University of Southern California Dissertations and Theses

Information sharing in work groups: A transactive memory approach

(USC Thesis Other)

Information sharing in work groups: A transactive memory approach

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION SHARING IN WORK GROUPS:
A TRANSACTIVE MEMORY APPROACH
Copyright 2005
by
YanXu
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
COMMUNICATION
August 2005
Yan Xu
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 3196915
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
®
UMI
UMI Microform 3196915
Copyright 2006 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
DEDICATION
TO MY LOVING HUSBAND SAGI
FOR SUPPORTING ME AND HELPING ME
KEEP A BALANCE IN LIFE
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGEMENTS
I would like to thank my academic advisor Janet Fulk and professors Paul
Adler and Patricia Riley for their helpful comments on earlier drafts of this
dissertation. Without their encouragement and input, I would not have been able to
complete this daunting project. I would also like to acknowledge my gratitude to the
faculty at Marshall School of Business and Leventhal School of Accounting in USC
who generously allowed me to collect data with their students. I would also like to
take this opportunity to extend my gratitude to the faculty and staff at the Annenberg
School for Communication in USC who have been always supportive of my personal
and professional growth.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
Page
DEDICATION................................................................................................................ii
ACKNOWLEDGEMENTS...........................................................................................ii
LIST OF TABLES..........................................................................................................v
LIST OF FIGURES.......................................................................................................vi
ABSTRACT..................................................................................................................vii
CHAPTER ONE - INTRODUCTION..........................................................................1
CHAPTER TWO - STUDY I: EVALUATING TM IN WORK GROUPS............. 9
CHAPTER THREE - STUDY II: EFFECTS OF TASK COMPLEXITY
AND CONFLICT ON TM AND ITS RELATION TO GROUP
PERFORMANCE................................................................................................. 45
CHAPTER FOUR - STUDY III: SHARING UNSHARED KNOWLEDGE IN
WORK GROUPS: A CHALLENGE POSED BY DIFFERENTIAL
KNOWLEDGE CRITICALITY..................................... 84
CHAPTER FIVE - GENERAL DISCUSSION AND CONCLUSION.................121
REFERENCES............................................................................................................ 130
APPENDICES
APPENDIX A: STUDY I SURVEY ITEMS................................................. 144
APPENDIX B: STUDY II SURVEY ITEMS................................................146
APPENDIX C: STUDY III SURVEY ITEMS...............................................148
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
V
LIST OF TABLES
Table Page
2.1. Descriptive Statistics and Correlations............................................................27
2.2. Analysis of Variance and Intraclass Correlation Coefficients
for Group-Level Scales.....................................................................................30
2.3. Summary of Model Revisions.......................................................................... 34
3.1. Analysis of Variance and Intraclass Correlation Coefficients
for Group-Level Scales.....................................................................................65
3.2. Descriptive Statistics and Correlations.............................................................68
3.3. Summary of Model Revisions .................................................................75
4.1. Analysis of Variance and Intraclass Correlation Coefficients
for Group-Level Scales...................................................................................108
4.2. Summary of Simultaneous Regression of Knowledge Diversity on
Members’ Disagreement on Knowledge Criticality Judgment...................109
4.3 Summary of Simultaneous Regression of Communication on
Members’ Disagreement on Knowledge Criticality Judgment...................I l l
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
vi
LIST OF FIGURES
Figure Page
1.1. Four Perspectives on Information Sharing........................................................6
1.2. Overview of Three Studies................................................................................. 8
2.1. Study I: A Schematic Representation of the TM Model................................20
2.2. Study I: Results of Original Theoretical Model and Modified
Measurement Models (*g < .05; N = 103).................................................... 32
2.3. Study I: Results of Modified Theoretical Model and Modified
Measurement Models (*g < .05; N = 103).................................................... 35
3.1. Study II: Model of Effects of Task Complexity and Conflict on
TM and on TM - Group Performance Interaction.........................................59
3.2. Study II: Results of Original Theoretical Model
(*P < .05, N = 154)........................................................................................... 72
3.3. Study II: Results of Modified Theoretical Model
(*g < .05, N = 154)........................................................................................... 76
4.1. Study III: Summary of Hypotheses................................................................100
5.1. Integrated Model: Combining Studies I, II, and III......................................126
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ABSTRACT
This dissertation project consists of three studies investigating issues related to
information sharing in work groups from the perspective of transactive memory
(TM). Study I examined how the antecedent, transactive processes, and the group
performance outcome were related to one another. The results suggested that,
inconsistent with TM theory, information allocation did not affect group
performance. The implication of this result is that TM theory needs to be revised and
to include contextual variables to capture information sharing processes in groups.
Studies II and III attempted to achieve this goal. These two studies looked at the
effects of two exogenous variables - task complexity and group knowledge
composition - on information sharing processes. The results in Study II suggested
that an increase in task complexity did not necessarily impair group performance.
Rather, group members performed better at complex tasks than simple tasks if they
had a well-constructed TM system; however, they performed worse if they
experienced high relationship conflict. Relationship conflict, not task conflict,
directly impaired group performance. Study III examined how group knowledge
diversity affected information sharing. Diversity research so far hasn’t been able to
draw a uniformly positive or negative link between diversity and performance. It is
likely that some intervening variables determine whether diversity would facilitate or
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
viii
impede performance. Study III argued that, although knowledge diversity would
potentially increase group performance by providing more resources to bear on the
task, it could also decrease performance by increasing disagreement on knowledge
criticality judgment among group members. The results supported the hypotheses.
More frequent communication among members with different expertise was
proposed as one solution to this problem. The results generated in the three studies
provide important managerial implications for knowledge management practices in
work groups. On the theoretical side, this dissertation encourages more detailed
specification of information sharing processes than what is currently prescribed by
TM. Additional contextual variables and information sharing processes should be
considered in order to produce the theoretical guidance that is actually actionable.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
CHAPTER ONE
INTRODUCTION
This dissertation project consists of three studies investigating issues related
to information sharing in work groups from the perspective of transactive memory
(TM). The topic of information sharing in work groups has gotten more and more
attention from researchers for two reasons primarily. First, work groups are
becoming the main units in organizations where most work gets done (Hackman,
1987). Second, with the continuing growth of the information economy, information
is now a critical asset to work groups and serves as a source of value for creating
competitive advantage (Spender, 1993). Effective information exchange, sharing,
and coordination have become primary determinants of performance (Snyder and
Morris, 1984). Despite the increasing awareness of the importance of information
sharing in work groups, group members often find that the information they have
been seeking is known by someone in the group, but information seekers are either
unaware of it or do not know where to find it, or coordination cost is so high that
neither information seekers nor information holders are motivated to share it.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Therefore, creating an effective system that facilitates sharing the information
residing in disparate individuals is essential to improving group performance.
Transactive Memory (TM) Theory (Wegner, 1987, 1995) has recently been
used to study the challenge of developing effective group information systems
(Hollingshead, 1998a, 1998b; Hollingshead, Fulk, & Monge, 2002; Moreland, 1999;
Moreland and Myaskovsky, 2000). TM theory identifies mechanisms for
specialization, communication, and coordination of information flows in groups. In
laboratory studies, groups with TM systems in place have been shown to be more
efficient and effective due to better coordination among group members
(Hollingshead, 1998a, 1998b, 1998c, 2000; Liang, Moreland, and Argote, 1995;
Moreland, Argote, and Krishnan, 1996; Wegner, Erber, and Raymond, 1991).
The notion of TM first appeared in the psychology literature two decades ago
(Wegner, Giuliano, and Hertel, 1985). The TM model is based on the foundation of
cognitive division of labor across persons. It argues that people don’t have to rely on
their own memories for all information. Instead, they can retrieve information from
other people’s memories when needed. With an aim to maximizing a group’s
cognitive capacity, each individual can serve as a repository for domain-specific
information by assuming or being assigned the responsibility for encoding, storing,
and retrieving that information. A differentiated and interdependent group memory
system effectively reduces the information load for each individual, promotes the
development of individual expertise, and enlarges the collective knowledge pool
available for all members.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3
The three studies included in this dissertation focus on distinct issues
associated with TM, or information sharing in general, in work groups. Study I tested
Wegner’s (1987, 1995) original TM model in natural group settings. It decomposed
the construct of TM and examined 1) how the antecedent of communication
frequency and contextual variables such as self-disclosure and group tenure impacted
the three TM processes including directory updating, information allocation, and
retrieval coordination; and 2) how these TM processes related to group performance.
Study I identified a significant insufficiency in the TM model when it was applied to
work groups - that is, information allocation was unrelated to group performance.
Possible explanations were proposed. Study II examined how the exogenous variable
of task complexity affected the development of TM in groups. Study II represented
one of the first endeavors to examine the role of task in TM functioning. It examined
1) direct effects of task complexity on TM functioning and indirect effects mediated
through task conflict and relationship conflict; and 2) how the effects of TM on
group performance differed in simple tasks and in complex tasks. It found that TM
had a strong performance effect in both complex tasks and simple tasks and the
effect was slightly stronger in complex tasks. In addition, contrary to earlier research
claiming that task conflict had a direct impact on group performance, task conflict
affected group performance through relationship conflict and TM.
TM processes, either in the form of allocation or retrieval, are essentially
processes of information sharing. We are all aware that not all information is created
equal. What information group members choose to share in a TM system is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
determined by their judgment on what information is critical to the focal task. Study
III thus examined how another exogenous variable - group knowledge composition -
affected information sharing processes. Specifically, 1) how group knowledge
composition affected the extent to which group members disagree on knowledge
criticality judgment; 2) how task uncertainty moderated this relationship; 3) the role
of communication; and 4) how group members’ disagreement on knowledge
criticality judgment affected group performance. The result showed that group
members differed in their prior knowledge (task knowledge and social knowledge)
which led to different knowledge criticality judgments. This result supported the
assumption that group members didn’t share unshared knowledge because they were
not aware that knowledge was critical to the focal task. Specifically, the effect of
knowledge diversity on disagreement on knowledge criticality judgment was found
stronger as task uncertainty increased. Disagreement on knowledge criticality
judgment had a negative effect on group performance. With respect to the role of
communication, communication structural diversity was found negatively associated
with members’ disagreement on knowledge criticality judgment and this association
was stronger in diverse groups than in homogeneous groups. Communication
strength was negatively related to members’ disagreement on knowledge criticality
judgment as well, but this relationship did not differ for diverse groups and for
homogeneous groups. Communication content diversity, however, was positively
related to members’ disagreement on knowledge criticality judgment and this
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
relationship was stronger in homogeneous groups than in diverse groups. Discussion,
limitations, and implications for future research were presented.
Most research on information sharing takes a behavioral perspective and
focuses on the behavioral pattern of sharing - that is, who retrieves what information
from whom (Comer, 1991; Morrison, 1993). Some researchers study information
sharing from the network perspective and examine how network features affect
information sharing (Borgatti and Cross, 2003). Miranda and Saunders (2003)
advance yet another view of information sharing and argue that one important
function of information sharing is social construction of meaning. However, little
research so far has taken a cognitive perspective on information sharing. Indeed,
group members often share information based on their perception of who has what
information. The current project centers around TM, a cognitive perspective on
information sharing. It examines how group members’ perception of information
location affects information sharing (see Figure 1.1 for a summary of the four
perspectives on information sharing in extant literature).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Information
Sharing
Figure 1.1 Four Perspectives on Information Sharing
Behavioral Perspective: Who shares what information with whom
(Comer, 1991; Morrison, 1993).
Cognitive Perspective: Information sharing based on perceptions of who knows what
(Wegner, 1987,1995).
Network Perspective: How do network features affect information sharing patterns?
(Borgatti & Cross, 2003)
Social Construction Perspective: Information sharing as social construction of meaning
(Miranda & Saunders, 2003).
Furthermore, little research has been conducted to examine the effects of
exogenous variables on information sharing, nor the way in which antecedents,
processes, and outcomes relate to one another in information sharing. The three
studies reported in this dissertation aimed to fill these gaps and examined
information sharing from the perspective of TM. These three studies looked at how
the TM model worked in natural groups, how the exogenous variables such as task
complexity and group knowledge composition affected information sharing
processes, and how information sharing processes influenced group performance.
Figure 1.2 displays how the three studies relate to one another. Specifically, Study I
examined the antecedent, processes, and outcome of TM. Study II investigated how
task complexity affected TM directly as well as indirectly through conflict, and how
TM affected group performance in complex tasks and simple tasks. Study III looked
at how group knowledge composition affected members’ disagreement on
knowledge criticality judgment, and how this disagreement influenced group
performance. In the next few chapters, I will discuss the rationale, methods, results,
and conclusion of each study in detail.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Figure 1.2 Overview of Three Studies
Outcome
Group
Performance
(Studies I, II,
& III)
Information Sharing
Processes
- TM Model
(Study I)
Exogenous Variables
Task Complexity
(Study II)
Group Knowledge
Composition
(Study III)
OO
9
CHAPTER TWO
STUDY I: EVALUATING TM IN WORK GROUPS
Most current research measured TM as one single construct and examined its
relationship with expertise recognition (Rau, 2000), communication (Hollingshead,
1998a, 1998b), and performance (Liang et al., 1995; Moreland and Myaskovsky,
2000). However, we are not sure how various processes or elements embodied in TM
relate to one another, nor do we know the conditions under which TM develops and
possible outcomes of such a development. Study I aimed to take a closer look at TM
by decomposing these elements to gain a better understanding of the conditions
under which TM evolves in work groups and the outcome of such as development.
We are interested in identifying antecedents, TM processes, possible performance
outcomes, and the pattern of the relationships among them in the context of natural
work groups.
Before we explain the reasons for evaluating Wegner’s TM model in work
groups, it is necessary to provide a brief overview of the TM model. TM consists of
a set of individual memories together with the communication links between
individuals (Wegner, Giuliano, and Hertel, 1985). TM comes into being when a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
person retrieves information stored in external storage, and/or when the information
stored in his or her own internal memory is retrieved and utilized by another person
(Wegner, 1987). This process is completed through communication. Wegner (1995)
identified three processes in the formation of TM: directory updating, information
allocation, and retrieval coordination. Directory updating is a process of learning
who knows what. Information allocation is a process of allocating incoming new
information to relevant experts for processing and storage. Information retrieval is a
process of bringing back uniquely stored information for task performance purposes.
The benefit of TM is to reduce individual’s cognitive load while enhance the
collective knowledge available to the whole group.
There are four reasons that warrant a detailed examination of the TM model
in work groups. First, Wegner, Giuliano, and Hertel (1985) initially conceptualized
TM in the context of intimate couples to describe processes for effective information
processing in intimate couples. Although Wegner (1995) attempted to theorize TM
processes in groups and organizations, the basic assumptions underlying his
theorizing effort were consistent with that of close relationships. Given the
substantial differences that exist in the nature of the relationship in intimate couples
and that in work groups, the assumptions on which the original TM model is erected
appear inappropriate when applied to work groups. For instance, the power
dimension of information that permeates work groups is largely missing in close
personal relationships. This difference would lead to different views about the
importance of information, different tendencies to withhold information, and more
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
11
importantly, how people collectively process and coordinate information in the two
settings. Recent research has applied TM to dyads other than couples (e.g.,
Hollingshead, 1998a, 1998b), work groups (e.g., Liang, Moreland, and Argote, 1995;
Moreland and Myaskovsky, 2000), and organizations (e.g., Anand, Manz, and Glick,
1998). Although the current literature has provided scattered empirical support of
TM in couples (Wegner, Giuliano, and Hertel, 1985) and groups (Watson,
Michaelsen, and Sharp, 1991), we are not confident enough to conclude that the TM
model described by Wegner (1995) accurately captures the cognitive aspect of group
collaboration.
Second, there are only two people involved in an intimate relationship
whereas a work group usually consists of more individuals. Wegner, Giuliano, and
Hertel (1985) predicted that analyzing TM processes in larger contexts might be too
hard to manage because their “framework for understanding transactive memory
would need to expand geometrically as additional individuals were added to the
system” (p. 257). Despite his initial pessimistic view, Wegner (1995) described the
three TM processes in groups and organizations without taking into account the issue
of size. Presumably, as a TM system involves more people, coordination cost
increases. This is because cognitive division of labor tends to be finer with the
increase in size especially in complex task conditions where no single individual has
all the information required to perform the task. The increased demand for
coordination in transactive encoding and retrieval in larger groups will discourage
them from engaging in these processes.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
Third, the variables described in the TM model and how they relate to one
another have not received adequate empirical assessment. Research on TM has been
focusing on communication (Hollingshead, 1998a, 1998b), expertise recognition
(Rau, 2000), and group performance (Liang, Moreland, and Argote, 1995; Moreland
and Myaskovsky, 2000). Many other variables remain unexplored. Research on
information seeking (e.g., Borgatti and Cross, 2003) and expertise coordination (e.g.,
Faraj and Sproull, 2000) draws on specific processes in TM, but no extant research
has provided an overall assessment of the TM model. Therefore, we are not clear
whether the relationships among the antecedent, the TM processes, and the outcome
variable described in the model are faithful representations of information exchange,
sharing, and coordination processes in work groups.
Fourth, most of the earlier research on group TM has been conducted in ad-
hoc groups in laboratories. Only recently Lewis (2003) successfully developed a TM
measure with field data. However, no research thus far has sought to evaluate the
theoretical validity of the TM model in natural settings. Therefore, we are not
confident whether the proposed relationships in Wegner’s (1995) TM model are
applicable to work groups.
With these questions in mind, Study I was designed to gather field data and to
use the structural equations modeling technique to evaluate the overall fitness of
Wegner’s TM model (1987, 1995) and revise it to better suit studies of group TM.
The TM model outlines the relationships among TM processes, their precursors, and
outcomes. The purpose of this study is to modify the original TM model to better
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
13
capture the cognitive aspect of group collaboration and to provide a theoretically
sound base model as a foundation for further research on group TM.
Antecedent to TM and Two Contextual Variables
Wegner (1987, 1995) identified an antecedent to TM processes -
communication frequency. A TM system consists of a knowledge network and a
communication network (Hollingshead, Fulk, and Monge, 2002). To utilize the
resources stored in the knowledge network, group members have to know who has
what knowledge. Communication is critical to acquiring such knowledge. Moreover,
individual memories are not physically connected, nor do people have the ability to
“read or write directly to each other’s memories” (Wegner, 1995). Therefore,
communication serves as a bridge between discrete memories to coordinate memory
sharing activities.
Communication Frequency and the Three Transactive Processes.
Communication has long been recognized as “the heart of the group interaction
process” (Shaw, 1964, p. 111). Communication is a multidimensional construct.
Wegner’s (1987, 1995) discussion on the role of communication in the functioning
of the three transactive processes - directory updating, information allocation, and
retrieval coordination - focuses on communication frequency.
An increase in communication frequency helps group members keep track of
changes in expertise directories. Greater amount of communication also leads to a
greater consensus on who knows what (Hollingshead, 1998a). Frequent and
extensive discussion on new information increases the likelihood that the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
information is remembered and allocated (Wegner, 1987) and improves group’s
transactive information processing capacity. Intense communication also allows
group members to develop “a common language for describing tasks” (Liang,
Moreland, and Argote, 1995). This shared language provides common labels for
various pieces of information. A shared understanding of the labels and the
information they stand for ease information allocation and the subsequent retrieval of
that information.
Hypothesis la: Communication frequency is positively related to
directory updating.
Hypothesis lb: Communication frequency is positively related to
information allocation.
Hypothesis lc: Communication frequency is positively related to
retrieval coordination.
Relate Two Contextual Variables to Directory Updating. Wegner (1987,
1995) identified two contextual variables - self-disclosure and group tenure - that
would affect members’ ability to recognize sources of expertise so as to update
expertise directories. All transactive information processing activities start with
expertise recognition. However, it is not always easy to correctly identify experts to
whom the group assigns information processing and storage responsibilities. It is
even more challenging to reach a consensus on expertise location. The difficulty
increases if we take a social perspective to expertise (Stein, 1997) which views
expertise perception as a joint product of the knowledge level of the expert and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15
social contextual variable, be it demographic, organizational (e.g., job title, span of
control), or social psychological (e.g., trust, accessibility). Different people allocate
different weights to these variables and make different judgment about who is the
expert in which area. Expertise identification does not necessarily require task
knowledge (Yetton and Bottger, 1982). Instead, self-disclosure and group tenure are
expected to enhance group members’ ability to identify expertise locations and
update expertise directories (Wegner, 1987).
Self-disclosure is defined as “the act of revealing personal information to
others” (Archer, 1980, p. 183). Self-disclosure has been found to be an important
precondition to the development of differentiated memory structures among intimate
couples (Wegner, Giuliano, and Hertel, 1985) and other dyads such as strangers
(Hollingshead, 1998b). In work settings, although people are unlikely to disclose
private thoughts to the same degree as they do with significant others, they
nonetheless share their hobbies, life experiences, values, preferences, opinions, and
other less intimate thoughts (Wegner, Erber, and Raymond, 1991). Self-disclosure of
this kind will help them know themselves better (Archer, 1980), which, in turn,
increase the likelihood of revealing such information to their work associates. Self-
disclosure brings about new information (Fisher, 1984) and generally follows the
norm of reciprocity (Derlega, Metts, Petronio, and Margulis, 1993). This implies that
with each self-disclosure encounter, all parties involved are likely to gain new
information about one another which leads to the formation and/or adjustment of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
16
expertise assessment. An increase in members’ ability to assess expertise will
improve the quality of directory updating.
Hypothesis 2a: Self-disclosure is positively related to directory
updating.
Group tenure is also likely to be positively correlated with directory updating.
Locating sources of expertise poses a challenge especially for groups with brief
tenure where members are presumably less familiar with one another. Such groups
are often found less capable of pooling unique information because of the absence of
accurate directory of who knows what (Gruenfeld, Mannix, Williams, and Neale,
1996). Longer group tenure will enhance members’ ability to keep their expertise
directories up-to-date for four reasons. First, as members work together for a longer
period of time, they will acquire a relatively accurate perception of who is good at
what through active verbal communication as well as passive observation.
Observation is particularly effective at learning sources of tacit knowledge (Comer,
1991; Morrison, 1993; Wegner, 1995). Group members who have been together
longer are likely to know more about one another’s background and past experience
that allows them to “discern with much greater precision just who is expert in each of
a variety of information domains” (Wegner, 1987, p. 191). Second, Katz (1978)
argues that workers usually focus their energy on relationship building in the initial
stage of their employment. Once they are socialized into the new environment, they
shift their attention to the task itself and become more attentive to the information
concerning one another’s expertise.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17
Third, a shared understanding of the various dimensions of the group is often
developed as members collaborate for a long period of time (Katz, 1982).
Disagreement on expertise locations is expected to decrease with group tenure as
shared understanding increases (Pelled, 1996). Fourth, members in groups with
longer standing may develop less ambiguous cognitive division of labor over time,
having designated individuals responsible for storing and retrieving information in
specific domains (Stasser, 1992; Wegner, 1987). Well-specified role structures
evolved in groups with longer tenure also facilitate accurate expertise recognition,
and hence directory updating (Libby, Trotman, and Zimmer, 1987).
Hypothesis 2b\ Group tenure is positively related to directory
updating.
Interrelations Among Three TM Processes
Wegner (1995) identified three transactive processes directory updating,
information allocation, and retrieval coordination. In this section, I will discuss the
interrelations of these three processes.
An important component of TM is directory updating. Directory updating
occurs when group members learn about one another’s expertise areas and develop
expertise directories of who knows what (Wegner, 1987). Expertise directories
“maximize both the speed of the search and its likelihood of finding the needed
information” (Wegner, 1995, p. 325). For instance, a customer is looking for a
computer desk at Office Depot. The staff at the store knows that John, an office
furniture salesman, knows best about computer desks and is most likely to make the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
18
right recommendation because computer desk is office furniture. According to their
expertise directory, store staff will direct customer’s inquiry about computer desks to
John right away. Utilizing store staffs expertise directories speeds the customer’s
information search process and the likelihood of finding a good computer desk.
Without an up-to-date and accurate directory, group members would not be able to
coordinate expertise through transactive encoding and retrieval.
Information allocation may facilitate retrieval coordination. Explicitly
assigning experts information processing and storage responsibilities helps to pool
nonredundant information (Stasser, Stewart, and Wittenbaum, 1995). According to
Wegner (1995), allocating information to designated experts increases the accuracy
and the speed of retrieval when there is more than one information source available.
Delegating information processing responsibilities inconsistent with individual
expertise leads to consistent failure in recalling unique information in groups.
Hypothesis 3a: Directory updating is positively related to information
allocation.
Hypothesis 3b: Directory updating is positively related to retrieval
coordination.
Hypothesis 3c: Information allocation is positively related to retrieval
coordination.
Transactive Processes and Group Performance Outcome
Transactive information processing increases the amount, quality, and
diversity of information groups have access to, and brings about positive
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19
performance effects (Moreland and Myaskovsky, 2000; Rau, 2000; Wegner, Erber,
and Raymond, 1991). Information allocation leads to specialization and the need for
cooperation among group members. A cooperative work environment helps
members develop positive perceptions of their group experience that will enhance
performance (Janz, Coquoitt, and Noe, 1997).
Whether members are able to utilize the information recalled by experts is
also essential to group performance. Research shows that in decision-making groups,
those adopting the “best-member strategy” (i.e., letting the “best” member make a
decision for the whole group) can reach a comparable performance level with groups
making decision collectively (Henry, 1995; Yetton and Bottger, 1982). Group’s
reliance on erroneous information provided by non-experts will potentially
undermine group performance (Rau, 2000).
Hypothesis 4a: Information allocation is positively related to group
performance.
Hypothesis 4b: Retrieval coordination is positively related to group
performance.
Based on the hypotheses, a structural equations representation of the TM
model was presented in Figure 2.1. All relationships are positive.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Figure 2.1 Study I: A Structural Equations Representation of TM Model
SD1
Self-
Disclosure
fSDl
DUl DU2 DU3 DU4
5
SD2
5
Directory
Updating
(DUl
SD3
8
SD4
Group
Tenure
(GT)
IA1 IA2 IA3
5 >
GT1
GP1
8 > GT2
Information
Allocation
(IA)
Group
Performance
(GrPl
GP2
8 CF1
GP3
8 *
CF2
Communication
Frequency
(CF)
CF3
Retrieval
Coordination
IRQ
8
8
CF4
CF5
8
8
RC2 RC1
CF6
s s
Method
Participants
In Study I, three hundred and eighty MBAs and MACCs (Masters of
Accounting) participated in the survey. These participants were from five sessions of
negotiation class and three sessions of accounting class. The participants worked in
3- to 5-person groups on semester-long group projects. Among all participants, 224
(58.9%) were male and 156 (41.1%) were female; 162 (42.6%) were first-years and
218 (57.4%) were second-years. Half of the participants were Caucasians, 38.4%
Asians, 5.8% Hispanics, 1.1% African Americans, and 4.7% of other ethnicities.
Most of the participants had much group work experience. Around 90% of them had
worked on more than 4 group projects before.
Procedures
With an aim to collecting data for Study I, the approval of research was
obtained from the University Internal Review Board before the survey. The
researcher contacted the instructors to see whether they would like to have the
researcher solicit participation in their sessions. After obtaining permissions from the
instructors, the researcher requested group membership information and randomly
assigned a numerical number to each group. The researcher then created a
transparency that listed group numbers and the names of the students that belonged
to each group. The survey was conducted at the beginning of the session. First, the
researcher handed out consent forms. The students decided whether they would like
to participate after reading the consent form. Then, those who agreed to participate
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
22
proceeded to the survey. Approximately 92% of the students participated in the
survey. The group membership information and group numbers were displayed on an
overhead projector. The participants were instructed to write down their group
number in the space provided on the top right comer of the questionnaire. The
researcher’s email address was provided in case the participants would like to learn
more about the study.
Instrumentation
The questionnaire for Study I consisted of demographic items and items
tapping seven constmcts: self-disclosure, group tenure, communication frequency,
information allocation, directory updating, retrieval coordination, and group
performance. Participants were asked to indicate on a five-point Likert scale (ranging
from “strongly agree” to “strongly disagree”) their responses to the items measuring
each constmct.
Self-Disclosure. Four items were employed to assess self-disclosure. Knecht,
Lippman, and Swap (1973) composed and empirically evaluated the validity of a
self-disclosure scale in the context of friendships. This scale included sixteen items,
four for each of the four levels of intimacy ranging from high intimacy to low
intimacy. Since collaborating on a group project is a work-related task, the students
were less likely to share such thoughts on parental relationships or drinking habits as
described in Knecht, Lippman, and Swap’s (1973) scale. Therefore, the researcher
took two items (i.e., “I tell others in my group what I need the most help with” and “I
tell others in my group my hobbies”) relevant to the present study and wrote another
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
23
two items “I tell others in my group things I am good at” and “I tell others in my
group things I am not good at” to measure self-disclosure.
Group Tenure. Group tenure was measured with two items: (1) the number of '
weeks the group worked together on the project; and (2) the average number of hours
the group spent on the project each week. Group tenure was the product of these two
indices. This measure should better capture group tenure in student work groups and
its effect on group performance than a single indicator scale. This is because students
usually wait until the last minute to work on group projects. The number of hours
they spend on the project can vary quite a bit at different stages with most hours
devoted to a couple of weeks before the project’s due date. A product of the two
items adjusts for this variation and is a more accurate measure of the actual time
students spend on the project.
Communication Frequency. The construct of communication frequency was
measured with Smith et al.’s (1994) scale. This scale was chosen because it has been
empirically proved to be a reliable scale, Cronbach’s Alpha = .73.
Directory Updating. Four items were included to measure the frequency and
accuracy of directory updating both within and outside the group. Among the four
items, two items were written to assess the frequency of updating expertise
directories of the members both within the group and outside the group with whom
the respondent had work relationship with and two items were written to evaluate the
accuracy of directory updating. The participants in this study were in their first term
in the program. Few of them knew one another before they took this class. Their
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
24
knowledge about who knew what in the group was close to zero when they started
the project. Therefore, the directory updating measure in Study I was not confounded
by the differences in their prior knowledge of one another’s expertise. Also, it is
important to assess directory updating both within and outside the group. Research
has found that both internal and external communication is critical to the successful
performance of knowledge-intensive project groups because the members of such
groups rarely have all the knowledge and skills required for the task and have to
obtain expert knowledge from both within and outside the group (Ancona and
Caldwell, 1992; Katz, 1982).
Information Allocation. Three measures were written by the researcher to
evaluate the extent to which the respondent allocated information to other group
members, other group members allocated information to the respondent, and the
group as a whole engaged in information allocation.
Retrieval Coordination. Two items from the TM scale (Hollingshead, 2001)
were used to measure retrieval coordination both within the group and outside the
group.
Group Performance. Hackman (1987) used three items to measure group
performance. The three items tapped three dimensions of group performance:
productivity (measured by the participants’ grades on the group project), members’
satisfaction with their group experience, and their ability to work together on future
projects. These three items collectively served as an index of group performance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
Appendix A includes the items used to measure the seven constructs in this
study.
Analysis
In Study I, three hundred and eighty observations were included in data
analysis. LISREL VIII was the statistical package used to analyze the data. LISREL
allows for an assessment of the model both on the local level by estimating the
individual structural coefficient between two latent variables or between a latent
variable and an observed variable and on the global level by providing a variety of
goodness-of-fit indices for the model as a whole. Bollen (1989) and Joreskog and
Sorbom (1996) suggested several rules be used to assess model fit and explained
reasons for using each criterion: (1) minimum fit function chi-square statistic divided
by degrees of freedom. A model is a good fit if this value is smaller than 5; (2) p
value. An insignificant p value (i. e.,> .05) suggests that the hypothetical model
adequately captures the real relationships between variables; (3) goodness-of-fit
index (GFI) and adjusted goodness-of-fit-index (AGFI). AGFI is GFI adjusted for
degrees of freedom and is usually lower than GFI. AGFI is preferred over GFI
because it is insensitive to sample size. AGFI and GFI of a well-specified model are
generally greater than .90; (4) root mean squared residual (RMR). RMR is an index
of residual average. A good model would show a RMR less than the .05 ceiling.
RMR is especially favorable in a situation with standardized indicators; (5)
comparative fit index (CFI) and incremental fit index (IFI). CFI and IFI represent the
difference between the hypothesized model and the null model. If both CFI and IFI
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
are greater than .90, the hypothesized model is a good fit; (6) coefficient of
determination. This index explains the amount of variances accounted for in each
dependent variable.
Since one criterion is not adequate for reaching a definite conclusion about
the general fitness of a model, multiple criteria were used in the analyses of
measurement models and the theoretical model in this paper. When assessing the
significance of each structural coefficient, t-values were reported.
Results
Table 2.1 summarizes the means, standard deviations, and correlations
among the variables.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 2.1 Study I: Descriptive Statistics and Correlations
Variable Mean s.d. DU IA RC GP SD GT CF
DU 2.78 .82
—
IA 2.91 .73 .100
_
RC 2.37 .69 .321** -.238*
_
GP 2.36 .79 .416** -.012 .404*
_
SD 2.38 .66 .451** -.018 .279** .147
_
GT 4.40 1.55 -.093 -.038 -.154 -.023 -.201*
CF 2.62 .64
417**
-.218* .385** .212* .293** -.148
Note: DU = Directory updating; IA = Information allocation; RC = Retrieval coordination; GP = Group performance; SD =
Self-disclosure; GT = Group tenure; CF = Communication frequency (**p < .001; *p < .05; N = 103).
ro
Measurement Models. Factor analyses of the measurement models were
conducted before testing the theoretical model. Thus, only reliable measures were
included in the assessment of the hypotheses. Self-disclosure was measured with
four items, t-ratios suggested that all four indicators were significantly loaded on the
latent factor. The minimum fit function chi-square was 6.71 with 2 degrees of
freedom. The chi-square divided by degrees of freedom was below 5 and the p value
was slightly greater than .05 (p = .053). Other goodness-of-fit indices (e.g., GFI =
.96, AGFI = .90, CFI = .91, IFI = .91, RMR = .057) suggested that the overall fitness
of the model was acceptable. The coefficients of determination of the indicators were
.05, .37, .37, and .47, respectively. All four items were kept for hypothesis testing.
The communication frequency measurement model was analyzed as a single
factor model which generated the following results, chi-square = 25.46, dfs = 9, p =
.0025, GFI = .92, AGFI = .82, CFI = .65, IFI = .69, RMR = .10. The overall
coefficients of determination ranged from .01 to .72. Two links showed non
significant t ratios and were excluded from the analysis of the theoretical model.
Directory updating was initially analyzed as a single-factor model. The
modification indices suggested that a two-factor orthogonal solution produce a better
fit. The two-factor orthogonal solution generated a chi-square of 5.92 with 2 degrees
of freedom, p = .08, GFI = .90, AGFI = .86, CFI = .92, IFI = .92, RMR = .06. The
coefficients of determination ranged from .19 to .57. All t-ratios were significant;
therefore, all items were included for hypothesis testing.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
29
Information allocation and group performance were each measured with three
items. These are saturated (i.e., just-identified) models and cannot be evaluated with
LISREL. The use of reference indicators did not solve the problem. Therefore,
Cronbach’s Alpha was computed to assess the reliability of each scale (Cronbach’s
Alphas = .74 and .78, respectively). Retrieval coordination was measured with two
items. The Cronbach’s Alpha for this scale was .12:
The Theoretical Model. LISREL VIII was employed to analyze the
theoretical model. LISREL VIII, like regression, assumes that the observations are
independent. However, the responses provided by the members of the same group
are not independent given their shared group experience. With an aim to avoiding the
violation of the independent observation assumption, 380 cases in the raw data were
aggregated into group data that produced 103 cases (i.e., groups). The satisfaction of
two criteria provided justification for this aggregation (Edmondson, 1999; Kenny
and LaVoie, 1985). First, the variables included in this study were meaningful
theoretically as group variables. Second, the intraclass correlations (ICC) for group
variables were greater than zero, a criterion that suggested a convergence among the
responses within a group and a systematic difference across groups (see Table 2.2 for
ICCs and analysis of variance).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 2.2 Study I: Analysis of Variance and Intraclass Correlation Coefficients
for Group-Level Scales
F (1 0 2 , 380) P
ICC
Self-Disclosure 2.29 <.001 .24
Communication Frequency 1.96 <.001 .09
Directory Updating 2.79 <.001 .31
Information Allocation 3.34 <.001 .44
Retrieval Coordination 2.37 <.001 .26
Group Performance 4.42 <.001 .64
O J
o
LISREL VIII was performed on group data (N=103). The results showed that
Hypotheses la, lb, lc, 2a, 3a, 3b, 3c, and 4b received full support while the other
hypotheses were not supported (see Figure 2.2 for details). Specifically,
communication frequency exhibited positive associations with all three transactive
memory processes: directory updating, information allocation, and retrieval
coordination; self-disclosure was found to be positively related to directory updating;
directory updating was positively related to information allocation as well as
retrieval coordination; information allocation was positively related to retrieval
coordination; and retrieval coordination was positively related to group performance.
Various goodness-of-fit statistics provided mixed support for the model fit. The
minimum fit function chi-square was 15.40 with 8 degrees of freedom (p = .052).
Clearly, the chi-square divided by degrees of freedom was lower than 5. This result
and other statistics such as GFI (.96), CFI (.94), and IFI (.94) suggested that the
model was a good fit. However, other goodness-of-fit indices were less satisfying,
AGFI = .86, RMR = .062. The overall coefficient of determination was .19.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Figure 2.2 Study I: Results of Original TM Model and Modified Measurement Models (*p < .05;N = 103)
5
DU2 DU1 DU3 DU4
SD1
Self-
Disclosure
(SD)
5 -+
SD2
Directory
Updating
(DU1
5
SD3
5 -► .03
SD4
.25* j s
IA3
ir'
5 -* Group
Tenure
(GT)
GTl
IA1 IA2
GP1
5 - ►
GT2
Information
Allocation
(IA1
Group
Performance
(G?)
.09
GP2
.31*,
.35^ GP3
CFI
9Q*T
.24*
.43*
Communication
Frequency
(CF)
CF2
Retrieval
Coordination
(RC)
8
8 *
CF4
RC1 RC2
CF5
8
8 8
U )
K )
33
Model Modification. Model revisions were conducted in two stages. In the
first stage, links with non-significant t-values were deleted. These links were deleted
one at a time until all non-significant links were removed, a procedure suggested by
Kaplan and Wenger (1993) to better capture the dynamic changes in the model. The
revised model fit the data better than the original one, chi-square = 16.35, degrees of
freedom = 10, p = .090, GFI = .96, AGFI = .86, CFI = .94, IFI = .95, RMR = .066.
The overall coefficient of determination was .27. Modification indices suggested that
a significant positive relationship exist between directory updating and group
performance. In the second stage, the link between directory updating and group
performance was added.
The modified model in stage 2 improved significantly. The minimum fit
function chi-square was 4.94 with 9 degrees of freedom, p = .840. The chi-square
divided by degrees of freedom was well below 5. Other goodness-of-fit indices
suggested a good model fit, GFI = .99, AGFI = .96, CFI = .99, IFI = .99, RMR =
.032. The average coefficient of determination was .34. The comparison between the
modified model and the null model suggested that the modified model was
significantly better than the null model. The goodness-of-fit indices for the null
model were as follows: chi-square = 112.70, degrees of freedom = 18, p = .000, GFI
= .76, AGFI = .63, CFI - .28, IFI = .30, RMR = .210. The overall goodness-of-fit
indices for the original theoretical model, modified models, and the null model were
summarized in Table 2.3. The final modified model was presented in Figure 2.3.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 2.3 Study I: Summary of Model Revisions
Revisions
(MI)
Chi-
square
Degrees
o f
freedom
p value GFI AGFI CFI IFI RMR CED
Hypothesized
Model
15.40 8 .052 .96 .86 .94 .94 .062 .19
Stage 1
Delete all non
significant
links
16.35 10 ..090 .96 .88 .94 .95 .066 .27
Stage 2
Add one link
4.94 9 .840 .99 .96 .99 .99 .032 .34
Null model 112.70 18 .000 .76 .63 .28 .30 .210 .00
Note: MI = Modification index; GFI = Goodness-of-fit index; AGFI = Adjusted goodness-of-fit index; CFI = Comparative fit
index; IFI = Incremental fit index; RMR = Root mean square residual; CED = Overall coefficient of determination.
C O
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Figure 2.3 Study I: Results of Modified Theoretical Model and Modified Measurement Models (*2 < .05; N = 103)
8
DU2 DU1 DU3 DU4
SD1
Self-
Disclosure
(SD)
5 -+ .36*
SD2
Directory
Updating
(DU1
8
SD3
SD4
8 32*
IA1 IA2 IA3
_ + - 1 T 2 5 * 4
GP1
Information
Allocation
(IA1
Group
Performance
(GP)
.31* GP2
.35* .24*
CFI
8 " H
GP3
Communication
Frequency
(CF)
.20*
CF2
8 .30*
Retrieval
Coordination
(RC)
CF4
8
CF5
8
RC1 RC2
s s
u >
36
Discussion
Study I was designed to examine how the contextual variables such as self
disclosure and group tenure affected directory updating, how the antecedent of
communication frequency was related to TM processes, the interconnections among
TM processes, and how TM processes affected group performance. Significant
positive relationships were found between self-disclosure and directory updating,
between communication frequency and each of the three transactive processes,
between directory updating and information allocation as well as retrieval
coordination, between information allocation and retrieval coordination, and between
retrieval coordination and group performance. The modification indices suggested a
significant positive relationship between directory updating and group performance.
It was reasonable to find that self-disclosure was positively related to
directory updating. If group members reveal to others in the group information on
their own strengths and weaknesses, others will have a better idea about what
information they have, attribute relevant expertise to them, and update their expertise
directories accordingly. According to Fisher and Ellis (1990), mutual self-disclosure
produces a “snowball effect.” In other words, if Person A initiates self-disclosure and
reveals personal information to Person B, Person B is likely to reciprocate with self
disclosure behavior and sometimes discloses even more personal information to
Person A. Thus, one person’s engagement in self-disclosure will lead to self
disclosure of everyone else in a group. Eventually all group members know one
another’s expertise and update their expertise directories.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
It was surprising to find that group tenure was not positively related to
directory updating. A post-hoc regression analysis showed that the best fitting line
would be described by the quadratic (curvilinear) function, f = -.02, Rj = .04, Rf
change = .02, overall F = 3.19, p = .04. In the early stage of group formation, group
members spend much time building relationships and then gradually shift their focus
to the task (Katz, 1982). In this early stage, group members learn more and more
about one another’s expertise as their communication shifts from relationship to task.
However, as groups continue to work together, they often experience a reduction in
task communication (Katz, 1982) and share less unique information (Mennecke and
Valacich, 1998). The decrease in task communication and information sharing
creates an obstacle to accurate expertise recognition because group members don’t
keep one another updated of the changes in the location of expertise any more. To
make the situation even worse, group members become more confident in their
judgment of who knows what overtime (Storey, 1991). As time goes on, inaccurate
expertise recognition impairs performance and forces group members to engage in
more task-related communication and share more unshared information. This effort
improves the members’ ability to identify sources of expertise and update their
directories with accurate and current expertise location information.
Communication frequency was positively related to each of the three
transactive processes: directory updating, information allocation, and retrieval
coordination. These findings are consistent with the hypotheses. For each transactive
process to function effectively, communication is the key because group members’
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
individual memories are not physically connected and need communication to link
them together.
Also consistent with the hypotheses, directory updating was positively related
to both information allocation and retrieval coordination. Information allocation was
positively related to retrieval coordination. For information allocation and retrieval
coordination to function smoothly, group members have to update their expertise
directories and know who knows what. When Person A allocates superior
information to Person B, Person A should know for sure that Person B is the expert
in that area and has expert information. When Person A needs information relevant
to that area, he or she may consider retrieving the information from Person B.
Inconsistent with the hypothesis, information allocation was not related to
group performance. There are multiple possible explanations for this finding. One
possible explanation is that the participants in this study were not aware of the
benefits of information allocation and simply did not allocate enough information to
other members in their group. The lack of information allocation might be the cause
of its non-significant effect on the performance outcome. Two reasons may account
for the absence of information allocation in groups. First, high coordination cost
involved in information allocation may discourage group members from engaging in
information allocation behavior. Second, job-related expertise is viewed as power in
the workplace. As Hardin (1995) argued, “coordination of a group is potentially
political” (p. 56). Job-related expertise or specialty is inherently power-laden
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
(Eisenberg and Phillips, 1991). Whether Person A is willing to pass the information
to the expert Person B to enhance Person B’s power is questionable.
Another possible explanation for the lack of a significant association between
information allocation and group performance is that information allocation may not
work as effectively as described in the theory. When non-expert Person A passes
information to the expert Person B, information degradation occurs. This process is
important because Person A may not be able to process the information as effectively
as Person B. The negative effect produced by information degradation may
counterbalance or even outweigh the benefits brought by information allocation.
Some people may argue that the lack of significant effects of information
allocation on group performance is due to the lack of knowledge differentiation in
student work groups. While this argument is theoretically plausible, the data from
Study III did not fully support this argument. Study III was conducted among student
work groups, same as Study I. The data in Study III suggested that there were some
variations in knowledge diversity. Among sixty-four groups, the lowest domain
diversity index was .00 and the highest was 2.24. The mean domain diversity was
.56, SD = .31. For expertise level diversity, the lowest score was .00 and the highest
was .62. The mean was .23, SD = .11. The lowest social knowledge diversity index
was 1.88 and the highest was 5.00. The mean of social knowledge diversity was
3.17, SD = .61. Whether these variations are great enough to permit valid tests of the
hypotheses on information allocation is unknown and needs further clarification.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
It was not surprising to find that retrieval coordination was positively related
to group performance outcome. Poor information retrieved from a person’s memory
is subject to public inspection and may be improved by other members. As a result,
this person may update his or her memory with more accurate information.
Continuously processing the information concerning a specific knowledge domain
also fosters the development of individual expertise and will potentially boost group
performance. The unique information recalled by each individual is integrated
through group discussion. New knowledge may be generated that will enlarge the
collective knowledge repository and enhance performance. The retrieval process also
brings together diverse expertise to complete the task. A group’s collective
information processing capacity is substantially greater than that of any single
individual. The group’s ability in carrying out the task is enhanced. Besides,
coordination involved in information retrieval will help group members develop a
sense of cooperation that will contribute to the positive experience of all members.
The results showed that there was an indirect link between directory updating
and group performance that was mediated by retrieval coordination. It is intuitively
plausible that if group members know other’s area of expertise and retrieve expert
information to bear on the task at hand, group performance is more likely to be
enhanced. However, the modification indices also suggested that there exists a direct
link between directory updating and group performance. It seemed that the
knowledge of who knows what would enhance performance even if group members
did not necessarily retrieve the information from the experts. Study II suggested that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
41
interpersonal frictions may prevent information seekers from retrieving information
from recognized experts. The fact that Person A recognized Person B as an expert
does not necessarily lead Person A to initiating an information request. As Borgatti
and Cross (2003) argued, information retrieval is often affected by interpersonal
relationships.
Limitations. Like any other study, Study I has its limitations. One limitation
concerns the sample. The current study was conducted among student work groups.
Student work groups differ from real work groups along two important dimensions.
First, the role structure in student work groups may not be as clearly specified as that
in real work groups. The lack of a well-defined role structure may cause ambiguity in
information allocation. Furthermore, a post-survey conversation with the participants
revealed that information allocation in the student groups in this study was mostly
based on shared workload rather than expertise levels. Thus, the hypothesized
relationship between information allocation and group performance could not be
adequately assessed by the current data.
Second, the results of factor analyses showed that the communication
frequency scale needed to be improved. Two items were not significantly loaded on
latent factors. In addition, the fitness indices provided mixed support for model fit.
The values of GFI and AGFI (GFI = .92, AGFI = .82) were much higher than the
values of CFI and IFI (CFI = .65, IFI = .69). This result suggested unstable
modeling. Future research is needed to develop a more stable scale to measure
communication frequency.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
42
Implications and Directions for Future Research. Although a number of
studies have provided empirical evidence for TM (e.g., Watson, Michaelsen, and
Sharp, 1991; Wegner, Erber, and Raymond, 1991), the specific mechanisms through
which TM operates differs across settings. The modified model in this study sheds
light on the working mechanisms of TM in student work groups. Although groups in
school settings differ from those in real work settings in some important ways,
information allocation appears problematic and demands our attention. Information
allocation was found unrelated to group performance outcome. This makes me
wonder whether information allocation is as effective as the TM theory has predicted
it to be. To what degree does information allocation contribute to group
performance? In work groups, information allocation occurs most frequently in the
form of assigning tasks according to role specifications. Task-related skills required
for performing the task are largely acquired by task performers themselves.
Allocating information may not be as effective as we have thought it to be.
High coordination cost, power dynamics, and information degradation also
discourage information allocation and decrease its performance benefits. For
instance, transactive processes described by Wegner (1987, 1995) are based on
intimate couples. They do not necessarily capture information coordination processes
in real work settings. Information may lose much of its political power in couples.
This is because the relationship between couples is often based on emotions whereas
the relationship between co-workers often has a stronger instrumental component.
Moreover, a couple only involves two people whereas a group or team has more
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
members. As the size of the group increases, the difficulty and cost of coordination
increases too. Much of the earlier work on TM was done in labs with dyads or three-
person groups. The issues of power and size have not been well investigated in lab
studies. More research is called for to examine effects of size and power on TM
functioning in natural groups. Also, as information is passed from non-experts to
experts, information degradation occurs and weakens the positive effect of
information allocation on group performance. Further research should gather TM
data in groups in real work settings to further examine information allocation process
and its relation to group performance outcome.
The results produced in this study have important implications in two areas:
theoretical development and practical guidance for management practices. On the
theoretical side, it cautions us about the importance of context in which theories are
formulated. When we apply a theory developed in one context to another context, we
have to take into consideration contextual differences that would potentially affect
the mechanisms through which the theory operates. The next step on the agenda will
be to develop a new model for group memory that will overcome the insufficiency of
the TM model.
On the practical side, TM has the potential to maximize a group’s
information processing capacity. Having an effective TM system in place is critical
to performance especially in groups performing knowledge-intensive tasks such as
product development and consulting. The TM model is framed in the context of
intimate couples. The nature of close personal relationship is substantially different
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
from that of work group relationship. Whether the TM processes described in the
context of dating or married couples apply to work groups remains unclear. The
results of this study have improved our understanding of the working mechanisms of
group TM to inform our practice in managing knowledge intensive groups.
Companies should seek to construct effective TM systems based on the model
outlined, modified, and validated in this study to enhance group performance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
CHAPTER THREE
STUDY II: EFFECTS OF TASK COMPLEXITY AND CONFLICT
ON TM AND ITS RELATION TO GROUP PERFORMANCE
Study I revealed the insufficiency of Wegner’s TM model when applied in
work groups and called for more detailed specifications especially on the context.
Study II extended the original TM theory and examined factors such as task and
conflict that would influence TM functioning and conditions under which TM would
contribute most to group performance. Earlier research on group TM has focused
heavily on expertise recognition (Rau, 2000) and group performance (Liang et al.,
1995; Moreland and Myaskovsky, 2000). Many other variables that potentially affect
TM functioning remain unexplored. For instance, many researchers studying groups
from a functional perspective argue that the type of task groups perform dictates the
processes needed for optimal performance (Hollingshead, et. al., 2005). In Study II, I
focused on task complexity. Presumably, tasks with differential complexities impose
differential demands on the knowledge and skills of task performers and on expertise
coordination among them. Compared with simple tasks, complex tasks require more
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
46
diverse expertise and/or more coordination. TM is expected to have stronger effects
on group performance in complex tasks than in simple tasks.
Rulke and Galaskiewicz (2000) found that groups of generalists
outperformed groups of specialists and mixed groups. They argued that this was
because generalists did not have to retrieve information from experts, and therefore
minimized coordination costs. They also found that specialist and mixed groups
performed as well as generalist groups if they had a decentralized communication
network. Compared with simple tasks, complex tasks require group members to
process more information. Given human limited cognitive capacity, information in
groups performing complex tasks is more likely to be distributed in disparate
individuals. In other words, groups performing complex tasks are more likely to
consist of specialists than generalists. Whether specialists in these groups are capable
of effectively retrieving information from experts is a major determinant of whether
specialist groups can perform well. Performing simple tasks, on the other hand, does
not require group members to process much information. Groups dealing with simple
tasks are more likely to consist of generalists rather than specialists. Everyone in the
group is capable of processing and storing all information related to simple tasks.
Information retrieval is optional. Therefore, TM should not matter much to the
performance of simple tasks in generalist groups.
Besides theoretical support, various practices by managers and speculations
by researchers have also suggested that TM be more important to the performance of
complex tasks than simple tasks. For instance, many consulting firms are building
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Al
knowledge bases where employees’ expertise areas are listed and contact
information is provided so that knowledge seekers know where to go for what
knowledge (Moreland, 1999). Lewis (2003) predicted that the performance effects of
TM would be especially conspicuous in knowledge-worker teams. She argued that a
team’s ability to locate, coordinate, and utilize distributed expertise was critical to
the successful completion of complex tasks such as management consulting.
However, she did not lay out the theoretical mechanism underlying her speculation;
nor did she provide empirical evidence to support her speculation. Study II was
conducted to examine the role of task complexity in TM functioning and in the TM -
- group performance interaction.
Furthermore, previous empirical research on TM did not look at the effects of
conflict (including both task conflict and relationship conflict) on TM. Wegner’s
(1987, 1995) formulation of TM theory proceeded on an overly rational basis by
assuming that experts would be willing to share their expertise and that knowledge
seekers would be delighted to turn to experts for superior knowledge. This
assumption is more likely to hold for romantic couples than for work groups where
job-related expertise is inherently power laden (Eisenberg and Phillips, 1991) and
coordination of expertise is frequently political (Hardin, 1995). Conflict would
potentially influence expertise coordination in TM, yet this dynamic has been
overlooked in both TM theory development and empirical research.
With an aim to filling these gaps in extant research, Study II has been
designed to address two questions: (1) how task complexity affects TM directly and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
48
indirectly through conflict; (2) how TM impacts group performance, and whether the
magnitude of this impact differs for complex tasks and for simple tasks. The model
developed and validated in this study is designed to advance theory building. Also it
is expected to produce specific guidelines on what managers need to do to improve
group performance in complex task conditions, such as product development and
management consulting.
Defining Task Complexity
Researchers have described tasks along various dimensions such as task
variety (Daft and Macintosh, 1981), task analyzability and routineness (Perrow,
1967), task uncertainty (Van de Ven, Delbecq, and Koenig, 1976), task complexity
(Campbell, 1988; Wood, 1986), task diversity, unpredictability, and interdependence
(Scott, 1998). These dimensions overlap to various degrees. For instance, both task
uncertainty and routineness have been construed as tapping two factors, task variety
and analyzability. Task complexity, an information processing approach to task
classification (Lawrence and Lorsch, 1967; Thompson, 1967), is closely associated
with all other dimensions. A task can be complex if it is uncertain, or non-routine, or
less analyzable, or highly variable. Study II employs Wood’s (1986) definition of
task complexity because this definition emphasizes the knowledge and skill demand
placed on task performers - the focus of TM theory, while more group-orientated
task classifications (e.g., McGrath, 1984; Shaw, 1973; Steiner, 1972) emphasize
other aspects such as performance processes or group product.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
According to Wood (1986), task complexity consists of three dimensions:
component complexity, coordinative complexity, and dynamic complexity.
Component complexity is the number of distinct acts to be executed and the number
of information cues to be processed in performing a task. Coordinative complexity
represents the nature of the relationship between task inputs and outputs. “The form
and the strength of the relationship between information cues, acts, products, as well
as the sequencing of inputs” (p. 68) constitute various dimensions of coordinative
complexity. The third dimension of task complexity is dynamic complexity, which
results from changes in the stability of the relationship between task inputs and
products over time during the course of performing a task. Wood (1986) argues that
although the three components of task complexity are not completely independent,
“the level of each complexity can vary quite considerably without affecting the
levels of the other two complexities” (p. 73).
For instance, dynamic complexity may result from a substantial increase in
component complexity, but not as much in coordinative complexity. An example of
this would be cooking in a restaurant. Usually between 3 pm and 5 pm, there are not
many diners and the orders are usually less diverse. Most people order only
appetizers and snacks rather than main courses. Cooking is relatively simple. At
dinner time, however, restaurants become a lot more crowded. Most diners order full
meals that include appetizers, main courses, and desserts. Orders become much more
diverse. Cooking turns into a much more complex task. A chef has to be able to
prepare a wider range of dishes. The component complexity of the cooking task
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
increases dramatically whereas the coordinative complexity remains about the same.
The dynamic complexity of cooking changes with the number of customers in a
restaurant and the time of the day.
Effects of Task Complexity on TM
Task complexity affects TM in two ways: (1) direct effects, and (2) indirect
effects through task conflict and relationship conflict.
Direct Effects of Task Complexity on TM
Task complexity is directly related to “the task attributes that increase
information load, diversity, or rate of change” (Campbell, 1988, p. 43). Zigurs and
Buckland (1998) argued that the level of these three factors in information
processing indicated the level of cognitive demand imposed on task performers.
Differential cognitive loads place differential demands on cognitive divisions of
labor, and hence differential incentives for the development of TM.
As component complexity increases, more sophisticated and diverse
knowledge and skills are required for the task (Wood, 1986). No single person
possesses all knowledge and skills to perform every aspect of the task. Weingart
(1992) found that planning had a more significant effect on performance when task
component complexity was high. Managing information is an important facet of the
planning process. With an aim to coping with high cognitive load, there is an
increasing demand for having each group member specialize in a unique knowledge
area, be responsible for gathering information pertinent to that area, and provide
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
51
expert knowledge when needed. Component complexity is likely to contribute to the
formation and development of group TM.
As coordinative complexity increases, group members need more
coordination of one another’s acts (timing, frequency, sequence, etc.) (Wood, 1986).
In the coordination process, group members acquire more knowledge about who has
expertise in which area and update their expertise directories accordingly. Accurate
directory information eases expertise coordination through allocation and retrieval
processes. TM thus evolves and matures.
Performance of a dynamically complex task requires knowledge about
changes in component and/or coordinative complexities over time (Wood, 1986).
Such knowledge is often acquired through intense communication. As dynamic
complexity increases, group members have to constantly coordinate individual acts
in order to adapt to the changes in the task environment. Their sensitive responses to
the task environment will provide the group a number of opportunities to correct
errors in expertise directories and identify the best person for allocation and retrieval
of domain-specific knowledge.
Hypothesis 1: Task complexity is positively related to TM.
Indirect Effects of Task Complexity on TM Through Conflict
Task Complexity and Task/Relationship Conflict. Conflict is defined as
“perceptions by the parties involved that they hold discrepant views” (Jehn, 1995).
Groups often experience two major types of conflict: task-related “substantive”
conflict (task conflict) and non-task related “affective” conflict (relationship conflict)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
52
(Guetzkow and Gyr, 1954, p. 370). Task conflict arises when group members fail to
achieve consensus on task-related matters due to different opinions on the content of
the task, while relationship conflict emerges from interpersonal incompatibility and
often arouses negative feelings toward one another in a group (Jehn, 1995).
Complexity produces conflict through tension and confusion (Wall and
Callister, 1995). Tasks with high component complexity bring people with diverse
knowledge backgrounds to work together in a group. There are increases in
informational diversity (differences in knowledge backgrounds), social category
diversity (differences in social category membership), and value diversity
(differences in conceptions of group’s real task, goal, target, or mission) in such
groups (Jehn, Northcraft, and Neale, 1999). Jehn et al. (1999) found that
informational diversity led to an increase in task conflict, that social category
diversity increased relationship conflict, while value diversity was positively
associated with both task conflict and relationship conflict (see also Jehn, Chadwick,
and Thatcher, 1997; Pelled, Eisenhardt, and Xin, 1999).
Higher coordinative complexity makes it more difficult to coordinate
individual acts. Greater coordination difficulty in the form of either capacity
problems or capability problems or both (Kelly, Futoran, and McGrath, 1990) leads
to different opinions about task-related issues. Task conflict arises. Some group
members may take these arguments personally and harbor hostility toward one
another (Amason, 1996). Relationship conflict arises. In addition, complex task
environments are often less certain and more ambiguous (Volkema, 1988). This
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
53
makes it difficult, if not impossible, to render a stable and thorough specification of
job responsibilities. Sometimes the responsibilities assigned are overlapping. At
other times, nobody takes on the responsibility that emerges from a new situation.
Confusions about one another’s responsibilities may create misunderstandings that
will lead to relationship conflict as well. Indeed, the task aspect and the relationship
aspect of conflict often go hand in hand (Van De Viliert and De Dreu, 1994).
As dynamic complexity increases, group members must frequently adapt to
changes in the “cause-effect chain or means-ends hierarchy” (Wood, 1986, p. 71).
During this ongoing adaptation process, group members may respond to changes
differently. The discrepancies in their responses require individuals adjust their
approach to the task and their respective responsibility for various aspects of the task
in a timely manner. Disagreements arising from this adjustment process may cause
both task conflict and relationship conflict.
Jehn and Mannix (2001) examined dynamic changes in the levels of task
conflict and relationship conflict in high-performing groups and low-performing
groups. They found that both types of conflict escalated over time in low-performing
groups. This finding implies that task conflict and relationship conflict may reinforce
each other.
Hypothesis 2a: Task complexity is positively related to task conflict.
Hypothesis 2b: Task complexity is positively related to relationship
conflict.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
Hypothesis 2c: Task conflict is positively related to relationship
conflict.
Effects of Task/Relationship Conflict on TM. Task conflict may be positively
related to the three transactive processes: directory updating, information allocation,
and retrieval coordination. Task conflict may facilitate extensive discussion (Jehn,
1995). Through in-depth discussion, group members are better at discerning one
another’s interest and expertise and updating their directories with accurate expertise
location information. More task-related discussion encourages self-reflection and
leads to more accurate assessments of one’s own strengths. Thus, the person with
real expertise is more likely to make the right claim about his or her own specialty.
With accurate and up-to-date expertise location information, group members are
more likely to allocate information to the real expert. Task conflict is positively
related to directory updating and information allocation.
Extensive discussion produced by task conflict helps to clear up
misunderstanding or ambiguity and foster accurate expertise recognition, a process
critical to retrieval coordination. Research shows that group’s ability to pool
unshared information depends on the degree to which information processing is
coordinated based on members’ mutual recognition of expertise (Stasser, Stewart,
and Wittenbaum, 1995; Stasser, Taylor, and Hanna, 1989). Detailed discussion
brought by task conflict also promotes critical evaluation of the information
contributed by experts and increase the likelihood that errors are detected and
corrected by others in the group (Turner and Pratkanis, 1997). Wittenbaum and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
Stasser (1995) also found that compared with the information shared by an
unrecognized expert, information contributed by a recognized expert was deemed
more accurate and credible and was more receptive to others in the group. Task
conflict is positively related to retrieval coordination. Moreover, task conflict also
produces a sense of engagement (Amason, 1996; Korsgaard, Schweiger, and
Sapienza, 1995) that encourages group members’ participation in all three stages of
transactive information processing.
Relationship conflict, on the other hand, may negatively impact TM. A
person’s perception of others’ expertise can be acquired through their self-disclosure
of personalities, experiences, opinions, and attitudes, many of which help a person
infer others’ interests and expertise (Wegner, Erber, and Raymond, 1991).
Relationship conflict reduces the amount of communication among group members
(Jehn, 1995) and makes it increasingly difficult to access such cues. Relationship
conflict impedes directory updating.
A number of studies have suggested that relationship conflict weakens task
performers’ cognitive processing capability (e.g., Kelly, 1979; Pelled, 1996;
Roseman, Wiest, and Swartz, 1994). A competent expert afflicted by relationship
conflict, thus, may not be able to provide expert knowledge in an effective fashion.
High level of relationship conflict also induces “contentious” communication that
leads to less critical evaluation of the information contributed by group members
(Lovelace, Shapiro, and Weingart, 2001) and lowers the quality of retrieval.
Furthermore, hostility incurred by relationship conflict makes group members
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
56
resistant to the information contributed by those whom they don’t like or who don’t
like them (Pelled, 1996). They are also less willing to pass new information to the
experts with whom they don’t get along to avoid communication. Relationship
conflict may obstruct information allocation and retrieval coordination.
Relationship conflict may also impact TM through turnover. Jehn (1995)
found a negative correlation between relationship conflict and members’ intent to
remain in the group. Changes of group composition may disrupt group TM. For
instance, if experts leave the group, they take away their expertise as well. It is often
difficult to replace them right away. The absence of requisite expert knowledge will
break down TM. Turnover may paralyze TM if expert knowledge stored in
individual memories has not been shared by all group members (Moreland, 1999;
Moreland and Argote, 2003; Rao and Argote, 1992, cited in Aldrich, 1999).
Hypothesis 3a: Task conflict is positively related to TM.
Hypothesis 3b: Relationship conflict is negatively related to TM.
TM and Group Performance
TM contributes to group performance through two mechanisms: enhancing
group’s information processing capacity (Rau, 2000) and improving coordination
(Liang et al, 1995). Expertise development is critical to the success of TM. Wegner
(1987) described an experiment conducted by Giuliano and Wegner among intimate
couples and found that experts were capable of processing a larger amount of
information than non-experts. Stasser et al. (1995) found similar results in groups.
They argued that expertise recognition and transactive retrieval, the two processes
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
enhancing group’s information processing capacity, were key reasons why TM could
boost group performance.
Liang et al. (1995) reported that TM also facilitated group performance by
improving coordination. If group members know who is good at what, they are able
to anticipate one another’s behavior and react to that behavior more quickly and
accurately. Janz, Coquoitt, and Noe (1997) found that efficient information
transmission led to more positive group process behaviors such as helping, sharing,
and innovating. Better coordination and cooperation will enhance group
performance.
Task complexity is likely to affect the strength of the association between
TM and group performance. In simple tasks, the amount of knowledge required for
performing the task is low. Group members tend to have homogeneous knowledge
backgrounds. There is a low demand for transactive information processing. TM
won’t have much effect on group’s performance outcome. In complex tasks,
however, task performers tend to have heterogeneous knowledge backgrounds.
Performing complex tasks effectively often requires smooth coordination of
individual expertise. Having a well-developed TM in place is essential to the
performance of complex tasks.
Also, as mentioned earlier, task conflict and relationship conflict increase
with an increase in task complexity. High relationship conflict in complex tasks leads
to high turnover (Jehn, 1995) and the disruption of TM. This is because groups
performing simple tasks are often consisted of generalists whereas groups
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
performing complex tasks are composed of specialists or mixed with specialists and
generalists (Rulke and Galaskiewicz, 2000). In simple task conditions, if a group
loses an expert, it is relatively easy to find a replacement. In complex task
conditions, it is much harder to replace an expert because fewer individuals have the
type and/or level of knowledge and skills required for the task as well as the
knowledge of the linkages across persons that are required for task completion. An
effective TM is more important to the successful completion of complex tasks than
simple tasks.
Hypothesis 4a: TM is positively related to group performance.
Hypothesis 4b: The positive relationship between TM and group
performance is stronger in complex tasks than in simple tasks.
All hypotheses are presented in Figure 3.1.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Figure 3.1 Study II: Model of Effects of Task Complexity and Conflict on TM and on TM - Group Performance Interaction
+
Task
Complexity
Task
Conflict
Transactive
Memory
Relationship
Conflict
Group
Performance
Moderator
1 The magnitude of the effect of TM on group performance is predicted to be stronger in complex tasks than in simple tasks.
Method
Participants
Four hundred and eighty undergraduates enrolled in a large introductory class
in a business school on the west coast participated in the survey. The participants
represented 85% of the total enrollment in the class. They worked in groups of 3-5
on a small-scale group exercise as well as a large-scale term project. Among all the
participants, 56% were male and 44% were female; 11% were between 18 and 19
years old, 65% between 20 and 21, 17% were between 22 and 23 years old, and 7%
were over 23 years old. More than 80% were business majors. A substantial
proportion of the participants (82%) had experience working on more than 4 group
projects prior to taking this class.
Procedures
The approval for research was obtained from the university IRB before the
survey was conducted. The researcher passed information sheets and questionnaires
to lab instructors and had them administer the survey at the end of lab sessions.
Potential participants were asked to read information sheets first to decide whether
they would like to participate. Those who decided to do the survey remained in the
classroom and proceeded to the survey. There were two versions of the survey.
Version A asked participants to respond to the items based on their experience in the
small-scale, simple group project. Version B asked about their experience in the
large-scale, complex term project. The two versions of the survey were randomly
distributed. Lab instructors had to make sure approximately half of the participants
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
61
were given Version A and the other half got Version B. Lab instructors also
randomly assigned a numerical number to each group and asked each participant to
write down this number in the space allocated in the survey. This procedure served
two purposes. First, it allowed the researcher to aggregate individual responses into
group data. Second, since members belonging to the same group got the same grade
for group projects, this procedure enabled the researcher to match grades (a
performance measure) with appropriate individuals without knowing their names.
The participants’ anonymity was thus protected. It took about 8-10 minutes to
complete the survey.
Instrumentation
The questionnaire included demographic items and items measuring five
constructs: task complexity, task conflict, relationship conflict, TM, and group
performance. Participants were asked to indicate on a five-point Likert scale (ranging
from “strongly agree” to “strongly disagree”) their responses to the items concerning
each of the five constructs.
Task Complexity Ten items were used to measure the three dimensions of
task complexity. Two items measuring component complexity and two items tapping
dynamic complexity were written according to Wood’s (1986) definition.
Coordinative complexity is very similar to task analyzability discussed by Daft and
Macintosh’s (1981). Therefore, coordinative complexity was measured by six items,
five of which were modified based on Daft and Macintosh’s (1981) task
analyzability scale and one of which (item 6, see Appendix B) was written by the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
62
researcher. A composite computed from these three dimensions was expected to
adequately capture the nature of task complexity.
Task Conflict Six items were used to measure task conflict. These items were
pooled from scales developed by Janssen, Van De Vliert, and Veenstra (1999) and
Jehn (1995), and tapped different opinions regarding the content of a task (Jehn,
1995). These items were chosen because of their empirically proved high reliability
(Cronbach’s Alpha > .70).
Relationship Conflict Six items measured relationship conflict. These items
came from scales written by Janssen et al. (1999) and Jehn (1995) to evaluate
conflicts resulting from interpersonal incompatibility that led to such psychological
discomforts as tension, annoyance, and hostility (Jehn, 1995). Similar to task conflict
scale, relationship conflict scale enjoyed high reliability as well in previous studies.
TM Hollingshead (2001) wrote nine items measuring various TM processes
based on her study on cognitive interdependence and TM development. These nine
items tapped three states of TM development: specialization, expertise recognition,
and cognitive interdependence. These three states correspond with Wegner’s (1995)
three transactive processes. This study adopted Hollingshead’s (2001) TM scale.
There are other TM scales available as well. For instance, Lewis (2003) developed
and validated a field measure on TM in work groups. Lewis’s (2003) scale consisted
of fifteen items tapping three dimensions: specialization, coordination, and
credibility. However, some researchers may argue that credibility is an antecedent
rather than an integral part of TM. Since the TM construct in Study II aimed to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
63
measure TM itself, not its antecedents, Lewis’s scale was not suitable for the current
study. Moynihan and Bratt (2000) also developed a TM scale, but its Cronbach’s
Alpha was below .70, a widely accepted threshold in social science research. A
pretest of Hollingshead’s (2001) scale, on the other hand, yielded a much higher
reliability, Cronbach’s Alpha = .78.
Group Performance Hackman (1987) constructed three items tapping three
dimensions of group performance: productivity, members’ satisfaction with their
group experience, and their ability to work together on future projects. In this study,
participants’ grades on group projects served as a productivity measure. The other
two dimensions of Hackman’s (1987) measures were assessed with self-report data.
All measures were presented in Appendix B.
Analysis
LISREL VIII was the statistical package employed to analyze the data. Like
regression, the structural equations modeling technique assumes that all data are
normally distributed and all observations are independent of one another (Hu and
Bentler, 1995). The Kolmogorov-Smimov test performed on each of the five
theoretical variables suggested that all data points were normally distributed. In this
study, the responses provided by the members in the same group were not
independent given their shared group experience. With an aim to avoiding the
violation of the independent observation assumption in LISREL VIII, 480 cases in
the raw data were aggregated into group data and produced 154 cases (i.e., groups)
that had more than one member from each group complete the survey. The
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
64
satisfaction of two criteria provided justification for this aggregation (Edmondson,
1999; Kenny and LaVoie, 1985). First, the constructs of interest were meaningful as
group constructs from a theoretical standpoint. For instance, TM is a group construct
because it is only meaningful on the collective level. Second, the intraclass
correlations (ICC) for group variables were greater than zero, a criterion suggesting a
convergence among the responses by the members within a group and a systematic
difference among the members between groups. One-way analysis of variance was
also performed on each variable to determine whether the differences between
groups were significantly greater than the differences within groups (Lovelace,
Shapiro, and Weingart, 2001) (see Table 3.1 for the results of ICCs and analysis of
variance). The results suggested that the individual responses within a group were
indeed interdependent and should be aggregated into one data point, the mean.
Factor analyses of the measurement models and the assessment of the theoretical
model were performed on the group data (N = 154).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 3.1 Study II: Analysis of Variance and Intraclass Correlation Coefficients for Group-Level Scales
F(153, 480)
P
ICC
Task Complexity 2.53 <.001 .26
Task Conflict 3.61 <.001 .31
Relationship Conflict 2.91 <.001 .27
Transactive Memory 3.28 <001 .28
Group Performance 2.83 <001 .27
ON
e /i
66
Bollen (1989) and Joreskog and Sorbom (1996) suggested several rules be
used to assess model fit and explained reasons for using each criterion: (1) minimum
fit function chi-square statistic divided by degrees of freedom. A model is a good fit
if this value is smaller than 5; (2) p value. An insignificant p value (i.e., >.05)
suggests that the hypothetical model adequately captures the real relationships
between variables; (3) goodness-of-fit index (GFI) and adjusted goodness-of-fit-
index (AGFI). AGFI is GFI adjusted for degrees of freedom and is usually lower
than GFI. AGFI is preferred over GFI because it is insensitive to sample size. AGFI
and GFI of a well-specified model are generally greater than .90; (4) root mean
squared residual (RMR). RMR is an index of residual average. A good model would
show a RMR less than the .05 ceiling. RMR is especially favorable in a situation
with standardized indicators; (5) comparative fit index (CFI) and incremental fit
index (IFI). CFI and IFI represent the difference between the hypothesized model
and the null model. If both CFI and IFI are greater than .90, the hypothesized model
is a good fit; (6) coefficient of determination. This index explains the amount of
variances accounted for in each dependent variable. Since one criterion is not
adequate for reaching a definite conclusion about the general fitness of a model,
multiple criteria were used to interpret the results of the measurement models and the
theoretical model reported below in Study II. When assessing the significance of
each structural coefficient, t-values were reported.
Independent samples T-Test was conducted on the means of task complexity
in the two versions of the survey to uncover the degree to which two tasks differ in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
67
their complexity. The results produced by this procedure showed that there was a
significant difference in the complexity of the tasks measured in the two versions of
the survey, t = 2.58, df = 152, p = .011 (two-tailed).
Results
Table 3.2 summarizes the means, standard deviations, and correlations
among all variables.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 3.2 Study II: Descriptive Statistics and Correlations (**p < .001; *p < .05; N = 154)
Variable Mean s.d. Task
Conflict
Relationship
Conflict
Transactive
Memory
Group
Performance
Task
Complexity
Task
Conflict
3.30 .69
-
Relationship
Conflict
4.08 .93 .256*
-
Transactive
Memory
2.62 .42 .030
. 234**
-
Group
Performance
2.34 .92 -.003 -.362** .480**
-
Task
Complexity
3.17 .49 .330** -.025 .338** .109
-
O n
oo
69
Analyzing Measurement Models
Factor analyses of the measurement models were conducted before testing
the theoretical model. Only reliable measures (i.e., measures that had significant t
values) were included in the assessment of the theoretical model.
Task complexity was analyzed as a three-dimensional construct. The three-
factor orthogonal solution generated a chi-square of 112.5 with 32 degrees of
freedom, p = .01, GFI = .87, AGFI = .78, CFI = .79, IFI = .79, RMR - .13. The
coefficients of determination ranged from .04 to .81. The chi-square divided by
degrees of freedom was much lower than 5, suggesting that the model was a good fit.
The index of GFI was also acceptable. However, other goodness-of-fit indices were
less satisfying. Since all t-ratios were still significant, all items were kept for
hypothesis testing.
The task conflict scale was evaluated as a single-factor scale, chi-square =
38.15, dfs = 9, p = .01, GFI = .92, AGFI = .82, CFI = .88, IFI = .88, RMR = .097.
The coefficients of determination were from .01 to .78. Chi-square divided by
degrees of freedom was well below 5, suggesting a good fit of the model. Goodness-
of-fit indices of GFI, CFI, and IFI also suggested a reasonable fit of the measurement
model although the results of AGFI and RMR were less satisfying. The t-values for
all links between the latent variable and the indicators were significant except one.
This item was excluded from the assessment of the theoretical model.
The relationship conflict submodel was analyzed as a single-factor submodel.
T-ratios suggested that all six indicators were significantly loaded on the latent
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
70
factor. The minimum fit function chi-square was 36.10 with 9 degrees of freedom.
The chi-square divided by degrees of freedom was below 5. Most goodness-of-fit
indices (e.g., GFI = .93, CFI = .96, IFI = .96, RMR = .042) suggested the overall
fitness of the model was acceptable except that the p value was lower than .05 (p =
.01) and AGFI did not exceed .90 (AGFI = .85). The coefficients of determination of
the indicators were from .39 to .85. Taken together, all six items were retained for
hypothesis testing.
The TM scale was treated as a three-factor scale and factor analyzed with the
orthogonal solution, chi-square = 198.65, dfs = 78, p = .006, GFI = .88, AGFI = .87,
CFI = .89, IFI = .88, RMR = .078. The overall coefficients of determination ranged
from .15 to .82. Chi-square divided by degrees of freedom was much lower than 5.
Although other goodness-of-fit indices were less satisfying, none of the links
between the indicators and latent variables showed non-significant t ratios.
Therefore, all measures were included in evaluating the theoretical model.
Group performance was measured with three items. This was a saturated (i.e.,
just-identified) model and could not be evaluated by confirmatory factor analysis in
LISREL. This problem couldn’t be resolved by using a reference indicator.
Therefore, Cronbach’s Alpha was computed to assess the reliability of this scale
(Cronbach’s Alpha = .81).
Analyzing the Theoretical Model
LISREL VIII was performed on the group data set (N = 154). The minimum
fit function chi-square was 14.49 with 3 degrees of freedom, p = .002. The chi-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
square divided by degrees of freedom was lower than 5, which suggested a good fit
of the model. Some goodness-of-fit indices also suggested that the model adequately
captured the real relationships among variables, GFI = .96, CFI = .90, IFI = .91,
although other indices were less satisfying, AGFI = .84, RMR = .066. The overall
coefficient of determination was .17. The LISREL results provided full support for
HI, H2a, H2c, H3b, and H4a. H2b and H3a were not supported. In other words, task
complexity was found positively related to TM and task conflict, but not relationship
conflict; task conflict positively related to relationship conflict, but not TM;
relationship conflict negatively related to TM; and TM positively related to group
performance. Two regression analyses were performed to determine different
strengths of the relationships between TM and group performance in simple tasks
and complex tasks. The results showed that in both conditions, TM was positively
related to group performance and that this relationship was stronger in complex tasks
(J3 = .545, p = .000) than in simple tasks (J3 = .335, p = .001). One-way ANOVA was
performed using group performance as the dependent variable and grouping by
simple tasks and complex tasks as factor, F (1, 153) = .239, p = .583. This result
suggested that although TM was more strongly related to group performance in
complex tasks than in simple tasks, the difference was not significant. Figure 3.2
presented the LISREL results of the theoretical model.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Figure 3.2 Study II: Results of Original Theoretical Model (*p < .05, N = 154)
.34*
Task
Complexity
.33* -.03
.30* .48*
- .22* -.12
Transactive
Memory
Task
Conflict
Group
Performance
Relationship
Conflict
- j
ISJ
73
Model Modification
Modification indices suggested ways to improve the theoretical model.
Revisions were subsequently conducted in two stages. In the first stage, links with
non-significant t-values were deleted. That is, the links between task conflict and TM
and between task complexity and relationship conflict were not estimated. These
links were deleted one at a time until both non-significant links were deleted, a
procedure suggested by Kaplan and Wenger (1993) to better capture the dynamic
changes in the model. The revised model was better than the original one, chi-square
= 16.02, degrees of freedom = 5, p < .007, GFI = .96, AGFI = .88, CFI = .91, IFI =
.91, RMR = .077. The overall coefficient of determination was .19.
Modification indices also suggested a significant causal relationship between
relationship conflict and group performance. De Dreu and Van Vianen (2001)
reviewed the current literature on conflict and found a negative correlation between
relationship conflict and group performance (r = -.48, p < .001). Conflict researchers
have suggested a number of reasons for this negative association. First, relationship
conflict often induces destructive conflict management which leads to even more
relationship conflict (De Dreu, 1997). Second, negative emotions such as anxiety
that accompany relationship conflict impair task performer’s cognitive processing
capacity and thus, impede group performance (Pelled, 1996). Third, relationship
conflict diverts task performer’s attention to non-task-related matters (Pelled, 1996).
Since much of the earlier research provided ample empirical evidence for the
existence of a negative association between relationship conflict and group
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
performance, this link was added in the second stage of model revision. The results
demonstrated a significant improvement in model fit with minimum fit function chi-
square = 3.75, degrees of freedom - 4, g = .441, GFI = .99, AGFI = .96, CFI = .99,
IFI = .99, RMR = .034. The overall coefficient of determination was .23. Table 3.3
summarizes the various goodness-of-fit indices of the theoretical model, two stages
of modification, and the null model. Figure 3.3 presented the modified theoretical
model and the structural coefficient of each link.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Table 3.3 Study II: Summary of Model Revisions
Revisions
(MI)
Chi-
square
Degrees
o f
freedom
P
value
GFI AGFI CFI IFI RMR CED
Hypothesized
Model
14.49 3 .002 .96 .82 .90 .91 .066 .17
Stage 1
Delete two non
significant links
16.02 5 .007 .96 .88 .91 .91 .077 .18
Stage 2
Add one link
3.75 4 .441 .99 .96 .99 .99 .034 .23
Null Model 113.25 10 .000 .78 .67 .01 .01 .220 .00
Note: MI = Modification index; GFI = Goodness-of-fit index; AGFI = Adjusted goodness-of-fit index; CFI = Comparative fit
index; IFI = Incremental fit index; RMR = Root mean square residual; CED = Overall coefficient of determination.
- j
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Figure 3.3 Study II: Results of Modified Theoretical Model (*p < .05, N = 154)
.33*
yr
.23* .42* .33* .26*
A V
Task
Conflict
Task
Complexity
Relationship
Conflict
Group
Performance
Transactive
Memory
-.26*
- j
a \
77
Discussion
Interpreting Results of The Theoretical Model and Their Implications
Study II was designed to examine the role of task complexity and conflict in
TM development and in TM - group performance interaction. Significant
relationships were detected between task complexity and TM, between task
complexity and task conflict, between task conflict and relationship conflict, between
relationship conflict and TM, and between TM and group performance. TM was
slightly more strongly related to group performance in complex tasks than in simple
tasks. Task complexity was found unrelated to relationship conflict and task conflict
unrelated to TM.
It was not surprising to find that task complexity was positively related to
TM. Earlier research suggested that group training (Liang et al., 1995; Moreland and
Myaskovsky, 2000), performance evaluation, and knowledge of one another’s
expertise (Littlepage, Robinson, and Reddington, 1997) foster TM in groups. The
present study suggested yet another contributing factor: task complexity. In complex
task conditions, people are more likely to depend on one another’s expertise to cope
with the high cognitive demand. To perform a complex task more effectively, they
have to divide cognitive labor based on individual interests and strengths. An
increase in task complexity increases the amount of interdependence and interaction
among group members, which nurtures the development of TM.
Task complexity was also found positively related to task conflict but not
relationship conflict. Task conflict was positively related to relationship conflict.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
78
Taken together, these findings suggested that, instead of a direct correlation, task
complexity was related to relationship conflict indirectly through task conflict. When
tasks get more complex, individuals with diverse knowledge backgrounds and
perspectives often work together to meet the high knowledge demand of the tasks.
The differences in knowledge backgrounds often lead to differences in priorities and
goals, both of which create task conflicts (Jehn et al., 1999). In complex task
conditions, people may feel ambiguous about the course of the task. The increased
ambiguity in task-related matters also increases task conflict. In dynamically
complex tasks, group members may respond to changes in different manners and/or
at different paces. Task conflicts arise. Task conflict may lead to relationship conflict
if group members take task-related disagreements personally. Task complexity was
not directly related to relationship conflict in Study II. On the contrary, the structural
coefficient between task complexity and relationship conflict was negative (-.12)
although not significant. This result suggested that task complexity might direct
people’s attention to task-related matters so that they overlook relationship issues.
Inconsistent with the prediction, task conflict was not significantly related to
TM although the structural coefficient between the two variables was positive.
Relationship conflict was negatively related to TM. This finding provided support
for a relational view on knowledge sharing (Borgatti and Cross, 2003).
Overwhelming relationship conflict preoccupies groups with interpersonal frictions
rather than task-related matters. Even if group members know who knows what, they
may not turn to the expert for the information because of the negative feelings they
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
hold against the expert. According to the theory of homophily, communication is
more frequent among people who are similar to one another (Ibarra, 1992; Westphal
and Milton, 2000). A well-established finding in the demography literature is that
people with similar background such as education and socio-economic status are
more likely to get along (Williams and O’Reilly, III, 1998). Thus, communication
may be more frequent among people who have favorable feelings toward one
another. Interpersonal hostility may reduce communication and obstruct expertise
coordination in TM.
Moreover, relationship conflict is positively associated with turnover (Jehn,
1995). Turnover changes the knowledge composition of a group and breaks down
group TM (Moreland, 1999; Moreland, Argote, and Krishnan, 1996). With an aim to
minimizing the disruptive effect of turnover on TM, managers should develop and
implement effective intervention strategies to build interpersonal relationships
among group members in addition to getting the work done. This is especially
important in the initial stage of group formation because the negative feelings group
members have developed in the initial stage may be carried over and enhanced over
time. This recommendation is also more important to groups with longer longevity.
Bradley, White, and Mennecke (2003) found that interpersonal interventions had a
stronger positive effect on performance in groups with longer history. This
managerial implication is also important given the detrimental effects relationship
conflict has on group performance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
80
Corroborating many prior research findings on TM (e.g., Liang et al., 1995;
Moreland and Myaskovsky, 2000), Study II provided further support for the
performance benefits of TM. Moreover, Study II suggested that TM was strongly
related to group performance in both complex tasks and simple tasks and this link
was even stronger in complex tasks than in simple tasks. This finding has important
practical implications. Even if groups perform simple tasks, it is still important to
construct an effective TM system to ensure good performance. In groups frequently
facing complex tasks, such as R&D groups and consulting groups, a well-developed
TM will facilitate group performance even more. Thus, it is even more important for
these groups to build effective TMs. They should create communication structures
that allow group members easy access to expertise directory information and expert
knowledge.
Some people may argue that there exists a negative relationship between task
complexity and group performance. Their logic is that group members tend to
perform simple tasks better than complex tasks. Although this hypothesis is
intuitively appealing, a closer look at the model reported in Study II makes it
problematic. Figure 8 illustrates two intervening processes involved in the
relationship between task complexity and group performance. In Process 1, task
complexity is positively related to TM and TM is positively related to group
performance. This result suggests a positive relationship between task complexity
and group performance if we incorporate TM as an intervening variable. In Process
2, task complexity is positively related to task conflict; task conflict is positively
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
81
related to relationship conflict; and relationship conflict is negatively related to group
performance. This result suggests that there is a negative relationship between task
complexity and group performance if we consider conflict as an intervening variable.
The two intervening processes lead to different predictions regarding the relationship
between task complexity and group performance. Therefore, the nature of the
association between task complexity and group performance is not uniformly
negative, but rather is mediated by TM and/or conflict.
Limitations and Direction for Future Research
The major limitation of the study lies in its sample. The participants of this
study were undergraduates enrolled in a large introductory business class. Student
work groups differ from groups in real work settings along several dimensions. First,
members in real work groups tend to interact with one another on a more regular
basis than those in student work groups. Members in student work groups often get
together to work on the project a couple of weeks before it is due rather than holding
regular meetings throughout the semester. This type of work pattern limits the
amount of interaction among students and affects such group processes as conflict
and TM development. Second, the issue of power is usually more salient in real work
groups than in student work groups. Differences in power dynamics may have an
impact on the amount of task conflict and relationship conflict present in a group and
how these two types of conflicts influence TM. For instance, task conflict was not
found to facilitate TM development in student work groups. This result may be due
to the lack of variation in task conflict across groups and may not generalize to real
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
work groups. Given the limitations of the student sample, it is highly desirable to
replicate these findings in real work groups.
Research on learning and knowledge management has gained an increasing
momentum in the information age. TM, a new perspective on management of
group’s cognitive assets, has much potential in this vibrantly growing area of
inquiry. For instance, previous empirical research on group and organizational
learning is far more behavior-centered than cognition-centered (Argote, 1999).
Cognitive learning in groups, however, is just as important as behavioral learning
because changes in behavior and changes in cognition often reinforce each other.
Fundamental shifts in behavioral patterns go hand in hand with cognitive shifts. TM
theory sheds light on a cognitive approach to group learning.
In addition, Study II examined the effect of task complexity on TM and
enriched our understanding about information sharing in general. Specifically, the
results suggested that task complexity affected information sharing patterns in work
groups and that the contributing effect of information sharing on performance was
statistically significant in both complex tasks and simple tasks and was slightly
stronger in complex tasks. The nature of a task often determines how much
information needs to be shared among group members and how important
information sharing is to performance. Task conflict was not related to information
sharing whereas relationship conflict was negatively associated with information
sharing. This finding suggests that relationship building is essential to effective
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
83
information sharing in groups whereas disagreements on tasks do not necessarily
impede information sharing unless such disagreements lead to interpersonal frictions.
Furthermore, there emerge more and more knowledge-intensive tasks with
the development of technology. TM is especially important to the collective
performance of groups performing knowledge-intensive tasks. Various variables
related to TM need to be examined systematically to produce more useful findings
that are applicable to complex field settings. Such variables include task complexity,
conflict, and performance investigated in Study II, as well as network structure,
personnel instability, group size, and so on so forth. Research incorporating key
variables influencing transactive information processing will enrich the theory and
provide more fine-grained analyses targeted toward specific situations. The model
developed in Study II extended the original TM theory and represents an initial effort
toward this end. More sophisticated models based on the original TM theory are to
be built and validated with empirical data.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
84
CHAPTER FOUR
STUDY III: SHARING UNSHARED KNOWLEDGE IN WORK
GROUPS: A CHALLENGE POSED BY DIFFERENTIAL
KNOWLEDGE CRITICALITY JUDGMENT
Study II examined how one exogenous variable - task complexity - affected
information sharing processes. Study III focused on another exogenous variable -
group knowledge composition. Specifically, Study III investigated how group
knowledge composition affected members’ knowledge criticality judgment. Each of
the three TM processes - directory updating, information allocation, and retrieval
coordination - requires sharing of information or knowledge. How group members
decide what knowledge is critical, and therefore share with others, has a direct
impact on the effective functioning of TM systems. Study III argued that one of the
difficulties group members would experience in sharing unshared knowledge would
be the difficulty posed by differential knowledge criticality judgment. Knowledge
criticality judgment here refers to the judgment of the extent to which given
knowledge is critical to the task at hand.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
85
There is a coherent body of literature in psychology that deals with sharing
shared (or common) versus unshared (or unique) knowledge in group discussion. A
well-established finding is that people are more likely to share common knowledge
than unique knowledge - that is, they tend to discuss what everyone already knows
rather than pooling unique knowledge known by one person (Gigone and Hastie,
1993; Stasser, Taylor, and Hanna, 1989). The extant literature provides some
explanations for why group members share shared knowledge more often than
unshared knowledge. For instance, Stasser (1992) offers a probability explanation.
He argues that shared knowledge is known by a greater number of individuals in a
group and thus, is more likely to be mentioned in group discussion. However, the
probability account does not apply to situations where group members are aware of
who knows what. The insufficiency of the probability account suggests that there
may be other processes that lead to groups’ tendency to share shared rather than
unshared knowledge. Wittenbaum and Park (2001) present a mutual enhancement
explanation to account for the social and psychological aspects of this phenomenon.
They argue that sharing shared knowledge enhances one another’s confidence in job-
related expertise as well as psychological comfort. This paper proposes an awareness
explanation that hasn’t been explored so far - that is, knowledge holders don’t share
unshared knowledge because they are not aware that knowledge is critical to task
performance.
The extant research on knowledge management has been mute on the
underlying mechanisms that cause the members in work groups to disagree on what
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
knowledge they consider critical to the performance of a focal task. The concept of
absorptive capacity highlights the importance of prior knowledge in the process of
assimilating new knowledge (Cohen and Levinthal, 1990), but it does not explain
why and how people differ in their perceptions of the criticality of new knowledge
before they decide to integrate that knowledge. This question is an important one
because people have to appreciate the value and the criticality of the knowledge
before they are motivated to learn (i.e., “absorb”) it. The performance benefit of
critical knowledge is well documented in earlier research. The decision making
literature suggests that the more critical knowledge is used in the decision making
process, the better decision people will make (see e.g., Blanchard, 1966; Porat and
Haas, 1969). O’Reilly (1980) cited Troutman and Shanteau’s (1977) work on judge’s
decision making to provide compelling empirical evidence that irrelevant knowledge
often misled judges to erroneous decision.
Study III addresses the knowledge criticality question. It argues that
knowledge criticality is not an objective construct and that people’s prior task
knowledge which consists of two dimensions: knowledge domain and expertise
level, as well as social knowledge, have an effect on what knowledge they perceive
critical to the focal task and thus, needs to be retrieved. Perceived task uncertainty is
likely to aggravate this effect. The roles of communication strength and diversity are
also discussed as well as the effect of disagreement on knowledge criticality
judgment on group performance. The arguments are made by drawing literature in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
87
multiple disciplines such as cognitive psychology, cybernetics, communication, and
management.
Task Knowledge and Knowledge Criticality Judgment
The proposition that different task knowledge leads to different judgment of
what knowledge is critical and valuable to the focal task is rooted in Bruner’s (1957)
analysis of “perceptual readiness,” a concept referring to “the relative accessibility of
categories to afferent stimulus inputs” (p. 148). Category accessibility differs among
individuals with diverse task knowledge. Task knowledge diversity can be classified
into two categories: domain diversity and expertise level diversity. Domain diversity
refers to the degree to which the members in a group possess knowledge in different
domains. The differential accessibility of various categories among individuals with
specialized knowledge in different domains may prime certain stimuli over others
and lead to different perceptions by different people. Expertise level diversity refers
to the extent to which expertise levels of knowledge holders vary in the same
domain. The prior knowledge of experts and non-experts varies in both content and
structure, which subsequently influences their judgment of knowledge criticality.
Diverse knowledge and skills among task performers are not as noticeable as diverse
“readily detectable ” attributes such as gender or race, yet yield substantial
performance consequence nonetheless (Sessa and Jackson, 1995).
Domain Diversity and Knowledge Criticality Judgment. Domain diversity
originates from diversity in functions, occupations, educational backgrounds, and so
on. Knowledge diversity has been reported to contribute to group performance
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
88
(Pelled, Eisenhardt, and Xin, 1999) through increased communication with experts
both within and outside the group (Ancona and Caldwell, 1992). This is especially
true in the area of innovation (Damanpour, 1991). However, knowledge diversity
brings about challenges groups have to deal with as well. Milliken and Martins
(1996) did a comprehensive review of the literature on group diversity and a
thorough analysis on the complex effects diversity has on group functioning. They
argued that subtle differences in knowledge structure and content might cause
substantial coordination problems. For instance, people with different functional
training may develop different beliefs about how various components in group life
are connected to one another. Pelled, Eisenhardt, and Xin (1999) reported that
functional diversity was one of the major sources of task conflicts because people
with different functional backgrounds tended to build different mental models and
had different opinions about work-related issues. Murray (1989) found that
functional diversity contributed to efficiency in a short run, but obstructed adaptation
in a long run.
Moreover, individuals with specialized knowledge in different domains
develop different “habits of thoughts,” as Hayek (1989) called it, that lead to
different evaluations of the criticality of the same knowledge item. For instance,
many economists assume human beings are perfectly rational, and therefore tend to
ignore the role emotion plays in human decision making. The decision models
developed by these scientists purposefully exclude the critical function of the
irrational aspect of human beings. Many decision scientists trained in the areas of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
89
psychology and sociology, on the other hand, regard irrational factors such as
emotion as essential to human decision making because, in their views, no one can
escape the influence of emotion no matter how rational they claim themselves to be.
Medin, Lynch, Coley, and Atran (1997) studied categorization and reasoning
among different types of tree experts such as taxonomists, landscape workers, parks
maintenance personnel and found that they used different reasoning strategies and
placed different weights on the same features. This result may be generalized to
other areas. Thus,
Hypothesis 1: Domain diversity is positively related to group
members’ disagreement on knowledge criticality judgment.
Expertise Level Diversity and Knowledge Criticality Judgment. Besides
domain diversity, the other dimension of task knowledge diversity is expertise level
diversity. In this paper, I use experts versus non-experts to differentiate individuals
with high levels of expertise from those with low levels of expertise. This distinction
enjoys greater validity than that between experts and novices because research has
found that the length of experience is not invariably correlated with the amount of
declarative knowledge, nor knowledge structures, nor performance (Sonnentag,
1998) - the three components of expertise (Ford and Kraiger, 1995; Schvaneveldt, et
al., 1985).
Knowledge criticality judgment may be dependent on both the content and
the structure of one’s prior knowledge. First, differences in the content of the
knowledge system among experts and non-experts may lead to different knowledge
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
90
criticality judgment. Chase and Simon’s (1973) study of chess players suggests that
compared with less skilled chess players, chess masters have their expert knowledge
stored in larger chunks, have more chunks, and all the chunks can be readily
retrieved. Lindsay and Norman’s (1977) study of memory development provides
further empirical support for Chase and Simon’s (1973) findings. They argue that
experts find it easier to retrieve knowledge items from their memories than non
experts because experts spend more time processing these items in the stage of
encoding. Kirschenbaum (1992) also reported that experts processed task-related
knowledge items in more depth than non-experts. The greater amount of knowledge
stored in experts’ memories and its ease for retrieval enhances experts’ ability to
recognize critical knowledge. In addition, the larger knowledge base possessed by
experts allows for easier connections between existing knowledge items in their
knowledge base and new knowledge items (Carley, 1986). This feature also helps
experts identify the criticality of a seemingly irrelevant knowledge item.
Second, differences in the structure of the knowledge system among experts
and non-experts may also lead to different knowledge criticality judgment. Day,
Arthur, and Gettman (2001) reported that expert’s knowledge structure differed from
that of a non-expert. The knowledge items in an expert’s knowledge system are
organized in a more coherent fashion than that in a non-expert’s knowledge system
(Wisniewski, 1995). The systematic organization of an expert’s knowledge system
enables experts to correctly capture the interconnectedness among knowledge items
(Sujan, Sujan, and Bettman, 1988) and integrate critical knowledge items into their
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
91
existing body of knowledge more effectively. Non-experts, however, often place
much weight on spurious associations and ignore the real associations among various
pieces of knowledge (McKeithen, Reitman, Rueter, and Hirtle, 1981). This is
probably another reason why experts are capable of recalling more critical
knowledge while non-experts often get distracted by irrelevant knowledge and delay
the problem solving process (Kirschenbaum, 1992).
The above discussion on knowledge structure has focused on patterns of
associations among knowledge items. Another dimension of knowledge structure is
the strength of the association. Schvaneveldt et al. (1985) reported that the items in
an expert’s knowledge system were connected with substantial key links while the
organization of a non-expert’s knowledge system was fragmented and characterized
by weak spurious links. People’s decision on what knowledge is critical is contingent
on the nature and the strength of associations between knowledge pieces
(Krippendorff, 1975). The lack of strong critical links and the existence of weak
inaccurate linkages in non-experts’ knowledge systems potentially impair their
knowledge criticality judgment.
In addition to differences in the content and the structure of knowledge
systems, differences in other aspects such as problem comprehension and task
representation may also lead experts and non-experts to different knowledge
criticality judgment. Sonnentag (1998) found that expert software designers were
better at capturing the essence of a problem very early in the problem solving stage
while non-expert designers improved their problem comprehension over time. Chi,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
92
Feltovich, and Glaser (1982) also found that experts had better representations of
tasks than non-experts. Experts are more likely to detect the underlying similarities
shared by a cluster of problems while non-experts focus on the ostensible
dissimilarities on the surface (Chi, Feltovich, and Glaser, 1981). Expert’s superior
ability in problem comprehension and task representation explains why experts,
compared with non-experts, have better intuitions that enable them to respond to a
problem solving situation more rapidly and more accurately (Simon, 1986) and
engage in task-related knowledge search (Bouman, 1980) in a more effective
manner. These differences suggest that experts and non-experts may disagree not
only on what knowledge is critical to the focal task, but also on when that knowledge
becomes critical.
Experts are also more likely to have access to diverse knowledge than non
experts. Morse and Gordon (1974) studied scientists working in private sectors and
found that high performing, innovative scientists tended to go outside their
immediate social circle for new knowledge while less innovative scientists tended to
limit their knowledge search within the circle. Due to differences in exposure, the
critical knowledge introduced by innovative scientists from outside sources may
appear novel to inward-focused scientists and may be dismissed as tangential by less
innovative scientists.
Based on these arguments, I derived two hypotheses regarding the
relationship between expertise level and group members’ disagreement on
knowledge criticality judgment.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
93
Hypothesis 2a: Average level of expertise in a group is negatively
related to group members’ disagreement on knowledge criticality
judgment.
Hypothesis 2b: Expertise level diversity in a group is positively
related to group members’ disagreement on knowledge criticality
judgment.
Social Knowledge and Knowledge Criticality Judgment
Hayek (1945) warned us against the sole reliance on scientific knowledge and
reminded us more than half a century ago that both scientific knowledge and
knowledge of the social context are critical to decision making and problem solving.
Social knowledge as well as task knowledge plays a critical role in determining how
group members use knowledge for task performance purposes (Levine and
Moreland, 1991). Social knowledge refers to “knowledge about the group’s
structure; its culture, norms, methods of coordination; and task performance
strategies” (Argote, Insko, Yovetich, and Romero, 1995, p. 525). For instance, role
expectation is one type of social knowledge. Role expectations are often dynamic
and multidimensional, especially in complex tasks (Brown, Ganesan, and
Challagalla, 2001). Group members must make sense of the constantly changing task
environment on an ongoing basis and figure out what they are expected to do given
their respective roles. Members with the same role functions may have different
understandings about the task due to differences in sense making. Members with
different role functions may harbor different priorities. These differences may lead
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
94
them to different judgment of the criticality of the same knowledge item.
Other social knowledge, including knowledge concerning “local norms,
values, language schemes, and subcultures,” has been proposed to affect how a
problem is defined and what is the appropriate solution (Katz and Tushman, 1979, p.
145). According to Katz and Tushman (1979), a problem is not definable only in
common (i.e., shared) terms; a “local orientation” is critical to problem definition as
well (p. 145). Cramton (2001) found that the lack of situational and contextual cues
was one of the major obstacles to coordinating geographically dispersed work
groups. She argued that groups distributed at different locations might be evaluated
according to different criteria and allocated different amounts of time to finish the
task. These differences may affect their definition of the task and what knowledge is
critical to accomplishing the task and when.
Hypothesis 3: Diversity of social knowledge in a group is positively
related to group members’ disagreement on knowledge criticality
judgment.
The Moderating Effect of Task Uncertainty
People’s perceptions of the criticality of a knowledge item vary especially in
complex task environments that involve much uncertainty (Halpin, Streufert, Steffey,
and Lanham, 1971). As tasks become more uncertain and complex, expert
knowledge is no longer concentrated in those high in the hierarchy but instead, more
evenly distributed across hierarchical levels (Katz and Tushman, 1979). This
increases the differentiation in task knowledge as well as social knowledge among
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
95
task performers, and hence disagreement on knowledge criticality judgment. When
the features of new knowledge are ambiguous (as in many uncertain task
environments), people tend to interpret new knowledge as congruent to their prior
knowledge, a phenomenon addressed by distortion theory (Heit, 1994). Since
different individuals possess different prior knowledge either due to differences in
their functional training, levels of expertise, or understandings of the social
environment, they are more likely to make biased interpretations of new knowledge
in uncertain tasks, which may lead to different assessments of the criticality of the
new knowledge to the focal task.
Hypothesis 4\ Task uncertainty moderates the strength of the
association between the diversity in group members’ task knowledge
and social knowledge and their disagreement on knowledge criticality
judgment with stronger associations expected in more uncertain task
environments.
Communication and Disagreement on Knowledge Criticality Judgment
Although knowledge diversity in a group increases its innovative capacity
(Cohen and Levinthal, 1990), its information search efficiency (Ginsberg, 1990), and
the number of choices available in collective decision making (Brunnson, 1982), it
creates communication and coordination problems and may not produce the optimal
outcome as expected. As Ancona and Caldwell (1992) have pointed out, the diverse
knowledge composition in a group does not have a uniformly positive effect on
performance; rather, its effects are contingent on the degree to which group
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
96
processes successfully maximize positive effects (e.g., enhanced creativity) and
minimize negative effects (e.g., difficulty in coordination). Communication is clearly
a key process component mediating the relationship between knowledge diversity
and group performance.
The connection between knowledge diversity and communication received
some attention in past research. Glick, Miller, and Huber (1993) reported that
functional diversity had a negative effect on the amount of person-to-person
communication among group members because of increased conflicts encountered
by members with different functional backgrounds. Sykes, Lamtz, and Cox (1976)
discovered that people tended to interact with similar others even if they were
physically far apart. Triandis (1960) found that shared mentality contributed to
communication effectiveness.
Ashby’s (1956) “law of requisite variety” provides an alternative view of the
link between knowledge diversity and communication. The law of requisite variety
stipulates that a match between the diversity of a system and the diversity of the
environment is a crucial condition for positive performance. When applied to the
discussion here, the diversity of communication among members should match the
diversity of the knowledge composition of the group to ensure optimal performance.
That is, when the members in a group have knowledge tapping a large number of
domains and/or highly differentiated expertise levels, communication should be
highly diverse for the group to perform effectively.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
97
In Study III, I analyze two dimensions of communication: communication
diversity and communication strength. Communication diversity consists of
structural diversity and content diversity. Communication structural diversity refers
to the extent to which individuals communicate with others who have task
knowledge in other domains, different levels of expert knowledge in the same
domain, and different understandings of the social environment. Communication
content diversity refers to the diversity of the knowledge content exchanged through
communication, be it task knowledge in various areas or social knowledge tapping
various dimensions of group life.
As discussed before, individuals with different task knowledge and social
knowledge are more likely to disagree on what knowledge is critical to the focal task,
the degree of criticality, and when it becomes critical. Communication patterns
characterized by high structural diversity and content diversity expose group
members to different knowledge criticality judgment and help them reach a sufficient
degree of consensus - a precondition for smooth coordination in the process of
performing the task. The first step to consensus building is to make everyone aware
of the divergent views present in a group. A simulation study conducted by Carley
(1991) found that a broad knowledge distribution would facilitate consensus
building. This is because people can only agree on a perspective if it is directly
related to the knowledge they already have (Carley, 1986). The greater exposure
brought by higher communication structural diversity and content diversity increases
the probability that group members understand others’ point of view, which fosters
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
98
consensus formation. Even if some members still disagree on certain issues after
communication, at least they are aware of the source of disagreement. The awareness
of alternate views will reduce coordination difficulties and improve performance.
Hypothesis 5a: Communication structural diversity is negatively
related to group members’ disagreement on knowledge criticality
judgment.
Hypothesis 5b: This relationship is stronger in groups with more
diverse knowledge composition (i.e., diverse groups) than groups with
less diverse knowledge composition (i.e., homogeneous groups).
Hypothesis 6a: Communication content diversity is negatively related
to group members’ disagreement on knowledge criticality judgment.
Hypothesis 6b: This relationship is stronger in groups with more
diverse knowledge composition (i.e., diverse groups) than groups with
less diverse knowledge composition (i.e., homogeneous groups).
Another method that would potentially increase consensus and improve
coordination is to have group members engage in intense communication, a feature
called “communication strength.” C. West Churchman wrote in his foreword to
Mitroff s (1983) book The Subjective Side o f Science that “truth is as much
psychological as it is logical” (p. xii). He argued that people had to be
psychologically prepared to accept a new way of thinking. Frequent and deep
communication will help to achieve mutual understanding and remove the
psychological barrier to acquiring a new perspective.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
99
Hypothesis 7a: Communication strength is negatively related to group
members’ disagreement on knowledge criticality judgment
Hypothesis 7b: This relationship is stronger in groups with more
diverse knowledge composition (i.e., diverse groups) than groups with
less diverse knowledge composition (i.e., homogeneous groups).
Disagreement on Knowledge Criticality Judgment and Group Performance
Disagreement on knowledge criticality judgment would presumably have a
negative effect on group performance. Although such disagreement may encourage
discussion of different opinions and would potentially lead to performance benefits,
it often takes too much time to reach a reasonable degree of consensus that allows for
an actionable solution. This is because the difference in knowledge criticality
judgment results from gaps in individual knowledge background that is hard to
change over night. Such difference may persist after extensive discussion. When
group members have much disagreement on what knowledge is critical to the focal
task, some of the disagreement may not be shared explicitly but may lead to the
withholding of critical knowledge nonetheless. This can be another contributing
factor to decreased performance.
Hypothesis 8: Disagreement on knowledge criticality judgment is
negatively associated with group performance.
All hypotheses were summarized in Figure 4.1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 4.1 Study III: Summary of Hypotheses
(Moderator)
Task Uncertainty
Disagreement on
Knowledge Criticality
Judgment
Communication
- Strength
- Structural Diversity
- Content Diversity
Knowledge Composition
- Diversity of Task Knowledge
• Domain Diversity (+)
• Expertise Level (-)
• Expertise Level Diversity
- Diversity of Social Knowledge
Group
* Performance
Method
Participants
Two hundred and three undergraduates enrolled in an introductory business
class in a large university on the west coast participated in the survey. The
participants represented 85% of the total enrollment in this class. The participants
worked in groups on a term project. Group size ranged from three to five. Altogether,
there were sixty four groups. About half of the participants (50.2%) were 20-21 years
old, 22.7% between 18 and 19, 18.2% between 22 and 23, and 8.4% were over 24
years old. Among all participants, 123 were males and 80 were females. Most of the
participants were Caucasians (40.4%) and Asians (44.3%). Approximately 80% of
the participants were sophomores and juniors.
Procedures
The approval of research was obtained from the University Internal Review
Board. The researcher contacted the course instructor and asked for permission to
solicit participation in his class. After obtaining the permission from the course
instructor, the researcher went to the class as scheduled and handed out consent
forms. Potential participants read the consent form to decide whether they would like
to do the survey. Those who chose to participate signed the consent form and
proceeded to the survey. Each group was assigned a group number by random and
this information was displayed on an overhead transparency during the course of the
survey. The participants were instructed to report their group number in the survey.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
102
This procedure allowed the researcher to aggregate individual responses into group
data in data analysis.
Instrumentation
Domain Diversity. Diversity is an index on the collective level. It describes
the degree of variation among the members in a group. Domain diversity is a
measure of the diversity of task-related knowledge domains among group members.
Knowledge domain is a categorical variable. According to Jackson et al. (1991),
Blau’s (1977) measure of heterogeneity/diversity is most appropriate for categorical
variables. Blau (1977, p. 9) proposed the following formula to compute
heterogeneity for categorical variables:
Heterogeneity = (1 - X/?,2 ).
In this formula,/? represents the ratio of the number of persons in a category (i.e.,
subgroup) to the total number of persons in the whole group and i is the number of
categories represented in a group. Blau’s heterogeneity index ranges from 0 (if all
members belong to the same category) to near 1 and increases as the number of
categories increases (assuming the total number of persons in a group remains
constant). Blau’s formula applied to cases where one person belongs to only one
category.
In measuring domain diversity, category refers to knowledge domain. The
course instructor identified three knowledge domains relevant to the project:
distribution management, operation management, and marketing. Since it is possible
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
103
that one person has expertise in more than one domain, a variant of Blau’s formula
was created to measure domain diversity.
Domain Diversity = 1/i [(1 - p i 2) + (1 - p 2) + (1 ~P32) + .. .(1 - p t2)]
In this formula, pi is the ratio of the number of experts in domain i to group size (n =
5). If there is 0 expert in every domain, the domain diversity index should have the
same value as all members (n = 5) are experts in all domains. Therefore, “0” was
recoded into “5” in data analysis. The domain diversity index ranged from 0 to 1
with greater values indicating greater diversity.
Expertise Level. Expertise level was measured by the mean of the expertise
levels in various knowledge domains reported by other members in the group.
Expertise Level Diversity. Expertise level diversity is defined as the variation
of the expertise levels among group members. Expertise level diversity is an interval
variable measured by participants’ ratings on a scale of 1 (low expertise) to 5 (high
expertise). According to Allison (1978), the coefficient of variation (an index
computed by dividing standard deviation by the mean) is a more appropriate
heterogeneity measure for interval variables than standard deviation because of its
scale-invariant feature.
Diversity of Social Knowledge. Diversity of social knowledge is a self-report
measure of the degree to which group members’ social knowledge differs from one
another. Social knowledge refers to knowledge about group structure, culture, norms,
coordination mechanisms, performance strategies, role expectations, and language
schemes. Items were written to assess each dimension of social knowledge.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
104
Communication Structural Diversity. Communication structural diversity
(CSD) refers to the degree to which members communicate with people who have
different task knowledge and social knowledge. Walsh (1988) borrowed
Hirschman’s (1964) formula of trade concentration to compute functional diversity
V £(Xi / X)2
where Xj is the length of experience (measured by the number of years) in function i;
X is the length of total work experience.
In a similar vein, I used this formula to compute communication structural
diversity (CSD), where Xi was the amount of time spent in communicating with
people who had expertise in a domain other than i, who possessed a different level of
expertise, and who had different social knowledge; X was the total amount of time
spent in communication with other members. The formula thus yielded three
statistics CSDi, CSD2, and CSD3 when applied to domain knowledge, expertise
levels, and social knowledge, respectively. Each of these statistics ranged from a
theoretically low 0 (i.e., group members only communicated with those who had task
knowledge in domains other than their own, on different level in the same domain, or
different social knowledge) to 1 (i.e., group members only communicated with those
who had task knowledge in the same domain, same level of expertise, or same social
knowledge). Communication structural diversity (CSD) was the mean of CSDi;
CSD2 > and CSD3. The correlations among the three measures ranged from .272 to
.450, all significant at .001 level.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
105
Communication Content Diversity. Communication content diversity refers
to the range of knowledge exchanged through communication among members. The
broader the range, the greater the communication content diversity. The participants
were presented with a set of knowledge domains the professor identified as relevant
to the project (see Appendix C). The participants were asked to choose the ones they
communicated with other group members. The number of distinct knowledge areas
(both task knowledge and social knowledge) discussed in communication served as
the measure of communication content diversity with larger numbers indicating
greater diversity.
Communication Strength. Communication strength describes the intensity of
communication. Stevenson and Gilly (1991) developed three items to measure tie
strength: “whether a respondent had a work tie to the next node, whether the
respondent went to the next node for technical advice, and whether the respondent
had social (not work-related) conversations with the next node.” Marsden and
Campbell (1984) proposed two dimensions for communication strength - time spent
on communication and depth of communication. They also found that closeness
rather than frequency was a better indicator of communication strength. Based on
these two articles, a scale measuring communication strength was constructed.
Task Uncertainty. Van de Ven, Delbecq, and Koenig (1976) described task
uncertainty as a two-dimensional construct consisting of task variety and task
analyzability. Perrow (1967) defined task variety as the extent to which work
activities change in different circumstances and task analyzability as the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
106
understandability of work activities and the availability of information search
procedures in cases of exceptions. Daft and Macintosh (1981) constructed two scales
to measure task variety and task analyzability based on Perrow’s (1967) conceptual
analysis. Daft and Macintosh’s (1981) scales on analyzability and variety was used
to assess task uncertainty in this study because of their high reliabilities (Cronbach’s
Alphas = .86 for the analyzability scale and .77 for the variety scale). Since
uncertainty is “perceptual” (Downey & Slocum, 1975), the participants were asked
to provide their perceptual evaluations of task variety and analyzability.
Disagreement on Knowledge Criticality Judgment. With an aim to assessing
the degree to which members in a group disagree on what knowledge is critical to the
focal task, participants were asked to indicate on a five-point Likert scale (ranging
from “strongly disagree” to “strongly agree”) their evaluations of the items tapping
three dimensions of the disagreement on knowledge criticality judgment - that is,
disagreement on what knowledge was critical to the focal task, the degree of
criticality, and at what stage the knowledge became critical to the focal task.
Group Performance. Group performance was an objective measure indicated
by grades on the group projects.
The items measuring each construct are included in Appendix C.
Analysis
Before testing the hypotheses, Cronbach’s Alpha was computed for each
scale to assess its reliability. This procedure produced reasonable reliability scores
for all scales, Cronbach’s Alphas = .78, .71, .65, .70, .89, .70, and .72 for scales
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
107
measuring diversity of social knowledge, communication strength, task variety, task
analyzability, disagreement on knowledge criticality judgment, communication
structural diversity, and communication content diversity. The scale analysis also
suggested ways to improve the reliability of two scales. For instance, deleting the
item “different subgroups have developed different terminologies and languages
unique to those groups” would bring up the reliability score of the diversity of social
knowledge scale up to .85 and deleting the item “my work on the group project is
routine” would increase the reliability of the task variety scale to .73. These two
items were thus excluded from subsequent hypothesis testing.
Regression was used to test the hypotheses. Regression assumes that data are
normally distributed. Kolmogorov-Smimov test performed on each variable
suggested that all data points were normally distributed.
Regression also assumes that all observations are independent. The members
in the same group had shared experience, and therefore, their responses were
interdependent. With an aim to avoiding the violation of the independent observation
assumption of regression, 203 cases in the raw data were aggregated into group data
that produced 64 cases (i.e., groups) with more than one member in each group
filling out the survey. The satisfaction of two criteria provided justification for this
aggregation (Edmondson, 1999; Kenny and LaVoie, 1985). First, the variables
included in this study were meaningful theoretically as group variables. For instance,
domain diversity is a meaningful construct only on the collective level. Second, the
intraclass correlations (ICC) for group variables were greater than zero, a criterion
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
108
that suggested a convergence among the responses within a group and a systematic
difference across groups (see Table 4.1 for ICCs and results of analysis of variance).
Table 4.1 Study III: Analysis of Variance and Intraclass Correlation Coefficients for
Group-Level Scales
F(63, 204) P
ICC
Diversity of Social Knowledge 4.48 <.001 .41
Domain Diversity 4.57 <.001 .54
Expertise Level Diversity 5.01 <.001 .40
Task Uncertainty 2.54 <.001 .13
Disagreement on Knowledge
Criticality Judgment
9.29 <.001 .43
Communication Strength 2.01 <.001 .20
Communication Content Diversity 1.97 <.001 .10
Communication Structural Diversity 2.71 <.001 .36
Group Performance 2.06 <.001 .25
Results
To assess Hypotheses 1, 2a, 2b, and 3, regression was performed on group
data by regressing disagreement on knowledge criticality judgment on domain
diversity, expertise level, expertise level diversity, and diversity of social knowledge,
respectively. The results supported Hypothesis 1, j3 = .15, t (62) = 1.99, p < .047,
Hypothesis 2a, j3 = -.19, t (62) = -2.78, p < .006, and Hypothesis 3, ]3 = .25, t (62) =
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
109
3.24, g < .001, but not Hypothesis 2b, J 3 = .12, t (62) = 1.51, p < .135 (see Table 4.2
for the summary). The results suggested that disagreement on knowledge criticality
judgment among group members was positively related to domain diversity,
expertise level, and diversity of social knowledge, but not expertise level diversity.
Table 4.2 Summary for Simultaneous Regression of Knowledge Diversity on
Members’ Disagreement on Knowledge Criticality Judgment
Predictor Beta t (62)
P
Domain Diversity .15* 1.99 .047
Expertise Level
19* *
-2.78 .006
Expertise Level Diversity .12 1.51 .135
Diversity of Social Knowledge 3.24 .001
Note: N = 64. For betas, *** p < .001, **_p < .01, *p < .05.
Hypothesis 4 predicted that task uncertainty would moderate the magnitude
of the relationship between the diversity of members’ task knowledge and social
knowledge and their disagreement on knowledge criticality judgment with a stronger
relationship expected for tasks with higher uncertainty. To test this hypothesis, the
data file was split into two parts, one with task uncertainty equal or greater than the
mean (high perceived task uncertainty) and the other lower than the mean (low
perceived task uncertainty). The results supported Hypothesis 4. Specifically, when
tasks were perceived highly uncertain, the regression procedure yielded the
following results: j3 = .10, t (62) = 1.39, p < .165 with domain diversity entered as the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
110
predictor, J 3 = .05, t (62) = .71, p < .479 with expertise level diversity as the
predictor, and J 3 = .301, t (62) = 4.39, g < .001 with diversity of social knowledge as
the predictor. When tasks were perceived less uncertain, the strength of the
relationship was weaker (indicated by lower p values), J 3 = .26, t (62) = .47, g < .660
for domain diversity, J 3 = .25, t (62) = .53, g < .625 for expertise level diversity, and
J 3 = .62, t (62) = 1.29, g < .267 for diversity of social knowledge.
This procedure is superior to moderated regression in this study. Russell and
Bobko (1992) have found that the use of a continuous scale rather than a discrete
Likert scale increases the effect size of the moderated regression analysis by an
average of 93%. Russell and Bobko (1992) argue that “the use of relatively coarse
Likert scales to measure fine dependent responses causes information loss that,
although varying widely across subjects, greatly reduces the probability of detecting
true interaction effects” (p. 336). They also argue that information loss poses a
problem especially in cases where the interaction itself is continuous rather than
constant. In this study, I used a discrete Likert scale to measure task uncertainty.
Moreover, the interaction was continuous, meaning that any changes in task
uncertainty would affect the relationship between group knowledge diversity and
members’ disagreement on knowledge criticality judgment. The use of moderated
regression to evaluate the moderating effect of task uncertainty in the relationship
between group knowledge diversity and disagreement on knowledge criticality
judgment may not be able to detect true effects. Splitting the data into high vs. low
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I l l
task uncertainty and performing regression analyses on these two sets of data is more
likely to find true effects.
Hypothesis 5a, 6a, and 7a predicted that communication structural diversity,
communication content diversity, and communication strength each negatively
impacted group members’ disagreement on knowledge criticality judgment. The
regression results (regressing disagreement on knowledge criticality judgment on
each of the three communication measures) provided support for communication
structural diversity, j) = -.14, t (62) = -1.87, p < .049, but not for communication
content diversity, J 3 = .08, t (62) = 1.01, p < .316, nor communication strength, J 3 = -
.05, t (62) = -.61, p < .548 (see Table 4.3 for the summary).
Table 4.3 Summary for Simultaneous Regression of Communication on Members’
Disagreement on Knowledge Criticality Judgment
Predictor Beta t(62)
P
Communication Strength -.05 -.61 .548
Communication Structural Diversity -.14* -1.87 .049
Communication Content Diversity .08 1.01 .316
Note: N = 64. For betas, *p < .05.
Hypothesis 5b, 6b, and 7b predicted that the three communication variables
would have a stronger effect on group’s disagreement on knowledge criticality
judgment when task knowledge and social knowledge were more diverse. The same
procedure used to test Hypothesis 4 was repeated here. The data file was split into
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
112
halves with the means of domain diversity, expertise level diversity, and social
knowledge. Those data points above the mean were for diverse groups and those
below the mean were for homogeneous groups. There were some variations in
within-group knowledge diversity. Among sixty-four groups, the lowest domain
diversity index was .00 and the highest was 2.24. The mean domain diversity was
.56, SD = .31. For expertise level diversity, the lowest score was .00 and the highest
was .62. The mean was .23, SD = .11. The lowest social knowledge diversity index
was 1.88 and the highest was 5.00. The mean of social knowledge diversity was
3.17, SD = .61.
For communication structural diversity and disagreement on knowledge
criticality judgment, stronger relationship was detected for diverse groups, £ = -1.75,
t (35) = -1.67, p < .042, than for homogeneous groups, J 3 = -.14, t (27) = -1.15, p <
.256. Contrary to the prediction, communication content diversity was found to have
a positive effect on group members’ disagreement on knowledge criticality judgment
and the relationship between these two variables was stronger in homogeneous
groups, J 3 = .14, t (35) = 1.12, p < .267, than in diverse groups, J 3 = .02, t (27) = 1.94,
P < .846, although neither of the relationships was significant. For communication
strength and disagreement on knowledge criticality judgment, the strength of the
association between these two variables was about the same in diverse groups, J 3 = -
.12, t (35) = -1.17, p < .246, as in homogeneous groups, J 3 = -.13, t (27) = -1.12, p <
.268. All significant regression coefficients were negative, which suggested that the
directions of these associations were negative as consistent with the prediction.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
113
To perform a post-doc analysis on the relationship between each of the three
communication variables and disagreement on knowledge criticality judgment with
respect to each type of knowledge diversity, the data file was split into two parts
according to the means of domain diversity, expertise level diversity, and diversity of
social knowledge, respectively. The regression procedure was performed by
regressing disagreement on knowledge criticality judgment on each communication
variable. The results showed that the patterns of associations for communication
structural diversity and communication content diversity in three types of knowledge
diversity and for communication strength in domain diversity and expertise level
diversity were similar to those reported earlier. For communication strength,
however, the results suggested that it had a stronger negative effect on group
members’ disagreement on knowledge criticality judgment in groups that had less
diverse social knowledge, J 3 = -.04, t (33) = -.45, p < .654, than in groups that had
more diverse social knowledge, J 3 = -.34, t (29) = -2.01, p < .048.
Hypothesis 8 predicted that disagreement on knowledge criticality judgment
would have a negative effect on group performance. The regression procedure
generated results in support of this hypothesis, J 3 = -.18, t (62) = -2.48, p < .032.
Discussion
Study III addresses a challenge that groups face in sharing unshared
knowledge. This challenge is posed, in part, by differential knowledge criticality
judgment - that is, different members in a group have different understandings of
what knowledge is critical to the focal task and therefore, needs to be retrieved. I
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
114
argue that the diversity of group members’ task knowledge, which consists of
domain diversity and expertise diversity, and social knowledge are positively related
to their disagreement on knowledge criticality judgment. The results suggested that
expertise level was negatively related to group members’ disagreement on
knowledge criticality judgment and that domain diversity and diversity of social
knowledge were positively related to group members’ disagreement on knowledge
criticality judgment, but expertise level diversity was not related.
Giyen the characteristics of the sample in Study III, we should exercise
caution in interpreting the result of no relationship between expertise level diversity
and disagreement on knowledge criticality judgment. The participants in this study
were students from various disciplines in business. They were either freshmen or
sophomores. There might not be enough variations in their expertise levels that
would permit an effective testing of the hypothesis. For instance, no participant in
Study III could claim to be an expert in any of the knowledge areas associated with
the group project. The arguments made in Study III emphasized the differences in
the content and the structure of experts’ knowledge systems and that of non-experts
and how these differences could potentially affect individual knowledge criticality
judgment. The participants in Study III might have some differentiation in their
expertise levels, but this differentiation might not be great enough to create
significant differences in the content and the structure of their knowledge systems,
and therefore differences in knowledge criticality judgment did not surface in the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
115
data. Hypothesis 2b should be re-evaluated with data collected among a sample of
individuals with more differentiated expertise levels.
Task uncertainty was found to aggravate the effects of knowledge diversity
on disagreement on knowledge criticality judgment. Although there was only one
task involved in Study III, participants’ perception of the uncertainty of the task
differed from one another nonetheless. The differences in perceived task uncertainty
would lead task performers to approaching the task differently and making different
judgments on the criticality of the same knowledge item. Therefore, the results on
the moderating effect of task uncertainty on the relationship between knowledge
diversity and disagreement on knowledge criticality judgment were tentative and
needed support from additional research.
With respect to the role of communication strength and diversity,
communication structural diversity was negatively associated with members’
disagreement on knowledge criticality judgment and this association was stronger in
diverse groups than in homogeneous groups. This is consistent with the prediction.
Communication structural diversity is one of the most important determinants of
what knowledge task performers are exposed to. As group members communicate
with others with expertise in a different domain, with expertise at another level, or
with different social knowledge, they will acquire knowledge possessed by others
that they themselves were not previously aware of. An increase in communication
structural diversity will thus increase the amount of shared (or common) knowledge
among task performers. The shared knowledge base will reduce the amount of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
116
disagreement on knowledge criticality judgment and contribute to consensus
formation. The contributing effect communication structural diversity has on the
formation of a shared knowledge base will be stronger in diverse groups than in
homogeneous groups because the members in homogeneous groups already have
much knowledge in common whereas those in diverse groups have little shared
knowledge given the differences they have in knowledge domains, expertise levels,
and social knowledge. The increase in shared knowledge is less in homogeneous
groups than in heterogeneous groups. It was not surprising that the effect of
communication structural diversity on members’ disagreement on knowledge
criticality judgment was found stronger in diverse groups than in homogeneous
groups.
Communication strength was negatively related to group members’
disagreement on knowledge criticality judgment, but this relationship was of similar
magnitude in diverse groups as in homogeneous groups. When this relationship was
examined with respect to domain diversity, expertise level diversity, and diversity of
social knowledge respectively, it was found that the only significant difference in the
strength of the association was with diversity of social knowledge. The negative
association between communication strength and disagreement on knowledge
criticality judgment was substantially stronger in groups where diversity of social
knowledge was low than in groups where diversity of social knowledge was high.
This finding suggests that shared social knowledge facilitates communication to a
greater extent than shared knowledge domains and similar expertise levels.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
117
Communication content diversity, however, was positively related to group
members’ disagreement on knowledge criticality judgment and this relationship was
stronger in homogeneous groups than in diverse groups. This finding was
contradictory to the prediction. However, a more careful deliberation would help to
illuminate this seemingly confusing finding. As group members share a greater
number of different types of knowledge with one another (i.e., communication
content diversity increases), they introduce more subjects on which they may
disagree. In diverse groups, disagreement on knowledge criticality judgment is likely
to be very high even when communication content diversity is low whereas in
homogeneous groups, disagreement on knowledge criticality judgment may be very
low if the content exchanged through communication is not diverse. As
communication content diversity increases, the increase in disagreement on
knowledge criticality judgment will be substantially greater in homogeneous groups
than in diverse groups. Thus, it made sense that communication content diversity
was found to be more strongly related to disagreement on knowledge criticality
judgment in homogeneous groups than in diverse groups.
Disagreement on knowledge criticality judgment was found negatively
correlated with group performance. Since communication, especially communication
strength and structural diversity, were proved to effectively reduce such a
disagreement especially in groups with more diverse knowledge composition,
stronger and more structurally diverse communication ties would potentially improve
group performance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
118
In Study III, I examined group members’ disagreement on knowledge
criticality judgment rather than knowledge relevance judgment. This is because the
term “criticality” is less ambiguous than “relevance.” If the participants were asked
to respond to the question “what knowledge do you think is relevant to the focal
task,” they would have included items that were both critical and those that were less
critical but relevant, even remotely. Substituting “relevant” with “critical” helped the
participants limit the scope of their responses to the most important items. Framing
the question in terms of criticality made it easier to detect the patterns of
disagreement among group members. That said, we should also be aware that the
choice of “critical” over “relevance” might have reduced the richness of the analysis.
For instance, it excluded cases where the discussion of relevant knowledge could
lead to the discovery of critical knowledge.
There are two major limitations in Study III. First, student work groups are
not the ideal setting for testing the hypotheses, especially the hypothesis on the effect
of expertise level diversity and disagreement on knowledge criticality judgment.
Future research should collect data among groups in real work place to replicate the
findings in this study. Second, the data gathered in this study were self-report on
various measures related to one task. Although different people had different
perceptions of the uncertainty of the same task and approached the task somewhat
differently, it would be better if the hypotheses were tested with data gathered among
groups that performed one of the two tasks that differed in uncertainty. This is
because groups that face uncertain task environments may structure themselves
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
119
differently from groups that face less uncertain task environments. The difference in
structure may lead to different knowledge distribution patterns in a group, and hence
disagreement on knowledge criticality judgment. Future research should measure
group structure and examine in detail the effect of group structure on knowledge
distribution and group members’ disagreement on knowledge criticality judgment.
Study III has implications for both research and practice. The findings
reported in this study have opened a new avenue for research on knowledge sharing
in groups that addresses both challenges and potential solutions. An important
question is: How can we build a group knowledge system that maintains a sufficient
degree of differentiation and at the same time allows for enough interconnections
among differentiated subsystems so that members can achieve similar evaluations of
the criticality of the same knowledge items and share them effectively. Study III also
has other practical implications. Managers of groups may use communication as an
intervention tool to help group members bring out unique knowledge to bear on the
task. As reported in this study, communication structural diversity is particularly
useful in producing agreement among members’ knowledge criticality judgment.
Managers of groups should create structures that encourage communication flow
among members with expertise in different domains, differentiated levels of
expertise in the same domain, and different social knowledge. Much earlier research
(e.g., Sykes et al., 1976) and theory of homophily (e.g., Ibarra, 1992; Westphal and
Milton, 2000) has shown that people like to interact more with similar others.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
120
Intervention programs that aim to enhance communication structural diversity will
prove useful to group knowledge sharing.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
121
CHAPTER FIVE
GENERAL DISCUSSION AND CONCLUSION
This dissertation includes three studies examining the effects of two
exogenous variables - task complexity and group knowledge composition - on
information sharing and how information sharing antecedent, processes, and
outcome are related to one another. Information sharing is viewed through the lens of
TM, an increasingly popular model in group research. TM views information sharing
as consisting of three processes: directory updating, information allocation, and
retrieval coordination. These three processes are critical to information sharing in
groups where members collaborate on a common task. With an aim to managing and
sharing information strategically, group members have to learn about sources of
expertise and assume or allocate information processing, storage, and retrieval
responsibilities. An important finding shared by all three studies was that
information sharing had performance consequences, especially in complex task
conditions. Effective information sharing was found positively related to group
performance. This finding suggests that like many other group processes,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
122
information sharing has to be well managed and utilized in order to achieve a
positive performance outcome.
The three studies reported in this dissertation are closely related to one
another (see Figure 10 for a summary of major findings). Study I examined the
relationship among antecedents, transactive processes, and performance outcome in
TM, an area overlooked in past research. Also, in addition to investigating the effect
of each transactive process on group performance, it also examined the interaction
among the three processes and how this interaction affected group performance. The
results suggested that information allocation was not related to group performance.
This finding was further examined in Study III. Information allocation sets its
premise on diverse knowledge composition in groups. In other words, information
allocation makes more sense in groups where members have knowledge in different
domains or expertise levels are highly differentiated. Study III offered an explanation
for the absence of the association between information allocation and group
performance - that is, group members who allocate the information and those who
receive the information may disagree on what information is critical to the focal task,
and this difference may create a discrepancy between what the sender wants to
allocate and what the receiver wants to absorb.
Study III also further elucidated the results from Study II. Study II found that
task complexity was positively related to task conflict. One possible consequence of
dealing with a complex task is that group members have more diverse knowledge
and skills, the independent variable examined in Study III. The dependent variable in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
123
Study III was disagreement on knowledge criticality judgment, one form of task
conflict examined in Study II. Overall, the negative relationship found between
group knowledge diversity and members’ disagreement on knowledge criticality
judgment in Study III provided a direct support for the negative correlation between
task complexity and task conflict reported in Study II.
Taken together, the three studies included in this dissertation offer us a better
understanding of a couple issues. First, the relationship between task complexity and
group performance is not as simple as some researchers have predicted. Intuitively,
one would imagine there is a negative association between task complexity and
group performance. The results in Study II suggested that this was not universally
true. The nature of the relationship between these two variables is contingent on
mediating processes. For instance, task complexity is positively related to group
performance if TM is the mediating variable. This is because task complexity is
positively related to TM and TM is positively related to group performance.
However, task complexity is negatively related to group performance if relationship
conflict is the mediating variable. This is because task complexity is positively
related to task conflict; task conflict is positively related to relationship conflict; and
relationship conflict is negatively related to group performance. The implication of
this finding is that, with an aim to better managing knowledge assets and improving
performance, group members in complex task conditions should spend their effort
constructing an effective TM system and minimizing relationship conflict given its
damaging effect on both TM and group performance.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
124
The destructive effect of relationship conflict on TM and group performance
received ample support in both Studies I and II. It was found that interpersonal
relationship, rather than expertise recognition, primarily determined from whom
group members retrieved information. Specifically, Study II found that relationship
conflict was negatively associated with information sharing processes in TM systems
and that task conflict affected information sharing through relationship conflict. No
direct relationship was detected between task conflict and information sharing in
TM. Study I reported that while directory updating could affect group performance
through retrieval coordination, it could also exert a direct effect on group
performance. This finding implies that even if group members know the location of
expertise, they may not go to the expert for the needed information. They may prefer
to retrieve information from the ones with whom they maintain a good relationship
rather than from the expert with whom they experience interpersonal friction. On the
other hand, the expert may not be willing to reveal expert information to people
whom they don’t have a good term with. Information sharing works in both
directions: information seekers seek information from experts and experts provide
information to information seekers. Interpersonal hostility impedes information
sharing in both processes.
The management implications of these findings are very clear. Since the
quality of information sharing is critical to the performance outcome and
interpersonal relationship plays a central role in ensuring smooth information flow,
managers should develop effective intervention programs to facilitate relationship
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
125
building. This is especially important in the early stage of group formation and in
groups who plan to work together for a long time (Katz, 1982).
Another important finding centers on homophily. The theory of homophily
states that group members have the tendency to communicate more with people who
are similar to them in educational background, socio-economic status, ethnicity, and
so on so forth. (Ibarra, 1992; Sykes et al., 1976; Westphal and Milton, 2000). Study
III examined the flip side of the theory of homophily - the theory of heterophily. It
looked at how group knowledge diversity affected the way in which group members
communicated with one another. Diversity research so far hasn’t been able to draw a
uniformly positive or negative link between diversity and performance. It is possible
that some intervening variables determine whether diversity would facilitate or
impede performance. Study III argued that, although knowledge diversity would
potentially increase group performance by providing more resources to bear on the
task, it could also decrease performance by increasing disagreement on knowledge
criticality judgment among group members. The results supported the hypotheses.
The theory heterophily argues that group members with diverse knowledge
background are less likely to communicate with one another. There should be a
negative correlation between knowledge diversity and communication. This may be
because knowledge diversity increases members’ disagreement on knowledge
criticality judgment (a form of task conflict), which induces relationship conflict,
which in turn inhibits communication (see Figure 5.1).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced w ith permission o f th e copyright owner. Further reproduction prohibited without permission.
Figure 5.1 Integrated Model: Combining Studies I, II, and III
+
i t i v
Task
Complexity
Knowledge
Diversity
Task
Conflict
Relationship
Conflict
Transactive
Memory
Group
Performance
Communication
127
Given diverse group’s difficulty in communicating and sharing unshared
information, diverse groups face more challenges in constructing a smooth TM
system. Study III suggested that communication be used as a tool to facilitate
information sharing. For instance, managers may contrive a structure that encourages
communication across functions and expertise levels, and hence, increase the
probability of sharing unique (unshared) knowledge rather than common (shared)
knowledge. Manipulating the physical structure of an office is one way to
accomplish this goal. An example of this would be the knowledge building at British
Airways (Earl, 2003). In this knowledge building, there is no elevator connected to
the parking garage. Employees have to walk to the first level, cross the lobby, and
take the elevator from there. The rationale behind this design is that having
employees walk across the lobby increases the probability of meeting someone from
another department or someone from the same department but with whom one does
not typically communicate with. When people run into one another in the lobby, they
may stop for a brief conversation and share information spontaneously. The
randomness in meeting people in this manner increases the likelihood of sharing
information that would otherwise not be shared.
Besides practical managerial implications, the three studies reported here also
have important theoretical implications. Study I suggested that TM’s description of
the way in which information sharing antecedents, processes, and outcome were
related to one another did not accurately capture what was really going on in natural
groups. The results supported a modified TM model for groups. Study II suggested
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
that various contextual variables be taken into consideration in examining
information sharing processes in TM systems. Only by developing comprehensive
models that encompass those variables, can we apply the models to concrete
situations. TM by itself is too rational and insensitive to contextual variations. Study
III argued that sometimes group members do not share unique information because
they are not aware of the importance of that information. This awareness account,
together with Stasser’s (1992) probability account, will provide a more
comprehensive explanation of information sharing in work groups.
These three studies report initial empirical results on the theories or models
they endorse. Future research is needed to replicate these findings to validate and
expand these models and to deepen and enhance our understanding of group
information sharing. Possible issues to be considered include motivation and
turnover. TM theory has been proposed as the basis for developing effective group
and organizational knowledge systems (e.g., Anand, Manz, & Glick, 1998;
Hollingshead, Fulk, & Monge, 2002). A precondition for TM systems to function
smoothly is that its members are willing to task responsibility for a portion of
knowledge and to share that knowledge with others. The reality, however, is that in
many cases knowledge is not shared because there are insufficient incentives for
sharing, or even dysfunctional reward systems that work against the distribution of
information (Orlikowski, 1993). Knowledge is also a resource for individuals, who
may want to hoard it in their own self-interest at the expense of the larger collective.
Any well-balanced theory of knowledge management must consider individual and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
129
collective motivation as well as the structure that an effective system should assume.
Xu, Fulk, Hollingshead, and Levitt (2005) proposed an expansion of TM theory to
consider (1) individual and collective incentives to participate effectively in a TM
system, and (2) factors that lead to development of an effective TM system. They
argued that a TM system is a “public good” to the members of the system, and make
predictions regarding participation and system development that are consistent with
the literature on collective action.
Regarding the turnover issue in TM systems, Xu (2005) described a task-
contingency perspective on the effects of turnover on TM systems. She argued that
(1) turnover has potentially adverse effects on TM systems; (2) these adverse effects
are dependent on or moderated by the structural characteristics of the knowledge
networks and the communication networks in TM systems; (3) the structural
configurations of the knowledge networks and the communication networks in TM
systems are contingent on task; and (4) therefore, the effects of turnover on TM
systems are task-contingent. The task-contingency perspective on turnover effects on
TM systems helps to reconcile the inconsistent findings reported in earlier research
and to provide an overarching theoretical framework for future empirical research on
turnover effects on TM.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
130
REFERENCES
Aldrich, H. (1999). Organizations evolving. CA, Thousand Oaks: Sage Publication.
Allison, P. D. (1978). Measures of inequality. American Sociological Review. 43
(6), 865-880.
Ashby, W. R. (1956). An introduction to cvbemetics. London: Chapman & Hall.
Amason, A. C. (1996). Distinguishing effects of functional and dysfunctional
conflict on strategic decision making: Resolving a paradox for top management
teams. Academy of Management Journal. 39, 123-148.
Anand, V., Manz, C. C., & Glick, W. H. (1998). An organizational memory
approach to information management. Academy of Management Review. 23 (4).
796-809.
Ancona, D. G., & Caldwell, D. F. (1992). Demography and design: Predictors of
new product team performance. Organization Science. 3 (3). 321-341.
Archer, R. L. (1980). Self-disclosure. In D. M. Wegner & R. R. Vallacher (Eds.),
The self in social psychology. New York: Oxford University Press.
Argote, L. (1999). Organizational learning: Creating, retaining, and transferring
knowledge. Boston: Kluwer.
Argote, L., Insko, C. A., Yovetich, N., & Romeo, A. A. (1995). Group learning
curves: The effects of turnover and task complexity on group performance. Journal
of Applied Social Psychology. 25 (6). 512-529.
Ashby, W. R. (1956). An introduction to cybernetics. London: Chapman & Hall.
Blanchard, W. (1966). Relevance of information and accuracy of interpersonal
prediction: A methodological note. Psychological Reports. 18, 379-382.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
131
Blau, P. M. (1977). Inequality and heterogeneity: A primitive theory of social
structure. New York, NY: The Free Press.
Bollen, K. A. (1989). Structural equations with latent variables. New York: John
Wiley & Sons.
Borgatti, S. P., & Cross, R. (2003). A relational view of information seeking and
learning in social networks. Management Science, 49 (41, 432-445.
Bouman, M. J. (1980). Application of information-processing and decision-making
research. In G. R. Ungson & D. N. Braunstein (Eds.), Decision making: An
interdisciplinary inquiry (pp. 129-167). Boston: Kent.
Bradley, J., White, B. J., & Mennecke, B. E. (2003). Teams and tasks: A temporal
framework for the effects of interpersonal interventions on team performance. Small
Group Research. 34 (31. 353-387.
Brown, S. P., Ganesan, S., & Challagalla, G. (2001). Self-efficacy as a moderator of
information-seeking effectiveness. Journal of Applied Psychology, 86 (51, 1043-
1051.
Bruner, J. S (1957). On perceptual readiness. Psychological Review, 64 (21. 123-152.
Brunson, N. (1982). The irrationality of action and action rationality: Decisions,
ideologies and organizational actions. Journal of Management Studies. 19. 29-44.
Campbell, D. J. (1988). Task complexity: A review and analysis. Academy of
Management Review, 13 (11. 40-52.
Comer, D. R. (1991). Organizational newcomers’ acquisition of information from
peers. Management Communication Quarterly, 5 (11. 64-89.
Carley, K. (1986). Knowledge acquisition as a social phenomenon. Instructional
Science. 14. 381-438.
Carley, K. (1991). A theory of group stability. American Sociological Review. 56
(3), 331-354.
Chase, W. C., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology. 4,
55-81.
Chi, M., Feltovich, P., & Glaser, R. (1981). Categorization and representation of
physics problems by experts and novices. Cognitive Science, 5, 121-152.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
132
Chi, M., Feltovich, P., & Glaser, R. (1982). Categorization in experts and novices. In
R. Sternberg (Ed.), The handbook of intelligence (pp. 66-84). Hillsdale, NJ:
Erlbaum.
Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective
on learning and innovation. Administrative Science Quarterly, 35, 128-152.
Cramton, C. D. (2001). The mutual knowledge problem and its consequence for
dispersed collaboration. Organization Science. 12 (31, 346-371.
Daft, R. L., & Macintosh, N. B. (1981). A tentative exploration into the amount and
equivocality of information processing in organizational work units. Administrative
Science Quarterly. 26, 207-224.
Damanpour, F. (1991). Organizational innovation: A meta-analysis of effects of
determinants and moderators. Academy of Management Journal, 34, 555-590.
Day, E. A., Arthur, W. Jr., & Gettman, D. (2001). Knowledge structures and the
acquisition of a complex skill. Journal of Applied Psychology. 86 15), 1022-1033.
De Dreu, C. K. W. (1997). Productive conflict: The importance of conflict
management and conflict issue. In C. K. W. De Dreu & E. Van de Vliert (Eds.),
Using conflict in organizations (pp. 9-22). London, UK: Sage.
De Dreu, C. K. W., & Van Vianen, A. E. M. (2001). Managing relationship conflict
and the effectiveness of organizational teams. Journal of Organizational Behavior,
22* 309-328.
Derlega, V. J., Metts, S., Petronio, S., & Margulis, S. T. (1993). Self-disclosure.
Newbury Park, CA: Sage.
Downey, H. K., & Slocum, J. W. (1975). Uncertainty: Measures, research, and
sources of variation. Academy of Management Journal. 18 (3), 562-578.
Earl, M. (2003). Knowledge management strategies in use. Presented at the Fourth
European Conference on Organizational Knowledge, Learning, and Capabilities,
OKLC 2003. April 13, Barcelona, Spain.
Edmondson, A. (1999). Psychological safety and learning behavior in work teams.
Administrative Science Quarterly. 44. 350-383.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
133
Eisenberg, E. M., & Phillips, S. R. (1991). Miscommunications in organizations. In
N. Coupland, H. Giles, & J. M. Wiemann (Eds.), Miscommunications and
problematic talk (pp. 244-258). Newbury Park, CA: Sage.
Faraj, S., & Sproull, L. (2000). Coordinating expertise in software development
teams. Management Science. 46 (12). 1554-1568.
Fisher, D. V. (1984). A conceptual analysis of self-disclosure. Journal for the Theory
of Social Behavior. 14 (3), 277-296.
Fisher, B. A., & Ellis, D. G. (1990). Small group decision making: Communication
and the group process. New York: McGraw Hill.
Ford, J. K., & Kraiger, K. (1995). The application of cognitive constructs and
principles to the instructional systems model of training: Implications for needs
assessment, design, and transfer. In C. L. Cooper & I. T. Robertson (Ed.),
International review of industrial and organizational psychology (pp. 1-48).
Chichester, England: Wiley.
Gigone, D., & Hastie, R. (1993). The common knowledge effect: Information
sharing and group judgment. Journal of Personality and Social Psychology. 65 (5),
959-974.
Ginsberg, A. (1990). Connecting diversification to performance: A sociocognitive
approach. Academy of Management Review. 15 (3). 514-543.
Glick, W. H., Miller, C. C., & Huber, G. P. (1993). The impact of upper-echelon
diversity on organizational performance. In G. P. Huber & W. H. Glick (Eds.),
Organizational change and redesign: Ideas and insights for imnroving performance
(pp. 176-214). New York, NY: Oxford University Press.
Grunfeld, D. H., Mannix, E. A., Williams, K. Y., & Neale, M. A. (1996). Group
composition and decision making: How member familiarity and information
distribution affect process and performance. Organizational Behavior and Human
Decision Processes. 67 (11. 1-15.
Guetzkow, H., & Gyr, J. (1954). An analysis of conflict in decision making groups.
Human Relations, 7* 367-381.
Hackman, J. (1987). The design of work teams. In J. Lorsch (Ed.), Handbook of
organizational behavior (pp. 315-342). Englewood Cliffs, NJ: Prentice-Hall.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
134
Halpin, S. M., Streufert, S., Steffey, J., & Lanham, N. (1971). Information load,
proportion of relevance, and relevance perception. Psvchonomic Science, 23 (6),
404-406.
Hardin, R. (1995). One for all: The logic of group conflict. Princeton, NJ: Princeton
University Press.
Hayek, F. A. (1945). The use of knowledge in society. American Economic Review,
35 141. 519-530.
Heit, E. (1994). Models of the effects of prior knowledge on category learning.
Journal of Experimental Psychology: Learning. Memory, and Cognition, 20 16).
1264-1282.
Henry, R. A. (1995). Improving group judgment accuracy: Information sharing and
determining the best member. Organizational Behavior and Human Decision
Processes, 62 (2), 190-197.
Hollingshead, A. B. (1998a). Communication, learning, and retrieval in transactive
memory systems. Experimental Social Psychology, 34, 423-442.
Hollingshead, A. B. (1998b). Retrieval processes in transactive memory systems.
Journal of Personality and Social Psychology, 74, 659-671.
Hollingshead, A. B. (1998c). Distributed expertise and transactive processes in
decision-making groups. In M. A. Neale, E. A. Mannix, & D. H Gruenfeld (Eds.),
Research on Managing in Groups and Teams. (Volume 1) (pp. 105-125). Greenwich,
CN: JAI Press.
Hollingshead, A. B. (2000). Perceptions of expertise and transactive memory in work
relationships. Group Process and Intergroup Relations. 3, 257-267.
Hollingshead, A. B. (2001). Cognitive interdependence and convergent expectations
in transactive memory. Journal of Personality and Social Psychology, 81, 1080-1089.
Hollingshead, A. B., Fulk, J., & Monge, P. (2002). Fostering intranet knowledge-
sharing: An integration of transactive memory and public goods approaches. In S.
Kiesler & P. Hinds (Eds.), Distributed work: New research on working across
distance using technology. Boston: MIT Press.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
135
Hollingshead, A. B., Wittenbaum, G. M., Paulus, P. B., Hirokawa, R. Y., Ancona, D.
G., Peterson, R. S., Jehn, K. A., & Yoon, K. (2005). A look at groups from the
functional perspective. In M. S. Pool, & A. B. Hollingshead (Eds.), Theories of small
groups: Interdisciplinary perspectives (pp. 21-62). Thousand Oaks, CA: Sage
Publications.
Hu, L. Z., & Bentler, P. M. (1995). Evaluating model fit. In R. H. Hoyle (Ed.),
Structural equation modeling: Concepts, issues, and applications (pp. 76-99).
Thousand Oaks, CA: Sage.
Ibarra, H. (1992). Homophily and differential returns: Sex difference in network
structure and access in an advertising firm. Administrative Science Quarterly. 37.
422-447.
Jackson, S. E., Brett, J. F., Sessa, V. I., Cooper, D. M., Julin, J. A., & Peyronnin, K.
(1991). Some differences make a difference: Individual dissimilarity and group
heterogeneity as correlates of recruitment, promotions, and turnover. Journal of
Applied Psychology. 7615). 675-689.
Janssen, O., Van De Vliert, E., & Veenstra, C. (1999). How task and person conflict
shape the role of positive interdependence in management teams. Journal of
Management. 25 (2), 117-141.
Janz, B. D., Colquitt, J. A., & Noe, R. A. (1997). Knowledge worker team
effectiveness: The role of autonomy, interdependence, team development, and
contextual support variables. Personnel Psychology. 50, 877-904.
Jehn, K. A. (1995). A multimethod examination of the benefits and detriments of
intragroup conflict. Administrative Science Quarterly, 40, 256-282.
Jehn, K. A., Chadwick, C., & Thatcher, S. M. B. (1997). To agree or not to agree:
The effects of value congruence, individual demographic dissimilarity, and conflict
on workgroup outcomes. International Journal of Conflict Management. 8 (4). 287-
305.
Jehn, K. A., & Mannix, E. A. (2001). The dynamic nature of conflict: A longitudinal
study of intragroup conflict and group performance. Academy of Management
Journal. 44 (2). 238-251.
Jehn, K. A., Northcraft,G. B., & Neale, M. A. (1999). Why differences make a
difference: A field study of diversity, conflict, and performance in work groups.
Administrative Science Quarterly, 44 (4). 741-763.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
136
Joreskog, K. G., & Sorbom, D. (1996). LISREL 8: User’s reference guide. Chicago,
IL: Scientific Software.
Kaplan, D., & Wenger, R. N. (1993). Asymptotic independence and separability in
covariance structure models: Implications for specification error, power, and model
modification. Multivariate Behavioral Research, 28, 483-498.
Katz, R. (1978). The influence of job longevity on employee relations to task
characteristics. Human Relations, 31 (8), 703-725.
Katz, R. (1982). The effects of group longevity on project communication and
performance. Administrative Science Quarterly, 27, 81-104.
Katz, R., & Tushman, M. (1979). Communication patterns, project performance, and
task characteristics: An empirical evaluation and integration in an R&D setting.
Organizational Behavior and Human Performance. 2 3 ,139-162.
Kelly, H. H. (1979). Personal relationships: Their structure and processes. Hillsdale,
NJ: Erlbaum.
Kelly, J. R., Futoran, G. C., & McGrath, J. E. (1990). Capacity and capability: Seven
studies of entertainment of task performance rates. Small Group Research. 21 (3).
283-314.
Kenny, D. A., & LaVoie, L. (1985). Separating individual and group effects. Journal
of Personality and Social Psychology, 48. 339-348.
Kirschenbaum, S. (1992). Influence of experience on information-gathering
strategies. Journal of Applied Psychology, 77 (3), 343-352.
Knecht, L., Lippman, D., & Swap, W. (1973). Similarity, attraction, and self
disclosure. Proceedings of the American Psychological Association. 8, 205-206.
(Summary)
Korsgaard, M. A., Schweiger, D. M., & Sapienza, H. J. (1995). Building
commitment, attachment, and trust in top management teams: The role of procedural
justice. Academy of Management Journal, 38, 60-84.
Larson, J. R. Jr., Foster-Fishman P. G., & Keys, C. B. (1994). Discussion of shared
and unshared information in decision-making groups. Journal of Personality and
Social Psychology. 67 (3), 446-461.
Lawrence, P., & Lorsch, J. (1967). Differentiation and integration in complex
organizations. Administrative Science Quarterly, 12. 1-47.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
137
Levine, J. M., & Moreland, R. L. (1991). Culture and socialization in work groups.
In L. B. Resnick, J. M. Levine, & S. D. Teasley (Eds.), Perspectives on socially
shared cognition (pp. 257-279). Washington DC: American Psychological
Association.
Lewis, K. (2003). Measuring transactive memory systems in the field: scale
development and validation. Journal of Applied Psychology. 88 (4), 587-604.
Liang, D. W. (1994). The effects of top management team formation on firm
performance and organizational effectiveness. Unpublished Doctoral Dissertation,
Carnegie Mellon University, Pittsburgh, PA.
Liang, D. W., Moreland, R. L., & Argote, L. (1995). Group versus individual
training and group performance : The mediating role of transactive memory.
Personality and Social Psychology Bulletin, 21, 384-393.
Libby, R., Trotman, K. T., & Zimmer, I. (1987). Member variation, recognition of
expertise, and group performance. Journal of Applied Psychology. 72, 81-87.
Lindsay, P. H., & Norman, D. A. (1972). Human information processing: An
introduction to psychology. New York, NY: Academic Press.
Littlepage, G., Robinson, W., & Reddington, K. (1997). Effects of task experience
and group experience on group performance, member ability, and recognition of
expertise. Organizational Behavior and Human Decision Processes. 69 (2), 133-147.
Lovelace, K., Shapiro, D. L., & Weingart, L. R. (2001). Maximizing cross-functional
new product teams’ innovativeness and constraint adherence: A conflict
communication perspective. Academy of Management Journal, 44 (4). 779-793.
Marsden, P. V., & Campbell, K. E. (1984). Measuring tie strength. Social Forces, 63,
482-501.
McGrath, J. E. (1984). Groups: Interaction and performance. Inglewood, NJ:
Prentice Hall.
McKeithen, K. B., Reitman, J. S., Rueter, H., & Hirtle, S. C. (1981). Knowledge
organization and skill differences in computer programmers. Cognitive Psychology.
13, 307-325.
Medin, D. L., Lynch, E. B., Coley, J. D., & Atran, S. (1997). Categorization and
reasoning among tree experts: Do all roads lead to Rome? Cognitive Psychology, 32,
49-96.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
138
Mennecke, B. E., & Valacich, J. S. (1998). Information is what you make of it: The
influence of group history and computer support on information sharing, decision
quality, and member perceptions. Journal of Management Information Systems. 15
(2), 173-197.
Milliken, F. J., & Martins, L. L. (1996). Searching for common threads:
Understanding the multiple effects of diversity in organizational groups. Academy of
Management Review, 21 (2). 402-433.
Miranda, S. M., & Saunders, C. S. (2003). The social construction of meaning: An
alternative perspective on information sharing. Information Systems Research. 14
(1), 87-108.
Mitroff, 1.1. (1983). The subjective side of science: A nhilosophical inquiry into the
psychology of the Apollo Moon scientists. Seaside, CA: Intersystems Publications.
Moreland, R. L. (1999). Transactive memory: Learning who knows what in work
groups and organizations. In L. Thompson, J. Levine, & D. Messick (Eds.), Shared
cognition on organizations (pp. 3-31): Lawrence Erlbaum Associates.
Moreland, R. L., Argote, L. (2003). Transactive memory in organizations. In R. S.
Peterson & E. A. Mannix (Eds.), Leading and managing people in the dynamic
organizations (pp. 135-162). Mahwah, NJ: Lawrence Erlbaum Associates.
Moreland, R. L., Argote, L., & Krishnan, R, (1996). Socially shared cognition at
work: Transactive memory and group performance. In J. L. Nye & A. M. Bower
(Eds.), What’s social about social cognition: Research on socially shared cognition in
small groups (pp. 57-84). Newbury Park, CA: Sage.
Moreland, R. L., & Myaskovsky, L. (2000). Exploring the performance benefits of
group training: Transactive memory or improved communication? Organizational
Behavior and Human Decision Processes, 82. 117-133.
Morrison, E. W. (1993). Newcomer information seeking: Exploring types, modes,
sources, and outcomes. Academy of Management Journal. 36 (3), 557-589.
Morse, E. V., & Gordon, G. (1974). Cognitive skills: A determinant of scientists’
local-cosmopolitan orientation. Academy of Management Journal. 17 (41. 709-723.
Moynihan, L., & Bratt, R. (2000). Antecedents and consequences of transactive
memory in service teams: Theory and scale development. Paper presented at the
2000 Annual Meeting of the Academy of Management.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
139
Murray, A. I. (1989). Top management group heterogeneity and firm performance.
Strategic Management Journal. Summer, 125-141.
O’Reilly, C. A. (1980). Individuals and information overload in organizations: Is
more necessarily better? Academy of Management Journal. 23 (4). 684-694.
Orlikowski, W.J. (1993). Learning from notes: Organizational issues in groupware
implementation. The Information Society Journal, 9, 237-250.
Pelled, L. H. (1996). Demographic diversity, conflict, and work group outcomes: An
intervening process theory. Organization Science, 7 (6), 615-631.
Pelled, L. H., Eisenhardt, K. M., & Xin, K. R. (1999). Exploring the black box: An
analysis of work group diversity, conflict, and performance. Administrative Science
Quarterly, 44. 1-28.
Perrow, C. (1967). A framework for the comparative analysis of organizations.
American Sociological Review, 32, 194-208.
Porat, A., & Hass, J. (1969). Information effects on decision making. Behavioral
Science, 14, 98-104.
Rao, R., & Argote, L. (1992). Collective learning and forgetting: The effects of
turnover and group structure. Paper presented at the Academy of Management
Meetings, Las Vegas.
Rau, D. (2000). The effect of transactive memory on the relationship between
diversity of expertise and performance in top management teams. Paper presented at
the Academy of Management Meetings, Toronto.
Roseman, I. J., Cynthia, W., & Swartz, T. S. (1994). Phenomenology, behaviors, and
goals differentiate discrete emotions. Journal of Personality and Social Psychology.
67,206-221.
Rulke, D. L., & Galaskiewicz, J. (2000). Distribution of knowledge, group network
structure, and group performance. Management Science, 46 (5). 612-625.
Russell, C. J., & Bobko, P. (1992). Moderated regression analysis and Likert scales:
Too coarse for comfort. Journal of Applied Psychology. 77 13), 336-342.
Schvaneveld, R. W., Durso, F. T., Goldsmith, T. E., Breen, T. J., Cooke, N. M.,
Tucker, R. G., & De Maio, J. C. (1985). International Journal of Man-Machine
Studies. 23. 699-728.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
140
Scott, R. W. (1998). Organization: Rational, natural, and open systems (4th ed.).
New York, NY: Prentice Hall.
Sessa, V. I., & Jackson, S. (1995). Diversity in decision-making teams: All
differences are not created equal. In M. M. Chemers, S. Oskamp, & M. A. Costanzo
(Eds.), Diversity in organizations: New perspectives for a changing workplace (pp.
133-156). Thousand Oaks, CA: Sage.
Shaw, M. E. (1964). Communication networks. In L. Berkowitz (Ed.), Advances in
experimental social psychology (pp. 111-147). New York, NY: Academic Press.
Shaw, M. E. (1973). Scaling group tasks: A method for dimensional analysis.
Catalog of Selected Documents in Psychology. Winter. 8.
Simon, H. A. (1986). What we know about the creative process. In .R L. Kuhn (Ed.),
Frontiers in creative and innovative management (pp. 3-20). Cambridge, MA:
Ballinger.
Smith, K. G., Smith, K. A., Olian, J. D., Sims, H. P. Jr., O’Bannonn, D. P., & Scully,
J. A. (1994). Top management team demography and process: The role of social
integration and communication. Administrative Science Quarterly. 39. 412-438.
Snyder, R. A., & Morris, J. H. (1984). Organizational communication and
performance. Journal of Applied Psychology, 69 (31. 461-465.
Sonnentag, S. (1998). Expertise in professional software design: A process study.
Journal of Applied Psychology. 83 (51. 703-715.
Spender, J. C. (1993). Competitive advantage from tacit knowledge? Unpacking the
concept and its strategic implications. Best Paper Proceedings, Academy of
Management Journal Annual Meeting, Atlanta.
Stasser, G. (1992). Pooling of unshared information during group discussions. In S.
Worchel, W. Wood, & J. A. Simpson (Eds.), Group process and productivity (pp. 48-
67). Newbury Park, CA: Sage.
Stasser, G., Stewart, D. D., & Wittenbaum, G. W. (1995). Expert roles and
information exchange during discussion: The importance of knowing who knows
what. Journal of Experimental Social Psychology. 31, 244-265.
Stasser, G., Taylor, L. A., & Hanna, C. (1989). Information sampling in structured
and unstructured discussions of three- and six-person groups. Journal of Personality
and Social Psychology. 57, 67-78.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
141
Steiner, I. (1972). Group process and productivity. New York, NY: Academic Press.
Stevenson, W., & Gilly, M. C. (1991). Information processing and problem solving:
The migration problems through formal positions and networks of ties. Academy of
Management Journal. 34 (41. 918-928.
Stewart, D. D., & Stasser, G. (1995). Expert role assignment and information
sampling during collective recall and decision making. Journal of Personality and
Social Psychology. 69, 619-628.
Storey, D. (1991). History and homogeneity: Effects of perceptions of membership
groups on interpersonal communication. Communication Research, 18 (21. 199-221.
Sujan, H., Sujan, M., & Bettman, J. R. (1988). Knowledge structure differences
between more effective and less effective salespeople. Journal of Marketing
Research. 25. 81-86.
Sykes, R. E., Lamtz, K., & Cox, J. C. (1976). Proximity and similarity effects on
frequency of interaction in a class of naval recruits. Sociometrv. 39 (3), 263-269.
Triandis, H. C. (1960). Cognitive similarity and communication in a dyad. Human
Relations. 13,175-183.
Troutman, C. M., & Shanteau, J. (1977). Inferences based on nondiagnostic
information. Organizational Behavior and Human Performance. 19. 43-55.
Turner, M. E., & Pratkanis, A. R. (1997). Mitigating groupthink by stimulating
constructive conflict. In C. K. W. De Dreu & E. Van de Vliert (Eds.), Using conflict
in organizations (pp. 53-71). London, UK: Sage.
Van de Ven, A. H., Delbecq, A., & Koenig, Jr., R. (1976). Determinants of
coordination modes within organizations. American Sociological Review. 41, 322-
338.
Van De Vliert, E., & De Dreu, C. K. W. (1994). Optimizing performance by conflict
stimulation. International Journal of Conflict Management. 5 (3). 211-222.
Volkema, R. J. (1988). Problem complexity and the formulation process in planning
and design. Behavioral Science. 33, 292-327.
Wall, J. A., & Callister, R. R. (1995). Conflict and its management. Journal of
Management. 21 (3). 515-558.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
142
Walsh, J. P. (1988). Selectivity and selective perception: An investigation of
managers’ belief structures and information processing. Academy of Management
Journal. 31 (41, 873-896.
Watson, W. E., Michaelsen, L. K., & Sharp, W. (1991). Member competence, group
interaction, and group decision making: A longitudinal study. Journal of Applied
Psychology, 76. 803-809.
Wegner, D. M. (1987). Transactive memory: A contemporary analysis of the group
mind. In B. Mullen & G. R. Goethals (Eds.), Theories of group behavior (pp. 185-
208). New York: Springer-Verlag.
Wegner, D. M. (1995). A computer network model of human transactive memory.
Social Cognition. 13 (31, 319-339.
Wegner, D. M., Erber, R., & Raymond, P. (1991). Transactive memory in close
relationships. Journal of Personality and Social Psychology. 61 (6). 923-929.
Wegner, D. M., Giuliano, T., & Hertel, P. (1985). Cognitive independence in close
relationships. In W. J. Ickes (Ed.), Compatible and incompatible relationships (pp.
252-276). New York: Springer-Verlag.
Weingart, L. R. (1992). Impact of group goals, task component complexity, effort,
and planning on group performance. Journal of Applied Psychology, 77 15). 682-
693.
Westphal, J.D., & Milton, L.P. (2000). How experience and network ties affect the
influence of demographic minorities on corporate boards. Administrative Science
Quarterly. 45, 366-398.
Williams, K. Y., & O’Reilly, III, C. A. (1998). Demography and diversity in
organizations: A review of 40 years of research. Research in Organizational
Behavior. 20. 77-140.
Wisnieski, E. J. (1995). Prior knowledge and functionally relevant features in
concept learning. Journal of Experimental Psychology: Learning. Memory, and
Cognition. 21 (2). 449-468.
Wittenbaum, G. M., & Park, E. S. (2001). The collective preference for shared
information. Current Directions in Psychological Science, 10 (2), 70-73.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
143
Wittenbaum, G. M., & Stasser, G. (1996). Management of information in small
groups. In J. L. Nye & A. M. Brower (Eds.), What’s social about social cognition?
Social cognition research in small groups (pp. 3-28). Thousand Oaks, CA: Sage.
Wood, R. E. (1986). Task complexity: Definition of the construct. Organizational
Behavior and Human Decision Processes, 37. 60-82.
Xu, Y., Fulk, J., Hollingshead, A. B., & Levitt, R. E. (2005). Knowledge resource
sharing as collective action: Integrating public goods theory and transactive memory.
Organization Science. Under revision.
Xu, Y. (2005). Exploring the effects of turnover on transactive memory systems: A
task-contingency perspective. Academy of Management Review. Under revision.
Yetton, P. W., & Bottger, P. C. (1982). Individual versus group problem solving: An
empirical test of a best-member strategy. Organizational Behavior and Human
Performance. 29. 307-321.
Zigurs, I. & Buckland, B. K. (1998). A theory of task/technology fit and group
support systems effectiveness. MIS Quarterly. 22 (3). 313-334.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
144
APPENDIX A
STUDY I SURVEY ITEMS
Self-Disclosure
1 .1 tell others in my group what I feel I need the most help with.
2 .1 tell others in my group things I am good at.
3 .1 tell others in my group things I am not good at.
4 .1 tell others in my group my hobbies.
Group Tenure
1. How long has your group worked together on this project (estimate the total
number of weeks)?
2. How many hours, on average, has your group spent on this project each week?
Communication Frequency
1. My group members frequently meet face-to-face at formally scheduled work
sessions.
2. My group members frequently meet face-to-face informally outside the meeting
room (e.g., in social settings).
3. My group members frequently exchange information through formal written
communication (e.g., memos).
4. My group members frequently exchange information through informal written
communication (e.g., personal notes).
5. My group members frequently communicate via telephone, email, or Instant
Messenger.
6. Our group meetings frequently involve more than one member but less than all
members.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
145
Directory Updating
1.1 frequently learn about the expertise of other members of my group.
2 .1 frequently learn about the expertise of other people outside my group whom I
have work relationship with.
3 .1 have accurate information about the expertise of other members of my group.
4 .1 have accurate information about the expertise of other people outside my group
whom I have work relationship with.
Information Allocation
1. When I come across information that is not closely related to my expertise, I’ll
pass it to a relevant expert and let the expert be responsible for processing and
storing that information.
2. When others come across information that is closely related to my expertise, they
will pass it to me and let me be responsible for processing and storing that
information.
3. To ensure quality, we always let the experts in respective knowledge domains
process and store the information pertinent to those domains.
Retrieval Coordination
1.1 work very closely with other group members.
2. My group coordinates knowledge well.
Group Performance
1. The members of the group have all developed the skills needed to work together
on future projects.
2. The members of the group all think that they have a fruitful experience with the
group.
3. Grades for the group project (to be obtained from the course instructor).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
146
APPENDIX B
STUDY II SURVEY ITEMS
Task Complexity
Component Complexity
1. There was a considerable amount of information that must be processed to
complete the task.
2. There were a large number of subtasks requiring specific knowledge and skills
that must be carried out to perform the major task.
Coordinative Complexity
1. The task was guided by standard procedures, directives, rules, etc. (R)
2. Group members knew a lot of procedures and standard practices to do the work
well. (R)
3. Understandable sequence of steps was followed in carrying out the task. (R)
4. People actually relied on established procedures and practices. (R)
5. Established materials (manuals, standards, directives, textbooks, and the like)
covered the task. (R)
6. It required a lot of coordination to complete the task.
Dynamic Complexity
1. Instead of following a set of fixed procedures, we must constantly adapt to
possible changes in order to perform the task well.
2. The task was described as routine. (R)
Task Conflict
1. Group members regularly took divergent viewpoints on the issues involved.
2. Group members often had different ideas on substantive matters.
3. Group members agreed on the course to take from the outset. (R)
4. Group members had predominantly identical ideas on the task. (R)
5. There were regularly different opinions on task-related issues.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
147
6. Diverse perspectives on substantive issues were the rule rather than the exception.
Relationship Conflict
1. The personal relationships were excellent. (R)
2. Some group members visibly disliked each other.
3. The tension among group members was painful.
4. Group members did not get on personally. (R)
5. The atmosphere was very pleasant. (R)
6. The friction among group members was annoying.
Transactive Memory
Specialization
1. Each member of my group has a specialized role.
2. Members of my group have interchangeable roles. (R)
3. Members of my group have a lot of overlapping knowledge. (R)
4. Each member has unique knowledge that they bring to our group.
Expertise Recognition
1.1 know a lot about the expertise of my group members.
2. My group members know a lot about my expertise.
3. My group members know a lot about one another's expertise.
Cognitive Interdependence
1.1 depend very much on the
my job.
2.1 depend very much on the
do my job.
Group Performance
1. The members of the group all developed the skills needed to work together on
future projects.
2. The members of the group all thought that they had a fruitful experience with the
group.
3. Grades for the group projects.
expertise of other members of my group in order to do
expertise of other people outside my group in order to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
148
APPENDIX C
STUDY III SURVEY ITEMS
Domain Diversity
Participants were asked to identify the number of distinct knowledge domains in
each group and the number of experts in each knowledge domain.
Expertise Level Diversity
Participants were asked to indicate on a five-point Likert scale (range from 1-
extremely low to 5-extremely high) the level of expertise of each member in a group
in each knowledge domain.
Diversity of Social Knowledge
1. The members in our group have different understandings about group structure.
2. The members in our group have different understandings about group culture and
norms.
3. The members in our group have different understandings about how we should
coordinate our work.
4. The members in our group have different understandings about how we should go
about performing the task.
5. Different subgroups have developed different terminologies and languages unique
to those groups.
Communication Structural Diversity
Among the total amount of time I spent in communicating with others in the group, I
spent
% of the time communicating with people who had knowledge in a different
domain,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
149
% of the time communicating with people who had expertise on a different
level, and
% of the time communicating with people who had different social
knowledge.
Communication Content Diversity
Please highlight the types of task knowledge and social knowledge you
communicated with others in a group.
Task knowledge: Distribution management, operation management, and marketing.
Social knowledge: Group structure, culture, norms, coordination modes, task
performance strategies, and language schemes.
Communication Strength
1.1 spent much time on work-related conversations with other members in my group.
2.1 spent much time on social (non work-related) conversations with other members
in my group.
3 .1 had deep discussions about work-related issues with other members in my group.
4 .1 had deep discussions about social (non work-related) issues with other members
in my group.
Task Uncertainty
Task Variety
1. The events that cause my work vary quite a bit.
2. My work is routine.
3. The decisions I make at work are differ from one day to the next.
4. It takes a lot of experience and training to know what to do when a problem arises.
5. My tasks require an extensive and demanding search for a solution.
Task Analyzability
1. My normal work activities are guided by standard procedures, directives, rules,
etc.
2.1 know a lot of procedures and standard practices to do the work well.
3. There is an understandable sequence of steps that can be followed in carrying out
the work.
4 .1 rely on established procedures and practices.
5. Established materials (manuals, standards, directives, statutes, technical and
professional books, and the like) cover the work.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
150
Disagreement on Knowledge Criticalitv Judgment
1. The members in our group frequently disagreed on what knowledge was critical to
the focal task.
2. The members in our group frequently disagreed on the degree of criticality of the
same knowledge item.
3. The members in our group frequently disagreed on at what stage the knowledge
became critical to the focal task.
Group Performance
Students’ grades for the group projects
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Knowledge sharing in Chinese surgical teams

PDF

Developing business communication skills: Leveraging stage versus global processes of change in skills improvement approaches

PDF

Coevolution of technology and organization: A case study of BrokerHouse

PDF

A work structure based approach to collaborative engineering design

PDF

Characterization of new factors involved in feline leukemia virus (FELV)-mediated leukemogenesis

PDF

From networks to Nickelodeon to Noggin: A communication networks perspective on the evolution of the children's television community

PDF

Investigation of degrading effects and performance optimization in long -haul WDM transmission systems and reconfigurable networks

PDF

Defining the advice -seeking request in electronic solicitations in boundary spanning environments

PDF

Health-related quality of life, occupation and prosthesis use in elderly people with lower extremity amputation in Taiwan

PDF

A system of participation and allocation preferences in organizations

PDF

Individual heterogeneity and program evaluation

PDF

A hierarchical co-evolutionary approach to conceptual design

PDF

Adaptive energy conservation protocols for wireless ad hoc routing

PDF

Foreign capital inflows and trade policy change: The automotive and consumer electronics industries in Brazil and China

PDF

Feeling good about control: Design considerations for accountability systems in schools

PDF

An organizational culture approach to the study of individual power and its use

PDF

A socio-technical approach to support collaborative engineering design

PDF

A critical analysis of ideology and discourse in two hotels

PDF

Internet connectedness and its social origins: An ecological approach to communication media and social inequality

PDF

Attachment, accommodation, and the outcomes of romantic relationships

Asset Metadata

Creator Xu, Yan (author)

Core Title Information sharing in work groups: A transactive memory approach

School Graduate School

Degree Doctor of Philosophy

Degree Program Communication

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,speech communication

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Fulk, Janet (committee chair), Adler, Paul (committee member), Riley, Patricia (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-468331

Unique identifier UC11336672

Identifier 3196915.pdf (filename),usctheses-c16-468331 (legacy record id)

Legacy Identifier 3196915.pdf

Dmrecord 468331

Document Type Dissertation

Rights Xu, Yan

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA