Weak Power and Simmelian Ties: Comparing

Analyses of Power and Centrality in Networks

James J. Chriss
Cleveland State University

ABSTRACT

Network Exchange Theorists, along with other network and social exchange
research programs utilizing the methodology of small group experiments, have
discovered that certain peculiar networks structures, such as the Kite, are
ambiguous with regard to the determination of the amount of power the central
position holds relative to peripheral actors in the network (the so-called “weak
power” condition). Likewise network organizational theorists such as David
Krackhardt, employing primarily survey and ethnographic methods in studying
concrete formal organizations, have noticed the Kite and have come to similar
(but not exactly the same) conclusions regarding centrality and power in this
particular structure. It is suggested that experiments that model the conditions
of Krackhardt’s Simmelian tie theory could potentially contribute to a reduction of
the remaining anomalies of the Kite.

INTRODUCTION

As Blau (1994, pp. 3-8) notes, there are two major varieties of structural analysis
in sociology. Macro-oriented structural analysis emphasizes macrostructural
influences—that is, population characteristics such as sex ratio, age or racial
composition, or income inequality—on people’s opportunities and constraints.
Micro-oriented structural analysis, on the other hand, is less apt to define social
structure in terms of differences between persons or groups, emphasizing
instead the pattern of relationships (“ties”) between positions (“nodes”)
constituting a network structure. In general, then, micro-oriented network
analysts are apt to take the larger, more formal structure and characteristics of
society for granted while seeking to explain the type of interactional relations
arising among individuals, groups, or organizations within that structure (Smelser
1988, p. 120).1
This paper analyzes a particular problematic that has arisen within
network analysis. This problematic, or theoretical object of analysis, concerns a
particular network structure known as the “Kite” (which will be explained more
fully shortly). What is interesting about the Kite is that two distinct traditions
1
There are also of course efforts to bridge the gap between macro and micro theorizing
in structural analysis, such as Lawler, Ridgeway, and Markovsky (1993), Stryker and
Burke (2000), and White (1992) to name a few.

1
within network analysis—one from the Network Exchange Theory (NET)
perspective headed by David Willer, Barry Markovsky, and associates, the other
by a group of network organizational scholars headed by David Krackhardt—have
hit upon a nearly simultaneous discovery of certain peculiar properties of this
structure. Although their characterization of the structure is quite close to one
another, the different conclusions the two groups have tentatively reached about
what’s ultimately “going on” can be traced to differences in metatheoretical
commitments, as well as differing methodological strategies. Specifically, the
Willer/Markovsky group has developed NET from the classical theory of Marx and
Weber,2 while Krackhardt and company have drawn upon another classical
tradition, that of Simmel. Up to this writing there is little evidence that the two
theory groups either know about or have been influenced by each other’s work.
By bringing together the theoretical and methodological insights of these two
groups, solutions to some of the problems associated with the Kite structure
come to light.
As will be discussed more fully below, the Kite is a five-node (or -position)
structure that has been particularly vexing for understanding the relationship
between centrality and power in social networks. Why and in what ways are
issues of centrality and power relevant to network analysts, organizational
scholars, and social scientists more generally? Krackhardt and Brass (1994, p.
210) explain that

Theoretically, actors in central network positions have greater access to, and
potential control over, relevant resources, such as information in the case of a
communication network. Actors who are able to control relevant resources, and
thereby increase others’ dependence on them, acquire power. In addition to
increasing others’ dependence on them, actors must also decrease their
dependence on others. They must have access to relevant resources that is not
controlled or mediated by others.

In the next section I analyze the NET researchers’ discovery of “weak power” as
it relates to the Kite specifically. After a discussion of processes of elaboration
and competition relating to NET specifically, I then turn to Krackhardt’s similar
discovery of the “Simmelian tie” as an important characteristic of the Kite. It will
be shown that these characterizations of the Kite structure—“weak power” as it
is called by the NET group and “Simmelian tie” as it is referred to by Krackhardt
—are virtually identical.

Willer and associates draw upon Marx and Weber at the metatheoretical level, that is,
as an “orienting strategy” (Berger and Zelditch 1997; Wagner 1984) for developing
formal theories of exchange, power, and group processes. In particular, the legacy of
Marx gives us a view of power as exploitation, whereas the legacy of Weber gives us a
view of power as influence or control (see Bell, Walker, and Willer 2000; Willer, Lovaglia,
and Markovsky 1997).

2
NETWORK EXCHANGE THEORY AND WEAK POWER

Network Exchange Theory is a formal theory of social structure that combines

some fundamental concepts of rational choice with the theoretical methods of
Marx and Weber.3 It is an elaboration and extension of an earlier theoretical
program known as Elementary Theory (ET) (see Willer and Anderson 1981). ET
was concerned with developing a formal analysis of exchange, conflict, and
coercive structures. NET goes beyond ET to include many more kinds of
structures (for recent summaries and reviews of NET, see Skvoretz 2000;
Walker, Thye, Simpson, Lovaglia, Willer, and Markovsky 2000). As Willer
(1999a, p. 3) explains, “ET and NET share the project of relating structure to
activity with the classical theories of Marx and Weber.” Marx and Weber are
particularly important because they both recognized that exclusion is a condition
of structural power, and according to Willer (1999a, p. 4) “exclusion is the
undoubtedly the single most important structural condition producing power.”
The significance of “exclusion” will become apparent as we move to a
discussion of the experimental and simulation research that has been conducted
to test various NET propositions over the years. This research has established
that exclusion produces power in various network configurations. Depending on
the type of network, power can be characterized as “strong,” “equal,” or “weak.”
The level of power here refers to the bargaining or exchange power of the
centrally located actor (or node) relative to the other four actors. 4 Up until
recently, network theorists had assumed that centrality in such networks was
linked to higher levels of power, status, and influence (Brass and Burkhardt
1992). That is to say, central actors—those tied to greater numbers of actors
than peripheral actors—were thought to accumulate more social benefits relative
to other actors as a result of their centrality. However, as Mizruchi and Potts
(1998, p. 354) note,

3
I characterize at least some of the assumptions of NET as “rational choice” because,
as we will see in the discussion of scope conditions, actors in exchange seek to
maximize their outcomes by accepting the best offer they receive. Markovsky (1997)
warns, however, that although a number of contemporary network exchange research
programs do indeed have explicit ties to rational choice theory (RCT) (such as
Bienenstock and Bonacich 1997; Bonacich 2000; Friedkin 1995; Marsden 1983;
Yamaguchi 1996), NET does not. Since NET relies on structural analyses to generate
predictions, NET does not explicitly invoke RCT; however, their assumptions do not
necessarily contradict RCT either (Markovsky 1997, p. 67; Willer and Skvoretz 1997).
4
A convention of network theory is to use the term “node” to refer to a position, that is,
a network location occupied by an actor (whether an individual, group, or organization).
Actors in this sense are “decision-making entities” that occupy positions (“nodes”) linked
by relations (“ties”) (Markovsky, Willer, and Patton 1988).

3
 Despite the once wide acceptance of the link between centrality and power,
recent research by experimental and mathematical sociologists has called this
link into question. The extent to which centrality leads to power is now an issue
of intense controversy.

It is on this controversy over centrality and power that both the NET and
Krackhardt groups have recently been focused. Figure 1 illustrates the three
types of power that accrue to the central node according to Markovsky and
company. In the 5-node configuration a central actor B has ties to four actors
A1, A2, A3, and A4. Figure 1a illustrates strong power because B is never excluded
in any combination of exchange relations. For example, B can decide to
exchange with only A1, thus excluding the other A’s from resource sharing.
Hence, B has more power relative to the other actors because B can exclude any
of the other actors from exchanging resources, and exchanges must always
include B.

(FIGURE 1 ABOUT HERE [figures to be added later])

Figure 1b is an illustration of equal power because no one actor has a monopoly

on the opportunity for exchange with other members of the network. This is an
equipower network because all its positions are isomorphic (Willer 1999b, pp.
110-111). Here, although B has ties with all four A’s just as in the case of strong
power, each A has ties to the other A’s as well. Hence, no actor is excluded from
exchanging with other actors in the network, and equal power results.
Figure 1c is an illustration of weak power. The strong power network of
1a can be transformed into the weak power network of 1c simply by adding ties
between A1 and A2 and between A3 and A4. By adding these two relations, the 4-
branch strong power relation (Figure 1a) is transformed into the weak power
“Kite” structure of Figure 1c. Now A1 can exchange with A2, and A3 with A4, thus
excluding B whenever they interact. Willer (1999b, p. 110) suggests that weak
power networks are intermediate to and between the structural conditions of
strong and equal power networks. That is, by adding or subtracting relations in
networks one can produce direct transformations from equal power to strong
power, from strong power to weak power, and so forth.
The peculiar thing about the Kite structure is that neither visual evidence
nor logical inference provides a clear indication of the power distribution of this
network. Clearly B’s power relative to the A’s is weakened when compared to
the strong power position B holds as illustrated in Figure 1a, but by how much?
Markovsky had earlier developed the Graph-theoretic Power Index (GPI) to
measure power and resource distributions in networks. As Markovsky et al.
(1988) note, it is not only the network structure, but also the (actor and position)
conditions under which exchanges transpire that affect such power and resource
distributions. The authors argue that the relative power for all positions in a
network can be determined by GPI assuming the following eight scope conditions
pertaining to actors and positions are met:

4
 1. All actors use identical strategies in negotiating exchanges.
2. Actors consistently excluded from exchanges raise their offers.
3. Those actors consistently included in exchanges lower their offers.
4. Actors accept the best offer they receive and choose randomly in deciding
among tied best offers.
5. Each position in a network is related to, and seeks exchange with, one or
more other positions.
6. At the start of an exchange round, equal pools of positively valued
resource units are available in every relation.
7. Two positions receive resources from their common pool if and only if
they exchange.
8. Each position exchanges with at most one other position per round
(Markovsky et al. 1988, p. 223).5

Power is assumed to derive, then, from the availability of alternative exchange

relations. For example, in Figure 1a the A’s have much less power than B
because the A’s must exchange with B and only B; they have no alternatives
(hence B is in a strong power position relative to the A’s). However, once ties or
relations between the A’s are added, as depicted in Figure 1c, the A’s now have
alternatives to exchanging only with B, and hence the A’s power increases
relative to B.
Measuring power with GPI also involves counting path lengths. Only
nonintersecting paths of each length stemming from a given position are
counted. Consider the T-shaped network of Figure 2. Here, three
nonintersecting two-path stem from C, namely C—B1—A1, C—B2—A2, and C—B3—
A3. However, only one nonintersecting two path stems from B1, either B 1—C—B3
or B1—C—B2. Imagine removing A1, B1, B2, and A2, leaving only the C—B3—A3
path. B3 is now obviously in a strong power position relative to C and A 3.
However, restoring B1 reduces B3’s power because C can now exchange with B1,
which means C will be excluded less frequently. Restoring B2 further improves
C’s power position; however, it does not further diminish the relative power
accruing to B3, because B3 can still exchange with A3 on an equal basis. This is
because B3—C—B1 and B3—C—B2 are intersecting two-paths from B3, so the
addition of B2, as was saw above, does not serve to further diminish B 3’s power.
This serves to illustrate, then, why only one nonintersecting path of a given
length is counted (Markovsky et al. 1988).

(FIGURE 2 ABOUT HERE)

Explicit tests of assumptions 2 and 3 concerning actors’ behavior when consistently

excluded from or included in exchanges were performed by Skvoretz and Zhang (1997).
Their data analysis supported these basic assumptions of NET.

5
 Additionally, odd length nonintersecting paths are weighted +1, while
even length paths are weighted –1. For example, from Figure 2, B 1 has 2 one
paths (+2), one nonintersecting two path (-1), and one nonintersecting three
path (+1), for a GPI index of +2. Positions with higher GPI scores have greater
power than those with lower scores. However, GPI failed to predict weak power
in certain network configurations (such as the Kite and the Stem), although it
was successful in determining whether a structure was a strong power network,
as well as the upper and lower bounds of the strong power position. Once the
strong power network is found, a further step involves breaking down the
network into its component “stems.”
The GPI procedure as just outlined now represents the first step of a two-
step process. In order to fully predict and measure weak power, the second step
involves the application of a “random-seek” method developed by Barry
Markovsky. As Willer (1999b, p. 117) explains, “Random seeks determine the
relative power of each position and, indeed, indicate whether the network is
weak power or equal power.” In effect, random-seek calculates the probability
(or likelihood) of a position being included in exchange relations. Although the
formula for random-seek need not concern us here, the uncertainty regarding
the power of the central position of the Kite structure relative to the peripheral
positions could now be reduced or eliminated. The values attached to each
position of a network once random-seek calculations are completed are termed
“exchange-seek” values, which represent the relative power of positions in a
network (Markovsky, Skvoretz, Willer, Lovaglia, and Erger 1993). 6
According to this procedure, Willer (1999b, p. 119) reports the exchange-
seek values of the Kite as follows: the central position B = .82, while the four A’s
= .79 each. Later experimental research studying negotiated exchange which
relaxed several of the previously stated scope conditions (for example, allowing
more than one exchange per round) supported these results (see, e.g., Lovaglia,
Skvoretz, Willer, and Markovsky 1995; Skvoretz and Willer 1991; for a critique of
this method, see Bienenstock and Bonacich 1997). What we are left to conclude,
according to the findings of NET researchers, is that the Kite is a weak power
network—indeed, just a step removed from being an equal power network—with
subtle but measurable power differences between the central position (.82
exchange-seek value) and the four peripheral positions (.79 exchange seek
value).

6
Further refinement of GPI by John Skvoretz led to the development of the Iterative
Likelihood Analysis (ILA) method for identifying the power type of a network. Rather
than counting paths and weighting them accordingly, ILA utilizes probability tree
analysis. For discussions of the ILA method, see Lovaglia, Skvoretz, Markovsky, and
Willer (1999), Simpson and Willer (1999), and Skvoretz (2000).

6
THE GROWTH OF NET: COMPETITION AND CRITICISM

Like any successful theoretical research program (Wagner 1984), NET has grown
and evolved through a series of elaborations (for example, the continual
refinement of methods for measuring and identifying types of power structures),
as well as from challenges from competing social exchange or network exchange
programs. Rather than providing an exhaustive review of the history of these
debates, I will be concerned here with highlighting several analyses that have
provided critical, and often constructive insights that have collectively had the
effect of pushing NET toward a less static, less rigidly structuralist, and more
context-sensitive view of power and centrality than earlier versions of NET and
ET.
First, several researchers have suggested that in such weak power
networks as the Kite, the attempt by the NET group to predict power solely on
the basis of structure would have to be augmented by explicit formulations of
the behavioral foundations of exchange relations in order to produce a more
fine-grained understanding of the small power differentials that exist in such
structures (Bienenstock and Bonacich 1997; Bonacich 1998; Yamaguchi 1996).
This notion of giving the cognitive element its due in network exchange theory,
for example, by incorporating concepts from game theory such as bargaining,
utility, rational choice, strategy, and even equity, would presumably help fill in
the gaps in our understanding of how and in what ways power is employed in
weak structural arrangements.
This particular criticism, however, is not wholly accurate in depicting NET
researchers as unwilling or unable to integrate negotiation into their small group
experiments. Lovaglia et al. (1995) set up experiments in which profit
expectations of the subjects were taken into account. Two factors that have
been found to affect profit expectations, and hence the actual forms of exchange
and negotiation taking place between actors in various network configurations,
are degree centrality, namely the number of actors to whom one is directly
connected in the network, and the likelihood of a person completing exchanges
with another (Lovaglia et al. 1995). Actors who perceive themselves to be
central as a result of many direct ties to others are more likely to expect that
they will do well in exchanges as a result of their central position, and this
expectation of profit affects negotiation strategy; for example, they are more
likely to resist an offer that appears to be lower than what they could reasonably
expect given their location and history of negotiations. 7 This issue of NET’s
inexorable move toward a more dynamic model of exchange that takes into
account the history of ongoing negotiations, rather than relying on initial offers
as is the case for experiments guided by Markovsky et al.’s (1988) original eight
scope conditions, will be returned to shortly.
7
Indeed, recently NET researchers have been conjecturing that degree centrality may
be more important in determining the nature of weak power structures than originally
thought (see Skvoretz 2000, p. 10).

7
 Second, from above, equity is an especially important and relatively recent
contribution to the discussion of power and centrality in networks. According to
Bonacich and Friedkin (1998), NET’s actor and position scope conditions—for
example, “all actors use identical strategies in negotiating exchanges”—generate
a very simple type of homogeneous exchange structure that may be inadequate
to the task of explaining the subtle power differentials evident in the Kite and
other weak power structures. In small group experiments, Bonacich and Friedkin
(1998) relaxed the restrictive scope conditions of NET to allow subjects a five-
minute period of bargaining after initial offers to all those actors to whom they
were directly tied. In NET experiments subjects are typically informed about the
form of the network and their positions within it, as well as being rotated
through all positions—which in the case of the Kite is five—over the course of
experimental trials (see, e.g., Szmatka and Willer 1995). They also know what
other players’ earnings are, since each round begins with a pre-established pool
of points or resources. Bonacich and Friedkin removed all these conditions by
assigning actors to fixed positions in networks of which they were unaware. The
authors in effect generated heterogeneously valued exchange networks to
augment the homogeneous networks studied by NET and other network
exchange researchers.
Even while restricting subjects’ information about the types of networks
they were part of and their locations in these networks, however, these
experimental conditions appeared to generate equity concerns among the
subjects. As Bonacich and Friedkin (1998, p. 168) note, in many cases it
appeared that subjects resisted the development of trading patterns in which
many of their available trading partners would be excluded from exchange. In
other words, subjects who take into account the values of fairness and equity
may confound the expectation that a particular position is more powerful by
foregoing, or refusing to take full advantage of, a presumed beneficial exchange
position. In light of this, Bonacich and Friedkin (1998, p. 170) suggest that
“Equity effects have been viewed as confounding the study of structural effects,
but they may be worth including as part of the specification of a general social
exchange model.”
Third, and extending somewhat the points of Bonacich and Friedkin
(1998) and Yamaguchi (1996), Mizruchi and Potts (1998) note that NET
experiments (as well as others, e.g., the power-dependence approach of Cook et
al. 1983) are set up such a way that a fixed number of points are divided
between exchange partners. These exchange relations can be characterized as
“zero-sum” insofar as one actor’s gain is another actor’s loss. When discreet,
tangible resources such as “points” in an experiment are exchanged, the zero-
sum nature of this activity seems obvious and unproblematic. However, power
can also be conceptualized as a problem involving the flow of information, that
is, the issue of who is in a position to control or decide what sorts of information
goes where, and in what form (Monge and Contractor 2001). Here, rather than
the zero-sum exchange network, we have a communication network where

8
whatever resources are considered to be important—information primarily—are
less readily quantifiable for purposes of manipulation in small group experimental
research or, at the very least, less apt to be of a zero-sum nature. 8 This
suggests that communication networks, the sorts of networks often studied by
organizational theorists such as Krackhardt, may be different in some respects in
comparison to the typical exchange networks studied by NET and other social
exchange theorists.
In any event, theoretical and methodological elaboration taking place
within the network of NET researchers, as well as competition from neighboring
theoretical research programs, have combined to move NET toward the
exploration and development of dynamic, rather than static, network models. At
the prodding of Friedkin (1995), Bonacich and Friedkin (1998), Bonacich (2000),
Macy and Flache (1995), and several of the other researchers mentioned above,
Willer and Willer (2000, p. 251) have recently admitted that

Networks, which are dynamic, have long been neglected by research in network
exchange: laboratory research has focused upon static structures. Yet,
exchange networks outside the laboratory or not static.

Willer and Willer (2000), following the lead of Leik (1992), attempt to show that
NET is not limited to static networks, and can accommodate the study of
dynamic networks. As NET has progressed from relatively simple models of
network structure characterized by “unrealistic” assumptions (Kanazawa 1998)
about social structure and exchange, to dynamic network analysis that
presumably gets closer to depicting “real” networks, it has moved closer to the
form of investigation that network organizational theorists such as David
Krackhardt have been developing. We turn now to a discussion of Krackhardt’s
program.

DAVID KRACKHARDT AND THE SIMMELIAN TIE

Now that we have summarized the way in which the NET group grappled with,
and appears to have explained, the subtle power differences of the Kite network,
we turn to a similar line of research and theorizing that has taken place roughly
during the same time—late 1980s through the present—by a different
researcher. Whereas Markovsky and associates have tended to employ the
methodology of computer simulation and experimentation to study networks as
an ideal type abstracted from any particular context or setting, David Krackhardt
has been oriented to studying networks primarily within the context of formal
8

Here Parsons’ (1978) notion of the cybernetic hierarchy of control with circulating and
generalized media of interchange—such as money, influence, political power, and value-
commitments—that go through cycles of inflation and deflation, comes to mind.

9
organizations. As such, Krackhardt often employs the methodology of the case
study and ethnographic and survey research, along with the help of computer
software programs to carry out such quantitative operations as clique analysis,
block modeling, multidimensional scaling for measuring the graph-theoretic path
distances between actors in a network, vector analysis, and matrix algebra (see
Krackhardt 1990; 1992, 1998; 1999).9 These differences in methodology and
substantive orientation between Krackhardt and the NET group are sufficient, I
would contend, to explain the nearly complete lack of mutual acknowledgment of
(that is, citation to) each other’s work.
Like many scholars working at the crossroads of organizational studies
and network analysis, Krackhardt has had an ongoing interest in issues of power
and influence within the organization. By the late 1970s Linton Freeman (1979;
Freeman, Radar, and Mulholland 1980) had pushed the issue of power and
centrality within social networks to the top of the research agenda for a growing
number of organizational theorists. Freeman’s studies examined structural
centrality and its effect on leadership selection, group problem solving, and
individual satisfaction. Employing small experimental groups, Linton and
associates found evidence that different conceptualizations of structural
centrality helped explain leadership nomination and role satisfaction, but not
group problem solving, in these groups. Specifically, “betweenness” centrality
was important for understanding leadership choice while role satisfaction was
dependent on both betweenness and “degree” centrality (Johnson and Parks
1998, pp. 223-224). 10
Although these were compelling findings, organizational scholars who
followed developments in network analysis were concerned that Linton’s
findings, being based as they were on experimental groups, might not be
generalizable to “real” organizations where such things as size, structure, and
goals could presumably affect the observed relationships (see Johnson and Parks
1998; Monge and Contractor 2001; Nohria 1992; Roberts and O’Reilly 1979).
Nevertheless, by the early 1980s organizational scholars such as David
Krackhardt, Daniel Brass, and Marlene Burkhardt began devoting time to
examining the implications of the network perspective for the understanding of
centrality and power within the context of naturally occurring organizations (for a
summary of these and other studies, see Johnson and Parks 1998; Monge and
Contractor 2001).

9
This is not to say that Krackhardt has never pursued the kind of simulational and
experimental research that is the bellwether of the NET group. For an example of such
research, see Krackhardt and Stern (1988).
10

“Centrality” is the extent to which an actor is central to a network. “Degree” is the

number of direct links with other actors. And “betweenness” is the extent to which an
actor mediates, or falls between any other two actors on the shortest path between
those actors (Brass 1995).

10
 These organizational theorists were especially concerned to discover
which aspects of centrality, as conceptualized and explained by network
theorists, could be confirmed via empirical testing in concrete organizations. Of
special interest were betweenness centrality, because of how these centrally-
located actors are able to control resources and information across the network;
degree centrality, because of how these positions are able to provide or withhold
access to valued information and resources; and closeness centrality, namely,
the extent to which an actor is close to, or can easily reach all the other actors in
the network (Brass 1995; Wasserman and Faust 1997, pp. 169-219). Hence,
betweenness centrality is concerned with control of valued resources, while both
degree and closeness centrality are concerned with access to such resources.

(FIGURE 3 ABOUT HERE)

These issues and trends within network and organizational studies serve
as a backdrop for analyzing the particular approach David Krackhardt has taken
in his work on power and centrality. These interests go back at least as far as
1987, the year in which he published an influential paper on cognitive social
structures (Krackhardt 1987).11 Following the formulations of Adams (1965) and
Burt (1982), Krackhardt argued that with regard to issues of power and influence
in organizations, it is important to understand how persons define the situation,
that is, how and why they come to perceive certain actors (“players” in Burt’s
terminology) as powerful or influential. This suggests that research should focus
on assessing individuals’ perceptions of the organizations within which they are
embedded, and then calculate “structural equivalence,” this being the degree to
which the various perceptions of individuals in the organization converge upon
one or several sociometric “stars” who are perceived as central, influential, or
powerful figures in the organization (Krackhardt 1987; Krackhardt and Brass
1994, p. 224). These “cognitive social structures” or “cognitive maps,”
Krackhardt argues, are often better at depicting the structural features of
organizations than the traditional network theoretical approach.
11
Michael Lovaglia and Barry Markovsky named weak power in 1990, but the seeds of
the concept may go as far back as 1987 to some informal communications Toshio
Yamagishi had with Markovsky and David Willer concerning a network structure (in this
case, the Stem) that GPI did not do a good job of covering (Markovsky 1997; Willer
1999b). Krackhardt’s explicit formulation of the Simmelian tie would come somewhat
later, with the seeds of the idea formed in 1987 and formulated for the first time more
explicitly in 1992.
So as to avoid terminological confusion, it should be pointed out that as far back
as 1990 Krackhardt had been referring to a particular network structure as a “Kite,”
although this Kite is NOT the same Kite which has been referred to throughout this
paper. Krackhardt’s “Kite” is aptly named, however, for it looks more like a kite than
what we are here calling the Kite (which some researchers, such as Bienenstock and
Bonacich 1997, also refer to as the “Hourglass”). Krackhardt’s (1990; Brass and
Burkhardt 1992) “Kite” is depicted in Figure 3.

11
 A few years later Krackhardt (1992) published a paper that further
explored the contested power-centrality link in organizations. He found
Granovetter’s (1973) theory of the strength of weak ties especially useful for his
purposes. Granovetter’s argument was that strong ties (such as informal
friendship ties that are characterized by affectivity and closeness) are just as
critical to the livelihood of formal organizations as the formal, publicly-sanctioned
weak ties (the classic pyramid shape of command and control where
officeholders view other officeholders in affectively-neutral, instrumental terms)
which had historically received the lion’s share of organizational theorists’
attention. However, Granovetter’s distinction between weak and strong ties was
somewhat limited as a guide for directing empirical organizational research.
Krackhardt sought to overcome Granovetter’s lack of conceptual clarity by
reworking the “strong” side of the weak tie/strong tie analytic. He did this by
developing the notion of a philos relationship to take the place of the strong tie.
For A and B to be philos, A and B must interact with each other on a regular
basis, A must feel affection for B, and A and B must have a history of
interactions extending over a significant period of time (Krackhardt 1992, pp.
218-219). Importantly, philos relationships are characterized by a strong and
abiding sense of trust between its members.
Staying consistent with his earlier theme of the importance of cognitive
structures, Krackhardt surveyed members of a small business firm, asking them
about their perceptions about who was related to whom in the firm. The philos,
or friendship network was operationalized by asking all 36 members of the firm
“Who would x consider to be a personal friend?” This was followed by a list of
the other 35 names from which the respondent could check as many as he or
she deemed appropriate. The other network studied by Krackhardt, analogous
to the weak tie of Granovetter, was the advice network, which in similar fashion
asked members “Who would x go to for help or advice at work?”
Krackhardt (1992, p. 234) discovered that the key players in the advice
network were not the key actors in the friendship network. Members who were
perceived to be central in the philos network were oftentimes not important to
the more formal advice network. Indeed, when actor A has philos ties to two
other actors B and C who may be at odds, the cross-pressures exerted from the
tension or conflict between B and C often leads to A dropping out of an advice
role altogether, thus losing influence (power) in such cases.
This discovery of the special constraints a central actor perceives or
experiences whenever he or she acts as a bridge between two groups to which
the central actor is closely (philos) tied was similar to the sorts of dilemmas that
the NET researchers had grappled with regarding centrality and power in weak
networks. Although Krackhardt had made a discovery about the types of ties
that contradict the power-centrality link, he still needed to examine the types of
networks within which such relations are embedded. This was accomplished
with the development of the notion of the Simmelian tie.

12
 Krackhardt’s discovery of the cross-pressures exerted on a central actor
from two conflicting philos ties had close affinities to Simmel’s (1950, pp. 135-
169) theory of the fundamental differences in characteristics between dyads and
triads. In comparison to dyads, triads are characterized by a reduction of
individuality among its members, a reduction in individual power of all of its
members (in that power is now spread between three persons instead of two),
and a heightened ability to moderate conflict, thus making the group more stable
(Krackhardt 1998, 1999; Simmel 1950).
Besides explicit attention being paid to Simmel at this time, Krackhardt
was also stimulated by Burt’s (1992) theory of structural holes. Burt suggests
that actors who act as bridges, who are connected to others who are not
themselves connected to each other, tend to be more powerful and command
more in exchanges with these others. In general, bridging ties are less
constraining on actors than non-bridging ties (Burt 1992). In response,
Krackhardt (1999, p. 184) argues that central, bridging positions may not be as
advantageous as Burt and other researchers have suggested, and that occupying
such a bridging role can be more constraining, especially if the tie is a
particularly strong kind, namely, a Simmelian tie.
Krackhardt (1999, p. 186) suggests that two people are “Simmelian tied”
to one another “if they are reciprocally and strongly tied to each other and if
they are each reciprocally and strongly tied to at least one third party in
common.” There is a strong isomorphism between the Simmelian tie and the
notion of a “clique,” where cliques are defined as a maximal set of three or more
nodes all of whom are directly and reciprocally connected to each other
(Krackhardt 1999, p. 186). There is an especially high level of constraint on the
central bridging position when it is located between two triadic cliques. As
Simmel notes, the nature of the group does not change appreciably beyond the
size of the triad. The key distinction is the important changes that take place
when group size increases from two (the dyad) to three (the triad) members.
Hence, Krackhardt uses triadic cliques to illustrate the concept of the Simmelian
tie.

(FIGURE 4 ABOUT HERE)

Figure 4 summarizes the idea of the Simmelian tie in comparison to Burt’s (1992)
structural hole theory. The key question being addressed here is: how much is
the central position (ego) constrained under varying tie conditions? (One should
immediately compare Figure 4 with Figure 1; it is the same 5-position network
that the NET researchers have investigated.) With regard to Figure 4a, both Burt
and Krackhardt would agree that ego is in the least constrained, or strongest
position. This also agrees with the NET researchers’ findings that such a
structure constitutes a strong power network.
Burt and Krackhardt diverge somewhat on the next two figures, however.
Regarding Figure 4b, Burt would suggest ego is maximally constrained because

13
since there are no structural holes in the network—because everybody is tied to
everyone else—there is no opportunity for ego to benefit relative to others
members of the network. On the other hand, Krackhardt suggests that although
ego is more constrained here than in Figure 4a, it does not represent the
maximal constraint condition insofar as ego must meet only one clique’s set of
norms. The NET position, which describes this structure as an equal power
network, would agree with Burt that this represents the maximal constraint
condition on ego. This is because, from the NET perspective, weak power falls
between the extremes of strong power (Figures 1a and 4a) and equal power
(Figures 1b and 4b).
Lastly, with regard to Figure 4c, Burt argues that ego is more constrained
than in Figure 4a but less constrained than in Figure 4b. In other words, ego is
somewhat constrained but also empowered in that ego is acting as a bridge
between two triadic cliques. On the other hand, Krackhardt would suggest that
this figure represents the maximal constraint condition, in that ego must now
meet the normative expectations of two different cliques. Notice that this figure
is the Kite structure from our earlier discussion of Network Exchange Theory.
The NET view of the Kite as a weak power structure is in closer agreement, then,
with Burt than with Krackhardt on the extent to which ego is constrained here.
We shall now examine the implications of the near convergence between
Krackhardt and the NET group with regard to this particular 5-position network.

POWER AND CENTRALITY REVISITED

We now have seen evidence of a convergence of findings between two

theoretical research programs utilizing different metatheoretical frameworks and
methodologies. Both the NET group and Krackhardt are seeking to explain how
power is related to structurally central positions in certain anomalous network
configurations. The Kite has arisen as an especially interesting network structure
in this regard because it represents the unique case of two triadic cliques sharing
one common member (the central position or “ego”), even though the NET
researchers have never explicitly studied the Kite in these terms. However, all
such Kite structures can be generalized to the problem of Simmelian ties and the
special constraints potentially placed on the central actor.
The use of “potential” in the above sentence deserves clarification. As
noted in Figure 4c, although Krackhardt (1999) appears emphatic in proclaiming
that the Kite represents the maximal constraint condition with regard to ego,
there are certain further situations or conditions that could relax this constraint.
In fact, it should be noted that the contingencies regarding just how much
constraint is indeed placed upon ego that have been discovered from the
“empirical” or “inductivist” side, represented by Krackhardt’s research on actual
organizations, reflect or mirror the same sorts of ambiguities and contingencies

14
that the “experimental” or “deductivist” side, represented by Willer and the NET
researchers, have grappled with.
We may move toward an understanding of what lies behind this
contingency by considering Bonacich’s (2000) recent work on patterns of
coalition formation in exchange networks. As Simmel (1950) noted long ago,
triads are stable to the extent that they are transitive, meaning that feelings and
relations among the members of the group are balanced or consistent (Caplow
1968). However, if for example one member of the triad begins accumulating
resources far above the level of the other two members, we now have an
intransitive triad in which the ties between members are qualitatively unequal, as
in Granovetter’s (1973) “forbidden triad.” Intransitive triads are unstable and
therefore susceptible to coalition formation, a situation in which the two weaker
or oppressed members of the triad combine their resources to oppose the
stronger member. Indeed, coalition formation is one method by which weak or
exploited members of coalitions seek to reduce power differences.
To illustrate aspects of coalition formation, Bonacich (2000) cites research
conducted by Bonacich and Applebaum (2000) on the basic structure of the
garment industry in Los Angeles. Bonacich and Applebaum (2000) discovered
that manufacturers try to ensure that the various contractors they work with are
not known to one another. By doing this, manufactures hope to strengthen their
bargaining position with contractors because they can always claim that some
unknown contractor has tendered a better offer. This prevents concerted action,
that is, coalition formation, on the part of contractors to nullify or reduce the
advantageous position manufactures enjoy. The authors go on to report that
manufactures ensure their edge by proclaiming their contractor lists are “trade
secrets,” a position backed and enforced by Federal trade law.
Obviously, then, the ability of positions to communicate and possibly form
coalitions is an important part of the explanation of power and centrality in
networks. Following the lead of Simmel (1950), Krackhardt (1999) also notes
the importance of communication in the situation involving a central actor acting
as a bridge between two triadic cliques. An important caveat regarding the
assertion of the maximal constraint placed on ego in the Kite configuration is
this: the maximal constraint condition holds only to the extent that the central
actor’s behavior is public, that is, is known to the other members of each clique .
In situations where the central actor’s conduct is private or secret—such as in
the case of the manufacturer guarding information about which contractors it is
working with—the cross-pressures emanating from the variable normative
expectations of the two cliques would be absent or considerably reduced. This in
analogous to what Simmel described as the tertius gaudens situation, in which a
bridging actor takes advantage of what the different actors know or don’t know
(Krackhardt 1999, p. 207).12
12
Another potentially divisive element or strategy in triads is divide et impera, in which
the third party actively seeks to keep the other two parties separate in order to maintain
the third party’s dominance or supremacy (Krackhardt 1999, p. 185). This is why Burt

15
CONCLUSION

What I am suggesting is that the near convergence between NET and Krackhardt
on centrality and power in networks, and the even closer convergence between
NET and Burt, can be seen as validating the general theoretical insights
concerning the relationship between power and centrality, including the
ambiguities that result when certain anomalous structures, such as the Kite, are
considered. That is to say, these ambiguities are being discovered from all sides,
whether in the form of slight or negligible advantages in power the central Kite
actor holds over other actors, or in cases where the central actor is actually
found to be disadvantaged relative to the other actors.
Let us summarize the major points of our comparison of the two
orientations, namely the experimental-deductive approach of NET and the
empirical-inductive approach of the two organizational theorists, Krackhardt and
Burt. All three agree that Figure 4a represents the minimal constraint imposed
on the central actor ego. NET and Burt agree that Figure 4b represents the
maximal constraint on ego, while Krackhardt suggests that Figure 4c—the Kite—
represents the maximal constraint condition. But what is crucial to note as well
is that NET’s GPI measures of the power of the central position (.82) relative to
the peripheral positions (.79) in the Kite is negligible, almost approaching the
equipower situation of Figure 1b. Recall as well NET’s actor and condition
assumptions, for example, the condition of equal resource pools, one exchange
maximum per round, and so forth.
NET’s finding of a small advantage for the central position in the Kite
network reflects only initial offers, insofar as subjects are limited to one
exchange maximum per round. In examining dynamic exchange networks where
subjects are allowed to complete as many trials as they can in an allotted time,
Friedkin (1995) reports that over time, over a series of exchanges, the predicted
payoff for the central actor fluctuates (somewhat unpredictably), and at
equilibrium the central actor is actually at a slight disadvantage in exchanges
with the peripheral actors (see also Bienenstock 1992). Later experiments on
dynamic networks employing similar scope conditions, as described above in the
earlier discussion of Bonacich and Friedkin (1998), found, however, that the
central actor in the Kite was able to maintain a slight power edge over peripheral
actors. This was also the case in Lovaglia et al.’s (1995) experiments, although
these experiments established somewhat different scope conditions.
What seems to be occurring here is that various experimental conditions
are producing variable findings regarding the amount of power of the central
position relative to peripheral positions in the Kite. In order to solve some of
these persistent ambiguities, experimental researchers, whether the NET group

(1992) suggests that Figure 4b represents the most constraint on ego: there are no
opportunities for tertius or divide plays.

16
or other social exchange programs, ought to consider modeling the conditions
under which Krackhardt’s central Kite actor is maximally constrained, that is,
under Simmelian tie conditions. As NET researchers continue to press forward
with experiments on dynamic networks, as suggested by Willer and Willer
(2000), perhaps a final solution to the puzzle of the Kite will emerge.

17
 REFERENCES

Adams, J.S. 1965. “Inequality in Social Change.” Pp. 267-300 in Advances in

Experimental Social Psychology, edited by L. Berkowitz. New York:
Academic Press.
Bell, Richard, Henry A. Walker, and David Willer. 2000. “Power, Influence, and
Legitimacy in Organizations: Implications of Three Theoretical Research
Programs.” Pp. 131-177 in Research in the Sociology of Organizations, vol.
17, edited by S. B. Bacharach and E.J. Lawler. Stamford, CT: JAI Press.
Berger, Joseph and Morris Zelditch, Jr. 1997. “Theoretical Research Programs: A
Reformulation.” Pp. 29-46 in Status, Network, and Structure: Theory
Development in Group Processes, edited by J. Szmatka, J. Skvoretz, and J.
Berger. Stanford, CA: Stanford University Press.
Bienenstock, Elisa Jayne. 1992. “Game Theory Models for Exchange Networks:
An Experimental Study.” Doctoral dissertation, Department of Sociology,
University of California, Los Angeles.
Bienenstock, Elisa Jayne and Phillip Bonacich. 1997. “Network Exchange as a
Cooperative Game.” Rationality and Society 9 (1):37-65.
Blau, Peter M. 1994. Structural Contexts of Opportunity. Chicago: University of
Chicago Press.
Bonacich, Edna and Richard Applebaum. 2000. Behind the Label: Inequality in
the Los Angeles Apparel Industry. Berkeley, CA: University of California
Press.
Bonacich, Phillip. 1998. “A Behavioral Foundation for a Structural Theory of
Power in Exchange Networks.” Social Psychology Quarterly 61 (3):185-
198.
Bonacich, Phillip. 2000. “Patterns of Coalitions in Exchange Networks: An
Experimental Study.” Rational and Society 12 (3):353-373.
Bonacich, Phillip and Noah E. Friedkin. 1998. “Unequally Valued Exchange
Relations.” Social Psychology Quarterly 61 (2):160-171.
Brass, Daniel J. 1995. “A Social Network Perspective on Human Resources
Management.” Pp. 39-79 in Research in Personnel and Human Resources
Management, vol. 13, edited by G.R. Ferris.
Brass, Daniel J. and Marlene E. Burkhardt. 1992. “Centrality and Power in
Organizations.” Pp. 191-215 in Networks and Organizations: Structure,
Form, and Action, edited by N. Nohria and R.G. Eccles. Boston, MA:
Harvard Business School Press.
Burt, Ronald S. 1982. Toward a Structural Theory of Action. New York: Academic
Press.
Burt, Ronald S. 1992. Structural Holes: The Social Structure of Competition.
Cambridge, MA: Harvard University Press.

18
Caplow, Theodore. 1968. Two Against One: Coalitions in Triads. Englewood
Cliffs, NJ: Prentice-Hall.
Cook, K.S., R.M. Emerson, M.R. Gillmore, and T. Yamagishi. 1983. “The
Distribution of Power in Exchange Networks: Theory and Experimental
Results.” American Journal of Sociology 87:275-305.
Freeman, Linton C. 1979. “Centrality in Social Networks: Conceptual
Clarification.” Social Networks 1:215-239.
Freeman, Linton C., D. Radar and R. Mulholland. 1980. “Centrality in Social
Networks II: Experimental Results.” Social Networks 2:119-142.
Friedkin, Noah E. 1995. “The Incidence of Exchange Networks.” Social
Psychology Quarterly 58 (3):213-222.
Granovetter, Mark S. 1973. “The Strength of Weak Ties.” American Journal of
Sociology 78 (6):1360-1380.
Johnson, Jeffrey C. and Dawn L. Parks. 1998. “Communication Roles, Perceived
Effectiveness, and Satisfaction in an Environmental Management
Program.” Computational and Mathematical Organization Theory 4
(3):223-239.
Kanazawa, Satoshi. 1998. “In Defense of Unrealistic Assumptions.” Sociological
Theory 16 (2):193-204.
Krackhardt, David. 1987. “Cognitive Social Structures.” Social Networks 9:109-
134.
Krackhardt, David. 1990. “Assessing the Political Landscape: Structure,
Cognition, and Power in Organizations.” Administrative Science Quarterly
35:342-369.
Krackhardt, David. 1992. “The Strength of Strong Ties: The Importance of Philos
in Organizations.” Pp. 216-239 in Networks and Organizations: Structure,
Form, and Action, edited by N. Nohria and R.G. Eccles. Boston, MA:
Harvard Business School Press.
Krackhardt, David. 1998. “Simmelian Ties: Super Strong and Sticky.” Pp. 21-38 in
Power and Influence in Organizations, edited by R.M. Kramer and M.A.
Neale. Thousand Oaks, CA: Sage.
Krackhardt, David. 1999. “The Ties that Torture: Simmelian Tie Analysis in
Organizations.” Pp. 183-210 in Research in the Sociology of Organizations,
vol. 16, edited by S.B. Andrews and D. Knoke. Stamford, CT: JAI Press.
Krackhardt, David and Daniel J. Brass. 1994. “Intraorganizational Networks: The
Micro Side.” Pp. 207-229 in Advances in Social Network Analysis:
Research in the Social and Behavioral Sciences, edited by S. Wasserman
and J. Galaskiewicz. Thousand Oaks, CA: Sage.

19
Krackhardt, David and Robert N. Stern. 1988. “Informal Networks and
Organizational Crises: An Experimental Simulation.” Social Psychology
Quarterly 51 (2):123-140.
Lawler, Edward J., Cecilia Ridgeway, and Barry Markovsky. 1993. “Structural
Social Psychology and the Micro-macro Problem.” Sociological Theory 11
(3):268-290.
Leik, R.K. 1992. “New Directions for Network Exchange Theory: Strategic
Manipulation of Network linkages.” Social Networks 14:309-324.
Lovaglia, Michael J., John Skvoretz, Barry Markovsky, and David Willer. 1999.
“Part 1: An Automated Approach to the Theoretical Analysis of Difficult
Problems.” Pp. 259-269 in Network Exchange Theory, edited by D, Willer.
Westport, CT: Praeger.
Lovaglia, Michael J., John Skvoretz, David Willer, and Barry Markovsky. 1995.
“Negotiated Exchanges in Social Networks.” Social Forces 74:123-155.
Macy, M. and A. Flache. 1995. “Beyond Rationality in Models of Choice.” Annual
Review of Sociology 21:73-91.
Markovsky, Barry. 1997. “Network Games.” Rationality and Society 9 (1):67-90.
Markovsky, Barry, John Skvoretz, David Willer, Michael Lovaglia, and Jeffrey
Erger. 1993. “The Seeds of Weak Power: An Extension of Network
Exchange Theory.” American Sociological Review 58:197-209.
Markovsky, Barry, David Willer, and Travis Patton. 1988. “Power Relations in
Exchange Networks.” American Sociological Review 53:220-236.
Marsden, Peter. 1983. “Restricted Access in Networks and Models of Power.”
American Journal of Sociology 88:686-717.
Mizruchi, Mark S. and Blyden B. Potts. 1998. “Centrality and Power Revisited:
Actor Success in Group Decision Making.” Social Networks 20:353-387.
Monge, Peter R. and Noshir S. Contractor. 2001. “Emergence of Communication
Networks.” Pp. 440-502 in The New Handbook of Organizational
Communication, edited by F.M. Jablin and L.L. Putnam. Thousand Oaks,
CA: Sage.
Nohria, Nitin. 1992. “Is a Network Perspective a Useful Way of Studying
Organizations?” Pp. 1-22 in Networks and Organizations: Structure, Form,
and Action, edited by N. Nohria and R.G. Eccles. Boston, MA: Harvard
Business School Press.
Parsons, Talcott. 1978. Action Theory and the Human Condition. New York: Free
Press.
Roberts, K.H. and C.A. O’Reilly. 1979. “Some Correlates of Communication Roles
in Organizations.” Academy of Management Journal 22:42-57.
Simmel, Georg. 1950. The Sociology of Georg Simmel, edited by K.H. Wolff. New
York: Free Press.
Simpson, Brent and David Willer. 1999. “A New Method for Finding Power
Structures.” Pp. 270-284 in Network Exchange Theory, edited by D. Willer.
Westport, CT: Praeger.

20
Skvoretz, John and David Willer. 1991. “Power in Exchange Networks: Setting
and Structure Variations.” Social Psychology Quarterly 54:224-238.
Skvoretz, John and Pidi Zhang. 1997. “Actors’ Responses to Outcomes in
Exchange Networks: The Process of Power Development.” Sociological
Perspectives 40 (2):183-197.
Skvoretz, John. 2000. “Fundamental Processes of Network Exchange: Results
from the South Carolina Laboratory for Small Group Sociological
Research.” Unpublished manuscript.
Smelser, Neil J. 1988. “Social Structure.” Pp. 103-129 in Handbook of Sociology,
edited by N.J. Smelser. Newbury Park, CA: Sage.
Stryker, Sheldon and Peter J. Burke. 2000. “The Past, Present, and Future of an
Identity Theory.” Social Psychology Quarterly 63 (4):284-297.
Szmatka, Jacek and David Willer. 1995. “Exclusion, Inclusion, and Compound
Connection in Exchange Networks.” Social Psychology Quarterly 58
(2):123-132.
Wagner, David G. 1984. The Growth of Sociological Theories. Beverly Hills, CA:
Sage.
Walker, Henry A., Shane R. Thye, Brent Simpson, Michael J. Lovaglia, David
Willer, and Barry Markovsky. 2000. “Network Exchange Theory: Recent
Developments and New Directions.” Social Psychology Quarterly 63
(4):324-337.
Wasserman, Stanley and Katherine Faust. 1997. Social Network Analysis:
Methods and Application. Cambridge, UK: Cambridge University Press.
White, Harrison. 1992. Identity and Control: A Structural Theory of Social Action .
Princeton: Princeton University Press.
Willer, David. 1999a. “Network Exchange Theory: Issues and Directions.” Pp. 1-
21 in Network Exchange Theory, edited by D. Willer. Westport, CT:
Praeger.
Willer, David. 1999b. “The Discovery of Weak Power.” Pp. 109-126 in Network
Exchange Theory, edited by D. Willer. Westport, CT: Praeger.
Willer, David and Bo Anderson (eds.) 1981. Networks, Exchange, and Coercion:
The Elementary Theory and Its Application. New York: Elsevier.
Willer, David, Michael J. Lovaglia, and Barry Markovsky. 1997. “Power and
Influence: A Theoretical Bridge.” Social Forces 76:571-603.
Willer, David and John Skvoretz. 1997. “Games and Structures.” Rationality and
Society 9 (1):5-35.
Willer, Robb and David Willer. 2000. “Exploring Dynamic Networks: Hypotheses
and Conjectures.” Social Networks 22:251-272.
Yamaguchi, Kazuo. 1996. “Power in Networks of Substitutable and
Complementary Exchange Relations: A Rational-Choice Model and An
Analysis of Power Centralization.” American Sociological Review 61
(2):308-332.

21