You are on page 1of 22

Group Decision and Negotiation 9: 275–296, 2000

THE IMPACT OF ROLE TRAINING 275


© 2000 Kluwer Academic Publishers. Printed in the Netherlands

The Impact of Role Training in a User-Driven Group


Support System Environment
CHELLEY VICIAN
School of Business and Economics, Michigan Technological University, 1400 Townsend Drive, Houghton, MI
49931-1295, USA

GERARDINE DESANCTIS
The Fuqua School of Business, Duke University, Box 90120, Durham, NC 27708-0120, USA

Abstract

We tested the proposition that training groups to use roles would improve performance in a user-driven,
GSS-supported meeting. Two methods of role training, fixed and rotated roles, were compared against a
control procedure in which no formal role training was provided. In a repeated measures experiment,
half of the groups completed four generate-creative tasks, while the remaining groups completed three
generate-creative tasks followed by one choice-intellective task. Both role training methods were
successful in reducing the time spent by groups orienting themselves to the technology, particularly in
the choice-intellective task context; however, decision quality or quantity did not improve with role
training.

Key words: group decision making, group decision support systems, group support systems, roles,
training

The implementation of group support systems (GSS) and other collaborative technologies
has led to a need to assure their effective use in organizational settings. GSS are integrated
software systems which provide electronic brainstorming, idea recording, voting, storage,
and display tools, usually within the context of a face-to-face meeting (see Jessup and
Valacich 1993). In a typical configuration, each group member is provided with a computer
workstation for entering ideas, votes, and the like; a shared viewing screen then displays
the results of individual inputs for all members to see (e.g., Bostrom, Watson and Kinney
1992). Although GSS have been shown to improve group performance, they most often
require external facilitation or other special resources because groups have difficulty in
learning how to operate the technology and apply the various decision aids to their specific
tasks (Anson, Bostrom and Wynne 1995; Bostrom, Anson and Clawson 1993; Dickson,
Lee-Partridge and Robinson 1993). Unlike typical office systems or desktop tools, GSS
require that users coordinate their use of the technology with those of other group members
and understand basic group decision modeling techniques.
276 VICIAN AND DESANCTIS

Improvements in group outcomes have been found when GSS groups are provided with
additional resources such as human facilitators, technical chauffeurs, or specially-developed
software that is embedded with on-line decisional guidance (Anson et al. 1995; Bostrom et
al. 1993; Dickson et al. 1993; Limayem 1992; Reagan-Cirincione 1994). A facilitator can help
the group to troubleshoot problems in operation of the technology, to decide when and
how to use specific decision tools, and to interpret results of decision analysis. Directing
groups to follow process-focused rules can also improve group decision performance
(DeSanctis, D’Onofrio, Sambamurthy and Poole 1989; Wheeler and Valacich 1996). But all
of these approaches require special resources, and little is known about what can be done
to assist a group in operating the GSS technology on its own, without reliance on such
resources. F urther, bo laboratory and field studies testing the efficacy of the strategies
listed above show that groups still encounter problems in applying the technology to their
work task(s) (DeSanctis, Poole, Lewis and Desharnais 1991; DeSanctis, Poole, Dickson and
Jackson 1993; Nunamaker, Briggs, Mittleman, Vogel and Balthazard 1996; Poole, Holmes,
Watson and DeSanctis 1993; Vician, DeSanctis, Poole and Jackson 1992; Watson, DeSanctis
and Poole 1988; Zigurs, Poole and DeSanctis 1988). No single strategy has proved
completely successful in enhancing group performance in GSS settings. How might group
performance be enhanced in a GSS session without external assistance, that is, in
“user-driven” GSS environments?
In this study we sought to identify whether groups could be taught to rely on their own
internal resources for effective GSS use. The group’s role system, which is a part of its
standing structure, is an available resource that might be shaped to manage a group’s
activities in a GSS session, thereby improving group performance (Bettenhausen 1991;
Biddle and Thomas 1966; Guetzkow 1968). We examined whether providing groups with
meeting roles and a training intervention for managing the roles in group interaction would
positively influence group performance. We were particularly interested in the impact of
role training across multiple trials with a GSS under the circumstances of a similar (same)
versus a novel (changed) task. In the next section we review prior theory and develop our
research model and hypotheses. We then describe our experiment and findings. We conclude
by discussing the implications of our results for further research.

Research framework and model

Our research framework is grounded in the concepts of role and role functions (Biddle and
Thomas 1966; Bormann 1990; Doyle and Strauss 1976; Hackman and Morris 1975; Katz
and Kahn 1978; McGrath 1984). Research on group-technology interaction suggests how
role specification might improve technology use and outcomes, and how repeated role
training will impact the use of technology over time and for different types of tasks (Benbasat
and Lim 1993; Dennis and Gallupe 1993; DeSanctis and Poole 1994; Nunamaker et al. 1996;
Sarbin and Allen 1968).
THE IMPACT OF ROLE TRAINING 277

What is a role?

Role theory assumes that a social setting embodies a structure consisting of parts, positions,
or offices which individual members of the collective occupy (Biddle and Thomas 1966;
Katz and Kahn 1978; Sarbin and Allen 1968). From this structural perspective, the concept
of role encompasses those activities normally associated with the conduct of a specific
position within the larger social system. A role embodies the particular range of potential
behaviors that an incumbent in the position can perform. The patterned expectations
associated with roles help to explain and predict the behavior of individuals in a collective
setting. Roles provide a bridge between individual actions and the larger collective context
(Katz and Kahn 1978; Roos and Starke 1981; Sarbin and Allen 1968). Roles are an integral
component of the standing group structure, a set of resources brought into the interaction
process (Bormann 1990; McGrath 1984). Role functions are those social and task activities
that must be accomplished in the course of productive group communication. Together,
roles and role functions act to link specific communication behaviors to individuals within
group interaction (Benne and Sheats 1948; Bormann 1990; Doyle and Strauss 1976).
Doyle and Strauss (1976) propose “the interaction method,” which is an organizing set
of roles and role functions designed to facilitate the effectiveness of decision-making
meetings. Role specialization is posited to improve performance by reducing the number
of role functions that are performed by any one group member and by improving attention
and performance for any one function performed by a member (Doyle and Strauss 1976).
A synopsis of four major meeting roles and attendant role functions is presented in Table
1. Within a GSS-supported meeting, specification of role functions should help to clarify
member interrelationships and technology responsibilities. Indeed, a central proposition
of our study is that group performance can benefit from specification of group roles and
role functions in GSS-supported meetings, particularly for novice users of the technology.

Table 1. Group meeting roles and corresponding role functions

Group meeting role Role functions

Facilitator Strictly procedural focus, not content; ensures that the group
follows any selected procedure, rules, guidelines; does not
engage in the content decisions of the meeting.
Recorder Provides documentation support for the group; keeps track of
meeting flow by using some form of group memory; provides
displays of group ideas or decisions when needed by the group
Member Actively follows the meeting content; works with other members
to make decisions; ensures that facilitator and recorder enact
behaviors commensurate with those roles so that member(s) can
pay attention to task at hand.
Leader/chairperson Also a member of the group; tries to discourage dominance by
any one member; encourages member participation; increases
group ownership of the meeting and its activities and decisions

(Adapted from Doyle and Strauss 1976).


278 VICIAN AND DESANCTIS

Effects of role specification in GSS sessions over time

A GSS-supported meeting presents a complex web of interactions to be managed: (a)


relationships between the group and its task; (b) relationships between the group and the
technology; and (c) relationships between the task and the technology (Applegate 1991;
DeSanctis and Poole 1994; McGrath and Hollingshead 1993). To the extent that novice
users of a GSS lack expertise in managing these relationships, techniques which enable
clarification of member responsibilities should serve as internal organizing resources for
the group, facilitating members’ interaction with each other, their task, and the technology.
Group support settings can be complex for users unaccustomed to invasive technology in
their meetings – formal decision procedures, computer screens, and the like (Chidambaram
1996; Watson et al. 1988). Role specification is expected to reduce uncertainty in group
interaction by specifying who is responsible for what actions at what times during the GSS
session. In particular, we expect that role specification will reduce the amount of time a
group spends communicating about its technology use, thus freeing up group attention to
focus on the task at hand.
Table 2 outlines the key meeting roles and their associated role functions and behaviors in
a GSS meeting context. These roles and behaviors are derived from prior research on small
group communication, GSS facilitation, and GSS use over time by actual work groups
(Ackerman 1996; Anson et al. 1995; Bostrom et al. 1993; Clawson, Bostrom and Anson
1903; DeSanctis et al. 1993; Dickson et al. 1993; Hirokawa and Gouran 1989; Vician et al.
1992). The four roles of chairperson, recorder, technology operator, and participant highlight
the necessary behaviors for managing group-task-technology relationships in a user-driven
(non-facilitated) GSS-setting. Our expectation is that if each role is attached to a particular
person within the group, clarity of process can result, which, in turn, can serve to enhance
group performance. Role specification acts as a structuring method for group interaction
patterns, and structuring methods have been shown in other settings to: (a) reduce confusion
regarding member responsibilities (Worchel and Schackelford 1991), and (b) provide a means
of proceduralizing group communication behaviors (Bavelas and Barrett 1951; Leavitt 1951)
thereby improving group performance (Hirokawa 1982; Shaw 1981).

Effects of role specification for new tasks

Role specification is expected to enable the group to adapt its use of the technology more
easily and rapidly, especially over time and for changes in task types (Guetzkow 1968;
Poole et al. 1993; Sarbin and Allen 1968; Worchel and Schackelford 1991). Without role
specification, novice groups face the stress and uncertainty of how to behave with each
new task type that comes along in the course of their GSS session. However, groups
receiving role specification as part of a role training intervention should have a performance
advantage when encountering a change in the type of decision task they confront, as member
responsibilities and communication behaviors have been structured by the presence of
roles. Thus, groups that have received role training can focus their efforts on task concerns
rather than group operating procedures (Mennecke, Hoffer and Wynne 1992; Tuckman
THE IMPACT OF ROLE TRAINING 279

Table 2. Meeting roles with corresponding role functions and GSS functional use behaviors for a
GSS-supported meeting

Meeting role Role functions GSS functional use behaviors

Chairperson Guides group in task Draws group attention to key


accomplishment; encourages elements of public screen; suggests
member participation; and encourages use of GSS to
coordinates group task support task work of group
activities

Recorder Provides documentation Displays group input on the GSS


support for the group; keeps public screen; controls scrolling
track of meeting flow by using functions; maintains knowledge of
some form of group memory; how to view group work with the
provides displays of group GSS; prints out key elements as
memory when needed by needed; may perform deletion and
group. modification functions for group;
saves meeting minutes; documents
key group decisions in the GSS.

Technology Guides group in GSS system Directs usage of GSS tools;


operator features use; responsible for provides leadership for use of tool
knowing how to use particular explains process and sequence of
feature(s) of the system; steps; troubleshoots group
coordinates group system use problems with technology.
activities.

Participant Actively follows the meeting Enters items/ideas into GSS; reads
contents; makes decisions; and interprets system outputs.
ensure that other role
incumbents fulfill their
positions.

1965). The positive effects of role training upon group performance should be especially
apparent when the task changes as procedural clarification has occurred from the overt
role specification and groups can use the GSS effectively to address the new task.

Effects of role training method

The precise effects of role training may vary as a function of the type of training the group
receives, and it is conceivable that some types of role training will be more effective than
others. Roles may be fixed, in which case each member enacts the same role regardless of
task, or roles may be rotated, which means roles are enacted by different members as time
progresses. Fixed role training facilitates individual learning and specialization of member
responsibilities for technology use and may provide group performance efficiencies for
repetitive tasks over time (Davis and Bostrum 1993; Johnson and Johnson 1994). Role
rotation provides a broader knowledge of member responsibilities for technology use to
280 VICIAN AND DESANCTIS

each individual member, thus expanding the group’s experiential learning and knowledge as
a whole (Johnson and Johnson 1994; Kolb 1984; Slavin 1991). The cross-training of member
responsibilities for technology use inherent in role rotation can increase the pool of internal
resources brought to group interaction (DeSanctis and Poole 1994; McGrath 1984) and may
better enable the group to adapt its resources for application to different task types (Hackman
1990; Kolb 1984; Louis and Sutton 1991; Poole et al. 1993; Sarbin and Allen 1968).
For our exploratory purposes, we are interested in whether role rotation reduces the
amount of time spent applying GSS technology to the task at hand when groups encounter
a change in the type of task they confront. Under circumstances where the task type stays
constant, we expect no performance difference between groups trained in fixed and rotated
role methods. Under circumstances where the task type changes, we expect groups with
rotated roles to have a performance advantage over groups with fixed roles due to the
shared knowledge about how to apply the technology to the task which has been experienced
by all group members.

Hypotheses

Figure 1 summarizes our study in terms of an input-process-output model of group behavior


(see McGrath 1984; Hackman and Morris 1975). We posit that a role training intervention
will yield benefits to novice users of GSS technology as follows:

(H1) GSS-supported groups receiving role training will produce a greater number of ideas
than GSS-supported groups not receiving role training.
(H2) GSS-supported groups receiving role training will have higher decision quality than
GSS-supported groups not receiving role training.
(H3) When encountering a change in type of task, GSS-supported groups receiving role
training will have lower amounts of orientation time than GSS-supported groups not
receiving role training.
(H4) When encountering a change in type of task, GSS-supported groups receiving role
training will have higher decision quality than GSS-supported groups not receiving
role training.
(H5) When encountering a change in type of task, GSS-supported groups receiving role
training with rotated roles will have lower orientation time and higher decision quality
than GSS-supported groups receiving role training with fixed roles.

Method

Overview

We conducted a laboratory experiment in which we varied three levels of role training


(none, fixed roles, rotated roles) over four experimental trials. Task type was constant in
trials 1, 2, and 3 (generate-creative) and varied in trial 4 (same: generate-creative; changed:
THE IMPACT OF ROLE TRAINING 281

GSS technology

Figure 1. Research model. (iv = independent variable; c = controlled variable).

choice-intellective), yielding a 3 X 2 factorial design with repeated observations on the


second factor. An overview of the experimental design is shown in Figure 2.
In all experimental conditions, each group member was assigned the role of meeting
participant. The three roles of chairperson, recorder, and technology operator were either
not formally assigned (control condition), assigned to the same person in each task trial
(fixed roles), or systematically rotated across members in each experimental trial (rotated
roles). We utilized two task types drawn from McGrath’s (1984) group task circumplex:
generate-creative and choice-intellective. All groups were given a generate-creative task
in trials one through three so that they could gain experience with a common task type
during role training. In the fourth trial, a choice-intellective task was given to half of the
groups (changed task treatment) and a generate-creative task was given to the remaining
groups (same task treatment), so that we could compare the relative performance of the
fixed and rotated training methods for a change in task type.
To operationalize the group performance construct, we utilized multiple dependent
variables appropriate to each type of task. For the generate-creative task (same task

Figure 2. Overview of the experimental design (C = control; FR = fixed roles; RR = rotated roles).
282 VICIAN AND DESANCTIS

treatment) we examined the quantity (number of ideas) and the quality (decision quality) of
the idea generation decision. For the choice-intellective task (changed task treatment), we
examined the quality (decision quality) of the idea evaluation decision. We also developed
a common measure of decision quality so that performance on the two tasks could be
compared, and for both task treatments we examined orientation time.
Control variables included group size, group history, and GSS technology. All groups
had three members. Groups were newly formed for purposes of this study, and the GSS
was typical of that used in laboratory and field settings (see Bostrom et al. 1992).

Subjects

Two-hundred and seventeen students, including 23 master’s students and 194


undergraduates, were drawn from information systems and speech communication classes
at a large Midwestern university. The students received a modest number of credits toward
the final course grade for participating in the study. On average, the participants were 24.8
years of age; reported “medium” experience working in groups; used a computer three
times a week; and had 3.8 years of full-time work experience. There were no statistically
significant differences in the background characteristics of the subjects across the
experimental conditions. Assignment of participants to three-person groups1 was random
within educational level (i.e., undergraduate students were grouped with other undergraduate
students). Groups were randomly assigned to experimental condition with 22 groups in
the control condition and 24 groups per each role training treatment (fixed and rotated).

Procedure

The four trials were conducted within a single experimental session. All tasks, role training
methods, time limits per session, and experimental procedures were tested and modified
through a series of pilot tests prior to conducting the actual experiment. (For details on the
pilot testing procedures see Vician 1994.) In the two treatment conditions group members
were formally assigned the roles of chairperson, recorder, or technology operator. In the
fixed role condition, group members were randomly assigned one of these three roles
(e.g., chairperson) in the first trial and retained this role throughout the remaining trials. In
the rotated roles condition, group members were randomly assigned a role in the first trial
and then rotated the role assignments in the second and third trials such that each member
performed a new role in each trial; in the fourth trial members in the rotated roles treatment
were again randomly assigned one of the three roles. Roles and corresponding role functions
were based on the descriptions shown in Table 2 and were communicated to the participants
via written role sheets and bright yellow tent cards listing functional use behaviors placed
at each participant’s seating area. An experiment administrator (following a written script)
led the group members through GSS training activities relevant to each group’s experimental
treatment or control condition.
THE IMPACT OF ROLE TRAINING 283

Tasks

Following Gallupe and colleagues (1991, 1992) in their studies of electronic brainstorming,
the Thumbs task (Bouchard and Hare 1970) was used as a warm-up exercise to familiarize
the groups with the GSS and experimental procedures. Groups were asked to generate
ideas regarding the practical benefits or difficulties that might arise if people developed a
second thumb on each hand. The generate-creative task used in the experimental trials
likewise was drawn from Gallupe’s work (1991, 1992) and was of the general form, “How can
[task context] be increased/improved at [your University/your geographical region]?” Four
contexts were used, one in each experimental trial: tourism, parking, cultural diversity, and
campus security. As an example, the tourism question read: “How can tourism in [your
geographical region] be increased?” Tasks were randomly presented for the first three trials
to avoid sequence effects.
In the fourth trial, groups assigned to the same task treatment were given the problem,
“How can campus security in [your university] be improved?” Groups assigned to the changed
task treatment were presented with a list of alternative methods for improving campus security
and were asked to rank the solutions by their feasibility for implementation. Group ideas or
ranks were then compared with those of an expert panel consisting of seven security officers
and managers from the same university in which the study was conducted.
For each task, group members could start by developing ideas or rankings individually,
but they were instructed to share their ideas, discuss them as a group, and “work together
to develop a group answer that is acceptable to all members of the group.” For
generate-creative tasks they were instructed to “develop as large a list as possible that
contains unique (i.e., not duplicates) and innovative/creative responses to the [task]
question.” For the choice-intellective task they were instructed to “provide a rank order of
items from highest impact ideas to the least impact ideas. It is important that your group
agrees on the final prioritization.” Groups were given 12 minutes to complete each of the
first three experimental trials and 20 minutes to complete the last trial.

Measurement

For each generate-creative task the number of ideas was measured by a count of the final
list of non-duplicate ideas generated by the group. In trial 4, decision quality was measured
by comparing the group outcomes (either the idea set for the generate-creative task or the
final ranking for the choice-intellective task) to the ideas and ranks generated by the expert
panel. For the same task treatment, the quality of each idea was determined by an itemized
comparison of the group list of ideas to the final list of ideas generated by the expert panel.
If a group’s non-duplicate idea appeared on the experts’ list, the idea was given a score
equivalent to the rank value from the experts’ list of ideas. A decision quality (ideas) score
was calculated for the group by summing all of the idea quality scores in trial 4. For the
changed task treatment, the quality of the group’s rankings was assessed by direct
comparison of group answers to those of the expert panel. Error scores were computed as
differences from the expert panel rankings. A sum of the error scores served as an overall
284 VICIAN AND DESANCTIS

measure of the decision quality. The decision quality (ranks) of the group’s rankings was
equal to the summed deviations from the ranks given by the experts, which follows the
scoring method used in the Lost on the Moon task developed by Hall and Watson
(1970).2 Finally, in order to compare decision quality across the two tasks, we then
constructed a common measure of decision quality by converting the trial 4 quality
measures for each task type (just described) to a common 100-point scale for
comparability purposes.
Our orientation time measure was based on techniques developed by Poole and
colleagues (1993) to study start-up and mechanical friction in use of GSS. Tapes and
speech acts were carefully time stamped. Coding followed a strict, objective protocol;
each tape was viewed twice to assure accuracy of coding. The start and stop times of
speech acts verbalizing concerns, questions, or problems with the use of the GSS were
noted on coding sheets and then summed to get a total indicator of the amount of time
spent in orientation activities by each group for each trial. These speech acts occurred
throughout the entire GSS session, not just in the first few minutes of group technology
use. Verbalization of system usage activities is known to serve as a good indicator of
cognitive load in the technology adoption process (Poole et al. 1993; Todd and Benbasat
1987), and a greater amount of orientation time suggests greater difficulty on the part of the
group in applying the technology to their work task.

GSS

All groups were provided with the Software Aided Meeting Management (SAMM) version
5.2 as described by Dickson, Poole and DeSanctis (1992). The system provided individual
computer workstations to each meeting participant for information input and a common
viewing screen to the entire group for information display. The system supports group
idea recording, commenting, editing, and idea evaluation. SAMM has been used in
numerous studies of electronic idea generation and choice making (e.g., DeSanctis et al.
1991, 1993; Dickson et al. 1993; Watson et al. 1988; Zigurs et al. 1988). Subjects were
given hands-on instruction on the use of system features relevant to their task situation
(e.g., idea generation and idea evaluation) and experimental treatment or control condition,
together with detailed help sheets for the use of system features during the work session.
Subjects had control over the initiation and use of these features at any time during each
trial.

Results

Manipulation check

During the course of the data collection, group members were observed for exhibition of
functional use behaviors (see Table 2) associated with the three assigned roles in the study
(chairperson, recorder, technology operator). Using prepared check sheets, observers
THE IMPACT OF ROLE TRAINING 285

recorded the presence or absence of functional use behaviors associated with each role
and then summarized this data for each treatment. We then performed a series of Chi-Square
tests comparing the observed and expected frequencies of chairperson, recorder, and
technology operator behaviors across experimental conditions. Results indicated that
functional use behaviors occurred more frequently than were expected by chance in the
fixed and rotated role conditions for six out of eight categories of functional use behaviors
for the generate-creative task, and for seven out of the eight categories for the
choice-intellective task. In the control condition, a significant Chi-Square statistic was
observed in only two out of eight categories for both tasks. Further, observed functional
use behaviors were associated with the experimentally assigned roles given to individual
group members. We concluded that the role training manipulations were achieved in the
experiment.

Analytic approach

Descriptive statistics for dependent variables in each trial across each experimental condition
are shown in Table 3, and a summary of the statistical analyses is presented in Tables 4 and
5. Note that group experiences in trials 1–3 were, in essence, training for the challenge of
trial 4, and we expected greater variation in group performance in trial 4. Groups were given
more time to complete trial 4; half of the groups received a new task type; all were compared
to an expert panel for objective measures of decision quality in trial 4. Hence, we analyzed
the results as follows. For trials 1–3, we compared groups on number of ideas and orientation
time. For trial 4, we compared groups on decision quality (ideas or rank scores), decision
quality (common scale), and orientation time. Finally, we ran a full factorial model, comparing
the three role training scenarios and two task treatments in trial 4 on decision quality
(common scale) and orientation time. For all analyses, we began with a multivariate analysis
of variance (MANOVA) to detect overall treatment and trial effects and then proceeded
with univariate ANOVA and planned contrasts (Bonferroni multiple comparison procedure
with overall family alpha level set at .05). Preliminary analyses of the data indicated that the

Table 3. Means (and standard deviations) of dependent variable for each experimental condition

Trial 1 Trial 2 Trial 3 Trial 4 Trial 4


(generate-creative tasks)(choice-intellective task)
Outcome C FR RR C FR RR C FR RR C FR RR C FR RR
measures

number of ideas 10 9.45 8.54 8.27 7.6 6.5 8 7.1 7.5 7.9 8.25 7.58 n/a n/a n/a
(4.2) (2.4) (3.1) (2.1) (3) (1.8) (2) (1.7) (1.6) (1.9) (2.3) (2)
decision quality n/a n/a n/a n/a n/a n/a n/a n/a n/a 64.3 62 64.3 11.7 11.6 11.9
(ideas or rank (11) (15) (22) (2.2) (2.1) (1.8)
scores)
decision quality n/a n/a n/a n/a n/a n/a n/a n/a n/a 38 36.2 37.6 55.7 55.5 56.6
(common scale) (6.4) (8.9) (13) (10) (0) (8.9)
orientation time 40.4 43.1 30.6 30 19.8 35.8 26.1 9.79 21 19.0 20.0 26.0 107 39.2 56.2
(41) (53) (26) (30) (24) (45) (28) (18) (25) (20) (23) (26) (65) (30) (51)

(C = control; FR = fixed roles; RR = rotated roles).


286 VICIAN AND DESANCTIS

Table 4. Repeated measures multivariate analyses of variance for effects of role treatment across
trials

Task MANOVA for Trials 1–3 MANOVA for Trials 1–4


type

All groups, Repeated measures MANOVA for effect of T1–T3 role N/A
regardless treatment on number of ideas, orientation time
of trial 4 Role [Wilks lambda = .890, F(4,132) = 1.98, p = . 101]
task Role by Trial [Wilks lambda = .938, F(8,266) = 1.08, p = .378]
treatment Trial [Wilks lambda = .784, F(4,266) = 8.58, p = .000**]

Generate- N/A Repeated measures MANOVA for effect of T1–T4 role


creative treatment on number of ideas
(same) Role [F(2,32) = 1. 11, p = .343]
Role by Trial [Wilks lambda = .809, F(6,60) = 1.11, p = .365]
Trial [Wilks lambda = .723, F(3,30) = 3.84, p = .019*]

Repeated measures MANOVA for effect of T1–T4 role


treatment on orientation time
Role [F(2,32) = .43, p = .655]
Role by Trial [Wilks lambda = .749, F(6,60) = 1.55, p = . 177]
Trial [Wilks lambda = .770, F(3,30) = 2.98, p = .047*]

Choice- N/A Repeated measures MANOVA for effect of T1–T4 role


intellective treatment on orientation time
(changed) Role [F(2,32) = 8.09, p = .001**]
Role by Trial [Wilks lambda = 737, F(6,60) = 1.65, p = .150]
Trial [Wilks lambda = 571, F(3,30) = 7.49, p = .001**]

*significant at .05; **significant at .01.

dependent variables met the assumption requirements of the parametric statistical


procedures.3

Table 5. Multivariate and univariate analyses for effects of role treatment for each of two tasks in
Trial 4

Task type MANOVAs ANOVAs


Orientation time Decision quality
(rank scores)

Both tasks, MANOVA for T4 on orientation time, decision quality by


decision quality role by task
(common scale)
R x T [Wilks lambda = .863, F(4,126) = 2.4, p = .05*]
Task [Wilks lambda = .421, F(2,63) = 43.4, p = .000**]
Role [Wilks lambda = .879, F(4,126) = 2.09, p = .086]

Generate- MANOVA for T4 on number of ideas, decision quality, – –


creative (same) orientation time
[Wilks lambda = .925, F(6,60) = .396, p = .879]a
Choice- MANOVA for T4 on decision quality, orientation time F(2,32) = 5.534, p = .009** F(2,32) = .048, p = .953
Intellective [Wilks lambda = .739, F(4,62) = 2.53, p = .050*]
(changed) pooled t = -.99, df = 22,
p = .332

a
Because the MANOVA model is not significant, univariate ANOVA tests for number of ideas, decision quality,
and orientation time are not reported; *significant at .05; **significant at .01.
THE IMPACT OF ROLE TRAINING 287

Overall effects of role training and task

Repeated measures MANOVA analysis based on trial 1 – trial 3 data indicated a significant
main effect for trial [Wilk’s lambda = .784, F(4,266) = 8.58, p = .000] on the number of ideas
generated and orientation time while interaction (role training x trial) and main effects (role
training) were absent [Wilk’s lambda = .938, F(8,266) = 1.08, p = .378; Wilk’s lambda = .890,
F(4,132) = 1.98, p = . 101]. Further analyses of the overall effects of role training are discussed
below within task type.

Same (generate-creative) task. Repeated measures MANOVA analyses for trials 1–4 data
show a significant main effect for trial both on the number of ideas generated [Wilk’s
lambda = .723, F(3,30) = 3.84, p = .019] and orientation time [Wilk’s lambda = .770, F(3,30) =
2.98, p = .047]; however, effects due to role training or role training x trial interaction were not
significant. The number of ideas generated decreased significantly for all groups (irrespective
of role training) between trials 1 and 2 and between trials 1 and 4. Groups produced more
unique ideas in the first trial, perhaps due to the novelty of the task or their use of a new
technology. Orientation time also decreased significantly for all groups (irrespective of role
training) between trial 1 and trial 4. Figure 3 provides a plot of cell means for groups in the
same (generate-creative) task treatment and depicts the observed patterns in orientation
behavior over time. Interestingly, the fixed role groups exhibited high orientation time in the
first trial but then apparently made gains in adapting their use of the technology over the
successive trials. In trial 4, both the control and the fixed role groups had lower levels of
orientation time than the rotated role groups, though not significantly as demonstrated by
the statistical testing presented earlier. Examination of the trial 4 MANOVA analysis on all
performance outcomes taken together (number of ideas generated, decision quality (ideas),
and orientation time) indicated no significant differences due to differing role training
treatments [Wilk’s lambda = .925, F(6,60) = .396, p = .879].

Changed (choice-intellective) task. Repeated measures MANOVA analyses for trials 1–4
on orientation time revealed significant main effects due to role training [F(2,32) = 8.09, p
= .001] and trial [Wilk’s lambda = .572, F(3,30) = 7.49, p = .001]. Planned contrasts indicated
that orientation time was significantly lower for groups with fixed roles regardless of trial
and increased significantly for all groups (irrespective of role training) between trials 1
and 4 and between trials 3 and 4. Examination of the trial 4 MANOVA analysis on decision
quality (rank scores) and orientation time indicated a significant difference across role
training treatments [Wilk’s lambda = .739, F(4,62) = 2.53, p = .050]. Inspection of the
univariate portion of this MANOVA indicated significant effects of role training on
orientation time [F(2,32) = 5.534, p = .009]. The plot of cell means in Figure 4 depicts
somewhat similar patterns of learning to adapt to the technology over time for both role
training treatments in contrast to control groups. These patterns will be explored further in
the next section.
To summarize our results thus far, we did not find a decision quality or idea quantity
performance advantage in groups with role training when compared to control groups in
our overall analyses.4 H1 and H2 are not supported. The trial 4 results (within task treatment)
288 VICIAN AND DESANCTIS

Figure 3. Orientation time (same task) by role training.

suggest that the role training effect on orientation time under the treatment of the changed
(choice-intellective) task was significant; thus, H3 is supported.

Effects of role training when encountering a change in task

MANOVA results comparing the performance effects (decision quality, orientation time)
due to role training treatment across two task types at trial 4 indicated a significant role
training x task interaction effect [Wilk’s lambda = .863, F(4,126) = 2.4, p = .05]. As noted
above, groups encountering the same (generate-creative) task in trial 4 did not show
significant performance effects due to role training [Wilk’s lambda = .925, F(6,60) = .396, p =
.879]. However, MANOVA for the changed (choice-intellective) task indicated a significant
THE IMPACT OF ROLE TRAINING 289

Figure 4. Orientation time (changed task) by role training.

effect on performance variables (orientation time, decision quality-tank scores) due to type
of role training [Wilk’s lambda .739, F(4,62) = 2.53, p = .050]. The univariate portion of the
MANOVA indicated a significant effect for type of role training upon orientation time
[F(2,32) = 5.534, p = .009] but not upon decision quality (rank scores) [F(2,32) = .048, p =
.953]. Though H3 is supported, H4 is not supported.
The orientation time variable at trial 4 had a moderate violation of the homogeneity of
variance assumption, so a non-parametric test was run to substantiate the results of the
previously reported parametric test, A Kruskal-Wallis ANOVA of ranks indicated a significant
effect due to type of role training and type of task upon orientation time at trial 4 (χ =
20.0459, df = 5, p = .0012). Planned contrasts indicated that, for the changed
(choice-intellective) task treatment, groups lacking role training (control groups) encountered
significantly greater amounts of orientation time than groups with role training (See Table
6), again providing support for H3. A directional t-test indicated that there was no significant
290 VICIAN AND DESANCTIS

Table 6. Planned contrasts of orientation time for groups receiving role training vs. groups not
receiving role training under condition of changed task

Planned contrast Significance at trial 4

Control groups > Fixed role groups p = .05


Control groups > Rotated role groups p < .10
Control groups > average of Fixed & Rotated role groups p = .05

difference for orientation time between the fixed role and rotated role treatments (pooled
t-value = –.99, df = 22, p = .332). H5 is not supported.

Follow up analysis

To further explore the observed differences in orientation time in trial 4, we conducted


ad hoc qualitative case analyses of six groups focusing on the fourth, changed-task
session. Using a median cutoff value to split the sample based on orientation time, we
randomly selected three groups with high levels of orientation time and three groups with
low levels of orientation time (16% of our sample) for this analysis. We reviewed the
videotapes of these sessions and prepared case write-ups detailing group interaction.
The case write-ups were then compared and contrasted to produce the following
summaries.

High levels of orientation time. The case descriptions of these groups evidenced relatively
less time spent in thorough task-related discussion. Each of the three groups took the task
very seriously and appeared motivated and involved in the discussions, but there was a
pronounced tendency for these groups to run out of time to address the task fully. This
“not-enough-time” predicament was especially noticeable in group interaction once a
problem applying the GSS technology was encountered. Groups did not tend to discuss
up-front which member would be responsible for utilizing the GSS technology to perform
the group functions (e.g., printing, sorting, deleting), yet would halt the entire group’s
actions on the task work and engage all group members in a discussion about how to use
the GSS technology to proceed. The actual use of the GSS technology was very choppy
and appeared to reduce the amount of time groups could spend critically discussing the
task requirements. Groups debated task-related ideas and worked hard to understand each
others’ viewpoints, but they also debated how to use the technology. Time spent discussing
use of the technology (e.g., “Are you fixing the numbers on the screen?” “Shall we delete
that item now or wait?”) detracted from their available time to debate specifics of the task
at hand. Overall, the flow of the discussion in these groups was lively but not smooth, as
task-related work was frequently interrupted with discussion about what to do, or not do,
with the GSS. Two of the three groups asked for help from the experimental administrator at
least once during the experimental period.

Low levels of orientation time. In contrast, the case descriptions of these groups evidenced
clear communication amongst members regarding responsibilities for GSS technology use
THE IMPACT OF ROLE TRAINING 291

(generally completed within the first few minutes of the session) and a high level of
confidence in member abilities to complete the GSS technology actions. Interestingly, these
groups joked and laughed at least once during their session whereas the high orientation
groups did not. None asked for help from the experiment administrator. The groups evidenced
substantial discussions about the parking problem task and a notable competence in their
utilization of the GSS technology. If problems with technology application occurred in
these groups, the problems were addressed quickly by the group members and with a
negligible detraction from task time. Most notable in these groups was a sense of rhythm,
or pacing, to their work. The groups worked in a rapid, highly coordinated way, moving
smoothly through the task in a steady manner. They rarely talked over one another or took
time away from the task to talk about the technology. For example, if one member proposed
re-ranking ideas, the others reacted quickly, either accepting or rejecting the motion and
moving on, together, to the next step in the task. They always seemed to be moving forward
in the task at a steady pace, never pausing or digressing to discuss the technology or other
matters.

Discussion

We found that role training reduced groups’ orientation time as they learned to apply the
technology to their work – a difficulty noted repeatedly in studies of GSS adoption
(Nunamaker et al. 1996; Poole et al. 1993; Vician et al. 1992; Watson et al. 1988; Zigurs et
al. 1988). The type of role training – whether fixed roles or rotated roles – did not influence
group outcomes as much as the presence or absence of role training. Reduction in group
orientation time meant that members spent less time talking about how to apply the
technology to the task at hand and, from our qualitative observations, more time focusing
on the task itself. The reduction in orientation time was particularly apparent when groups
faced a change in task, a common and frequent occurrence in most organizational usage of
GSS (DeSanctis et al. 1991, 1993). The effects on orientation time were persistent and
strong in all of our multivariate analyses. Our qualitative analysis indicated that groups
with low orientation time had a rhythmic pace to their work; they moved through the
decision task in synchrony focusing much more on the task at hand than on the operation
of the technology. The technology faded into the background. Groups with high orientation
time, on the other hand, distributed their discussion across task and technology matters
with the latter appearing to interrupt the flow of the group’s work. It may be that low
orientation time is related to a sense of urgency and team energy that provides groups with
momentum. Brown and Eisenhardt (1998) and Gersick (1989) describe how groups are better
able to handle transitions and change in their work when they act in synchrony and with a
highly, coordinated sense of pacing. Further research might examine how role training
promotes pacing in groups and how such pacing can help or hinder effective use of GSS
technology.
Our findings can be interpreted as empirical evidence of the theoretical arguments
advanced in prior research on group technology adoption that effective use of technological
resources may require group learning of new methods or strategies for applying these
292 VICIAN AND DESANCTIS

technologies to ongoing tasks (Hiltz and Johnson 1990; Leidner and Jarvenpaa 1993; Poole
et al. 1993). Our findings suggest that the introduction of a role training intervention into
the initial GSS sessions of novice users may enhance the longer term use of such
technologies for multiple and differing decision tasks. One implication is that groups using
GSS over time may wish to incorporate group roles as a safeguard against group resistance
to the technology borne out of premature frustration with applying the GSS application to
the task at hand.
We noted different patterns of orientation time for groups with roles in contrast to groups
without roles, particularly when the groups encountered a change in task. Under the
circumstances of the same task, groups with fixed roles experienced a higher amount of
orientation behavior in the first trial, but when the task subsequently changed, groups with
fixed roles experienced lower amounts of orientation behavior. The results suggest that
some groups may initially resist the imposition of a prescribed group structure upon the
group interaction process, but in the course of working within assigned roles, a certain
level of interaction effectiveness with the technology begins to emerge. This interpretation
is consistent with prior scholarship on group role enactment which states that roles emerge
through group interaction over time, shaped through dynamic processes of meeting the
needs of the group (Bormann 1990; McGrath 1984). This interpretation is also consistent
with the message of adaptive structuration theory: groups must develop their own ways of
working with technology and appropriating structural resources into group interaction to
achieve group performance advantages (DeSanctis and Poole 1994; Poole and Baldwin
1996).
Theoretically, the reduction in orientation time associated with role training should
have resulted in greater quantity and quality of decision outcomes for the group (DeSanctis
and Poole 1994; Hirokawa 1982; McGrath 1984). The quantity and quality effects might
have been more pronounced had the sample size been larger, the measurement more
extensive, or the role training period extended over the course of days rather than just a
few hours within the confines of our laboratory. These are avenues for further research.
Additional research is also required to fully investigate how variations in time spent orienting
the group to technology use are related to group productivity and decision quality, as other
studies have suggested that some threshold of technology orientation behavior within a
group is useful to group performance (Poole et al. 1993). As global, virtual teams relying
upon group communication and decision technologies become commonplace in
organizational life, further laboratory and field research is clearly necessary to explore the
ramifications of varying levels of technology orientation behavior and its effects upon
performance outcomes in group support system environments.
Our study had the strengths of manipulating role training method, group task, and
undertaking repeated observation over several trials. Future research in this area should
recognize the difficulty in demonstrating subtle role effects in groups, the complexity of
design and results analysis for such studies, and that process analysis should be used along
with outcome analysis to observe the effects of roles on the performance outcomes of
groups as they use information technology.
THE IMPACT OF ROLE TRAINING 293

Notes

1. Actual distribution of 217 subjects to groups resulted in the creation of seven 4-person sets. In each of these
4-person sets, the additional member was assigned the role of session observer and was seated away from
the work session area behind a panel. The session observer watched the work session with the experiment
administrator via a TV monitor and logged his/her observations of the group’s functional role behavior
enactments. No interaction between the session observer and the 3-person working group was permitted
during the experimental session.
2. The value of the error score was defined to be inversely related to decision quality. These overall scores
were then reverse-coded for analysis purposes, making a higher error score equivalent to higher quality.
Thus, according to Hall and Watson (1970) a score of 0 would note a complete match with expert values. In
this study, a score of 0 was reverse-scored to the maximum error of 21 in our dataset and other scores were
accordingly re-coded to remain consistent with the same task quality values.
3. Examination of the preliminary analyses indicated a slight violation of the univariate homogeneity of variance
assumption for the number of ideas (same task treatment at trials 1 and 2, and changed task treatment at all
trials) and orientation time (changed task treatment at trials 1 and 2; same task at trial 1; changed tasks at
trial 4) variables according to Cochran’s C and Bartlett Box-F criteria. Inspection of plotted data indicated a
fairly symmetrical distribution of error sources and did not suggest major outliers in the dataset. The ANOVA
F-test is robust to the situation as described (Norton 1952), and any outlier measurement in our dataset may
inflate the error mean square resulting in an F-test that is more conservative relative to Type I error (p. 25,
Barker and Barker 1984). Square root transformations of the data resolved the assumption violation problem
but did not change the individual test results or the conclusions drawn although the transformations introduced
interpretation problems. Hence, we decided to report all data and results in raw data format. In terms of the
variance-covariance assumption underlying the MANOVA test, preliminary analyses showed a mild violation
with the doubly multivariate repeated measures MANOVA over trials 1–3 for number of ideas and orientation
time and for the repeated measures MANOVA over trials 1–4 on orientation time for the changed task
treatment. Under these circumstances, both Olson (1976) and Hand and Taylor (1987) suggest the use of the
Pillai’s Trace as the MANOVA test statistic as it is most robust to departures from MANOVA assumptions.
In these two instances, we reviewed both the Pillai and Wilk’s statistics and relied on the Pillai as an indicator
of significance. We further found that the statistics (and significance values) were essentially equivalent
thus we report Wilk’s statistic in the main body of the paper. Where a planned comparison was necessary
and the assumption violation was present, we used the Huynh-Feldt modification of the univariate
results to compute an adjusted MSE for the Bonferroni calculation, thus ensuring appropriate safeguards
against Type I error (Barker and Barker 1984; Norusis 1990). Further, in the case of the full-factorial
MANOVA on decision quality and orientation time at trial 4, a non-parametric Kruskal-Wallis
one-way ANOVA for orientation time was run as a substantiation of the significant effect reported
in the parametric test. Overall, we determined that our statistical analyses and interpretation procedures
have ensured against Type I error, and thus, the assumption conditions of the parametric tests are
essentially met for the purposes of stating research conclusions. Further details and analytical
comparisons that support our analysis process are available upon request from the first author.
4. At the advice of reviewers, we attempted additional measures of decision quality (ideas) and idea quantity to
address the common research situation of these variables being highly correlated such that only one measure
is necessary to depict differences in group performance. Our attempts at re-measurement (“selecting”
a group’s idea if it matched the expert’s top 3 ideas; re-counting unique ideas per Gallupe et al’s.
(1991, 1992) instructions; and re-counting unique ideas if they matched the expert top 3 list, thus
allowing for a maximum score of 3) were unsuccessful in identifying significant differences.

References

Ackerman, F. (1996). “Participants’ Perceptions on the Role of Facilitators Using Group Decision
Support Systems,” Group Decision and Negotiation 5(1), 93–112.
294 VICIAN AND DESANCTIS

Anson, R., R. Bostrom, and B. Wynne. (1995). “An Experiment Assessing Group Support Systems and
Facilitator Effects on Meeting Outcomes,” Management Science 41(2), 189–208.
Applegate, L.M. (1991). “Technology Support for Cooperative Work: A Framework for Studying Introduction
and Assimilation in Organizations,” Journal of Organizational Computing 1, 11–39.
Barker, H.R., and B.M. Barker. (1984). Multivariate Analysis of Variance: A Practical Guide to Its Use in
Scientific Decision Making. Alabama: University of Alabama Press.
Bavelas, A., and D. Barrett. (1951). “An Experimental Approach to Organization Communication.” Personnel
27, 366–371.
Benbasat, I., and L.H. Lim. (1993). “The Effects of Group, Task, Context, and Technology Variables on the
Usefulness of Group Support Systems.” Small Group Research 24, 430–462.
Benne, K., and P. Sheats. (1948). “Functional Roles of Group Members.” Journal of Social Issues 4, 41–
49.
Bettenhausen, K.L. (1991). “Five Years of Group Research: What We Have Learned and What Needs to
be Addressed.” Journal of Management 17, 345–381.
Biddle, B.J., and E.J. Thomas. (1966). Role Theory: Concepts and Research. New York: Wiley.
Bormann, E.G. (1990). Small Group Communication: Theory and Practice (3rd edition). New York: Harper & Row.
Bostrum, R.P., R. Anson, and V. Clawson. (1993). “Group Facilitation and Group Support Systems.” In L.M.
Jessup and J. Valacich (eds.), Group Support Systems: New Perspectives (pp. 146–168). New York:
Macmillan.
Bostrum, R.P., R.T. Watson, and S.T. Kinney. (1992). Computer Augmented Teamwork: A Guided Tour. New
York: Van Nostrand Reinhold.
Bouchard, T.J. Jr., and M. Hare. (1970). “Size, Performance, and Potential in Brainstorming Groups.” Journal
of Applied Psychology 54(1), 51–55.
Brown, S., and K. Eisenhardt. (1998). Competing on the Edge: Strategy as Structured Chaos. Boston: Harvard
Business School Press.
Chidambaram, L. (1996). “A Study of Relational Development in Computer Supported Groups.” MIS Quarterly
20(2), 143–165.
Clawson, V., R.P. Bostrom, and R. Anson. (1993). “The Role of the Facilitator in Computer-supported Meetings.”
Small Group Research 24, 547–565.
Davis, S.A., and R.P. Bostrum. (1993). “Training End Users: An Experimental Investigation of the Roles of the
Computer Interface and Training Methods.” MIS Quarterly 17, 61–85.
Dennis, A.R., and B. Gallupe. (1993). “A History of Group Support Systems Empirical Research: Lessons
Learned and Future Directions.” In L.M. Jessup and J. Valacich (eds.), Group Support Systems: New
Perspectives (pp. 59–76). New York: Macmillan.
DeSanctis, G., M.J. D’Onofrio, V. Sambamurthy, and M.S. Poole. (1989). “Comprehensiveness and
Restrictiveness in Group Decision Heuristics: Effects of Computer Support on Consensus Decision Making.”
In J.I. DeGross, J.C. Henderson, and B.R. Lonsynski (eds.), Proceedings of the Tenth International
Conference on Information Systems (pp. 131–140). Boston, MA.
DeSanctis, G., and M.S. Poole. (1994). “Capturing the Complexity in Advanced Technology Use: Adaptive
Structuration Theory.” Organization Science 5(2), 121–147.
DeSanctis, G., M.S. Poole, G. Dickson, and B. Jackson. (1993). “An Interpretive Analysis of Team Use of
Group Technologies.” Journal of Organizational Computing 3(1), 1–30.
DeSanctis, G., M.S. Poole, H. Lewis, and G. Desharnais. (1991). “Using Computing to Improve the Quality
Team Process: Some Initial Observations from the IRS-Minnesota Project.” Proceedings of the Twenty-
Fourth International Conference on System Sciences, Kauai, HI, Vol. III, 547–553.
Dickson, G.W., J. Lee-Partridge, and L. Robinson. (1993). “Exploring Modes of Facilitative Support for GDSS
Technology.” MIS Quarterly 17, 173–194.
Dickson, G., M.S. Poole, and G. DeSanctis. (1992). “An Overview of the Minnesota GDSS Research Project
and the SAMM System.” In R.P. Bostrom, R. Watson, and S.T. Kinney (eds.), (pp. 163–179).
Computer Augmented Teamwork: A Guided Tour. New York: Van Nostrand Reinhold.
Doyle, M., and D. Strauss. (1976). How to Make Meetings Work. New York: Wyden Books.
Gallupe, R.B., L.M. Bastianutti, and W.H. Cooper. (1991). “Unblocking Brainstorms.” Journal of
Applied Psychology 76(1), 137–142.
THE IMPACT OF ROLE TRAINING 295

Gallupe, R.B., A.R. Dennis, W.H. Cooper, J.S. Valacich, L.M. Bastianutti, and J.R. Nunamaker, Jr.
(1992). “Electronic Brainstorming and Group Size.” Academy of Management Journal 35, 350–369.
Gersick, C.J.G. (1989). “Making Time: Predictable Transitions in Task Groups.” Academy of Management
Journal 31, 9–41.
Guetzkow, H. (1968). “Differentiation of Roles in Task-Oriented Groups.” In D. Cartwright, and A.
Zander (eds.), Group Dynamics: Research and Theory (3rd edition, pp. 512–526). New York: Harper
& Row.
Hackman, J.R. (1990). Groups That Work (And Those That Don’t). San Francisco: Jossey-Bass.
Hackman, J.R., and C.G. Morris. (1975). “Group Tasks, Group Interaction Process, and Group Performance
Effectiveness: A Review and Proposed Integration.” In L. Berkowitz (ed.), Advances in Experimental
Social Psychology (Vol. 8, pp. 45–99). New York, Academic Press.
Hall, J., and W.H. Watson. (1970). “The Effects of a Normative Intervention on Group Decision-making
Performance.” Human Relations 4, 299–317.
Hand, D.J., and C.C. Taylor. (1987). Multivariate Analysis of Variance and Repeated Measures. New York:
Chapman and Hall.
Hiltz, S.R., and K. Johnson. (1990). “User Satisfaction with Computer-mediated Communication Systems.”
Management Science 36, 739–764.
Hirokawa, R. (1982). “Group Communication and Problem-solving Effectiveness I: A Critical Review of
Inconsistent Findings.” Communication Quarterly 30, 134–141.
Hirokawa, R., and D. Gouran. (1989). “Facilitation of Group Communication: A Critique of Prior Research and
an Agenda for Future Research.” Management Communication Quarterly 3, 71–92.
Jessup, L.M., and J. Valacich (eds.). (1993). Group Support Systems: New Perspectives. New York: Macmillan.
Johnson, D.W., and R.T. Johnson. (1994). Learning Together and Alone (4th edition). Boston: Allyn &
Bacon.
Katz, D., and R.L. Kahn. (1978). The Social Psychology of Organizations (2nd edition). New York:
Wiley.
Kolb, D. (1984). Experiential Learning: Experience as the Source of Learning and Development.
Engelwood Cliffs, NJ: Prentice-Hall.
Leavitt, H.J. (1951). “Some Effects of Certain Communication Patterns on Group Performance.” Journal of
Abnormal and Social Psychology 46, 38–50.
Leidner, D.E., and S.L. Jarvenpaa. (1993). “The Information Age Confronts Education: Case Studies on Electronic
Classrooms.” Information Systems Research 4, 24–54.
Limayem, M. (1992). “Automating Decision Guidance: Design and Impacts in a Group Decision Support
Environment.” Unpublished Ph.D. dissertation, University of Minnesota, Minneapolis.
Louis, M.R., and R.I. Sutton. (1991). “Switching Cognitive Gears: From Habits of Mind to Active Thinking.”
Human Relations 44(1), 55–76.
McGrath, J.E. (1984). Groups: Interaction and Performance. Englewood Cliffs, NJ: Prentice-Hall.
McGrath, J.E., and A.B. Hollingshead. (1993). “Putting the ‘Group’ Back in Group Support Systems: Some
Theoretical Issues About Dynamic Processes in Groups with Technological Enhancements.” In L.M. Jessup,
and J. Valacich (eds.), Group Support Systems: New Perspectives (pp. 78–96). New York: Macmillan.
Mennecke, B.E., J.A. Hoffer, and B.E. Wynne. (1992). “The Implications of Group Development and History
for Group Support System Theory and Practice.” Small Group Research 23(4), 524–572.
Norton, D.W. (1952). “An Empirical Investigation of Some Effects of Non-normality and Heterogeneity on the
F-distribution.” Unpublished doctoral dissertation, State University of Iowa.
Norusis, M.J. (1990). SPSS Advanced Statistics User’s Guide. Chicago: SPSS Inc.
Nunamaker, J.F. Jr., R.O. Briggs, D. Mittleman, D. Vogel, and P. Balthazard. (1996). “Lessons From a
Dozen Years of Group Support Systems Research: A Discussion of Lab and Field Findings.” Journal of
Management Information Systems 13(3), 163–207.
Olson, C.L. (1976). “On Choosing a Test Statistic in Multivariate Analysis of Variance.” Psychological Bulletin
83, 579–585.
Poole, M.S., M. Holmes, R. Watson, and G. DeSanctis. (1993). “Group Decision Support Systems and Group
Communication: A Comparison of Decision Making in Computer-supported and Non-supported groups.”
Communication Research 20, 176–213.
296 VICIAN AND DESANCTIS

Poole, M.S., and C. Baldwin. (1996). “Developmental Processes in Group Decision Making.” In R.
Hirokawa, and M.S. Poole (eds.), Communication and Group Decision Making (2nd edition). Thousand
Oaks, CA: Sage.
Reagan-Cirincione, P. (1994). “Improving the Accuracy of Group Judgment: A Process Intervention Combining
Group Facilitation, Social Judgment Analysis, and Information Technology.” Organizational Behavior and
Human Decision Processes 58, 246–270.
Roos, L.L. Jr., and F.A. Starke. (1981). “Organizational Roles.” In P.C. Nystrom, and W.H. Starbuck
(eds.), Handbook of Organizational Design, Volume I: Adapting Organizations to Their Environments
(pp. 290–308). New York: Oxford University Press.
Sarbin, T.R., and V.L. Allen. (1968). “Role Theory.” In G. Lindzey, and E. Aronsen (eds.), Handbook of
Social Psychology. (Vol. 1, 2nd edition, pp. 488–567). Reading, MA: Addison-Wesley.
Shaw, M.E. (1981). Group Dynamics: The Psychology of Small Group Behavior (3rd edition). New York:
McGraw-Hill.
Slavin, R.E. (1991). “Synthesis of Research on Cooperative Learning.” Educational Leadership 47, 52–55.
Todd, P., and I. Benbasat. (1987). “Process Tracing Method in Decision Support Systems Research: Exploring
the Black Box.” MIS Quarterly 11, 493–512.
Tuckman, B.W. (1965). “Developmental Sequence in Small Groups.” Psychological Bulletin 63, 384–399.
Vician, M. (1994). “Coordination of Interdependent Activities: The Impacts of Role Assignment and Rotation
on the Use of a Group Decision Support System.” Unpublished Ph.D. Dissertation, University of Minnesota,
Minneapolis.
Vician, C., G. DeSanctis, M.S. Poole, and B. Jackson. (1992). “Using Group Technologies to Support the Design
of ‘Lights Out’ Computing Systems: A Case Study.” In K.E. Kendall, K. Lyytinen, and J.I. DeGross
(eds.), Proceedings of the IFIP WG 82. Working Conference on The Impact of Computer Supported
Technologies on Information Systems Development, Minneapolis. MN, June 1417, 151–178.
Watson, R.T., G. DeSanctis, and M.S. Poole. (1988). “Using a GDSS to Facilitate Group Consensus: Some
Intended and Unintended Consequences.” MIS Quarterly 12, 463–477.
Wheeler, B., and J. Valacich. (1996). “Facilitation, GSS, and Training as Sources of Process Restrictiveness and
Guidance for Structured Group Decision Making: An Empirical Assessment.” Information Systems Research
7, 429–450.
Worchel, S., and S.L. Schackelford. (1991). “Groups Under Stress: The Influence of Group Structure and
Environment on Process and Performance.” Personality and Social Psychology Bulletin 17, 40–
647.

You might also like