgin: On Network Theory

1171

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

Homophily is weakly transitive because if A is similar
to B, and B is similar to C, then A and C are likely
to be somewhat (i.e., weakly) similar as well. To the
extent that similarity causes ties, this will induce weak
transitivity in the tie structure as well.
The second premise of SWT is that bridging ties are
a potential source of novel ideas. A bridging tie is a tie
that links a person to someone who is not connected
to his or her other friends.6 The idea is that, through a
bridging tie, a person can hear things that are not already
circulating among his close friends. In Figure 3, A’s tie
with G is a bridging tie. Because A is the only person
in her social group with a tie outside the group, A has
the benefit of hearing things from G that the rest of A’s
group has not yet heard.
Putting the two premises together, Granovetter reasons that strong ties are unlikely to be the sources of
novel information. The reason is as follows. First, bridging ties are unlikely to be strong. According to the first
premise, if A and G have a strong tie, then G should
have at least a weak tie to A’s other strong friends,
which would imply that the A–G tie was not a bridge,
because there would be multiple short paths from A to
G via their common acquaintances. Therefore, it is only
weak ties that are likely to be bridges. Second, because
bridges are the sources of novel information, and only
weak ties are bridges, it is the weak ties that are the
best potential sources of novel information.7 Granovetter
uses this theory to explain why people often get or at
least hear about jobs through acquaintances rather than
close friends. In this sense, the theory is one of individual social capital, where people with more weak ties
(i.e., more social capital) are more successful. Granovetter also applies the theory at the group level, arguing
that communities with many strong ties have pockets of
strong local cohesion but weak global cohesion, whereas
communities with many weak ties have weak local cohesion but strong global cohesion. Using the case study
of Boston, in which the city assimilated one adjacent
community but failed to assimilate another, he suggests
that a community’s diffuse, weak-tie structure constitutes
group-level social capital that enables the group to work
together to achieve goals, such as mobilizing resources
and organizing community action to respond to an outside threat.
Figure 3

Another well-known network theory is Burt’s (1992)
structural holes theory of social capital. The theory of
SH is concerned with ego networks—the cloud of nodes
surrounding a given node, along with all the ties among
them. Burt argues that if we compare nodes A and B in
Figure 3, the shape of A’s ego network is likely to afford
A more novel information than B’s ego network does for
B, and as a result, A may perform better in a given setting, such as an employee in a firm. Both have the same
number of ties, and we can stipulate that their ties are
of the same strength. However, because B’s contacts are
connected with each other, the information B gets from,
say, X may well be the same information B gets from Y.
In contrast, A’s ties connect to three different pools of
information (represented by the circles in Figure 4). Burt
argues that, as a result, A is likely to receive more nonredundant information at any given time than B, which
in turn can provide A with the capability of performing
better or being perceived as the source of new ideas.
Kilduff (2010) argues that Burt’s portrayal of
the social world differs significantly from that of
Granovetter along a variety of dimensions. For example,
Kilduff sees Granovetter as embracing a serendipitous
world in which people form ties that only incidentally
prove useful, whereas Burt embraces a more strategic
and instrumental view. However, at the level of the specific theories of SWT and SH, it should be obvious
that Burt’s theory is closely related to Granovetter’s. In
Burt’s language, A has more structural holes than B,
which means A has more nonredundant ties. In Granovetter’s language, A has more bridges than B. But
whether we call them nonredundant ties or bridges, the
concept is the same and so are the consequences: more
novel information. Where Granovetter and Burt differ
is that Granovetter further argues that a tie’s strength
determines whether it will serve as a bridge. Burt does
not disagree and even provides empirical evidence that
bridging ties are weaker in that they are more subject to decay (Burt 1992, 2002). However, Burt sees tie
Figure 4

Node A’s Ego Network Has More Structural Holes
than B’s

X

Bridging Tie from A to G

B

Y
A
A

G

C
Note. Removing the A–G tie disconnects the network.

B

Borgatti and Halgin: On Network Theory

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

1172

strength as a mere “correlate” of the underlying principle, which is nonredundancy (1992, p. 27). Thus, the difference is between preferring the distal cause (strength
of ties), as Granovetter does, and the proximal cause
(bridging ties), as Burt does. The first yields an appealingly ironic and counterintuitive story line; the second
“captures the causal agent directly and thus provides a
stronger foundation for theory” (Burt 1992, p. 28). In
addition, Granovetter uses getting jobs as an outcome
of having nonredundant information, whereas Burt uses
getting promoted. In our view, these are small differences in ornamentation. Both theories are based on the
same underlying model of how networks work.

Characterizing Network Theory
Examining SWT and SH from a metatheoretical point
of view, we can see two features of network theory that
are highly characteristic. First, the twin notions of structure and position play fundamental roles. For example,
in SWT, the reason weak ties are useful is not because
they are inherently so but because it is the weak ties that
tend to bridge network clusters. It is their structural role
that makes them advantageous. Similarly, in SH, it is the
shape of the ego network around a person that confers
advantages to that person. Note that the theory ignores
egos’ own attributes (such as how creative they are) and
also the attributes of egos’ contacts (e.g., how smart they
are, or how gullible or powerful) and only looks to see
whether the alters are numerous and unconnected. This
is not to say that ego and alter attributes are not important (they may well be much more important); it is just
that the agenda of the theory—and the charter of network theory in general—is to explicate the connection
between structure and outcome, and one aspect of this
agenda is the study of the pure effects of structure. To be
clear, the general agenda of examining the consequences
of network structure includes the examination of how
structure and attributes interact to yield outcomes. But
a piece of that investigation is the exploration of how
structural differences alone have effects.8
Second, there is an implicit theory of network function; in the case of SWT and SH (but not all network
theories), the network function is the flow or distribution
of information. In effect, SWT and SH rely on an underlying model of a social system as a network of paths that
act as conduits for information to flow. We refer to this
as the flow or pipes model. The abstract flow model carries with it some basic assumptions, such as the longer a
path is, the longer it takes something to traverse it. From
this general model, we can readily derive a number of
theoretical propositions that form the core of theories
like SWT and SH. For instance, nodes that are far from
all others will, on average, receive flows later than nodes
that are more centrally positioned. Similarly, nodes that
are embedded in locally dense parts of a network will

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

often receive the same bits of flow from their various
contacts, because the contacts are tied to each other as
well. These flow outcomes (time until arrival; amount of
nonredundant flow received) are then related to a variety
of more general outcomes, such as creativity, likelihood
of promotion, getting a job, etc.
As an aside, by specifying additional features of the
central process in the model, we can generate additional theoretical implications. For example, for the flow
model, we can specify different variations for how flows
move through the network (Borgatti 2005). For instance,
a dollar bill moving through a network transfers from
actor to actor in such a way that it is never in two places
at the same time. In contrast, a virus or bit of news
duplicates from actor to actor so that when A passes it
to B, A retains a copy. Another dimension of how things
can flow is what kinds of paths or trajectories they trace
through the network. Some viruses, for instance, tend not
to reinfect a node, either because the node has become
immune or because it is now isolated or dead. In network
theory, this kind of trajectory is known as a true path. In
contrast, gossip can easily pass through a node multiple
times, because transmitters do not necessarily know who
has already received it. However, we might argue that
gossip does not revisit ties. That is, once I have told you
a story, I am unlikely to tell you the same story again
(forgetfulness just adds a bit of noise to the system). In
network theory, this kind of traversal is known as a trail.
A dollar bill illustrates yet another type of path, technically called a walk, which is unrestricted with respect
to whether it reuses nodes or ties (e.g., on a given day,
a person might give the bill to a store and the next day
receive it back in change for another purchase. It could
then be given by that person to that store yet again in a
third transaction).
Given that things flow through the network according
to certain rules, some obvious outcomes can be predicted
as consequences of the network structure. For instance,
at the node level, we may be interested in the expected
time until (first) arrival of whatever is flowing through
the network. Certain (central) nodes are positioned in
such a way that, on average, they receive the flow sooner
than other nodes. We may also be interested in how often
or with what level of certainty a node receives a given
bit of flow. It should be noted that both of these flow
outcomes are fully defined within the model but are not
necessarily the empirical outcomes that we actually measure. Network research consists of equating these model
outcomes with other constructs such as the likelihood
of getting a job (Granovetter 1973, 1974), being promoted (Brass 1984, 1985; Burt 1992), or being creative
(Burt 2004, Perry-Smith 2006). Thus, hypotheses that
are actually tested in empirical studies relate features of
the observed network to outcomes such as performance
in an organizational setting, and network theory consists
of elaborating how a given network structure interacts

Borgatti and Halgin: On Network Theory

1173

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

with a given process (such as information flow) to generate outcomes for the nodes or the network as a whole.
The flow model is the most developed theoretical
platform in network theory, but it is not the only one.
The field has clearly identified phenomena and developed theoretical explanations that cannot be reduced to
the flow model. One such area is the study of power.
Cook and Emerson (1978) pioneered the experimental
study of the exercise of power in exchange networks. In
their experiments, subjects occupied nodes in a network
designed by the researcher. The subjects played a game
in which, at each round, they had the option to negotiate a deal with someone they were connected to. At
each round, each subject could only close on one deal.
Across rounds, the subjects’ objective was to make as
many deals at the best possible terms as possible. For
example, for the network in Figure 5, Cook and Emerson found that the subject in position B was able to
negotiate the best deals, even though subjects were not
shown the structure of the network they were embedded in. From Cook and Emerson’s point of view, the
fundamental advantage that B enjoys is the dependency
of others, which is a function of the (lack of) availability of alternatives on the part of B’s potential partners.
Node B has two alternatives available for making a deal,
whereas A and C have no alternatives to B and are therefore wholly dependent on B. This positional advantage
is very different from the concept of centrality, which
largely emerges from the flow model. This can be seen
in the experimental results for the network in Figure 6,
in which B and D emerge as high-power positions and
A, C, and E have very low power. This might seem surprising given that B, C, and D all have two potential
trading partners, but the difference is that C’s partners
B and D both have better alternatives to C, namely, the
wholly dependent A and E. Thus, whereas a basic principle in centrality phenomena is that being connected
to well-connected others implies greater centrality, in
power phenomena it can be the other way around: being
connected to weak others makes one powerful, and being
connected to powerful others makes one weak (Bonacich
1987, Markovsky et al. 1988, Marsden 1983).
What is especially interesting about network power is
that network structure (and location within that structure) matters, and yet the basis for network power is not
Figure 5

Experimental Exchange Network in Which Node B
Has the Most Power
A

B

the accumulation or early reception of a resource that is
flowing over well-positioned nodes, as it is in the flow
model. This is especially clear in the experimental setting because the rules of the game explicitly prohibit the
flow of resources. Nor is power itself flowing, because
if it were, nodes adjacent to a powerful node would be
empowered.
Another way to look at network power is in terms of
coordination and virtual amalgamation. Consider node E
negotiating with a set of alters A1–A4, as shown in Figure 7. Because there is only one E and several As, one
might expect E to have a difficult time; however, this
would only be true if the As worked together as a unit.
One way this can happen is if the As are bound together
by ties of solidarity. In the extreme, this can be seen as
converting the As into a single node that can deal with E
on an equal basis—that is, a mechanism of virtual amalgamation, as shown in Figure 8. This is the principle
behind unionization.
Figure 7

Negotiation Network

A1

A2
E

A3

A4

Figure 8

Negotiation Network with Unionization of the As

A1

A2

C

E

Figure 6

A

Experimental Exchange Network in Which Nodes
B and D Have the Most Power

B

C

D

E

A3

A4

Borgatti and Halgin: On Network Theory

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

1174

A closer look at the unionization example suggests
that two kinds of relationships among nodes are implicit.
One is the potential exchange tie that exists between E
and its alters. The other is the solidarity ties that may
exist among the alters. A key point is that the alters
in Figures 7 and 8 are essentially of a type, with the
same interests (e.g., to exchange with E) and capabilities (making them interchangeable from E’s point of
view). This effectively allows E to induce competition
(itself a kind of tie) between the As (which can be mitigated by ties of solidarity, as in unionization). A different sort of case is the so-called “network organization,”
in which a set of autonomous organizations coordinate
closely, as if comprising a single superordinate entity
(Jones et al. 1997, Powell 1990). By working together
they can accomplish more than they could alone. We can
view this phenomenon as virtual capitalization, meaning
that the bonds between the nodes enable the nodes to
act as if they were transferring the capabilities of the
other nodes to each other, but without actually doing so.
Supply chain networks have a similar character. Rather
than vertically integrating and taking on the tasks and
abilities of upstream suppliers (as in the simple capitalization process that occurs in the flow model), the firm
has bonds with those suppliers that enable it to behave
as if it had those capabilities. More generally, this is the
same phenomenon studied by principal/agent theory, in
which the agent acts in the interests of a principal without the principal having to do the work itself.
In all these examples—from exchange experiments to
principals and agents—a common underlying theme is
that the network tie serves as a bond that aligns and
coordinates action, enabling groups of nodes to act as a
single node, often with greater capabilities. The bonding
function serves as the basis for what we call the bond
or coordination model, and it is the analogue of the flow
function in the flow model.
An interesting question is whether the work on experimental exchange networks can be derived from the bond
model. There is a point of commonality, which is that
when a pair of nodes makes a deal in a given round, the
nodes become, momentarily, a unit that excludes those
not part of the deal. From this perspective, a node’s
advantage derives from its inexcludability. For example,
consider the positions of B and D in Figure 6: if C and D
make a deal, B can make a deal with A. If D makes
a deal with E, then B can make a deal with either A
or C. There is no combination of outcomes in any round
that does not give both B and its twin D the option of
making a deal.
Like the flow model, the bond model permits a number of derivations, which in turn enable us to construct
measures of power and predict power-based outcomes.
For example, the logic of dependency and excludability dictates that the existence of node A on the other
side of node B is detrimental for node C, who would

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

Figure 9

Nodes A and H Are Structurally Isomorphic

B

A

E

C

D

I

H

F

G

J

rather that B not have any alternatives to itself. Generalizing this a bit, paths of an even length emanating
from a node reduce its power, whereas paths of an odd
length increase its power. This theorem is the basis for
several measures of network power, including the graphtheoretic power index (Markovsky et al. 1988) and beta
centrality (Bonacich 1987, 2007).
Another derivation from the bond model is that isomorphic nodes will have similar outcomes even if they
are not reachable from each other (as flow-based processes would require). For example, in Figure 9, nodes A
and H are structurally isomorphic and therefore must
have the same structural advantages and disadvantages.
Holding constant individual differences in the abilities
of actors occupying network positions, we can expect
that structurally isomorphic nodes will have similar
outcomes.
The third thing to note about network theory is that
the core concept of the field—the network—is not only
a sociological construct but also a mathematical object.
As a result, it is sometimes possible to use the machinery of mathematics to generate new theory. For instance,
Rapoport (1963) and others show that transitivity tends
to create highly clustered graphs that have many long
paths or disconnected components, which means that
networks with high transitivity are slow or incomplete
diffusers. This, of course, is the basis for SWT and SH
theory. More generally, the coincidence of sociological
networks and mathematical networks makes it easy to
generate formal theory that is expressed in mathematical form. This can be a blessing, but it also carries
with it the danger that the nonmathematically inclined
will not see it as theory at all, but rather as some form
of statistics. A good example is the notion of betweenness centrality, which is defined by the formula shown
in Equation (1). It has been shown (Borgatti 2005) that

Borgatti and Halgin: On Network Theory

1175

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

the betweenness formula gives the expected values of
the number of times something reaches a node in a certain flow process (namely, one in which the things flow
along shortest paths, and when there are multiple equally
short paths, they choose one of them with equal probability). Thus, what looks like methodology is in fact
formal theory based on the flow model:
bk =

X gikj
i1 j

gij

1

(1)

where bk is the betweenness of node k, gij is the number
of geodesic paths from i to j, and gikj is the number of
geodesic paths from i to j that pass through k.
It is worth noting that even things as technical as
the notions of structural equivalence (Lorrain and White
1971) and regular equivalence (Everett and Borgatti
1994, White and Reitz 1983) were explicitly developed in an effort to formalize the social role theory of
Linton (1936), Nadel (1957), Merton (1959), and others. Similarly, the notions of clique (Luce and Perry
1949), n-clique (Luce 1950), k-plex (Seidman and Foster
1978), and other subgroups, which sound so methodological, were actually attempts to state with mathematical precision the concept of group that Cooley (1909),
Homans (1950), and others had discussed at a more intuitive level.9

Goals of Network Theorizing
So far, we have focused on characterizing modes of
explanation in network models. In this section we focus
on characterizing the kinds of outcomes that these models are used to explain. As in much of social science,
there are two generic types of outcomes that network
research has sought to explain. The first can be broadly
termed choice and includes behaviors, attitudes, beliefs,
and (in the case of collective actors like organizations)
internal structural characteristics. Network research on
choice has often been framed in terms of similarity of
choice, as in explaining which pairs of nodes make
similar choices. For this reason, work in this area is
often referred to as the social homogeneity literature, as
noted by Borgatti and Foster (2003). The second generic
outcome is success, which includes performance and
rewards, whether at the node or whole network level.
Work in this area is known as the social capital literature. Combining these two generic outcomes with the
two explanatory models we have outlined, we get a simple typology of network theorizing.
As shown in Table 2, the top right quadrant, contagion, consists of flow-based explanations of (similarity of) choice, which is a well-populated segment of the
literature. The principal example of this kind of work
is diffusion or adoption of innovation studies in which
nodes are conceptualized as influencing each other to

Table 2

Network Functions (Mechanisms) by Model and
Research Tradition
Research tradition

Model

Social capital

Social homogeneity

Network flow model
(ties as pipes)
Network coordination
model (ties as bonds)

Capitalization

Contagion

Cooperation

Convergence

adopt their traits. For example, work in the organizational theory literature posits that one reason organizations have similar structures is diffusion (Davis 1991,
DiMaggio and Powell 1983). Extending Dimaggio and
Powell (1983), we can use locus of agency to distinguish four different types of diffusion. As shown in
Table 3, DiMaggio and Powell discuss mimetic processes, in which the adopter actively seeks to copy a trait
from a node in its environment, and coercive processes,
in which the node is forced by a node in its environment
to adopt a trait (such as a certain accounting system). In
addition, we identify two other processes, the apprentice
process, in which both the ego and its environment are
actively trying to help the ego get what the alter has,
and the osmotic process, where neither party is actively
expending energy to enable the transfer, but it happens
anyway (as when the ego learns a new term or concept
simply by listening to the alter).
The bottom right quadrant of Table 2, convergence,
contains bond-based explanations of homogeneity. Work
in this area includes research on structural equivalence
(Lorrain and White 1971, Burt 1976), which posits that
nodes adapt to their environments, and as a result nodes
with similar structural environments will demonstrate
similarities (Erickson 1988). For example, suppose two
people in different parts of the world are highly central in the advice networks around them; that is, everybody is constantly seeking their advice. As a result,
both of them develop a dislike of the phone, because
it so often brings work for them. Hence, sameness
in terms of centrality level leads to similar attitudes.
Work in this area can also be seen as a special case
of coordination in which nodes behave similarly rather
than simply in concert (as in the bottom left quadrant),
which is similar to the sociological concept of Gemeinschaft (Tönnies 1912). Other work we would classify
Table 3

Breakdown of Diffusion Processes by Locus of
Agency
Alter (the social environment)

Ego (the one
adopting)
Active
Passive

Active

Passive

Apprentice processes
Coercive processes

Mimetic processes
Osmotic processes

Borgatti and Halgin: On Network Theory

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

1176

in this quadrant includes the networks-as-prisms concept of Podolny (2001), along with the empirical work
of Kilduff and Krackhardt (1994) and the identity-based
network research of Podolny and Baron (1997) and
Halgin (2009), which suggests that network ties provide
informational clues to audiences regarding the quality
and identity of an actor.
The top left quadrant of Table 2, capitalization, contains flow-based explanations of achievement. The basic
concept here is that social position in a network provides
access to resources. Work in this area is exemplified by
strength of weak ties theory (Granovetter 1973), Lin’s
(1988) social resource theory, and the information benefits theory of structural holes (Burt 1992).
Finally, the bottom left quadrant of Table 2, cooperation, consists of bond-based explanations of achievement. Here, combinations of nodes act as a unit, excluding others and exploiting divisions among them. This is
exemplified by the stream of research on experimental
exchange networks (Bonacich 1987, Cook and Emerson
1978, Markovsky et al. 1988), as well as the control benefits theory of structural holes (Burt 1992). For a more
detailed discussion of the work falling into each quadrant, see the review by Borgatti and Foster (2003).10

Discussion
In this section, we comment on the analysis presented in
this paper, organizing our comments into two sections,
entitled model-based theorizing and endogeneity.
Model-Based Theorizing
In this paper, we have argued that at least some portions
of network analysis can be described as model-based
theorizing and have outlined two fundamental models,
the flow and bond models, that underlie extant network
theorizing. According to Lave and March (1975), modelbased theorizing is one of the strongest forms of theorizing. In model-based theorizing, we imagine an observed
state of affairs as the outcome of an unseen process,
which is what is specified by the model. Given the
model, you can derive testable implications, including
the original observations that led you to postulate the
model. Ideally, a model can also be expressed formally
so that the machinery of mathematics and/or simulation
can be used to derive additional implications that might
be difficult to develop by simple intuition (e.g., Everett
and Borgatti 1994, Lorrain and White 1971, Luce and
Perry 1949, White and Reitz 1983). The implications are
used to test the theory as well as to apply the theory to
new situations.
One feature of model-based theorizing is the separation between the abstract elements of the model and
the mapping of those elements to the real world. Hence,
we should write network theory at the level of, say,
the function of enabling something to flow from one

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

node to another, not at the level of, say, who-likes-whom
ties. For example, in SWT, Granovetter (1973, p. 1361)
specifies quite clearly what a strong tie is (namely, a
combination of time, emotional intensity, intimacy, and
reciprocal services). However, this definition is open to
debate and is not appropriate in all settings, such as
when the nodes are firms. A closer look at the theory
shows that a specific definition is actually unnecessary:
any type of tie that has the property of generating gtransitivity will do. The rest of the theory does not make
use in any way of the fact that strong ties were defined in
terms of emotional intensity and the rest. The only property of strong ties that is actually utilized is the property
of g-transitivity.
There is an analogy here to object-oriented computer
programming (OOP), in which real-world entities are
modeled as classes of “objects” that consist of data
along with procedures (called “methods”) that operate
on them. A key principle of OOP is that one should
program to an interface rather than to an implementation. What this means is that higher-level code should
not have to know the details of how lower-level code
works—the functions of the lower-level code should be
encapsulated so that higher-level code deals only with
the functions of the code, not the means by which they
are accomplished. For example, if we are modeling interactions among animals, our main code should not have
to know how, exactly, a cow sounds or a dog moves.
Rather, it should be able to issue to the object representing a particular animal a general command such
as “make sound” or “move” and have this interpreted
appropriately by the object, which knows how to make
its own sound and how to execute its own way of moving. In this way, any changes to how a specific kind
of animal makes sounds or moves will not affect the
main program, and new types of animals can easily be
incorporated.
In network theory, the concept corresponding to
OOP’s object is the network, and what corresponds to
OOP’s methods is the set of processes or functions
that we define on the network, such as flow of information. In our view, this analogy helps point the way
toward dealing with issues of context and culture. For
example, a theory built on a particular definition of tie
(e.g., friendship) will run into problems when we try to
apply it cross-culturally, because friendship has different
implications in different cultures and settings. A better
approach is to build theory at the level of abstract ties
that have certain properties needed by the theory (e.g.,
ties create shared identity, or ties transfer resources).
Then to apply the theory in a given setting, we use our
situated ethnographic knowledge to find an appropriate,
specific type of tie that, in that context, entails the functions needed by the theory.
The analogy also helps clarify the question of whether
we can apply the same network theories to collective

Borgatti and Halgin: On Network Theory

1177

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

and/or nonhuman actors—such as firms—as we do to
actors that are individual persons (Madhavan 2010). For
example, if we wish to apply SWT theory to firms, we
need only ensure that the kind of ties we study have
the property of g-transitivity and serve as pipes through
which resources flow. We need not worry that ties among
firms do not have “emotional intensity” or “intimacy”
as long as there is a kind of interfirm tie that has the
two properties that Granovetter’s model depends on—
namely, transitivity and enabling the flow of information. Of course, it should be noted that different kinds
of nodes have different capabilities, which needs to be
taken account of in generating the auxiliary theorizing
that links model outcomes to such outcome variables as,
say, performance or creativity. For example, when an
individual hears two bits of information, he has a fighting chance of integrating them, but when a firm hears
two bits of information, it may be different parts of the
organization that house them, and the bits may never
come together in the same space to be integrated.11
Endogeneity
In this paper, we have separated network theory from
theory of networks in an attempt to trade breadth for
depth. However, reviewers of this paper have raised
some concerns about this separation. First, there is the
question of whether the distinction is “merely” analytical, because it might be expected that, in reality, the
two kinds of processes occur together. Second, there is
the concern that we cannot correctly predict outcomes
of network structure if we have not taken account of
how the network got there—that is, the trajectory matters. Third, there is the question of endogeneity. Endogeneity means different things in different contexts, but
one sense of the term is that factors seen as causing the
outcome are in some part dependent on the outcome.
Finally, the issue of agency comes to mind. If actors
deliberately shape the networks around them for their
benefit, can it really be said that it was network structure
that led to the benefit?
To begin our discussion, let us make clear on a semantic level that network theory and theory of networks
are not disjoint sets. Recall that we defined the domain
of network theory as the consequences of network processes and structures. In our examples, these consequences were things such as performance or reward.
However, it is obvious that the consequences of network processes can include other network phenomena,
in which case network theory is simultaneously theory of networks, which is to say we have a network
theory of networks (see Table 4). In a network theory
of networks, both independent and dependent variables
involve network properties. An example is the cascade
of effects that can be produced by the formation of a
positive or negative tie between two actors. For instance,
suppose spouses Bill and Nancy develop a negative tie

Table 4

Network Theory and Theory of Networks
Dependent variable

Independent variable

Nonnetwork variable
as outcome

Network variable
as outcome

Nonnetwork variable
as antecedent

(Nonnetwork
theory)

Theory of
networks

Network variable
as antecedent

Network
theory

Network theory of
networks

between them, culminating in an acrimonious divorce.
According to balance theory, we can expect that a third
person, Sally, with a strong positive tie to both parties, will experience stress and be likely to weaken the
tie with one of them—i.e., choose sides. This in turn
has a ripple effect on Sally’s friends, who may also be
forced to choose sides. Another example is the interaction between homophily and centrality (Ibarra 1992).
If actors have a marked tendency to be homophilous
with respect to race, and one race has a clear numerical
majority, we can expect that members of the majority
race will be more central.
A more interesting question is whether, as Salancik
(1995) seems to feel, a network theory must include
a theory of networks. In particular, are there any circumstances where we must take into account how a
network reached a given structure in order to understand the consequences of that structure? Common sense
would suggest that the answer is “yes.” For example,
consider two nodes who occupy identical positions in a
network (both have many structural holes) and have similar motivations. However, node A reached that position
through a long campaign of strategic relationship building, whereas node B arrived at it serendipitously and in
fact is unaware of the potentialities of its position. We
can readily imagine that, in a population of nodes like
A, the correlation between structural holes and power
will be higher than in a population of nodes like B, who
do not think to exploit their position. Thus, the causal
link between holes and power varies depending on how
the nodes got their holes. Or does it? The key difference
between A and B is that A’s journey to that position
implies awareness of its value, allowing A to exploit it.
But suppose there other ways of becoming aware of the
value of one’s structural holes. For example, suppose
node B attends an executive education class on social
network analysis. Given that B has the same position
and same awareness as A, shouldn’t the consequences
for B be the same as the consequences for A, all else
being equal?
Thus, on closer inspection, the answer to whether network theory must include theory of networks would
seem to be “no.” If a model has been constructed that
embodies the mechanisms that convert a given set of
inputs at time T to an output at T + 1, then given that
input, nothing else is needed to explain the outcome. In

Borgatti and Halgin: On Network Theory

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

1178

practice, however, it is a little more complicated. For one
thing, knowing the input at T may involve a longitudinal
analysis. As a very simplistic example, suppose an outcome is a function of whether a network is increasing
or decreasing in density (e.g., the nodes make certain
choices when they perceive the density to be changing
in a particular direction). A snapshot of the network at
a single point in time does not tell us whether the density is waxing or waning. However, once we have determined, via longitudinal analysis, whether it is waxing or
waning at time T , we can set the “momentum” variable
at T to the observed value, and we then have all the
information we need to understand what happens next.
As a more substantive example, consider the strength
of weak ties theory. An appealing feature of the theory is that it spans both the theory of networks domain
and the network theory domain. As discussed, a key
premise of SWT is that networks form in such a way
that they exhibit g-transitivity. It can then be derived
that bridging ties are unlikely to be strong ties. This is
the “theory of networks” portion of the theory. If we
then combine another premise (that bridging ties are the
most likely source of novel information), we can conclude that the structural property of having many weak
ties is likely to be associated with access to more novel
information, which in turn may be associated with performance gains. This is the “network theory” portion.
The combination of the two portions is both satisfying
and elegant. But do we need the first part to get the
second part right? Strictly speaking, the answer is “no.”
To derive the hypothesis that weak ties will be associated with strong performance and that this is mediated
by access to novel information, we do not need to know
why networks have g-transitivity—merely that they do.
However, there is also the matter of how satisfying a
theory feels. In any theory in which X leads to Y, we can
wonder what leads to X. In some cases this feels like an
urgent and necessary question. For example, if the explanation for why people divorce is “because they want to,”
we are likely to demand an explanation of why they want
to. In other cases, there is enough of a sense of process
or mechanism in the theory that we are willing to back
off of the chain of infinite regress. For example, among
other arguments, Granovetter (1973) uses balance theory
to explain g-transitivity. According to balance theory, a
person seeks to be congruent with those she likes. When
she is not, she feels dissonance and seeks to reduce it.
We could ask why, but most of us in the management
field are willing to let that one go and let the psychologists deal with it. Ultimately, at what point we feel
enough explanation has been given to be satisfying in
a given context is a question for the sociology of science and not a question about a particular field such as
network theory.
It should be noted that, in principle, the ability to
theorize about consequences of networks independently

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

of antecedents does not absolve the field from resolving issues of endogeneity in a given empirical inquiry.
For example, Lee (2010) finds that in a biotech setting,
the cross-sectional correlation between structural holes
and innovative performance disappears when controlling
for inventors’ past performance. Thus, in that particular case, it appears that it is performance that creates
holes rather than the other way around, and whatever
is responsible for performance is stable over time so
that past performance predicts future performance. Thus,
it could be an individual characteristic such as skill or
personality that causes both structural holes and performance. This is an important result but should not be
misread as saying something fundamental about network
theorizing. In every field study we must be concerned
about whether A causes B, or the other way around,
or whether both are caused by an uncontrolled third
variable.12
Finally, we take up the issue of agency as it relates
to endogeneity. One of the legacies of the social capital approach in social network research is the notion
that ties and position can be “good,” that is, associated
with positive outcomes such as performance or reward.
Inevitably, this leads to the following bit of reasoning: if occupying a certain position in the network is
rewarding, we can expect actors to take steps to achieve
that position. Thus, the network structure is not a given
in the sense of an exogenous variable, but rather it is
shaped by the actors specifically to achieve the very outcomes that we researchers associate with those structures. Therefore, any theory of social networks must take
into account actors’ agency in creating those networks.
The problem with this, as we have pointed out, is that it
is not the actors intentions and actions leading to occupying a certain position that creates the outcome but the
actual occupation of the position. A rock dropped from
the same place in the same way has the same outcomes
regardless of whether it was dropped on purpose or by
accident. Given the same conditions, the outcomes are
the same.13
One thing this discussion highlights is the importance
of node attributes and contextual factors in network
research. Occupying a certain structural position carries certain potentialities, but the actual outcomes may
depend on a number of additional factors, including how
the actor plays it. How they play it may be a function
of how they got there, and so knowing how they got
there could give our predictive ability a boost. However,
it is not the journey itself that is the theoretical variable
but rather the complex of conditions (e.g., state of mind,
skills, motivations) at the end of the journey that is the
causal agent. If we can measure that condition directly,
there is no need to code the journey. In this sense, if we
find that we cannot predict how X leads to Y without
knowing how X came about, it is evidence that our theory of how X leads to Y is incomplete: we are missing

Borgatti and Halgin: On Network Theory

1179

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

a node attribute or other contextual factor that interacts
with network position to bring about the outcome being
modeled.

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

Conclusion
Our principal goal in this paper has been to dissect and
characterize network theorizing. In doing so, we have
argued that much of network theory (and methodology)
is based on the flow model, which is now well elaborated and serves to unify large portions of network theory. We have also argued that another model, the bond
model, is under development and can potentially unify
several other areas of inquiry. More generally, we hope
that our discussion of network concepts and model-based
theorizing in network research will help clarify existing
theory as well as facilitate the generation of new theory.
Acknowledgments
The authors are grateful to the Organization Science editorial
team and the members of the LINKS Center for Social Network Analysis at the University of Kentucky, especially Dan
Brass, Ginny Kidwell, Joe Labianca, and Ajay Mehra, for their
help in shaping the authors’ thinking. This work was funded in
part by Grant HDTRA1-08-1-0002-P00002 from the Defense
Threat Reduction Agency.

Endnotes
1

Source. Google Scholar.
Etic versus emic is a distinction made in cognitive anthropology between organizing the world using researcher-driven
criteria (etic) and organizing things the way natives do (emic).
The terms come from the linguistic distinction between phonetic (how things sound) and phonemic (what things mean).
3
It should be acknowledged, however, that there is a literature
that labels organizational forms intermediate between hierarchies and markets as “networks.” In this literature, a network
refers to a group of organizations working closely together,
almost as if they were one superordinate organization.
4
This is Atkin’s (1972) “backcloth/traffic” distinction.
5
Granovetter (1973, p. 1361) provides a definition of strength
of tie, but it is useful to realize that any definition of
tie strength that preserves the first premise can be used
(Freeman 1979).
6
More technically, a bridge is a tie between A and B that, if
removed, would leave a very long path (if any at all) connecting A to B. A bridge, then, is a shortcut in the network.
7
Note that there is no claim that all weak ties are sources of
novel information—just the ones that happen to be bridges.
Granovetter’s point is simply that it is weak ties rather than
strong ties that are more likely to be bridges.
8
We are grateful to an anonymous reviewer for pointing
this out.
9
An explanation of these terms is beyond the scope of this
paper. For a review, consult Wasserman and Faust (1994).
10
The terminology in Borgatti and Foster (2003) is somewhat
different, but the underlying ideas are the same.
11
We are grateful to an anonymous reviewer for making
this point.
2

12

Note that the possibility that A and B both cause each
other should not concern us: if our theorizing suggests that A
causes B, and we find that A and B cause each other, then
our theory is supported. The fact that we have also learned
something about the causes of A is a side benefit.
13
It might be argued that this is not true in a court of
law, where the consequences for the rock-dropper may differ
depending on the court’s perception of the dropper’s intentions. But then the conditions are not the same. From a legal
perspective, a rock dropped by accident versus with an intent
to kill are two different events.

References
Allen, T. J. 1977. Managing the Flow of Technology. MIT Press,
Cambridge, MA.
Anklam, P. 2007. Net Work: A Practical Guide to Creating and
Sustaining Networks at Work and in the World. ButterworthHeinemann, Woburn, MA.
Atkin, R. H. 1972. From cohomology in physics to q-connectivity in
social science. Internat. J. Man-Machine Stud. 4(2) 139–167.
Baker, W. 2000. Achieving Success Through Social Capital: Tapping
the Hidden Resources in Your Personal and Business Networks.
Jossey-Bass, San Francisco.
Bonabeau, E., V. Krebs. 2002. Model behavior. Optimize (12) 1–3.
Bonacich, P. 1987. Power and centrality: A family of measures. Amer.
J. Sociol. 92(5) 1170–1182.
Bonacich, P. 2007. Some unique properties of eigenvector centrality.
Soc. Networks 29(4) 555–564.
Borgatti, S. P. 2005. Centrality and network flow. Soc. Networks 27(1)
55–71.
Borgatti, S. P., P. Foster. 2003. The network paradigm in organizational research: A review and typology. J. Management 29(6)
991–1013.
Brass, D. J. 1984. Being in the right place: A structural analysis of
individual influence in an organization. Admin. Sci. Quart. 29(4)
518–539.
Brass, D. J. 1985. Men’s and women’s networks: A study of interaction patterns and influence in an organization. Acad. Management J. 28(2) 327–343.
Brass, D. J. 2002. Social networks in organizations: Antecedents and
consequences. Unpublished manuscript, University of Kentucky,
Lexington.
Brass, D. J., K. D. Butterfield, B. C. Skaggs. 1998. Relationships and
unethical behavior: A social network perspective. Acad. Management Rev. 23(1) 14–31.
Burt, R. S. 1976. Positions in networks. Soc. Forces 55(1) 93–122.
Burt, R. S. 1992. Structural Holes: The Social Structure of Competition. Harvard University Press, Cambridge, MA.
Burt, R. S. 2002. Bridge decay. Soc. Networks 24(4) 333–363.
Burt, R. S. 2004. Structural holes and good ideas. Amer. J. Sociol.
110(2) 349–399.
Cook, K. S., R. M. Emerson. 1978. Power, equity and commitment
in exchange networks. Amer. Sociol. Rev. 43 721–739.
Cooley, C. H. 1909. Social Organization: A Study of the Larger Mind.
Shocken Books, New York.
Cross, R. L., A. Parker, S. P. Borgatti. 2000. A birds-eye view: Using
social network analysis to improve knowledge creation and sharing. Knowledge Directions 2(1) 48–61.

Borgatti and Halgin: On Network Theory

1180
Davis, G. F. 1991. Agents without principles? The spread of the poison pill through the intercorporate network. Admin. Sci. Quart.
36(4) 583–613.

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

DiMaggio, P. L., W. W. Powell. 1983. The iron cage revisited: Institutional isomorphism and collective rationality in organizational
fields. Amer. Sociol. Rev. 48(2) 147–160.
Dubini, P., H. Aldrich. 1991. Personal and extended networks
are central to the entrepreneurial process. J. Bus. Venturing
6(5) 305–313.
Ebers, M. 1997. Explaining inter-organizational network formation.
M. Ebers, ed. The Formation of Inter-Organizational Networks.
Oxford University Press, Oxford, UK, 3–40.
Erickson, B. 1988. The relational basis of attitudes. S. D. Berkowitz,
B. Wellman, eds. Social Structures: A Network Approach. Cambridge University Press, Cambridge, UK, 99–121.

Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

Lazarsfeld, P., R. K. Merton. 1954. Friendship as a social process:
A substantive and methodological analysis. M. Berger, T. Abel,
C. H. Page, eds. Freedom and Control in Modern Society. Van
Nostrand, New York, 18–66.
Lee, J. 2010. Heterogeneity, brokerage, and innovative performance: Endogenous formation of collaborative inventor networks. Organ. Sci. 21(4) 804–822.
Lin, N. 1982. Social resources and instrumental action. P. Marsden, N.
Lin, eds. Social Structure and Network Analysis. Sage, Beverly
Hills, CA, 131–145.
Linton, R. 1936. The Study of Man. D. Appleton-Century, New York.
Lorrain, F., H. C. White. 1971. The structural equivalence of individuals in social networks. J. Math. Sociol. 1(1) 49–80.
Luce, R. D. 1950. Connectivity and generalized cliques in sociometric
group structure. Psychometrika 15(2) 169–190.

Everett, M. G., S. P. Borgatti. 1994. Regular equivalence: General
theory. J. Math. Sociol. 19(1) 29–52.

Luce, R., A. Perry. 1949. A method of matrix analysis of group structure. Psychometrika 14(2) 95–116.

Freeman, L. C. 1979. Centrality in social networks: Conceptual clarification. Soc. Networks 1(3) 215–239.
Granovetter, M. S. 1973. The strength of weak ties. Amer. J. Sociol.
78(6) 1360–1380.

Madahaven, R. 2010. Intra-organizational and inter-organizational
network effects: Using meta-analysis to illuminate distinctive
mechanisms. Presentation, Intra-Organizational Networks Conference, April 10, University of Kentucky, Lexington.

Granovetter, M. S. 1974. Getting a Job: A Study of Contacts and
Careers. Harvard University, Cambridge, MA.

Markovsky, B., D. Willer, T. Patton. 1988. Power relations in
exchange networks. Amer. Sociol. Rev. 53(2) 220–236.

Halgin, D. 2009. The effects of social identity, network connectivity, and prior performance on career progression and resilience:
A study of NCAA basketball coaches. UMI Proquest Doctoral
Dissertation 3344938, Boston College, Chestnut Hill, MA.

Marsden, P. V. 1983. Restricted access in networks and models of
power. Amer. J. Sociol. 88(4) 686–717.

Homans, G. C. 1950. The Human Group. Harcourt-Brace, New York.

McPherson, J. M., L. Smith-Lovin. 1987. Homophily in voluntary
organizations: Status distance and the composition of face-toface groups. Amer. Sociol. Rev. 52(3) 370–379.

Ibarra, H. 1992. Homophily and differential returns: Sex differences
in network structure and access in an advertising firm. Admin.
Sci. Quart. 37(3) 422–444.

McPherson, J. M., L. Smith-Lovin, J. M. Cook. 2001. Birds of a
feather: Homophily in social networks. Annual Rev. Soc. 27
415–444.

Jones, C., W. S. Hesterly, S. P. Borgatti. 1997. A general theory of network governance: Exchange conditions and social mechanisms.
Acad. Management J. 22(4) 911–945.
Kahler, M. 2009. Collective action and clandestine networks: The
case of Al Qaeda. M. Kahler, ed. Networked Politics: Agency,
Power, and Governance. Cornell University Press, Ithaca, NY,
103–126.
Kilduff, M. 2010. Serendipity vs. strategy: A tale of two theories. Presentation, Intra-Organizational Networks Conference, April 10,
University of Kentucky, Lexington.
Kilduff, M., D. Krackhardt. 1994. Bringing the individual back in:
A structural analysis of the internal market for reputation in
organizations. Acad. Management J. 37(1) 87–108.
Krackhardt, D., L. W. Porter. 1985. When friends leave: A structural
analysis of relationship between turnover and stayers’ attitudes.
Admin. Sci. Quart. 30(2) 242–261.
Krackhardt, D., L. W. Porter. 1986. The snowball effect: Turnover
embedded in communication networks. Appl. Psych. 71(1)
50–55.
Lakatos, I. 1980. The Methodology of Scientific Research Programmes: Philosophical Papers, Vol. 1. Cambridge University
Press, Cambridge, UK.
Laumann, E. O., P. V. Marsden, D. Prensky. 1983. The boundary specification problem in network analysis. R. S. Burt, M. J. Minor,
eds. Applied Network Analysis: A Methodological Introduction.
Sage, Beverly Hills, CA, 18–34.
Lave, C. A., J. G. March. 1975. An Introduction to Models in the
Social Sciences. Harper & Row, New York.

Merton, R. K. 1959. Social Theory and Social Structure. Free Press,
New York.
Nadel, S. F. 1957. Theory of Social Structure. Free Press, New York.
Obstfeld, D. 2005. Social networks, the tertius iungens orientation, and involvement in innovation. Admin. Sci. Quart. 50(1)
100–130.
Perry-Smith, J. E. 2006. Social yet creative: The role of social relationships in facilitating individual creativity. Acad. Management
J. 49(1) 85–101.
Podolny, J. M. 2001. Networks as the pipes and prisms of the market.
Amer. J. Sociol. 107(1) 22–60.
Podolny, J. M., J. N. Baron. 1997. Resources and relationship: Social
networks and mobility in the workplace. Amer. J. Sociol. 62(5)
673–693.
Powell, W. W. 1990. Neither market nor hierarchy: Network forms
of organization. B. Staw, L. L. Cummings, eds. Research in
Organizational Behavior, Vol. 12. JAI Press, Greenwich, CT,
295–336.
Rapoport, A. 1963. Mathematical models of social interaction. R. D.
Luce, R. R. Bush, E. Galanter, eds. Handbook of Mathematical
Psychology, Vol. 2. John Wiley & Sons, New York, 493–579.
Salancik, G. R. 1995. WANTED: A good network theory of organization. Admin. Sci. Quart. 40(2) 345–349.
Seidman, S. B., B. L. Foster. 1978. A graph theoretic generalization
of the clique concept. J. Math. Sociol. 6(1) 139–154.
Sparrowe, R. T., R. C. Liden, S. J. Wayne, M. L. Kraimer. 2001.
Social networks and the performance of individuals and groups.
Acad. Management J. 44(2) 316–325.

Borgatti and Halgin: On Network Theory
Organization Science 22(5), pp. 1168–1181, © 2011 INFORMS

Tönnies, F. 1912. Gemeinschaft Und Gesellschaft, 2nd ed. Fues’s Verlag, Leipzig. [Orig. pub. 1887.]
Wasserman, S., K. Faust. 1994. Social Network Analysis: Methods
and Applications. Cambridge University Press, Cambridge, UK.

INFORMS holds copyright to this article and distributed this copy as a courtesy to the author(s).
Additional information, including rights and permission policies, is available at http://journals.informs.org/.

White, D. R., K. P. Reitz. 1983. Graph and semigroup homomorphisms on networks of relations. Soc. Networks 5(2) 193–234.

Stephen P. Borgatti holds the Paul Chellgren Chair
of Management at the LINKS Center for Social Network

1181
Analysis, Gatton College of Business and Economics, University of Kentucky. He received his Ph.D. in 1989 from the
University of California, Irvine. His research interests include
social network theory and methodology, knowledge management, and social cognition.
Daniel S. Halgin is an assistant professor of management at
the LINKS Center for Social Network Analysis. He received
his Ph.D. in management from Boston College. His research
interests include social network theory, identity dynamics, and
career trajectories.

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close