 

 

 

 

“1‘ Michigan State
on”; University

 

 

This is to certify that the
thesis entitled

SOCIAL INTERACTION AND COLLECTIVE EFFICACY
DISPERSION: A SOCIAL NETWORK ANALYSIS

presented by

Graig Michael Chow

has been accepted towards fulﬁllment
of the requirements for the

Ph.D. degree in Kinesiology

 

 

”@034 :2. “Mia

Major Professor’s Sigiiature
G I

Date

 

MSU is an Afﬁrmative Action/Equal Opportunity Employer

.- ‘----l-0-o--‘-

.--I-l-I-t-I-l-I-n-A-l-l-O-t-O-I-I-I-O-O-I-t-I-I-I‘0‘.-.—

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 ICIProj/AccaPreleIRCIDateDuehdd

SOCIAL INTERACTION AND COLLECTIVE EFFICACY DISPERSION:
A SOCIAL NETWORK ANALYSIS

By

Graig Michael Chow

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
KINESIOLOGY

2009

ABSTRACT

SOCIAL INTERACTION AND COLLECTIVE EFFICACY DISPERSION:
A SOCIAL NETWORK ANALYSIS

By
Graig Michael Chow
The purpose of this study was to investigate the inﬂuence of social interaction on
collective efﬁcacy dispersion, and to examine the moderating inﬂuence of dispersion on

team-related outcomes. The sample consisted of 46 intercollegiate women’s softball

 

teams. A longitudinal design was employed with two time points: beginning of the
season and middle of the season. At Time 1, participants completed demographic,
collective efﬁcacy, and team commitment questionnaires. At Time 2, participants
completed demographic, collective efﬁcacy, team commitment, and sociometric
questionnaires. Based on the sociometric data, group-level social network metrics were
calculated and cohesive subgroups were identiﬁed within teams using Klique Finder
(Frank, 1995). A series of hypotheses were made regarding the inﬂuence of social
network variables on collective efﬁcacy dispersion and the moderating inﬂuence of
collective efﬁcacy dispersion on team performance and commitment. The majority of
teams in the study did not demonstrate evidence of subgroups with only 33% having
communication subgroups and 28% having friendship subgroups. Membership in
cohesive Subgroups was related to number of years on the team and to a lesser degree,
starting status and collective efﬁcacy beliefs. Results demonstrated that teams with
communication subgroups had more collective efﬁcacy dispersion than teams without
communication subgroups. However, there was a lack of support regarding the

inﬂuence of subgroup’s prior mean levels of collective efﬁcacy on collective efﬁcacy

 

beliefs at Time 2, indicating that individuals’ collective efﬁcacy beliefs were not
inﬂuenced by members of their subgroup. Centralization of the communication network
emerged as a positive predictor of collective efﬁcacy dispersion. Teams with more
centralized communication ties had higher levels of dispersion than team with less
centralized communication ties. Limited support was found for collective efﬁcacy
dispersion as a predictor and moderator. Collective efﬁcacy dispersion at the middle of
the season was negatively related to run differential at the end of the season, and
collective efficacy dispersion moderated the relationship between aggregated collective

efﬁcacy and run differential.

DEDICATION

To my mom, Lissa, for always believing in me and to Leena
for all of her love and many sacriﬁces.

iv

ACKNOWLEDGMENTS

A special thank you to my advisor and dissertation committee chair Dr. Deborah
Feltz. Thank you to my dissertation committee members, Drs. Ken Frank, Dan Gould,
and Marty Ewing for providing guidance and direction on this project.

My appreciation goes to the athletes and coaches who volunteered their time and

effort to make this study possible.

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................... viii

CHAPTER 1

INTRODUCTION
Nature of the Problem ............................................................................................. 1
Statement of the Problem ........................................................................................ 6
Contextual Factors. ............................................................................................... 12
Purpose of the Study ............................................................................................. 14
Hypotheses ............................................................................................................ 14
Research Questions ............................................................................................... 15
Delirnitations ......................................................................................................... l 5
Deﬁnitions ............................................................................................................. 16

CHAPTER 2

REVIEW OF LITERATURE
Dispersion as a Group—Level Construct ................................................................ 18
Social Interaction and Shared Beliefs ................................................................... 22
Social Networks .................................................................................................... 26
Social Network Measures ..................................................................................... 29
Subgroups. ............................................................................................................ 37
Collective Efﬁcacy Dispersion as a Moderator .................................................... 40
Summary ............................................................................................................... 43

CHAPTER 3

METHOD
Participants ............................................................................................................ 45
Procedure .............................................................................................................. 46
Measures ............................................................................................................... 47
Data Analyses ....................................................................................................... 54

CHAPTER 4

RESULTS
Descriptives ........................................................................................................... 58
Social Network Variables and Dispersion ............................................................ 60
Multilevel Model of Collective Efﬁcacy Based on Subgroups ............................ 61
Moderating Inﬂuence of Dispersion on Team Performance ................................. 63
Moderating Inﬂuence of Dispersion on Team Commitment ................................ 64
Summary of Hypotheses and Results ................................................................... 64

vi

CHAPTER 5

DISCUSSION
The Emergence of Collective Efﬁcacy Dispersion ............................................... 66
Collective Efﬁcacy Dispersion as an Independent and Moderating Variable ...... 70
Implications ........................................................................................................... 73
Strengths and Limitations ..................................................................................... 74
Future Research Directions. .................................................................................. 78
Conclusions ............................................................................... _ ............................ 82

APPENDICES
Appendix A. Athlete Demographic Questionnaire (Time 1) ................................ 83
Appendix B. Athlete Demographic Questionnaire (Time 2) ................................ 84
Appendix C. Coach Demographic Questionnaire ................................................. 85
Appendix D. Collective Efﬁcacy Questionnaire ................................................... 86
Appendix E. Team Commitment Questionnaire ................................................... 87
Appendix F. Social Interaction Questionnaire ...................................................... 88

REFERENCES ................................................................................................................ 101

vii

LIST OF TABLES

Table 1. Athlete Demographics ................................................................................. 89
Table 2. Coach Demographics ................................................................................... 90
Table 3. Means and Standard Deviations for Team Variables. ................................. 91
Table 4. Bivariate Correlations for Team Variables. . . . .. .......................................... 92
Table 5. Dispersion Regressed on Network Variables .............................................. 93
Table 6. Multilevel Model Based on Communication Subgroups. . . . .. ..................... 94
Table 7. Multilevel Model Based on Friendship Subgroups. .................................. 95

Table 8. Dispersion as a Moderator of Collective Efﬁcacy-Team Performance
Relationship ............................................................................................................... 96

Table 9. Dispersion as a Moderator of Collective Efﬁcacy-Team Commitment
Relationship ............................................................................................................... 97

viii

LIST OF FIGURES

Figure 1. Example Data of Dispersion ....................................................................... 98
Figure 2. Illustration of Network Density... .. . . .. ...................................................... 99
Figure 3. Illustration of Network Centralization. .. .................................................. 100

 

ix

CHAPTER 1
INTRODUCTION
Nature of the Problem

The nature of intact teams involves members performing together in dynamic
and complex environments where individuals develop beliefs about the team that may
be similar or different ﬁ'om their teammates. While shared experiences provide a
common frame of reference that draws team members’ beliefs closer together, the
importance given to the structure of group interaction is often overlooked. Depending
on the structure of interpersonal interaction, group members’ beliefs may converge or
diverge over time creating dispersion at the team level, which has implications for team
functioning, performance, and affective states.

The purpose of this dissertation was to examine the emergence of collective
efﬁcacy dispersion and to investigate the team-related consequences of collective
efﬁcacy dispersion. This chapter begins with a brief review of the collective efﬁcacy
research, discusses the concept of collective efﬁcacy dispersion, introduces social
network analysis as a tool for examining interpersonal interaction and social inﬂuence,
and concludes with a discussion of the tenets and limitations.

Collective efﬁcacy is “a group’s shared belief in their conjoint capabilities to
organize and execute the courses of action required to produce given levels of
attainments” (Bandura, 1997, p. 476). Theorists have asserted that collective efﬁcacy is
an important determinant of team performance (Bandura, 1997; Gist, 1987; Lindsley,
Brass, & Thomas, 1995; Zaccaro, Blair, Peterson, & Zazanis, 1995). Judgments of

collective efﬁcacy inﬂuence the tasks that teams select to engage in, the amount of

effort members put forth together, the degree to which members persevere and remain
task-oriented when the team is struggling, and the level of resiliency to bounce back
from wrenching defeats (Bandura, 1997). Empirical research in sport and organizational
psychology has provided support for the performance-oriented consequences of
collective efﬁcacy beliefs. Teams with a stronger sense of collective efﬁcacy
outperform teams with a weaker sense of collective efﬁcacy across various group
environments including sport (Feltz & Lirgg, 1998), executive training (Gibson, Randel,
& Barley, 2000), health (Gibson, 1999), ﬁnancial organizations (Campion, Medsker, &
Higgs, 1993), academics (Parker, 1994), and military units (Chen, Webber, Bliese,
Mathieu, Payne, Born, & Zaccaro, 2002). A meta-analysis of 53 studies found that
collective efﬁcacy was positively related to team performance with a corrected mean
correlation of .41 (Gully, Incalcaterra, J oshi, & Beaubien, 2002).

How teams develop their sense of collective efﬁcacy is somewhat similar to how
self-efﬁcacy beliefs are formed, that is through sources of efﬁcacy information.
Because perceptions of collective efﬁcacy are rooted in perceptions of self-efﬁcacy, the
sources of efﬁcacy beliefs are hypothesized to be isomorphic across levels (Bandura,
1997; Lindsley et a1., 1995). Based on the tenets of self-efﬁcacy theory (Bandura, 1977,
1997), researchers have investigated sources of collective efﬁcacy by extending the
antecedents of personal efﬁcacy to the group level. Consistent with self-efficacy theory,
studies have found that collective efﬁcacy beliefs are inﬂuenced by previous team I
performance (Feltz & Lirgg, 1998), vicarious experience (Chase, Feltz, & Lirgg, 2003),
verbal persuasion (Hodges & Carron, 1992), and physiological/emotional states

(Greenlees, Nunn, Graydon, & Maynard, 1999). However, collective efﬁcacy is not a

mere extension of self-efﬁcacy theory. While there are similarities between the
antecedents of self- and collective efﬁcacy beliefs, the assumption of homology may be
inaccurate (Chen & Bliese, 2002; Chen et al., 2002). Consequently, researchers have
identiﬁed antecedents that are unique to collective efﬁcacy such as group cohesion
(Paskevich, Brawley, Dorsch, & Widmeyer, 1999), leadership (Watson, Chemers, &
Preiser, 2001), and motivational climate (Magyar, Feltz, & Simpson, 2004).

Although collective efﬁcacy is conceptualized as a group level attribute, the
construct is measured at the individual level. A team level estimate is derived by
aggregating group members’ individual perceptions of collective efﬁcacy. In order to
justify aggregating individual perceptions of collective efﬁcacy to form the higher level
construct (i.e., shared collective efﬁcacy), researchers calculate the degree of within-
team variability in members’ responses. If a relatively high level of within-team
agreement is demonstrated, collective efﬁcacy is considered a shared group level
phenomenon which provides justiﬁcation for aggregating the lower level responses.
However, if members’ perceptions of collective efﬁcacy demonstrate insufﬁcient levels
of consensus, some researchers contend that the construct is operating at the individual
level (Gibson, 1999; Gully et al., 2002) or that there is no distinguishable collective
efﬁcacy for the team as a whole (Bandura, 1997). In contrast, Moritz and Watson
(1998) argue that focusing only on shared collective efﬁcacy beliefs disregards within-
team variability, which may provide insight into team functioning and performance. For
instance, a team with relatively high collective efﬁcacy (i.e., group mean) may be
comprised of members who hold different beliefs pertaining to the team’s capabilities.

These within-team differences in collective efﬁcacy may contribute to conﬂict among

members, a lack of cohesiveness, and varying amounts of effort exhibited by certain
members.

The absence of perceptual consensus does not imply that a reasonable estimate
of collective efﬁcacy does not exist (Myers & Feltz, 2007), just as a substantial degree
of consensus within teams does not imply that there is a total lack of within-team
variability. In fact, complete unanimity of group members’ collective efﬁcacy beliefs is
rare (Bandura, 1997). Individuals occupying different positions or serving different
functions within the same social system may differ in their sense of collective
competence (Bandura, 1993). For instance, a veteran player who has experienced
constant failure throughout her tenure may possess very different beliefs about the
team’s functioning than a ﬁrst-year player who has had few opportunities to perform
collectively with her teammates. When individual perceptions of collective efﬁcacy are
aggregated to the group level, these differences are discarded and within-team
variability is treated as constant across all teams that achieve a sufﬁcient level of
consensus. However, teams with similar aggregated collective efﬁcacy beliefs may have
different levels of within-group variation (see Figure 1). Furthermore, differences in
group members’ collective efficacy beliefs at the individual level create dispersion at
the team level, and current theory does not address how these differences affect
motivational and behavioral consequences.

A dispersion theory moves beyond the dichotomous view of perceptual
consensus and conceptualizes within-team variability as a focal unit level variable
(Brown, Kozlowski, & Hattrup, 1996; Chan, 1998). That is, rather than being used as a

statistical prerequisite for aggregation, dispersion theories posit that the variance lost

 

during aggregation procedures may be of substantive interest. Thus, the degreeof
within-group variation can be investigated as an independent variable, moderating
variable, or dependent variable (Brown et al., 1996; Chan, 1998). Collective efﬁcacy
represents a useful theoretical construct in which to examine dispersion as a focal
variable because it is characterized by a representative value of members’ perceptions
and the variability around that central belief (Bandura, 1997). A dispersion theory of
collective efﬁcacy assumes that within-team variation in group members’ beliefs is as
meaningful as the central tendency around that belief.

While conceptual arguments for investigating collective efﬁcacy dispersion as a
focal variable have been presented in the literature (Chan, 1998; DeRue, Hollenbeck,
Ilgen, & Feltz, in press; Moritz & Watson, 1998), only one empirical study has
investigated this topic. Arthur, Bell, and Edwards (2007) examined the degree of
within-team variability in efﬁcacy beliefs with dyadic teams performing a perceptual-
motor skill task that involved information processing and was highly interdependent.
The purpose of this study was to examine the moderating inﬂuence of collective
efficacy dispersion on the aggregated collective efﬁcacy-team performance relationship.
They found that the relationship between aggregated collective efﬁcacy and team
performance was stronger for dyads with higher levels of perceptual consensus than for
dyads with lower levels of consensus. Further, dyads with shared collective efficacy
beliefs outperformed dyads with dispersed collective efﬁcacy beliefs, regardless of the
magnitude of collective efﬁcacy. This study represents an important shift in the
conceptualization of collective efﬁcacy and provides preliminary support for the

usefulness of collective efﬁcacy dispersion as a focal variable in team research.

Statement of the Problem

The overemphasis of collective efﬁcacy as a shared belief has contributed to a
lack of understanding regarding the emergence of collective efﬁcacy. Thus, little is
known about why some teams develop more interrelated beliefs, while other teams
develop more dispersed beliefs. Theorists have assumed that common experiences such
as outcomes of team performance are sufﬁcient for creating shared collective efﬁcacy
beliefs. This perspective is limited however, because it assumes that teams that have
received performance feedback always achieve the same level of perceptual consensus,
and it fails to adequately explain why teams with similar experiences attain different
levels of dispersion. Another perspective suggests that substantial social interaction is a
necessary precondition for the development of shared collective efﬁcacy beliefs (Jung
& Sosik, 2003), though the role of group interaction in the formation of collective
efﬁcacy has largely been ignored. Gibson (1999) proposed that collective efﬁcacy
develops “as group members collectively acquire, store, manipulate, and exchange
information about each other and about their task, context, process, and prior
performance” (p. 138). Consistent with this observation, the current study examines
social interaction as the foundation for consensus formation within teams. High levels
of social interaction foster similarity among group members and promote the emergence
of shared beliefs (Klein, Conn, Smith, & Sorra, 2001; Gonzalez-Roma, Peiro, &
Tordera, 2002), whereas a lack of group interaction constrains information ﬂow and
produces dispersed beliefs.

A social network perspective is concerned with the structure or pattern of social

relations between individuals and assumes that group interactions will explain outcomes

at multiple levels of analysis beyond the attributes of individuals (e. g., experience,
number of years on the team) or organizational wntext (Wasserman & Faust, 1994).
Housed within social network analysis, network theories of social inﬂuence suggest that
social interaction provides a conduit along which social inﬂuence or contagion ﬂows.
Lindsley et a1. (1995) observed the social effects of collective efﬁcacy and have argued
that a shared sense of collective efﬁcacy emerges as individuals interact and compare
themselves with others to test and conﬁrm their own perceptions. According to network
theorists, social inﬂuence is a function of two independent processes — communication
exchanges and interpersonal comparison based on friendship. Communication ties
reﬂect instrumental interactions or task-related exchanges where information provides a
mechanism of social inﬂuence. Information exchanges are likely to produce efﬁcacy
contagion as group members share and discuss aspects of team functioning such as
strengths and weaknesses of the team, previous performances and goals. Social
comparison reﬂects expressive interaction or affective relationships where similarity
and closeness provide a pathway of inﬂuence. Comparison is similar to social modeling
and is likely to produce contagion when individuals modify their collective efﬁcacy
beliefs to be consistent with others with whom they are affectively close. The social
effects of communication and social comparison are captured by verbal persuasion and
vicarious experience, respectively, in efficacy theory (Bandura, 1997).

An advantage of social network analysis is that group level metrics can be
calculated that describe different features of a network’s interaction structure. Perhaps
the most common index of network structure used to compare different networks is

structural density (Scott, 2000). Density is a measure of social proximity and refers to

the proportion of ties in a network relative to the total number of possible ties in a
network (Scott, 2000; Wasserman & Faust, 1994). Networks can be characterized as
either dense or sparse depending on the density metric. Figure 2 illustrates a dense
network on the left and a sparse network on the right. Communication network density
reﬂects the extent to which actors share task-related information, while friendship
network density reﬂects the extent to which actors are involved in personal
relationships. Teams with densely conﬁgured instrumental and expressive ties
outperform and are more committed to staying together than teams with sparsely
conﬁgured ties (Balkundi & Harrison, 2006). In addition, density facilitates social
diffusion and comparison processes which promote the emergence of shared beliefs in
teams (Zohar & Tenne-Gazit, 2008).

While density quantiﬁes the average of actor tendencies, centralization is a
measure of variability or spread of social ties and reﬂects the extent to which a network
revolves around a single central actor (Wasserman & Faust, 1994). Based on the
centralization metric, networks can be characterized as centralized or decentralized.
Figure 3 presents a centralized network on the left compared to a decentralized network
on the right. In a centralized communication network, a central actor acquires
information pertaining to the beliefs of all other members, while peripheral members
are starved for information because they have limited interaction with each other. As
centralization declines however, group members become equally central in the
interaction structure and information ﬂows more uniformly throughout the system.
Consequently, decentralized communication networks are beneﬁcial for complex tasks

(Cummings & Cross, 2003; Sparrowe, Liden, Wayne, & Kraimer, 2001) and provide

better opportunities for consensus formation (Zohar & Tenne-Gazit, 2008). Whereas
centralized communication networks hinder social inﬂuence processes, centralized
friendship structures promote interpersonal inﬂuence and the emergence of shared
beliefs. Friendship centralization reﬂects the extent to which a central actor is involved
in the majority of personal relationships and is based on affective visibility where the
central actor provides a common ﬂame of reference for interpreting the situation (Zohar
& Tenne-Gazit, 2008). Group members who are engaged in a dyadic comparison
process to few others tend to hold similar perceptions because they agree with the same
signiﬁcant other (Zohar & Tenne—Gazit, 2008). Accordingly, centralized
communication networks are likely to produce higher levels of collective efﬁcacy
dispersion, while centralized ﬁiendship structures are likely to promote similar
collective efﬁcacy beliefs within teams.

Another way of characterizing a network is by determining whether subgroups
are present within a team. In contrast to research that investigates subgroups based on
demographic attributes (Lau & Mumighan, 1998), a social network approach examines
subgroups according to patterns of social interaction. Thus, subgroups represent non-
overlapping cohesive clusters and are deﬁned in terms of a concentration of interactions
within subgroups relative to the extent of interaction between subgroups (Frank et al.,
2008). Although it has been assumed that the team is the appropriate unit of analysis in
collective efﬁcacy research, formal boundaries may be misleading and inappropriate
when clusters or subgroups exist. Indeed, Brown et a1. (1996) suggested that dispersion
may emerge as a result of subgroups within the team, and recommended investigating

how often subgroups occur in teams along communication and social interaction

patterns. Although there is a lack of research on subgroups and collective efﬁcacy,
teams with communication and ﬁiendship subgroups are likely to have members with
dispersed collective efﬁcacy beliefs because information pertaining to the team’s
functioning and social comparison is constrained within cohesive clusters.

If there is evidence that subgroups based on communication and friendship ties
exist within teams and that such subgroups inﬂuence collective efficacy dispersion, then
it is important to understand the extent to which subgroups align according to collective
efficacy beliefs. That is, dispersion at the team level may be a result of cohesive
subgroups comprised of interacting members who develop shared collective efﬁcacy
beliefs over time. Previous studies have indicated that members of cohesive subgroups
tend to have similar beliefs as a result of interaction patterns (Bavelas 1950; Leavitt
1951; Smith, 1973). Accordingly, members of subgroups are likely to develop similar
collective efﬁcacy beliefs as frequent discussions about the team’s functioning and
affective relationships provide members with access to the beliefs of others. This
implies that individuals’ collective efﬁcacy beliefs may be inﬂuenced by members of
their subgroup. Indeed, conforming to norms in the subgroup contributes to
homogeneity within subgroups and heterogeneity between subgroups (F rank et al.,
2008). In this study, collective efficacy is modeled as a function of subgroup’s prior
mean level of collective efﬁcacy to determine the relative inﬂuence of subgroup
processes.

In contrast to typical social interaction measures that ask individuals to rate the
general level of interaction among team members, social network techniques provide a

comprehensive description of group interaction. Traditional measures of social

10

interaction assess individuals’ perceptions regarding how frequently team members as a
whole communicate or engage in personal relationships with each other. In contrast, a
social network approach captures the complete network of relationships by asking each
individual to indicate their interactions or relations with every member of the social
system. An added advantage is that social network analysis examines structural
properties of teams in relation to perceived outcomes, which avoids the problem of
using percept-percept relations (Totterdell, Wall, Holman, Diamond, & Epitropaki,
2004). The application of social network methods to the study of collective efﬁcacy has
been proposed by researchers (Chow & Feltz, 2007; Gibson, 1999). For instance,
Gibson (1999) suggested that “applying techniques from social network analysis would
shed additional light on the patterns of connections between members and factors
inﬂuencing how group members perceive, weight, and combine information in forming
efficacy beliefs” (p. 150). In response, the present study uses group level social network
measures (density, centralization, and subgroups) to predict collective efﬁcacy
dispersion.

To establish collective efﬁcacy dispersion as a meaningful unit level variable,
the construct must be related to important group outcomes. Previous studies have found
that aggregated collective efﬁcacy is a positive predictor of team performance (Gully et
al., 2002). However, a ﬁrller understanding of this relationship could be achieved
through the simultaneous examination of within-team variability in group members’
judgments. Researchers have proposed that collective efﬁcacy dispersion moderates the
relationship between aggregated collective efﬁcacy and team performance such that, the

relationship is stronger when members’ beliefs display higher levels of perceptual

ll

consensus (Gully et al., 2002; Kozlowski, Gully, Nason, & Smith, 1999). Support for
the moderating inﬂuence of collective efﬁcacy dispersion on team effectiveness has
been provided with dyads performing an experimental task (Arthur et al., 2007). A
logical extension that supports the generalizability of Arthur and colleagues (2007)
ﬁndings is to examine collective efﬁcacy dispersion as a moderator of the aggregated
collective efﬁcacy-team performance relationship with teams comprised of multiple
members performing meaningful tasks in their natural environment.

In addition to team performance, collective efﬁcacy dispersion may moderate
the relationship between aggregated collective efﬁcacy and group affective states such
as team commitment. Team commitment reﬂects individuals’ desire and resolve to
continue playing with their team (Scanlan, Carpenter, Schmidt, Simons, & Keeler,
1993). It is a motivational force that has implications for the amount of effort put forth
and psychological withdrawal. The relationship between aggregated collective efﬁcacy
and team commitment should be stronger when dispersion is low rather than high
because perceptual consensus fosters identiﬁcation with the team and facilitates
consistent affective responses (Gonzalez-Roma et al., 2002; Lindell & Brandt, 2000).
Contextual Factors

The context chosen for the current investigation was sport. Team sports provide
an optimal context in which to study the development of collective efﬁcacy through
social interaction patterns because athletic teams represent intact units performing
meaningful tasks in complex environments where members share a common purpose
and experience similar events. Furthermore, unlike many teams found in organizational

settings, sports teams receive frequent feedback where individuals attempt to make

sense of previous performances by engaging in communication and social comparison
processes with other teammates. Another advantage of using sports teams to better
understand collective efﬁcacy dispersion is that there are identiﬁable and relevant
stages of team development. For instance, a competitive season is characterized by a
beginning, middle, and end. Previous research suggests that team members’ collective
efficacy beliefs become more homogenous over time (Jung & Sosik, 2003). Thus, using
team sport as a context to investigate why some teams develop more interrelated beliefs
than others allows for an examination of this process among teams at similar points in
time.

Women’s intercollegiate softball was selected for this study for conceptual
reasons and to further advance this line of inquiry with an underrepresented type of
sport. Softball is a team sport in which only women participate at the intercollegiate
level. Therefore, differences due to gender and sport are essentially controlled for by
using women’s softball teams. While this may limit the generalizability of the ﬁndings,
the internal validity is strengthened. Previous research on collective efﬁcacy has
focused almost exclusively on highly interdependent sports such as football (Myers,
Feltz, & Short, 2004), hockey (Feltz & Lirgg, 1998), and basketball (Watson et al.,
2001 ). Unfortunately, little is known about the development of collective efﬁcacy in
sports that are characterized by lower levels of system interdependence. Team sports
can be viewed along a task interdependence continuum with coactive sports (e.g., golf)
at the low end and highly interdependent sports (e. g., soccer) at the high end. Located in
the middle of the continuum are sports that require moderate levels of coordination and

interaction among members such as baseball and softball. Finally, intercollegiate

l3

softball teams were selected over other competitive levels such as high school or club
because rosters and team performance statistics are easily accessible from college and
conference websites.
Purpose of the Study
The primary purpose of this study was to examine communication and
ﬁiendship networks as antecedents of collective efﬁcacy dispersion. This study extends
previous research by examining collective efficacy dispersion as a focal variable and
employs a social network framework to understand how collective efﬁcacy beliefs
develop as a function of interpersonal relations. A secondary purpose of this study was
to establish the construct validity of collective efﬁcacy dispersion by investigating the
moderating inﬂuence on team performance and commitment.
Hypotheses
1. The density of a team’s communication network is negatively related to
collective efﬁcacy dispersion, while controlling for prior levels of dispersion.
2. The density of a team’s friendship network is negatively related to collective
efﬁcacy dispersion, while controlling for prior levels of dispersion.
3. Centralization of a team’s communication network is positively related to
collective efﬁcacy dispersion, while controlling for prior levels of dispersion.
4. Centralization of a team’s friendship network is negatively related to collective
efﬁcacy dispersion, while controlling for prior levels of dispersion.
5. Subgroup communication is positively related to collective efﬁcacy dispersion,

while controlling for prior levels of dispersion.

l4

6.

Subgroup friendship is positively related to collective efﬁcacy dispersion, while
controlling for prior levels of dispersion.

Collective efﬁcacy dispersion moderates the relationship between aggregated
collective efﬁcacy and team performance such that, the relationship will be
stronger when collective efﬁcacy dispersion is low rather than high.

Collective efﬁcacy dispersion moderates the relationship between aggregated
collective efﬁcacy and team commitment such that, the relationship will be

stronger when collective efﬁcacy dispersion is low rather than high.

Research Questions

1. Do subgroups based on communication ties exist within teams?

2. Do subgroups based on ﬁiendship ties exist within teams?

3. Do communication subgroups align by collective efﬁcacy beliefs?

4. Do friendship subgroups align by collective efﬁcacy beliefs?

5. Are individuals’ collective efﬁcacy beliefs inﬂuenced by members of their
communication subgroup?

6. Are individuals’ collective efﬁcacy beliefs inﬂuenced by members of their
ﬁiendship subgroup?

Delimitations

The ﬁndings are limited to intercollegiate women’s softball teams, and thus may

not generalize to other populations. For instance, the results may not be applicable to

softball teams at different competitive levels, sports characterized by lower or higher

task interdependence, or men’s sports in general. Communication and friendship

interactions on women’s teams may be different from interactions on men’s teams.

15

Further, some of the measures employed such as the collective efﬁcacy questionnaire

may only be appropriate for intercollegiate softball teams.

Definitions

1.

Centralization: the extent to which a network revolves around a central actor;
the sum of the differences between the largest centrality score and all other
observed centrality scores, divided by the maximum possible sum of differences
in actor centrality.

Collective efﬁcacy: a team’s belief in their capabilities to organize and execute
the courses of action necessary to produce certain levels of attainments.
Collective eﬁ‘icacy dispersion: the degree of within-team variability in group
members’ collective efﬁcacy beliefs.

Communication network: task-related exchanges among team members
pertaining to the team’s functioning such as discussing the team’s strengths and
weaknesses, performances, or goals.

Density: a measure of the interconnectedness of actors in a network; the
proportion of ties in a network relative to the total number of possible ties in a
network.

Friendship network: expressive interaction or affective relationships among
team members; very good ﬁiends, seen socially outside of team-related
activities.

Klique Finder: a program that is based on a general algorithm for identifying

cliques (clusters or subgroups) of actors in a network.

16

8. Run differential: the number of runs scored minus the number of runs allowed
divided by the number of games played.

9. Social network analysis: a set of techniques that map and measure relationships
and ﬂows between actors in a network.

10. Subgroups: non-overlapping cohesive clusters in terms of a concentration of
interactions within subgroups relative to the extent of interaction between
subgroups.

11. Team commitment: a team’s desire and resolve to continue playing with their
team.

12. Team performance: a team’s winning percentage; the number of wins divided

by the number of games played

17

CHAPTER 2
REVIEW OF LITERATURE

The purpose of this chapter is to provide a review of literature that is relevant to
the variables and procedures in this study. This chapter begins by providing a rationale
for conceptualizing collective efﬁcacy dispersion as a theoretical construct based on
conceptual contributions and empirical research. Next, research that has examined
social interaction and shared beliefs is presented and integrated with collective efﬁcacy.
This is followed by a summary of social network analysis including research on
communication and friendship networks, as well as group level metrics that describe the
structure of various networks. Finally, research pertaining to collective efﬁcacy
dispersion as a moderating variable is discussed.
Dispersion as a Group-Level Construct

Elemental composition models are concerned with situations in which data
collected at lower level units are combined to establish the higher level construct (Chan,
1998). For instance, collective efﬁcacy is an emergent group-level construct that is
derived or composed from members’ individual perceptions. Bandura (1997) has
suggested two methods for assessing collective efﬁcacy that reﬂect different forms of
composition. The ﬁrst method aggregates team members’ individual perceptions of self-
efﬁcacy (i.e., direct consensus), while the second method aggregates team members’
individual perceptions of collective efﬁcacy (i.e., referent-shift consensus). While both
methods have been used in collective efficacy research, the majority of studies have
preferred the individual perceptions of collective efﬁcacy method because it is a better

predictor of group performance (Feltz & Lirgg, 1998; Myers, Feltz et al., 2004).

18

In composition based on referent-shift consensus, the basic content of the
efﬁcacy perception is unchanged, but the referent of the content has changed from the
self to the team (Chan, 1998). Researchers using referent-shift consensus will assess
team members’ individual perceptions of collective efﬁcacy and then calculate the
degree of within-team agreement as an index of consensus. If a sufﬁcient level of
within-team consensus is demonstrated, the individual perceptions are aggregated to
form the higher level construct of shared collective efficacy. Thus, collective efﬁcacy is
operationalized as the mean of team members’ collective efﬁcacy beliefs and the degree
of within-team variability is used as a statistical prerequisite to justify aggregation (e.g.,
Feltz & Lirgg, 1998; Prussia & Kinicki, 1996). A limitation of this approach is that it
dichotomizes within-group variance and assumes that dispersion is the same across
groups. However, teams that reach a sufﬁcient of level for aggregation do not
necessarily have the same degree of within-team variation.

A dispersion theory moves beyond the dichotomous view of aggregation and is
deﬁned as “an argument for the construct validity of dispersion along a speciﬁc
variable” (Brown et al., 1996, p. 10). That is, dispersion is conceptualized as a focal
construct instead of merely a statistical prerequisite for aggregation. Whereas referent-
shift composition uses within-group consensus as a necessary precondition for the
construct validity of the higher level construct, dispersion composition uses within-team
variability as the operationalization of the group level construct (Chan, 1998). Thus,
groups can be characterized along a continuum from high dispersion to low dispersion.

An implicit assumption of dispersion theories is that the degree of within-team

l9

variability may be of theoretical interest and explain individual and team-related
outcomes, or may moderate group-level effects.

Despite conceptual arguments (Brown et al., 1996; Chan, 1998; James,
Demaree, & Wolf, 1984; Lindell & Brandt, 1997), dispersion constructs have received
little empirical examination. An exception is perceived climate, which shares similar
characteristics to collective efﬁcacy in that both constructs require composition models
where consensus is used to justify aggregating lower level responses (i.e., psychological
climate) to represent the higher level construct (i.e., organizational climate). Climate
strength, which refers to the degree of within-team variability in members’ perceptions
of climate (stronger climates have lower levels of within-group variability), has been
found to account for unique variance in important organizational outcomes beyond the
magnitude or aggregate of climate perceptions (Bliese & Halverson, 1998; Gonzalez-
Roma et al., 2002; Schneider, Salvaggio, & Subirats, 2002). For instance, Bliese and
Halverson (1998) examined the inﬂuence of climate strength on units’ average
psychological well-being with US. Army companies. They found that after controlling
for aggregated climate perceptions, climate strength in leadership and peer relations
predicted companies’ average psychological well-being. Schneider and colleagues
(2002) investigated whether climate strength moderated the relationship between bank
employee ratings of service climate and customer perceptions of service quality. They
found that the relationship between average climate perceptions and customer
satisfaction services was stronger for units with high climate strength (i.e., lower

within-team variability) for managerial practices.

20

Similar to climate strength, theorists have asserted that collective efﬁcacy
dispersion may provide insight into team functioning and performance beyond the mean
of collective efﬁcacy (Chan, 1998; DeRue et al., in press; Moritz & Watson, 1998).‘To
date, only one empirical study has investigated collective efﬁcacy dispersion as a focal
variable. Arthur et a1. (2007) explored the relationship between the degree of within-
team variability in collective efﬁcacy and team performance with 85 male dyadic teams
performing an information processing task that was highly interdependent and complex.
A videogame-based aviation task (i.e., Space Fortress) was employed where
participants performed simultaneously each half of the task components (pilot-gunner
functions and copilot mine-missile manger functions). Participants were administered a
collective efﬁcacy questionnaire at three time points over a 2-week period, which asked
“How conﬁdent are you in the ability of your team to play Space Fortress?” and “If
your team played Space Fortress in competition with 10 other teams, how do you think
your team will place?” At each administration occasion, dyads performed two test
games in which trainees alternated task roles. Team performance was calculated by
taking the mean of the total scores from the two test games. The authors found partial
support for the predictive validity of collective efﬁcacy dispersion. Dyads with higher
levels of within—team agreement outperformed dyads with lower levels of agreement,
after controlling for collective efﬁcacy magnitude. Findings from this study provide
preliminary empirical support for the construct validity of collective efﬁcacy dispersion,
though it is difﬁcult to draw substantive conclusions based on one study. Whether the
ﬁndings generalize to intact teams comprised of several members performing in their

natural environment is unknown. More importantly, if the degree of within-team

21

variability is of substantive interest to researchers and practitioners, then it is important
to understand the factors that contribute to perceptual consensus or dispersion in group
members’ collective efﬁcacy beliefs.
Social Interaction and Shared Beliefs

Several theoretical frameworks suggest that social interaction underlies
agreement or disagreement within a group. One of the ﬁrst theories to recognize the
importance of interpersonal interaction in the development of shared perceptions was
social comparison theory (F estinger, 1954). Social comparison theory explains how
individuals evaluate their own opinions and attitudes by comparing themselves to
others. The theory posits that attitude formation results from a social comparison
process where individuals weigh and integrate the attitudes of similar others. Other
frameworks such as action theory (Silverrnan, 1971) suggest that interpersonal
interactions with others cause people to modify and transform social meanings. An
action theory perspective suggests that people who interact with each other in a group
tend to attach similar meanings to organizational events, while people involved in
different interaction groups attach different meanings to the same events (Rentsch,
1990). Symbolic interactionism posits that meanings of environmental events are
formed in the context of social interaction and modiﬁed through interpretation
processes (Blumer, 1969). Studies using symbolic interactionism suggest that
individuals interacting with each other in work place develop similar perceptions of the
environment over time (Schneider & Reichers, 1983). Symbolic interaction involves
sense-making interaction where members engage in inductive and exploratory

exchanges concerning the meaning of events in a quest for perceptual consensus (Zohar

22

& Tenne-Gazit, 2008). Finally, social identity theory (Tajfel & Turner, 1979) was
developed to understand the psychological basis of intergroup discrimination and is
concerned with explaining why individuals in groups adopt shared attitudes. The theory
posits that group identiﬁcation promotes shared social agreement among group
members. Taken together, these different theoretical perspectives suggest that shared
perceptions develop from prolonged interaction, whereas divergent beliefs arise from a
lack of social interaction.

Previous research on climate strength (i.e., within-group variability in climate
perceptions) has provided evidence for the inﬂuence of social interaction on shared
perceptions. Social interaction among unit members fosters similarity in psychological
climates and promotes the emergence of shared climate perceptions (Klein et al., 2001;
Gonzalez-Roma et al., 2002). Klein and colleagues (2001) investigated social and work-
related interaction as an antecedent of within-group variability in group members’
perceptions of their work environment. The social interaction scale asked group
members to describe how frequently they interacted with each other as ﬁiends, whereas
the work interdependence instrument required group members to rate the extent to
which they coordinated with and depended on each other to accomplish their work
tasks. Group level analyses revealed that units with greater social and work-related
interaction had less within-group variability in responses to ﬁnancial resource
availability and plant innovativeness than units with lesser amounts of interaction.
Gonzalez-Roma and colleagues (2002) examined social interaction as an antecedent of
climate strength with 197 work units of the regional public health service. Social

interaction was measured by asking unit members to report how often they talked about

23

the work unit’s goals, work planning, and functioning with their unit work mates. The
authors found that social interaction signiﬁcantly correlated with climate strength in
goals orientation and innovation. Units with higher levels of social interaction among
group members regarding work-related issues (i.e., goals, planning, and functioning)
had higher levels of perceptual consensus in climate perceptions. Results on climate
strength suggest that group level social interaction is an antecedent of within-team
variability in group members’ climate perception. Teams in which members perceive
high levels of within-team interaction develop similar perceptions, whereas teams in
which group members perceive a lack of social interaction develop dispersed
perceptions. An alternative explanation is that individuals choose to interact with other
group members who have similar perceptions. Previous studies that have examined the
relationship between group interaction and the development of shared perceptions have
failed to obtain an initial measure of the perception of interest which makes it difﬁcult
to ascertain the direction of causality.

Several theorists have suggested that shared collective efﬁcacy beliefs emerge
from interpersonal interaction. Gibson (1999) proposed that collective efﬁcacy deve10ps
“as group members collectively acquire, store, manipulate, and exchange information
about each other and about their task, context, process, and prior performance” (p. 138).
Conversely, differences in group members’ perceptions of the team’s capability to
succeed would arise when individuals fail to obtain and exchange information with each
other. Jung and Sosik (2003) have also noted the role of social interaction in the
emergence of shared collective efﬁcacy beliefs. In discussing their ﬁnding that student

groups working on decision-making projects developed more homogenous perceptions

24

of collective efﬁcacy over the course of a semester, they speculated that substantial
social interaction and perspective taking was a necessary precondition for the
emergence of shared beliefs. Based on these assertions, the extent to which group
members discuss and share their perspectives about the team’s functioning may
determine differences in collective efﬁcacy dispersion among teams. Teams in which
group members engage in frequent communication should develop more interrelated
beliefs, whereas teams that lack social interaction should develop more dispersed
beliefs. However, because the degree of social interaction has not been directly assessed
in relation to collective efﬁcacy, these propositions are only speculative.

The notion that collective efﬁcacy beliefs can be inﬂuenced through social
interaction has been implied by researchers who have critiqued the group discussion
method of collective efﬁcacy assessment. In contrast to the standard assessment
procedure discussed earlier in this chapter, where members’ individual perceptions of
collective efﬁcacy are aggregated (i.e., rate your conﬁdence in the team’s ability), the
group consensus approach uses group discussion to obtain a single estimate (e.g.,
Gibson etal., 2000; Jung & Sosik, 2003). Researchers that use the group discussion
method ﬁrst proceed by having team members rate their conﬁdence in the team
individually (as in the individual collective efﬁcacy method), and then as a collective
group. During group assessment, team members are allowed to interact with each other
to discuss their personal beliefs regarding the team’s ability to be successful until a
forced consensus is reached. Bandura (1997) has criticized the group-based approach
because social persuasion and pressure to conform may produce inaccurate estimates of

collective efﬁcacy. For instance, an individual may agree with the team estimate that is

25

derived based on interactions with vocal members when in reality their perception
remains unchanged. While the group consensus approach may reveal that individuals
have deviated from their held belief erroneously, it is also plausible that individuals
modify or change their perception of collective efﬁcacy as a result of information
exchanges with other group members. Convincing arguments or access to new
information are weighed against the individual’s initial belief. Studies that have
employed both methods have not measured individual perceptions of collective efﬁcacy
a second time following group discussion and thus, it is unknown whether group
members actually changed their belief or simply conformed to reduce conﬂict.
Social Networks

Although social network analysis has been used since the mid-19303, interest in
social networks has begun to grow at an increasing rate. Fields such as sociology,
psychology, epidemiology, and physics are now adopting a network perspective to
study complex issues. In organizational psychology alone, social network techniques
have been used to examine a wide range of organizational phenomena such as social
capital, ernbeddedness, network organizations, board interlocks, joint ventures and
inter-ﬁrm alliances, knowledge management, social cognition, and group processes (see
Borgatti & Foster, 2003). Surprisingly, the application of social network analysis in
sport psychology research has failed to match the progress of other disciplines. This is
unfortunate considering that social network techniques are particularly useful for stating
social properties and processes and rigorously deﬁning theoretical constructs (Freeman,

1984).

26

A social network consists of a set of actors and the ties between these actors
(Wasserman & Faust, 1994). The actors may represent different levels of analysis
including individuals, teams, organizations, societies, or concepts. A wide range of ties
have been examined that reﬂect different types of relations or interactions among actors
such as communication (e.g., who talks to whom), affective (e. g., who is ﬁiends with
whom), proximal (e. g., who is physically close to whom), and cognitive (e.g., who
knows whom). The critical difference between social network techniques and traditional
analytic methods is that it uses structural or relational information of a group rather than
attribute data. Based on the pattern of relationships between individuals in a network,
measures of social structure can be calculated that may explain the transfer of
information, beliefs, and behaviors.

While actors are often involved in multiple types of relations with other
members in a network, certain types of ties are more relevant to social inﬂuence than
others. Two types of ties that have been studied in relation to important individual and
organizational outcomes are communication and friendship networks. Communication
networks reﬂect instrumental interactions that involve exchanges between members that
are task-related. Examples of communication ties include receiving advice about work-
related problems, discussing work-related topics, and interacting with people in order to
complete a task. Task-related information exchanges provide pathways to coordinate
existing activities and reinforce organizational norms. They have been found to predict
inﬂuence acquisition (Brass, 1984), organizational power (Burkhardt & Brass, 1990),
risk taking, acceptance and information access (Ibarra & Andrews, 1993), work-related

knowledge (Morrison, 2002), performance (Sparrowe et al., 2001), perceptions of

27

learning and enjoyment (Baldwin, Bedell, & Johnson, 1997) and self-efﬁcacy
(Burkhardt, 1994).

Friendship networks represent affective or expressive ties among members. In
contrast to communication networks which are sources of information and advice,
ﬁiendship networks are based on intimacy and trust, and are important sources of social
support (Ibarra, 1993; Lincoln & Miller, 1979). In addition, ﬁiendship ties tend to be
more stable and enduring than communication relations (Shah, 2000). Close ﬁiends or
individuals who one consults or gets help from about personal issues deﬁne the
ﬁiendship network. Because of the level of affective closeness reﬂected in a friendship
network, group members tend to share acceptable or attitude-reinforcing information to
reduce stress and conﬂict (Krackhardt, 1999). Similar to communication ties, ﬁiendship
networks predict important outcomes such as organizational commitment (Morrison,
2002), team-based learning satisfaction (Baldwin et al., 1997), resource sharing during
crisis (Krackhardt & Stern, 1988), and career-related decision making (Kilduff, 1990).

There tends to be some overlap between communication ties and ﬁiendship ties
in a network (Borgatti & Foster, 2003). That is, people often communicate with
individuals who are also considered their close ﬁiends. However, the two types of ties
are not only conceptually distinct (a person may communicate with someone who is not
their friend), but the theoretical mechanisms of social inﬂuence are also different.
Communication ties are pathways for the transfer of knowledge and information.
Discussion and perspective sharing about task-related issues such as previous
performance resolve uncertainty and provide actors with access to new information and

a better understanding of team events. Friendship ties are pathways that produce social

28

inﬂuence via interpersonal comparison and social modeling where similar or affectiver
close signiﬁcant others provide a frame of reference for managing uncertainty (Burt,
1987).
Social Network Measures

While communication and ﬁiendship ties describe particular types of relations
between actors in a network, group level social network metrics describe the structure
of these ties. An advantage of this approach is that the metrics can be used to
characterize teams according to their pattern of communication and ﬁiendship
interactions. Consequently, the structure of teams’ communication and friendship
networks can be examined in relation to differences in group level outcomes such as
collective efﬁcacy dispersion. The most common group level network measure is
structural density (Scott, 2000). Density is a measure of the interconnectedness of actors
in a network and is deﬁned as the proportion of ties in a network relative to the total
number of possible ties. For instance, if Team A and Team B both had 5 group
members, there would be 20 possible ties within each team (assuming directional
relations). If Team A had 10 pairs of ties and Team B had 6 pairs of ties, then Team A
would have a denser network. The density coefﬁcient is at a minimum when no direct
ties exist between actors in a network and at a maximum when every actor has a tie with
all other members.

Previous organizational studies that have employed social network methods
have demonstrated that teams with higher levels of communication and ﬁiendship
density outperform teams with lower levels of density (Baldwin et al., 1997; Reagans &

Zuckerman, 2001). Reagans and Zuckerman (2001) examined the relationship between

29

communication density (i.e., the average frequency of communication among team
members) and team performance with 224 corporate research and development teams.
The social network data were obtained through a sociometric instrument in which
members were asked to indicate how frequently they communicated with other
members of the same team. They found that teams averaging more frequent
communication among members achieved higher levels of productivity because dense
patterns of local interaction provided a basis for coordination and collective action.
Baldwin and colleagues (1997) examined the inﬂuence of communication and
friendship network ties on team performance with MBA student teams. Communication
relations were assessed by asking students to indicate the classmates who were
important sources of school-related advice or whom they approached for school-related
problems, while friendship ties were measured by asking respondents to indicate the
individuals who were very good friends, people whom they saw socially outside of
school. Results demonstrated that groups that reported higher levels of communication
density had stronger perceptions of shared work load and team interaction effectiveness
which in turn, predicted team grade. In addition, teams that reported denser friendship
ties had stronger perceptions of team interaction effectiveness which in turn, also
contributed to student teams’ grades. A meta-analysis of 37 studies involving 3,098
intact teams revealed that teams with densely conﬁgured instrumental and expressive
ties performed higher and were more committed to staying together than teams with
sparsely conﬁgured ties (Balkundi & Harrison, 2006).

Not only does network density improve team performance, but it may also

facilitate the development of shared beliefs (Baerveldt & Snijders, 1994; Friedkin,

30

1984; Zohar & Tenne-Gazit, 2008). Indeed, Brass, Butterﬁeld, and Skaggs (1998) have
argued that dense networks may be a necessary requirement for the development of
shared norms and values. Network density fosters identiﬁcation with the group,
enhances coordination and collective action, and promotes mutual trust (Coleman,
1988; Portes & Sensenbrenner, 1993). In dense networks, group members frequently
share information and compare themselves to others, which thereby facilitates the
diffusion of beliefs and social modeling processes. Conversely, in sparse networks,
information sharing and interpersonal comparisons are constrained which produces
isolated opinions and provides fewer opportunities for social comparison. Indeed, a
team of isolates will have difﬁculty exchanging task-related issues because there are no
established patterns of ties to convey the information (Balkundi & Harrison, 2006).
Evidence supporting a potential relationship between density and collective
efﬁcacy dispersion comes from the group cohesion literature. Although conceptually
distinct, researchers have used density as a group-level measure of cohesion (Festinger,
Schachter, & Black, 1950; Frank 1996, Frank & Yasumoto 1998). Although several
deﬁnitions of group cohesion have been proposed (Mudrack, 1989), sport and
organizational research typically conceptualize group cohesion as a cognitive,
motivational, or affective group-level attribute. For instance, Carron, Brawley, and
Widmeyer (1998) have deﬁned group cohesion as “a dynamic process that is reﬂected
in the tendency for a group to stick together and remain united in the pursuit of its
instrumental objectives and/or for the satisfaction of member affective needs” (p. 213).
Measures of group cohesion based on such deﬁnitions use attitudinal scales where team

members rate their level of attractiveness to the group (e. g., Carron, Widmeyer, &

3]

Brawley, 1985). In contrast, network density provides a measure of cohesion based on
direct interactions and relationships within teams.

Theorists have asserted that group cohesion is an antecedent of collective
efﬁcacy (Zaccaro at al., 1995). Group cohesion is a multidimensional construct that
involves both task and social factors. Task cohesion reﬂects members’ feelings about
the similarity, closeness, and bonding within the team as a whole around the group’s
goals and objectives. Previous studies have found that individuals with higher levels of
task cohesion have stronger perceptions of collective efﬁcacy than individuals with
lower levels of task cohesion (Kozub & McDonnell, 2000; Paskevich et al., 1999;
Spink, 1990). For example, Paskevich and colleagues (1999) investigated the inﬂuence
of group cohesion on collective efﬁcacy with university and club volleyball teams.
They found that the task cohesion subscales (individual attraction to the group-task and
group integration-task) distinguished athletes that were either high or low in collective
efﬁcacy. Players with higher levels of task cohesion held stronger beliefs about the
team’s ability to be successful than individuals with lower levels of task cohesion.
Kozub and McDonnell (2000) reported similar ﬁndings with rugby players in that task
cohesion accounted for 32% of the variance in members’ collective efﬁcacy beliefs.

Social cohesion reﬂects group members’ feelings about the similarity, closeness,
and bonding within the team as a whole around the group as a social unit. Although a
weaker source of collective efﬁcacy than task cohesion, studies have shown that social
cohesion inﬂuences beliefs regarding team effectiveness (Kozub & McDonnell, 2000;
Spink, 1990). In one of the ﬁrst studies to examine collective efﬁcacy, Spink (1990)

proposed a relationship between group cohesion and collective efﬁcacy based on

32

previous research suggesting that cohesion and efﬁcacy were important determinants of
sport performance. Participants were volleyball players from recreational and elite
teams participating in an annual tournament. Separate analyses were conducted, which
revealed that group cohesion was a positive predictor of collective efﬁcacy for players
of elite teams, but not for players of recreational teams. That is, athletes of elite
volleyball teams who reported that their team was socially cohesive held stronger
beliefs about their team’s capabilities. Further evidence for the inﬂuence of social
cohesion on collective efﬁcacy was demonstrated by Kozub and McDonnell (2000)
who found that rugby players with higher perceptions of social cohesion had a stronger
sense of collective efﬁcacy than players with lower perceptions of social cohesion.

Clearly, research suggests that higher levels of task and social cohesion are
associated with stronger perceptions of collective efﬁcacy. Further, although
conceptually different, communication and friendship density may serve as proxies for
task and social cohesion, respectively. For instance, task cohesion implies that members
are involved in frequent task-oriented exchanges, while social cohesion suggests that
members view their team as an important social group. Indeed, Zaccaro at al. (1995) has
speculated that groups may not be as cohesive if some members perceive that the team
is strong, while others believe that it is weak.

Whereas density reﬂects the general level of cohesion in a network,
centralization describes the extent to which the cohesion is organized around central
positions (Scott, 2000). Centralization is a group-level social network metric that is a
measure of variability, dispersion, or spread of social ties (Wasserman & Faust, 1994).

It represents the extent to which a network revolves around a highly central actor

33

(Freeman, 1979). Completely centralized networks resemble a perfect star where one
central actor is tied to all other members who are not tied to each other. Conversely,
decentralized networks exist when all network members have the same centrality score.

Centrality is an individual level variable that reﬂects the extent to which an actor
is directly tied to other members in a network. All actors in a social network have a
centrality score. Actors who occupy more central positions in a network have greater
levels of power (Brass, 1984), job satisfaction (Dean & Brass, 1985), inﬂuence in
decision-making (Friedkin, 1993), and performance (Baldwin et al., 1997; Sparrowe et
al., 2001). While centrality reﬂects an individual’s position in a network, centralization
describes the network structure as a whole. However, the positive consequences
associated with centrality at the individual level may not necessarily extend to
centralization at the group level. Early research on centralization employed
experimental designs where the pattern of communication among group members was
manipulated by controlling who could send information to whom (Bavelas, 1950;
Leavitt, 1951; Shaw, 1954, 1964). Results ﬁom these studies found that centralized
communication structures were beneﬁcial for simple tasks, while decentralized
communication networks were superior for complex tasks. Recent studies on
centralization have replicated earlier ﬁndings with intact groups in the ﬁeld. For
instance, Sparrowe and colleagues (2001) investigated the inﬂuence of advice
centralization and density on performance with 38 work groups. While density was
unrelated to team performance, groups with decentralized advice patterns performed
better than groups with centralized advice patterns. Cummings and Cross (2003)

investigated structural properties of a network with 182 work groups in a global

34

organization. They found that groups constrained by centralized communication
structures performed worse than groups with more integrative structures.

Similar to structural density, centralization may affect the emergence of shared
or dispersed perceptions at the team level, though the direction of effect may differ
depending on the type of tie. Zohar and Tenne—Gazit (2008) examined communication
and friendship centralization as an antecedent of climate strength for safety with male
infantry soldiers on 45 platoons undergoing advanced training camp for army ﬁeld
units. A sociometric instrument was administered in which participants were asked how
much they talked with each of the platoon members on subjects that were activity or
mission related (i.e., communication network) and with which of their platoon members
they consulted or received help from about personal issues (i.e., friendship network).
The authors hypothesized that units with decentralized communication structures would
have less within-team variability in climate perceptions because a wider spread of
member exchanges offers better opportunities for social diffusion of information.
Conversely, they hypothesized that platoons with more centralized ﬁiendship networks
would have less climate dispersion because a central actor would provide a common
referent for peripheral members to compare and model themselves after. They reasoned
that individuals who were involved in a dyadic comparison process with few members
were more likely to hold similar beliefs because they agreed with the same signiﬁcant
other. Consistent with their hypotheses, unit members’ safety climate perceptions
demonstrated higher levels of agreement when communication ties were less centralized

and when friendship ties were more centralized.

35

The opposing effects of communication and friendship centralization can be
explained by the different mechanisms of social inﬂuence (Zohar & Tenne-Gazit,
2008). In a communication network, information serves as the conduit in which the
beliefs of group members transfer to others. Decentralized communication structures
facilitate information ﬂow throughout the network (Cummings & Cross, 2003), and
increase the likelihood that members will receive access to the views of all others. In
contrast, centralized networks constrain the ﬂow of information which is concentrated
around a dominant actor. When information is transmitted to relatively few members,
the central actors obtain access to the perspectives of the peripheral members, while the
network as a whole receives limited information. Whereas information deﬁnes the
communication network, the pathway of social inﬂuence in a ﬁiendship network centers
on interpersonal comparison or social modeling. Centralized ﬁiendship networks
revolve around a central actor who is ﬁiends with the majority of team members who
are not ﬁiends with each other. Because friendship ties are based on closeness and the
central actor has the highest affective visibility, peripheral members tend to conform to
the central belief (Zohar & Tenne-Gazit, 2008).

While the ﬁndings of Zohar and Tenne-Gazit (2008) demonstrate support for the
predictive validity of group level structural properties on the development of perceptual
consensus and provide a general framework for the examination of social interaction as
a source of collective efﬁcacy dispersion, there are some limitations that should be
noted. The primary limitation was that a previous measure of within-team variability in
unit members’ perceptions of safety climate was not obtained, which makes it difﬁcult

to determine the direction of inﬂuence. For instance, did social interaction produce

36

shared beliefs or did individuals choose to interact with group members who held
similar perceptions? This limitation has been observed in the social interaction-shared
perception literature (e. g., Gonzalez-Roma et al., 2002) as well as social network
research (e.g., Rentsch, 1990). The lack of longitudinal research makes it difﬁcult to
determine the inﬂuence of the network on the hypothesized effects (Lazer, 2001). While
multiple assessments of the social network may provide the clearest understanding
regarding the issue of inﬂuence or selection, the difﬁculty of collecting social network
data poses a problem. An alternative involves measuring the hypothesized outcome
(e. g., initial collective efﬁcacy dispersion) prior to the social network to control for
previous differences between teams. Another limitation of the Zohar and Tenne-Gazit
(2008) study was that networks were deﬁned as the unit members only and thus, did not
include group leaders. However, social interactions between individual group members
and the team leader are just as important as interactions among group members, and
may inﬂuence group-level social network metrics.
Subgroups

Previous studies that have examined subgroups have focused on differences
between group members on demographic attributes such as age, sex, ethnicity, tenure,
and functional area (Jackson, May, & Whitney, 1995). This line of research has
produced equivocal ﬁndings regarding the consequences of subgroups. One perspective
suggests that subgroups have a negative inﬂuence on team functioning because
differences among members hinder communication and social integration. Supporting
this perspective, research has found that teams with subgroups perform at lower levels

than both highly homogeneous and highly heterogeneous teams as a result of

37

communication problems, increased conﬂict, and lower levels of team identity (Barley
& Mosakowski, 2000). The presence of subgroups may create social factions within
teams where individuals only interact with other members of their subgroup with little
interaction between members of different subgroups. An alternative perspective is that
subgroups have a positive inﬂuence on team functioning. Research supporting this
perspective suggests that subgroups contribute to higher levels of creativity, richer
information processing, and higher quality decision-making (McGrath, 1984; Jackson,
1992; Lovelace, Shapiro, & Weingart, 2001). Teams with subgroups engage in more
learning behavior than teams that lack subgroups because differences across subgroups
ensure that a diversity of insights is considered (Gibson & Vermeulen, 2003).

A social network approach to identifying cohesive subgroups focuses on social
interactions rather than demographic characteristics (Frank, 1995). This perspective
deﬁnes subgroups as clusters of actors who engage in more frequent relations with
members internal to their subgroup than with members external to their subgroup.
Similar to density and centralization, subgroups are examined according to
communication and friendship ties in this study. Subgroups may produce lower levels
of emotional attachment to the group as a whole if members identify with their
subgroup more strongly than their team (Paxton & Moody, 2003). In addition,
consensus formation may depend on whether subgroups exist within teams (Brown et
al., 1996). For instance, teams with subgroups based on communication and friendship
interactions may have higher levels of collective efﬁcacy dispersion than teams without
subgroups. Although there may be clusters of local agreement within subgroups, the

differences between subgroups are likely to create dispersion at the team level. In

38

subgroups, there tends to be informational redundancy among members, which
reinforces certain attitudes and beliefs, eventually becoming part of the subgroup’s
normative structure. While the dense pattern of interactions within cohesive subgroups
facilitate information ﬂow and social comparison processes among actors of the same
subgroup, members of different subgroups have limited access to the beliefs and
perspectives of others. The lack of information sharing and knowledge transmission
produces homogenous beliefs within clusters and heterogeneous beliefs between
clusters.

Although a social network approach to identifying subgroups focuses on dense
patterns of social interaction, research suggests that subgroups align by individuals’
background characteristics, sentiments, and behaviors (Frank, 1995). That is, members
of cohesive subgroups tend to have similar demographic characteristics, hold similar
beliefs, and exhibit similar patterns of behavior. One belief that may be related to
subgroup membership is collective efﬁcacy. Indeed, researchers have speculated that
subgroups based on collective efﬁcacy may form within teams as a result of social
interaction patterns among members (DeRue et al., in press). The formation of
subgroups based on collective efﬁcacy likely emerges through frequent discussions
about the team’s functioning and affective relationships which provide members with
access to the beliefs of others.

An underlying premise of subgroups is that members of particular clusters
develop similar attitudes and beliefs as a result of dense social interactions within their
subgroup. Individuals may modify or change their personal beliefs through social

inﬂuence processes exerted by proximal members. This implies that actors may be

39

inﬂuenced by members of their subgroup. Previous research suggests that individuals,
particularly females, are highly responsive to social norms in their local clusters (Frank
et al., 2008). Just as the pattern of group interaction at the network level may explain
between-team diﬁerences in collective efficacy dispersion, team members’ perceptions
of collective efﬁcacy may depend on subgroup processes. Through communication
exchanges and social modeling, individuals may modify or change their beliefs to
match the perceptions of their subgroup members. This suggests that individuals’
collective efﬁcacy beliefs will be determined by their subgroup’s prior mean levels of
collective efﬁcacy.
Collective Eﬂicacy Dispersion as a Moderator

The relationship between collective efﬁcacy and team performance has been
well documented in the literature. Results from both laboratory (Hodges & Carton,
1992; Lichacz, & Partington, 1996) and ﬁeld studies (Feltz & Lirgg, 1998; Myers, Feltz
et al., 2004; Myers, Payment, & Feltz, 2004) have demonstrated that teams with
stronger judgments regarding their coordinative capabilities outperform and persist
longer than teams with weaker judgments. Feltz and Lirgg (1998) investigated
relationships among aggregated self-efﬁcacy, aggregated collective efﬁcacy, and team
performance with six male intercollegiate ice hockey teams over the course of a 32
game season. They found that aggregated collective efﬁcacy predicted team
performance within teams and across games, but aggregated self-efﬁcacy did not.
Myers, Feltz et a1. (2004) replicated these ﬁndings with 10 intercollegiate American
football teams and further extended this line of inquiry by demonstrating that

aggregated collective efﬁcacy was a positive predictor of subsequent offensive

4o

performance within weeks and across teams. Myers, Payment et a1. (2004) examined
the inﬂuence of collective efﬁcacy on team performance over the course of a
competitive ice hockey season within weekends with 12 female intercollegiate teams.
The authors improved upon earlier research by statistically controlling for previous
team performance. After controlling for Friday team performance, the inﬂuence of
Saturday collective efﬁcacy on Saturday team performance was positive and moderate.
Taken together, these ﬁndings indicate that the magnitude of collective efﬁcacy is
positively related to team performance.

A primary limitation of previous collective efﬁcacy—team performance studies is
that the magnitude of collective efﬁcacy has been emphasized, while the degree of
within-team variability around the central belief has been neglected in the primary
analyses. It has been proposed that collective efﬁcacy dispersion may moderate the
relationship between collective efﬁcacy magnitude and team performance such that, the
relationship will be stronger when members’ beliefs are more interrelated than when
they are dispersed (Gully et al., 2002; Kozlowski et al., 1999). Research has
demonstrated that the relationship between aggregated group perceptions and team
performance is moderated by the degree of within-team dispersion. As discussed
previously, Schneider et a1. (2002) examined whether climate strength moderated the
relationship between bank employee ratings of service climate and customer
perceptions of service quality. They found that interaction between climate strength for
managerial practices and average climate perceptions explained unique variance in
customer satisfaction services after controlling for main effects. As discussed earlier in

this chapter, there is also some empirical evidence that suggests collective efﬁcacy

41

dispersion moderates the collective efﬁcacy-team performance relationship (Arthur et
al., 2007). The implication of this study is that the relationship between collective
efﬁcacy and team performance cannot be fully understood without consideration of
collective efﬁcacy dispersion.

While the majority of research has examined behavioral outcomes of collective
efﬁcacy such as team effectiveness, beliefs regarding the group’s capabilities are also
posited to inﬂuence team cognitive and affective states. For instance, collective efﬁcacy
inﬂuences team attributions (Chow & Feltz, 2008), shared mental models (Peterson,
Mitchell, Thompson, & Burr, 2000), and precompetitive cognitive anxiety (Greenlees et
al., 1999). In addition, teams with a stronger sense of efﬁcacy set more difﬁcult group
goals and are more committed to those goals (Mulvey & Klein, 1998).

Just as teams with high collective efﬁcacy tend to have members who are more
committed to collective goals, such teams should also have members who are more
attached to playing with their team. Conversely, on teams with low collective efficacy,
group members may be less committed to their team and as a result, invest less time and
effort into group endeavors. Previous research has found that teams with higher
collective efﬁcacy have greater team viability, which reﬂects group members’
willingness to remain with the group (Pescosolido, 2003). A construct that is similar to
team viability is commitment. Sport commitment has been deﬁned as “a psychological
construct representing the desire and resolve to continue sport participation” (Scanlan et
al., 1993, p. 6). It is a motivational force for continued involvement, and reﬂects an
important psychological underpinning of persistence. While the magnitude of collective

efﬁcacy may exert a direct inﬂuence on team commitment, the relationship may depend

42

on the degree of within-team variability in group members’ beliefs. Shared beliefs
strengthen team identity and the value of team membership, which reduces
psychological withdrawal (Roberson & Colquitt, 2005). Previous studies have found
that perceptual consensus interacts with aggregated perceptions to affect the
predictability of team affective states. For instance, Gonzalez-Roma et al. (2002)
investigated whether climate strength moderated the relationship between work units’
aggregated climate perceptions and their collective affective responses. They found that
climate strength in innovation moderated the inﬂuence of work units’ climate on
average satisfaction and commitment, and that climate strength in goals orientation
moderated the inﬂuence of work units’ climate on average commitment. The authors
reasoned that strong consensus fosters uniform affective responses, whereas weak
consensus yields a larger variability in the associated affective responses.
Summary

Previous research has focused exclusively on the magnitude of collective
efﬁcacy, while treating the degree of within-team variability as measurement error.
However, collective efﬁcacy dispersion may provide insight into team functioning and
performance beyond the magnitude of collective efﬁcacy. A dispersion theory examines
collective efﬁcacy dispersion as a dependent variable with hypothesized antecedents
and as a moderating variable with proposed team-related outcomes. Social network
analysis provides techniques to measure the patterns of social interaction within teams
which can be used to compare differences in collective efﬁcacy dispersion between
teams. Not only is there a lack of understanding regarding the emergence of within-

team variability in group members’ collective efﬁcacy beliefs, but also the

43

consequences of collective efﬁcacy dispersion. The usefulness of a dispersion theory of
collective efﬁcacy can be recognized by demonstrating that it predicts important team

outcomes such as performance and commitment.

44

CHAPTER 3
METHOD
Participants

Teams from the Midwestern and Eastern regions of the United States were
contacted for recruitment in the study. Initially, 50 teams agreed to participate in this
study and completed Time 1 measures. Of these teams, 46 completed both the Time 1
and Time 2 questionnaires. A MANOVA was run to determine whether the four teams
that did not return the Time 2 questionnaires differed from the teams that completed
both sets of questionnaires on various demographic, psychological, and performance
variables (e.g., experience, years on the team, team size, collective efﬁcacy, team
commitment, and winning percentage). Results revealed that there were no signiﬁcant
differences on any of these variables. Thus, only the 46 teams that completed both Time
1 and Time 2 measures were used in the study and subsequent analyses.

The participants in this study were athletes (N = 763) and their head coaches (N
= 46) from intercollegiate women’s softball teams. Starters and nonstarters as well as
position players and pitchers were included in this study. Athletes ranged in age ﬁom
18 to 24 years (M = 19.79, SD = 1.30), were members of their respective teams for 1 to
5 years (M = 2.09, SD = 1.08), and had played softball for 1 to 20 years (M = 12.20, SD
= 3.49). Coaches (Female = 38, Male = 8) ranged in age from 24 to 68 years (M =
39.58, SD = 11.94), had coached softball for 2 to 35 years (M = 14.09, SD = 9.26), and
were with their respective teams for 1 to 27 years (M = 8.37, SD = 7.56. The majority of
teams in the study were from Division II and 111 (Division I = 1, Division II = 12,

Division III = 30, NAIA = 3). Team size ranged from 10 to 24 players (M = 16.59, SD =

45

3.16), team mean playing experience ranged from 8.27 to 15.13 (M= 12.16, SD = 1.33),
and team mean number of years on the team ranged from 1.57 to 2.58 (M = 2.09, SD =
.24).

Procedure

Permission to conduct this study was obtained from the Institutional Review
Board for Human Subject Research. Following approval, coaches were contacted via
email requesting permission for their team to participate in this study. An explanation of
the purpose and procedure of the research was provided to coaches of teams who agreed
to participate in this study. Coaches were asked to provide information regarding the
date of their ﬁrst competition (to ensure that the Time 1 surveys were delivered) and the
number of athletes on their team. In addition, because the sociometric instrument
requires team members’ names, coaches were asked to submit a team roster. Informed
consent was obtained from all players and head coaches prior to data collection.

Data collection involved two time points. Time 1 data were collected at the
beginning of the season, prior to the ﬁrst scheduled competition. This time frame was
selected to allow team members to develop an initial sense of collective efﬁcacy that
was not inﬂuenced by competitive feedback or team performance from the current
season. It provided a baseline measure that was constant across all teams. At Time 1,
athletes were administered demographics, collective efﬁcacy, and team commitment
questionnaires, while coaches completed demographics and collective efﬁcacy. Time 2
data were collected at the middle of the season, after teams had played half of their
scheduled competitions. Questionnaires that were administered to athletes at Time 2

included demographics, collective efﬁcacy, team commitment, and the social network

46

instrument. At Time 2, coaches were administered demographics, collective efficacy,
and social network questionnaires. Team performance data in terms of winning
percentage were obtained at two time points: current winning percentage (collected at
Time 2: middle of season) and overall winning percentage (collected after all scheduled
competitions had been completed). In addition, team statistics pertaining to runs scored
and runs allowed were collected at the same time points (middle of the season and end
of the season).

Questionnaires were administered before or after a selected practice, or at a team
meeting, which was neither immediately before nor after a competition in order to avoid
competition-speciﬁc responses. An athletic trainer was responsible for administering the
questionnaires to athletes and the head coach. Participants were guaranteed
conﬁdentiality of their responses and were instructed to complete the questionnaires
individually without conversing with teammates. Following completion, questionnaires
were collected by the athletic trainer, placed in a sealed envelope to further ensure
conﬁdentiality, and mailed to the principal investigator using a prepaid postage return
envelope.

Measures

Demographics. A demographic questionnaire was used to obtain background
information from players and coaches. The athlete demographic questionnaire at Time 1
is presented in Appendix A. Items included age, NCAA Division, year in school (e.g.,
ﬁrst-year, sophomore, junior, senior, ﬁfth-year), total number of years playing softball,
position(s) played, number of years on the team, and starter status. In addition, athletes

were asked to indicate whether they were a team captain and whether they considered

47

themselves to be a leader. The Time 2 athlete demographic questionnaire (see Appendix
B) was similar to Time 1 with the exception of three additional items that assessed
players’ conﬁdence in their coach’s ability to: (a) communicate effectively with players,
(b) build the mental skills of players, and (c) make critical coaching decisions. These
three items were selected because of their congruence with the collective efﬁcacy
measure used in this study. Responses were made on a 5-point scale ranging ﬁ'om 1
(cannot do at all) to 5 (highly certain can do). A conﬁdence in coach score for each
individual was calculated by averaging each athlete’s responses to the 3-items, while a
conﬁdence in coach score for each team was calculated by aggregating team members’
scores to the group-level. The internal consistency reliability for the conﬁdence in
coach scale was .92.

The coach demographic questionnaire is presented in Appendix C. Items
included gender, age, number of years coaching the team, total number of years
coaching softball, number of years playing experience, highest level of playing
experience, and team’s win/loss record. In addition, coaches were asked to rate the
physical ability and teamwork ability of the athletes on their team this year, and the
overall ability of the teams on their schedule on a scale from 1 (very poor) to 5
(excellent).

Collective eﬁicacy. The collective efﬁcacy measure comprised lO-items and was
developed in accordance with Bandura’s (2006) recommendations. A conceptual
analysis was conducted in consultation with players and coaches to identify important
team performance competencies relevant to success in softball. The collective efﬁcacy

questionnaire is presented in Appendix D. Players and coaches were asked to rate how

48

certain they were that their team as a whole could: (a) communicate well as a unit, (b)
regain mental focus after an error/mistake, (c) avoid walking opposing batters, (d)
consistently throw strikes, (6) outscore opponents, (f) consistently put the ball in play,
(g) have a high ﬁelding percentage, (h) hit with runners in scoring position, (i)
successfully lay down bunts, and (i) make good decisions on the base paths. Although
ll-point rating scales (i.e., 0% to 100% with intervals of 10%, or 0 to 10 with intervals
of 1) have been recommended (Bandura, 2006), studies have found that efﬁcacy scales
employ too many categories (Zhu & Kang, 1998; Zhu, Updyke, & Lewandowski,
1997). An Optimal rating scale offers a more accurate true score estimate of within-team
variability (Linacre, 2002). Scales with more options can produce higher levels of
variance simply because of properties of the scale (Brown et al., 1996). Collective
efﬁcacy instruments are particularly susceptible to this problem because most
respondents only use the upper end of the rating scale (Myers & Feltz, 2007). Thus,
collective efﬁcacy ratings were made on a 5-point scale ranging ﬁ'om 1 (cannot do at
all) to 5 (highly certain can do). A collective efﬁcacy score for each athlete and coach
was calculated by averaging each person’s responses to the lO-items, while a collective
efﬁcacy score for each team was calculated by aggregating team members’ scores to the
group-level. lntemal consistency reliabilities for the collective scale were .89 at Time 1
and .89 at Time 2 for athletes. For the collective efﬁcacy scale rated by coaches, the
internal consistency reliabilities were .85 at Time 1 and .88 at Time 2.

Research that has conceptualized dispersion or within-team agreement as a focal

variable has indexed the construct using rwg (Arthur et al., 2007) and standard deviation

(Klein et al., 2001; Schneider et al., 2002; Zohar & Tenne-Gazit, 2008). While cogent

49

arguments have been provided in support of both conceptualizations, collective efﬁcacy

dispersion was operationalized as the standard deviation of members’ collective

efﬁcacy beliefs because rwg can exceed 1.00 on occasion and because most people think

of dispersion in terms of standard deviation (Lindell & Brandt, 2000). Following
procedures advanced by Klein and colleagues (2001), collective efﬁcacy dispersion was
calculated by averaging the standard deviation of group members’ responses to each
collective efﬁcacy item. Averaging the standard deviations across the collective
efﬁcacy items rather than simply calculating the standard deviation of members’ mean
collective efﬁcacy scores provides a more precise and comprehensive measure of
dispersion. The dispersion index was normalized to account for team size differences by
using the following equation: Log (SD + 1/2n), with higher scores reﬂecting higher
levels of dispersion and lower scores reﬂecting higher levels of consensus.
Commitment. The team commitment questionnaire is presented in Appendix E.
Team commitment was assessed using a modiﬁed 4-item measure of sport commitment
(Scanlan et al., 1993). Sport commitment can be assessed at various levels of analysis
such as commitment to a particular team, to a particular sport, or to sport in general. For
purposes of this study, sport commitment was conceptualized as commitment to the
team. Items included (a) “I am dedicated to playing with this team” (b) “It would be
hard for me to quit playing with this team” (0) “I am determined to keep playing with
this team” and (d) “I am willing to do almost anything to keep playing with this team.”
Ratings were made on a 5-point scale ranging from 1 (not true at all for me) to 5
(completely true for me) with higher scores representing greater commitment to the

team. A commitment score for each athlete was calculated by averaging each player’s

50

 

responses to the 4-items, while a commitment score for each team was calculated by
aggregating team members’ scores to the group-level. Evidence for the factorial and
discriminant validity and reliability of the commitment scale has been demonstrated
with youth sport participants (Scanlan, Simons, Carpenter, Schmidt, & Keeler, 1993).
The sport commitment scale has also been used with an adult sample of university staff
and students engaging in exercise classes (Wilson, Rodgers, Carpenter, Hall, Hardy, &
Fraser, 2004) and adult rugby players of recreational, amateur, university, and
professional clubs (Boardley, Kavussanu, & Ring, 2008). lntemal consistency
reliabilities for the team commitment scale were .90 at Time 1 and .93 at Time 2.

Social network variables. The network data were collected through a
sociometric instrument and measured as complete networks with each participant
referring to all other team members when responding to a network item. Complete
network analysis obtains all interactions among actors in a group which produces an
actor-by-actor matrix of relational values. Data were recorded in the form of an
adjacency matrix where each actor was assigned both a column and a row. Athletes and
head coaches received an alphabetized list of every team member (including the head
coach for athletes) and asked to identify those with whom they shared each kind of
relation. The sociometric questionnaire is presented in Appendix F. The communication
network item asked participants to rate how ﬁequently they talked with each of their
team members on subjects that were related to the team’s functioning such as discussing
the team’s strengths and weaknesses, performances, or goals. Respondents were asked
to circle the appropriate number next to each member’s name using a scale ranging

from 1 (very little) to 5 (a great deal). The friendship network item asked participants to

51

indicate who they considered to be a very good, someone they saw socially outside of
team—related activities. Respondents were asked to place an X next to the names of each
team member who they considered a ﬁiend. For the friendship item, participants were
allowed to select as many members as applicable.

Structural density is the ratio between the number of ties in a network and the
total number of possible ties in a network. Because directed relations were assessed
(e.g., Person A might indicate that Person B was a ﬁiend, but Person B might not
indicate that Person A is a friend), density was calculated by the equation: l/n(n — 1), '
where l was the number of lines present and n was the number of actors within the
network. The density coefﬁcient can range from 0 (no density) to 1 (complete density).
A binary format was used to calculate network density. Because the communication
scale was value-based, the ratings were dichotomized. Ratings of 1, 2, and 3 received a
value of 0, while ratings of 4 and 5 received a value of l. Ucinet software was used to
compute the density coefﬁcient for each team (Borgatti, Everett, & Freeman, 2002).

Network centralization is the sum of the differences between the largest
centrality score and all other observed centrality scores, divided by the maximum
possible sum of differences in actor centrality (W asserman & Faust, 1994).
Centralization was calculated using Freeman’s (1979) degree-based centralization

index:
* 2
CD = 2 [C001 )- C0010] / (3-1) ,
it
where C D(n ) is the centrality of the most central node, C D(n,-) is the centrality of actor

i, and g is the number of actors. Because relations were measured as directional, the

. 2 . . .
denominator was (g-l) instead of (g-l) x (g-2). The centralization score can range from

52

0 (every member is connected to every other member) to 1 (all members are connected
to only 1 member). Ucinet software was used to compute the centralization coefﬁcient
for each team (Borgatti et al., 2002).

Team performance. Team performance was conceptualized as a team’s winning
percentage and was calculated by dividing the number of wins by the number of games
a team had played. It is a standardized measure that takes into account the number of
games a team had played prior to completing the self-report questionnaires. Win/loss
records for each team were obtained from the head coach and veriﬁed through team and
conference websites.

Run diﬂerential. Although winning and losing are absolute measures of team
performance, winning percentage may not necessarily be the best method of assessing
how well a team performs because not all wins and losses are equal. For instance, losing
by one run is seemingly better than losing by ten runs. Thus, in addition to winning
percentage, run differential was calculated for all teams by subtracting the number of
runs allowed from the number of runs scored. This score was divided by the number of
games played to produce a run differential per game score because certain teams played
a different number of competitions than other teams. The number of runs scored and
runs allowed for each team were obtained from team and conference websites.

Team size. The number of players on a team can affect group level social
network metrics such as density. Therefore, Friedkin (1981) recommends that group
size and density should be analyzed simultaneously. Team size may also inﬂuence
perceptions of collective efﬁcacy. According to Zaccaro et al. (1995), teams with more

members may have higher perceptions of collective efﬁcacy because there are greater

53

amounts of resources available in larger groups. Conversely, team size may be
negatively related to collective efﬁcacy because coordination becomes more difﬁcult
and the likelihood of social loaﬁng and clique formation increases as the size of the
group increases. Consequently, team size was assessed to examine relationships
between network variables and collective efﬁcacy dispersion. Team size was measured
as the total number of players on a given team as reported by the coach and was veriﬁed
through team rosters.

Data Analyses

Descriptive statistics (i.e., mean, standard deviation) and bivariate correlations
were calculated for all variables. lntemal consistency reliabilities were computed for
scales that had multiple items.

A multiple regression was run to test Hypotheses 1-6 (i.e., communication
density, friendship density, communication centralization, ﬁiendship centralization,
communication subgroup, and friendship subgroup). Conﬁdence in the coach was
included as a predictor variable in the preliminary model, but it was unrelated to
collective efﬁcacy dispersion and thus, was removed ﬁ'om the ﬁnal model. Collective
efﬁcacy dispersion at Time 2 was regressed on collective efﬁcacy dispersion at Time 1
and the six network variables. To test Hypothesis 7 (the moderating inﬂuence of
dispersion on the relationship between collective efﬁcacy and team performance), a
multiple regression was run at two time points (middle of the season and end of the
season). For instance, middle of the season winning percentage was regressed on
collective efﬁcacy, dispersion, and the interaction between collective efﬁcacy and

dispersion assessed at the beginning of the season. Similar analyses were conducted for

54

run differential. The collective efﬁcacy and dispersion variables were centered to reduce
multicollinearity with the interaction term. Similarly, Hypothesis 8 (the moderating
inﬂuence of dispersion on the relationship between collective efﬁcacy and team
commitment) was tested using multiple regression analyses with centered collective
efﬁcacy, centered dispersion, and the collective efﬁcacy by dispersion interaction as the
independent variables. Separate analyses were conducted for team commitment at Time
1 and Time 2. Because the team commitment scale was signiﬁcantly skewed at both
time points, a negative base-10 logarithm was applied in order to normalize the
distributions.

Research Questions 1 and 2 pertain to the occurrence of subgroups within teams.
In order to identify non-overlapping subgroups based on communication and friendship
ties within teams, Klique Finder was used. Klique Finder is based on Frank’s (1995,
1996) network clustering algorithm and determines whether there is evidence that actors
engage in exchanges within subgroups at a rate that is unlikely to have occurred by

chance alone. To test for subgroup processes, a likelihood ratio test between two models

is compared: 1) Log (p[w,-,w= 1]/1-p[w,~,-' = 1]) = 90 + 91b0,, same groupi," and 2) Log
(P[Wii'=1]/1’P[Wii’:1]) = 90 + eIbase same groupi,” + elsubgroup processes same

groupii '. A small p-value indicates that the null hypothesis that 6 1 subgroup processes is

zero can be rejected, which provides evidence that there are subgroups based on

interaction patterns. Using the approximate test of concentration of ties within

subgroups based on the size of 61 subgroup processes, each team was examined separately

to determine whether it contained subgroups. This analysis was performed for both

55

communication and ﬁiendship ties. A dichotomous variable was created where teams
with subgroups were coded as 1, while teams without subgroups were coded as 0.
Research Questions 3 and 4 relate to whether subgroups align by collective
efﬁcacy, while Research Questions 5 and 6 relate to whether individuals’ collective
efﬁcacy beliefs are inﬂuenced by members of their communication and friendship
subgroup. Based on the Klique Finder results, only teams that demonstrated evidence of
subgroups were included in the analyses examining these research questions. Because
of the nested nature of the data (i.e., individuals within subgroups within teams),
multilevel modeling was used. Multilevel statistical techniques (e. g., hierarchical linear
modeling - HLM) reduce the problems associated with single level analyses by
enabling the researcher to simultaneously examine relationships at each level and across
levels, while determining the amount of variation at each level (Raudenbush & Bryk,
2002). Separate multilevel analyses were conducted for communication and ﬁiendship
subgroups where collective efﬁcacy at Time 2 was the dependent variable. The ﬁrst step
of model building involved imposing an unconditional model where no individual,
subgroup, or team level predictors were entered into the model. The primary purpose of
ﬁtting the unconditional model was to determine the amount of variation in collective
efﬁcacy that existed at each level (Research Questions 3 and 4). After running the
unconditional model, prior levels of collective efﬁcacy were entered as predictors at
each level (individual, subgroup, and team) to examine whether collective efﬁcacy
beliefs were inﬂuenced by subgroup processes (Research Questions 5 and 6). All
predictor variables were grand-mean centered. The multilevel model for collective

efﬁcacy at Time 2 was:

56

Level 1:
collective efﬁcacy 2icj = “ocj + rtlcjcollective efﬁcacy licj + eijk
Level-2:

710cj = BOOj + [ion-collective efﬁcacy 1 subgroup meancj + T0cj

7ticj = I310j
Level-3:

BOOj + 7000 + y001collective efﬁcacy 1 team meanj + uOOj

Borj =Y010

Bioj =7100

57

CHAPTER 4
RESULTS

Descriptives

Descriptive statistics pertaining to athlete and coach demographics are presented
in Table l and Table 2, respectively. Means and standard deviations for selected team
level variables are presented in Table 3. Collective efﬁcacy scores ranged from 3.38 to
4.64 (M= 3.89, SD = .32) at Time 1 and ﬁom 2.95 to 4.79 (M= 3.82, SD = .41) at Time
2. Dispersion scores ranged from -.28 to —.01 (M = -.15, SD = .05) at Time 1 and from -
.44 to .00 (M = -.17, SD = .07) at Time 2. Team commitment scores ranged from 3.52 to
5.00 (M = 4.74, SD = .25) at Time 1 and ﬁ'om 3.44 to 5.00 at Time 2. Conﬁdence in the
coach scores ranged from 2.27 to 4.98 (M = 3.94, SD = .64). For the social network
variables, teams tended to have moderately dense communication networks (M = .47,
SD = .11) and friendship networks (M = .65, SD = .12), and more decentralized
communication (M = .19, SD = .05) and ﬁiendship (M = .26. SD = .08) networks.
Klique Finder identiﬁed 15 (32.6%) teams that had communication subgroups and 13
(28.3%) teams that had friendship subgroups. Each communication subgroup contained
3 to 14 members, whereas each ﬁiendship subgroup contained 3 to 16 members. Similar
to the multilevel analysis procedures described in the data analyses section of Chapter 3,
an unconditional model was run with years on the team as the outcome variable to
determine if subgroups aligned by number of years on team. Results revealed that 34%
of the variance was due to communication subgroup membership (p < .001), whereas

32% of the variance was due to ﬁiendship subgroup membership (p < .001 ). Thus,

7
subgroups tended to align by number of years on the team. In addition, a x" test of

58

independence was conducted to determine whether subgroups also aligned by starting
status (starter, non-starter). Results revealed that communication subgroups aligned by
starting status (p < .01), but friendship subgroups did not.

Bivariate correlations for selected team level variables are presented in Table 4.
Pearson Product Moment correlations revealed that collective efﬁcacy was signiﬁcantly
and positively related to team performance and run differential. In addition, there was a
signiﬁcant and positive relationship between collective efﬁcacy and team commitment
at both time points. Collective efﬁcacy was signiﬁcantly and negatively associated with
dispersion with stronger correlations occurring at similar time points (e. g., Collective
Efﬁcacy 1 and Dispersion 1; Collective Efﬁcacy 2 and Dispersion 2). Interestingly, the
correlation between Dispersion l and Dispersion 2 (r = .56) was weaker than the
correlation between Collective Efﬁcacy 1 and CollectiveEfﬁcacy 2 (r = .77), indicating
that collective efﬁcacy dispersion was less stable over time. There was a signiﬁcant and
negative correlation between dispersion and team performance, as well as dispersion
and run differential. For relationships between dispersion and team commitment, only
Dispersion 2 and Team Commitment 2 were signiﬁcantly related. Conﬁdence in the
coach was signiﬁcantly and negatively associated with collective efﬁcacy dispersion.
Although modest correlations were found between the network variables and
dispersion, only communication subgroup was signiﬁcantly associated. There was a
signiﬁcant and negative relationship between ﬁiendship centralization and collective
efﬁcacy at Time 1 and 2. Low to moderate correlations were found between the

network measures, with signiﬁcant correlations ranging from .36 to .65.

59

Bivariate correlations between team level demographic variables, dispersion,
and network variables were also performed to determine whether certain demographic
variables should be included as control variables in subsequent analyses. Team size and
mean years on the team did not signiﬁcantly correlate with dispersion or any of the
social network variables.

For the sociometric questionnaire, team response rates ranged from 65% to
100% (M = 96%, SD = .07) with 27 teams (59%) providing complete data. Consensus
estimates were performed for all variables that were aggregated to the team level.

Estimates were calculated across the items of each scale using the within-group

interrater agreement index, rwg, (James et al., 1984), assuming no response bias and

discrete data. The number of response categories for collective efﬁcacy, team

commitment, and conﬁdence in the coach met the discrete assumption of seven plus or

minus two (James et al., 1984). ng values for collective efﬁcacy ranged from .75 to .89

(M= .82, SD = .03) at Time 1 and from .75 to .91 (M= .83, SD = .04) at Time 2. For

team commitment, rwg values ranged from .76 to 1.24 (M = .90, SD = .09) at Time 1

and from .76 to 1.16 (M = .86, SD = .09) at Time 2. ng values for conﬁdence in the
coach. ranged from .75 to .99 (M = .81, SD = .06). Furthermore, correlations between
rwg and SD for collective efﬁcacy beliefs were -.99 at Time 1 and -.97 at Time 2,

indicating that the choice of dispersion index was inconsequential.
Social Network Variables and Dispersion
Hypotheses 1-6 posit that network variables inﬂuence collective efﬁcacy

dispersion, while controlling for prior levels of collective efﬁcacy dispersion. To test

60

these hypotheses, dispersion at Time 2 was regressed on dispersion at Time 1,
communication density, friendship density, communication centralization, ﬁiendship
centralization, communication subgroup, and ﬁiendship subgroup. Results are presented
in Table 5. Prior collective efﬁcacy dispersion signiﬁcantly predicted dispersion at
Time 2 (ﬂ = .49, p < .001). For the network variables, communication and ﬁiendship
density, as well as friendship centralization did not signiﬁcantly predict collective
efﬁcacy dispersion. Thus, Hypotheses 1, 2, and 4 were not supported. However, the
inﬂuence of communication centralization on collective efﬁcacy dispersion (Hypothesis
3) approached statistical signiﬁcance (,6 = .22, p =. .08), indicating that teams with less
centralized communication networks had lower levels of collective efﬁcacy dispersion
than teams with more centralized communication networks. Subgroups based on
communication and friendship ties demonstrated signiﬁcant, but differential effects.
Communication subgroups were a signiﬁcant and positive predictor of dispersion (,8 =
.45, p < .01 ), whereas ﬁiendship subgroups were a signiﬁcant and negative predictor of
dispersion (,8 = -.31, p < .05). Thus, Hypothesis 5 was supported and Hypothesis 6 was
in the unexpected direction. Teams with communication subgroups had higher levels of
dispersion, while teams with ﬁiendship subgroups had lower levels of dispersion. The

regression model explained 43% of the variance in collective efﬁcacy dispersion at

Time 2 (AR2 = .20, F change = 2.66, p < .05).

Multilevel Model of Collective Eﬁicacy Based on Subgroups
In order to determine the amount of variation in collective efﬁcacy that existed
at the subgroup level (Research Questions 3 and 4) and to examine whether collective

efﬁcacy beliefs were inﬂuenced by subgroup’s prior mean levels of collective efﬁcacy

61

(Research Questions 5 and 6), multilevel modeling was used. As discussed in the
method section, separate multilevel analyses were conducted for communication and
friendship subgroups, respectively. For the multilevel model of collective efﬁcacy based
on communication subgroups, there were 274 individuals within 47 subgroups within

15 teams. Results from the unconditional model revealed that 62% of the variance in
collective efﬁcacy was at the individual level, 5% was at the subgroup level, and 33%

was at the team level. While only 5% of the variance was due to communication

subgroup membership, the variance component was signiﬁcant (x2 = 46.78, p < .05),

providing some support that communication subgroup membership was related to
collective efﬁcacy beliefs. After running the unconditional model, collective efﬁcacy at
Time 1 was included as a predictor variable at each level of analysis to examine
whether collective efﬁcacy beliefs at Time 2 were inﬂuenced by subgroup’s prior mean
level of collective efﬁcacy. That is, collective efﬁcacy at Time 2 was predicted by
individuals’ prior collective efﬁcacy, subgroups’ prior collective efﬁcacy, and teams’
prior collective efﬁcacy. Results ﬁ'om the multilevel model based on communication
subgroups are presented in Table 6. Prior collective efﬁcacy at the individual (,8 = .57, p
< .001) and team (,8 = .62, p < .05) level were signiﬁcant and positive predictors of
collective efﬁcacy at Time 2. However, subgroup’s prior mean level of collective
efﬁcacy did not predict collective efﬁcacy at Time 2, after controlling for prior levels of
collective efﬁcacy at the individual and team level.

Similar analyses were performed for ﬁiendship subgroups. For the multilevel
model of collective efﬁcacy based on friendship subgroups, there were 204 individuals

within 30 subgroups within 13 teams. Results from the unconditional model revealed

62

that 51% of the variance in collective efﬁcacy was at the individual level, 13% was at

the subgroup level, and 36 % was at the team level. The variance component associated
with ﬁiendship subgroups was signiﬁcant (x2 = 42.65, p < .01), providing support that

ﬁiendship subgroup membership was related to collective efﬁcacy beliefs. Results from
the multilevel model based on friendship subgroups with prior levels of collective
efﬁcacy included at each level are presented in Table 7. Prior collective efﬁcacy at the
individual level was a signiﬁcant and positive predictor of collective efﬁcacy at Time 2
(,8 = .43, p < .001), while prior collective efﬁcacy at the team level was unrelated to
collective efﬁcacy at Time 2. Further, the inﬂuence of subgroups’ prior mean levels of
collective efﬁcacy approached statistical signiﬁcance (,8 = .25, p = .13).
Moderating Inﬂuence of Dispersion on Team Performance

To test whether dispersion moderated the relationship between collective
efﬁcacy and team performance (Hypothesis 7), a regression was run for each time point
(Collective Efﬁcacy 1 and Team Performance 2/Run Differential 2; Collective Efﬁcacy
2 and Team Performance 3/Run Differential 3). In order to reduce multicollinearity, the
predictors (collective efficacy, dispersion, and the collective efﬁcacy by dispersion
interaction) were centered. The results of the regression analyses are presented in Table
8. For Team Performance 2 and 3, the collective efﬁcacy by dispersion interaction was
not signiﬁcant, indicating that dispersion did not moderate the relationship between
collective efﬁcacy and team performance. Instead, collective efﬁcacy was the sole
predictor ofTeam Performance 2 (8 = .55, p < .01) and 3 (,8 = .67,p < .001). The
regression models for Team Performance 2 and 3 accounted for 32% and 58% of the

variance, respectively. For Run Differential 3, Collective Efﬁcacy 2 was a signiﬁcant

63

and positive predictor (8 = .5 7, p < .001). Furthermore, collective efﬁcacy dispersion
emerged as a signiﬁcant predictor (8 = -.31, p < .05) of run differential, while the
collective efﬁcacy by dispersion interaction approach statistical signiﬁcance (8 = .22, p
= .08). Thus, Hypothesis 7 was partially supported, though in the unexpected direction.
The model for Run Differential 3 explained 52% of the variance.
Moderating Inﬂuence of Dispersion on Team Commitment

To test whether dispersion moderated the relationship between collective
efﬁcacy and team commitment (Hypothesis 8), a regression was run for each time point
(Collective Efﬁcacy 1 and Team Commitment 1; Collective Efﬁcacy 2 and Team
Commitment 2). Similar to the team performance/run differential regression analyses,
the predictors (collective efficacy, dispersion, and the collective efﬁcacy by dispersion
interaction) were centered to reduce multicollinearity. The results of the regression
analyses are presented in Table 9. For Team Commitment 1 and 2, the collective
efﬁcacy by dispersion interaction was not signiﬁcant, indicating that dispersion did not
moderate the relationship between collective efﬁcacy and team commitment. Thus,
Hypothesis 8 was not supported. Instead, collective efﬁcacy was the sole predictor of
Team Commitment 1 (8 = .43,p < .05) and 2 (8 = .55,p < .01). The regression models
explained 14% and 43% of the variance in team commitment, respectively.
Summary of Hypotheses and Results

Hypotheses 1-6 investigated the inﬂuence of social network variables on
collective efﬁcacy dispersion, while controlling for prior levels of collective efﬁcacy
dispersion. Hypothesis 1 posited that communication density would be negatively

related to collective efﬁcacy dispersion. No support was found for Hypothesis 1.

64

Hypothesis 2 posited that friendship density would be negatively related to collective
efﬁcacy dispersion. No support was found for Hypothesis 2. Hypothesis 3 was that
communication centralization would be positively related to collective efﬁcacy
dispersion. Partial support was found for Hypothesis 3 in that the inﬂuence of
communication centralization approached statistical signiﬁcance. Hypothesis 4 was that
ﬁiendship centralization would be negatively related to collective efﬁcacy dispersion.
No support was found for Hypothesis 4. Hypothesis 5 posited that ﬁiendship subgroups
would be positively related to collective efﬁcacy dispersion. Although statistically
signiﬁcant, the inﬂuence of friendship subgroups was in the unexpected direction.
Hypothesis 6 posited that communication subgroups would be positively related to
collective efﬁcacy dispersion. Support for Hypothesis 6 was found. Hypotheses 7 and 8
examined the moderating inﬂuence of collective efﬁcacy dispersion. Hypothesis 7 was
that collective efﬁcacy dispersion would moderate the relationship between aggregated
collective efﬁcacy and team performance. Partial support was found for Hypothesis 7,
though in the unexpected direction. The relationship between aggregated collective
efﬁcacy and run differential was stronger for teams with high dispersion. Hypothesis 8
was that collective efﬁcacy dispersion would moderate the relationship between
aggregated collective efﬁcacy and team commitment. No support was found for

Hypothesis 8.

65

CHAPTER 5
DISCUSSION

Despite ample research on collective efﬁcacy in sport and organizational
psychology, little is known as to why some teams develop more interrelated collective
efﬁcacy beliefs over time and others do not, and the further implications this has on
team ﬁmctioning. In this study, a dispersion theory of collective efﬁcacy was developed
and investigated by examining the antecedents and moderating inﬂuence of collective
efﬁcacy dispersion. Theorists have suggested that interaction patterns represented as
social networks contribute to the emergence of shared beliefs, and that the relationship
between collective efﬁcacy and team-related outcomes depend on the degree of within-
team variability surrounding members’ judgments. Findings from this study contribute
to a better understanding of collective efﬁcacy dispersion. This chapter discusses the
ﬁndings of the current study, identiﬁes implications of these results, discusses strengths
and limitations of the study, and presents future research directions.
The Emergence of Collective Eﬂicacy Dispersion

The novel ﬁnding of this study was that subgroups based on communication and
ﬁiendship ties inﬂuenced collective efﬁcacy dispersion, even after controlling for prior
levels of dispersion. However, the direction of effect differed for communication and
friendship subgroups, respectively. Teams with communication subgroups had members
with more dispersed collective efﬁcacy beliefs than teams that did not have
communication subgroups. This ﬁnding is consistent with the theorizing of Brown et al.
(1996) who speculated that dispersion at the group level may result from subgroups

based on social interaction. While subgroups create patterns of local agreement, there

66

tends to be collective efﬁcacy dispersion in the team as a whole. In contrast, teams with
ﬁiendship subgroups had group members with less dispersed collective efﬁcacy beliefs
than teams without friendship subgroups. However, this ﬁnding appears to be a
statistical artifact. When examined separately from communication subgroups, the
inﬂuence of ﬁiendship subgroups on collective efﬁcacy dispersion was not signiﬁcant.
This is further supported by the bivariate correlations presented in Table 4, which
demonstrate that there was no relationship between ﬁiendship subgroups and collective
efﬁcacy dispersion (r = -.01).

There was some support for the notion that communication and friendship
subgroups aligned according to collective efﬁcacy beliefs. That is, individuals of
cohesive subgroups tended to have similar collective efﬁcacy beliefs in that there was a
signiﬁcant amount of variance in collective efﬁcacy at the subgroup level. However, the
amount of variance attributed to subgroup membership was rather low (5% for
communication subgroups and 13% for ﬁiendship subgroups). Further substantiating
this point was that individuals’ perceptions of collective efﬁcacy were not inﬂuenced by
their subgroup’s prior mean levels of collective efﬁcacy. If subgroup effects were
salient in determining collective efﬁcacy, then it would be expected that individuals
would conform to their subgroup’s normative belief over time as a result of frequent
interactions with proximal members. Indeed, a social network approach to subgroups
suggests that members of cohesive subgroups develop similar beliefs over time through
dense patterns of social interaction. However, results from the current study indicate

that collective efﬁcacy may not be a belief that is inﬂuenced by subgroup processes.

67

Perhaps having a small number of teams with the presence of subgroups provided a
limited test of subgroup effects.

Although descriptive in nature, an interesting ﬁnding was that the majority of
teams in this study did not have communication or friendship subgroups. Among the
teams that participated in the study (N = 46), 15 (33%) teams demonstrated evidence of
communication subgroups, while 13 (28%) teams demonstrated evidence of ﬁiendship
subgroups. Teams that had subgroups were similar to teams that did not have subgroups
in terms of coaching demographics (gender, age, years coaching the team, and coaching
experience) and division. In addition, the presence of subgroups was not associated with
team size, team experience, or team mean number of years on the team. Because
subgroups based on social interaction have not been examined in sport, it is difﬁcult to
know whether the percentages found are representative of sports teams in general. In
comparison to other settings such as schools or business, subgroups may be less
common among athletic teams because of increased opportunities for social interaction.
For instance, members of competitive sports teams often interact with each other on a
daily basis, either at strength training sessions, practice sessions, team meetings, or
competitions. The frequent opportunities to communicate and develop personal
relationships, which are inherent to team sports, may explain why most teams in this
study did not have subgroups. In addition, coaches and team captains try to build team
unity intentionally, as exempliﬁed by the popular slogan, “there is no I in team.”
Another explanation may be that communication and ﬁiendship subgroups are more
prevalent in different types of sports. For instance, sports characterized by larger team

sizes where team members are required to perform speciﬁc roles (e.g., football) may be

68

more likely to develop subgroups. Although the majority of teams in this study did not
have subgroups, the importance of investigating subgroups was demonstrated in that
teams with subgroups had different levels of collective efﬁcacy dispersion than teams
without subgroups. Thus, future studies should continue to examine subgroups based on
different social relations with athletic teams.

Results from this study provided partial evidence that the structure of
communication ties inﬂuences the degree of within-team variability in members’
collective efﬁcacy beliefs. Teams with more decentralized communication patterns had
less collective efﬁcacy dispersion than teams with more centralized communication
patterns. This ﬁnding is consistent with previous research that suggests decentralized
communication networks facilitate the spread of information and promote shared
understandings among group members (Cummings & Cross, 2003; Zohar & Tenne-
Gazit, 2008). In contrast to centralized communication networks which revolve around
a highly central actor, decentralized communication networks offer better social
diffusion processes because communication exchanges are unconstrained. Thus,
information regarding the team’s strengths and weaknesses, performances, and goals is
discussed openly among all network members, which contributes to a shared sense of
collective efﬁcacy at the team level. Alternatively, in teams with centralized
communication structures, central actors have disproportionate access to members’
attitudes and beliefs. A centralized communication conﬁguration may produce
collective efﬁcacy dispersion because peripheral members’ beliefs are isolated, and
only the central actor is involved in the majority of discussions about the team’s

functioning.

69

The proposed hypotheses related to network density and collective efﬁcacy
dispersion were not supported. Teams with dense communication and friendship
networks did not have less collective efﬁcacy dispersion than teams with sparse
networks. This ﬁnding is in contrast to previous research that suggests dense networks
promote homogenous perceptions among team members, whereas sparse networks
contribute to heterogeneous perceptions among team members (Zohar & Tenne-Gazit,
2008). An explanation for the inconsistent results may be attributed to the analysis used
in this study. Previous network studies in organizational psychology have conducted
separate analyses for density and centralization, and have employed cross-sectional
designs (e.g., Zohar & Tenne-Gazit, 2008). The current investigation simultaneously
examined several network variables (e.g., density, centralization, subgroups) and
controlled for prior levels of dispersion. When separate analyses are conducted, the
independent contributions of the predictor variables (i.e., density and centralization) can
not be determined. For instance, Sparrowe and colleagues (2001) found that network
density was unrelated to team performance, after accounting for network centralization.
Accordingly, in this study, network density did not inﬂuence collective efﬁcacy
dispersion, beyond the effects of network centralization and subgroups. Findings from
this study indicate that the structural patterns of ties (i.e., centralization and subgroups)
may be more salient antecedents of collective efﬁcacy dispersion than the magnitude of
ties (i.e., density).

Collective Eﬂicacy Dispersion as an Independent and Moderating Variable
The majority of studies that have examined the inﬂuence of collective efﬁcacy

on team performance have neglected the degree of within-team variability around the

70

group level mean. However, theorists have proposed that within-team variation in
members’ collective efﬁcacy beliefs moderates the relationship between collective
efﬁcacy magnitude and team performance (Gully et al., 2002). This study tested this
proposition and found partial support for the moderating inﬂuence of collective efﬁcacy
dispersion. For midseason and end of the season winning percentage, neither dispersion
nor the collective efﬁcacy by dispersion interaction was a signiﬁcant predictor beyond
the magnitude of collective efﬁcacy beliefs. However, teams with less collective
efﬁcacy dispersion at the middle of the season outscored their opponents (at the end of
the season) to a higher degree than teams with more collective efﬁcacy dispersion,
regardless of the magnitude of collective efﬁcacy beliefs. This suggests that in addition
to enhancing collective efficacy through sources of efﬁcacy information (i.e., mastery
experiences, vicarious experiences, verbal persuasion, and physiological/emotional
states), coaches and practitioners should focus on developing shared beliefs among team
members. Even when the magnitude of collective efﬁcacy is moderate, teams may
beneﬁt by achieving higher levels of agreement. As collective efﬁcacy beliefs become
more interrelated, conﬂict among teammates is reduced, which allows teams to focus on
coordinative processes that impact team ﬁrnctioning. Partial evidence was found for
collective efﬁcacy dispersion as a moderator of the collective efﬁcacy-run differential
relationship. However, the moderating effect was in the unexpected direction. Although
the relationship between middle of the season collective efﬁcacy and end of the season
run differential was positive for teams with low and high dispersion, the relationship
was stronger for teams with higher levels of dispersion. This ﬁnding is inconsistent with

previous research that found that the relationship between aggregated collective efﬁcacy

71

and team performance was positive only for teams with low dispersion (Arthur et al.,
2007) and suggests that the collective efﬁcacy-team performance relationship is robust.
Furthermore, this study addressed several of the limitations associated with previous
collective efﬁcacy dispersion research. For instance, Arthur and colleagues employed
an experimental design with all-male dyads, and used a 2—item collective efﬁcacy
questionnaire. This study was conducted in a natural setting with intact women’s
softball teams comprised of multiple members, and used a lO-item sport-speciﬁc
measure of collective efﬁcacy. Therefore, ﬁndings ﬁom the current investigation extend
previous research, and provide evidence for the construct validity of collective efficacy
dispersion.

In addition to team performance, dispersion was hypothesized to moderate the
relationship between collective efﬁcacy and team commitment such that, the association
would be stronger when group members’ beliefs were more interrelated. However, no
support was found for the moderating inﬂuence of dispersion, as only collective
efﬁcacy magnitude emerged as a signiﬁcant predictor of team commitment at the
beginning of the season and middle of the season. This ﬁnding is in contrast to previous
dispersion studies that suggest interrelated beliefs foster uniform affective responses
such as team commitment (Gonzalez-Roma et al., 2002). An explanation for the
inconsistent results may be the sample used in this study. Whereas previous studies
have used work units, the current study used softball teams where the majority of
athletes reported strong feelings of commitment to their team. As a result, there was
little variation between teams. Although a negative base-10 logarithm was applied to

the team commitment scale to normalize the distribution, there may not have been

72

enough differences between teams for dispersion to emerge as an important factor
beyond the magnitude of collective efﬁcacy. Nevertheless, the ﬁnding that teams with
stronger collective efﬁcacy beliefs were more committed than teams with weaker
collective efﬁcacy beliefs contributes to a fuller understanding regarding the
consequences of collective efﬁcacy. While previous studies have investigated the
inﬂuence of collective efﬁcacy on various team affective states (e.g., Chow & Feltz,
2008; Pescosolido, 2003; Peterson et al., 2000), none have examined team commitment.
When members are committed to continue playing with their team, they are more likely
to identify with their group and invest time and effort into group endeavors. Thus, team
commitment has implications for group cohesion, motivation, and
psychological/physical withdrawal. Findings from this study suggest that these factors
can be modiﬁed by fostering collective efﬁcacy beliefs within teams.
Implications

The current study has demonstrated that not only are aggregated collective
efﬁcacy beliefs important, but also the degree of within-team variability in group
members’ judgments. As team members’ collective efﬁcacy beliefs become more
similar over time, team performance tends to improve in a linear direction. However, it
cannot be assumed that the sources of aggregated collective efﬁcacy extend to
collective efﬁcacy dispersion. This study uncovered an underlying mechanism into why
some teams develop more interrelated beliefs, while others do not, that is, through the
pattern of social interaction. An implication of this ﬁnding is that coaches and applied
consultants can increase homogenous collective efﬁcacy beliefs within teams by

providing opportunities for interpersonal interaction and by focusing on social activities

73

that involve active participation from all team members. Employing activities and
techniques that facilitate the spread of information ﬂow such as team goal-setting,
where each member discusses her perspective about the team’s functioning, should
promote shared collective efﬁcacy beliefs at the team level. Misinterpretations of team
goals and previous performances may arise when members have limited access to the
attitudes and beliefs of others. However, it is important to note that extremely low levels
of collective efﬁcacy dispersion may be detrimental to team functioning and
performance. When team members hold identical beliefs about the team’s functioning,
they are less likely to engage in effective decision making processes and reappraise
strategies after performance setbacks. Having some doubt, especially in the preparatory
stages of performance, may actually increase motivational levels and facilitate the
development of skills (Bandura, 1997). Another implication of this study pertains to the
coach’s role in providing accurate feedback. Although performance outcomes provide a
common experience for members to judge the capabilities of their team, coaches are
often responsible for framing the results through encouragement and feedback. Coaches
who provide inconsistent feedback or only share their perspectives with certain
individuals such as assistant coaches or captains are likely to create within-team
variation in members’ collective efficacy beliefs which in turn, may impair team
functioning.
Strengths and Limitations

This study used a social network approach to understand how team members’
collective efﬁcacy beliefs become shared or dispersed over time. Scholars have
assumed that social interaction underlies consensus formation within teams (Jung &

Sosik, 2003), and have recommended that social network techniques be employed to

74

 

further advance this line of inquiry (Gibson, 1999; Chow & F eltz, 2008). An advantage
of social network techniques is that they describe the structure and pattern of social ties
within teams, and provide a more comprehensive measure of group interaction than
traditional instruments. Indeed, by utilizing a social network approach, I found that the
structural conﬁgurations of networks (i.e., centralization and subgroups) were more
salient in determining collective efﬁcacy dispersion than the mere intensity of social
interaction (i.e., density).

Another strength associated with the current investigation pertains to the
longitudinal nature of the design. The methodology used in this study improved upon
previous dispersion and network studies by accounting for prior levels of dispersion,
while examining the inﬂuence of social interaction. This limitation was addressed by
obtaining an initial measure of collective efﬁcacy at a time point that was constant
across all teams (i.e., beginning of the season). Previous organizational research has
employed cross-sectional designs, which made it difﬁcult to determine whether social
interaction contributed to within-team variability or whether within-team variability
contributed to social interaction. By including a prior estimate, I was able to determine
that social interaction patterns within teams inﬂuenced subsequent levels of collective
efﬁcacy dispersion above and beyond initial levels of dispersion.

Although there are several strengths associated with this study, there are some
limitations that should be mentioned. The ﬁrst limitation was the sample size. While the
number of teams that participated in this study (N = 46) could certainly be considered
large for sport psychology research, the sample size was less than ideal for statistical

purposes. Small sample sizes are a common problem in group research because a team

75

comprised of several members is essentially reduced to one participant. Some of the
proposed hypotheses tested in this study approached statistical signiﬁcance (e. g.,
communication centralization and dispersion, collective efﬁcacy by dispersion
interaction and run differential). Perhaps these hypotheses would have been supported
had there been more teams in the study. In addition, approximately 33% of the teams in
this study had subgroups. This signiﬁcantly reduced the sample size and degrees of
ﬁ'eedom in the multilevel model examining whether subgroups aligned by collective
efﬁcacy and whether subgroup processes inﬂuenced collective efficacy beliefs. Future
studies with larger sample sizes may have the power necessary to detect stronger team
differences, and should increase the likelihood of identifying a sufﬁcient number of
teams with subgroups.

The second limitation of the study pertains to the temporal proximity between
collective efﬁcacy and team performance measurement. Studies that have investigated
the relationship between collective efﬁcacy and team performance have typically
assessed collective efﬁcacy within 24 hr. prior to competition (e.g., Feltz & Lirgg,
1998; Myers, Feltz et al., 2004; Myers, Payment et al., 2004). This line of work has
used competition-speciﬁc measures of collective efﬁcacy (e.g., How conﬁdent are you
that your team can perform in the upcoming competition) and has examined its
inﬂuence on team performance in the competition referenced. An alternative to
competition-speciﬁc measures of collective efﬁcacy are general measures of collective
efﬁcacy. A general sense of collective efﬁcacy focuses on team’s beliefs in their
coordinative capabilities to perform at certain points in time. General measures have

been used in studies that have assessed collective efﬁcacy at multiple time points (e. g.,

76

 

Paskevich et al., 1999; Watson et al., 2001), rather than studies that have assessed
collective efﬁcacy prior to all competitions throughout a season (e.g., F eltz & Lirgg,
1998; Myers, Feltz et al., 2004; Myers, Payment et al., 2004). Because measures of
collective efﬁcacy could not be obtained prior to each competition, a general measure
was used in this study. As a result, ﬁndings related to the moderating inﬂuence of
collective efﬁcacy dispersion on team performance may be limited. Although
competition-speciﬁc measures are recommended in collective efﬁcacy-team
performance research (Myers & Feltz, 2007), the regression models in this study
explained between 32% and 58% of the variance in team performance. However, a
better understanding of the moderating inﬂuence of collective efﬁcacy dispersion could
be achieved by assessing collective efﬁcacy prior to selected competitions.

A ﬁnal limitation of the current investigation was that complete networks were
not assessed. In this study, a social network was deﬁned as the players and head coach
within the team. However, not only are teams comprised of players and head coaches,
but also trainers and assistant coaches. By not including these individuals as part of the
social network, it was assumed that the interactions of players and head coaches with
trainers and assistant coaches were constant. This assumption is problematic because
certain players are likely to have different social interactions with trainers and assistant
coaches. For purposes of this study however, it would have been difﬁcult to assess the
complete network of relationships within teams. Unlike players and head coaches,
trainers and assistant coaches are not always listed on team rosters. Furthermore, the
number of trainers and assistant coaches signiﬁcantly varies between teams. Thus, the

current study focused on the most visible and perhaps relevant members of the social

77

system. Future studies could address this limitation by examining the social ties that
occur among all individuals on the team.
Future Research Directions

Because this study represents an initial attempt to examine the antecedents and
moderating inﬂuences of collective efﬁcacy dispersion, there are several areas that
could be investigated in future studies. First, the current study focused on the degree of
collective efﬁcacy dispersion as measured by within-team variability (i.e., SD).
However, DeRue et al. (in press) have proposed that in addition to the degree of within-
tearn variation, the pattern or form of efﬁcacy dispersion should be considered. They
suggest that there are four forms of collective efﬁcacy dispersion that may exist within
teams, which differentially affect team processes and effectiveness. The ﬁrst is a shared
efﬁcacy conﬁguration which is similar to the conceptualization of previous collective
efﬁcacy research, and reﬂects a general lack of variability among team members’
beliefs. A second form is minority dissent where a single team member has a
meaningfully different belief relative to the rest of the team. Theoretically, the
dissenting member can have a relatively higher or lower collective efﬁcacy belief than
the rest of the team. The third is a bimodal form of dispersion which represents a
distribution of beliefs along which subgroups form within the team. Finally, a
fragmented form of dispersion is where all team members have meaningfully different
beliefs. Future researchers could examine whether the forms of collective efﬁcacy
dispersion proposed by DeRue and colleagues actually exist in intact groups such as
athletic teams. If such conﬁgurations are found to exist, then researchers could examine

the conditions under which they develop, the factors that create the different forms of

78

dispersion, and the team-related consequences of various efﬁcacy patterns. The various
forms of collective efﬁcacy dispersion proposed by DeRue and colleagues were based
on 4-person teams, which made it difﬁcult to detect these forms with the data in this
study because of the relatively large number of players within teams. Future studies
interested in the forms of collective efﬁcacy dispersion should sample teams that are
smaller in size such as basketball teams of starters, 4-man bobsled teams, or curling
teams. A suggestion related to the DeRue et al. study is that researchers should continue
to examine different measurement procedures for assessing various forms of dispersion.
For instance, although DeRue and colleagues recommend that skewness and kurtosis
estimates be used as measures of efﬁcacy dispersion, recent research suggests that there
are several problems associated with these measures (Chow, Dithurbide, & Feltz, 2009).
In order to more fully understand how social networks develop and their effects
on group processes, team emergent motivational states, and team functioning, future
studies could assess social networks at multiple time points over the course of a
competitive season. Due to the time consuming nature of sociometric questionnaires,
social network data was collected at only one time point in this study (middle of the
season). Thus, it is unknown whether communication and ﬁiendship ties within teams at
the middle of the season were similar or different from those at the beginning of the
season or at the end of the season. Further, an advantage of using intercollegiate sports
teams in future social network research is that players’ social interactions could be
assessed throughout their tenure with the team to examine how these interactions

develop or change over time. Intercollegiate athletics provides a context where new

79

members are constantly entering and old members are constantly leaving and as a result,
social interactions and relationships within teams are likely to change.

Another future research direction involves using qualitative methods to examine
the pattern of social ties within teams. An interesting ﬁnding of the current study was
that most teams did not have subgroups. However, the teams that demonstrated
evidence of communication subgroups had higher levels of collective efﬁcacy
dispersion. In order to understand why some teams develop subgroups, while other
teams do not, a researcher could follow select teams over the course of a competitive
season and conduct interviews or focus groups with players and coaches. Further,
qualitative methods may provide further insight into how individuals deﬁne
communication and ﬁiendship ties and whether they are aware of the social structure
that exists within the team.

Future studies could also examine other antecedents of collective efﬁcacy
dispersion. Understanding the factors that contribute to shared or dispersed collective
efﬁcacy beliefs is critical for construct development. Additionally, if further evidence is
provided that collective efﬁcacy dispersion predicts important team-related outcomes,
then it is imperative that researchers understand how to foster its development and
maintenance. This study focused on social interaction as the basis for why teams
develop different levels of agreement, though there are certainly other factors that are
likely to affect the extent to which members agree about the team’s capabilities to be
successful. For instance, the stage of group development may impact collective efﬁcacy
dispersion. Previous research has indicated that collective efﬁcacy beliefs become more

homogeneous over time (Jung & Sosik, 2003), which suggests that levels of dispersion

80

may be more susceptible to change during earlier stages of team development such as
the beginning of the season. Examining the factors that contribute to a quick
development of interrelated collective efﬁcacy beliefs would be particularly beneﬁcial
to coaches and applied consultants. Characteristics of the coach may be another
antecedent of collective efﬁcacy dispersion. Leaders who foster close relationships with
players, create more opportunities for sharing and clarifying perceptions, and exhibit
greater consistency across situations should have teams with less collective efﬁcacy
dispersion. In contrast, leaders who create social factions by favoring certain players,
employ an autocratic approach to coaching where player input is limited, and provide
inconsistent feedback are likely to have teams with dispersed collective efﬁcacy beliefs.
Further, there is some evidence of gender differences between male and female coaches
of female teams regarding the relationship between coaches’ perceptions of motivation
efﬁcacy and team satisfaction with male coaches showing no relationship and female
coaches showing a signiﬁcant and positive relationship (Myers, Vargas-Tonsing, &
Feltz, 2005). This gender difference in coaching efﬁcacy could inﬂuence collective
efﬁcacy dispersion.

A ﬁnal recommendation for future research is to examine gender differences in
social interaction patterns. For instance, there may be different communication and ’
friendship dynamics along gender lines. Research has found that females tend to view
affectively oriented communication skills as more important, whereas males tend to
view instrumentally oriented communication skills as more important (Burleson,
Kunkel, Samter, & Werking, 1996). The current study examined women’s teams and

thus, it is unknown whether the ﬁndings generalize to men’s teams. In order to

81

investigate gender differences, researchers could compare men’s and women’s athletic
teams in sports that are played by both sexes with similar rules such as'soccer,
volleyball, or basketball. Comparing women’s softball teams with men’s baseballtearns
would provide an inaccurate test because the strategies and rules associated with each
sport are fundamentally different.
Conclusions

Although collective efﬁcacy is characterized by both a representative team
estimate and the degree of within-team variability around this central belief, previous
studies have treated the degree of within-team variability as a statistical precondition for
aggregation. In accordance with dispersion theorists (Brown et al., 1996; Chan, 1998), a
dispersion theory of collective efﬁcacy was developed and tested with intercollegiate
women’s softball teams. This study contributes to a better understanding of how the
construct emerges and provides support for conceptualizing collective efﬁcacy
dispersion as a meaningful group level variable. Social network techniques were used to
describe the structure of social interactions within teams, which explained why some
teams developed more interrelated collective efﬁcacy beliefs than other teams. In
addition, collective efﬁcacy dispersion predicted run differential and moderated the

relationship between collective efﬁcacy and run differential.

82

Appendix A

Athlete Demographic Questionnaire (Time 1)

 

 

1. Name (Last, First):

2. Today’s Date:

3. Age:

4. College Attended:

5. NCAA Division:

6. Year in School: I First-Year
E Sophomore
[2 Junior
C. Senior
1: Fifth-Year

7. Total number of years playing this sport:

Answer the following questions regarding the current season only:

 

8. Current position(s) played:

9. Jersey number:

10. Number of years on this team (including this year):

11. Are you typically a: f STARTER C NON-STARTER TN/A
12. Are you a team captain this year? F YES [7 NO

13. Do you consider yourself to be a leader on this team? F YES 7: NO

83

5.

Appendix B

Athlete Demographic Questionnaire (Time 2)

Name (Last, First):

 

Today’s Date:

Age:

Year in School: E First-Year
C Sophomore
1: Junior
17 Senior
C Fifth-Year

Total number of years playing this sport:

Answer the following questions regarding the current season only:

6.

7.

10.

ll.

12.

13.

Current position(s) played:

 

Number of years on this team (including this year):

Are you typically a: C STARTER E NON-STARTER DN/A
Are you a team captain this year? C. YES 1: NO

If not a captain, do you consider yourself to be a leader on this team? I: YES I NO

How certain are you that YOUR COACH CAN communicate effectively with players?

Carmot Moderately Highly certain
do at all can do can do
1 2 3 4 5

How certain are you that YOUR COACH CAN build the mental skills of players?

Cannot - Moderately Highly certain
do at all can do can do
1 2 3 4 5

How certain are you that YOUR COACH CAN make critical coaching decisions?

Cannot Moderately Highly certain
do at all can do can do
1 2 3 4 5

84

10.

ll.

12.

Appendix C
Coach Demographic Questionnaire

Today’s Date:

Gender". E Female F Male

Age:

College currently coaching at:

 

Number of years as coach of this team (including this year):

Total number of years coaching this sport:

Number of years playing experience in this sport:

Highest level of playing experience:

 

What is your team’s current (W-L) record:

 

How would you rate the physical ability of the athletes on your team this year?

Very
poor Excellent
1 2 3 4 5

How would you rate the teamwork ability of the athletes on your team this year?

Very
poor Excellent
1 2 3 4 5

How would you rate the overall ability of the teams on your schedule this year?

Very
poor Excellent
1 2 3 4 5

85

Appendix D
Collective Efﬁcacy Questionnaire
Directions: Listed below are different team performance skills. Please rate how certain you are
that YOUR TEAM AS A WHOLE can execute these skills. Circle the appropriate number to
the right of each statement.

Rate your degree of confidence by recording a numberfrom 1 to 5 using the scale given below:

How certain are you that YOUR TEAM CAN...”

 

 

 

 

 

 

 

 

 

 

 

Cannot Moderately Highly certain

do at all can do can do
1. communicate well as a unit 1 2 3 4 5
2. regain mental focus after an 1 2 3 4 5

error/mistake

3. avoid walking opposing batters 1 2 3 4 5
4. consistently throw strikes 1 2 3 4 5
5. outscore opponents 1 2 3 4 5
6. consistently put the ball in play 1 2 3 4 5
7. have a high ﬁelding percentage 1 2 3 4 5
8. hit with runners in scoring position 1 2 3 4 5
9. successfully lay down bunts 1 2 3 4 5
10. make good decisions on the base paths 1 2 3 4 5

 

86

 

Appendix E
Team Commitment Questionnaire
Directions: Think about the current season and playing with this team. Please rate how true

each statement is to you by recording a number from 1 to 5 using the scale given below. Circle
the appropriate number to the right of each statement.

 

 

 

 

 

Not at all Completely
true for me true for me
1. I am dedicated to playing with this team. 1 2 3 4 5
2. It would be hard for me to quit playing with l 2 3 4 5
this team.
3. I am determined to keep playing with this 1 2 3 4 5
team.
4. I am willing to do almost anything to keep 1 2 3 4 5
playing with this team.

 

87

 

Appendix F
Social Interaction Questionnaire

Directions: Think about the interactions that you have with other members on this team in the current
season.

1. Please rate how frequently you talk with each of your team members on subjects that are related to the
team’s ﬁmctioning such as discussing the team’s strengths and weaknesses, performances, or goals.
Circle the appropriate number next to each member’s name in the talk about team’s functioning
column.

2. Place an X in the friend column next to each member’s name who you consider to be a very good
friend of yours, someone whom you see socially outside of team-related activities. You may select as
many members as applicable.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Talk about Team’s Functioning Friend
(circle appropriate number for each member, (select as many
except for ﬂurself) as applicable)
Player Very little Sometimes A great deal
EXAMPLE 1 2 (a) 4 5 x
Player 1 l 2 3 4 5
Player 2 1 2 3 4 5
Player 3 1 2 3 4 5
Player 4 l 2 3 4 5
Player 5 1 2 3 4 5
Player.6 _ l 2 3 4 5
Player 7 1 2 3 4 5
Player 8 1 2 3 4 5
Player 9 1 2 3 4 5
Player 10 1 2 3 4 5
Player 1 1 l 2 3 4 5
Player 12 1 2 3 4 5
Player 13 1 2 3 4 5
Player 14 1 2 3 4 5
Coach 1 2 3 4 5

 

 

 

 

 

88

 

Table 1

Athlete Demographics (N = 763)

 

 

Variable Frequency M SD
Year in School 2.25 1.13
Freshman 255
Sophomore 214
Junior 152
Senior 132
Fifth-Year 10
Years on Team 2.09 1.08
One 297
Two 217
Three 135
Four 112
Five 2
Starter Status
Starters 482
Non-Starters 247
N/A 34
Captain Status
Captain 101
Non-Captain 662
Leader Status
Leader 513 ‘
Non-Leader 250
Age 1 9.79 1 .30
Playing Experience 12.20 3.49

 

89

 

 

Table 2

Coach Demographics (N = 46)

 

 

Variable Frequency M SD
Gender

Female 38

Male 8
Years Coaching Team 8.37 7.56
Age 39.58 11.94
Coaching Experience 14.09 9.26
Playing Experience 17.04 8.53

 

9O

Table 3

Means and Standard Deviations for Team Variables (N = 46)

 

 

Variable M SD
Collective Efﬁcacy 1 3.89 .32
Collective Efﬁcacy 2 3.82 .41

Dispersion 1 -0.15 .05
Dispersion 2 -O.17 .08
Communication Density 0.47 .11
Friendship Density 0.65 .12
Communication Centralization 0.19 .05
Friendship Centralization 0.26 .08
Communication Subgroup 0.33 .47
Friendship Subgroup 0.28 .46
Performance 2 0.55 .20
Performance 3 0.55 .18
Run Differential 2 0.87 2.80
Run Differential 3 0.88 2.46
Team Commitment 1 4.74 .25
Team Commitment 2 4.56 .35
Conﬁdence in Coach 3.94 .64

 

91

 

 

Haw—o A

$5388 ﬁes-£8858: 88:: 88-3294 «2 H «3

 

 

_ N 8 s u a 8 m o a : 3- .... E a a
ram. -.|
N23 8...... l.-
88: -8... 8...... l
9mg 8...... -8: 8...... l
m. 8 -8 .8 a: -8 l-
a. mo .8 .8 -8 -8 8...... .l
s on 8 a; 8 .8 -8... -.8 .l
8. mo -8: -8... .3 .8 -8... -8: .8 l-
emma .8 -8 .3 8.... -8 -8 -8 -8 II
8.80 .8 -8 .5 so. .3 a... -8 8 8...... II
:8: 8...... 8...... -8... 3...... -8 8 -8 -8 -8 -8 I
88.2.8 8...... 8...... -8... -8... -8 -8 -8 -8 -8 -8 8.... |
8.85 8...... 3...... -8: -8: -8 -8 -8 Lo -8 .8 8...... 8...... l
3.68 8...... 8...... -8... -8... -8 -8 .8 -8 -8 -8 8...... 8...... 8......

3. H0 _ .wo: 5%.... Ag Awe .3 .w _ ... .No 33...... J; J: .3» 00...... .Nm .3... I
3. H0 N .33.. .mm: 3.3 13...... .mm .3... A: 23...... a; now .3... .33.. .3... .mm... 6%.... |
3. Oh. we: Qua... 33... 3.3.3.. Mama... .3... 33... £3... 33 .3 ha»... km...» .33.. .33. .3... .33..

92

 

28m. mm H no=noz<o min-.83. m0 u Oo=ooa<o maomov. Cinnamon. 00 u 0085:3830: 09.me- mU H man-8:336 U955.- 00 H 0033:3820:
Gong—www.mo? m0 u magma—.6 035.523... 0mm n Goa—523839. mccmnocu. mmO u manna-ﬂan mccwacu. wan. n Ham... $233338. WU n We:
9323:3- A-O u Ham... 0038:303. CO H Oosmaosoo 5 demo:- ...h A .3. ...... h A .3.

Table 5

Dispersion Regressed on Network Variables

 

 

Variable B SE B ,B
Constant -0.203 0.1 1 l

Dispersion 1 0.777 0.188 0487*“
Communication Density 0.090 0.101 0.127
Freindship Density -0.022 0.101 -0.036
Communication Centralization 0.003 0.002 0.220
Friendship Centralization 0.002 0.001 0.244
Communication Subgroup 0.071 0.022 0.445"
Friendship Subgroup -0.051 0.023 -0.307*

 

*p < .05. **p < .01. ***p < .001.

93

Table 6

Multilevel Model Based on Communication Subgroups

 

 

 

 

Fixed Effect Coefﬁcient SE t
Intercept 3.71 .04 82.85***
CE] Team .62 .21 2.92*
CEl Subgroup -.05 .15 -.33
CE] Individual .57 .06 9.89***
Random Effect Variance Component df ;2
Intercept 1 .01 31 4628*
Level-1 .16

Interceptl/Intercept 2 .02 13 35.39M

 

*p < .05. **p < .01. *"p <.

001.

94

Table 7

Multilevel MOdel Based on Friendship Subgroups

 

 

 

 

Fixed Effect Coefﬁcient SE T
Intercept 3.80 .08 47.36***
CEl Team -.09 .26 -.32
CE] Subgroup .25 .16 1.57
CEl Individual .43 .06 734*“
Random Effect Variance Component df £7-
lntercept l .01 16 22.28
Level-1 .12

Interceptl/Intercept 2 .07 1 1 85.13***

 

*p < .05. "p < .01. ***p <.

001.

95

Table 8

Dispersion as a Moderator of Collective Efﬁcacy-Team Performance Relationship

 

ﬂ R2

DV: Team Performance 2

Collective Efficacy 1 .55“

Dispersion 1 -.10

Collective Efﬁcacy x Dispersion .03 32*"
DV: Team Perfomance 3

Collective Efﬁcacy 2 .67***

Dispersion 2 -20

Collective Efﬁcacy x Dispersion .10 .58***
DV: Run Differential 2

Collective Efﬁcacy 1 .52"

Dispersion 1 -.16

Collective Efﬁcacy x Dispersion .06 031*”
DV: Run Differential 3

Collective Efficacy 2 ,57***

Dispersion 2 331*

Collective Efﬁcacy x Dispersion .22 .52***

 

*p < .05. "p < .01. ***p < .001.

96

Table 9

Dispersion as a Moderator of Collective Efficacy-Team Commitment Relationship

 

I3 R2

DV: Team Commitment 1

Collective Efﬁcacy 1 .43*

Dispersion 1 -.09

Collective Efﬁcacy x Dispersion .16 .14*
DV: Team Commitment 2

Collective Efﬁcacy 2 .55**

Dispersion 2 -.19

Collective Efﬁcacy x Dispersion .03 .43***

 

*p < .05. **p < .01. ***p < .001.

97

Ewan A.

@85th GAS ossbahmamo:

:v Om_ Omw 0mm. OmA 0mm

A A u A w A
N u A u A u
w A A A w u
A w w u A A
m w A A w w
o A A u u A
Mb .mm .mm .mm .3 .mm
.38: 3

4.83 .mb

Po

PA

Pm

PA

PA

ud

Pm

.mA

HO

a

MD

H883

A83 Mb

Om—

Tom

0mm Omu OmA 0mm

N

Abw

Tom

N N

w w

.mA row

NN

Nm

WA

Pm

A.~

Ad

PM

_.oo

98

 

 

\A
\.
, A .
\\\\ . A ./ /l
l.\.\ A /.,
\x . A I
\ .. /
.\\ A / /./
\ x. . .. 4
xx , A ., //
. A //
\ . I.
x A /
\. A A /
\..\ A //
\
\\. A I //
\. I;
.\\ /
..\x A /.
u! tlttllllstltlf L, P: ill» /I
1.. ., . .. .‘.
. ./ . x
. . . _.
. /, A \x\ _
\ ..
A /./.. A, \\ A
,A / A \\.. A
. A .\\ ..
A .1 A R .
A . . A A
A I: , y\. A
A /- A v .A. A A
./ A \\ .. A
A .l. . \
A. . A \ . .
, I, A t
. ./ A t ..
A 11!... A ‘\ A
, ,. 1./.r.\., . A
A \\ . _ /// A
A x...\ .A /, A
. m A
.A x, A //x . A
A t A ./1 z A
\\ / A
A A // m A A
A I\\ A i K. ,. A
A . , . / A
A . . / I
_ ., I/ A
.. x. x/ A
A . A l/ / A
A A // _
. / . A
A x A // .. .
. . A /,r, _
...I.. 1.! l i 1 l I It; «Ill III. l lily! llll‘ltvlu/r» Pl! .1- .llyvuilt‘llltlillllltul llllltllllilll..l\l‘
4! « \\.
./ s.\
. A .\
_ \\
/ A \ \s
A .
A s
A \
.
.l A
. ./. x

99

 

EmEd w

NNAEASQQA Qs>§€cla QmESNANEAd:

 

 

100

 

REFERENCES

Arthur, W., Bell, S. T., & Edwards, B. D. (2007). A longitudinal examination of the
comparative criterion-related validity of additive and referent-shift consensus

operationalizations of team efﬁcacy. Organizational Research Methods, 10, 35-
58.

Baerveldt, C., & Snijders, T. A. B. (1994). Inﬂuences on and from the segmentation of
networks: Hypotheses and tests. Social Networks, 16, 213-232.

Baldwin, T. T., Bedell, M. D., & Johnson, J. L. (1997). The social fabric of a team-
based M.B.A. program: Network effects on student satisfaction and
performance. Academy of Management Journal, 40, 1369-1397.

Balkundi, P., & Harrison, D. A. (2006). Ties, leaders, and time in teams: Strong
inference about network structure’s effects on team viability and performance.
Academy of Management Journal, 49, 49-68.

Bandura, A. (1977). Self-efﬁcacy: Toward a unifying theory of behavioral change.
Psychological Review, 84, 191-21 5.

Bandura, A. (1993). Perceived self-efﬁcacy in cognitive-deveIOpment and functioning.
Educational Psychologist, 28, 117-148.

Bandura, A. (1997). Self-eﬁicacy: The exercise of control. New York: Freeman.

Bandura, A. (2006). Guide for creating self-efﬁcacy scales. In F. Pajares & T. Urdan
(Eds), Self-eﬂicacy beliefs of adolescents (pp. 307-337). Greenwich, CT:
Information Age Publishing.

Bavelas, A. (1950). Communication patterns in task-oriented groups. Journal of the
Acoustical Society of America, 22, 723-730.

Bliese, P. D., & Halverson, R. R. (1998). Group consensus and psychological well-
being: A large ﬁeld study. Journal of Applied Social Psychology, 28, 563-580.

Blumer, H. (1969). Symbolic interactionism: Perspective and method. Englewood
Cliffs, NJ: Prentice-Hall.

Boardley, I. D., Kavussanu, M., & Ring, C. (2008). Athletes’ perceptions of coaching
effectiveness and athlete-related outcomes in rugby union: An investigation

based on the coaching efﬁcacy model. The Sport Psychologist, 22, 269-287.

Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). Ucinet for Windows: Software
for social network analysis. Harvard, MA: Analytic Technologies.

101

Borgatti, S. P., & Foster, P. C. (2003). The network paradigm in organizational
research: A review and typology. Journal of Management, 29, 991-1013.

Brass, D. J. (1984). Being in the right place: A structural analysis of individual
inﬂuence in an organization. Administrative Science Quarterly, 29, 51 8-53 9.

Brass, D. J ., Butterﬁeld, K. D., & Skaggs, B. C. (1998). Relationships and unethical
behavior: A social network perspective. Academy of Management Review, 23,
14-3 1 .

Brown, K. G., Kozlowski, S. W., & Hattrup, K. (1996, August). Theory, issues, and
recommendations in conceptualizing agreement as a construct in organizational
research: The search for consensus regarding consensus. Paper presented at the
Academy of Management Annual Meeting, Cincinnati, Ohio.

Burkhardt, M. E. (1994). Social interaction effects following a technological change: A
longitudinal investigation. Academy of Management Journal, 3 7, 869-898.

Burkhardt, M. E., & Brass, D. J. (1990). Changing patterns or patterns of change: The
effect of a change in technology on social network structure and power.
Administrative Science Quarterly, 35, 104-127.

Burleson, B. R., Kunkel, A. W., Samter, W., & Werking, K. J. (1996). Men’s and
women’s evaluations of communication skills in personal relationships: When

sex differences make a difference and when they don’t. Journal of Social and
Personal Relationships, 13, 201 -224.

Burt, R. S. (1987). Social contagion and innovation: Cohesion vs. structural
equivalence. American Journal of Sociology, 92, 1287-1335.

Campion, M. A., Medsker, J. G., & Higgs, A. C. (1993). Relations between work group
characteristics and effectiveness: Implications for designing effective work
groups. Personnel Psychology, 46, 823-850.

Carron, A. V., Brawley, L. R., & Widmeyer, W. N. (1998). The measurement of
' cohesiveness in sport groups. In J. L. Duda (Ed.), Advances in sport and
exercise psychology measurement (pp. 213-226). Morgantown, WV: Fitness
Information Technology.

Carron, A. V., Widmeyer, W. N., & Brawley, L. R. (1985). The development of an
instrument to assess cohesion in sport teams: The Group Environment
Questionnaire. Journal of Sport Psychology, 7, 244-266.

Chan, D. (1998). Functional relationships among constructs in the same content domain

at different levels of analysis: A typology of composition models. Journal of
Applied Psychology, 83, 234-246.

102

Chase, M. A., Feltz, D. L., & Lirgg, C. D. (2003). Sources of collective efﬁcacy and
individual efﬁcacy of collegiate athletes. International Journal of Sport and
Exercise Psychology, 1, 180-191.

Chen, G., & Bliese, P. D. (2002). The role of different levels of leadership in predicting
self- and collective efﬁcacy: Evidence for discontinuity. Journal of Applied
Psychology, 87, 549-556.

Chen, G., Webber, S. S., Bliese, P. D., Mathieu, J. E., Payne, S. C., Born, D. H., &
Zaccaro, S. J. (2002). Simultaneous examination of the antecedents and

consequences of efﬁcacy beliefs at multiple levels of analysis. Human
Performance, 15, 3 8 l -409.

Chow, G. M., Dithurbide, L., & Feltz, D. L. (2009, June). Going beyond level of
analysis: Examining dispersion as a team-level construct. Paper presented at the
North American Society for the Psychology of Sport and Physical Activity
Annual Conference, Austin, Texas.

Chow, G. M., & Feltz, D. L. (2007). Exploring new directions in collective efﬁcacy and
sport. In M. R. Beauchamp & M. Eys (Eds.), Group Dynamics Advances in
Sport and Exercise Psychology: Contemporary Themes. UK: Routledge.

Chow, G. M., & Feltz, D. L. (2008). Exploring the relationships between collective
efﬁcacy, perceptions of success, and team attributions. Journal of Sports
Sciences, 26, 1 179-1189.

Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal
of Sociology, 94, 895-8120.

Cummings, J. N., & Cross, R. (2003). Structural properties of work groups and their
consequences for performance. Social Networks, 25, 197-281.

Dean, J. W., & Brass, D. J. (1985). Social interaction and the perception of job
characteristics in an organization. Human Relations, 38, 571-582.

DeRue, D. S., Hollenbeck, J. R., Ilgen, D. R., & Feltz, D. L. (in press). Efﬁcacy
dispersion in teams: Moving beyond agreement and aggregation. Personnel
Psychology.

Barley, P. C., & Mosakowski, E. (2000). Creating hybrid team cultures: An empirical
test of transitional team functioning. Academy of Management Journal, 43, 26-
49.

Feltz, D. L., & Lirgg, C. D. (1998). Perceived team and player efﬁcacy in hockey.
Journal of Applied Psychology, 83, 557-564.

103

Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7,
117-140.

Festinger, L., Schachter, S., & Black, K. W. (1950). Social Pressures in Informal
Groups: A Study of Human Factors in Housing. New York: Harper.

Frank, K. A. (1995). Identifying cohesive subgroups. Social Networks, 1 7, 27-56.

Frank, K. A. (1996). Mapping interactions within and between cohesive subgroups.
Social Networks, 18, 93-1 19.

Frank, K. A., Muller, C., Schiller, K. S., Riegle-Crumb, C., Mueller, A. S., Crosnoe, R.,
& Pearson, J. (2008). The social dynamics of mathematics course-taking in high
school. American Journal of Sociology, 113, 1645-1696.

Frank, K. A., & Yasumoto, J. Y. (1998). Linking action to social structure within a
system: Social capital within and between subgroups. American Journal of
Sociology, 104, 642-686.

Freeman, L. (1979). Centrality in social networks: Conceptual clariﬁcation. Social
Networks, 1, 215-239.

Freeman, L. C. (1984). The impact of computer based communication on the social
structure of an emerging scientiﬁc specialty. Social Networks, 6, 201-221

Friedkin, N. E. (1981). The development of structure in random networks: An analysis
of the effects of increasing network density on ﬁve measures of structure. Social
Networks, 3, 41-52.

Friedkin, N. E. (1984). Structural cohesion and equivalence explanations of social
homogeneity. Sociological Methods & Research, 12, 235-261.

Friedkin, N. E. (1993). Structural bases of interpersonal inﬂuence in groups: A
longitudinal case study. American Sociological Review, 58, 861-872.

Friedkin, N. E. (2001). Norm formation in social inﬂuence networks. Social networks,
23, 167-189.

Friedkin, N. E., & Cook, K. S. (1990). Peer group inﬂuence. Sociological Methods &
Research, 19, 122-143.

Gibson, C. B. (1999). Do they do what they believe they can? Group efﬁcacy and group

effectiveness across tasks and cultures. Academy of Management Journal, 42,
l 3 8-1 52.

104

Gibson, C. B., Randel, A., & Barley, C. (2000). Understanding group efﬁcacy: An
empirical test of multiple assessment methods. Group & Organization
Management, 25, 67-97.

Gibson, C. B., & Vermeulen, F. (2003). A healthy divide: Subgroups as a stimulus for
team learning behavior. Administrative Science Quarterly, 48, 202-239.

Gist, M. E. (1987). Self-efﬁcacy: Implications for organizational behavior and human
resource management. Academy of Management Review, 12, 472-485.

Gonzalez-Roma, V., Peiro, J. M., & Tordera, N. (2002). An examination of the
antecedents and moderator inﬂuences of climate strength. Journal of Applied
Psychology, 8 7, 465—473.

Greenlees, I. A., Nunn, R. L., Graydon, J. K, & Maynard, I. W. (1999). The
relationship
between collective efﬁcacy and precompetitive affect in rugby players: Testing
Bandura’s model of collective efﬁcacy. Perceptual and Motor Skills, 89, 431-
440.

Gully, S. M., Incalcaterra, K. A., Joshi, A., & Beaubien, J. M. (2002). A meta-analysis
of team-efﬁcacy, potency, and performance: Interdependence and level of
analysis as moderators of observed relationships. Journal of Applied
Psychology, 8 7, 819-832.

Hodges, L., & Carron, A. (1992). Collective efﬁcacy and group performance.
International Journal of Sport Psychology, 23, 48-59.

Ibarra, H. (1993). Network centrality, power, and innovation involvement:
Determinants of technical and administrative roles. Academy of Management
Journal, 36, 471-501.

Ibarra, H., & Andrews, S. B. (1993). Power, social inﬂuence, and sense making: Effects
of network centrality and proximity on employee perceptions. Administrative
Science Quarterly, 38, 277-303.

Jackson, S. (1992). Team composition in organizations. In S. Worchel, W. Wood, & J.
Simpson (Eds.), Group Process and Productivity. London: Sage.

Jackson, S., May, K. E., & Whitney, K. (1995). Understanding the diversity of
dynamics in decision making teams. In R. A. Guzzo, E. Salas, & Associates
(Eds.), Team Eﬂectiveness and Decision Making in Organizations. San
Francisco: J ossey-Bass.

105

James, L. R., Demaree, R. G., & Wolf, G. (1984). Estimating within-group rater
reliability with and without response bias. Journal of Applied Psychology, 69,
85-98. .

Jung, D. 1., & Sosik, J. J. (2003). Group potency and collective efﬁcacy: Examining
their predictive validity, level of analysis, and effects of performance feedback
on future group performance. Group & Organization Management, 28, 366-391.

Kilduff, M. (1990). The interpersonal structure of decision making: A social
comparison approach to organizational choice. Organizational Behavior and
Human Decision Processes, 4 7, 1-19.

Klein, K. J ., Conn, A. B., Smith, D. B., & Sorra, J. S. (2001). Is everyone in agreement?
An exploration of within-group agreement in employee perceptions of work
environment. Journal of Applied Psychology, 86, 3-16.

Kozlowski, S. W. J ., Gully, S. M., Nason, E. R., & Smith, E. M. (1999). Developing
adaptive teams: A theory of compilation and performance across levels and
time. In D. R. Ilgen, & E. D. Pulakos (Eds.), The Changing Nature of
Performance: Implications for Staﬁing, Motivation, and Development, 240-292.
San Francisco: Jossey-Bass Publishers.

Kozub, S. A., & McDonnell, J. F. (2000). Exploring the relationship between cohesion
and collective efﬁcacy in rugby teams. Journal of Sport Behavior, 23, 120-129.

Krackhardt, D. (1999). The ties that torture: Simmelian tie analysis in organizations. In
S. B. Bacharach, S. B. Andrews, & D. Knoke (Eds.), Research in the sociology
of organizations, vol. 16, 183-210. Stamford, CT: JAI Press.

Krackhardt, D., & Stern, R. N. (1988). Informal networks and organizational crisis: An
experimental simulation. Social Psychology Quarterly, 51, 123-140.

Lau, D. C., & Mumighan, J. K. (1998). Demographic diversity and faultlines: The
compositional dynamics of organizational groups. Academy of Management
Review, 23, 325-340.

Lazer, D. M. (2001). The co-evolution of individual and‘network. Journal of
Mathematical Sociology, 25 , 69-108.

Leavitt, H. J. (1951). Some effects of certain communication patterns on group
performance. Journal of Abnormal and Social Psychology, 46, 38-50.

Lichacz, F. M., & Partington, J. T. (1996). Collective efﬁcacy and true performance.
International Journal of Sport Psychology, 2 7, 146-158.

106

Linacre, J. M. (2002). Optimizing rating scale category effectiveness. Journal of
Applied Measurement, 3, 85-106.

Lincoln, J. R., & Miller, J. (1979). Work and friendship ties in organizations: A
comparative analysis of relational networks. Administrative Science Quarterly,
24, 181-199.

Lindell, M. K., & Brandt, C. J. (1997). Measuring interrater agreement for ratings of a
single target. Applied Psychological Measurement, 21, 271-278.

Lindell, M. K., & Brandt, C. J. (2000). Climate quality and climate consensus as
mediators of the relationship between organizational antecedents and outcomes.
Journal of Applied Psychology, 85, 331-348.

Lindsley, D. H., Brass, D. J ., & Thomas, J. B. (1995). Efﬁcacy-performance spirals: A
multilevel perspective. Academy of Management Review, 20, 645-678.

Lovelace, D., Shapiro, D. L., & Weingart, L. R. (2001). Maximizing cross-functional
new product teams' innovativeness and constraint adherence: A conﬂict
communications perspective. Academy of Management Journal, 44, 779-79.

Magyar, T. M., Feltz, D. L., & Simpson, I. P. (2004). Individual and crew level
determinants of collective efﬁcacy in rowing. Journal of Sport & Exercise
Psychology, 26, 136-153.

Marsden, P. V., & Friedkin, N. E. (1993). Network studies of social inﬂuence.
Sociological Methods & Research, 22, 127-151.

McGrath, J. E. (1984). Groups, interaction, and performance. Englewood Cliffs, NJ:
Prentice-Hall.

Moritz, S. E., & Watson, C. B. (1998). Levels of analysis issues in group psychology:
Using efﬁcacy as an example of a multilevel model. Group Dynamics: Theory,
Research, and Practice, 2, 285-298.

Morrison, E. W. (2002). Newcomers’ relationships: The role of social network ties
during socialization. Academy of Management Journal, 45, 1 149-1160.

Mulvey, P. W., & Klein, H. J. (1998). The impact of perceived loaﬁng and collective
efﬁcacy on group goal processes and group performance. Organizational
Behavior and Human Decision Processes, 74, 62-87.

Mudrack, P. E. (1989). Deﬁning group cohesiveness: A legacy of confusion? Small
Group Behavior, 20, 37-49.

107

 

 

Myers, N. D., & Feltz, D. L. (2007). From self-efﬁcacy to collective efﬁcacy in sport:
Transitional issues. In G. Tenenbaum, & R. C. Eklund (Eds.), Handbook of

sport psychology, 3 ed. New York: Wiley.

Myers, N. D., F eltz, D. L., & Short, S. E. (2004). Collective efﬁcacy and team
performance: A longitudinal study of collegiate football teams. Group Dynamics:
Theory, Research, and Practice, 8, 126-138.

Myers, N. D., Payment, C. A., & F eltz, D. L. (2004). Reciprocal relationships between
collective efﬁcacy and team performance in women’s ice hockey. Group
Dynamics: Theory, Research, and Practice, 8, 182-195.

Myers, N. D., Vargas-Tonsing, T. M., & Feltz, D. L. (2005). Coaching efﬁcacy in
intercollegiate coaches: Sources, coaching behavior, and team variables.
Psychology of Sport and Exercise, 6, 129-143.

Parker, L. E. (1994). Perceived self-efﬁcacy and collective-efﬁcacy at the workplace.
Journal of Applied Social Psychology, 24, 43-59.

Paskevich, D. M., Brawley, L. R., Dorsch, K. D., & Widmeyer, W. N. (1999).
Relationship between collective efﬁcacy and team cohesion: Conceptual and

measurement issues. Group Dynamics: Theory, Research, and Practice, 3, 210-
222.

Paxton, P., & Moody, J. (2003). Structure and sentiment: Explaining emotional
attachment to group. Social Psychology Quarterly, 66, 34—47.

Pescosolido, A. T. (2003). Group efﬁcacy ad group effectiveness: The effects of group
efﬁcacy over time on group performance and development. Small Group
Research, 34, 20-42.

Peterson, E., Mitchell, T. R., Thompson, L., & Burr, R. (2000). Collective efﬁcacy and
aspects of shared mental models as predictors of performance over time in work
groups. Group Processes & Intergroup Relations, 3, 296-316.

Portes, A., & Sensenbrenner, J. (1993). Embeddedness and immigration: Notes on the
social determinants of economic action. American Journal of Sociology, 98,
' 1320-1350.

Prussia, G. E., & Kinicki, A. J. (1996). A motivational investigation of group
effectiveness using social cognitive theory. Journal of Applied Psychology, 81,
1 87-1 98.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications

and
data analysis methods. Newbury Park, CA: Sage.

108

 

Reagans, R., & Zuckerman, E. W. (2001). Networks, diversity, and productivity: The
social capital of corporate R&D teams. Organization Science, I 2, 502-517.

Rentsch, J. R. (1990). Climate and culture: Interaction and qualitative differences in
organizational meanings. Journal of Applied Psychology, 75, 668-681.

Roberson, Q. A., & Colquitt, J. A. (2005). Shared and conﬁgural justice: A social
network model of justice in teams. Academy of Management Journal, 30, 595-
607.

Scanlan, T. K., Carpenter, P. J., Schmidt, G. W., Simons, J. P., & Keeler, B. (1993). An
introduction to the sport commitment model. Journal of Sport & Exercise
Psychology, 15, 1-15.

Scanlan, T. K., Simons, J. P., Carpenter, P. J., Schmidt, G. W., & Keeler, B. (1993).
The sport commitment model: Measurement and development for the youth
sport domain. Journal of Sport & Exercise Psychology, 15, 16-38.

Schneider, B., & Reichers, A. E. (1983). On the etiology of climates. Personnel
Psychology, 36, 19-39.

Schneider, B., Salvaggio, A. N., & Subirats, M. (2002). Climate strength: A new
direction for climate research. Journal of Applied Psychology, 8 7, 220-229.

Scott, J. (2000). Social network analysis: A handbook (2nd ed.). Thousands Oaks, CA:
Sage.

Shah, P. P. (2000). Network destruction: The structural implications of downsizing.
Academy of Management Journal, 43, 101-1 12.

Shaw, M. (1964). Communication networks. In L. Berkowitz (Ed.), Advances in
experimental psychology (pp. 111-147). New York: Academic Press.

Shaw, M. E. (1954). Some effects of unequal distribution of information upon group
performance in various communication nets. Journal of Abnormal and Social
Psychology, 49, 547-553.

Silverrnan, D. (1971). The theory of organizations: A sociologicalframework. New
York: Basic Books.

Smith, P. (1973). Groups within organizations: Applications of social psychology. New
York: Harper and Row.

109

Sparrowe, R. T., Liden, R. C., Wayne, S. J ., & Kraimer, M. L. (2001). Social networks
and the performance of individuals and groups. Academy of Management
Journal, 44, 316-325.

Spink, K. S. (1990). Group cohesion and collective efﬁcacy of volleyball teams.
Journal of Sport & Exercise Psychology, 12, 301-311.

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.
Monterey, CA: Brooks-Cole.

Totterdell, P., Wall, T., Holman, D., Diamond, H., & Epitropaki, O. (2004). Affect
networks: A structural analysis of the relationship between work ties and job-
related affect. Journal of Applied Psychology, 89, 854-867.

Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications.
New York, Cambridge University Press.

Watson, C. B., Chemers, M. M., & Preiser, N. (2001). Collective efﬁcacy: A multilevel
analysis. Personality and Social Psychology Bulletin, 2 7, 1056-1068.

Wilson, P. M., Rodgers, W. M., Carpenter, P. J ., Hall, C., Hardy, J ., & Fraser, S. N.
(2004). The relationship between commitment and exercise behavior.
Psychology of Sport & Exercise, 5, 405-421 .

Zaccaro, S. J ., Blair, V., Peterson, C., & Zazanis, M. (1995). Collective efﬁcacy. In J. E.
Maddux (Ed.), Self-eﬂicacy, adaptation and adjustment: Theory, research and
application (pp. 308-330). New York: Plenum.

Zhu, W., & Kang, S. J. (1998). Cross-cultural stability of the optimal categorization of a
self-efﬁcacy scale: A Rasch analysis. Measurement of Physical Education and
Exercise Science, 2, 225-241.

Zhu, W., Updyke, W. F., & Lewandowski, C. (1997). Post-hoe Rasch analysis of
optimal categorization of ordered-response scale. Journal of Outcome
Measurement, 1, 286-304.

Zohar, D., & Tenne-Gazit, O. (2008). Transformational leadership and group interaction
as climate antecedents: A social network analysis. Journal of Applied
Psychology, 93, 744-757.

110

      

MMMMMMM

flirt/1,11 li/Iﬂllﬁlﬂljllﬂlﬂlﬂlﬂlﬂlf