Citation
The impact of telecommuting on team effectiveness

Material Information

Title:
The impact of telecommuting on team effectiveness
Creator:
Bach, Reina
Publication Date:
Language:
English
Physical Description:
ix, 70 leaves : ; 28 cm

Thesis/Dissertation Information

Degree:
Master's ( Master of Arts)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
Department of Psychology, CU Denver
Degree Disciplines:
Psychology

Subjects

Subjects / Keywords:
Telecommuting ( lcsh )
Teams in the workplace ( lcsh )
Teams in the workplace ( fast )
Telecommuting ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 62-70).
General Note:
Department of Psychology
Statement of Responsibility:
by Reina Bach.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
50727776 ( OCLC )
ocm50727776
Classification:
LD1190.L645 2002m .B32 ( lcc )

Full Text
THE IMPACT OF TELECOMMUTING ON TEAM EFFECTIVENESS
by
Reina Bach
B.A., Lafayette College, 1986
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Arts
Psychology
2002


This thesis for the Master of Arts
degree by
Reina Bach
has been approved by
Donna Chrobot-Mason
Annette Towler

Date


Bach, Reina (M.A., Psychology)
The Impact of Telecommuting on Team Effectiveness
Thesis directed by Professor Kurt Kraiger
ABSTRACT
Technological advancements have facilitated an increase in
telecommuting as a work alternative, offering benefits not only to individuals
but also to organizations. As the incidence of telecommuting among
individuals increases, the teams on which these individuals work may be
affected.
The current study examined data from 52 teams at a global software
development company containing a minimum of telecommuter. Data
analysis performed at the team level showed a significant positive
relationship between team composition, defined as the percentage of
telecommuters, and team commitment to telecommuting. There was also a
significant positive relationship between team commitment to telecommuting
and team innovation. Mutual trust and team cohesion both acted as
moderators to the relationship between team commitment to telecommuting


and team innovation. Finally, teamwork behaviors were found to mediate
the relationship between team commitment to telecommuting and team
innovation.
As communications technology continues to evolve and organizations
continue to operate in team-based structures, the need for applied research
in this area will exist. The intent of this study was to contribute to the field
of team research in a telecommuting environment.
This abstract accurately represents the content of the candidate's thesis. I
recommend its publication.
Signed
IV


CONTENTS
Figures.................................................--viii
Tables....................................................ix
CHAPTER
1. INTRODUCTION.......................................... 1
Purpose of the Study.................................1
Scope of the Study..................................7
Arrangement of the Thesis............................7
2. REVIEW OF THE LITERATURE................................9
Research at the Individual Level.....................9
Research at the Managerial Level....................12
Research at the Organizational Level................13
Non-Team Research............................ 13
Team Research.................................15
Hypotheses....................................22
3. METHOD.................................................24
Procedure...........................................24
Participants........................................26
Measures.......................................... 27
v


Commitment to Telecommuting..................27
Mutual Trust.................................28
Team Cohesion.............................. 29
Teamwork Behaviors...........................30
Innovation...................................31
Task Interdependence.........................31
Team Composition.............................32
Demographic Variables........................33
Aggregation of Indices.............................33
4. RESULTS.............................................. 37
Descriptives..................................... 37
Hypothesis Testing.................................37
Exploratory Analysis............................. 44
5. DISCUSSION............................................46
Summary of Main Points........................... 46
Contributions of the Current Study.................47
Limitations and Directions for Future Research.....48
APPENDIX
A. TELECOMMUTING INTERVIEW FORM .........................50
B. COMMITMENT TO TELECOMMUTING SURVEY....................52
VI


C. PARTICIPANT TELECOMMUTING SURVEY.................54
BIBLIOGRAPHY.............................................62
vii


FIGURES
Figure
2.1 Conceptual Model of the Current Study
VIII


TABLES
Table
3.1 Within Team Agreement and Intraclass Correlation Coefficients... 35
4.1 Means, Standard Deviations, and Zero Order Correlations......38
- Team Level Analysis
4.2 Hierarchical Regression of Commitment to Telecommuting,......39
Mutual Trust, and the Product of Commitment to
Telecommuting and Trust on Innovation
4.3 Hierarchical Regression of Commitment to Telecommuting,......41
Team Cohesion, and the Product of Commitment to
Telecommuting and Team Cohesion on Innovation
4.4 Zero Order Correlations Centered Data......................42
4.5 Hierarchical Regression of Centered Data.....................42
4.6 Simple Regression Equation for Centered Data -...............43
Regression of Zy on Zt at Particular Values of Zc
IX


CHAPTER 1
INTRODUCTION
Purpose of the Study
Increasingly sophisticated technology allows individuals to
communicate in a distributed environment. There are a number of reasons
why telecommuting is expected to increase as an alternative work
arrangement. The intent of this study is to further research efforts in this
area and to provide insight specifically related to the impact of telecommuting
on team effectiveness.
Technological advancements are making a dramatic impact on
competition among organizations on a global basis. Computer technology
and telecommunications, particularly the Internet, are launching a global
economy. The birth of e-business, whether it is business-to-business or
business-to-consumer commerce over the Internet, has forever changed the
world economy. This evolution is supported by the fact that capital spent on
information technology in the U.S. has more than tripled since 1960, from
10% to 35% (Byrne, 2000).
1


As technology continues to increase the pace of business, operational
efficiency and establishing a competitive advantage are key to organizational
success (Buhler, 1995). To drive new ideas and products and to stay ahead
of the steep new-product curve, it will be critical for businesses to attract,
cultivate, and retain the best thinkers. Management expert Gary Hamel once
said "We have moved from an economy of hands to an economy of heads;
therefore, the price of imagination, the premium for it, will go up" (Buhler,
1995).
Not only are organizations facing pressure to improve operational
processes, they also face the issue of a shrinking labor pool. The supply of
available workers is expected to decrease and yet demand continues to
increase. Despite the recent economic downturn in the U.S., the demand for
labor is still projected to exceed supply within five and by 2031, the U.S.
workplace may be nearly 35 million short of the estimated 58 million new
employees required years (Employment Policy Foundation, 8/23/01).
McKinsey, a global organizational consulting firm, recently updated their 1997
research study. Results showed that "89 percent of the 6,900 managers
surveyed thought it is more difficult to attract talented people now than it was
three years ago and 90 percent thought it is now more difficult to retain
them" (Axelrod, Handfield-Jones, & Welsh, 2001).
2


Organizations are devising creative new ways to win the battle of
attracting and retaining top talent. One method, although perhaps not
utilized as much in light of recent economic conditions, is through the use of
financial incentives and perquisites. National hotel chains offer managers free
luxury vehicles and one of the top public accounting firms gives away new
Jeep Wranglers and $15,000 checks to employees for referring new
candidates who are hired on. (Conlin, Palmer, & Saveri, 1999) Handing out
one-time incentives, until recently, was more prevalent than increasing
employee salaries. Companies know that large across-the-board salary
increases would adversely impact the bottom line, so they are reluctant to
hand them out. Annual salary increases have actually decreased from 5.2%
in 1990 to 4.2% in 1998 and 1999 (Conlin, Palmer & Saveri, 1999). Some
organizations are still implementing variable pay type of programs such as
stock options and bonuses that are linked to performance goals, at individual,
group, and organizational levels.
Another mechanism organizations are utilizing to attract and retain
talent is offering flexible work schedules to their employees. Some
employees work four ten-hour days per week, affording them an extended
weekend. Other employees alter the their daily work schedule, so instead of
9 a.m. to 5 p.m., they might work 6:30 a.m. to 3:30 p.m.
3


Telecommuting and teleworking are other methods that organizations
are using as a way to attract and retain employees and they are gaining
momentum. Telecommuting has allowed organizations to realize benefits
such as lower overhead costs and an enhanced ability to attract and retain
certain highly valued employees (Olson & Primps, 1984). A 2001 General
Accounting Office report estimated that 16.5 million employees telecommute
a minimum of one day per month and that 9.3 million employees
telecommute a minimum of one day per week (Wells, 2001). AT&T
sponsored a study by the International Telework Association & Council.
Results of that study found that the number of teleworkers in the United
States is up nearly 17 percent to 28.8 million people and of this amount, 21.7
percent work from home (International Telework Association & Council,
2000). Logitech Inc., a computer peripherals manufacturer, conducted a
study that showed 80 percent of Americans believe it is important to have an
office workplace at home" (Employment Policy Foundation Newswatch,
11/9/01). It seems that regardless of how one measures telecommuting or
teleworking, all indications are that the number of people participating in
these programs is expected to rise.
Domestic U.S. and foreign governments are not only recognizing, but
promoting telework. The U.S. government provided federal income tax
4


credits to businesses that participated in a voluntary pilot e-commute
program launched before August 17, 2001 in Denver, Colorado and four other
major metropolitan areas (Employment Policy Foundation Newswatch,
8/23/01). European Union employers and unions entered a voluntary
agreement that establishes guidelines to ensure that employees performing
telework enjoy the same protections as other employees, including
employment rights, salary structures and career opportunities.(Employment
Policy Foundation Newswatch, 5/9/01).
Telework and telecommuting are sometimes used interchangeably.
The term telework is more widely used in Europe, but the term
telecommuting is still widely used today in North America (Li, 1998). The
concept of telework goes back several many years and is a broad term that
means utilizing telecommunications to work wherever necessary to in order to
satisfy internal or external client needs. A teleworker may work from a home
office, a telework center, a satellite office, a client's office, an airport lounge,
a hotel room, or the local coffee shop. Telecommuting, on the other hand, is
considered to be a subset of teleworking and is traditionally defined as
"salaried employees who work at home during the day for part or all of the
work week instead of going into the office" (International Telework
Association & Council, 2000). The U.S. Department of Labor considers
5


telecommuting working from home on a regular basis, meaning working a
minimum average of 1.5 days per week from home (Wendell, 2001).
Why are telecommuting and teleworking so popular? There are
several reasons that prompt organizations to embrace them. Some
anticipated benefits include enhanced ability to recruit and retain employees,
reduced absenteeism and turnover, and improved work productivity. Some
studies have shown that employees are more productive and when the labor
market is tight, employers depend on telecommuting as a means to attract
and retain skilled workers (Employment Policy Foundation Newswatch,
6/7/01). According to the International Telework Association & Council,
employees working at home report they are on average 15% more
productive than those who do not work from home. Additionally,
administrative expenses are reduced through fewer office space
requirements. From an environmental standpoint, teleworking and
telecommuting help to reduce traffic and overcrowding. Fewer people
commuting on the highways helps to reduce air pollution.
Teleworking and telecommuting offer benefits to the employee as well.
Employees can avoid the drudgery and risks associated with the traditional
commute to a traditional office (International Telework Asosciation & Council,
2000). Working from home can be more enticing to employees since
6


commute times are increasing. The U.S. Census Bureau's Census 2000
Supplementary Survey reported that work commutes increased by an average
of two minutes between 1990 and 2000 and that the average U.S. worker
took 24.3 minutes to get to work in 2000 (Cree & Sorenson, 2001).
Scope of the Study
The present study was conducted at a global software development
company with approximately 5,000 employees. The definition of
"telecommuter" was defined as an employee working at least one day per
week from home. Teams consisting of at least one telecommuter were
surveyed, which was approximately 1,000 employees located throughout the
United States on 167 teams.
. Arrangement of the Thesis
An overview of the current environment in which organizations
compete is presented as well as specific approaches that organizations are
taking in order to survive in this competitive climate. Then, a broad
framework of empirical research in the area of teleworking and
telecommuting is presented. One of the variables affecting telecommuting
that has received less research attention is the effect of teams. In the next
7


chapter, a methodology for evaluating teamwork in a telecommuting setting
is discussed with specific hypotheses. And finally, analysis results and a
discussion are presented.
8


CHAPTER 2
REVIEW OF THE LITERATURE
Although organizational researchers are only beginning to study
teleworking and telecommuting, a framework for that research can be
identified. There are three major areas into which the research can be
clustered: an individual context, a managerial context, and an organizational
context. Each area contains its own sub-categories.
Research at the Individual Level
The general research themes that fall under the individual context of
telecommuting include the following: work/family balance and boundaries,
interest in and attitudes toward telecommuting, productivity/job satisfaction,
telecommuter and job characteristics, and psychological impacts on
telecommuters. Duxbury, Higgins, and Neufeld (1998) examined how
telework arrangements affect an employee's ability to balance work and
family demands and what employees perceive to be the work and non-work
related advantages of telework arrangements. Their study involved four
groups in three Canadian organizations: 1. teleworkers (n=54); 2. managers
of teleworkers (n=26); 3. co-workers of teleworkers (n=22); and, 4. a control
9


group (n=36). Results indicated that teleworkers had significantly lower
levels of interference from work to family, significantly lower levels of
interference from family to work, and significantly fewer problems managing
their family time than they did prior to being teleworkers. Data supported the
positive view of telework in that they suggest working from home helps
employed parents balance work and family demands. Cree (1998) also
examined work and family balance by analyzing survey data from 7,586
respondents. Results from that study found that work/family balance
increase as telecommuting frequency increases when total hours worked per
week is held constant. There appeared to be a potential positive influence on
work/family balance when people telecommute, if overworking is avoided.
Fireman (1999) developed and tested a model of telecommuting
withdrawal behaviors by taking samples from a Fortune 100 company
(n=1,154) and a smaller company (n=139). Fireman found that community,
the desire for work location social interactions, was associated with
decreasing telecommuting, and that discomfort, a lack of support from the
supervisor and/or from the organization, was associated with ceasing
telecommuting.
Hartman, Stoner, and Arora (1991) collected data from 97
telecommuters at 11 different public and private organizations relative to
10


productivity and satisfaction. Results showed that satisfaction with the
performance evaluation system was significantly correlated with both
productivity and satisfaction, technical and emotional support received from
supervisors was significantly correlated to satisfaction, and family disruptions
and satisfaction were negatively correlated. The ratio of telecommuting
hours to total hours worked was negatively correlated to productivity.
Neufeld (1997) investigated the most significant individual consequences of
telecommuting at three organizations. Results of his study revealed that both
satisfaction and productivity increased due to work arrangement regardless of
level of job demands.
Huws, Korte, and Robinson (1990) analyzed telework survey data
relative to individual and occupational characteristics, although these analyses
were not the main focus of their research efforts. Results suggest that
telework arrangements seem to attract a particularly high proportion of well-
qualified women and that most of the telework occupations are generally
regarded as suitable for decentralization.
Olson and Primps (1984) collected data from 14 companies running
formal pilot programs and 6 high-technology firms. Their results found that
in general, the male professional employees in their study reported that
working at home helped to reduce their stress levels while women
11


consistently reported stress associated with working from home. According
to Gurstein (1991), telecommuters may experience increased social isolation
and increased home- and work-role conflicts.
Research at the individual level showed that telecommuting
arrangements tend to attract a high proportion of highly qualified females and
that it had a positive affect on work/family balance. Although employees may
experience social isolation, they tend to feel more productive and satisfied.
Research at the Managerial Level
Major themes covered under the managerial context include the
relationship between managers and employees, trust, and attributes of
successful managers. Speeth (1992) studied the attributes of successful
managers of telecommuters by interviewing 30 managers, each with a
minimum of six months experience managing telecommuters and at least 25
percent of their staff in a telecommuting situation. Speeth concluded that
telemanagers are experienced, effective, highly educated, and exhibit high
levels of personal achievement. In a study conducted by Lallande (1984),
results showed that managers need to be more disciplined in establishing
clear objectives and maintaining records of instructions provided to their
employees and of the decisions they make. Reinsch (1999) found among 104
12


telecommuting workers that when managers responded constructively to
criticism and when managers displayed high loyalty toward employees,
respondents forecasted a successful relationship between the worker and his
or her manager.
Research at the managerial level showed that although managers are
experienced, highly effective, and highly educated, they need to be more
disciplined in setting clear objectives and documenting instructions for their
telecommuting employees.
Research at the Organizational Level
Finally, the following themes fall under the organizational context of
empirical research on telecommuting: business outcomes and benefits of
telecommuting, organizational culture, organizational commitment and
identification, and teams. There has been little research looking specifically at
the effect of telecommuting on teams. However, telecommuting research at
the individual level has implications for team-level research.
Non-Team Research
Jackson (1999) collected data from 305 employees from a large
telecommunication service organization. Results showed that employees had
13


lower intentions to turnover when flexible work arrangements, including
telecommuting, were available and the then individual had a preference for
flexibility. Olson and Primps (1984) studied 20 companies and reported that
organizations that viewed working at home as a way to empower employees
to have greater control over their work improved the quality of work life for
those employees. Those organizations also realized benefits such as lower
overhead costs and an enhanced ability to attract and retain certain highly
valued employees.
Taveras (1998) analyzed survey data from 250 telecommuters and 250
non-telecommuters and found no difference between the two groups on
perceived organizational support, perceived value congruence, job satisfaction
and commitment. Buessing and Broome (1999) studied 33 teleworkers at
two different companies, examining trust in the company, in supervisors, and
in the work system as well as organizational commitment. Results indicated
that trust in others and in the systems predicted both feelings of alienation
and moral commitment, but only trust in the company predicted job
involvement. Whiting (1998) compared a group of virtual office employees
to a group of similar traditional office employees. Results of this study
suggested virtual office employees had lower organizational commitment
during the transition to the virtual office. Once the transition was complete,
14


however, organizational commitment was equal to that of traditional office
employees. Hill, Miller, Weiner, and Colihan (1998) studied 157 teleworkers
and 89 traditional office workers at IBM. Results of their qualitative analyses
revealed a negative influence on teamwork, supported by comments stating
that camaraderie, mentoring, networking among co-workers and esprit de
corps had suffered due to telework.
Research on telecommuting at the organizational level showed that
employees are less inclined to want to leave an organization when
telecommuting is an available and valued work option. Telecommuting helps
to reduce overhead costs and helps organizations to attract and retain
valuable employees.
Team Research
Why study telecommuting and teams? The organizational structures
of today are becoming more team based, autonomous structures (Platt &
Page, 2001). Organizations are increasingly using teams as a key strategy for
successful management (Hackman, 1986; Peters, 1988). People tend more
to work together in teams rather than as individuals to fulfil complex tasks (Li,
1998). Technology enables geographically dispersed team members to work
together across time zones, allowing organizations to build effective teams
15


from individuals who might not otherwise be available to work together.
Organizations are able to capitalize on availability of resources, both internal
and external, and expertise (Platt & Page, 2001). With the increase of both
telecommuting and team-based work structures, one might anticipate
telecommuting having a growing impact on teams and team performance.
More research is needed at the team level, studying the effects of
telecommuting on task productivity, team cohesion, morale, and member
retention (Tavares, 1998).
Researchers have been studying teams for many years and the
traditional model for conceptualizing performance at the team level is an
input-process-output (I-P-O) model (Guzzo & Shea, 1992). The current study
uses their framework to understand the effects of telecommuting on team
outcomes through the following conceptual model in Figure 2.1.
Figure 2.1. Conceptual Model of the Current Study
16


Inputs. Team composition is the input for the I-P-0 model illustrated
in Figure 2.1. On a team containing traditional office and home office
employees, the proportion of telecommuters on a team may affect the
team's commitment to telecommuting as a work option. If team members
support telecommuting, perhaps the team would operate more effectively.
Group Process. Process refers to aspects of how a team operates.
Three attitudinal aspects relative to telecommuting teams are the teams'
commitment level to telecommuting, mutual trust level among team
members, and team cohesion. Although there are several attitudes that
exist at the team level, these three were selected based on previous
research discussed below. Teamwork behavior can also be categorized
under team process and is also included in this study.
Commitment to an organization has been used in research studies as
both a predictor and criterion variable. Commitment to telecommuting is a
similar concept, but it describes a team's support for telecommuting. It is
the relative strength of an individual's identification with and involvement in,
either direct or indirect, telecommuting. It can be characterized by: a
strong belief in and acceptance of telecommuting as an option for getting
work done, a willingness to exert considerable effort in order for
17


telecommuting to exist as a work option, and a strong desire to maintain
telecommuting as an option for getting work done. Team commitment has
been linked to team performance (Bishop et al., 1997; Hackman, 1987;
Scott & Townsend, 1994). It could seem reasonable that if there were at
least one telecommuter on a team, the team's overall commitment level to
telecommuting may also impact their effectiveness as a team. On such
teams, telecommuting is a means of getting working together and
completing work. Teams that are committed not only to the team itself, but
committed to how team members complete their work (e.g.,
telecommuting), may be more effective.
In general, team attitudes facilitate team behaviors that lead to team
effectiveness. Mutual trust between team members is one such attitude that
is important team process in order for teamwork to exist and it is increasingly
recognized as an essential to examine in applied research (Jones & George,
1998). McIntyre and Salas (1994) suggest that if teamwork exists, team
members monitor one another's performance and in order for such
monitoring to exist, a psychological contract of trust among team members
must exist. Platt and Page (2001) found that high performance teams must
build trust. Dirks (1999) also found that in high-trust teams, motivation was
transformed into joint efforts and hence higher performance, in low-trust
18


teams, motivation was transformed into individual efforts. Jarvenpaa, Knoll,
& Leidner (1998) studied virtual teams, those that "require at least two team
members working at a distance" (Platt & Page, 2001). Their research focused
more on behaviors, such as volunteering for roles and exhibiting individual
initiative, associated with virtual teams that developed trust. This study
focused on behavioral outcomes of virtual teams where trust is present, but
not on specific team attitudes toward trust. Trust is important for team
performance and may actually moderate the relationship between a team's
commitment to telecommuting and their effectiveness as a team, measured
by innovation.
Team cohesion is also an important process variable relative to team
effectiveness (Bettenhausen, 1991). Team cohesion is the desire to stay or
be associated with a given set of team members (Festinger, 1950). Several
studies have found significant relationships between team cohesion and team
performance (Evans & Dion, 1991; Mullen & Copper, 1994; Oliver, 1990).
Team cohesion may moderate the relationship between a team's commitment
to telecommuting and team effectiveness, measured by innovation.
Teamwork behavior is another team process in the I-P-0 model.
McIntyre and Salas (1994) showed that teamwork is a complex of behavioral
characteristics (e.g., Oser, McCalium, Salas, & Morgan, 1989). The key
19


complex behavioral characteristics they found were performance monitoring,
feedback, closed-loop communication, and backing-up behaviors.
McIntyre and Salas (1994) proceeded to identify four principles that
describe essential teamwork behaviors. Principle 1: Teamwork means that
members monitor one another's performance. Effective team members
perform their own tasks while keeping track of their team members' and this
performance monitoring becomes an accepted part of an implicit contract
among team members.
McIntyre and Salas' (1994) identified Principle 2: Teamwork implies
that members provide feedback to and accept it from one another. As team
members monitor one another's performance, they provide feedback to them
on their effectiveness. In their research, they found that often status, rank,
or tenure impeded the free flow necessary for feedback (Driskell & Salas,
1992). In order for teamwork to be effective and teams to be high-
performing, team members should not feel constrained by such impediments.
The next principle is Principle 3: Teamwork involves effective
communication among members, which often involves closed-loop
communication.
Closed-loop communication involves the following sequence of
behaviors: (1) the sender initiates the message; (2) the receiver
20


accepts the message and provides feedback to indicate that the
message has been received; and (3) the sender double-checks to
ensure that the intended message was received (McIntyre & Salas,
1994).
The last principle of essential teamwork behavior they identified is
Principle 4: Teamwork implies the willingness, preparedness, and inclination
to back fellow members up during operations. Back-up behavior helps a
team operate as one, as more than the sum efforts of its individuals.
Stronger teams have members that exhibit a willingness to jump in and assist
when needed, and they accept help without fear of being viewed as weak
(McIntyre & Salas, 1994).
Technology may affect how team members interact and how they
behave and therefore, may affect team effectiveness. The current study
examined potential mediating effects of teamwork behaviors on the
relationship between team commitment to telecommuting and team
effectiveness.
Task interdependence is the degree to which employees feel that their
tasks depend on interacting with others and on others' tasks being completed
(Campion etal., 1993; Kiggundu, 1981,1983; Pearch &Gregersen, 1991).
High task interdependence is one characteristic of teams (Salas et al., 1992;
21


Sundstrom, de Meuse, & Futrell, 1990). Some tasks completed by teams
containing telecommuters cannot be fulfilled on a face-to-face basis, which
may impact team effectiveness. Mediator effects of task interdependence on
the relationship between commitment to telecommuting and innovation was
examined through exploratory analysis in the current study.
Outputs. Team effectiveness is an output, tangible or intangible, of
team processes and refers to any indicator of how successfully a team fulfills
its tasks, mission, or objectives (Kraiger & Wenzel, 1997). Team innovation,
the successful implementation of creative ideas (Amabile, Conti, Coon,
Lazenby, & Herron, 1996), is also an output of team process and is one
measure of team effectiveness (West & Farr, 1990). Team innovation was
used as a criterion variable to measure team effectiveness because
participants included in the study were knowledge worker employees of a
software development company. In that environment, better teams are
those that are more innovative problem-solving.
Hypotheses
Using the I-P-0 model components identified above, this study
analyzed the impact of telecommuting on teams. One might expect that if a
greater proportion of team members telecommutes, then momentum to
22


support telecommuting might be greater. Specifically, the following
hypotheses were made. There is a positive relationship between the
proportion of telecommuters to the commitment to telecommuting
(Hypothesis 1). If a team with telecommuting members is committed to
telecommuting, then the team is more effective. There is a positive
relationship between commitment to telecommuting and innovation
(Hypothesis 2). Trust and cohesion are important aspects of high performing
teams. The relationship strength between commitment to telecommuting
and innovation is moderated by the level of trust (Hypothesis 3) and the
relationship strength between commitment to telecommuting and innovation
is moderated by the level of cohesion (Hypothesis 4). Although a team may
be committed to telecommuting, teamwork behaviors may also impact team
performance. Teamwork behaviors mediate the relationship between
commitment to telecommuting and innovation (Hypothesis 5). Mediating
teamwork behaviors are: monitoring of teammate performance, team
members provide feedback to and accept it from one another, team members
utilize closed-loop communication, and team members are willing, prepared,
and inclined to back their teammates.
23


CHAPTER 3
METHOD
A combination of qualitative and quantitative data collection and
analysis, or mixed-methods approach, were applied in this study. "Such
combination of research approaches will maximize knowledge yield and widen
the scope of research contributions by management and organizational
researchers" (Currali &Towler, 2001).
Procedure
The first step in was to interview two telecommuter program
managers and one manager of telecommuters using a form, appearing in
Appendix A, designed to elicit open-ended comments. One program manager
was particularly interested in gathering information on how managers; were
handling the denial of the request to telecommute by employees. The other
program manager wanted to know the differences in how employees spent
their time at home versus in the office. Although these issues were not
addressed in the current study, items were added to the survey instrument to
collect data for the program manager for informational purposes only. In
discussions with these three individuals, two challenges surfaced. One issue
24


was that remote accessibility varies based on available technology (e.g., DSL,
ISDN, or telephone line), which can affect an employee's ability to do their
job. A second issue was that managers have a difficult time allowing
employees to telecommute, especially on a full-time basis. Many managers
feel that if they cannot physically see an employee working at his or her desk,
the employee must not be working. Neither of the two challenges mentioned
during interviews was addressed in the current study, but are offered as a
guide for future research.
The next step was to identify teams having at least one telecommuter,
defined in this study as an employee who works from home at least one day
per week. A list of approximately telecommuters at the software
development company was provided by the telecommuting program
manager. Along with this list, organizational charts were used to identify all
teammates of the telecommuter reporting to the same manager. For the
purposes of this study, "team" was defined as all individuals reporting up one
organizational level to the same manager. A total of 167 teams comprised of
at least one telecommuter were identified. Individuals on these teams
included various job functions across several business functions including
software development, sales, consulting, customer support, finance, human
resources, information technology, legal, and marketing.
25


The final step was to distribute the survey shown in Appendix C to 983
employees. Each of the 167 teams was assigned a number, which appeared
in the footer of the document (e.g. 0202-115, date followed by team
number). Such coding allowed surveys to be identified by team with the
intent to aggregate data at the team level. Members of the same team
received a survey with the same coding as an attachment to an email from
the researcher. In an attempt to preserve anonymity, the researcher sent the
survey to herself and blind carbon copied members of each team. A total of
167 emails were sent in order to distribute the survey. Employees emailed
completed surveys directly to the researcher, who subsequently deleted the
email and entered data into an SPSS database. Survey response rate was
37%.
Participants
The intent of this study was to analyze data from teams containing at
least one telecommuter. This condition was satisfied if at least one
respondent per team indicated that he telecommutes or that one of his team
members telecommutes. The number of teams satisfying this condition was
88. The researcher further limited participation in the study to include only
those teams with a minimum of three survey responses per team, which
26


resulted in 53 teams. Team researchers recommend using a higher number
of responses per team, if possible.
Measures
Commitment to Telecommuting
A commitment to telecommuting scale was created by modifying the
Organizational Commitment Questionnaire (Mowday, Steers, & Porter, 1979).
This technique has been used successfully in organizational research (Bishop
& Scott, 2000). The commitment to telecommuting is the relative strength of
an individual's identification with and involvement in, either directly or
indirectly, telecommuting.
A pilot study was conducted to validate the new 11-item 5-point Likert
scale where 1 = strongly disagree and 6 = strongly agree. The survey
instrument (Appendix B) was distributed to 11 telecommuters and 21 non-
telecommuters, as identified using the list of telecommuters provided by the
program manager. Exploratory principal component unrotated factor analysis
revealed a two factor solution. Factor one included all items except item 9
and it accounted for 70.53% of the variance in commitment to
telecommuting. Factor two included item 9 "I would accept almost any type
27


of job assignment in order to keep telecommuting as an option for getting
work done" and accounted for an additional 9.32% variance in commitment
to telecommuting. Examining factor one, reliability analysis for the iO-item
scale showed Chronbach's standardized alpha (a) = .96. The decision was
made that factor one was commitment to telecommuting.
Factor analysis on data from participants included in the survey
revealed a one factor using a principal axis solution which explained 69.82%
of the variance in commitment to telecommuting (a = .96). Sample items
include "I value telecommuting as a work arrangement" and "I am willing to
put in a great deal of effort beyond that normally expected in order to
support my telecommuting teammates".
Mutual Trust
Mutual trust was measured by five items (Smith-Jentsch, Kraiger,
Cannon-Bowers, & Salas, 2002) using a 6-point Likert scale where 1 =
strongly disagree and 6 = strongly agree. Factor analysis results using a
principal axis oblimin rotation revealed a two factor solution with factor one
explaining 43.49% of the variance in mutual trust and factor two explaining
an additional 12.40% of the variance in mutual trust. Reliability analysis on
factor one (items 1,2, and 5) showed a =. .71 and for factor 2 (items 3 and 4)
28


a = .75. Reliability analysis on all five items a = .76, but removing item 5
showed a = .77. The researcher made a decision to include items 1 4 to
measure mutual trust based on reliability analysis results. Sample items
included "My teammates work out their differences in an honest manner" and
"I believe that things I say in confidence to my teammates will not be
repeated to others or used against me".
Team Cohesion
Team cohesion is the active participation and commitment driving the
desire to stay and freely interact on a team (Fiore, Salas, & Bowers, in press).
An eight-item 6-point Likert scale was used to measure team cohesion where
1 = strongly disagree and 6 = strongly agree.
Factor analysis using a principal axis oblimin rotation revealed a two
factor solution where the two reverse scored items (1 and 5) loaded onto
factor two. Factor one explained 51.52% of the variance in team cohesion
and factor two accounted for an additional 10.07% of the variance in team
cohesion. Reliability analysis showed a = .82 for factor one and a = .55 for
factor two. Given that only the reverse scored items loaded onto the second
factor and that the reliability was lower, survey respondents may have been
29


confused with the item wording. The researcher focused on factor one to
measure team cohesion.
Reliability analysis on the remaining six items was conducted, showing
alpha increased from .62 to .83 when item 8 was deleted. The remaining five
items loaded onto one factor and explained 50.46% of the variance in team
cohesion, as revealed through a principal axis factor analysis solution.
Sample items included "I get along well with others of this team" and "My
team is a close team".
Teamwork Behaviors
Teamwork behaviors were measured by utilizing the work on team
behaviors of McIntyre and Salas (1994), which focused on feedback,
communication, and backing up fellow teammates. Eight items were
measured on a 6-point Likert scale where 1 = strongly disagree and 6 =
strongly agree.
Principal axis factor analysis results revealed one factor, which
accounted for 51.52% of the variance in team behavior. Reliability analysis
showed that when item 8 was removed, alpha increased from .87 to .89.
Only items 1-7 were used to measure teamwork behavior. Samples of items
30


included "Members of my team monitor one another's performance" and
"Members of my team are prepared to back up fellow team members.
Innovation
The West and Anderson (1996) five-item innovation scale was used to
measure the perceived level of innovation. Respondents were asked to
compare innovation of their team with other similar teams on various
activities such as "Setting work targets or objectives" and "Developing
innovative ways of accomplishing targets/objectives". On the 1-5 scale, 1
and 2 indicated highly stable: few changes introduced, 3 indicated highly
innovative: some changes introduced, and 4 and 5 indicated highly
innovative: many changes introduced.
Principal axis factor analysis results revealed a one factor solution that
accounted for 70.84% of the variance in innovation. Reliability analysis
showed a = .92.
Task Interdependence
Four items were used to measure task interdependence. One item
was taken from each for the three scales: Pearce and Gregersen (1991),
31


Campion et al. (1993), and Kiggundu (1981). One item was also taken from
Bishop and Scott (2000).
Principal axis factor analysis results revealed a one factor solution that
explained 50.15% of the variance in task interdependence. Reliability
analysis showed alpha increased from .78 to .81 when item 1 was deleted.
As a result, the three remaining items were used to measure task
interdependence. Sample items included "Jobs performed by team members
are related to one another" and "To achieve high performance it is important
to rely on each other".
Team Composition
Utilizing the list of telecommuters and organizational charts, the
researcher was able to compute the percentage of telecommuters on each
team ((number of telecommuters/total number of team members)* 100).
Surveys returned by five teams, however, indicated a higher number of
telecommuters than indicated on the list of telecommuters provided by the
telecommuting program manager. For example, the list indicated that team
156 had 2 telecommuters, but 3 members of team 156 returned surveys
indicating they telecommute. In these situations, the researcher utilized the
higher number of telecommuters to compute the percentage of
32


telecommuters on that team because it is feasible that additional employees
became telecommuters after the list was obtained from the program
manager.
Demographic Variables
All demographic variables were measured by self-report.
Tenure with the Organization. Tenure categories included less than 6
months, 6 to 12 months, 1 to 2 years, 2 to 5 years, 5 to 8 years, and more
than 8 years.
Tenure as a Telecommuter. Response categories included not
applicable, less than 6 months, 6 to 12 months, 1 to 2 years, and more than
2 years.
Other Variables. Gender, race, education were included. An item
asking if the respondent telecommutes was included and an item asking if
any of the person's teammates telecommutes was included.
Aggregation of Indices
There are several procedures used to justify aggregation of individual
data to the group or team level. First, intraclass correlation coefficients were
calculated. Both ICC(l) and ICC(2) were calculated to assess within-group
33


homogeneity (Bartko, 1976; James, 1982). ICC(l) is interpreted as the
proportion of total variance accounted for by group membership (Bryk &.
Raudenbush, 1982). ICC(2) estimates group mean reliability (Bartko, 1976;
James, 1982; McGraw &. Wong, 1996; Shrout 8i Fleiss, 1979). To calculate
ICC(l) and ICC(2) statistics, a series of one-way ANOVAs were conducted
where each variable of interest (the scale means) was the dependent variable
and team membership was the independent variable. Mean square between
and mean square within values from ANOVA results were used in the
following formula: ICC(l) = (MSB-MSW)/(MSB+((k-l)*MSW)) and ICC(2) =
(MSB MSW)/MSB. Since team size varied between 3 and 9, the Blalock
(1972) formula to calculate the average group size and used in place of k,
group size. Generally, significant F-tests indicate support for aggregation
using the ICC(l) statistic. In the current study, p < .10 for only commitment
to telecommuting, task interdependence, and innovation. ICC(2) values less
than .50 are considered poor support for aggregation. The maximum ICC(2)
value for the current study was .39. Results from ICC analyses appear in
Table 3.1
34


Table 3.1
Within Team Agreement and Intraclass Correlation Coefficients
Scale Twq ICC(l) ICC(2)
Commitment to Telecommuting 1.01 .08 F=1.34, p=.09 .25 F=1.34,p=.09
Mutual Trust .92 .05 F=1.22, p=.18 .18 F=1.22, p=.18
Team Cohesion .88 -.01 F=.95, p=.57 -.05 F=.95, p=.57
Teamwork Behavior .92 .03 F=l.ll, p=.30 .10 F=l.ll, p=.30
Task Interdependence .84 .08 F=1.35, p=.08 .26 F=1.35, p=,08
Innovation .66 .14 F=1.65, p=.01 .39 F=1.65, p=.01
Average .87 .06 .19
Different techniques to support aggregation can result in different
conclusions, but if within group agreement is considered high enough then a
researcher will conclude that aggregation to the group level is appropriate
using only rwg (Klein, et al., 2000). The rwg statistic was calculated to assess
within-group agreement by comparing observed group variance to expected
random variance by calculating (James, Demaree, & Wolf, 1984). The rwg
35


value for each team was calculated to assess within-team agreement for each
variable of interest. One team was deleted from the analysis because rwg = -
17.50 for innovation. Removing this team from the analysis increased the rwg
for innovation from .32 to .66. Although removing cases with rwg values
below .70 is debatable (Klein, Bliese, Kozlowski, Dansereau, Gavin, Griffin,
Hofmann, James, Yammarino, & Bligh, 2000), the researcher made the
decision to do so in this study. The average rwg for the 52 teams was .87 and
the range was .66 to 1.01. Generally, average rwg exceeding .70 are
interpreted as support for aggregation (Ibid). Results from the within team
agreement analysis appear in Table 3.1.
It is possible to have high within group agreement, but low reliability
demonstrated by low ICC values. This situation can occur if respondents
tended to restrict their range of item responses (Bliese, 2000). Consequently,
the rwg statistic was used to assess within-group agreement to support
aggregation at the team level using 52 teams.
36


CHAPTER 4
RESULTS
DescriPtives
There were 216 individuals on 52 teams included in the study. The
majority of respondents were non-telecommuters (61%). Most of the
respondents worked with telecommuting teammates (83%). Of those who
responded, 54% were male and 84% were white. The largest portion (48%)
of individuals have been employed at the organization between 2 and 5 years
and 48% also have a 4-year college degree. For those that telecommute, the
largest percentage (22%) have telecommuted for more than 2 years.
Hypothesis Testing
Zero-order correlations, descriptive statistics, and reliability coefficients
for the predictor and criterion measures examined at the team level are
presented in Table 4.1. Hypothesis 1 suggested that the percentage of
telecommuters on a team would be positively related to commitment to
telecommuting. This hypothesis was tested by a zero-order correlation
between the two variables. A significant correlation was found (r = .42, p <
37


.01), providing support for the hypothesis. Hypothesis 2 suggested that
commitment to telecommuting would be positively related to innovation. This
hypothesis was tested by a zero-order correlation between the variables. A
significant correlation was found (r = .33, p < .05), providing support for the
hypothesis.
Table 4.1
Means. Standard Deviations, and Zero Order Correlations Team Level
Analysis flM = 521
Variable M SD 1 2 3 4 5 6 7
1. Commitment to Telecommuting 4.72 .63 _ .56** .43** .34* .16 .33* 42**
2. Mutual Trust 4.57 .44 .55** .57** .20 .30* .33*
3. Team Cohesion 4.88 .37 .57** 49** .46** .08
4. Teamwork 4.51 .39 .38** .35** .17
Behavior 5. Task 4.94 .50 47** -.18
Interdependence 6. Innovation 3.33 .54 .33**
7. Percentage of 25.5718.98
Telecommuters
** Indicates significance at the 0.01 level (2-tailed).
* Indicates significance at the 0.05 level (2-tailed).
Hypothesis 3 suggested that the relationship strength between
commitment to telecommuting and jnnovation would be moderated by the
level of trust. The Baron and Kenny (1986) method of conducting a series of
38


regression equations was used to test the hypothesis. First, the interaction
term XZ was computed as the product of the predictor X (commitment to
telecommuting) and the moderator Z (trust). Moderator effects were
indicated by the significant effect of the XZ while X and Z are controlled in a
hierarchical regression. Table 4.2 shows R-squared was significant when the
moderator was entered into the model (F(l,48) = 7.85, p < .01), providing
support for Flypothesis 3.
Table 4.2
Hierarchical Regression of Commitment to Telecommuting, Mutual Trust, and
the Product of Commitment to Telecommuting and Trust on Innovation
Step Variable Added R- Squared A R- Squared F (A)
Step 1 Commitment to Telecommuting .11 .11 6.30*
Step 2 Mutual Trust .13 .02 1.09
Step 3 Commitment to Telecommuting Mutual Trust Product .25 .12 7.85**
** Indicates significance at the 0.01 level.
* Indicates significance at the 0.05 level.
Hypothesis 4 suggested that the relationship strength between
commitment to telecommuting and innovation would be moderated by the
level of cohesion. The method suggested by Baron and Kenny (1986) was
also used to test this hypothesis. First, the interaction term XZ was computed
39


as the product of the predictor X (commitment to telecommuting) and the
moderator Z (cohesion). Moderator effects were indicated by the significant
effect of the XZ while X and Z are controlled in a hierarchical regression.
Table 4.3 shows R-squared was significant when the moderator was entered
into the model (F (1,48) = 4.03, p = .05). The multicollinearity between
commitment to telecommuting and the product of commitment to
telecommuting and team cohesion (r=.93, p < .001) posed a problem since
high multicollinearity can lead to technical issues in estimating the regression
coefficients (Aiken & West, 1991). Centering predictor variables, and thus the
interaction term, often helps to minimize these issues (Neter, Wasserman, &
Kutner, 1989).
40


Table 4.3
i\v,v|iwjivii vi wiiiiimum^.h. w f .-ww and the Product of Commitment to Telecommutina and Team Cohesion on
Innovation
Step Variable Added R- Squared AR- Squared F (A)
Step 1 Commitment to Telecommuting .11 .11 6.30*
Step 2 Team Cohesion .23 .12 7.75**
Step 3 Commitment to Telecommuting Team Cohesion Product .29 .06 4.03*
** Indicates significance at the 0.01 level.
* Indicates significance at the 0.05 level.
Commitment to telecommuting and team cohesion were centered by
computing new variables equal to the original value minus the mean for that
original value. A new interaction term was then computed using centered
commitment to telecommuting and centered team cohesion. Table 4.4 shows
no multicollinearity between the three predictor variables.
41


Table 4.4
Zero Order Correlations Centered Data
Variable 1 2 3
1. Centered Commitment to Telecommuting - . .43 -.12
2. Centered Team Cohesion -.15
3. Product of Centered Commitment to Telecommuting and Centered Team Cohesion
A hierarchical regression was conducted using the same procedure
outlined above for the non-centered variables. Table 4.5 shows a significant
R-squared (p < .05) when the interaction term was entered into the
regression equation, providing support for Hypothesis 4.
Table 4.5
Hierarchical Regression of Centered Data
Step Variable Added R- Squared AR- Squared F (A)
Step 1 Commitment to Telecommuting (t) .11 .11 6.30*
Step 2 Team Cohesion (c) .23 .12 7.75**
Step 3 Commitment to Telecommuting Team Cohesion Product (tc) .29 .06 4.03**
** Indicates significance at the 0.01 level.
* Indicates significance at the 0.05 level.
42


A simple regression equation was created using centered data. Team
cohesion values for ZcM, and ZcH, and ZcL were entered into the equation to
create a series of simple regression equations.
Table 4.6
Simple Regression'Equation for Centered Data Regression of Zy on Zt at
Particular Values of Zc
In general: Predicted Zy = (.15 .25 Zc) Zt + .37 Zc
At Zch = .37: Predicted Zy = .06 Zt + .13
At Zcm = 0.00: Predicted Zy = .15 Zt
At ZcL = -.37:________Predicted Zy = .24 Zt .13 _______________
Note, y is the predicted value for innovation, t is commitment to
telecommuting, and c is team cohesion.
Hypothesis 5 suggested that teamwork behaviors would mediate the
relationship between commitment to telecommuting and innovation. This
hypothesis was tested using the Baron and Kenny (1986) method of
conducting a series of regression analyses to test for a mediating variable
between a predictor and criterion variable. The first regression equation was
to regress the mediator (team behavior) on the predictor (commitment to
43


telecommuting). The second equation was to regress the criterion
(innovation) on the predictor (commitment to telecommuting). The third
equation was to regress the criterion (innovation) on both the predictor
(commitment to telecommuting) and the mediator (teamwork behavior).
According to Baron and Kenny, there is mediation if all of the following are
true: The predictor affects the mediator in equation 1 (R-Squared = .12, F
(1.50) = 6.59, p = .01, beta = .34), the predictor affects the criterion in
equation 2 (R-Square = .11, F (1,50) = 6.30, p < .02, beta = .33), and the
mediator affects the criterion in equation 3 (beta = .26, p < .07). Since all
three held true in the predicted direction, the effect of the predictor on the
criterion was less in Equation 3 (beta = .26) than in Equation 2 (beta = .33),
providing Support for Hypothesis 5.
Exploratory Analysis
The Baron and Kenny (1986) strategy was also used to test if task
interdependence was a mediator of the relationship between commitment to
telecommuting and innovation. Results indicated that commitment to
telecommuting did not affect task interdependence (R-Squared = .03, F
(1.50) = 1.23, p = .27, beta = .18). Commitment to telecommuting affected
innovation (R-Square = .11, F (1,50) = 6.30, p < .02, beta = .33). Finally,
44


task interdependence affected innovation (beta = .43, p < .01). Although all
three did not hold true, the effect of commitment to telecommuting on
innovation was less in Equation 3 (beta = .27) than in Equation 2 (beta =
.33).
Based on this analysis, task interdependence did not mediate the
relationship between commitment to telecommuting and innovation. There
was, however, a main effect on innovation. Since task interdependence was
not related to commitment to telecommuting, it could not be a mediator.
45


CHAPTER 5
DISCUSSION
Summary of Main Points
The intent of this study was to provide additional insight into the
relationship between telecommuting and team effectiveness. Teams
containing at least one telecommuter were identified and surveyed at a global
software development company. Using the rwg statistic, individual data were
aggregated into 52 teams to analyze the interrelationship of several variables.
The relationship between team composition and commitment to
telecommuting was analyzed. Mutual trust and cohesion were analyzed as
moderators of the relationship between commitment to telecommuting and
innovation. Teamwork behaviors and task interdependence were examined
as potential mediators of the relationship between commitment to
telecommuting and innovation.
Results showed a significant moderate relationship between team
composition, as measured by the percentage of telecommuters, and
commitment to telecommuting. There was also a significant moderate
relationship between commitment to telecommuting and innovation. The
46


relationship between commitment to telecommuting and innovation was
moderated by both mutual trust and team cohesion. Teamwork behaviors
mediated the relationship between commitment to telecommuting and
innovation, however task interdependence did not mediate the relationship.
Contributions of the Current Study
Telecommuting is an area of organizational research that has not been
studied to a great extent at this point. It is a relatively new phenomenon
with the evolution of technology, particularly in the past ten years. There has
been, however, much research focused on teams. This study not only
examined a new area of organizational research, but offered a unique
approach in the examination process by leveraging existing research of
traditional office-based teams. Through this study, we recognize the
importance of trust and cohesion on teams with telecommuters. The
commitment to how work is accomplished, specifically the commitment to
telecommuting, relates to how innovative team members perceive themselves
to be and thus, how effective they are as a team. Teamwork behaviors on
telecommuting teams also relate to team effectiveness.
A team's commitment level to telecommuting is a new construct developed
through this study that can be used by other researchers in the future.
47


There are also practical implications resulting from this study. While
telecommuting is feasible for some people, it may not be feasible for
everyone. In reality, there will probably be teams containing a mixture of
traditional office-based employees and telecommuting employees in many
organizations. From this standpoint, the current study provided relevant
results that those organizations might utilize. Commitment to telecommuting
is an important construct for a team as a unit, not just for those individuals
that telecommute. Organizations might look educate all team members on
the benefits of telecommuting in order to raise the commitment level on the
team. Trust, cohesion, and teamwork behaviors relate to the effectiveness of
telecommuting teams. Organizations might also train managers on how to
foster these attitudes and behaviors among members of their teams.
Limitations and Directions for Future Research
The current study contributed to the understanding of telecommuting
and teams, but there were limitations of the study. Although in team-level
research a sample size of 25 teams may be sufficient (McIntyre & Salas,
1992), the sample size of 52 teams in this study could still be viewed as a
limitation. Data were collected from only one organization, posing
generalizability issues to other organizations and jobs. A third limitation was
48


that all measures were self-reported as opposed to using a more objective
measurement of the variables.
Telecommuting and its impact on teams is an area of research that
requires further exploration. Researchers should gather data from more
teams within multiple organizations across different industries. The current
study examined a just a few of the numerous team attitudes, behaviors and
measures of team effectiveness that have been studied in previous research
Future could investigate other team attitudes such as team efficacy, other
behaviors such electronic communication, and alternative measures of team
effectiveness such as project delivery metrics. Many opportunities exist for
organizational researchers to explore telecommuting and teams further.
49


APPENDIX
APPENDIX A
TELECOMMUTING INTERVIEW FORM'
Person Interviewed:_______________ Date:___________________________
Organization:_____________________ Interviewer:______________________
1. What is your current title and what are your current responsibilities
relative to telecommuting?
2. How long has your organization offered the telecommuting to
employees?
3. Please describe your telecommuting program.
Formal or informal?
How do you define "telecommuter'?
How many telecommuters does your organization have?
At what rate does the number of telecommuters increase or decrease
per year?
What is your projected growth or decline rate of the number of
telecommuters?
4. Through your own observation and information gathered from those
affected by the program, what do you see as the biggest benefits for:
Telecommuting employees
Managers of telecommuters
The organization
Teams
50


5. Through your own observation and information gathered from those
affected by the program, what do you see as the biggest challenges or
issues for:
Telecommuting employees
Managers of telecommuters
The organization
Teams
6. Have you made any changes to the program since its inception and if so,
what were the reasons for the changes?
7. Do you plan to make any changes in the program and if so, why?
8. Is there anything else you would like to share with me about the
program or about telecommuting in general?
9. Are there any specific topics or aspects of telecommuting that you would
like me to include on my survey to telecommuters and their non-
telecommuting teammates, if possible?
51


APPENDIX B
COMMITMENT TO TELECOMMUTING SURVEY
Please read each statement and indicate your agreement level by replacing
the appropriate number with an "X". Please return to
reina bach@yahoo.com. Feel free to contact Reina at 303-475-3185 if you
have any questions.
For this survey, "telecommuting" is defined as working from home at least
one day per week. Thank you for taking the time to complete
this survey!
Question Strongly Disaaree Strongly Aaree
1. I talk up telecommuting to my friends as a great work arrangement. 1 2 3 4 5 6
2. I am proud to tell others about telecommutinq. 1 2 3 4 5 6
3. I am extremely glad that telecommuting is available as an option for getting work done. 1 2 3 4 5 6
4. Iam willing to put in a great deal of effort beyond that normally expected in order for telecommuting to be successful. 1 2 3 4 5 6
5. I value telecommuting as a work arrangement. 1 2 3 4 5 6
52


Question Strongly Disagree Strongly Agree
6. The option to telecommute really inspires the very best in me in the way of job performance. 1 2 3 4 5 6
7. I really care about the fate of telecommuting as a work option. 1 2 3 4 5 6
8. For me telecommuting is the best possible way to qet work done. 1 2 3 4 5 6
9. I would accept almost any type of job assignment in order to keep telecommuting as an option for getting work done 1 2 3 4 5 6
10. I plan to telecommute in the future. 1 2 3 4 5 6
11. Iam willing to put in a great deal of effort beyond that normally expected in order to support my teammates. 1 2 3 4 5 6
53


APPENDIX C
PARTICIPANT TELECOMMUTING SURVEY
Instructions
In an effort to improve our telecommuting (teleworking) program, we are
asking employees to complete this brief survey.
Please forward the completed survey via email back to Reina Bach. Should
you have any questions, please contact Reina at 303-475-3185.
Please note that for this survey, "team" refers to all people reporting to the
same manager that you report to.
We value your input and appreciate your time. Thank you!
1. On average, I work from home at least one day per week. (TELECOMM)
____Yes _______No
2. Excluding myself, at least one member of my team works from home at
least one day per week. (TEAMTELE) _____Yes ______No
Please read each statement below and indicate your agreement level by
replacing the appropriate number with an "X".
Question Strongly Disagree Strongly Agree
3.1 talk up telecommuting to my 1 2 3 4 5 6
54


Question Strongly Disagree Strongly Agree
friends as a great work arrangement. (COMMUl)
4.1 am proud to tell others about telecommuting. (COMMIT2) 1 2 3 4 5 6
5.1 am extremely glad that telecommuting is available as an option for getting work done. CCOMMIT3) 1 2 3 4 5 6
6.1 am willing to put in a great deal of effort beyond that normally expected in order for telecommuting to be successful. (C0MMIT4) 1 2 3 4 5. 6
7.1 value telecommuting as a work arrangement. (COMM1T5) 1 2 3 4 5 6
8. The option to telecommute really inspires the very best in me in the way of job performance. (COMM]T6j 1 2 3 4 5 6
9.1 really care about the fate of telecommuting as a work option. (C0MMIT7) 1 2 3 4 5 6
10. For me telecommuting is the best possible way to get work done. (COMMIT8) 1 2 3 4 5 6
11.1 plan to telecommute in the 1 2 3 4 5 6
55


Question i Strongly Disagree Strongly Agree
future. (COMMITS
12.1 am willing to put in a great deal of effort beyond that normally expected in order to support my telecommuting teammates. (COMMITIO) 1 2 3 4 5 6
13. My teammates work out their differences in an honest manner. (TRUST1) 1 2 3 4 5 6
14. I believe that things I say in confidence to my teammates will not be repeated to others or used against me. (TRUST2) 1 2 3 4 5 6
15.1 feel comfortable letting my teammates coordinate for me. (TRUST3) 1 2 3 4 5 6
16.1 feel comfortable coordinating special requests for my teammates without needing to immediately know why. (TRUST4) 1 2 3 4 5 6
17. Some of my teammates have been known to speak negatively about one another to others outside the team. (TRUSTS reverse scored) 1 2 3 4 5 6
56


Question Strongly Disagree Strongly Agree
18. If given the chance, I would prefer not to work with this team. (COHESN1 reverse scored) 1 2 3 4 5 6
19.1 get along well with others of this team. (COHESN2) 1 2 3 4 5 6
20.1 believe the members of this team would readily defend me from criticism by others outside the team. (COHESN3) 1 2 3 4 5 6
21.1 feel that I am really part of my team. (COHESN4) 1 2 3 4 5 6
221 find that I generally do not get along with the other members of my team. (COHESN5 reverse scored) 1 2 3 4 5 6
23.1 feel like the members of my team are my friends. (COHESN6) 1 2 3 4 5 6
24. My team is a close team. (COHESN7) 1 2 3 4 5 6
25.1 socialize with members of my team outside of work. (COHESN8) 1 2 3 4 5 6
26. Members of my team monitor one another's performance. (BEHAV1) 1 2 3 4 5 6
27. Members of my team 1 2 3 4 5 6
57


Question Strongly Disagree Strongly Agree
provide feedback to one another. (BEHAV2)
28. Members of my team accept feedback from one another. CBEHAV3") 1 2 3 4 5 6
29. Communication between members of my team is effective. (BEHAV4) 1 2 3 4 5 6
30. When members of my team communicate, the following occurs: the sender initiates the message, the receiver accepts the message and provides feedback to indicate that the message was received, and the sender double-checks to ensure that the intended message was received. (BEHAV5) 1 2 3 4 5 6
31. Members of my team are willing to back up fellow team members. (BEHAV6) 1 2 3 4 5 6
32. Members of my team are prepared to back up fellow team members. (BEHAV7) 1 2 3 4 5 6
33. Members of my team are inclined to back up fellow team members. (BEHAV8) 1 2 3 4 5 6
34.1 frequently must 1 2 3 4 5 6
58


Question Strongly Disagree Strongly Agree
coordinate my efforts with others. (TASKINT1)
35. Jobs performed by team members are related to one another. (TASKINT2) 1 2 3 4 5 6
36. For the team to perform well members must communicate well. (TASKINT3) 1 2 3 4 5 6
37. To achieve high performance it is important to rely on each other. (TASKINT4) 1 2 3 4 5 6
38. My morale is high. (MORALE) 1 2 3 4 5 6
39.1 am satisfied with my iob. (JOBSAT) 1 2 3 4 5 6
40. My request to work from home at least one day per week was denied by
my manager. (REQDENY)_______Yes ____No _____Not Applicable
If you answered yes to question #40, please answer questions #41 and
#42. Otherwise, please skip to question #43.
Strongly Strongly
Disagree____________________________ Agree
41.1 was satisfied with how my manager told me I could not work from home at least one day per week. (DENYSAT) 1 2 3 4 5
59


42. If you answered YES to question #40, please provide any additional
comments about how your manager communicated his/her decision not to
approve your request.
43. If you work at least one day from home per week, please take a
moment to describe your typical day:
(a) At Home
(b) At The Office
44. Gender: (GENDER) ______Female ________Male
45. Race: (RACE) ______Hispanic _____Pacific Islander _____White
____Black _______Native American _______Other
46. Number of years employed by this organization: (ORGTENUR)
___Less Than 6 Months _____6-12 Months ,___i_l-2 Years ___2-5
Years ____5-8 Years ______ More Than 8 Years
47. Highest level of education attained: (EDUC)
____High School _____Some College _____4-year College Degree
____Some Graduate _______Graduate Degree
48. Length of time I have worked at least one day per week from home with
this organization: (TELETEN)
____Not Applicable ____Less Than 6 Months ______6-12 Months
1-2 Years More Than 2 Years
Compared with other similar teams how innovative do you consider your team to be? Please indicate your response by replacing the appropriate number with an "X".
highly stable: few changes introduced highly innovative: some changes introduced highly innovative: many changes introduced
49. Setting work targets or objectives. (INNOV1) 1 2 3 4 5
60


Compared with other similar teams how innovative do you consider your team to be? Please indicate your response by replacing the appropriate number with an "X".
highly stable: few changes introduced highly innovative: some changes introduced highly innovative: many changes introduced
50. Deciding the methods used to achieve objectives/targets. CINNOV2) 1 2 3 4 5
51. Initiating new procedures or information systems. (INNOV3) 1 2 3 4 5
52. Developing innovative ways of accomplishing targets/objectives. (INNOV4) 1 2 3 4 5
53. Initiating changes in the job content and work methods of your staff. (INNOV5j 1 2 3 4 5
61


BIBLIOGRAPHY
Aiken, L.S., & West, S.G. (1991). Multiple regression: Testing and interpreting
interactions. Thousand Oaks, CA: Sage Publications.
Amabile, T.M., Conti, R., Coon, H., Lazenby, & J., Herron, M. (1996).
Assessing the work environment for creativity. Academy of
Management Journal 39.1154-1184.
Axelrod, E.L., Handfield-Jones, & H., Welsh, T.A. (2001). The McKinsev
Quarterly. [On-linel. Available: www.mckinsev.com
Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychology,
51* 1173-1182.
Bartko, JJ. (1976). On various intraclass correlation reliability coefficients.
Psychological Bulletin. 83. 762-765.
Bettenhausen, K.L. (1991). Five years of group research: What have we
learned and what needs to be addressed? Journal of Management. 17.
345-381.
Bishop, J.W., Scott, K.D., & Casino, L.S. (1997, August). The differential
effects of team commitment and organizational commitment on job
performance and intention to quit. (Paper presented at the annual
meeting of the Academy of Management, Boston).
Bishop, J.W., & Scott, K.D. (2000). An examination of organizational and
team commitment in a self-directed team environment. Journal of
Applied Psychology. 85. 439-450.
62


Blalock, H.M., Jr. (1972). Social Statistics (2nd ed.). New York: McGraw-Hill.
Bliese, P.D. (2000). Within-group agreement, non-independence, and
reliability: Implications for data aggregation. In K. Klein &
S.WJ. Kozlowski (Eds.), Multilevel theory, research, and methods
in organizations: Foundations, extensions, and new directions
(pp. 349-381). San Francisco: Jossey-Bass.
Bryk, A.S., & Raudenbush, S.W. (1992). Hierarchical linear models. Thousand
Oaks, CA: Sage.
Buessing, A., & Broome, P. (1999). Trust under telework. Zeitschrift Fuer
Arbeits- und Oraanisationspsvcholoaie. 43.122-133.
Buhler, P. (Nov. 1995). Creating flexibility in today's workplace. Supervision.
56,24-26.
Byrne, J.A. (2000, Aug. 28). Management by web. Businessweek Online [On-
line]. Available: www.businessweek.com
Campion, M.A., Medsker, G.J., & Higgs, A.C. (1993). Relations between work
group characteristics and effectiveness: Implications for designing
effective work groups. Personnel Psychology. 46. 823-850.
Cannon-Bowers, J.A., Salas, E., & Converse, S.A. (1993). Shared mental
models in expert team decision making. In N J. Castellan, Jr. (Ed.),
Current issues in individual and group decision making (pp. 221-246).
Hillsdale, NJ: Erlbaum.
Conlin, M., Coy, P., Palmer, A.T., & Saveri, G. (1999, Dec. 6). The wild new
workforce. Businessweek Online [On-line]. Available:
www.businessweek.com
Cree, L.H. (1998). Work/familv balance of telecommuters. Unpublished
doctoral dissertation, The California School of Professional Psychology,
San Diego.
63


Cree, L.H., & Sorenson, R.C. (2001). What happens when employees
telecommute? some research findings. Performance Management
Library. [On-line]. Available: www.Derformaworks.com
Currall, S.C., & Towler, A.J. (in press). Research methods in management and
organizational research: toward integration of qualitative and
quantitative techniques. 1-21.
Dirks, K.T. (1999). The effects on interpersonal trust on work group
performance. Journal of Applied Psychology. 84. 445-455.
Driskell, J.E., & Salas, E. (1992). Can you study real teams in contrived
settings? The value of small-group research to understanding teams.
In R.W. Swezey & E. Salas (Eds.), Teams: their training and
performance (pp. 101-124). Norwood, NJ: Ablex.
Duxbury, L., Higgins, C., & Neufeld, D. (1998). Telework and the balance
between work and family: Is telework part of the problem or part of
the solution? In M.K. Khosrowpur, J. Travers & B. Klinder (Eds.), The
virtual workplace (pp. 218-250). Hershey, PA: Idea Group Publishing.
Employment Policy Foundation Newswatch. (2001, May 9). EU, unions agree
on voluntary rules to protect telecommuters in retail, trade [4
paragraphs]. EPF Newswatch [On-line]. Available E-mail: info@epf.org
Employment Policy Foundation Newswatch. (2001, June 7). Employers see
advantages to telecommuting [3 paragraphs]. EPF Newswatch
[On-line]. Available E-mail: info@epf.org
Employment Policy Foundation Newswatch. (2001, Aug. 23). The American
workplace report 2001 to be released Aug. 29 [7 paragraphs]. EPF
Newswatch [On-line]. Available E-mail: info@epf.oro
Employment Policy Foundation Newswatch. (2001, Aug. 23). Telecommuting
tax credit program started in five cities [2 paragraphs]. EPF
Newswatch [On-line]. Available E-mail: info@epf.oro
64


Employment Policy Foundation Newswatch. (2001, Aug. 23). 2000 census
shows longer U.S. commutes [2 paragraphs]. EPF Newswatch
[On-line]. Available E-mail: info@epf.org
Employment Policy Foundation Newswatch. (2001, Nov. 9). Home offices are
in [1 paragraph]. EPF Newswatch [On-line]. Available E-mail:
info@epf.ora
Evans, C.R., & Dion, K.L. (1991). Group cohesion and performance: A meta-
analysis. Small Group Research. 22.175-186.
Festinger, L. (1950). Informal social communication. Psychological Review,
52x 271-282.
Fiore, S.M., Salas, E. & Bowers, C.A, (in press). Distributed coordination
space: Toward a theory of distributed team performance. 1-38.
Fireman, S.M. (1999). A model of telecommuting withdrawal: Employee
perceptions predicting the reduction or stopping of telework.
Unpublished doctoral dissertation, University of Washington.
Gurstein, P. (1991). Working at home and living at home: emerging
scenarios. Journal of Architecture and Planning Research. 8.165-180.
Guzzo, R.A., & Shea, G.P. (1992). Group performance and intergroup
relations in organizations. In M.S. Dunnette & L.M. Hough (Eds.),
Handbook of industrial and organizational psychology (Vol. 3,
pp. 269-313). Palo Alto, CA: Consulting Psychologists Press.
Hackman, J.R. (1986). The psychology of self-management in organizations.
In M.S. Pallak & R. Perloff (Eds.). Psychology and work (pp. 89-136).
Washington, DC: American Psychological Association.
Hackman, J.R. (1987). The design of work teams. In J.W. Lorsch (Ed.),
Handbook of organizational behavior (pp. 315-342). Englewood Cliffs,
NJ: Prentice Hall.
65


Hartman, R.I., Stoner, C.R., & Arora, R. (1991). An investigation of selected
variables affecting telecommuting productivity and satisfaction. Journal
of Business & Psychology. 62. 207-225.
Hill, E.J., Miller, B.C., Weiner, S.P., &Colihan, J. (1998). Influences of the
virtual office on aspects of work and work/life balance. Personnel
Psychology. 51. 667-681.
Huws, U., Korte, W.B., & Robinson, E. (1990). Telework: towards the elusive
office. (R. Boland & R. Hirschheim). Chichester, England: John Wiley &
Sons.
International Telework Association & Council Telework America 2000 Study.
(2001). [On-line]. Available:
www.telecommute.ora/twa2000/research results summarv.shtml
Jackson, S.E. (1999). Flexibility in work arrangements: how availability.
preferences, and use affect business outcomes. Unpublished doctoral
dissertation, Rutgers, The State University of New Jersey, New
Brunswick.
James, L.R. (1982). Aggregation bias in estimates of perceptual agreement.
Journal of Applied Psychology. 67. 219-229.
James, L.R., Demaree, R.J., & Wolf, G. (1984). Estimating within-group
interrater reliability with and without response bias. Journal of Applied
Psychology. 69. 85-98.
Jarvenpaa, S.L., Knoll, K., & Leidner, D.E. (1998). Is anybody out there?:
Antecedents of trust in global virtual teams. Journal of Management
Information Systems. 14. 29-64.
Jones, G.R., & George, J.M. (1998). The experience and evolution of trust:
Implications for cooperation and teamwork. Academy of Management
Review. 23. 531-546.
Kiggundu, M.N. (1983). Task interdependence and job design: Test of a
theory. Organizational Behavior and Human Performance. 31.145-172.
66


Klein, K.J., Bliese, P.D., Kozlowski, S.W.J., Dansereau, F., Gavin, M.B., Griffin,
M.A., Hofmann, D.A., James, L.R., Yammarino, F.Y., & Bligh, M.C.
(2000). Multilevel analytic techniques: Commonalities, differences,
and continuing questions. In K. Klein & S.WJ. Kozlowski (Eds.),
Multilevel .theory, research, and methods in organizations:
Foundations, extensions, and new directions (pp. 512-553).
San Francisco: Jossey-Bass.
Kraiger, K., & Wenzel, L.H. (1997). Conceptual development and empirical
evaluation of measures of shared mental models as indicators of team
effectiveness. In M.T. Brannick, E. Salas, & C. Prince (Eds.), Team
performance assessment and measurement: Theory, methods, and
applications. Mahwah, NJ: Erlbaum.
Lallande, A. (1984). Probing the telecommuting debate. Business Computer:
April.
Li, F. (1998). Team-telework and the new geographical flexibility for
information workers. In M.K. Khosrowpur, J. Travers, 8i B. Klinder
(Eds.), The virtual workplace (pp. 301-315). Hershey, PA: Idea Group
Publishing.
McGraw, K.O., 8i Wong, S.P. (1996). Forming inferences about some
intraclass correlations coefficients. Psychological Methods. 1. 30-46.
McIntyre, R.M., & Salas, E. (1994). Measuring and managing for team
performance: emerging principles from complex environments. In R.A.
Guzzo & E. Salas (Eds.), Team effectiveness and decision making (pp.
9-43). San Francisco: Jossey Bass.
Mowday, R.T., Porter, L.W. & Steers, R.M. (1982). Employee organizational
linkages: The psychology of commitment, absenteeism, and turnover.
New York: Academic Press.
Mullen, B., & Copper, C. (1994). The relation between group cohesiveness
and performance: An integration. Psychological Bulletin. 115 (2), 210-
227.
67


Neter, J., Wasserman, W., & Kutner, M.H. (1989). Applied linear regression
models (2nd ed.). Homewood, IL: Irwin.
Neufeld, DJ. (1997). Individual consequences of telecommuting. Unpublished
doctoral dissertation, The University of Western Ontario, London.
Neuman, G.A., & Wright, J. (1999). Team effectiveness: beyond skills and
cognitive ability. Journal of Applied Psychology. 84. 376-389.
Oliver, L.W. (1990). The relationship of cohesion and performance: A meta-
analvsis and critique of the literature. Unpublished manuscript, U.S.
Army Research Institute.
Olson, M.H., & Primps, S.B. (1984). Working at home with computers: Work
and nonwork issues. Journal of Social Issues. 40. 97-112.
Oser, R., McCallum, G.A., Salas, E., & Morgan, B.B., Jr. (1989). Toward a
definition of teamwork: An analysis of critical team behaviors (NTSC
Report NO. 89-004). Orlando, FL: Naval Training Systems Center.
Pearce, J.L., & Gregersen, H.B. (1991). Task interdependence and extrarole
behavior: A test of the mediating effects of felt responsibility. Journal
of Applied Psychology. 76. 838-844.
Peters, TJ. (1988). Thriving on chaos. New York: Knopf.
Platt, R.G., & Page, D. (2001). Managing the virtual team: Critical skills and
knowledge for successful performance. In M.K. Khosrowpur, J.
Travers, M. Rossi, & N. Tonelli (Eds.), Telecommuting and virtual
offices: Issues & opportunities (pp. 130-145). Hershey, PA: Idea
Group Publishing.
Prince, C., & Salas, E. (1993). Training and research for teamwork in the
military aircrew. In E. Wiener, B. Kanki, & R. Helmreich (Eds.),
Cockpit resource management (pp. 337-366). San Diego, CA:
Academic Press.
68


Reinsch, N.L. Jr., (1999). Selecting communication variables and
telecommuting participation decisions: Data from telecommuting
workers. Journal of Business Communication, 36r 247-260.
Salas, E., Dickinson, T.L., Converse, S.A., & Tannenbaum, S.L. (1992).
Toward an understanding of team performance and traiing. In R.W.
Swezy & E. Salas (Eds.), Teams: Their training and performance
(pp.2-29). Norwood, NJ: Ablex.
Scott, K.D., & Townsend, A.M. (1994, August). Teams: Why some perform
and others do not. HR Magazine. 39. 62-67.
Shrout, P.E., & Fleiss, J.L. (1979). Intraclass correlations: Uses in assessing
rater reliability. Psychological Bulletin. 86. 420-428.
Smith-Jentsch, K.A., Kraiger, K., Cannon-Bowers, J.A., & Salas, E. (2002).
Familiarity breeds teamwork. 1-39.
Speeth, L.M. (1992). The attributes of successful managers of telecommuters
and successful telecommuting programs. Unpublished doctoral
dissertation, Golden Gate University, San Francisco.
Sundstrum, E., de Meuse, K.P., & Futrell, D. (1990). Work teams: Applications
and effectiveness. American Psychologist. 45.120-133.
Taveras, L.E. (1998). The effects of telecommuting on attitudinal and
behavioral characteristics of organizational culture. Unpublished
doctoral dissertation, University of Sarasota, Sarasota.
Wells, S.J. (2001). Making telecommuting work. HR Magazine. 46. 34-45.
Wendell, J. (2001). Federal telework topics. U.S. Department of Labor
Telework and the New Workplace of the 21st Century.
[On-line]. Available: www.dol.gov/dol/asp/public/telework
West, M.A., & Farr, J.L (Eds.). (1990). Innovation and creativity at work:
Psychological and organizational strategies. Chichester, England:
Wiley.
69


Whiting, V.R. (1998). Behavioral outcomes of a change from a traditional
office to a virtual office: A longitudinal analysis. Dissertation Abstracts
International. 58(11A), 4355.
70


The capture range of the correct optimum is the portion of the parameter
space in which an algorithm is more likely to converge to the correct optimum [7].
The greedy algorithm does not work well if the initial transformation vector is
not in the capture range of the correct optimum in that there is divergence from
the optimal solution. A local optimum is eventually attained and the program
terminates. For the following examples, the transformation that results in the
optimal solution (for 0 .5 or integer .5 rotation values for the 256x256 case)
is an rr-shift of 1 pixel, a y-shift of 0 pixels, and a rotation of -10 degrees, denoted
(1 0-10).
Table 4.1 and Figure 4.7 show convergence to the approximate optimum
resulting from a starting transformation of (11 11 10). Table 4.2 and Figure 4.8
show convergence to an incorrect solution from the starting transformation of
(12 12 12). A rough estimate of the capture range for this algorithm can be
derived from Table 4.3 which gives the starting and final transformation vectors
for 22 trials, where there is convergence to the correct transformation, (10-
10), for initial transformations of (-10 -10 -18) and (25 25 11). Therefore, the
capture range is roughly within 35 pixels in the horizontal direction, 36 pixels
in the vertical direction, and 20 degrees of rotation. Unfortunately, the size of
the capture range depends on the features in the images and cannot be known
a priori [7]. However, visual inspection of the registered images can reveal
convergence outside the capture range.
The problem of the patient positioning has to be solved using an algorithm
that is not only sufficiently accurate, but one that is also fast enough to fulfill the
requirements of daily clinical use. For this application, the requirement that the
67


1
DRR
EPI
2
3
4
6
7
8 9 10
11
12 13 14
15
16 17
18
19 20 21
22
23 24
68
Figure 4.7: Greedy algorithm, run 1, 256x256 images.


iteration x-shift y-shift rotation MI
1 11 11 10 0.7638
2 13 9 9 0.7780
3 11 7 8 0.7871
4 9 5 7 0.7969
5 9 3 6 0.8043
6 9 3 5 0.8087
7 9 3 4 0.8149
8 9 3 3 0.8177
9 7 1 2 0.8203
10 5 1 1 0.8253
11 4 1 0.5 0.8311
12 2 1 -0.5 0.8349
13 2 -1 -1.5 0.8419
14 0 -1 -2.5 0.8542
15 -2 1 -3.5 0.8845
16 -4 -1 -4.5 0.9112
17 -4 1 -5.5 0.9466
18 -2 1 -6.5 0.9869
19 -2 1 -7.5 1.0420
20 0 1 -8.5 1.1204
21 0 1 -9.5 1.1916
22 1 0 -10 1.2176
23 0 0 -9.5 1.2021
24 1 0 -10 1.2176
69
Table 4.1: Sample run (Figure 4-7) converges to the optimum transformation
in 8.712 minutes.


DRR
EPI
7
11
12 13 14
&
i
Greedy algorithm, run 2, 256x256 images.
70
Figure 4.8:


iteration x-shift y-shift rotation MI
1 12 12 12 0.7663
2 14 10 13 0.7734
3 14 8 14 0.7821
4 14 G 14 0.7890
5 12 4 14 0.7922
6 10 2 15 0.7982
7 8 2 1G 0.8029
8 8 2 16.5 0.8052
9 8 0 1G.5 0.8058
10 6 -2 1G.5 0.80G7
11 6 -4 17.5 0.8086
12 6 -6 17.5 0.8135
13 7 -6 18 0.8150
14 8 -6 18 0.8152
15 7 -6 18 0.8150
Table 4.2: Sample run (Figure 4-8) converges to a suboptimal solution in
5.0828 minutes.
71


T0{x,y, angle) Tfinai{x,y, angle) MI time(min)
(11 11 10) ( 1 0-10) 1.217601 8.788317
(12 12 10) ( 1 0-10) 1.217601 8.709550
(13 13 10) ( 1 0-10) 1.217601 8.498200
(16 16 10) ( 1 0-10) 1.217601 9.150317
(20 20 10) ( 1 0-10) 1.217601 10.609733
(30 30 10) ( 42 46 7.5) 0.750654 4.397467
(11 11 11) ( 8 -6 18) 0.815214 4.548517
(20 20 11) ( 1 0-10) 1.217601 10.912367
(25 25 11) ( 1 0-10) 1.217601 11.816650
(26 26 11) ( 8-6 18) 0.815214 8.016017
(29 29 11) ( 51 2 15.5) 0.829537 7.151950
(00 00 12) (-14 -9 18) 0.833644 4.187017
(11 11 12) ( 8 -6 18) 0.815214 4.644167
(-10 -10 -19) ( 6 -25 -15.5) 0.883105 5.377883
(-10 -10 -18) ( 1 0-10) 1.217601 4.953967
(-11 -11 -18) ( 6 -25 -15.5) 0.883105 5.203483
(-12 -12 -18) ( 6 -25 -15.5) 0.883105 5.360033
(-13 -13 -18) ( 4 -50 -20.5) 0.916436 9.069867
(-17-17-18) (-11 -33 -26) 0.879511 4.321567
(-18 -18 -18) ( -3 -35 -27) 0.888810 5.214000
(-19 -19 -18) (-23 -44 -19) 0.909333 6.070050
(-20 -20 -18) (-23 -44 -19) 0.909333 5.909833
Table 4.3: Convergence data for 22 vans, 256x256 images.
72


Figure 4.9: 64 x 64 DRR and EPI.
Figure 4.10: 128x128 DR.R and EPI.
73


Figure 4.11: 250x256 DRR and EPI.
initial set of parameters be in a relatively narrow capture range of the optimum
does not seem unreasonable assuming that external markers are used for prelim-
inary patient positioning. However, the speed of convergence can be improved
by finding the optimal transformation for low resolution versions of the images
to be registered. Figures 4.9, 4.10, and 4.11 show the sets of images used in these
examples at resolutions of 64 x 64, 128 x 128, and 256 x 256, respectively. Tables
4.4, 4.5, and 4.6 respectively, show the progression toward convergence for these
sets. Table 4.7 summarizes the data. As expected, it requires the least amount
of time for the lowest, 64 x 64, resolution. Also, the 64 x 64 and 128 x 128 reso-
lutions converge to a slightly different transformation solution, (0 0 -9.5), than
that of the 256 x 256 resolution. However, it is very close and can be used as the
starting transformation for a higher resolution case. From Table 4.6, iterations
11 and 12, it can be seen that this would require only two iterations compared to
74


12, a significant, reduction in computation time. So, a multi-resolution approach,
that is, starting with a series of relatively coarse resolutions with the objective
being to find one or more approximate solutions, or candidate local maxima, as
suitable starting solutions for refinement at progressively higher resolutions, is
one way to speed convergence to the global optimum. Of course, whether or not
the global optimum, or something reasonably close to it, has been attained, has
to be determined, for the type of registration problem described here, by visual
inspection.
The multi-resolution method was implemented using first the 64 x 64 res-
olution, starting with (0 0 0), which converged to the solution (0 0 -9.5). This
solution was then used as the starting transformation vector with the 256 x 256
resolution. The algorithm converged to (1 0 -10) in two iterations with total
time for both resolutions equal to 1.5311 minutes.
The program can be improved for the case of the possibility of a poorly cho-
sen initial starting point. For example, the algorithm can be called repeatedly
while a record of the transformation vectors that have been checked is main-
tained. A new, not previously checked, transformation vector can be created
for a new iteration of the algorithm. The maximum mutual information might
then be retrieved from the stored list of values. This is basically multistart
optimization and was implemented and apparently works well. Using a set of
64 x 64 resolution images, with an initial transformation vector of (-50 50 -50),
and the specification that the greedy algorithm be called 10 times, the program
converged to the expected optimum of (0 0 -9.5) for the 64 x 64 resolution. This
occurred at iteration 63 of 104 total iterations and in approximately 5.7 minutes.
75


iteration x-shift y-shift rot (degrees) MI
1 0 0 0 1.904037
2 -2 0 -1 1.932668
3 -2 0 -2 1.946724
4 -2 0 -3 1.947779
5 -2 0 -4 1.974409
6 0 0 -5 1.981745
7 0 0 -6 2.010966
8 0 0 -7 2.027811
9 0 0 -8 2.059447
10 0 0 -9 2.101459
11 0 0 -9.5 2.110144
Table 4.4: 6'^x 64 run converges in 0.6504 minutes.
4.1.2 Genetic Algorithm
Unlike the greedy algorithm, the genetic algorithm, created in MATLAB for
this part of the study, begins with a random set of solutions, or a population,
consisting of a number of trial solutions. A trial solution consists of a trans-
formation vector as described in the previous section, that is, a lateral shift, a
vertical shift, and a rotation. The mutual information of each is evaluated. The
transformations that yield a mutual information values that meets or exceed a
76


iteration x-shift y-shift rot (degrees) MI
1 0 0 0 1.200664
2 -2 -2 -1 1.236596
3 -4 0 -2 1.252961
4 -4 0 -3 1.272571
5 -4 0 -4 1.291002
0 -2 0 -5 1.314651
7 -2 0 -6 1.353761
8 -2 0 -7 1.393065
9 0 0 -8 1.440195
10 0 0 -9 1.527838
11 0 0 -9.5 1.549610
Table 4.5: 128x128 run converges in 1.2058 minutes.
77


iteration x-shift y-shift rot (degrees) MI
1 0 0 0 0.795863
2 2 -2 -1 0.842388
3 0 -2 -2 0.850167
4 -2 -2 -3 0.872867
5 -4 -2 -4 0.899841
6 -4 0 -5 0.929021
7 -4 0 -6 0.965123
8 -2 0 -7 1.013159
9 -2 0 -8 1.076260
10 0 0 -9 1.170099
11 0 0 -9.5 1.202078
12 1 0 -10 1.217601
Table 4.6: 256x256 run converges in 4-3367 minutes.
resolution T0(x,y, angle) Tfinai{x, y, angle) MI tiine(min)
64 x 64 (0 0 0) (0 0 -9.5) 2.1101 0.6504
128x 128 (0 0 0) (0 0 -9.5) 1.5496 1.2058
256x 256 (0 0 0) (1 0-10) 1.2176 4.3367
Table 4.7: Summary of Tables 4-3, 4-4, and 4-5-
78


user-specified value above the average value are selected. Each of these is ran-
domly paired with another member in this set. A new set of vectors is created
by randomly exchanging vector elements between the pairs. Additionally, for
each iteration, another random set of trials half the size of the original set is
added, an immigration step. Every trial is recorded so that none is repeated.
After a user-specified number of iterations, the program terminates and the vec-
tor that yields the maximum value for mutual information is considered to be
the solution. Obviously, the larger the population and the larger the number of
iterations, the greater the chance that this will be the optimal solution.
An experiment was run two times with the 256 x 256 DRR and EPI image
shown in Figure 4.11 and yielded the results shown in Table 4.8. The approxi-
mate optimal solution was found in run 2. Each run began with a population of
100 and an initial transformation vector (0 0 0). Forty iterations were specified
as well as a maximum shift, in absolute value, of 11 pixels, and maximum rota-
tion, also in absolute value, of 10 degrees. Clearly, more iterations are required
to guarantee an approximately optimal solution.
Repeating the same experiment, but using maximum shifts, in absolute
value, of 2 pixels, and maximum rotation, also in absolute value, of 10 degrees
yielded the results shown in Table 4.8. The logic here is to assume that the actual
transformation lies within, in absolute value, the range of values determined
by these maxima. Note the shorter runtime and that both runs yielded an
approximately optimal solution. This is no doubt due to the fact that the
respective vector values were from a relatively narrow range and close to the
optimum.
79


If, say, 10 2 had been designated as the range of values to be considered for
runs 3 and 4, then one could expect a shorter runtime. For these runs, rotation
values from 0 to 10 were possible, so a wider range of values generated resulted
in longer runtime.
The algorithm was adjusted to account for a range of values between user-
specified minimum and maximum values for each parameter. The results for
runs 5 and 6 are displayed in Table 4.8. The only changes in the experimental
parameters (Table 4.9) were the addition of a minimum transformation vector
of values, (0 0 9), and a change from 10 to 11 for the upper bound on rotation
giving a maximum transformation vector of (2 2 11). Not only was the run-
time significantly reduced, but mutual information values closer to optimal were
achieved. Note that where the greedy search allows additions to integer rotation
values from the set {-1,-.5,0,.5,1}, the genetic algorithm allows for a continuous
range, that is, randomly generated decimal values plus or minus an integer or
zero, of values for rotation.
In all of these experiments, fitness was defined as having a mutual infor-
mation value 1.1 and above times the average value for an iteration. Keeping
all parameters the same and increasing this factor to 1.2, yielded the results
shown in Table 4.8. As expected, mutual information values near optimum were
achieved and there was a significant reduction in runtime apparently due to the
fact that fewer vectors deemed as fit, can be chosen from a narrower range of
values that are also within the capture range of the optimum.
The parameters used for these experiments are tabulated in Table 4.9.
4.2 Nelder-Mead (MATLABs fminsearch)
80


parameters 1,2 run 3,4 pairs 5,6 7,8
To{x,y. angle) (0 0 0) (0 0 0) (0 0 0) (0 0 0)
initial population 100 100 100 100
iterations 40 40 40 40
above average factor 1.1 1.1 1.1 1.2
min x-shift - - 0 0
min y-shift - - 0 0
minimum rotation - - 9 9
max x-shift 11 2 2 2
max y-shift 11 2 2 2
max rot 10 10 11 11
Table 4.9: Parameter list for genetic run pairs.
x-shift y-shift rot (degrees) MI time(min)
1 0 -9.8673 1.2213 1.1455
Table 4.10: MATLABs fminsearch (Nelder-Mead) algorithm. 256x256 im-
ages.
82


of the angle of rotation of -9. The integral part of the value was not changed,
since, from the previously described experiments and visual inspection, it was
established that the solution for rotation is 9 a decimal value. The results
from over 7500 decimal values generated are shown in Figures 4.12 and 4.13. A
value of -9.867333 as the angle of rotation with a mutual information value of
1.221330 was obtained. Figure 4.13 shows that a range of angle values gives the
same mutual information value.
A major advantage of simulated annealing is its ability to avoid becoming
trapped at local minima. The algorithm employs a random search that can
accept changes that decrease objective function as well as those that increase it.
However, for this application, acceptance of a transformation that results in a
smaller value for mutual information makes no sense, so only those that result
in a higher mutual information value are accepted.
4.4 Other Experiments
In addition to the EPI example used in this chapter, 18 other EPIs of resolution
64x64 were tested using the greedy algorithm which was modified to limit the
search space. In each case, the search space included x- and y-shifts of 0 to 8
pixels, and rotations of 0 to 10 degrees. Additionally, in every case, the greedy
algorithm was called 20 times with 50 iterations maximum specified per run. It
was determined, by the data and visual inspection of the output graphics, that
the resultant transformations were successful in all cases.
83


Figure 4.12: Plot of MI versus Angle of Rotation from simulated annealing
data.
84


Figure 4.13: Detail of Figure 4.12, plot of MI versus Angle of Rotation from
simulated annealing data.
85


5. Conclusion
In the experiments described in Chapter 4, mutual information-based image reg-
istration was found to be a robust and easily implemented technique. In every
case the entire, unpreprocessed, image was registered. Because the setup errors
were unknown, the registrations were judged to be successful or not based on
visual inspection. In all nineteen cases, the use of the mutual information tech-
nique resulted in successful transformations as determined by visual inspection.
It appears that the greedy and genetic algorithms perform well, and in a rea-
sonable amount of time from a clinical perspective, given that there is assumed
to be a good estimate of the general location of the optimum. If such an esti-
mate could not be made, then the genetic algorithm would probably outperform
the greedy algorithm since the greedy algorithm tends to terminate at a local
optimum. Of course, this problem is resolved by multistart optimization, that
is, doing multiple searches starting with a variety of transformations resulting
in multiple solutions, and then choosing the solution which yields the highest
value of mutual information.
Given a full range of possible transformation values, the genetic algorithm
can find the global optimum given enough time. The time required may be
more than is practicable, but the problem can be made practicable, as was
demonstrated in this experiment, by constraining the search area to within the
capture range of the optimum. As mentioned in Section 4.1.1, the size of the
capture range depends on the features in the images and cannot be known a
86


priori, but visual inspection of the registered images can reveal convergence
outside of the capture range.
The advantage of mutual information-based image registration is that it
can be fully automatic in that it makes no assumption of the functional form or
relationship between image intensities in the image to be registered. However,
it is clearly important that a method of quality assurance be used to ensure that
only well registered images are used for clinical decision making.
87


Appendix A. Brief Discrete Probability Theory
Definition A.l Given an experiment whose outcome is unpredictable, such an
outcome, say X, is called a random variable, or trial. The sample space
of the experiment is the set of all possible trials. If the sample space is either
finite or countably infinite, the random variable is said to be discrete.
Example 5. Roll of a die.
Suppose that a die is rolled once, and let X denote the outcome of the experi-
ment. The sample space for this experiment is the 6-element set
n = {1,2,3,4,5,6},
where each outcome i, for i 1,..., 6, corresponds to the number of dots on
the face that turns up. The event
£ = {1,3,5}
corresponds to the statement that the result of the roll is an odd number. As-
suming that the die is fair, or unloaded, the assumption is that every outcome
is equally likely. Therefore, a probability of | is assigned to each of the six
outcomes in il, that is, m(i) = for 1 < i < 6.
Definition A.2 Let X be a random variable which denotes the outcome, of
finitely many possible outcomes, of an experiment. Let il be the sample space of
the experiment. A probability distribution function, or probability mass
88