Content uploaded by Peter Kremer
Author content
All content in this area was uploaded by Peter Kremer on Feb 02, 2016
Content may be subject to copyright.
Small Group Research
2014, Vol. 45(1) 3 –36
© The Author(s) 2013
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1046496413510362
sgr.sagepub.com
Article
Cooperative and
Competitive Structures
of Trust Relations in
Teams
Dean Lusher1, Peter Kremer2, and Garry Robins3
Abstract
This article argues that it is not just trust-generating but also trust-
inhibiting mechanisms that operate in teams, and that these cooperative and
competitive structures of interpersonal relations of trust within teams may
affect team performance. Specifically, we propose that the presence of trust-
generating structures (e.g., reciprocity, trusting in the referrals of others
we trust, trusting in high performers and more experienced people) and
the absence of trust-inhibiting structures (e.g., not trusting in the referrals
of others we trust) are more likely to be associated with successful teams.
Using exponential random graph models, a particular class of statistical
model for social networks, we examine three professional sporting teams
from the Australian Football League for the presence and absence of these
mechanisms of interpersonal relations of trust. Quantitative network
results indicate a differential presence of these postulated structures of
trust relations in line with our hypotheses. Qualitative comparisons of
these quantitative findings with team performance measures suggest a link
between trust-generating and trust-inhibiting mechanisms of trust and team
performance. Further theorization on other trust-inhibiting structures of
trust relations and related empirical work is likely to shed further light on
these connections.
1Swinburne University of Technology, Hawthorn, Victoria, Australia
2Deakin University, Geelong, Victoria, Australia
3University of Melbourne, Victoria, Australia
Corresponding Author:
Dean Lusher, Swinburne University of Technology, Hawthorn, Victoria 3122, Australia.
Email: dlusher@swin.edu.au
510362SGR45110.1177/1046496413510362Small Group ResearchLusher et al.
research-article2013
by guest on January 7, 2016sgr.sagepub.comDownloaded from
4 Small Group Research 45(1)
Keywords
social network analysis, intra-group relations, methodology, exponential
random graph models
The concept of trust resonates with lay and academic audiences alike because
it is so pervasive in our social worlds. Trust is an almost universal part of
human social relations, from friends to family to economic exchanges (Fehr,
2009). Trust is seen as an enabler, a lubricant that reduces transaction costs
(Arrow, 1974; Gambetta, 1988; Uzzi, 1997) because it can be used as an
“alternative control mechanism” to “hierarchical contracts” (Gulati, 1995, p.
93) as well as to coercion and surveillance (Gambetta, 1988). Trust is com-
monly accepted to be a driving force in teams and by which goals can be
obtained, not just individually but collectively. This is especially so in the
context of social groups such as teams or organizations where the complexi-
ties of trust are multiplied not only by increased numbers of people but also
by the need for coordinated social action to achieve group outcomes. The
importance of trust and coordinated group action for collective outcomes is
evident when considering how trust is required between members of medical
teams involved in the care of patients (e.g., surgical teams), or emergency
teams responding to natural or human-induced disasters.
Research on the connection between trust and team effectiveness is
therefore of great importance, and much has been done already to unpack
the ways in which various elements of trust are related to a wide range of
team performance measures (Costa, 2003; Costa, Roe, & Taillieu, 2001; De
Jong & Elfring, 2010; Jones & George, 1998). However, one avenue that
we see missing from the current literature is an understanding of not just
how specific structures of trust relations may be associated with team effec-
tiveness (van de Bunt, Wittek, & de Klepper, 2005) but more particularly
whether there are specific structures of trust networks that might be associ-
ated with hampering team outcomes. By structure of trust we take an
explicitly social network approach, referring to the patterns of social net-
work ties (in this case interpersonal social ties of trust between team mem-
bers) that recur throughout the network. We examine well-accepted patterns
of trust that are important for successful group outcomes, such as reciproc-
ity, transitive closure (i.e., trusting in others that those we trust also trust in,
such that A trusts B, B trusts C, so A also trusts C), trust in those high in
experience/performance, and those of similar experience/performance. We
also examine patterns that are to be avoided, such as the non-closure. We do
this using a statistical models for social networks approach to help us
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 5
determine whether such patterns of trust are present more (or less) than we
would expect by chance in our data.
To this end, the current study addresses and extends the “need to think of
mechanisms and processes which can reinforce and help sustain team-based
initiatives” (McHugh, Niehaus, & Swiercz, 1997, p. 47) but also thinking
through those mechanisms that undermine them, and articulates such mecha-
nisms and processes theoretically in terms of structures of intra-team trust
relations. We examine interpersonal dyadic relations of trust between all team
members within a team that together constitute a social network of trust. We
argue an array of specific structures of trust relations that we consider may be
related to better team performance. We present four trust-generating or coop-
erative structures of trust that reflect widely accepted mechanisms of trust
(reciprocity, transitive closure, trust in high experience/performance, and
trust in similar experience/performance). Crucially, we also present a specific
structure of trust ties that may be associated with reduced team performance,
which we call a trust-inhibiting or competitive structure of trust. The article
then examines three teams to search for the presence of such theoretical
structures of trust. We use a particular statistical modeling approach to social
networks known as exponential random graph models (ERGMs), which are
tie-based models for predicting network relations. We then take these quanti-
tative empirical results and qualitatively compare them with team perfor-
mance measures to see if there is an association between them.
Importantly, it has been noted that “studying multiple mediating mecha-
nisms provides a more nuanced understanding of how trust affects team per-
formance than studying them one at a time” (De Jong & Elfring, 2010, p.
543). Using ERGM, we simultaneously investigate this range of trust-gener-
ating and trust-inhibiting network substructures (and as such, the theoretical
concepts these substructures represent, such as reciprocity or transitive clo-
sure) and thereby assess the importance of all of these concepts, one against
the other, within a single analysis for each team. A further decided advantage
of ERGMs is that they can specifically take the dependencies inherent in
social relations into account in principled ways that make them superior to
models that assume independent observations (Snijders, 2011), making them
cutting-edge models for social network data (Lusher, Koskinen, & Robins,
2013). There are very few studies that have studied trust using an ERGM
methodological approach, and so this method offers potentially new insights
into the structure of trust relations. Foreshadowing our results, we have evi-
dence that the presence of cooperative structures of trust and the absence of
competitive structures of trust are positively associated with better team
performance.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
6 Small Group Research 45(1)
Cooperative and Competitive Structures of Trust
Relations in Teams
Trust is a term used knowingly in general parlance but one that also has con-
siderably different definitions depending on the research context, making a
unitary definition of trust a difficult task (Costa et al., 2001; Dirks & Ferrin,
2002; Earle & Siegrist, 2006; Krackhardt, 1996; McEvily, Perrone, & Zaheer,
2003). Trust can be seen as a general willingness to be vulnerable (Mayer,
Davis, & Schoorman, 1995) involving both risk and interdependence
(Rousseau, Sitkin, Burt, & Camerer, 1998). In a recent review, it was found
that of the 171 articles examined on trust in organizations, there were 129
different measurements of the concept (McEvily & Tortoriello, 2011). This
review found that trust is seen generally in two ways: as a psychological state
mainly in the social psychology literature and as a general organizing prin-
ciple. Recently, Lusher, Robins, Pattison, and Lomi (2012) have argued that
it is useful to view trust through both of these lenses at the same time.
In this study, we take an explicitly relational (and social network) perspec-
tive on trust, in line with previous research (Becerra & Gupta, 2003; Burt,
2001; Burt & Knez, 1995; Buskens, 1998; Ferrin, Dirks, & Shah, 2006;
Mayer et al., 1995; van de Bunt et al., 2005). We do so because this permits
us to look for specific patterns of trust relations—patterns that represent
social processes of trust—and assess whether these structures of trust are
associated with team performance. Importantly, the notion of mixed-motive
interactions highlights the dilemma of some situations in which people are
conflicted between desires to cooperate or compete with one another (Davis,
Laughlin, & Komorita, 1976; Komorita & Parks, 1995; Schelling, 1960). In
the case of sporting teams, teams that are ongoing (De Jong & Elfring, 2010),
the simultaneous operation of cooperation and competition among team
members is quite evident:
The public face of the team attempts to present to the rest of the world that of a
“family,” whose shared goal of winning games and championships bonds its
individual members together. But the structure of athletic careers is such that
individuals on teams are constantly competing against each other—first for a place
on the team, then for playing time, for public recognition and star status, and
eventually, just to stay on the team. (Messner, 1997, p. 344)
Teams are not purely cooperative enterprises (Messner, 1997) and have
resultant competition between players, which others have noted as antagonis-
tic cooperation (Riesman, 1953). While we could restrict such sentiments to
sporting teams, it is clear that within any group that people have agendas
which clash with those of others, and so the notion of antagonistic cooperation
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 7
can possibly be seen to reflect more general conceptions of teams or groups
(e.g., Labianca & Brass, 2006; Sampson, 1968; Tsai, 2002). Clearly some
aspects of competition are fundamental to team development, improvement,
and outcomes. Yet competition so fierce that it has an impact on trust is likely
to have a deleterious effect on the team (or on the other hand, be indicative of
existing problems within the team). An example of this could be two players
in which many others trust both who do not trust one another. There is an obvi-
ous tension to balancing these cooperative and competitive relations between
team members.
While it is obvious that trust relations can be viewed as collaborative,
what is less clear is how they might be seen as competitive. To be clear, we
are not so much suggesting that team members are competing for trust, but
that competition between team members is associated with a lack of trust
between them, and this inhibits trust not just between these competing team
members but may also cause split loyalties among their supporters and lessen
trust throughout the team. We argue that certain structures of trust relations
can be evaluated as competitive, and this begins to answer the question of
how the structure of trust relations may impinge on team performance. Such
a view is informed by the insight that it is not just important where social ties
are but also where they are not (White, Boorman, & Breiger, 1976). In the
case where social ties are expected to be present, but instead are absent, we
can interpret this as indicative of a lack of trust. And it is this lack of trust that
we argue is indicative of competition among team members and has trust-
inhibiting effects.
We note here that we are not examining negative ties, such as distrust,
negative gossip, work difficulty or conflict, on which much is written
(Ellwardt, Labianca, & Wittek, 2012; Huitsing et al., 2012; Labianca &
Brass, 2006; Labianca, Brass, & Gray, 1998). We are instead examining posi-
tive ties, but noting that the absence of some trust ties in specific structures
can be informative of trust within a team. As such, this is a univariate analysis
of a standard positive tie network of trust relations, but one that can inform
the sorts of issues that a simultaneous analysis of negative and positive ties
can bring. We do not argue it as a replacement for studying negative ties, but
simply as another avenue into this terrain.
Given the multifaceted nature of trust (Mayer et al., 1995), it is expected
that there are differing explanatory causes or social mechanisms (Hedström
& Swedberg, 1998) that explain the presence of interpersonal trust relations
in teams. Drawing on the trust literature from organizational research and the
sporting literature on cooperation and competition within sporting teams, we
hypothesize five such social mechanisms reflective of cooperative and also
competitive structures of trust relations within teams. We note that of these
by guest on January 7, 2016sgr.sagepub.comDownloaded from
8 Small Group Research 45(1)
five mechanisms, one corresponds to a competitive (or trust-inhibiting) struc-
ture of trust (not trusting in the referrals of others we trust) and the other four
to cooperative (or trust-generating) structures of trust (reciprocity, trusting in
the referrals of others we trust, trusting in high performers and more experi-
enced people, and trusting in others of similar performance/experience). Our
hypotheses relate to network structures (or network configurations) presented
later in Table 1 and explained in more detail in the section on social network
analysis (SNA), and in the methods. The expectation of the presence of these
structures is not just that we will see some of them in a network of trust rela-
tions, but more of such structures than would be expected than by chance
(and thus a statistically significant number of such network structures—see
the “Methods” section). Our major argument here is that trust relations that
are of value to the team should be reflected by an absence of an effect for
trust-inhibiting structures (i.e., Hypothesis 3 [H3]) and a presence of trust-
generating structures (i.e., Hypothesis 1 [H1], Hypothesis 2 [H2], Hypothesis
4 [H4], and Hypothesis 5 [H5]). Trust is equally about the absence of nega-
tives as it is the presence of positives. We now outline these social mecha-
nisms of trust relations.
First, trust is commonly seen as reciprocal (Burt & Knez, 1995; Gambetta,
1988; McEvily et al., 2003), specifically as an expectation that others will
reciprocate such behavior (Tyler & Kramer, 1996). According to Blau (1964),
reciprocity is a universal human activity, representing social exchange.
Indeed, as trust is seen as risky or creating uncertainty (Jones & George,
1998), the expectation that trust is reciprocated is a psychological heuristic
that can be used to ameliorate such concerns (Lusher et al., 2012). Within a
team, we argue it would be expected that if trust is conferred upon another
that it will be reciprocated. We note that such an explanation relates to the
structure of social ties regardless of the individual qualities of team members.
In reciprocal relations, a social tie (in this case, trust) comes about because of
another already existing social tie. That is, it is because “you scratch my
back” that “I will scratch yours.” Trust may be reciprocated due to signaling
by one partner in the dyad to the other, or it may simply be inferred. We do
not claim to know the underlying psychology of such relationship formation
(though for an interesting take on this issue, see Lusher et al., 2012). This
leads us to our first hypothesis:
H1: Trust relations will be reciprocated.
A second social mechanism of trust refers to the fact that trust has also
been viewed as involving transitive closure or triadic relations; that is, if A
trusts B, and B trusts C, then A will also trust C, or more colloquially referred
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 9
Table 1. Summary of Network Effects Evaluated Using ERGM.
Parameter Image Explanation
Purely structural effects
1. Arc One actor nominating
another actor (baseline
propensity to form ties)
2. Mutual Mutual ties between two
actors (models the
tendency for reciprocation
across the graph)
3. Simple
connectivity
(2-path)
Chain-like or simple
hierarchy formation of ties
4. Popularity
spread
Indicative of the presence
of highly nominated
individuals within a
network (models the
indegree distribution)
5. Activity spread Indicative of the activity of
actors to engage many
others (models the
outdegree distribution)
6. Path closure Triadic closure (i.e., a friend
of a friend is a friend)
7. Popularity
closure
Triadic closure where
two nodes share high
popularity
8. Cyclic closure The propensity for ties to
form as part of a cyclic
triad or a multiple cyclic
configuration
9. Multiple
connectivity
The propensity for ties to
form as part of formations
involving multiple short
paths between actors
10. Shared
popularity
The propensity for
popularity based structural
equivalence involving
multiple short paths
between actors
(continued)
by guest on January 7, 2016sgr.sagepub.comDownloaded from
10 Small Group Research 45(1)
Parameter Image Explanation
Actor-relation effects
11. Sender The attribute of the sender
of the tie, which may be
continuous, categorical
or binary (models the
propensity of an actor with
the attribute to send ties,
i.e., to be active in network
terms)
12. Receiver The attribute of the receiver
of the tie, which may be
continuous, categorical
or binary (models the
propensity of an actor
with the attribute to be
popular)
13. Homophily A selection of a person with
similar levels of the same
continuous attribute (e.g.,
age)
Note. ERGMs = exponential random graph models.
Table 1. (continued)
to as a friend of a friend is a friend (Burt, 2005; Buskens, Raub, & van der
Veer, 2010; Granovetter, 1985; Kipnis, 1996; McEvily et al., 2003). Of
course, the importance of triadic relations goes back to ideas of the famous
sociologist Simmel (1950), and the fact that this constitutes the smallest
group with a majority. Heider’s (1958) well-known balance theory, as too
Cartwright and Harary’s (1956) elucidation of Heider’s idea into structural
balance theory (though with less of a focus on the psychological aspects),
have demonstrated the importance of triadic relations for over 50 years.
Granovetter’s (1973) strength of weak ties argument further illustrated the
importance of triadic relations, highlighting the discomfort faced when two
of our friends are not at least connected by a weak tie. Trust triads therefore
indicate trust beyond the dyad and at a group level. Robins, Pattison, and
Wang (2009) demonstrated for a trust network within an organization in the
presence of transitive closure (but not cyclic closure), concluding that trust
was structured hierarchically for triadic relations. Yet Robins et al. (2009)
also noted the presence of reciprocity, and hence of balance with regard to
trust relations. As such, trust may be both balanced and hierarchical
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 11
simultaneously within the one social context. Both of the above hypotheses,
for reciprocal trust and transitive closure trust, we view as indicative of coop-
erative structures of trust. While one mechanism is balanced and the other is
hierarchical, both mechanisms demonstrate cooperation—the adherence to
mutual structures of trust or to hierarchical structures of trust.
H2: Trust relations will involve transitive closure of ties.
The third social mechanism of trust can be seen as the opposite of the
second, and as indicative of a competitive structure of trust relations. This
third social mechanism draws on the absence of social ties—that is, where
trust relations may be expected but are not present. While it is noted above
that trust should be transitive (i.e., if A trusts B, and B trusts C, then A should
also trust C), where it is not (i.e., where A does not also trust C) there is likely
to be tension (Granovetter, 1973; Heider, 1958). As an example, consider two
prominent and influential team members who are both trusted by many team-
mates (indeed, these two influential players are trusted by the same team-
mates). If these two prominent team members do not trust one another when
it is expected they should (i.e., a trust relation is absent when it is expected to
be present), this is likely to indicate a division in the team, and result in ten-
sion. Krackhardt’s (1987) cognitive social structures (CSS) are informative
here. CSS refers to the ways in which individuals within a network perceive
of the social ties of others within that network, such that actor i reports on
social ties from actor j to k. People are aware of social ties beyond their own
specific relations and these are important to consider (Ferrin et al., 2006).
H3: Trust-inhibiting relations will involve the non-closure of trust ties.
A fourth social mechanism of trust refers to the fact that some team mem-
bers with certain individual-level characteristics are also likely to be trusted
more than others because of these qualities. Clearly, the uncertainty or risk
inherent in trusting another can be alleviated by investing trust in those who
are high performers or who have status within the team (Jones & George,
1998; Mayer et al., 1995). Gorgenyi (1998) highlights the importance of per-
formance and experience to informal hierarchical relations in sporting teams,
though clearly in teams more general factors such as ability (Mayer et al.,
1995) and reputation (Burt, 2005) are valued. We would expect that those
team members who are high on performance are more likely to be trusted
because it is these people who play a critical role in the success of the team.
Furthermore, we expect that team members who are high in experience are
also likely to garner trust because they can pass on knowledge to younger
by guest on January 7, 2016sgr.sagepub.comDownloaded from
12 Small Group Research 45(1)
team members and provide the cool head in a crisis. Furthermore, their lon-
gevity in the team shows a commitment to organization (Costa, 2003).
Drawing on social identity theory and the related self-categorization theory,
there is evidence to suggest that identification with group causes preferential
choices of others (Tajfel & Turner, 1979; Turner, Hogg, Oakes, Reicher, &
Wetherell, 1987). Self-categorization theory details the process by which
people cognitively represent social groups in terms of prototypes. A prototype is a
subjective representation of the defining attributes (e.g., beliefs, attitudes,
behaviours) of a social category, which is actively constructed from relevant social
information in the immediate or more enduring interactive context. (Hogg, Terry,
& White, 1995, p. 261)
People resembling the prototype are more likely to be liked, admired, and
indeed trusted than those who do not fit the prototype. It is through this iden-
tification process that individuals and groups align, so trust in the prototype
is trust in the group itself. Freeman’s (1977) concept of central people within
networks (i.e., those who receive more ties) being advantaged can also tie in
here. In this case, such people are central due to their individual-level attri-
butes—they receive more ties due to their individual-level qualities.
H4a: Trust relations will be afforded to team members high in
experience.
H4b: Trust relations will be afforded to team members high on
performance.
Finally, the concept of birds of a feather flock together is known as homoph-
ily (McPherson, Smith-Lovin, & Cook, 2001) that appears to be a universal
human tendency for people to connect with similar others. Research suggests
that shared membership of a group makes people more trustworthy toward one
another because of this social similarity in characteristics (Earle & Siegrist,
2006; Lincoln & Miller, 1979; McAllister, 1995; Sheppard & Tuchinsky,
1996). Again, the cognitive heuristic of choosing people like me reduces uncer-
tainty with regard to trust, and is a way of overcoming tensions involved in the
risk of trusting others, and a way of expressing one’s confidence in others
(Axelrod, 1984). As a result, we expect that team members are more likely to
trust others with similar experience to themselves. Furthermore, team members
will also trust others of similar performance level to themselves.
H5a: Trust relations will be afforded to other team members of similar
experience.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 13
H5b: Trust relations will be afforded to other team members of similar
performance.
In summary, we have hypothesized multiple reasons of how trust relations
are structured within teams that represent competing social mechanisms
regarding how interpersonal trust occurs in teams. The first, second, fourth,
and fifth hypotheses relate to cooperative structures of trust relations because
they are seen as trust-generating mechanisms. Conversely, the third hypoth-
esis is seen as a competitive structure of trust because it is a trust-inhibiting
mechanism—resisting the formation of trust ties. We expect that teams that
have more trust-generating cooperative structures of trust and an absence of
trust-inhibiting competitive structures of trust are likely to be better function-
ing teams and be more successful. Conversely, those teams that lack coopera-
tive structures of trust and/or have competitive structures of trust are less
likely to be functioning well and perform more poorly.
To be able to make inferences about such social mechanisms we need to
test them, one against the other, empirically. It is for this reason that we
employ SNA and use a particular class of statistical model for social net-
works, ERGMs, so that we can determine statistically that explanations for
trust relations are more (or less) likely. Finally, we compare the quantitative
network models results of these structures of trust qualitatively with team
performance measures to see if this thesis is supported.
SNA and ERGMs
If we are interested in trust as a relational construct, then it is necessary to
have a methodology capable of dealing with relational data. One such meth-
odology is SNA, which by formal definition is a set of techniques that focus
on the “relationships [sic] among social entities, and on the patterns and
implications of these relationships” (Wasserman & Faust, 1994, p. 3). As
such, the term social network can be moved beyond use as a metaphor of the
social world and implemented as a specific, local-level relational methodol-
ogy (Emirbayer & Goodwin, 1994). So what we examine within network
analysis is not individuals but the collection of relations between the
individuals.
In a social network, individuals (or actors) are represented as nodes in a
graph, and the relations between them are represented as edges or lines. In a
very practical sense, a social network can be measured by asking all individu-
als in a particular social context (e.g., a team) about a particular social rela-
tion with others in the network (e.g., “Who do you trust?”). The network is
by guest on January 7, 2016sgr.sagepub.comDownloaded from
14 Small Group Research 45(1)
then the combination of all possible nominations among all pairs of actors in
the network. In addition, individual-level variables (e.g., years with the team)
can be incorporated and investigated within a social network.
Our primary aim in using SNA is to predict how and why social relations
occur—in this case, trust. Networks are relational (i.e., dependent) by nature,
but standard statistical tests assume the independence of observations, which
necessarily results in a disjunction between theory and method. ERGMs pro-
vide a means to resolve this tension between theory and method. ERGM are a
particular class of statistical model for social networks that were originally
developed by Frank and Strauss (1986), and further refined by Wasserman and
Pattison (1996) and Pattison and Wasserman (1999). These models have the
capacity to address complex social structures. Recent model derivations have
the capability of examining both individual-level variables and structural rela-
tions simultaneously (Daraganova & Robins, 2013; Robins & Daraganova,
2013; Robins, Elliott, & Pattison, 2001; Robins, Pattison, & Elliott, 2001).
New specifications for ERGM allow for the examination of higher order
social structures (Robins et al., 2009; Snijders, Pattison, Robins, & Handcock,
2006) and have permitted a more detailed investigation of social structures
and individual attributes together, providing a possible means to handle both
structure and agency. The assumption of interdependency which underpins the
ERGM framework, and which suggests that in social contexts individuals (or
groups) are not independent of one another, is seen as advantageous and com-
plex dependency assumptions have been proposed for this purpose (Pattison
& Robins, 2002; Snijders et al., 2006). More detailed introductions to ERGMs
are provided by Contractor, Wasserman, and Faust (2006) and Robins,
Pattison, Kalish, and Lusher (2007) and Lusher et al. (2013).
The ERGM class of statistical models for social networks essentially
works as a pattern recognition device for predicting why social network ties
occur. We call these patterns of social network ties network configurations,
and a range of such configurations are presented in Table 1. Network con-
figurations (or effects) give us an understanding of those structural processes
necessary to explain how the network came about. A network configuration
is a consequential pattern that may represent an underlying social process (or
social mechanisms), and as such it is claimed that a network structure is the
consequence of a dynamic process. The cooperative and competitive struc-
tures of trust relations outlined previously are themselves social mechanisms
that can be examined in a network by including network configurations that
represent such processes. So the processes of H1 through H5 are aligned with
network substructures that we then look for in our model to see if such sub-
structures occur at greater (or less) than chance levels.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 15
While one could simply count the number of reciprocated ties or the num-
ber of triangles in a network, one would not know if there were more triangles
(or less) than expected by chance. It is important to note that networks formed
completely at random (i.e., networks without structure) still have some recip-
rocated ties, and some triangles. For example, in the simulation of Bernoulli
random graphs, in which ties are added to a graph independently of one
another (i.e., ties are added with equal probability between any pair of actors),
we are still likely to observe a number of triangles. So to claim that there is
an effect for transitive path closure (the notion that a friend of a friend is a
friend, the formation of a triad), we would need to see more triangles than we
would expect to see by chance (i.e., if the network ties were formed ran-
domly). ERGM therefore permits the statistical evaluation of competing
hypotheses for why social ties may be present. Using ERGM, we can exam-
ine a trust network for reciprocity, transitive closure, and homophily all at the
same time and evaluate if these structures occur at greater (or less than) we
might expect in a random network.
Finally, ERGMs permit differentiation between endogenous network pro-
cesses and processes related to actor attributes (Lusher et al., 2013).
Endogenous (or purely structural network effects) reflect the ways in which
network ties self-organize. As such, purely structural effects reflect processes
in which ties form due to the presence or absence of other ties. Conversely, ties
may form due to actor attributes, and are known as actor-relation effects, high-
lighting the association of a tie with the qualities of an actor. This is particu-
larly important for H4 and H5 in which we state that trust occurs due to
experience and performance of team members. It is known that when endog-
enous processes are not controlled for, the effects of actor attributes may be
overestimated and therefore spurious claims may be made about the impact of
individual-level qualities (Lusher & Ackland, 2011). The ERGM framework
therefore provides a more principled way of making inferences about the asso-
ciation of actor attributes and network ties because it can distinguish between
whether ties were formed due to the attributes of the actors or whether simply
the actor’s popularity is the result of being embedded within many purely
structural network structures. Other approaches, such as linear regression
using attributes and degree centrality, cannot delineate these processes.
Finally, we reiterate that ERGMs are models for predicting the presence of
social ties. This makes the unit of analysis not the number of individuals within
the team (n) but the number of possible ties between all players, or n(n − 1),
with n − 1 accounting for the fact that people cannot make self-nominations. In
the case of a 34-player team as in Club A, there are 1,122 observations, thereby
providing plenty of statistical power for the analysis.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
16 Small Group Research 45(1)
Methods
Participants
Three Australian Football League (AFL) teams from Melbourne, Australia,
participated in this research. Excluding missing data, there were 107 male ath-
letes from three AFL clubs (Club A, n = 34; Club B, n = 36; Club C, n = 37).1
Based on a potential 40-person playing list, participation rates ranged from
85% to 92.5%, with data collected between December 2004 and January 2005.
The AFL has representative teams in most states of Australia making it a
national competition and the pre-eminent sport in the country. These clubs
were chosen due to their perceived variability in performance and culture.
Materials
A pen-and-paper survey instrument was administered as part of a larger study
that focused on social structures among sports teams (Lusher, Robins, &
Kremer, 2005). A staff roster that contained a list of names of all playing and
coaching staff with assigned numeric identifiers was provided so responses
for network questions could be indicated. The trust network was measured by
asking participants with respect to their particular team “Who do you trust?”
This network is binary (so a relationship is either present or absent) and
directed (so A trusts B is different from B trusts A), and self-nominations
(i.e., I trust myself) were not permitted. We have deliberately kept the defini-
tion of trust open here to include instrumental and expressive trust because in
these intense team environments where players are constantly interacting,
there is good reason to believe that both may be important. Restricting trust
nominations only to work-related matters might possibly discount the emo-
tional bonding and strong interpersonal ties that should be promoted to
develop trust among team members (De Jong & Elfring, 2010). Trust net-
works for the three clubs are graphically presented in Figures 1, 2, and 3.
A number of individual-level measures were also included. A measure of
AFL playing ability (hereafter labeled performance) was included as a con-
tinuous variable for each player. Performance was djudged by asking the
social network question “Who are the best players at your club?” of the play-
ers themselves and where a player could nominate more than one player. As
such, each player scored between 0 (for no nominations) and n − 1 (for nomi-
nations by all others in the team, except himself, as no self-nominations were
permitted; n = number of players in the team). This resulted in a count of the
number of nominations received for each player that was then included as a
continuous node-level variable for each player. Players who received greater
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 17
numbers of best player network nominations were considered to have higher
playing performance or, in other words, to be better or more skilled players.
Experience was measured as the self-reported number of AFL games played
by respondents.
As a measure of general team performance, we examined the ladder (or
points table) for the 16-team competition across the years 2004 and 2005,
which followed and preceded the data collection, respectively. Our lack of
precision, for example, by not mentioning which quartile of the ladder a team
finished in, is deliberate and required so that the identities of teams are pro-
tected and remain confidential. We found that Club A finished near the bot-
tom in both the years, Club B can be described as finishing mid-range on the
ladder, and finally Club C can be described as finishing near the top of the
ladder (and so having higher team performance).
The profile for each club is presented in Table 2. Preliminary analyses
(ANOVA) indicated that there were no differences in individual performance
between clubs for experience, performance, and age.
Figure 1. Trust network for Club A (n = 34).
by guest on January 7, 2016sgr.sagepub.comDownloaded from
18 Small Group Research 45(1)
Procedure
Participants were briefed on the purpose of the study and then completed the
survey in a single session of approximately 30 min duration. The club-spe-
cific rosters were used by players to record responses (i.e., numeric codes
for individual players) to the network questions. All surveys were completed
in a club meeting room during the off-season period from December 2004 to
January 2005. Participants were able to spread out to some degree when
responding to the survey, in much the same way that school students in a
classroom are spread out in an exam. Two researchers were present, oversee-
ing data collection and encouraging self-completion and non-monitoring of
others. Players gave their de-identified completed surveys directly to the
researchers to ensure confidentiality. Ethics approval for the study was pro-
vided by the University of Melbourne Human Research Ethics Committee.
Data Analysis and ERGM Specification
Data analyses of interpersonal trust relations (i.e., the trust network) involved
ERGMs using the PNet freeware program (Wang, Robins, & Pattison, 2009).
Figure 2. Trust network for Club B (n = 36).
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 19
Figure 3. Trust network for Club C (n = 37).
Table 2. Summary of Attributes for the Three AFL Clubs (M ± SD).
Club A Club B Club C
M (SD)M (SD)M (SD)
Age 22.8 (3.7) 22.6 (3.2) 22.7 (3.7)
Experience (games played) 73.3 (70.0) 60.6 (71.0) 65.0 (71.4)
Performance (peer-rated nominations) 4.4 (8.5) 3.9 (7.8) 4.9 (10.3)
Overall team performance Low Medium High
Density 0.1791 0.0833 0.1111
Highest in degree 14 14 10
Note. AFL = Australian Football League.
We included the network effects from Table 1 into our models. All 10 of the
purely structural parameters were included to examine the self-organizing
properties of network ties. For the actor-relation parameters, we included
by guest on January 7, 2016sgr.sagepub.comDownloaded from
20 Small Group Research 45(1)
separate sender, receiver, and homophily2 effects for the variables’ perfor-
mance and experience. Broadly, this class of effects examines the formation
of relations of trust due to individual qualities of players.
We fixed (or conditioned on) the outgoing (sent) trust ties of three players
in Club A as these players indicated that they trusted everyone (or nearly
everyone) in their team. For excessive network nominations for trust, it is a
reasonable approach to fix3 the outgoing ties of such nodes (i.e., making them
exogenous to the modeling process) as a method to deal with such high num-
bers of nominations (see Lusher et al., 2012).
Importantly, each team is statistically examined separately. Therefore, we
do not make statistical comparisons across teams, but rather examine statisti-
cally whether the presence of reciprocity, transitivity, or receiving of ties due
to experience occurs at greater than chance levels. As such, we essentially
compare each club with other possibilities of how its social ties might be
arranged. We note that as age and experience were highly correlated (r = .93,
p < .05), we only included experience in our social network models to avoid
problems associated with co-linearity. The correlation between experience
and performance was also significant but more moderate, r = .57, p < .05. The
densities of the three trust networks were Club A = .1791, Club B = .0833,
and Club C = .1111.
Finally, we note that we do not include a measure of team performance in
our models, so there is no statistical relationship between the ERGM esti-
mates and team-level success. It would be possible to run some form of mul-
tilevel analysis of the networks to determine this if there were enough teams
to justify it (e.g., Lubbers, 2003). However, with only three teams, this was
not possible. Rather, in the current study, we qualitatively align team-level
success data with the network measures of Hypotheses 1 through 5 that arise
from the ERGMs and look for associations between the two.
Results
We separate the results here into (a) model estimates and (b) qualitative
implications.
Model Estimates
Network parameter estimates for the ERGMs are presented in Table 3 (with
associated standard errors in parentheses). As a general rule, an estimate
greater (in absolute value) than 2 times standard error is regarded as statisti-
cally significant (denoted by an asterisk *). Estimates are provided in square
brackets [], and their standard errors in parentheses () within these brackets.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 21
We do not go through each of the effects in the model, but instead focus on
those effects directly related to our hypotheses. However, we do note that we
have included a number of control effects so that we account for complex
interdependencies and self-organizing tendencies of network ties (Rank,
Robins, & Pattison, 2010). The specific variables selected for the set of con-
trol variables follows guidelines on ERGM (Lusher et al., 2013, p. 175). We
note all parameters in our models indicated good convergence of the Markov
chain Monte Carlo maximum likelihood estimation (MCMCMLE) algo-
rithm, and that in the goodness of fit the convergence t-ratios for all included
effects <.1 and for all non-included effects <2.0. This suggests that these
three models represent our data well, and that other network effects that were
not specifically modeled are not extreme and are therefore well captured by
the model (Koskinen & Snijders, 2013).
Trust relations will be reciprocated. There was a significant and positive effect
for reciprocity for all three clubs [A: 1.43 (.45)*, B: 2.02 (.50)*, and C: 1.91
(0.46)*], which means that, given the other effects in the model, players
reciprocate their trust nominations more than expected by chance. This sup-
ports H1 regarding structures of mutual trust, and aligns with a considerable
body of literature on trust that notes the importance of reciprocity.
Table 3. ERGM Parameter Estimates for Trust Relations and Player
Characteristics for Three AFL Clubs (Standard Errors in Parentheses).
Club A Club B Club C
Hypothesized effects
(H1) Mutual 1.43 (0.45)* 2.02 (0.50)* 1.91 (0.46)*
(H2) Path closure 1.01 (0.27)* 0.37 (0.37) 0.539 (0.266)*
(H3) Shared popularity 0.82 (0.35)* 0.10 (0.03)* 0.07 (0.06)
(H4a) Receiver–experience 0.0053 (0.0025)* 0.002 (0.002) 0.007 (0.002)*
(H4b) Receiver–performance 0.026 (0.021) 0.06 (0.02)* 0.025 (0.012)*
(H5a) Homophily–experience 0.006 (0.002)* 0.005 (0.002)* 0.009 (0.002)*
(H5b) Homophily–performance −0.009 (0.020) 0.02 (0.02) 0.005 (0.01)
Control effects
Arc −11.44 (2.85)* −4.11 (0.62)* −3.99 (0.65)*
Simple connectivity −0.12 (0.08) −0.47 (0.23) −0.12 (0.09)
Popularity spread 4.60 (1.45)* 0.61 (0.29)* 0.52 (0.30)
Activity spread −0.49 (0.39) 0.19 (0.28) 0.45 (0.30)
Popularity closure −0.31 (0.73) 0.74 (0.38) 0.20 (0.25)
Cyclic closure −0.259 (0.127)* −0.19 (0.19) −0.07 (0.14)
Multiple connectivity 0.08 (0.09) 0.31 (0.26) −0.05 (0.11)
Sender-performance 0.016 (0.017) 0.01 (0.02) 0.01 (0.01)
Sender-experience 0.0037 (0.0020) 0.000 (0.001) −0.0033 (0.0018)
Note. Each club is modeled separately. ERGM = exponential random graph models; AFL = Australian
Football League.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
22 Small Group Research 45(1)
Trust relations will involve transitive closure of ties. In Club A, the significant and
positive effect for path closure [1.01 (.27)*], and the significant and negative
effect for cyclic closure [−.259 (.127)*], indicate that hierarchical multiple
triadic closure is very likely and generalized (or cyclic) triads are very unlikely.
As such, trust is not given in a general sense to others in the team, but is given
in a hierarchical manner, and this is support for H2. In Club B, there was no
significant effect for path closure. However, Club C did show a significant and
positive effect [.539 (.266)*] for path closure, indicating support for H2.
Trust-inhibiting relations will involve the non-closure of trust ties. The third hypoth-
esis examines a trust-inhibiting structure of trust, as opposed to all other
hypotheses that examine trust-generating structures. For Club A, there is a
significant and positive shared popularity effect [.82 (.35)*], indicating that it
is common for players to agree on the same players as team members they
trust (see Table 1 effect 10 for a visualization of this effect). This is support
for the third structure of trust (H3) and indicates that some highly popular
teammates do not trust one another. For Club B, we also find a significant and
positive effect for shared popularity [.10 (.03)*], demonstrating support for a
trust-inhibiting structure of H3. Unlike Clubs A and B, for Club C there is no
shared popularity effect (H3), and importantly this indicates the absence of
trust-inhibiting structures of trust relations for Club C.
Trust relations will be afforded to team members high in experience and perfor-
mance. With regard to trusting team members high in experience (H4a), Club
A demonstrates a positive and significant receiver effect for experience [.0053
(.0025)*], indicating that players with more experience are significantly more
likely to be trusted than others. For Club B, there is a significant and positive
receiver effect for performance (H4b) [.06 (.02)*], indicating that players of
higher rated performance were more likely to be trusted. Finally, for Club C,
there are two significant and positive receiver effects: one for experience
(H4a) [.007 (.002)*] and one for performance (H4b) [.025 (.012)*]. These
separate main effects for receiving ties indicate that players high in perfor-
mance are more likely to be trusted, and also players high in experience are
more likely to be trusted by others. Again, these are separate effects, and it
may be that they are not the same people (i.e., the people who are trusted for
high performance may not have high experience, and vice versa).
Trust relations will be afforded to other team members of similar experience and
performance. For Club A there is a significant and positive homophily effect
for experience [.006 (.002)*], indicating that players trust others of a similar
level of experience to themselves (supporting H5a). This effect is over and
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 23
above the receiver effect for experience, and each effect must be interpreted
in light of the other effect, and so in conjunction these two effects are inter-
preted as players trust others of similar or greater experience to themselves.
However, there are no such effects for peer-rated individual performance.
Club B also has a significant and positive homophily effect for experience
[.005 (.002)*], indicating that players trust others with similar levels of expe-
rience to themselves. Furthermore, Club C also shows a positive and signifi-
cant homophily effect for experience [.009 (.002)*]. We note that there was
no support for H5b regarding trust homophily due to performance in any of
the clubs.
Qualitative Implications
In this section, the quantitative results from the ERGMs are compared qualita-
tively with the team performance measures we have for the three AFL teams. A
summary of the results for each hypothesis by club is given in Table 4 below.
The most interesting effect is the difference between teams with respect to
the trust-inhibiting structure of trust relations. While it may be seen as only
Table 4. Summary of Support for Each of the 5 Hypotheses for Each Club
(O = Hypothesis Supported) for Cooperative and Competitive Structures of Trust.
Club
Network configuration
graphic
Club A Club B Club C
Overall team performance Low Medium High
Trust generating structures
H1 (mutual) O O O
H2 (path closure) O O
H4a (receiver experience) O O
H4b (receiver performance) O O
H5a (homophily experience) O O O
H5b (homophily performance)
Trust-inhibiting structures
H3 (shared popularity) O O
by guest on January 7, 2016sgr.sagepub.comDownloaded from
24 Small Group Research 45(1)
one effect among many, to make this importance of such a trust-inhibiting
effect more concrete, consider the effects on a team if the captain and the vice
captain do not trust one another. The presence of network structures like this
for Clubs A and B suggest that such a situation is actually taking place in both
of these clubs. The importance of the perception of social ties between alters
has been demonstrated by Krackhardt (1987). If in the above example ten-
sions between captain and vice captain were not masked and those other
players who trusted both were aware of the absence of trust between the two,
this might create psychological tension for these other players (Heider, 1958)
and create a situation of conflicted loyalties. This situation of trusting two
other people who do not trust one another is typically resolved by either (a)
trust occurring between captain and vice captain or (b) by players who trust
the captain and vice captain withdrawing trust for one of these two. It is for
this reason that we refer to such a structure as trust-inhibiting and competitive
because it drives divisions between team members.
Importantly, for Club C, there was an absence of the hypothesized trust-
inhibiting structure (H3), which differentiates this club from the other two.
Furthermore, it is worth reiterating that with respect to their overall team
performance, Club C was rated highest, followed by Club B and then Club A.
We note that team performance was not measured quantitatively against the
presence/absence of these structures of trust, but rather that such judgments
were made qualitatively and based on comparing the quantitative network
results with team performance. In any case, while a myriad of factors can
contribute to the overall team performance, there is some face validity to the
claim that teams that do well are not hampered by trust-inhibiting structures
of trust relations. The fact that the club who finished highest overall of the
three teams on the points table did not have trust-inhibiting mechanisms is
promising. Clearly, it may not simply be just the presence of ties but also in
some cases the absence of ties that is important (White et al., 1976).
With regard to trust-generating structures of trust relations, it is again impor-
tant to note that Club C has all four types of such structures present with respect
to trust relations. In all clubs, there is support for reciprocity of trust relations,
and thus H1 relating to trust-generating mechanisms. It is no surprise that we
found the significant presence of mutual trust ties in all three clubs, especially
given that the reciprocal nature of trust is seen as fundamental to the definition
of trust itself. Indeed, reciprocity can almost be seen as the fundamental struc-
ture for trust relations, and the absence of such patterns from a network of trust
would indicate an incredible lack of trust within the network.
However, for transitive path closure (H2) —or the idea that a friend of a
friend is a friend—we find evidence of this in Clubs A and C, but not Club B.
We noted earlier that path closure (otherwise known as triadic closure, which
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 25
is about the formation of triads/triangles) is another fundamental element of
groups (Heider, 1958; Simmel, 1950), and a structure that has been shown to
be extremely important in social network (Burt & Knez, 1995; Granovetter,
1973; Robins et al., 2009). That it is not significantly present in Club B is
concerning, because as Granovetter (1985) notes, “Better than the statement
that someone is known to be reliable is information from a trusted informant
that he has dealt with that individual and found him so” (p. 490). A triad (in
the form of path closure4) essentially represents Granovetter’s sentiments
above regarding the referral of trust. The absence of such an effect indicates
that team members in Club B are reluctant to believe the recommendations of
their teammates regarding trust, indicating that while team members may be
willing to take risks in trusting at the level of the dyad (i.e., reciprocity), they
are unwilling to risk referrals, or such referrals are not being made. In any
case, team members are uncomfortable investing trust based on the views of
another, or offering their view of another. This suggests trust in the form of
path closure involves more risk than reciprocal trust, where judgments are
made by oneself and not through the investment of third parties.
It is almost a given, within the context of these sporting teams, that you
would trust your best players and also your more experienced colleagues
because their value to the team in their ability to help win games (the team’s
primary group outcome) is higher. These structures of trust seem to be funda-
mental along the lines of mutual trust ties, such that if players in a team did
not trust those high in performance or experience you would be worried—
such lack of trust would be conspicuous in its absence. Again, Club C meets
both of these expectations, with separate effects showing players investing
trust in those high in experience (H4a) and those high on performance (H4b).
Yet, once again indicating some issues, Club A players trust in experience,
but not performance, and Club B players trust in performance, but not experi-
ence. The implication is that for Club A high performers may be unreliable,
and that for Club B more experienced players may be seen as past it, rather
than older and wiser.
Finally, indicating that similarity breeds trust among the players, in all
three clubs players trusted others of similar levels of experience to them-
selves. This effect is over and above the receiver effect for experience, so that
beyond trusting highly experienced team members, team members also trust
other team members who are similar in experience. It is also over and above
effects of reciprocity and transitive closure. Given the correlation between
experience and age, this may not be so surprising. Yet curiously, this is not the
case for performance, and in this case it seems that team members trust up the
hierarchy toward better performing players, not across the hierarchy to those
similar to themselves. This may be because experience is a more easily
by guest on January 7, 2016sgr.sagepub.comDownloaded from
26 Small Group Research 45(1)
defined commodity than performance. However, experience is also some-
what less contested than performance, with players competing with one
another in their performances to be a part of the team (Messner, 1997).
In summary, our method of analysis (ERGM) has been used to differenti-
ate different structures of trust relations to unpack the subtle ways that trust
is afforded to teammates. It is only Club C that demonstrates the presence of
all four types of trust-generating structures of trust relations but also the
absence of trust-inhibiting structures of trust relations. This same club has the
highest overall team performance. As such, the qualitative comparison of the
quantitative network model findings for these structures of trust with team
performance does suggest an interesting association between these mecha-
nisms of trust and team performance.
Discussion
The current study is primarily concerned with the examination of a trust-
inhibiting (or competitive) structure of trust relations in teams, and its asso-
ciation with team performance. Trust-inhibiting structures were not found in
the club that had the highest overall team performance, but were found in
teams that performed more poorly. The inclusion of trust-generating struc-
tures of trust adds validity to our findings regarding the trust-inhibiting struc-
ture. All types of trust-generating structures were also found in the team with
the highest overall team performance, though in the other two teams these
trust-generating structures occurred to a lesser extent. The trust-generating
hypotheses were indeed nothing new and replicate well-established findings
regarding trust. The observed trust-generating and trust-inhibiting effects
align, and give credence to the suggestion that negative relations are indeed
powerful and are worth avoiding (Labianca & Brass, 2006).
The issue of interpersonal trust in teams is not merely an academic ques-
tion. Determining how and why some teams work effectively together while
others do not is of wider interest given that many teams, or groups more
generally, place significant importance on collective outcomes that depend
on the cooperation and trust of its members. Many teams institute mentor
programs for the purposes of team building and cohesion. If trust is based
more on experience then matching older team members with younger ones
would seem a sensible approach to mentoring. On the other hand, if trust is
(peer-referenced) performance related then it may well be worth using this
quality as a characteristic for matching team members. Indeed, there may
also be other qualities on which one is trusted, such as attitudes like identifi-
cation to the team. The statistical ERGM approach is able to uncover conse-
quential patterns of network trust ties and their association with personal-level
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 27
attributes, thereby giving insights into potential underlying social processes
that are not visible with standard statistical techniques. The ERGM approach
gives us the ability to look, one team at a time, at social mechanisms that
inhibit or generate trust in teams. This ability to postulate nuanced theory
about trust relations and test it empirically within the context of teams is an
important contribution of this article.
Jones and George (1998) discuss the utility of conceiving of trust as being
in one of three forms: distrust, conditional trust, and unconditional trust. It is
particularly unconditional trust that they suggest can “fundamentally change
the quality of the exchange relationship and convert a group into a team” (Jones
& George, 1998, p. 539). Where trust is expected but withheld is suggestive
that unconditional trust is not evident, particularly in cases where it is between
prominent and highly trusted members of the team. By conceptualizing trust in
network terms, and thinking of certain micro structures as describing “distinct
states or forms of the trust experience” (Jones & George, 1998, p. 537) we have
potentially shown here that distinct states may operate simultaneously within
the one team. In Club C we would argue that there are many example structures
of trust (H1, H2, H4, and H5) and an absence of distrust (H3). Together, these
effects could be interpreted as representing unconditional trust, and the value of
the current study shows that we are able to measure empirically structures of
trust that relate to this concept and evaluate teams on trust using a network
approach. There are of course other possible structures that trust can take, and
theoretical work needs to be done to articulate these and then examine trust
relations in teams empirically for their presence.
This article has not delved directly into values, attitudes, moods and emo-
tions, other psychological mechanisms involved with trust by including such
actor-level measures on these issues (Mayer et al., 1995). However, our
results do offer possibilities to understand trust within teams from a psycho-
logical perspective. Building on the work of Heider (1958), Krackhardt’s
(1992) seminal study on the strength of strong ties among managers clearly
shows the psychological strain that a person endures when he is pulled in dif-
fering directions from others with differing views. We have shown here that
it is possible to measure such tensions at a level that includes multiple indi-
viduals who might be in the same conflicted position by including a param-
eter in an ERGM to see if such structures occur more than we might expect.
So we are able to include measures of tension as parameters in a model and
test them statistically. The results of this study suggest that within the
observed networks there are multiple people in potentially tense positions
regarding trust relations.
When using an ERGM to examine social network data, the researcher is
made to consider the multiple and intersecting reasons why network ties (i.e.,
by guest on January 7, 2016sgr.sagepub.comDownloaded from
28 Small Group Research 45(1)
trust) occur. This multiple explanation approach sits very comfortably with
agreed conceptions of the multifaceted nature of trust (Mayer et al., 1995;
McEvily et al., 2003). As such, different structures of trust ties can relate to
different theoretical conceptions of trust, potentially accommodating a range
of theoretical perspectives simultaneously, and comparing them one against
the other, statistically. Such an approach allows for a complex analysis of
trust that does not take one particular process at a time, single it out, and
potentially overestimate its importance (De Jong & Elfring, 2010), but
instead forces an explanation to appear above and beyond the importance of
competing reasons. Within the ERGM framework, network ties are seen as
locally emergent and the agglomeration of these local patterns forms the
global structure of trust relations observed for each network/team. This is
both a methodological viewpoint and also a theoretical statement about net-
work tie formation (Robins & Lusher, 2013). As noted, there are at least two
broad types of local patterns. In one of these types, network effects occur
because of the presence of other network ties known as purely structural
network effects, or network self-organization (e.g., “you scratch my back, I’ll
scratch yours,” or reciprocity). Such ties occur regardless of their personal
qualities of the people involved—one tie comes about because of the pres-
ence of another tie; that is, one tie depends on the other tie. H1, H2, and H3
are such explanations of how trust ties occur, and give insights into the
dependencies of trust relations. Other network effects occur due to the indi-
vidual-level qualities of individuals and are called actor-relation effects (e.g.,
a team member may be popular because she is a high performer). H4 and H5
are examples of such actor-relation effects for trust. This study has shown
evidence that both of these types of reasons for trust to occur are important,
and that purely structural explanations do not wash away actor-relation
explanations, or vice versa. A focus only on endogenous process or only on
individual attributes is likely to insufficiently explain trust ties.
A further strength of the study is that ERGMs are appropriate and fine-
grained methods for investigating specific structures of trust relations among
team members. We have detailed specific ways in which trust is structured
within each of the three teams, including how trust relations align with indi-
vidual attributes. We reiterate that we have conducted separate analyses for
each club, in a similar vein to running separate regression analyses for each
club. However, it is also worth noting that in the current research we have
controlled for the structure of social ties and their dependencies in these mod-
els in a way that a regression cannot (because it assumes independent
observations).
There are of course certain limitations to this article. The observations
made are specific to the teams we have observed and are not generalizable
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 29
to other contexts and further investigation in other settings would be useful.
We do not make causal claims here, such that trust relations have impacted
on performance, for it could also be the case that team performance is gen-
erative of trust. Clearly longitudinal analysis is required to unpack such
issues. Furthermore, we do not have detailed information about the clubs
themselves. Further research is needed to better clarify how the proposed
network structures for trust relations are associated with team cohesion and
other metrics on which teams can be measured. And there are surely other
structures of trust relations that may be of consequence for teams, and these
need to be elaborated. In addition, the measurement of trust through a single
item might be improved on using a multi-item scale that more specifically
defines trust for the participants instead of leaving the definition up to the
participant to work out for themselves. We have not delineated instrumental
and expressive trust in our network measure. Certainly it may be informative
to do so in future research, and we see possibilities here with regard to using
a multivariate ERGM to unpack the structures related to each type of trust
network and how these two might align and diverge. The current research
also opens up other possibilities. For instance, just as experience and perfor-
mance have been analyzed with respect to their co-occurrence with interper-
sonal trust relations, it is also possible to investigate how individual-level
perceptions of trust co-occur with such structures. Specifically, in relation to
the link between trust and team performance, it would be possible to mea-
sure, at the level of the individual, team satisfaction or perceived task perfor-
mance (Costa, 2003), or team reflexivity and team effort (De Jong & Elfring,
2010), and how these may be related to structures of trust. This would allow
a further unpacking of the “mechanisms and processes which can reinforce
and help sustain team-based initiatives” (McHugh et al., 1997, p. 47), and
to develop a greater understanding of the organizing principles of trust
(Gambetta, 1988).
Conclusion
Many insights into trust in teams point to mechanisms that can be seen as
trust generating, but less attention is directed toward trust-inhibiting struc-
tures that may undermine trust within teams. The current research argued that
trust-inhibiting (or competitive) structures of trust may indeed be deleterious
to team performance. We hypothesized one such trust-inhibiting structure,
which entails highly trusted team members not trusting one another, and
examined three teams for its presence. An analysis of trust networks using
ERGMs indicates that this trust-inhibiting structure of trust was absent in the
most successful team but present in the two less successful teams. Furthermore,
by guest on January 7, 2016sgr.sagepub.comDownloaded from
30 Small Group Research 45(1)
that well-accepted trust-generating mechanisms of trust (e.g., reciprocity)
were highly present in the most successful team, but less so in the less suc-
cessful teams. We qualify our findings by noting that our analyses of the
mechanisms of trust were quantitatively derived, and then compared qualita-
tively with team performance. Nonetheless, we do suggest there is an asso-
ciation between the presence of trust-generating mechanisms of trust, the
absence of trust-inhibiting mechanisms of trust, and team performance. We
suggest that trust-inhibiting structures offer an importance to pursue a line of
inquiry through which to understand tensions regarding trust in teams because
these may have a significant relationship with team performance. Further
theorization on other trust-inhibiting structures of trust relations and related
empirical work is likely to shed further light on the connection of trust in
teams and team performance.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research,
authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research,
authorship, and/or publication of this article: Funding by the Australian Football
League (AFL) Research and Development Board that assisted this research is grate-
fully acknowledged.
Notes
1. One player was transferred from one club to another. He is included in the social
network analyses twice as there are separate models for each club, but in the
individual-level analyses his second responses have been removed to avoid vio-
lation of assumptions.
2. Homophily is implemented as the negative of absolute difference for continuous
variables. The valences of results have been reversed so that positive scores are
indicative of homophily.
3. The implication is that we are modeling the rest of the trust network conditional
on the trust nominations by individuals who were not prepared to differentiate
in choices of their fellow team members. This leaves open whether the indi-
viduals concerned did actually trust everybody, or instead wished to impress the
researchers with their high levels of trusting. The decision not to fix the nomina-
tions by these players rather than exclude them enables us to retain important
information of how these fixed ties may participate in effects that explain the
occurrence of other ties in the networks.
4. There are indeed other forms of triads that are important—see Holland and
Leinhardt (1970).
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 31
References
Arrow, K. J. (1974). The limits of organization. New York, NY: W. W. Norton.
Axelrod, R. (1984). The evolution of cooperation. New York, NY: Basic Books.
Becerra, M., & Gupta, A. K. (2003). Perceived trustworthiness within the organiza-
tion: The moderating impact of communication frequency on trustor and trustee
effects. Organization Science, 14, 32-44. doi:10.1287/orsc.14.1.32.12815
Blau, P. M. (1964). Exchange and power in social life. New York, NY: Wiley.
Burt, R. S. (2001). Bandwidth and echo: Trust, information, and gossip in social net-
works. In J. E. Rauch & A. Casella (Eds.), Networks and markets (pp. 30-74).
New York, NY: Russell Sage Foundation.
Burt, R. S. (2005). Brokerage and closure. Oxford, UK: Oxford University Press.
Burt, R. S., & Knez, M. (1995). Kinds of 3rd-party effects on trust. Rationality and
Society, 7, 255-292. doi:10.1177/104346396008001005
Buskens, V. (1998). The social structure of trust. Social Networks, 20, 265-289.
Buskens, V., Raub, W., & van der Veer, J. (2010). Trust in triads: An experimental
study. Social Networks, 32, 301-312. doi:10.1016/j.socnet.2010.05.001
Cartwright, D., & Harary, F. (1956). Structural balance: A generalization of Heider’s
theory. Psychological Review, 63, 277-293.
Contractor, N. S., Wasserman, S., & Faust, K. (2006). Testing multi-theoretical mul-
tilevel hypotheses about organizational networks: An analytic framework and
empirical example. Academy of Management Review, 31, 681-703.
Costa, A. C. (2003). Work team trust and effectiveness. Personnel Review, 32,
605-622. doi:10.1108/00483480310488360
Costa, A. C., Roe, R. A., & Taillieu, T. (2001). Trust within teams: The relation
with performance effectiveness. European Journal of Work and Organizational
Psychology, 10, 225-244. doi:10.1080/13594320143000654
Daraganova, G., & Robins, G. (2013). Autologistic actor attribute models. In D.
Lusher, J. Koskinen, & G. Robins (Eds.), Exponential random graph models for
social networks: Theory, methods and applications (pp. 102-114). New York,
NY: Cambridge University Press.
Davis, J. H., Laughlin, P. R., & Komorita, S. S. (1976). Social psychology of small
groups cooperative and mixed-motive interaction. Annual Review of Psychology,
27, 501-541. doi:10.1146/annurev.ps.27.020176.002441
De Jong, B. A., & Elfring, T. (2010). How does trust affect the performance of ongo-
ing teams? The mediating role of reflexivity, monitoring, and effort. Academy of
Management Journal, 53, 535-549. doi:10.5465/amj.2010.51468649
Dirks, K. T., & Ferrin, D. L. (2002). Trust in leadership: Meta-analytic findings
and implications for research and practice. Journal of Applied Psychology, 87,
611-628. doi:10.1037/0021-9010.87.4.611
Earle, T. C., & Siegrist, M. (2006). Morality information, performance informa-
tion, and the distinction between trust and confidence. Journal of Applied Social
Psychology, 36, 383-416. doi:10.1111/j.0021-9029.2006.00012.x
Ellwardt, L., Labianca, G., & Wittek, R. (2012). Who are the objects of positive and
negative gossip at work? A social network perspective on workplace gossip.
Social Networks, 34, 193-205. doi:10.1016/j.socnet.2011.11.003
by guest on January 7, 2016sgr.sagepub.comDownloaded from
32 Small Group Research 45(1)
Emirbayer, M., & Goodwin, J. (1994). Network analysis, culture, and the problem of
agency. American Journal of Sociology, 99, 1411-1454. doi:10.2307/2782580
Fehr, E. (2009). On the economics and biology of trust. Journal of the European
Economic Association, 7, 235-266. doi:10.1162/JEEA.2009.7.2-3.235
Ferrin, D. L., Dirks, K. T., & Shah, P. P. (2006). Direct and indirect effects of third-
party relationships on interpersonal trust. Journal of Applied Psychology, 91,
870-883. doi:0.1037/0021-9010.91.4.870
Frank, O., & Strauss, D. (1986). Markov graphs. Journal of the American Statistical
Association, 81, 832-842. doi:10.1080/01621459.1986.10478342
Freeman, L. C. (1977). Set of measures of centrality based on betweenness.
Sociometry, 40, 35-41.
Gambetta, D. (1988). Can we trust trust? In D. Gambetta (Ed.), Trust: Making and
breaking cooperative relations (pp. 213-237). New York, NY: Blackwell.
Gorgenyi, I. (1998, Summer). Hunting territory: The structure of team life–Part 1.
Sports Coach, 18-21.
Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology,
78, 1360-1380.
Granovetter, M. (1985). Economic action and social structure: The problem of embed-
dedness. American Journal of Sociology, 91, 481-510.
Gulati, R. (1995). Does familiarity breed trust: The implications of repeated ties for
contractual choice in alliances. Academy of Management Journal, 38, 85-112.
doi:10.2307/256729
Hedström, P., & Swedberg, R. (1998). Social mechanisms: An analytical approach to
social theory. Cambridge, UK: Cambridge University Press.
Heider, F. (1958). The psychology of interpersonal relations. New York, NY: Wiley.
Hogg, M. A., Terry, D. J., & White, K. M. (1995). A tale of two theories: A criti-
cal comparison of identity theory with social identity theory. Social Psychology
Quarterly, 58, 255-269.
Holland, P. W., & Leinhardt, S. (1970). A method for detecting structure in sociomet-
ric data. American Journal of Sociology, 76, 492-513.
Huitsing, G., van Duijn, M. A. J., Snijders, T. A. B., Wang, P., Sainio, M., Salmivalli,
C., & Veenstra, R. (2012). Univariate and multivariate models of positive and
negative networks: Liking, disliking, and bully–victim relationships. Social
Networks, 34, 645-657. doi:http://dx.doi.org/10.1016/j.socnet.2012.08.001
Jones, G. R., & George, J. M. (1998). The experience and evolution of trust:
Implications for cooperation and teamwork. Academy of Management Review,
23, 531-546. doi:10.2307/259293
Kipnis, D. (1996). Trust and technology. In T. R. Tyler & R. M. Kramer (Eds.), Trust
in organizations: Frontiers of theory and research (pp. 39-50). Thousand Oaks,
CA: Sage.
Komorita, S. S., & Parks, C. D. (1995). Interpersonal relations mixed-motive inter-
action. Annual Review of Psychology, 46, 183-207. doi:10.1146/annurev.
ps.46.020195.001151
Koskinen, J., & Snijders, T. A. B. (2013). Simulation, estimation and goodness of fit.
In D. Lusher, J. Koskinen, & G. Robins (Eds.), Exponential random graph mod-
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 33
els for social networks: Theory, methods and applications (pp. 141-166). New
York, NY: Cambridge University Press.
Krackhardt, D. (1987). Cognitive social structures. Social Networks, 9, 109-134.
doi:10.1016/0378-8733(87)90009-8
Krackhardt, D. (1992). The strength of strong ties: The importance of philos in orga-
nizations. In N. Nohria & R. G. Eccles (Eds.), Networks and organizations:
Structure, form and action (pp. 216-239). Boston, MA: Harvard University Press.
Krackhardt, D. (1996). Comment on Burt and Knez’s third-party effects on trust.
Rationality and Society, 8, 111-116. doi:10.1177/104346396008001005
Labianca, G., & Brass, D. J. (2006). Exploring the social ledger: Negative relation-
ships and negative asymmetry in social networks in organizations. Academy of
Management Review, 31, 596-614. doi:10.5465/AMR.2006.21318920
Labianca, G., Brass, D. J., & Gray, B. (1998). Social networks and perceptions of
intergroup conflict: The role of negative relationships and third parties. Academy
of Management Journal, 41, 55-67. doi:10.2307/256897
Lincoln, J. R., & Miller, J. (1979). Work and friendship ties in organizations: A com-
parative analysis of relational networks. Administrative Science Quarterly, 24,
181-199.
Lubbers, M. J. (2003). Group composition and network structure in school classes:
A multilevel application of the p* model. Social Networks, 25, 309-332.
doi:10.1016/S0378-8733(03)00013-3
Lusher, D., & Ackland, R. (2011). A relational hyperlink analysis of an online social
movement. Journal of Social Structure, 12. Retrieved from http://www.cmu.edu/
joss/content/articles/volume12/Lusher/
Lusher, D., Koskinen, J., & Robins, G. L. (Eds.). (2013). Exponential random graph
models for social networks: Theory, methods and applications. New York, NY:
Cambridge University Press.
Lusher, D., Robins, G. L., & Kremer, P. (2005). Masculine behaviour and social
networks in team structures. Melbourne, Australia: A Report for the Australian
Football League.
Lusher, D., Robins, G. L., Pattison, P. E., & Lomi, A. (2012). “Trust me”: Differences
in expressed and perceived trust relations in an organization. Social Networks,
34, 410-424. doi:http://dx.doi.org/10.1016/j.socnet.2012.01.004
Mayer, R. C., Davis, J. H., & Schoorman, F. D. (1995). An integrative model of organi-
zational trust. Academy of Management Review, 20, 709-734. doi:10.2307/258792
McAllister, D. J. (1995). Affect- and cognition-based trust as foundations for interper-
sonal cooperation in organizations. Academy of Management Journal, 38, 24-59.
doi:10.2307/256727
McEvily, B., Perrone, V., & Zaheer, A. (2003). Trust as an organizing principle.
Organization Science, 14, 91-103. doi:10.1287/orsc.14.1.91.12814
McEvily, B., & Tortoriello, M. (2011). Measuring trust in organisational research:
Review and recommendations. Journal of Trust Research, 1, 23-63. doi:10.1080/
21515581.2011.552424
McHugh, P. P., Niehaus, R., & Swiercz, P. M. (1997). Team-based work systems: Lessons
from the industrial relations literature. Human Resource Planning, 20, 44-47.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
34 Small Group Research 45(1)
McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily
in social networks. Annual Review of Sociology, 27, 415-444. doi:10.1146/
annurev.soc.27.1.415
Messner, M. A. (1997). Like family: Power, intimacy, and sexuality in male athletes’
friendships. In M. M. Gergen & S. N. Davis (Eds.), Toward a new psychology of
gender (pp. 341-360). Florence, KY: Taylor & Francis.
Pattison, P. E., & Robins, G. L. (2002). Neighbourhood-based models for social net-
works. Sociological Methodology, 32, 301-337. doi:10.1111/1467-9531.00119
Pattison, P. E., & Wasserman, S. (1999). Logit models and logistic regressions for
social networks: II. Multivariate relations. British Journal of Mathematical and
Statistical Psychology, 52, 169-193. doi:10.1348/000711099159053
Rank, O., Robins, G. L., & Pattison, P. E. (2010). Structural logic of network organi-
zations. Organizational Science, 21, 745-764. doi:10.1287/orsc.1090.0450
Riesman, D. (1953). The lonely crowd: A study of the changing American character.
New Haven, CT: Yale University Press.
Robins, G., & Daraganova, G. (2013). Social selection, dyadic covariates, and geo-
spatial effects. In D. Lusher, J. Koskinen, & G. Robins (Eds.), Exponential ran-
dom graph models for social networks: Theory, methods and applications (pp.
91-101). New York, NY: Cambridge University Press.
Robins, G., Elliott, P., & Pattison, P. (2001). Network models for social selection
processes. Social Networks, 23, 1-30. doi:10.1016/S0378-8733(01)00029-6
Robins, G., & Lusher, D. (2013). What are exponential random graph models? In D.
Lusher, J. Koskinen, & G. Robins (Eds.), Exponential random graph models for
social networks: Theory, methods and applications (pp. 9-15). New York, NY:
Cambridge University Press.
Robins, G., Pattison, P., & Elliott, P. (2001). Network models for social influence
processes. Psychometrika, 66, 161-189. doi:10.1007/BF02294834
Robins, G., Pattison, P., Kalish, Y., & Lusher, D. (2007). An introduction to exponen-
tial random graph (p*) models for social networks. Social Networks, 29, 173-191.
doi:10.1016/j.socnet.2006.08.002
Robins, G., Pattison, P., & Wang, P. (2009). Closure, connectivity and degree dis-
tributions: Exponential random graph (p*) models for directed social networks.
Social Networks, 31, 105-117. doi:10.1016/j.socnet.2008.10.006
Rousseau, D. M., Sitkin, S. B., Burt, R. S., & Camerer, C. (1998). Not so different
after all: A cross-discipline view of trust. Academy of Management Review, 23,
393-404. doi:10.5465/AMR.1998.926617
Sampson, S. A. (1968). A novitiate in a period of change: An experimental and case
study of relationships (Unpublished doctoral dissertation). Cornell University,
New York, NY.
Schelling, T. C. (1960). The strategy of conflict. Cambridge, MA: Harvard University
Press.
Sheppard, B. H., & Tuchinsky, M. (1996). Micro-OB and the network organization.
In T. R. Tyler & R. M. Kramer (Eds.), Trust in organizations: Frontiers of theory
and research (pp. 140-165). Thousand Oaks, CA: Sage.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
Lusher et al. 35
Simmel, G. (1950). The sociology of Georg Simmel. New York, NY: Free Press.
Snijders, T. A. B. (2011). Statistical models for social networks. Annual Review of
Sociology, 37, 131-153. doi:10.1146/annurev.soc.012809.102709
Snijders, T. A. B., Pattison, P. E., Robins, G. L., & Handcock, M. (2006). New speci-
fications for exponential random graph models. Sociological Methodology, 36,
99-153. doi:10.1111/j.1467-9531.2006.00176.x
Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W.
G. Austin & S. Worchel (Eds.), The social psychology of intergroup relations
(pp. 33-47). Monterey, CA: Brooks.
Tsai, W. P. (2002). Social structure of “coopetition” within a multiunit organiza-
tion: Coordination, competition, and intraorganizational knowledge sharing.
Organization Science, 13, 179-190. doi:10.1287/orsc.13.2.179.536
Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987).
Rediscovering the social group: A self-categorization theory. Oxford, UK: Blackwell.
Tyler, T. R., & Kramer, R. M. (1996). Whither trust? In T. R. Tyler & R. M. Kramer
(Eds.), Trust in organizations: Frontiers of theory and research (pp. 1-15).
Thousand Oaks, CA: Sage.
Uzzi, B. (1997). Social structure and competition in interfirm networks: The
paradox of embeddedness. Administrative Science Quarterly, 42, 35-67.
doi:10.2307/2393808
van de Bunt, G. G., Wittek, R. P. M., & de Klepper, M. C. (2005). The evolution
of intra-organizational trust networks. The case of a German paper factory: An
empirical test of six trust mechanisms. International Sociology, 20, 339-369.
doi:10.1177/0268580905055480
Wang, P., Robins, G. L., & Pattison, P. E. (2009). PNet: A program for the simula-
tion and estimation of exponential random graph models. Melbourne, Australia:
University of Melbourne.
Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applica-
tions. Cambridge, UK: Cambridge University Press.
Wasserman, S., & Pattison, P. E. (1996). Logit models and logistic regressions for
social networks. 1. An introduction to Markov graphs and p. Psychometrika, 61,
401-425. doi:10.1007/BF02294547
White, H. C., Boorman, S. A., & Breiger, R. L. (1976). Social structure from multiple
networks. I. Blockmodels of roles and positions. American Journal of Sociology,
81, 730-780. doi:10.2307/2777596
Author Biographies
Dean Lusher is a Senior Research Fellow in the Swinburne Business School at
Swinburne University of Technology, Australia. His research interests are in the theory
and application of statistical models for social networks, in regard to organizations,
innovation and organizational culture, and issues of social inclusion.
Peter Kremer is a Senior Lecturer in Sport Behaviour in the School of Exercise and
Nutrition Sciences at Deakin University, Australia. His research interests focus on
by guest on January 7, 2016sgr.sagepub.comDownloaded from
36 Small Group Research 45(1)
behavioral and psychological factors associated with participating in sport and exer-
cise and the role of sport and exercise in promoting positive mental health outcomes.
Garry Robins is a mathematical psychologist in the School of Psychological
Sciences, University of Melbourne, Australia, whose research focuses on statistical
models for social networks. He has been involved in a wide variety of empirical stud-
ies dealing with social networks.
by guest on January 7, 2016sgr.sagepub.comDownloaded from
