Wagman

Interdependence and Group Effectiveness
Author(s): Ruth Wageman

Source: Administrative Science Quarterly, Vol. 40, No. 1 (Mar., 1995), pp. 145-180
Published by: Sage Publications, Inc. on behalf of the Johnson Graduate School of Management,
Cornell University
Stable URL: http://www.jstor.org/stable/2393703 .
Accessed: 07/02/2015 06:49
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Sage Publications, Inc. and Johnson Graduate School of Management, Cornell University are collaborating
with JSTOR to digitize, preserve and extend access to Administrative Science Quarterly.
http://www.jstor.org
This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

All use subject to JSTOR Terms and Conditions
Interdependence and This study investigated the differential effects of task
GroupEffectiveness design and reward system design on group functioning;
the effectiveness of "hybrid" groups, in which groups'
tasks and/or rewards have both individual and group
Ruth Wageman elements; and how individuals' preferences for
ColumbiaUniversity autonomy moderate their responses to interdependence
at work. An intervention in the reward system at a large
U.S. corporation created group, individual, and hybrid
rewards for 150 existing teams of technicians that had
group, hybrid, or individual tasks. Groups performed
best when their tasks and outcomes were either pure
group or pure-individual. Hybridgroups performed quite
poorly, had low-quality interaction processes, and low
member satisfaction. Task and outcome interdependence
affected different aspects of group functioning: Tasks
influenced variables related to cooperation, while
outcomes influenced variables related to effort.
Individuals' autonomy preferences did not moderate the
effects of task and reward interdependence but, instead,
were themselves influenced by the amount of
interdependence in the work. These findings have
implications for the design of work and reward systems
for work groups.'
Two contrastingmodels dominate the way both managers

and researchers thinkabout the design of work. On the one
hand, work can be designed to be highlyinterdependent,
requiringthe inputof several people to complete it. One
example of such a design is a team responsible for creating
a new advertisingcampaign.The team might include copy
writers, graphicartists, and projectmanagers, all of whose
contributionsare necessary for completingthe task, with
members held collectivelyaccountablefor the qualityof the
new promotionstrategy. Alternatively,work can be
structuredto be highlyindependentand to be performedby
individuals.The rewardsystem then reinforces individual
excellence. An example is a sales team in which each
member is given responsibilityfor sales in one specific
territoryand is paid a commission based solely on his or her
individualsales performance.
There also exists a thirdmodel that has received relatively
little attention: a "hybrid"design that combines elements of
interdependentand independentwork. One example of such
a design is a group of researchers in a development
laboratory,each of whom pursues independent research
projects and, in addition,collaborateson some largershared
? 1995 by Cornell University. enterprise. Members of such hybridgroups sometimes
0001-8392/95/4001-0145/$1 .00. operate entirelyindependentlyand sometimes as a team.
0
I would like to thank Chris Argyris, Inthe above examples, both the means by which the work
Richard Hackman, and Mike Jensen for is accomplished (the task) and the ways in which
help throughout this research project.
Thanks also to Suzy Fenwick, Tom
performanceis assessed and rewarded(the work outcomes)
Ruddy, and Chuck Ray at Xerox for their vary in interdependence.These distinct forms of
support and for help in making the interdependencecan be designed independentlyof each
research possible. And special thanks to
all the managers and technicians at Xerox
other and be combined in differentways. The research
for making me welcome in their described here examines the separate and joint effects of
workplace. This paper is based on a differentlevels of task interdependenceand outcome
doctoral dissertation; the research was
generously funded by the Harvard interdependence-individual,hybrid,and group-on the
Business School and Xerox Corporation. effectiveness of workinggroups in organizations.
145/AdministrativeScience Quarterly,40 (1995): 145-180

The effects of task and outcome interdependence on group
effectiveness also may depend on the characteristics of the
individuals doing the work. The technicians studied here are
organized into groups that vary in task interdependence. I
changed the nature of the rewards they received to create
variance in outcome interdependence. And I measured work
preferences of individual technicians to explore the role of
individual differences in moderating the effects of
interdependence on group effectiveness.
Forms of Interdependence
Interdependence among people in organizations can derive
from several sources: (1) task inputs, such as the
distribution of skills and resources and the technology that
define the work (e.g., individuals on an assembly line vs.
teams building whole products), (2) the processes by which
members execute the work (e.g., people who make sales
calls alone vs. people who sell as teams), (3) the way that
goals are defined and achieved (e.g., measures of collective
vs. individual performance), and (4) the way that
performance is rewarded (e.g., rewards contingent on group
vs. individual performance). In this research, I focus on two
forms of interdependence. The first derives from inputs into
the work and from the process by which the work is carried
out, which I refer to as task interdependence. The second
derives from the degree to which significant consequences
of the work-such as goal attainment and tangible
rewards-are contingent on collective performance, which I
refer to as outcome interdependence.
Task interdependence. Some researchers (e.g., Johnson
and Johnson, 1989) distinguish task interdependence, in
which each member must take action for other members to
do any part of their work (as in a basketball team) from
resource interdependence, in which each member can
complete his or her part of the whole, but resources such as
information are distributed among members and the whole
task is not complete until each member has completed his
or her part (e.g., a design team). Other scholars (e.g.,
Thompson, 1967; Van de Ven and Ferry, 1980), by contrast,
focus on the different processes by which inputs can be
combined to complete a whole piece of work-e.g., pooled
interdependence, in which subtasks are performed
separately and in any order, vs. sequential interdependence,
in which subtasks are completed in a specified sequence.
What these various types of interdependence have in
common is that each describes the degree to which the task
requires collective action.
There are several different perspectives on task
interdependence. Thompson (1967) viewed task
interdependence as a characteristic of work that is inherent
in the technology of the task (e.g., assembly line work is
inherently sequentially interdependent). Others (e.g., Shea
and Guzzo, 1989) have viewed task interdependence as a
characteristic of the way people behave in executing their
work (e.g., assembly line workers who help each other are
more task interdependent than those who do not). This
paper takes a perspective that falls between these two
extremes: Task interdependence is a structural feature of
146/ASQ, March 1995

Interdependence
work, but tasks can be designed to be performed at varying

levels of interdependence (e.g., workers formed into groups
and given the tools and instructions to build subassemblies
are more interdependent than those who are assigned
individual tasks on an assembly line). The instructions and
materials that define a task create a level of
interdependence that in turn may influence how much unit
members interact in executing the task (Hackman, 1969).
Task design is manipulable. Manufacturing work may be
designed so that individuals with distinct skills execute their
part of the task-one input into the final product-
independent of bther workers. Alternatively, group members
may be cross-trained and work simultaneously and, at times,
interchangeably, on completing the whole. And, finally, one
might create a hybrid form in which members sometimes
work alone at independent tasks and sometimes work
together as a team. This research investigates the
performance and processes of groups that experience these
three distinct forms of task interdependence.
Outcome interdependence. Outcome interdependence is
defined as the degree to which the significant outcomes an
individual receives depend on the performance of others. A
noninterdependent reward is one given exclusively for
individual excellence, such as a commission paid to
individual salespeople. A maximally interdependent reward,
by contrast, is one given to individuals based exclusively on
group performance, such as a gainsharing plan. A hybrid
reward is one in which a significant proportion of the reward
is based on group performance, and another significant
proportion is based on individual performance.
Goal achievement also may be seen as a significant
outcome of group work. The term goal interdependence
sometimes has been used synonymously with task
interdependence (Sayles, 1958), but task and goal
interdependence are conceptually and empirically
distinguishable: Goal interdependence can exist without any
interdependence in the means of accomplishing the work
(e.g., a room full of telemarketers may be held accountable
for a collective goal, but they complete individual tasks), and
vice versa (e.g., Mitchell and Silver, 1990). Like rewards,
goals can be established for groups, for individuals, or both.
This research focuses on how these three forms of outcome
interdependence, especially in conjunction with different
forms of task interdependence-individual, hybrid, and
group-affect group functioning. I develop the argument
with a set of research questions that I return to as I report
the results of the field experiment.
Interdependence and Performers' Work Experiences
Both task and outcome interdependence have been
identified by scholars as potentially key to group
effectiveness (Guzzo and Shea, 1987), and there has been
some debate about which is more critical (Deutsch, 1962;
Slavin, 1983), yet while these two forms of interdependence
can be conceptually distinguished, it is not clear that they
are experienced differentlyby group members. One
possibilityis that an increase in either form of
interdependenceincreases members' general sense that
147/ASQ, March 1995

they must depend on others at work, but they may not
differentiate between interdependence in the work itself and
interdependence in the outcomes of work. Alternatively,
members may experience task and outcome
interdependence differently, but changes in one form also
may influence experience of the other. Guzzo (1992), for
example, suggested that increasing outcome
interdependence may alter group members' experienced
task interdependence and, consequently, change the way
that they approach the work. Finally, group members may
perceive the design of the task and the nature of work
outcomes as distinct, and changes in one form may alter
their experience of interdependence in that form only.
Whether the two forms of interdependence are experienced
differentially has important implications for the design of
groups: If they act as substitutes for one another, for
example, then changes in either will alter team members'
experienced interdependence. This research examines
empirically the relative influence of task and outcome
interdependence on experienced interdependence through
the following question:
Research Question 1: To what extent do task and outcome
interdependencedifferentiallyaffect the interdependence
individualsexperience in theirwork?
Interdependence and Group Effectiveness
It also is not known how much interdependence among
group members is best for group effectiveness. For this
study, effectiveness is defined as having three components:
(1) the group's product meets or exceeds the needs of the
users of that product; (2) group members interact in ways
that allow the group to work more effectively over time,
learning from each other and developing norms of operating
that support high-quality performance; and (3) the group
experience, on balance, satisfies rather than frustrates the
needs of members (Hackman, 1987).
There are at least three distinct possible relationships
between total (task plus outcome) interdependence and
group effectiveness: (1) the level of interdependence does
not influence group effectiveness at all; (2) there is a
monotonic increase in group effectiveness as
interdependence increases; and (3) the relationship is
curvilinear, such that groups with very low and very high
levels of interdependence are less effective than groups
with moderate levels of interdependence.
The more the better. Group effectiveness may increase
with increasing levels of interdependence. In their
meta-analysis of the effects of cooperative vs. individualistic
and competitive forms of interdependence, Johnson and
Johnson (1989) made a strongly stated case for the benefits
of high interdependence, citing positive effects on learning,
achievement, cognitive complexity of thought, and
interpersonal relations. In a similar vein, Mesch et al. (1988)
found that high task interdependence in classrooms
enhanced learning relative to an independent task structure,
and this effect became even more pronounced when a
group rewardwas added to the interdependenttask.
148/ASQ, March 1995

Interdependence
Two problems arise in extrapolatingfrom these findings to

group performancein general. First,both the Johnsons'
(1989) meta-analysisand Mesch et al.'s (1988) work focus
on groups in classrooms, where the task is learning,rather
than on workinggroups in organizations.Second, in Mesch
et al.'s study, no comparisongroup experienced an
interdependenttask with rewards based on individual
performance.It may be that addingany performance-
contingent reward-even a noninterdependent
one-enhances interdependentlearning.To test for an
additiveeffect of task and outcome interdependenceon
workinggroups, tasks with differentlevels of
interdependence must be combined with both independent
and interdependentoutcomes for groups in organizations,as
was done in the study reportedhere.
Eithertoo much or too little is detrimental. Alternatively,
one might predictthat either very high or very low levels of
interdependencewill underminegroup effectiveness. Too
little interdependence may result in very low levels of
interactionamong group members, thus preventingmutual
learningand the development of collective norms of
operating(Johnson and Johnson, 1989). By contrast, very
high levels of interdependencein the task and outcomes
may raise the level of coordinationto the point where its
costs outweigh its benefits. Much energy may be expended
coordinatingmembers and regulatingcollective behaviorthat
might otherwise be expended on task performanceitself.
No relationship. Finally,the qualityof task performance,
coordinationamong group members, collective effort
directed at the task, and satisfactionwith the work all may
be uninfluencedby the level of interdependenceamong
members. Each of these outcomes may be controlledby
other variables,such as individualcharacteristicsor the
organizationalcontext. Or perhapswork can be performed
effectively at any level of interdependence,and group
members can adapt their interactionsto performwell at that
level of interdependence.The above arguments lead to the
following question:
Research Question 2: What are the effects of varyinglevels of
total (task plus outcome) interdependenceon group effectiveness?
Differential Effects of Task and Outcome Interdependence
on Group Functioning
While some previous research suggests task and outcome
interdependence influence behaviorsimilarly,they may
influence differentaspects of group functioning.Three
particularaspects of group functioningmay be differentially
affected by tasks and outcomes: cooperationamong group
members, individuals'effort levels, and group norms about
cooperationand effort.
Cooperation. Studies of task interdependence have
demonstratedthat higherlevels of task interdependence
result in more communication,helping,and information
sharingthan do individualistictasks (Crawfordand Haaland,
1972; Johnson, 1973). Studies of outcome interdependence
also have reportedthat group rewards increase cooperative
behavior(Millerand Hamblin,1963; Rosenbaumet al., 1980;
Shea and Guzzo, 1989). Thus both forms of interdependence
149/ASQ, March 1995

appear to influence the level of cooperative social interaction
in groups.
Effort. The level of individual members' effort has important
effects on the quality of group performance (Steiner, 1972).
Varying levels of outcome interdependence may differentially
influence motivation and thus affect task effort, as may be
the case with group versus individual rewards. People tend
to perceive the link between individual behavior and
outcomes to be more direct for individual rewards than for
group rewards, especially when the group is large. Because
people perceive group rewards to be less directly
influenceable by individual behavior than individual rewards,
group rewards may be slightly less potent motivators than
rewards for individual performance (Hayes, 1976). Thus,
other things being equal, highly interdependent outcomes
should result in lower motivation levels than independent
ones.
Norms. Task and outcome interdependence also may
influence different group norms, the informal rules that
groups use to regulate members' behavior. Norms provide
group members with a means for directing and anticipating
the behavior of others and for making quick and appropriate
responses (Kiesler and Kiesler, 1970; Shaw, 1981). Norms
emerge in groups to control the behaviors that are important
for the group's survival, its task success, and smooth
interpersonal relationships (Feldman, 1984). Two kinds of
behavior that may become codified as norms in
interdependent groups are the amount of cooperation
displayed by members and the level of effort exerted in the
task. Greater task interdependence makes mutual helping,
information sharing, and other cooperative behaviors more
important to completing the task, and it enhances members'
expectations of help and information sharing from others
(Thomas, 1957; Spilerman, 1971). Thus high task
interdependence may result in strong norms promoting
cooperation.
Effort norms, by contrast, may be more influenced by
outcome than by task interdependence. High outcome
interdependence makes high member effort more important
because all members' efforts contribute to an individual
member's chances to receive valued rewards (Deutsch,
1949). Group-based outcomes, then, may enhance group
norms that regulate members' efforts. Berkowitz's (1957)
study of the effects of different levels of outcome
interdependence on group member effort levels on an
interdependent task suggest this may be true. He found that
high outcome interdependence promoted the best
performance and concluded that group rewards enhanced
collective expectations (norms) about member effort and
heightened individual motivation, but his conclusions about
group effort norms and motivation were inferred from
performance scores, rather than being drawn from direct
measures of group norms. I examine this issue more directly
with the following research question:
Research Question 3: Do task and outcome interdependence
differentiallyinfluenceinterpersonalinteractions,motivation,and
normativecontrolof cooperationand effort?
150/ASQ, March 1995

Interdependence
Hybrid Forms of Interdependence

Research on interdependencein groups has focused on
comparisonsof the two extreme models: individualand group
work. The thirdmodel, the hybriddesign, has received
relativelylittle attention in the literature,despite the
prevalence of jobs that requireboth collective action and
independentaction. Johnson and Johnson (1989), who
examined the effectiveness of mixed models of
interdependence,found that they resulted in poorer
performanceand lower-qualitygroup processes than pure
cooperative (interdependent)forms. The mixturesthey
studied, however, Combineddifferenttask structures
(cooperativeor competitive)with contrastingforms of
outcome interdependence;there were no cases in which
both the task itself and the work outcomes combined
elements of both interdependenceand independence.
More attention has been given to hybridrewardsthan to
hybridtasks (e.g., Lawler,1990). Rosenbaumand his
colleagues (1980) offered individual,group, or mixed
rewardsto subjects workingon interdependentand
independenttasks. Groupsperformedrelativelypoorlywhen
hybridor individualrewards, ratherthan group rewards,were
given for performanceon the highlyinterdependenttask.
The authors concluded that even a small proportionof
individualrewards underminesperformanceof a group task.
Even less is known about what happens when hybrid
outcomes are combined with a hybridtask. The few studies
that have simultaneouslymanipulatedtask and outcome
interdependencegenerallysuggest that performanceis best
when both tasks and rewards cue the same level of
interdependence(Millerand Hamblin,1963; Wageman and
Baker,1994). To examine the impact of hybridtasks and
outcomes on group effectiveness, then, it may be most
useful to contrast the effectiveness of the three congruent
designs: group tasks and group outcomes, individualtasks
and individualoutcomes, and hybridtasks and hybrid
outcomes.
There are at least three distinct potentialeffects for hybrid
forms. First,hybridtasks and outcomes may additively
combine the basic effects of group and individualdesigns.
Logically,for example, one might expect the level of
cooperationin a hybridtask to be greater than that for
individualtasks. At the same time, because they cue
significantindependentaction as partof the work, hybrids
should result in less cooperationthan group tasks. The same
logic might be appliedto hybridrewards.The individualpart
of a hybridrewardmay result in clear relationshipsbetween
individualaction and rewards (andthus strong motivation);at
the same time, the collective partof the rewardsignals
strong dependence on the actions of others for outcomes,
thus obscuringthe relationshipbetween individualactions
and rewards (and underminingmotivation)(Vroom,1964).
Hybridforms of interdependencethus may show effects on
groups and their members that are intermediatebetween
pure group and pure individualforms.
Second, hybridforms may capturethe best of both designs
without their respective disadvantages, creating positive
151/IASQ,March 1995

synergy. Hybridtasks may afford the benefits of high
interdependence: Team members can learn from each other,
develop norms that promote high cooperation, and arrive at
more creative solutions to work problems than they might
have working independently (Maier, 1983). At the same
time, hybrid tasks may afford the benefits of independent
designs, such as giving each performer high personal control
over some portion of the work. Hybridtasks also may be
productive and satisfying both for people who are
comfortable in groups and people with strong preferences to
work autonomously.
Finally, hybrid forms may result in negative synergy,
capturing the disadvantages of group and individual work
without their respective benefits. By allowing members to
concentrate energy on individualtasks, hybrid tasks may
prevent members from developing the collective processes
that facilitate group achievement. At the same time, by
making members partiallydependent on each other's actions
for important outcomes, hybrid tasks and rewards may
undermine the sense of personal control that comes from
individual designs (Bird and Brame, 1978). Moreover,
because they demand both interdependence and
independence, hybrid tasks may be equally unpleasant for
people who prefer autonomous work and those who are
most productive in highly interdependent settings. I examine
this problem through the following research question:
Research Question 4: How do people performunder hybridforms
of interdependencerelativeto pure groupand individualforms?
Individual Differences
Some individual differences influence behavior across a wide
variety of situations, while others are activated only by
particularsituations (Price, 1974). One such individual
difference that surely is activated by interdependence is the
degree to which members have strong preferences to work
independently. I call these autonomy preferences and use
the term to refer to the degree to which people dislike and
avoid settings in which they must depend on others.
Preference, in this sense, is conceptually distinct both from
personality traits and from transitory states. A prototypical
trait is a characteristic that is stable over time, resides within
the person, and may be observed across different situations,
while a prototypical state refers to a characteristic of a
person that is caused by the immediate situation and is of
short duration (Zuckerman, 1983). Many personal
characteristics belong to neither category (Chaplin, John, and
Goldberg, 1988), and preferences are among these: They
are more stable than mere transitory states but can be
influenced by experience. Because preferences are
constructed through multiple experiences of positive
outcomes, they are stable enough to condition how people
respond to a particularsituation, but they can be altered over
time in response to counterpreference experience (Breer and
Locke, 1965).
The effects of task and outcome interdependence on work
satisfaction and on group processes and performance may
be moderated by the autonomy preferences of group
members. People with strong autonomy preferences are
152/ASQ, March 1995

Interdependence
attracted to situations in which they can work alone and

control their own work pace (Birch and Veroff, 1966). Their
performance suffers when they cannot determine the timing
and sequence of their own tasks (Vroom, 1959). Thus
individual differences in preference for autonomy should
moderate the impact of high or low interdependence on
performers. Operating in highly interdependent groups, for
example, often requires high levels of interaction and close
coordination of members' actions in timing and sequence.
Members of such groups depend on one another to achieve
important outcomes. For people who have strong autonomy
preferences, therefore, group work can be constraining and
frustrating and may undermine their satisfaction with their
work. They may avoid working collaboratively and
concentrate on aspects of the task that allow them to work
alone, free of the constraints of others. When members
prefer autonomy, the level of interaction may be low,
resulting in weak norms regulating members' behavior and
low levels of learning from others.
Some people with strong autonomy preferences may have
accumulated relatively little experience with teamwork in
their work lives. If they were to experience a highly
interdependent group, their preferences might be altered.
Placed in a group with others who are comfortable with
highly interdependent work, people with strong autonomy
preferences might gradually come to tolerate and be
effective in such settings as well (Breer and Locke, 1965).
Over time, then, a highly task-interdependent group with
group-based rewards would produce more of a decrease in
autonomy preferences than any other combination of task
and outcome interdependence. Thus autonomy preferences
may moderate the effects of task interdependence on
motivation and satisfaction, and they may also be changed
as a result of experience with interdependence. I examine
the effects of preferences for autonomy with the following
research question:
Research Question 5: What are the jointeffects of
interdependenceand individualdifferences in preference for
autonomyon groups and their members?
Research Strategy
To address the research questions, I conducted a
longitudinal quasi-experimental field study with a selection of
existing groups that had individual, hybrid, and group tasks.
The organizational reward system was then altered such that
one third of the groups received rewards contingent on
group, individual, or both group and individual performance,
respectively (see Figure 1, below, for the research design).
In addition, I assessed individual differences in preference
for autonomy at work. Thus, of the three factors in the
design, one was a selection factor (task interdependence),
one was manipulated (outcome interdependence), and one
was a covariate (individualdifferences in preference for
autonomy). Surveys, observation, and archival sources
provided measures of the dependent variables, which
included group performance, group norms and interpersonal
processes, and individualsatisfaction, motivation,and
learning.
153/ASQ, March 1995

METHOD
Overview
Over 800 service technicians (152 groups) at Xerox
Corporation participated in the research. A pre-intervention
survey provided baseline measures of group and individual
performance, behaviors, and perceptions. First-line managers
were interviewed to assess task interdependence for the
participating groups. The outcome interdependence
intervention was then implemented, creating the three
different reward conditions. When the intervention had been
in place for four months, the survey was reissued, and
archival data gathered about post-intervention performance.
Field observations during the four months after the
intervention provided qualitative data that elaborated the
quantitative findings. Finally, individual differences were
reassessed eight months after the intervention.
Research Site
The research was conducted in the U.S. Customer Services
division of Xerox Corporation. The division employs more
than 15,000 men and women, of whom about 12,000 are
technicians who repair machines. The service organization is
divided geographically into nine areas, which are, in turn,
subdivided into districts. Each district is then broken into
approximately eight to ten subdistricts-some on the basis
of geographical distribution, others on the types of machines
serviced. At the head of each subdistrict is a field manager,
to whom 15-30 service technicians report. These
technicians are organized into work groups of between three
and nine individuals, forming, on average, between two and
five work groups per field manager.
Task. Service technicians repair complex machines, they
respond to customer calls about machine breakdowns, and,
to a lesser extent, initiate visits to customer sites for
preventive maintenance. Customer calls come in through a
central dispatching office that technicians call throughout the
day for repair calls that have come in. Technicians prioritize
calls based on the seriousness of the problem and the type
of machine or customer. Because replaced machine parts
are paid for by Xerox and not billed to customers, one
aspect of the technicians' task that is critical to the
organization is efficient use of parts. Technicians make their
own decisions about when to replace and when to repair
parts and which parts to replace preventively.
There is a great deal of variability in the amount and type of
maintenance different machines require, which is reflected
in the level of technicians' experience, training, and salary
and in the number of machines technicians are called upon
to service. Still, all technicians have a high level of control
over their work in many areas, including their work pace, the
order in which they handle customers, their strategies for
minimizing parts expenses, and their maintenance practices.
For all groups, excellence means providing quick, high-quality
service to customers while minimizing parts expenditures.
Outcomes. There are a number of important outcomes for
technicians: ongoing feedback about performance, merit
increases based on performanceappraisals,a gainsharing
154/ASQ, March 1995

Interdependence
program,and district-levelmonetaryperformance-contingent
rewards.Techniciansreceive periodicfeedback about many
aspects of their performance.Typically,groups receive
monthlygraphs of their response times, their call rate, their
parts expenditures,and reportsof any machines performing
especially poorly.They also receive customer satisfaction
surveys. Some managers have begun providinggroup-level
performancegoals and feedback, and some provideno
individual-leveldata at all to their groups. Informalfeedback,
from machines and customers, accrues almost exclusively to
individuals.Thus, the pre-interventionlevel of
interdependenceof goals and of performancefeedback
about goal attainmentvariedto some extent but was largely
individual.
Yearlysalaryincreases were based on a performance
appraisalconsisting of an interimreview in June and an
end-of-yearratingthat determinedthe technicians' merit
increase in salaryfor the subsequent year. Managers rated
technicians on their individualperformancein a numberof
areas, includingprofessionalismwith customers,
achievement of performancetargets (e.g., "92.5 percent
customer satisfaction"),and teamwork. In some cases, field
managers were experimentingwith group performance
appraisals;most, however, were based 50 percent or less
on group performance.Thus the level of interdependencefor
merit increases was generallyquite low, with some cases of
hybriddesigns and a few cases of group designs.
Three furthersources of rewards exist. First,all districtsare
participantsin a corporategainsharingprogramthat
distributesquarterlybonuses to technicians based on their
district'sperformance.In addition,a yearly President's Club
Award, usuallyconsisting of a four-to-five-dayvacationtripat
the company's expense, is awarded to a fixed percentage of
technicians in each district.President's ClubAwards accrue
almost exclusively to individualsbased on individual
performance.A few districts have set aside enough
individualallocationsto give an awardto an entire work
group. Other managers have redesigned the award to allow
all districtmembers to participatein a weekend outing, thus
removingthe performancecontingency.
The finalsource of rewardswas a fund availableto field
managers, to be used at their discretionfor rewards.
Discretionaryfunds allocatedto field managers were most
typicallyused noncontingentlyon end-of-yearevents, such
as a picnicor Christmasparty.More than 50 percent of
groups for which pre-interventionrewarddata were available
(N = 48) received some of these funds in the form of a
social event that was exclusively noncontingenton
performance.
Sampling Strategy
The sample selection process was designed to maximizethe
commitment of participatingdistrictmanagers to the
research project,increasingthe chances that the outcome
interdependenceinterventionwould be thoroughly
implemented. Only managers who expressed strong interest
in participatingand were willingto alter the ways rewards
were distributedin their districtswere included.Twenty-four
155/ASQ, March 1995

senior managers and their staffs in the customer services
organizationwere asked to nominatedistricts in which the
districtmanagerwould welcome new ideas about reward
strategies and therefore would be likelyto participate.Of the
first eight districtmanagers interviewed,five agreed to
participate,two in the group-rewardconditionand three in
the hybridcondition.None of them expressed interest in the
individual-reward condition.Consequently,two additional
districtmanagers were invitedexplicitlyto participatein the
individualcondition,and both accepted. While this selection
strategy did not allow randomassignment of districtsto
rewardconditions, it did substantiallyincrease the chances
of producinglarge differences among rewardconditions by
maximizingthe managers' motivationto participatein the
intended changes. Further,this process did not mean that
managers would be using more of the same strategies
alreadyin use; the outcome interdependence manipulation
producedsubstantialpre- and post-interventiondifferences in
rewardsystem design, as detailed below.
Withinany one district,all field managers participatedin the
same outcome condition.Two districtsparticipatedin the
group outcome interdependencecondition,two in the
individualcondition,and three in the hybridcondition.The
seven districtswere distributedthroughoutthe eastern half
of the United States. Average tenure, percentage of women,
and percentage of minoritiesdid not differamong outcome
conditions,althoughthey did vary among districts. Mean
group size was 6.2 and did not vary by outcome condition. In
all, 60 groups (353 technicians)participatedin the
group-outcomecondition,77 groups (398 technicians)in the
hybrid-outcomecondition,and 55 groups (369 technicians)in
the individual-outcomecondition.
Independence Variables
Taskinterdependence. Groupsof technicians, who were
historicallyhighlyindependent, had been created in 1984,
but the extent to which these so-called teams actually
operated as such variedwidely. In many instances, the work
remainedan individualtask, with individualcustomer
responsibilitiesand no collective decision makingor action.
Manygroups had hybridtasks that entailed some significant
individualresponsibilitiesbut also some significantgroup
decisions and actions. Finally,many groups had no individual
assignments but shared responsibilityequallyamong
themselves for repaircalls and collectivelymade decisions
about their work processes. Thus individual,hybrid,and
group tasks were all represented in the organizationpriorto
the start of the research. I categorizedthe level of task
interdependenceof the participatinggroups in two stages: I
first assigned groups to categories based on a range of
descriptionsof task features by managers, and then I did a
detailed inspection of data about the groups' tasks and
adjusted the categorizations,where necessary, to fit the
realityof the groups' work.
Initialcategorization. I collected data about group task
interdependenceby interviewingeach first-linemanager
about their groups' tasks. Managersdescribed their own
instructionsto groups on three dimensions: (1) whether
156/ASQ, March 1995

Interdependence
groups managed their parts expenses collectivelyor

independently,(2) types of decisions made as a group, and
(3) frequency of requiredmeetings. They also described
other task inputs such as (1) the extent to which group
members were trainedto work on the same products; (2)
the complexityof productsassigned to the group; (3)
groups' geographicaldispersion (measured in maximum
minutes of travelacross the group territory);(4) the
communicationstechnology availableto members (e.g.,
pagers or voice mail);and (5) how repaircalls were queued
for the group-whether they were assigned to specific
individualsor to the group as a whole.
To create the provisionalmeasure of task interdependence, I
coded each of these eight task dimensions as low, mixed, or
high interdependenceand assigned values from 1 (individual)
to 3 (group).The exception was meeting frequency, which I
scored simply as the numberof requiredmeetings each
group held per month. I standardizedeach dimension across
groups to eliminatedifferences in variancedue to scaling;
these standardizedvariableswere summed to produce the
provisionalestimate. Then, to determine which of these
eight variablesmost stronglydifferentiatedamong groups,
the unstandardizedvariableswere entered into a
discriminantanalysis on the overallinterdependencescore.
The five variablesthat emerged from the discriminant
analysis as the most powerfulpredictorsof task
interdependenceare described in detail below.1
Repaircall queuing. In some groups, technicians took repair
calls in orderof urgency, in any member's nominalterritory.
Thus they were highlyinterdependentfor the care of
customers in the collective territory.In other groups,
managers maintainedrigidindividualterritories,and individual
members worked on their own machines only (an
independentstrategy).There were few moderately
interdependentstrategies, though in some groups members
shared collective responsibilityfor repaircalls but maintained
individualresponsibilityfor communicationwith customers to
determine the natureof the problemand give the customer
an estimate of the arrivaltime of a technician. In almost all
cases, techniciansworked on the machines individually.
Thus interdependencein machine care varied mostly in
terms of how customer responsibilitieswere defined for the
groups and the extent to which group members were
assigned collective responsibilityfor respondingto repair
calls from all the group's customers.
Parts expenses. Some managers promoted interdependence
by requiringthe group to manage parts expenses for the
I group as a whole. Other managers providedonly
The discriminantanalysisproducedtwo individual-levelbudgets and data about individual
canonicaldiscriminantfunctions,only the expenditures,thus treatingparts usage as an individualtask.
firstwith an eigenvalueabove 1. The
second (nonsignificant) function
comprisedgeography,product Maintenancepractices. Technicianswere requiredto design
complexity,and communications and implement a standardset of maintenancepractices to
technology;each correlatedpoorly(<.10) be followed at every service call. Standardmaintenance
with the first function.In addition,
geographyand communications practices includeddecisions about how frequentlyto replace
technologyprovedto have relativelylittle certainparts; which parts of the machine to clean at each
variance.Forthese reasons, these three visit; how to informthe customer about what was done to
variableswere droppedfromthe task
interdependencemeasure. the machine; and how to test the machine's functioning
157/ASQ, March 1995

before departing.Manygroups developed these practices for
the group as a whole. In addition,some groups developed a
method for "auditing"each other to determine members'
complianceto the agreed-uponmaintenance practices. Thus
they both designed and inspected their maintenance
practices as groups. In many groups in which technicians
were more independent, individualsdetermined their own
maintenancepracticeswithout consultingthe group.
Cross-training.Some groups had members who worked on
many differentproducts;others did not. To the extent that
members were trainedto repairsimilarmachines, they
shared a common set-of problems and a language for
discussing them, which promoted interdependence in the
work. By contrast, groups whose members were trainedon
differentproducts had no such commonality.The degree of
cross-trainingalso significantlyinfluencedthe group's
queuing strategy: Groupswhose members worked on
differentproductlines could not share collective
responsibilityfor the group's machines.
Meetings. The frequency of formalmeetings of entire work
groups (as opposed to informalgatheringsof two or three
technicians)variedfrom never to once a week. Some
managers promotedtask interdependenceby requiring
weekly or biweekly meetings at which members discussed
group decisions and areas of collective responsibility.Other
groups were not requiredto meet and either did not meet at
all or met only rarely.
I obtained a continuous measure of task interdependence by
takingthe sum of the standardizedscores on the five
variablesthat emerged from the discriminantanalysis. To
produce the provisionalmeasure of task interdependence-
individual,hybrid,and group levels-the distributioncreated
by summing the standardizedscores on the five variables
was then trichotomizedat naturalbreakpoints.
Categorization adjustments. I examined groups in each
category of task interdependencequalitativelyto assure that
the empiricalmeasure trulyreflected the conceptual
definitionsof individual,hybrid,and group tasks. A detailed
inspection of the task descriptionsof groups in each
category showed that only 6.5 percent of all groups had
been misclassified by the empiricalprocess. Allwere groups
that had fallen into the hybridcategory but were properly
associated with the individualtask condition.The
misclassified groups had significanttrainingand time
resources invested in off-lineactivities (meeting time and
cross-training),but their day-to-daywork was designed to be
executed entirelyindependently.
A typicalgroup in the hybridcondition(N = 51) had high
levels of interdependencearounda set of tasks-for
example, they designed their maintenancepractices as a
collective, met frequently,and managed parts expenses for
the group as a whole-and low interdependencearound
others (e.g., members worked on widely differingproducts
and were responsible for machines and customers only in
their own individualterritories).Thus members of hybrid
groups sometimes operated solo and sometimes in groups.
By contrast, members of a typicalgroup in the individual
158/ASQ, March 1995

Interdependence
task condition(N = 56) operated almost exclusively solo:

They cared for machines and customers in their own
individualterritories,designed their maintenancepractices
without group input,and managed parts expenses for their
machines alone, not as partof a collective. Members met
occasionallyas a group, and in many groups were required
to do so, but their actual day-to-daywork was not
accomplishedwith any reference to other group members.
Finally,members of a typicalgroup in the group task
condition(N = 45) designed their maintenancepractices
collectivelyand monitoredmembers' compliancewith those
practices; they responded to calls from any of the group's
customers, often activelyconsultingabout which member
should handle a particularcall; and they managed their parts
expenses as a group. Thus all the significanttasks of group
members in the group task conditionwere conducted
collectively.
Outcome interdependence manipulation. Unliketask
interdependence,the full range of levels of outcome
interdependencedid not exist priorto the start of the
research. The purpose of the interventionwas to establish
three distinctforms of outcome interdependence,thus
allowingthe creationof the full 3 x 3 design. In the group
outcome condition,managers providedoutcomes to
technicians contingent on group behaviorand performance;
in the hybridcondition,managers providedoutcomes to
technicians contingent on both group and individualbehavior
and performance;in the individualcondition,managers
providedoutcomes to technicians contingent exclusively on
individualbehaviorand performance.To structurethe
outcome interdependenceintervention,I trainedmanagers
to provideongoing reinforcement,helped set up a public
recognitionprogram,and worked with managers to make
performanceappraisalsand awards consistent with the
rewardconditionof each district.
First-linemanager trainingin ongoing rewardpractices. In
each district,I conducted one-day trainingsessions in
"plannedspontaneous" rewardpractices, developed
specificallyfor this study. In the course of this training,field
managers learnedways to use their already-existing
discretionaryfunds to provideongoing reinforcementto
groups and/orindividuals,contingent on specific behaviors.
The trainingprovidedmanagers with tools to plan their
opportunitiesto observe their technicians; it involved
exercises that helped them identifyspecific behavioralsigns
that technicianswere managingtheir own performance;and
it allowed the managers to practiceprovidingintangible
rewards such as verbalfeedback. Managersdeveloped
monthlyplans for deliveringrewardsto groups and/ortheir
members. Fieldmanagers in the individualconditionwere
instructedto dispense rewardsto individualmembers,
contingent on individualbehavior.In the group condition,
field managers were trainedto rewardthe group as a whole
contingent on group behavior.Finally,field managers in the
hybridconditionwere trainedto providerewards at times to
individualsand at times to entire groups. Field managers
recordedall rewardsdeliveredeach month on a special
research form and sent these records to me.
159/ASQ, March 1995

Public recognition program. Participating district managers
and I jointlydesigned a publicrecognitionprogramto
institutewithin the districts. Districtmanagers provided
publicrecognitionof groups and/orindividualsfor
outstandingaccomplishments and contributionsto the
districtas a whole. Managersin the group conditionpublicly
recognizedentire groups; managers in the hybridcondition
used half their resources to recognize groups and half to
recognize individuals;managers in the individualcondition
publiclyrecognized individuals.Examplesof recognized
accomplishments and contributionsincluded(1) a technician
volunteeringto take oina special projectin the district,(2) a
work group offeringone of its members to another group
with an unmanageablyhigh call rate, and (3) an individual
takingover the responsibilitiesof an absent field manager.
Districtmanagers recognizedsuch contributionsin their
monthlynewsletter, and these groups and individuals
received some tangible reward,such as a gift certificate
presented to them before their peers at a districtwideevent.
Performance appraisals and major awards. Performance
data, performanceappraisals,and cash achievement awards
not funded by the districts'discretionaryrecognitionbudget
were also made as consistent as possible with the reward
conditionof each district.Because performanceappraisals
are year-endevents and did not occur duringthe course of
this study, they are unlikelyto be a significantinfluence on
the overallexperienced level of rewardinterdependence. In
the group and hybridconditions, however, all technicians
were providedwith group-leveldata about customer
satisfaction and machine reliability,and the planned
performanceappraisalratingswere based in part (and in
some cases, entirely)on group performance.
In all cases, the interventionproducedsignificantdifferences
in the rewardpractices of participatingdistricts. Funds
previouslydistributednoncontingentlywere distributed
contingent on performance,in accord with the outcome
interdependenceconditionof the district;publicrecognition
programswere created where none existed before; and field
managers providedfrequent rewardsto groups and/ortheir
members that had not been providedbefore. Thus the
trainingmanipulationdid alter how managers distributed
rewards, and it created three distinct levels of outcome
interdependence-group, hybrid,and individual.
Individualpreference for autonomy. Preference for autonomy
was assessed using a scale I developed (see AppendixA for
items). To distinguishbetween more general measures of
need for autonomy, I call this scale "preferencefor
autonomy at work." Technicianscompleted the
five-item-scalemeasure twice, once priorto the outcome
interdependence interventionand again when the
interventionhad been in place for eight months.
Dependent Variables
Groupeffectiveness was assessed using (1) archivaldata
about group performance;(2) survey and archivalmeasures
of the degree to which members interactin ways that
increase the chances of workingmore effectively over
time-including member learning,qualityof interaction
160/ASQ, March 1995

Interdependence
processes, and group norms; and (3) survey measures of

individualwork motivationand satisfaction.
Sources of data. Archivaldata were collected for the four
months followingthe rewardintervention.These monthly
assessments of technicians'performancewere combined
across the four months to produce a single post-intervention
measure. In some cases, performancedata were available
only at the individuallevel. When it was necessary to
compute group-levelmeasures, individualdata were
combined within work groups, either by averagingindividual
data (e.g., average response time) or by combiningraw
numbers to calculate proportions(e.g., percent of customers
satisfied).
Both pre- and post-interventionsurvey data were collected.
Surveys were distributedone month priorto the intervention
and again four months after the interventionbegan.
Response rates were as follows: 816 technicians (73
percent) completed the first survey; 573 (51 percent)
completed the second.2 The survey requiredabout twenty
minutes and includedmeasures of both group-and
individual-levelvariables.Group-levelmeasures assessed
characteristicsof the group as a whole, includingnorms and
the qualityof group processes. Individual-level variables
described individualperceptions and affective reactions, such
as motivationand work satisfaction.To create group-level
measures, individuals'responses were averaged within work
groups. To ensure that these aggregated scores were
meaningfulgroup-levelmeasures, intraclasscorrelations
were computed for all group-levelvariables;all were
significantat p < .01 or better (ICCsrangedfrom .19 to .49).
I collected observationaldata over the course of four
months, beginningtwo months after the outcome
interventionwas in place, and coded it before conducting
quantitativeanalyses. I selected 17 groups to observe to
explore the dynamics of the nine differentcombinationsof
task and outcome interdependence. Foreach group
observed, I first attended a regularlyscheduled work-group
meeting. Followingeach meeting, I spent the remainderof
the day travellingwith a technicianfrom that group to each
of his or her service calls. Duringthis time, I took detailed
notes on such items as the technician's behaviorwith
customers, the repairand maintenanceproceduresfollowed
by the technician(forexample, in orderto determine the
extent to which the technicianconformedto established
group procedures),and communicationsamong group
members. I used these data to help elaboratethe
quantitativefindings.
In addition,I interviewedtechnicians about variousaspects
of the group and its work. Forexample, I asked technicians
which work processes had been designed by group
consensus and which had been left to the individual
technician'sdiscretion. I asked them to describe any
rewardsthey had received from the field manager duringthe
preceding month. These questions served, in conjunction
2 with more formalmeasures, as manipulationchecks.
Therewere no significantdifferencesin
response rate by task or outcome Measures. Each dependent variableand its data source is
interdependenceconditions. described below. Group-levelmeasures includegroup
161/ASQ, March1995

performance,group norms, and the qualityof member
interactions;individual-level variablesinclude individuals'
experienced interdependence,their work satisfaction,
internalwork motivation,and amount of learningfrom other
group members. Table 1 provides means and standard
deviationsfor all dependent variablesand internal
consistency reliabilities(Cronbach'salpha)and discriminant
validities(averagecorrelationswithin scale versus average
correlationwith relatedscales) for all survey scales.
Groupperformance.Groupperformancewas assessed using
five measures obtainedfrom company archives.The first
was a measure of customer satisfactionwith machine
reliability,derivedfrom a customer survey distributed
periodicallyby the organization.Based on their survey
responses, customers were coded by the organizationas
"satisfied"or "not satisfied." Customersatisfaction
performancewas the percent of the group's customers
surveyed in the four months after the interventioncoded by
the organizationas "satisfied."This measure was included
only for groups that had received at least 10 surveys during
the four post-interventionmonths (67 percent).
A second measure of group performancewas the cost of
parts used in maintainingmachines. Xerox provides an
expected level of parts expendituresto groups/individuals
based on the type of machine and the historicalvolume of
usage on each machine. Parts expense is then measured as
expected expendituresdividedby actual expenditures.
Three additionalmeasures of performanceconcerned the
qualityof the groups' repairpractices: response time, repair
time, and machine reliability.Response time is the
percentage of calls taken by group members that falls within
specified time limits. It is a measure of how well technicians
are doing at getting to their customers quicklyenough to
satisfy them. Repairtime is assessed as the amount of
Table 1
Descriptive Statistics
Cronbach's r within r with

Variable Mean S.D. alpha scale related scales
Preference for autonomy 3.21 .90 .78 .38 .09
Performance measures
Customer satisfaction 90.74 14.56
Parts expenses 111.35 22.71
Response time 86.73 7.09
Repair time 102.61 56.43
Machine reliability 101.53 12.47
Cooperation norms .72 .41 .74 .36 .25

Effort norms .50 .44 .63 .31 .17
Quality of group processes 4.75 .86 .85 .46 .25
Time out of territory 40.45 11.88
Experienced task interdependence 5.10 .95 .76 .40 .22
Experienced outcome interdependence 4.60 1.07 .73 .35 .30
General work satisfaction 5.28 1.15 .65 .32 .10
Work motivation 5.55 .86 .79 .56 .34
Learning 5.53 .98 .76 .51 .37
Note: Scale statistics reported are for time 1; scale reliabilities and discriminant validities did not differ between time
1 and time 2.
162/ASQ, March 1995

Interdependence
repair time a machine is expected to require (based on

historical data), divided by the actual number of minutes
spent repairing a machine. Repair time is a measure of
technicians' technical proficiency-how quickly they can fix
problems. Machine reliability is measured as the number of
repair calls expected on a machine (again, based on historical
data) divided by the actual number of repair calls. Machine
reliabilityassesses how well the machines are repaired and
maintained. Expected levels for response time, repair time,
and machine reliabilityall vary with the type of machine and
the volume of usage on each machine. By dividing expected
levels (derived from national averages for each machine
type) by actual levels, measures are produced that assess
performance as a proportion of national averages and are
reasonably comparable across groups that differ widely in
the types of products they maintain. To create a measure of
overall performance, the cells of the design were ranked,
from best to worst, on all five performance measures.
Group norms. Survey data assessed the strength and

direction of group norms about cooperation and effort.
Technicians were asked to describe, on a scale from -2
("group strongly discourages") to +2 ("group strongly
supports") their group's expectations about a number of
behaviors, including acts of cooperation (or refusal to
cooperate), and actions that represent high (and low) levels
of effort (see Appendix A for items). The mean of group
members' ratings served as a measure of norm direction
and strength (i.e., a mean of 1.5 represents stronger
pro-cooperation norms than a mean of .5), while the inverse
of the variance of group members' ratings served as a
measure of consensus among group members (consistent
with Jackson, 1965).
Quality of group process. A seven-item survey scale

assessed group members' perceptions that group members
were working together in ways that increased their
effectiveness over time (Allmendinger et al., 1992).
Technicians rated, on a scale ranging from 1 ("strongly
disagree") to 7 ("strongly agree") their agreement with
statements describing positive and negative group
processes, such as "Every time someone tries to straighten
out a work group member whose behavior is not acceptable,
things seem to get worse rather than better." A second
measure of group process was taken from company archival
data. Xerox keeps records of which technician takes each
repair call and to which the machine is officially assigned.
Using these data, I constructed a measure of the percentage
of calls taken within groups by technicians outside their own
official territories. This measure provides a behavioral
assessment of how much group members help each other
in the course of their work.
Experienced interdependence. Two survey scales assessed

respondents' experienced task and outcome
interdependence. A five-item survey scale asked members
to assess how much they needed other members'
resources and actions to complete their task (e.g., "My work
is not done untileveryone in the group has done his or her
163/ASQ, March 1995

part.").A second five-item measure assessed technicians'
beliefs that contingent outcomes were influencedby the
actions and outcomes of their fellow group members (e.g.,
"My teammates' performanceaffects my rewards.").While
correlated,these two scales have good internalconsistency
reliabilities;moreover,the average intercorrelationamong
items within scale is substantiallyhigherthan with items
from the other scale (see Table 1). Consequently,they were
retainedas two distinct measures of experienced
interdependence.
Individualsatisfaction, motivation, and learning. Three survey
items assessed technicians'overallsatisfactionwith their
work (e.g., "Generallyspeaking, I am very satisfied with this
job."),and three measured the extent of their internal
motivationto performwell at their work (e.g., "My opinion
of myself goes up when I do this job well.") (Allmendinger
et al., 1992). A finalthree-itemscale asked group members
to assess the extent to which they learnedskills and
strategies from each other at work (e.g., "I learnthings in
my group that make me a better technician.").
Data Analysis
Analyses were conducted using the SAS GeneralLinear
Models (GLM)program.Because individualsubjects are
nested within groups, groups within field managers, and
field managers within districts, hierarchicalmodels were
used, and these differedslightlyfor group-leveland
individual-levelanalyses. Forgroup-levelvariables,GLMs
were runentering task condition,outcome condition,the
interactionof task and outcome, district,and field manager,
in that order.The errorterm used for testing higher-level
effects was thus the mean squared errorfor field manager
within district.Forindividual-leveldependent variables,GLMs
(thatdid not include individuals'preference for autonomy)
were runentering task condition,outcome condition,the
interactionof task and outcome, district,field manager, and
group, in that order.The errorterm used for testing
higher-leveleffects was thus the mean squared errorfor
groupwithin field manager. Foranalyses that included
preference for autonomy, autonomywas entered as a
continuous measure, and groupwas droppedfrom the
design to avoid overspecificationof the model.
RESULTS
AppendixB presents the omnibus 3 x 3 analyses of
variancefor all majordependent variables,showing Fs and
significance levels for the main effects of task
interdependenceconditionand outcome interdependence
condition,and their interaction.
Manipulation Checks
The followingtwo sets of analyses assess whether (1) the
measure of task interdependencederivedfrom managers'
descriptionspredictedreal differences in members' task
behavior,and (2) the experimentalmanipulationof outcomes
resulted in the intended levels of outcome interdependence.
Task interdependence. Highlyinterdependentgroups were
those described as having(among other features) a
164/ASQ, March 1995

Interdependence
call-queuingprocess that placed responsibilityfor all

machines on all group members, whereas less
interdependentgroups had more assignment of individual
responsibilityfor machines in specified territories.
Consequently, I expected to find that technicians in highly
task-interdependentgroups took a significantlygreater
numberof repaircalls in other members' nominalterritories
than did members of less interdependentgroups.
A GLMwas performedon the mean percentage of calls
taken outside the technicians'own territories,averaged
across the four months before the task interdependence
data were collected. Members of groups in the group task
conditionspent significantlymore time out of their territory
(mean = 46.33) than did technicians in the hybridcondition
(mean = 39.71), who in turn spent significantlymore time
out of their territorythan did technicians in the individual
task condition(mean = 38.05), [F(2,44) = 3.26, p < .05].
Thus, the measure of task interdependencederived from
managers' descriptionsdoes predictan importantbehavioral
difference among groups.
Outcome interdependence. I used the monthlyworksheets
providedby field managers to count the number,type, and
target (i.e., group or individual)of rewardsthey administered
and thus assessed the percentage of the total numberof
rewardsthat each field managergave to groups versus
individuals.Managerswho participatedin the group outcome
conditiondelivereda mean of 99 percent of rewards
contingent on group behavior.Those who participatedin the
hybridconditiondelivereda mean of 48 percent to groups,
and those who participatedin the individualcondition
delivereda mean of 8 percent to groups and 92 percent to
individuals.
These results show that the manipulationhad its intended
effects. Nevertheless, it is useful to know something about
managers' behaviorpriorto the interventionto ensure that
any effects of outcome interdependencewere due to the
interventionratherthan to initialdifferences among manager
preferences. Interviewswith 26 of the 57 participating
managers priorto the interventionrevealed that slightly
more than 37 percent were spending their rewardfunds
noncontingently(e.g., technicianswere given gifts or invited
to parties regardless of their performanceor behavior).
Another40 percent were contingentlyrewardingindividual
behaviorexclusively; fewer than 23 percent were rewarding
group behaviorexclusively; and none of the managers was
using a hybridrewardstrategy. Forfield managers'
discretionaryrewards,the correlationbetween the
pre-interventionand post-interventionproportionof rewards
given to groups is .29. Clearly,the manipulationdid alter
how managers distributedrewards.
Still,althoughmanagers in the hybridconditiondid divide
their rewards roughlyevenly between groups and
individuals,they may have done so by providingonly
group-levelrewardsto some of their groups and only
individual-levelrewardsto others. Because this research
explores the effects of the level of outcome
interdependenceexperienced by groups, it is importantto
know how rewardswere distributedwithin groups.
l165/ASQ,March 1995

The worksheets again providedthe relevantdata. Foreach
group, I dividedthe total numberof rewardsdelivered to the
group as a whole by the total numberof rewardsdelivered
to members. Withinthe hybridtrainingcondition,the
proportionof group rewards received by each group varied
from 0 percent to 100 percent. While variancein the group
and individualoutcome conditionswas lower, a few groups
in the individualconditionreceived some group-level
rewards,and a few groups in the group conditionreceived
some individual-levelrewards.This means that some groups
reportingto managers in each trainingconditiondid not
experience the intended level of outcome interdependence.
Therefore,for the main analyses, I reassigned groups to
outcome conditions based on the rewardsgroups actually
received. Those groups who received 90 percent or more of
their rewardsas groups (N = 61) were includedin the group
condition,regardless of the traininginterventionin the
district.Groupsthat received fewer than 20 percent of their
rewardsas groups were assigned to the individualcondition
(N = 50); and groups that received between 30 percent and
85 percent of their rewardsfor group behaviorwere
assigned to the hybridcondition(N = 41). Only 12 percent
of groups were reassigned. For reassigned outcome
conditions,the mean percent of rewards deliveredfor group
behaviorwas 99.9 percent, 49.3 percent, and 0.4 percent
for group, hybrid,and individualconditions, respectively.
Havingestablished that the task conditions reflect real
differences in task interdependenceand that the reward
manipulationresulted in real changes in outcome
interdependence, I now turnto the research questions I
examined throughthe analyses.
Research Question 1: Effects of Tasks and Outcomes on
Experienced Interdependence
Effects of tasks on experienced interdependence. To
assess the influence of task interdependenceon
experienced interdependenceat time 2, two GLMswere
performedpredictingthe two dependent variablesfrom task
condition,outcome condition,and their interaction.The top
panel of Table 2 shows the results. The lineartrend for the
effect of task conditionon experienced task
interdependencewas significant;members of groups with
highlyinterdependenttasks reportedsignificantlygreater
experienced task interdependencethan did those in the
hybridtask condition;members of hybridtask groups, in
turn, reportedsignificantlygreater experienced task
interdependencethan did those in the individualtask
condition.Task conditionalso significantlyinfluenced
outcome interdependenceexperienced by groups,
independentof the actual outcome condition:Groupswith
highertask interdependenceexperienced their outcomes as
more interdependentthan did those whose tasks required
less interdependence.
Effects of outcomes on experienced interdependence.
Because outcome interdependencewas experimentally
manipulated,a more controlledtest of its effects on
experienced interdependencewas possible by assessing its
effects on changes in interdependencebetween time 1,
l166/ASQ,March 1995

Interdependence
prior to the start of the study, and time 2, four months after
the outcome intervention was implemented. The bottom
panel of Table 2 shows mean changes in experienced task
and outcome interdependence by outcome and task
conditions. Individual-rewardgroups showed no significant
change in experienced outcome interdependence; groups
that received group-level rewards showed a significantly
greater increase in experienced outcome interdependence
than those in the individual condition. Hybrid groups fell
between individual and group conditions. Thus the outcome
intervention did alter technicians' experienced outcome
interdependence. By contrast, Table 2 shows that the
manipulation of outcome interdependence did not alter
technicians' experienced task interdependence. Thus the
influences of task and outcome interdependence on
experiences of interdependence are not symmetric. Task
characteristics influenced experienced interdependence both
in the work and in important reward outcomes, while
outcome interdependence affected experienced outcome
interdependence but had no influence on how technicians
experienced their work.
Research Question 2: Effects of Total Interdependence on

Group Effectiveness
Total interdependence was calculated by summing task and

outcome interdependence (where individual = low, hybrid
= moderate, group = high) for each cell. Figure 1 shows
the total interdependence scores. Cells with the same total
interdependence score were combined. Table 3 presents the
effects of total interdependence at time 2 on group
Table2
Effects of Task and Outcome Conditions on Experienced
Interdependence at Time 2*
Experienced Experienced
task outcome
interdependence interdependence
Task
condition Mean S.D. Mean S.D.
Individual 4.73a .80 4.61b 1.07
Hybrid 4.93a .97 4.76c .90
Group 5.31a 1.04 5.24bc .92
F(2,44) = 14.85, p < .05 F(2,44) = 26.34, p < .05
Change in Change in
experienced experienced
outcome task
Outcome interdependencet interdependence
condition Mean S.D. Mean S.D.
Individual .15d .76 - .07 .78
Hybrid .30 1.04 .07 .80
Group .45 1.07 .05 .98
F(2,44) = 3.97, p < .05 F(2,44) = 1.01, n.s.
* Means withincolumns that share a superscriptdiffersignificantlyaccording
to Tukey'sHSD.
t Changein experiencedoutcome interdependencedifferssignificantlyfrom0
for the group and hybridoutcome conditions. For the individualoutcome
condition,it does not.
167/ASQ,March 1995

Figure 1. Research design.
TASK INTERDEPENDENCE
Individual Hybrid Group
A B C
Individual
Z Very Low Low Moderate
z
LAJI D E F
CLA
0 0
cc
a:Hybrid
ILU
Low Moderate High
G H
Group
Moderate High Very High
Note: Total interdependence scores are noted in the bottom left of each cell.
performance, processes, and norms, and individual

satisfaction, motivation, and learning. Total interdependence
does significantly influence all these aspects of group
functioning, but in different ways for different variables. For
some variables, the effect is best described as "more is
better." As Table 3 shows, cooperation norms [F(4,42) =
2.64, p < .051, self-reported quality of interpersonal
processes [F(4,42) = 2.80, p < .051, the degree to which
members learn from each other [F(4,42) = 3.38, p < .05]
and work satisfaction [F(4,561) = 1.43, p < .05] all increase
with total interdependence. For two variables, the
relationship is "less is better." Stronger effort norms [F(4,42)
= 3.42, p < .05] and greater work motivation [F(1,557) =
1.55, p < .05] are associated with very low degrees of
interdependence. Finally, two effects showed curvilinear
relationships. First, although the differences among mean
performance ranks are not significant, performance tends to
be better at very high and very low levels of
interdependence. Second, the level of cross-territory helping
is significantly greater when interdependence is either very
low or very high [F(4,42) = 3.97, p < .051. In no case was
any aspect of group effectiveness best at moderate rather
than extreme levels of interdependence.
Research Question 3: Differential Effects of Task and

Outcome Interdependence
The omnibus analyses of variance (reported in Appendix B)
show that task and outcome interdependence significantly
influence different aspects of group functioning. Table 4
presents the means of the effects for task and outcome
168/ASQ, March 1995

Interdependence
Table3
Effects of Total Interdependence on Group Effectiveness
Interdependence
Very low Low Moderate High Very high
Effects (S.D.) (S.D.) (S.D.) (S.D.) (S.D.)
"Moreis better"
Cooperationnorms* .62 (.39) .70 (.45) .65 (.53) .91 (.36) .98 (.39)
Qualityof group process* 4.67 (1.12) 4.64 (.97) 4.56 (.83) 4.93 (.94) 5.37 (.77)
Learning* 5.37 (1.04) 5.32 (.92) 5.36 (1.00) 5.48 (.80) 5.81 (.72)
Generalsatisfaction 5.22 (1.11) 5.22 (1.10) 5.25 (1.24) 5.42 (1.03) 5.32 (1.18)
"Less is better"
Effortnorms .91 (.49) .46 (.38) .56 (.43) .63 (.37) .69 (.40)
Workmotivation 5.59 (.92) 5.54 (.74) 5.54 (.82) 5.42 (.84) 5.45 (.99)
Curvilinear
Performancerankt 2.4 3.8 3.3 3.3 2.2
Time out of territoryt 42.7 (13.0) 37.7 (14.7) 40.1 (18.4) 40.3 (12.2) 49.9 (12.8)
* Lineartrend is significantat p < .05.
t 1 = high rank, 5 = low rank.
$ Verylow and very high interdependencediffersignificantlyfrom all others.
conditions on each dependent measure at Time 2. The

differential effects of task and outcome interdependence
help account for the patterns observed in the analyses of
total interdependence.3 The observed more-is-better effects
of total interdependence are due solely to main effects of
task interdependence (with the exception of effects on
learning, discussed below). The top panel of Table 4 shows
that groups with group tasks have stronger norms promoting
cooperation, higher quality group processes, and greater
member satisfaction with work than do groups with hybrid
or individual tasks. Main effects of outcome
interdependence for these variables were not significant.
The observed less-is-better effects are due, by contrast, to
main effects of outcome interdependence. The middle panel
of Table 4 shows that groups that receive individual
outcomes show significantly stronger norms promoting
effort and significantly higher work motivation than do
groups that receive either hybrid or group outcomes. In each
case, the main effect of task was nonsignificant.
The observed curvilinear relationship between total
interdependence and out-of-territory helping comes from two
significant effects (see Table 4, bottom panel): (1) an effect
of task, already described above, such that group tasks
result in significantly greater cross-territory helping than do
hybrid or individualtasks and (2) an effect of outcome
interdependence, such that relatively low levels of helping
occur under hybrid outcomes. The simple effects of task and
outcome conditions on learning parallel the findings for
out-of-territory helping (Table 4, bottom panel): High task
interdependence increases learning, and hybrid outcomes
undermine learning.
Finally, the bottom panel of Table 4 shows the main effects

3 of task and outcome interdependence on overall
Therewere no significanteffects on performance. GLMs performed on the five separate
normativeconsensus-the variance
among members' ratingsof norm performance measures indicated that both task and outcome
directionand strength. interdependence significantly influence some measures of
169/ASQ, March 1995

Table4
Means and Differential Effects of Task and Outcome Interdependence*
Task Condition Outcome Condition
Effects Individual Hybrid Group F Individual Hybrid Group F
"Moreis better"
Cooperationnorms .66a .76 1.01a 4.26 .88 .78 .75 2.64
(.47) (.45) (.39) (p < .05) (.40) (.54) (.43) (p = .08)
Qualityof groupprocess 4.45a 475b 5.24ab 7.32 4.94 4.70 4.80 1.50
(1.06) (,81) (.76) (p < .05) (.93) (.99) (.96) (n.s.)
Generalsatisfaction 5.12a 5.27 5.50a 3.96 5.42 5.22 5.26 1.95
(1.20) (1.11) (1.10) (p < .05) (1.06) (1.20) (1.20) (n.s.)
"Less is better"
Effortnorms .56 .64 .62 .02 .69cd .55C .59d 2.21
(.55) (.34) (.36) (n.s.) (.25) (.37) (.46) (p < .05)
Internalwork motivation 5.46 5.58 5.66 .98 5.66c 5.58 5.46c 3.74
(.84) (.83) (.90) (n.s.) (.82) (.84) (.90) (p < .05)
Curvilinear
Proportionof calls out 38.05a 39.71b 4633ab 7.97 42.19c 38.29cd 43.60d 2.53
of territory (16.97) (13.38) (12.85) (p < .05) (14.05) (12.03) (17.76) (p = .05)
Learning 5.20a 5.44b 5.65ab 14.45 5.51 c 5,30cd 5.49d 4.02
(1.04) (.91) (.75) (p < .05) (.88) (1.08) (.86) (p < .05)
Mean performancerankt 4.8 5.4 4.6 6.12 4.5 6.7 3.6 9.04
(p < .05) (p < .05)
* Means withinrows that share a superscriptdiffersignificantly
accordingto Tukey'sHSD.Standarddeviationsare in parentheses.
t Note: 1 = highrank,9 = low rank;allninecells were ranked,andmean rankstakenacross cells for maineffects. The numberreported
in the F columnfor performanceis a Kruskal-WallisxV.
performance,and a Kruskal-Wallis nonparametricanalysis of

varianceon the mean ranksconfirmsthis findingfor the
measure of overallperformance(see AppendixB). An
inspection of the mean ranksshows that the specific
effects detected are that both hybridtasks and hybrid
outcomes are associated with poor group performance
relativeto individualand group tasks and outcomes. Thus
the trend observed in the relationshipbetween total
interdependenceand performancebecomes clearer: Groups
with moderate levels of total interdependence have hybrid
tasks, hybridoutcomes, or both, and these are performing
much worse than those with either group or individual
designs.
Research Question 4: Effects of HybridDesigns
As seen in Table 4, both hybridtasks and outcomes lead to
poorerperformancethan do group and individualdesigns.
Moreover,hybridsalso undermineother measures of group
effectiveness-such as helping behaviorand
learning-relative to group and individualdesigns. It remains
possible, nonetheless, that congruence between the levels
of interdependenceof tasks and outcomes may be critical
for group performance(e.g., Wageman and Baker,1994).
Consequently,it may be that when hybridtasks are
combined with hybridoutcomes, groups performbetter than
when hybridtasks and outcomes are combined with group
or individualoutcomes and tasks.
Table 5 shows the performanceranksfor all nine cells of the
design. A Mann-WhitneyU test was performedto test the
hypothesis that the mean performanceof the congruent
cells (cells A, E, and I in Figure1) was significantlybetter
than that of the moderatelyand extremely incongruentcells.
170/ASQ, March 1995

Interdependence
Table5
Mean Rank on Performance Measures by Task and Outcome Condition
Task Interdependence
Outcome
Interdependence Individual Hybrid Group Total
Individual 3.6 5.6 4.3 4.5
Hybrid 7.6 6.6 5.9 6.7
Group 3.2 3.9 3.7 3.6
Total 4.8 5.4 4.6
Note: 1 = high rank, 9 = low rank.
This test was nonsignificant [U(3,6) = 7, p = .36],

consistent with the lack of significant interaction effects
found in the analysis reported in Appendix B. An inspection
of the mean ranks for all nine cells shows once again that
both hybrid tasks and hybrid outcomes, including the
congruent hybrid, produce poor performance relative to
group and individual tasks and outcomes. A Mann-Whitney U
test of ranks confirms this observation when the four corner
cells (cells A, C, G, and I in Figure 1) are compared with all
cells that contain either hybrid outcomes or a hybrid task.
This test is significant, U(4,5) = 1, p = .016, indicating that
pure group or individual outcomes, and pure group or
individual tasks in any combination lead to better
performance than all hybrid task and outcome designs.
Research Question 5: Individual Differences
To test the degree to which relationships between
interdependence and group effectiveness are moderated by
individual differences, preference for autonomy was entered
both separately and in interaction with task and outcome
interdependence. For group-level analyses, group
composition around preference for autonomy was entered
as the proportion of group members scoring in the top
quintile of the preference for autonomy scale.
Only one analysis of the moderating effects of autonomy
preferences showed a significant interaction with
interdependence. The analysis of work motivation indicated
that autonomy preferences interact significantly with
outcome interdependence (P = -.06, p < .05), such that
the decrement in work motivation due to group rewards is
even greater for those individuals who strongly prefer
autonomy. Beyond this one finding, there was no evidence
that individual differences in preference for autonomy
moderate the effects of interdependence on group
effectiveness [median F(2,38) = .45, n.s.]. This lack of
significant moderating effects is not due to problems with
the preference for autonomy measure per se, because there
were significant main effects of autonomy such that people
with high preferences for autonomy helped each other less
and learned less from others.
The final analysis examined the effects of task and outcome
interdependence on individuals' preferences. Because task
interdependence was a selection variable, any changes in
autonomy preferences from experience with group tasks
should alreadybe observable in the pre-intervention
171/ASQ, March 1995

preference for autonomy scores. The upper panel of Table 6
displays the preference for autonomy means by task
condition at time 1. This table shows that those individuals
who worked on group tasks show significantly lower
preferences for autonomy than individuals who worked on
hybrid tasks, who in turn show lower preferences than those
who worked on individualtasks.
To test the influence of outcome interdependence on
autonomy preferences, mean change in preference for
autonomy (post-intervention score minus pre-intervention
score) was broken down by outcome interdependence
condition. Autonomy preferences changed with the level of
interdependence in group outcomes. The lower panel of
Table 6 indicates that technicians who experienced individual
outcomes showed significantly greater increases in their
preference for autonomy than did technicians who
experienced hybrid outcomes. Technicians in groups that
received group outcomes showed the least change. Thus
preference for autonomy is itself influenced by the level of
interdependence in the task and in reward outcomes.
Table 6
Effects of Interdependence on IndividualDifferences in Preference
for Autonomy*
Preference for
autonomy
Condition Mean S.D.
Task interdependence
Group 3.02a .58
Hybrid 3.18a .64
Individual 3 35a .74
F(2,575) = 3.14, p < .05
Change in preference
for autonomy
Outcome interdependence
Group .26b .66
Hybrid .46b .85
Individual .69b ,74
F(2,311) = 1.82, p < .05
* Means withincolumnsthat share a superscriptdiffersignificantlyaccording
to Tukey'sHSD.
DISCUSSION
The findings of this research illuminate three aspects of the
interrelations among task interdependence, outcome
interdependence, and the characteristics of individuals in
influencing group behavior and performance: (1) the relative
potency of the three sources of influence, (2) the
comparative dynamics of pure individual, pure group, and
hybrid designs, and (3) how the interrelationships among
tasks, rewards, and individuals evolve over time.
Potency of Task, Outcomes, and Preferences
Individualdifferences in autonomy preferences did not
moderate personal reactionsto tasks and to rewards.
172/ASQ, March 1995

Interdependence
Rather, individuals' preferences came into congruence over

time with the kinds of tasks and rewards they experienced.
While people who preferred autonomy in their work did tend
to help each other less and to learn less from their fellow
group members, their work performance depended almost
entirely on the design of the work they did and the kinds of
rewards they received. This research shows that individual
differences are malleable: As people gain experience with
interdependence, they grow more accepting of it, and even
come to prefer it.
The design of the work had strong effects on cooperation,

helping, and learning, regardless of reward system design.
By contrast, group rewards had no independent influence on
cooperative behavior. Moreover, the design of the group
task strongly influenced members' perceptions of their
outcome interdependence; outcomes, by contrast, had no
such effect on experiences of the task.
Reward outcomes appear to affect the character of

members' motivation rather than to influence group behavior
directly. Collective rewards helped motivate highly
interdependent groups to perform well, whereas rewards for
individual excellence energized members of independent
groups, demonstrating that outcomes either enhance or
undermine motivation for the kinds of work behaviors that
are elicited by the task. The pivot, then, is how the work is
structured. Work design shapes individuals' preferences,
their behavior, how they experience their rewards, and the
impact of those rewards on their performance. Not all work,
perhaps, permits as much choice about how tasks are
designed as was the case for the service technicians in this
study. But whenever cooperative behavior is critical to
excellent task performance, it is most essential to create real
task interdependence and then support the task design with
interdependent rewards.
Individual, Group, and Hybrid Designs
Findings showed that the work performed by the technicians

in this study can be done well in two different ways:
independently and interdependently. Service technicians with
an individual work design are principally responsible for
machines in their own territories, they work alone, and they
have long worked in this way. With no one to share the
responsibility, these technicians develop a strong sense of
personal responsibility for how well their machines run. They
know their machines and customers extremely well, and
when they arrive at customer sites, they often know from
memory what preventive maintenance practices are due on
each machine. The individual reward system reinforces the
cues they get from their tasks and increases their work
motivation and efforts to improve their own performance.
Technicians operating individuallyalso reported a stronger
perceived link between their individual actions and positive
outcomes than did those working in group and hybrid
designs. Individualdesigns enhance norms that support high
levels of independent, rather than coordinated effort, as
shown in this exchange among members of a group in the
individualrewardconditionwith a highlyindividualistictask
173/ASQ, March 1995

design, who, at the behest of management, were tryingto
work out some collective work strategies:
Firsttechnician: If I have one call in the queue and I'm out on a
call, I don't want help unless I ask for it. I'llback you up if you
want, but it shouldn'tbe automatic.
Second technician: OK,then should we go throughall our group's
machines and just get retrofitsinstalledon all of them [regardless
of whose machine it is]?
Firsttechnician: No, I thinkit's the responsibilityof each tech rep
to take care of retrofitson his own machines duringregularcall
time. We shouldn'thave to do extra visits for that.
The group continuedthe discussion for an hour and never
agreed on anything.But if one looks at this group's
performance,its machine repairrecordis extraordinary.
Members were more than 20 percent faster than average at
fixing machines, their machine reliabilitywas well over
nationalstandards,and 100 percent of this group's
customers who were surveyed described themselves as
satisfied with service.
Highlyinterdependentgroups use the collective knowledge
and skills of their members to get their work done. They
exhibit high-qualitysocial processes, extensive mutual
learning,and a sense of collective responsibilityfor
performanceoutcomes. The demands of the task and cues
from the rewardsystem to operate as a collective were
potent enough to encourage strong norms supportinghigh
levels of cooperationand effort. Positive interpersonal
interactionsdeveloped most strongly in task-interdependent
groups, suggesting that such groups must develop
constructiveways of interactingto survive as teams. Group
outcomes may foster a different kindof motivationthan that
which fuels the independentprocess. Members of
interdependentgroups often talk about excellent
performancein collective terms. One technician reported
that when members of her group took calls on machines
that were assigned to other groups, they often felt they had
to clean up after others who had been sloppy in their care
practices. "Actually,we're pretty pleased with ourselves
about this. We were rated as 'role models' for our
preventivemaintenancepractices." Groupswith high task
interdependencesee their rewardsas highlyinterdependent:
Ratherthan a strong sense of personal responsibilityfor how
satisfied an individual'sown customers are, they feel
responsible to other group members for all the group's
customers: "There'sno 'mine' and 'yours'for us. We go
where we're needed, and we take care of each other."
Because technicians generallyrepairmachines alone, even
the highlyinterdependenttask in this setting does not
represent an extreme form of high task interdependence,
like the task of surgicalteams, which are reciprocally
interdependentand work together constantly (Steiner, 1972).
Nevertheless, by creatinginstructionsand processes that
cued groupwork, it was still possible to make the task
interdependentenough to fuel a collective process.
By contrast, a look at the dynamics of hybridgroups shows
that both the independentand the interdependentprocesses
174/ASQ, March 1995

Interdependence
were barelyhalf-fueledin the hybrid-task/hybrid-outcome

condition.Hybridtasks requiredgroups to act sometimes as
groups, sometimes as individuals.While the individualpart
came naturally,acting as a groupwas relativelynew.
Individualsperceived the introductionof group-level
elements to theirwork and rewardsas an add-on, not a
fundamentalchange in the work, and it undermined
attentionto the basic aspects of the task. When technicians
in hybridgroups were requiredto spend time meeting as a
group, they interpretedit as a signal that "street-time"was
no longer as importantas lookinglike a team. Ratherthan
using meetings as an opportunityto solve collective
problems, they treated them as an escape from their
individualresponsibilities:"Ourmanagercomes to all our
meetings now, because we used to spend two hours a
week in meetings just crackingjokes and telling stories."
Nor do hybridgroups experience the benefits of the

interdependentprocess. Theircooperationnorms are weak,
and the qualityof their interpersonalprocesses is relatively
low. They work alone, with loose and ineffective
coordination.Groupswith hybridtasks and outcomes rarely
cross territoriesto help other technicians, even though their
call-queuingstrategies often requirecollective care of
machines. Because hybridgroups have only half the
opportunityto learnto become teams that highly
interdependentgroups have, they have less time to develop
norms, work throughprocess problems, and develop
collective strategies. As a result, group norms do not
supportgood performance.Reportedone technicianfrom a
hybridgroup: "There'salways a few customer accounts that
everybody hates going to. When we had individual
territories,those people got taken care of by their assigned
tech rep. Now, if I comply with the no territorystrategy, I
get stuck with accounts the other reps are avoiding."
The problems that arise with hybridgroups may be due to

the mixed signals they receive from the task instructions
they are given. One technicianin a hybridgroup described
his manageras "veryexpense-conscious. He sends us voice
mail about month-to-dateexpense performanceevery week
for the whole group and brokendown by individual.He gets
after [individualtechnicians]pretty quick if they're running
over budget." Yet in this group, managingparts expenses
was a group task. Hybridoutcomes only add to the
confusion. Members of hybridgroups see their rewards as
dependent neitheron individualperformancenor on group
performance.Hybridoutcomes do not lead to the strong
perceived behavior-outcomelinkthat comes from individual
rewardsfor excellence; neither do they promote the
collective motivationengendered by group outcomes. As
one techniciansaid, "We're on this 50/50 [reward]thing,
which I thinkjust gets in my way. I can concentrate on
doing my job well, but in the end it doesn't do me much
good because others don't do theirs." Introducingsome
group-levelrewards underminedtechnicians' sense of
individualresponsibility,without providingstrong enough
collective motivationto fullydevelop the interdependent
process.
175/ASQ, March 1995

Congruence and Change over Time
The results of this study also providesome insight into what
happens when the task and rewards are incongruent.People
in the individual-task/group-outcome and group-task/
individual-outcomeconditions received different mixed
messages. Groupsin both of these conditionsexperienced
tasks that were either consistently independentor
consistently interdependent,but the outcomes they
experienced were incongruentwith the cues in their tasks.
Despite such incongruence,these groups performed
considerablybetter than did groups in the matched hybrid
condition.Groupswith individualtask and group outcomes
had generallypoor group and individualprocesses, yet they
performedquite well, both in absolute terms and relativeto
the other task and outcome conditions.What may have
helped these groups is a long historyof operating
independently.What was new for them were the group-level
outcomes introducedby this research. The interaction
process problems and low work satisfactionthey
experienced may have been due to the disruptionof their
familiarindependentwork routines by the introductionof
interdependentoutcomes. Althoughthese difficultieswere
not so severe that they destroyed members' capabilityto
get the work done well, these are groups in transition.
Observationaldata on several of these groups suggest that
collective outcomes from their managers had prompted
some of them to increase the interdependencein their work.
At one meeting I attended, the group of technicians
developed a planfor cross-traininggroup members on all
productsso that they could share collective responsibilityfor
all machines. If they succeed, the task could become highly
interdependent.By contrast, a second group in the same
conditioncould not alter its task because the manager
viewed cross-trainingas too expensive. Variancein the
degree to which groups can change the structureof their
tasks may be an importantpredictorof how the instabilityof
such groups resolves itself.
The performanceof groups with group task and individual
outcomes presents a differentpuzzle. These groups exhibit
all the benefits of group tasks and individualoutcomes in
their social processes and in members' satisfactionwith the
work, yet their performancefalls exactly in the middle of all
the task and outcome conditions.A similarpatternemerged
in a laboratorystudy of the jointeffects of task and reward
interdependence, in which pairsof subjects workingon an
interdependentediting task showed high levels of
cooperationregardless of outcome interdependence but
performedless well when outcome interdependencewas
low (Wagemanand Baker,1994). Grouptasks tend to
increase members' perceptions that their outcomes are
interdependent,even when they actuallyare not. Because
outcome interdependencein these groups was perceived
ratherthan real, it may have been insufficientto generate
the level of collective task motivationthat was exhibited by
groups that had both interdependenttasks and
interdependentoutcomes. By contrast to the above
mismatched groups, these groups had less opportunityto be
groups in transition;they had little latitudeto change the
176/ASQ, March 1995

Interdependence
rewardsthey get from management to more interdependent

outcomes. Moreover,given their comfortablesocial
relationships,they had little incentive to alter their tasks and
thus may continue to operate in this fashion without direct
interventionfrom outside the group.
Change over time. Findingsfrom this research suggest that
the design of the work that people do, the types of rewards
they receive for good performance,and their personal
characteristicsmay come into congruence over time. Groups
experiencingincongruencechange in ways that reduce it.
These changes happen dynamically:Tasks alter individuals'
experiences of their outcomes, outcome contingencies
promptindividualsto alter their tasks, and both tasks and
outcomes shape the work preferences of the individuals
themselves. But whether congruence leads to favorable
outcomes or unfavorableoutcomes may depend a great deal
on the motivationalpropertiesof the tasks and the
characteristicsof the people who performthem. In this
study, group members worked on a task that was
motivationallywell designed, whether it was performed
independentlyor interdependently.Technicianshad ample
opportunityto use their own initiativeand judgment, they
faced challenges from a wide range of engaging technical
problems, and they received richand detailed feedback on
their performanceboth from customers and from the
machines themselves. Moreover,the technicians were
generallycompetent, well trained,and self-starters. In this
study, both a strong independentprocess and a strong
interdependentprocess led to stable, long-termgood
performance.Congruenceof task and outcome
interdependence-with both being either independent or
interdependent-can lead to excellent performanceand
well-satisfied individuals,as it did in this research. But that
may only be the case as long as the work itself is
motivationallywell designed, the groups well composed, and
the individualmembers competent.
Groupswith hybridtasks and outcomes may be unstable
over time, in part because they are likelyto suffer from
performanceproblems. Such hybridgroups have not been
much studied, but examples of problemswith hybridgroups
are easy to find in organizations.Task forces often turntheir
group tasks into hybridtasks by meeting once and dividing
subtasks among individuals,and people proceed
independently,with little contact with other members of the
group. In the interests of saving time, they may not meet
again as a group untiltheir individualcontributionsmust be
combined. Such teams often run into disasters because
some members have not done their partto acceptable
standardsor have taken disparateviews of what the group
was tryingto accomplish-problems that the group could
not deal with duringits life, because after initialmeetings it
never met or acted as a group.
Popularbooks on management extol the benefits to
organizationsof using groups to performwork. But if
managers inadvertantlycreate hybridgroups by importing
group processes into a high-performingsystem with
individualtasks and rewards,they may find that what they
actuallyhave broughtin is a TrojanHorse.
177/ASQ, March 1995

REFERENCES
Allmendinger, Jutta, J. Richard Guzzo, RichardA. Mesch, Debra, MarvinLevy,David
Hackman,Linda Kowal-Wolk,and 1992 "Taskinterdependenceand W. Johnson, and Roger T.
ErinV. Lehman rewardinterdependence: Johnson
1992 "Methodsand measures for Multidisciplinary perspectives 1988 "Impactof positive
the cross-nationalstudy of on the interaction." interdependenceand
symphonyorchestras."Report Symposiumpresented at the academic groupcontingencies
No. 4, Cross-NationalStudy of meeting of the Academyof on achievement."Journalof
SymphonyOrchestras, Management,LasVegas. Social Psychology,128:
HarvardUniversity. Guzzo, RichardA., and Gregory P. 345-352.
Berkowitz,Leonard Shea Miller,L., and Robert L. Hamblin
1957 "Effects of perceived 1987 "Groupeffectiveness: What 1963 "Interdependence,differential
dependency relationsupon reallymatters."Sloan rewarding,and productivity."
conformityto group ManagementReview, Spring: AmericanSociologicalReview,
expectations."Journalof 25-31. 28: 768-778.
Abnormaland Social Hackman,J. Richard
Psychology,55: 350-354. Mitchell,Terence R., and William
1969 "Towardunderstandingthe S. Silver
Bird,A. J., and Joan Brame role of tasks in behavioral 1990 "Individual and groupgoals
1978 "Self vs. team attributions:A research."Acta Psychologica, when workersare
test of the 'I'm OK,but the 31: 97-128. interdependent:Effects on
team's so-so' phenomenon." 1987 "Thedesign of work teams." task strategies and
ResearchQuarterly,49: InJ. W. Lorsch(ed.), performance."Journalof
260-268. Handbookof Organizational AppliedPsychology,75:
Birch, David, and Joseph Veroff Behavior:314-342. 185-193.
1966 Motivation:A Study of Action. EnglewoodCliffs,NJ: Price, RichardH.
Monterey,CA: Brooks-Cole. Prentice-Hall. 1974 "Thetaxonomicclassification
Breer, Paul E., and Edwin A. Locke Hayes, Louise A. of behaviorsand situations
1965 Task Experienceas a Source 1976 "Theuse of group and the problemof
of Attitudes. Homewood, IL: contingenciesfor behavioral behavior-environment
Dorsey. control:A review." congruence."Human
PsychologicalBulletin,83: Relations,27: 567-585.
Chaplin, William F., Oliver P. 628-648.
John, and Lewis R. Goldberg Rosenbaum, Milton E., Danny I.
1988 "Conceptionsof states and Jackson, Joseph Moore, John L. Cotton, Michael S.
traits: Dimensionalattributes 1965 "Structuralcharacteristicsof Cook, Rex A. Hieser, M. Nicki
with ideals as prototypes." norms." In I. D. Steinerand Shovar, and MorrisJ. Gray
Journalof Personalityand M. Fishbein(eds.), Current 1980 "Groupproductivityand
Social Psychology,54: Studies in Social Psychology: process: Pureand mixed
541-557. 301-309. New York:Holt. rewardstructuresand task
Johnson, David W. interdependence."Journalof
Crawford,Jeffrey L., and Gordon Personalityand Social
A. Haaland 1973 "Communicationin conflict
situations:A criticalreview of Psychology,39: 626-642.
1972 "Predecisional
information-seeking and the research."International Sayles, Leonard R.
subsequent conformityin the Journalof GroupTensions, 3: 1958 Behaviorof IndustrialWork
social influenceprocess." 46-67. Groups.New York:Wiley.
Journalof Personalityand Johnson, David W., and Roger T. Shaw, MarvinE.
Social Psychology,23: Johnson 1981 GroupDynamics:The
112-119. 1989 Cooperationand Competition: Psychologyof SmallGroup
Deutsch, Morton Theoryand Research.Edina, Behavior.New York:Harper.
1949 "Anexperimentalstudy of the MN: InteractionBook Co. Shea, Gregory P., and RichardA.
effects of cooperationand Kiesler, Charles A., and Sara B. Guzzo
competitionupon group Kiesler 1989 "Groupsas human
process." HumanRelations, 1970 Conformity.Reading,MA: resources." In G. R. Ferris,
2: 129-152. Addison-Wesley. and K.M. Rowlands(eds.),
1962 "Cooperationand trust: Some Lawler,Edward E., III Researchin Personneland
theoreticalnotes." In M. R. 1990 StrategicPay:Aligning HumanResources
Jones (ed.), Nebraska Management,5: 323-356.
Organizational Strategiesand
Symposiumon Motivation: Pay Systems. San Francisco: Greenwich,CT:JAI Press.
275-319. Lincoln,NE: Jossey-Bass. Slavin, Robert E.
Universityof NebraskaPress. 1983 CooperativeLearning.New
Maier, Norman R. R.
Feldman, Daniel C. 1983 "Assets and liabilitiesin group York:Longman.
1984 "Thedevelopment and
enforcement of group problem-solving:The need for Spilerman, Seymour
an integrativefunction."In 1971 "Raisingacademic motivation
norms."Academy of J. R. Hackman,L.W. Porter, in lower class adolescents: A
ManagementReview, 9: and E. E. Lawler(eds.),
47-53. convergence of two research
Perspectiveson Behaviorin traditions."Sociologyof
Organizations,2d ed.: Education,44: 101-108.
385-392. New York:
McGraw-Hill.
178/ASQ, March 1995

Interdependence
Steiner, Ivan D. Vroom, Victor H. Zuckerman,Marvin
1972 GroupProcess and 1959 "Some personality 1983 "Thedistinctionbetween trait
Productivity.New York: determinants of the effects of and state is not arbitrary:
Academic Press. participation." Journal of Commenton Allenand
Thomas, Edwin J. Abnormal and Social Potkay's'On the arbitrary
1957 "Effects of facilitativerole Psychology, 59: 322-327. distinctionbetween traitsand
interdependenceon group 1964 Work and Motivation. New states.' " Journalof
functioning."Human York: Wiley. Personalityand Social
Relations,10: 347-366. Psychology,44: 1083-1086.
Wageman, Ruth, and George P.
Thompson, James D. Baker
1967 Organizationsin Action. New 1994 "Incentives and cooperation:
York:McGraw-Hill. The joint effects of task and
Van de Ven, Andrew H., and reward interdependence on
Diane L. Ferry group performance."
1980 Measuringand Assessing Unpublished manuscript,
Organizations.New York: Columbia Business School.
Wiley.
APPENDIXA: Survey Items on Preference for Autonomy and Norms

Preference for Autonomy at Work
I like my work best when I do it all myself.
I prefertasks that allow me to work with others.*
I would ratherwork alone than with other people.
The less I have to relyon others at work, the happierI am.
I would ratherwork througha work problemmyself than ask for advice.
Workingin small groups is better than workingalone.*
Group Norms
Cooperation
Goingout of one's way to help a group memberwith a difficultcustomer.
Helpinga group memberwithout being asked.
Keepingin touch with other group members duringthe day.
Not takingcalls outside one's own territory.*
Effort
Lookingfor ways to beat last month's response time performance.
Spendingtime in the partsdrop/coffeeshop when anothergroup member
still has calls on the board.*
Takingrepaircalls after hours.
* Reverse-scored.
179/ASQ, March 1995

APPENDIXB: Omnibus GLMsfor All Dependent Variables
Dependent variable Factor x2 p

Performanceranking Task 6.12 p < .05
Outcome 9.04 p < .05
TxOinteraction .74 n.s.
F P
Cooperationnorms Task 4.26 p < .05
Outcome 2.64 p = .08
TxOinteraction 1.53 n.s.
Effortnorms Task .02 n.s.
Qualityof groupprocess Task 7.32 p < .05
Outcome 1.50 n.s.
Out-of-territory
helping Task 7.97 p < .05
Outcome 2.53 p = .05
Learning Task 14.45 p < .05
Generalsatisfaction Task 3.96 p < .05
Outcome 1.95 n.s.
TxOinteraction .93 n.s.
Workmotivation Task .98 n.s.
Note: Regressionanalyses on these dependent variablesemployingcontinu-
ous versions of task and outcome interdependenceshowed no significant
differences from these ANOVAfindings,with one exception: The effect of
outcome interdependenceon cooperationnorms (p = .08 in the above anal-
ysis) becomes significantat p < .05.
180/ASQ, March 1995


Wagman

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Wagman

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Wagman

Uploaded by

Copyright:

Available Formats

Interdependence and Group Effectiveness

Author(s): Ruth Wageman

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

Two contrastingmodels dominate the way both managers

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

work, but tasks can be designed to be performed at varying

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

Two problems arise in extrapolatingfrom these findings to

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

Hybrid Forms of Interdependence

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

attracted to situations in which they can work alone and

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

groups managed their parts expenses collectivelyor

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

task condition(N = 56) operated almost exclusively solo:

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

processes, and group norms; and (3) survey measures of

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

Cronbach's r within r with

Preference for autonomy 3.21 .90 .78 .38 .09

Cooperation norms .72 .41 .74 .36 .25

162/ASQ, March 1995

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

repair time a machine is expected to require (based on

Group norms. Survey data assessed the strength and

Quality of group process. A seven-item survey scale

Experienced interdependence. Two survey scales assessed

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

call-queuingprocess that placed responsibilityfor all

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

Research Question 2: Effects of Total Interdependence on

Total interdependence was calculated by summing task and

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

Individual Hybrid Group

Z Very Low Low Moderate

Moderate High Very High

performance, processes, and norms, and individual

Research Question 3: Differential Effects of Task and

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

conditions on each dependent measure at Time 2. The

Finally, the bottom panel of Table 4 shows the main effects

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

performance,and a Kruskal-Wallis nonparametricanalysis of

This content downloaded from 193.255.248.150 on Sat, 7 Feb 2015 06:49:31 AM

This test was nonsignificant [U(3,6) = 7, p = .36],