when private beliefs shape collective reality
TRANSCRIPT
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 1/15
When Private Beliefs Shape CollectiveReality: The Effects of Beliefs About
Coworkers on Group Discussion andPerformance
Peter H. KimDepartment of Management and Organization, 307F Bridge Hall, Marshall School of Business,
University of Southern California, Los Angeles, California 90089-1421
The study presented in this paper examines how beliefs about coworkers affect groupdiscussion and performance. Two beliefs are considered: (1) Perceptions of coworker
task competence, and (2) achievement motivation. This study investigates whether these
perceptions can actually hinder group discussion and performance, and considers the con-
texts in which these detrimental effects are more or less likely to arise. Results indicate that
although perceptions of higher achievement motivation in coworkers lowered performance
when task information was partially shared, they raised performance when task informa-
tion was fully shared. A content analysis of group discussions reveals that the discussion
behaviors examined by this study, rather than the more frequently examined bias toward
discussing common information, mediated these results.
(Group; Information Sharing; Perception; Competence; Motivation )
IntroductionAs organizations more heavily rely on groups to make
decisions, researchers have placed increasing attention
on the ways in which groups manage information.
Organizations rely on groups because they can access
a larger amount and greater diversity of information
than individuals. A group’s potential to draw on this
information does not automatically imply, however,
that it will be accessed or effectively incorporated into
the group decision (Wittenbaum et al. 1996). Research
suggests that members often fail to exchange their
uniquely held information and that this problem can
even worsen over time (Kim 1997). The study pre-
sented in this paper seeks to extend our understand-
ing of these issues by exploring the reasons why such
deterioration in information sharing, and ultimately
performance, might occur. To address this question,
this study (1) investigates how beliefs about cowork-
ers may affect the group information-sharing process,
(2) considers whether their implications for perfor-
mance may depend on the information-sharing con-
text in which these beliefs arise, and (3) explores the
mechanisms underlying these relationships through a
content analysis of group discussions.
Changes Over Time
A group’s ability to manage information has been
shown to undergo profound changes over time. The
literature on transactive memory, for example, has
shown that experience can enable groups to encode
and retrieve information in a manner that allows each
member to focus on a specific aspect of the task and
0025-1909/03/4906/0801$05.001526-5501 electronic ISSN
Management Science © 2003 INFORMS
Vol. 49, No. 6, June 2003, pp. 801–815
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 2/15
KIM
When Private Beliefs Shape Collective Reality
on how his or her role relates to those of others in
their group (Wegner 1987, Liang et al. 1995, Moreland
1999). This research contends that, as a result of
this change, groups are able to reduce their informa-tion loads and ultimately improve their performance.
Similarly, research on shared mental models indicates
that, over time, groups can develop collective under-
standings or shared mental models of their situation
and that these organized bodies of knowledge can
facilitate communication, improve coordination, and
improve group performance (e.g., Orasanu and Salas
1993).
Recent evidence suggests, however, that a group’s
ability to manage knowledge does not always
improve over time. This evidence is provided by a
small number of studies that have investigated theimplications of experience for group discussion and
performance. Although the exchange of information
among members can be essential for performance,
groups often focus on information they share in com-
mon rather than the potentially valuable information
any individual member may hold (e.g., Stasser and
Stewart 1992, Winquist and Larson 1998). While one
might expect that this discussion bias would diminish
over time, initial findings suggest that it may some-
times worsen as groups gain experience with the task
and/or team (Kim 1997, Wittenbaum 1998, see also
Gruenfeld et al. 1996 for results that depend on the
distribution of information).
Kim (1997), for example, found that groups in
which members had previously worked together on
a similar task displayed a larger discussion bias and
achieved lower task performance than groups with
no prior experience with coworkers or the task and
groups whose prior experience was limited to either
coworkers or the task. One explanation for this effect
is the “curse of knowledge,” which refers to peo-
ple’s tendency to overestimate the ability of others
to accurately solve a problem (e.g., Camerer et al.1989, Fussell and Krauss 1992, Nickerson et al. 1987).
Kim (1997) suggested that members of experienced
groups may have exhibited a larger discussion bias
and achieved lower performance because their greater
familiarity with both the task and team may have
made them more susceptible to the “curse of knowl-
edge,” which would lead them to believe that their
partners were already aware of their privately held
information and, thus, lead them to exert less effort
to communicate it than members of inexperienced
groups.A more thorough explanation for Kim’s (1997) find-
ings, however, is that experience may shape a range
of beliefs about coworkers (e.g., due to increased
familiarity with the task and/or team, members’
schemas about how relationships and/or capabilities
should develop over time, and so on) and that each of
these beliefs can affect the sharing of information in
groups. This notion is based on research in which per-
ceptions of high-coworker competence and motiva-
tion were found to lower member efforts and collec-
tive performance (Williams and Karau 1991) as well
as on the potential implications of such perceptionsfor group coordination and discussion (Wittenbaum
et al. 1998, 1999). In this study, I consider these two
perceptions and their likely effects on the discus-
sion bias (including the mentioning, repetition, and
proportion of critical versus noncritical information),
teamwork (including performance monitoring, feed-
back, closed-loop communications, and backing-up
behaviors), and task performance.
Task Competence
Although many kinds of ability can be identified, this
study focuses on a type of ability that would be most
relevant for the work group conditions under investi-
gation (i.e., task competence). Competence is defined
as the degree to which one possesses the technical
and interpersonal skills required for one’s job (e.g.,
Butler 1991). Thus, we might expect that perceptions
of competence in coworkers should affect members’
beliefs that their partners will be able to successfully
complete the task.
Perceptions of coworker competence may affect
group information sharing and performance in two
ways. First, perceptions of coworker competence mayaffect member orientations toward their task. Whereas
perceptions of high-coworker competence should fos-
ter the belief that one’s own contributions are less
necessary and entice members to exert less effort on
their task (i.e., engage in social loafing), perceptions of
low-coworker competence should compel members to
increase their efforts to compensate for the inadequate
802 Management Science/Vol. 49, No. 6, June 2003
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 3/15
KIM
When Private Beliefs Shape Collective Reality
contributions of their partners (i.e., engage in social
compensation) (Williams and Karau 1991). Second,
perceptions of coworker competence may affect group
member orientations toward their team. Membersmay feel a greater need to obtain positive evaluations
from their coworkers when the perceived competence
of their coworkers is high versus low, and research
suggests that the desire for such enhancement may
increase their focus on commonly held information
(Wittenbaum et al. 1999).
These differences should, furthermore, have impor-
tant implications for group discussion and per-
formance. When information is partially shared,
members’ reduced efforts and increased need for
positive evaluations, due to perceptions of high-
coworker competence, should impair the information-
sharing process by increasing the discussion bias,
reducing teamwork, and lowering task performance.
When information is fully shared, however, percep-
tions of coworker competence may still affect mem-
bers’ efforts and need for positive evaluations, but
this should be less likely to influence information
sharing in a manner that would increase the discus-
sion bias, reduce teamwork, or lower performance,
because group members already possess full access to
this information. Thus, we might expect that whereas
perceptions of high-coworker competence should fos-ter a larger discussion bias, less teamwork, and
lower task performance relative to perceptions of low-
coworker competence when task information is par-
tially shared, these effects should be less likely when
task information is fully shared.
Hypothesis 1a. Perceptions of high-coworker compe-
tence will foster a larger discussion bias than percep-
tions of low-coworker competence when task information is
partially shared.
Hypothesis 1b. Perceptions of high-coworker compe-
tence will foster less teamwork than perceptions of low-coworker competence when task information is partially
shared.
Hypothesis 1c. Perceptions of high-coworker compe-
tence will foster lower task performance than percep-
tions of low-coworker competence when task information is
partially shared.
Achievement Motivation
As with competence, motivation can be exhibited
in many ways. This study focuses, however, on the
type of motivation that would be most relevantfor the work group conditions under investigation
(i.e., achievement motivation). Motivation is, there-
fore, defined as the degree to which one seeks to pur-
sue high standards of achievement rather than being
influenced by factors that might hinder achievement.
Williams and Karau (1991) examined the effects of
perceived coworker motivation on the incidence of
social loafing versus social compensation in groups
by considering a context in which participants inde-
pendently worked to contribute to the collective goal.
The implications of this perception might differ, how-
ever, in contexts where group members must inter-act with one another to develop a collective solution
(i.e., when achievement depends on the performance
of the entire group). In such cases, the perception
of high motivation in coworkers suggests that these
coworkers will not only increase their own efforts
but also demand increased efforts from others. Per-
ceptions of high-coworker motivation should, there-
fore, foster the belief that coworkers will monitor
their efforts, deter attempts to social loaf, and expect
them to make substantive contributions. Thus, group
members should exert greater efforts to develop well-
reasoned opinions when perceptions of coworker
motivation are high rather than low.
The implications of this perception for information
sharing and performance may further depend, how-
ever, on the distribution of task information, because
this distribution may affect member assessments of
whether subsequent group discussions would help
them reach their goal. When information is fully
shared, group members are not only able to develop
fully informed prediscussion opinions but are also
able to more easily evaluate each of their partners’ jus-
tifications for these opinions. These conditions shouldencourage members who perceive high motivation in
their coworkers to engage in a productive form of
task conflict (Jehn 1995), which would help groups
thoroughly consider and use critical task informa-
tion. Perceptions of high-coworker motivation should,
therefore, promote a smaller discussion bias, greater
teamwork, and higher performance than perceptions
Management Science/Vol. 49, No. 6, June 2003 803
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 4/15
KIM
When Private Beliefs Shape Collective Reality
of low-coworker motivation when task information is
fully shared.
When information is only partially shared, how-
ever, members are not only unable to develop fullyinformed prediscussion opinions (given that none of
them would possess all the necessary facts), but will
also find it much more difficult to evaluate their part-
ners’ justifications for these opinions. Such conditions
should lead members who perceive high motivation
in their coworkers to believe that attempts to resolve
such differences would foster a fruitless debate over
whose facts are more correct, given that they were
each highly motivated to succeed, yet could not even
agree on the basic facts of the case. This belief should,
in turn, lead members to lower their assessments
of the potential value of group discussion and, con-sequently, engage in social loafing (e.g., by simply
adopting a majority rule voting scheme). Thus, per-
ceptions of high-coworker motivation should foster a
larger discussion bias, less teamwork, and lower task
performance than perceptions of low-coworker moti-
vation when task information is partially shared. We
might, therefore, predict the following interactions.
Hypothesis 2a. Perceptions of high-coworker motiva-
tion will foster a larger discussion bias than perceptions of
low-coworker motivation when task information is partially
shared, but perceptions of high-coworker motivation willfoster a smaller discussion bias than perceptions of low-
coworker motivation when task information is fully shared.
Hypothesis 2b. Perceptions of high-coworker motiva-
tion will foster less teamwork than perceptions of low-
coworker motivation when task information is partially
shared, but perceptions of high-coworker motivation will
foster greater teamwork than perceptions of low-coworker
motivation when task information is fully shared.
Hypothesis 2c. Perceptions of high-coworker motiva-
tion will foster lower task performance than perceptions of
low-coworker motivation when task information is partiallyshared, but perceptions of high-coworker motivation will
foster greater task performance than perceptions of low-
coworker motivation when task information is fully shared.
Finally, given that (1) perceived coworker com-
petence and motivation are expected to influence
the discussion bias and teamwork and (2) both
the discussion bias and teamwork are expected to
influence task performance, we might expect that
(3) the predicted effects of perceived coworker com-
petence and motivation on task performance wouldoperate through their influence on the discussion bias
and teamwork. In other words, the discussion bias
and teamwork are each predicted to mediate the
effects of perceived coworker competence and moti-
vation on task performance.
Hypothesis 3. The discussion bias will mediate the
effects of perceived coworker competence and perceived
coworker motivation on task performance.
Hypothesis 4. The amount of teamwork will mediate
the effects of perceived coworker competence and perceivedcoworker motivation on task performance.
Overview
In summary, this study examines two percep-
tions that may help explain why experience may
adversely influence group information sharing and
performance. Moreover, it explores the contexts in
which these detrimental effects are more or less
likely to arise. Whereas perceptions of high-coworker
task competence and achievement motivation should
increase the discussion bias, reduce teamwork, andlower performance when information is partially
shared, these effects should not occur and even
reverse (for perceived achievement motivation) when
information is fully shared. The study also inves-
tigates the mechanisms underlying these relation-
ships by assessing several teamwork dimensions (i.e.,
performance monitoring, feedback, closed-loop com-
munications, and backing-up behaviors) through a
content analysis of group discussions. These objec-
tives are pursued through the laboratory experiment
described below.
MethodThis study employed a 2(high vs. low competence)×
2(high vs. low motivation)×2(full vs. partial informa-
tion) between-subjects design. Three-person groups
were used as the unit of analysis.
804 Management Science/Vol. 49, No. 6, June 2003
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 5/15
KIM
When Private Beliefs Shape Collective Reality
Participants
The study involved 276 undergraduates from a large
public university, randomly assigned to 92 groups.
They took part in the exercise for course credit andthe chance to win cash prizes of $90, $60, and $30 that
were awarded through a lottery to three of the best
performing groups.
Procedure
Participants were informed that the purpose of the
study was to investigate the effects of group com-
position on task performance and that, for this rea-
son, some participants would be assigned to groups
with members who were similar, while other partic-
ipants would be assigned to groups with members
who were dissimilar. Participants completed an initialtask, comprised of true or false and multiple-choice
questions that ostensibly assessed them on their com-
petence and motivation. Before receiving feedback,
participants were randomly assigned to three-person
groups and given materials for a murder mystery,
which they were asked to read and review as individ-
uals. During this time, they were asked to complete
a prediscussion questionnaire, which asked partici-
pants to select the suspect they believed to be the
culprit and write a brief rationale for their decision.
After returning their packets and questionnaires, par-
ticipants were informed that both of their cowork-
ers were found to possess high/low competence and
high/low motivation based on their performance on
the initial task (i.e., participants received the same
feedback about each of their partners, but were not
given feedback about themselves). Participants were
then given 20 minutes to solve the murder mystery
as a group.
Decision Task
Participants read a series of interviews from a homi-
cide investigation. These interviews were providedin a booklet that included supporting materials such
as a list of characters, a map, a handwritten note,
and a newspaper article (Stasser and Stewart 1992).
The interviews contained 24 clues that were either
incriminating or exonerating for each of three sus-
pects (Suspect 1, Suspect 2, and Suspect 3). Specifi-
cally, each suspect was incriminated by 6 clues, but
there were also 3 clues that exonerated Suspect 2
and 3 clues that exonerated Suspect 3. Therefore, the
set of 24 clues was designed so that Suspects 2 and 3
could be ruled out. Furthermore, all the clues consid-ered together supported the conclusion that Suspect 1
had both motive and opportunity to commit the crime
(i.e., was the correct choice).
Experimental Manipulations
Information Distribution. Each participant’s pac-
ket of information contained several pages of evi-
dence. Some of this evidence was critical to solving
the task, and some of it was not critical. Nine critical
clues were necessary to determine that Suspect 1 was
guilty. Three of those clues incriminated Suspect 1,and the other 6 exonerated Suspects 2 and 3.
In the full information condition, each of the three
group members was given a complete set of infor-
mation, including all 9 incriminating and exonerat-
ing clues (i.e., every member received each of the
24 total clues). However, in the partial information
condition, the 9 critical clues were distributed among
group members so each member received all 15 non-
critical clues but only one-third of the critical evidence
(i.e., each member received 21 of the 24 total clues).1
One group member received 3 exonerating clues for
Suspect 2; a second received 3 exonerating clues forSuspect 3; and the third received 3 incriminating clues
for Suspect 1. In this manner, the critical clues pos-
sessed by each member were completely nonredun-
dant, but all of them were available for use by the
group. Group members were not informed about the
distribution of information.
Perceived Task Competence. The level of per-
ceived task competence was manipulated by pro-
viding false feedback to participants regarding their
partners’ performance on the deductive reasoning
component of the initial task, which they were toldcould predict their ability to successfully solve the
homicide investigation. Perceptions of high (low)
1 Although study participants received different numbers of clues in
these two information distribution conditions, these specific main
effects would not account for the predictions and results that are
of primary interest in this study.
Management Science/Vol. 49, No. 6, June 2003 805
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 6/15
KIM
When Private Beliefs Shape Collective Reality
competence were fostered by informing participants
that they had been assigned to a group with members
who each performed well (poorly) on the deductive
reasoning component of the initial task. Participantswere never informed of their own competence. More-
over, the purported purpose of the experiment (i.e., to
study the effects of group composition), and explicit
instructions not to discuss this feedback, reduced the
likelihood that they would acquire this information
from their partners.2
Perceived Achievement Motivation. The level of
perceived motivation was manipulated by providing
false feedback to participants regarding their part-
ners’ results on the personality profile component
of the initial task. Perceptions of high (low) motiva-tion were fostered by informing participants that they
had been assigned to a group with members who
were each highly motivated toward achievement and,
thus, likely to support and defend answers that they
believed to be correct (easily influenced by factors
other than achievement and, thus, likely to avoid dis-
putes and support popular opinions). As with compe-
tence, participants were never informed of their own
motivation. Moreover, the purported purpose of the
experiment (i.e., to study the effects of group compo-
sition), and instructions not to discuss this feedback,
reduced the likelihood that they would acquire thisinformation from their partners.
Manipulation Checks
A pilot study was conducted with 36 additional
undergraduate participants to determine whether the
manipulations of coworker competence and motiva-
tion would affect participants’ perceptions of their
partners’ competence and motivation, as well as their
own competence and motivation, before the group
discussion. Separate measures of perceived coworker
competence, perceived coworker motivation, percep-
tions of own competence, and perceptions of own
motivation were obtained with 7-point scales, where
1= very low and 7 = very high. Results revealed that
the high-competence manipulation fostered higher
2 None of the groups were found to violate instructions to refrain
from discussing their feedback.
perceptions of coworker competence M = 581 sd =
027 than the low-competence manipulation M =
306 sd = 028 (F 1 34 = 5077 p < 00005, and
that the high-motivation manipulation fostered higherperceptions of coworker motivation M = 581
sd = 023 than the low-motivation manipulation
M = 272 sd = 023 F 1 34 = 9121 p < 00005.
The competence manipulation did not affect percep-
tions of coworker motivation F 1 34 = 113ns,
nor did the motivation manipulation affect percep-
tions of coworker competence F 1 34 = 024ns.
Finally, neither the competence nor the motivation
manipulations were found to affect participants’ per-
ceptions of their own competence or motivation
F s < 06. Thus, each of the study manipulations was
found to operate as intended.
Dependent Measures
Task Performance. The group decision concerning
the culprit was obtained by asking groups to deter-
mine which suspect was most likely to have com-
mitted the crime. Decisions that named the correct
suspect as most likely received a score of 1, and deci-
sions that named an incorrect suspect as most likely
received a score of 0.
Content Analysis. Group discussions were tape-
recorded and transcribed for the content analysis.
The content analysis coded for the discussion bias
in two ways. This approach is consistent with prior
discussion bias studies, which not only assess the
mentioning and repetition of critical and noncritical
information (e.g., Larson et al. 1996, Stasser et al.
1995) but also consider the possibility that the focus
of discussion on critical rather than noncritical infor-
mation may be more important for group perfor-
mance than the sheer number of times these clues
are mentioned or repeated (e.g., Stasser and Stewart
1992). These measures should provide complemen-tary lenses into the discussion bias by revealing the
extent to which this information-sharing problem
arises because groups fail to notice information (i.e.,
the mentioning of clues), do not sufficiently evalu-
ate information (i.e., the repetition of clues), or do
not place sufficient emphasis on the right kind of
information (i.e., the discussion focus).
806 Management Science/Vol. 49, No. 6, June 2003
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 7/15
KIM
When Private Beliefs Shape Collective Reality
The mentioning of critical information counted thenumber of critical clues that were stated at least once.
For example, if a group stated 3 different critical clues
during its discussion, “X,” “Y,” and “Z,” the group isconsidered to have mentioned 3 critical clues, regard-less of how often each of these 3 critical clues wasstated. The repetition of critical information counted
the number of times critical clues were stated afterthey were first mentioned, and then divided thesetotals by the number of critical clues that were men-
tioned at least once. This measure yielded a per-mentioned-item repetition rate for critical information
(see Winquist and Larson 1998). In the above example,if the group stated clue “X” one time after it was firstmentioned, clue “Y” two times after it was first men-
tioned, and clue “Z” three times after it was first men-tioned, the per-mentioned-item repetition rate for crit-ical information would be two (i.e., 1+2+3/3= 2).
Analogous procedures were used to calculate the men-tioning and repetition of noncritical clues.
The content analysis assessed the discussion focus
by examining the degree to which groups discussedcritical versus noncritical information. Thus, two vari-
ables were coded: (1) The total number of times groupmembers stated critical clues (i.e., by counting eachtime any critical clue was mentioned or repeated), and
(2) the total number of times group members stated
noncritical clues (i.e., by counting each time any non-critical clue was mentioned or repeated). The discus-
sion focus was obtained by dividing the total numberof times group members stated critical informationby the total number of times they stated noncritical
information.Finally, the content analysis coded 9 additional
discussion behaviors. These behaviors were found,through a perusal of transcripts, not only to cap-ture the essence of these group discussions but also
to illustrate the teamwork dimensions identified byMorgan et al. (1986)—performance monitoring, feed-
back, closed-loop communications, and backing-upbehaviors. These behaviors included the frequencywith which group members contributed an opinion,asked for an opinion, contributed a justification, asked
for a justification, confirmed a partner’s statement,dissented with a partner’s statement, asked a fact-
based question, stated that they had no opinion, andcontributed a task-relevant phrase.
Members exhibited performance monitoring by
asking for an opinion (e.g., “Who do you think commit-
ted the crime?”) or asking for a justification (“Why do
you think it was Billy Prentice?”). Members providedfeedback by dissenting with a partner’s statement (e.g.,
“I don’t think he did it.”) or confirming a partner’s
statement (e.g., “That’s right.”). Members displayed
closed-loop communications by confirming a partner’s
statement (see previous example) or asking a fact-
based question (e.g., “Eddie had a hearing aid?”). Mem-
bers exhibited backing-up behaviors by contributing
an opinion (e.g., “I thought it was Mickey Malone.”) or
contributing a justification (e.g., “Mickey did it because
Mr. Guion was going to ruin his business.”). Members
also indicated the need for backing-up behaviors by
stating that they had no opinion (e.g., “I don’t know.”).Finally, each of these behaviors was aggregated to
determine the frequency with which members con-
tributed a task-relevant phrase, which provided a gen-
eral indication of the time and effort groups devoted
to their task.
Two independent raters were trained to perform
these content analyses. The raters were given cod-
ing sheets that identified all the discussion bias and
discussion behavior variables and were instructed to
record, for each line of the transcripts, every occa-
sion in which any of these variables were found. The
desired measures were then obtained through fre-
quency counts. Some utterances qualified as both a
discussion bias and discussion behavior variable and
were coded for both. The utterance, “Eddie had a
hearing aid?,” for example, represents the mention-
ing of critical information as well as the asking of a
fact-based question and was, thus, counted for each
category. These kinds of dual-coded utterances helped
foster significant correlations between the discussion
bias and discussion behavior measures. Despite these
correlations, the measures still differed in the roles
they played in this study (see Results).The raters were blind to the hypotheses and treat-
ment conditions. Nevertheless, the content analy-
ses from the two independent raters demonstrated
high interrater reliability, as indicated by a ran-
dom sample of 26 transcripts (noncritical informa-
tion (Kappa = 094), critical information (Kappa =
085), asked for opinion (Kappa = 091), asked for
Management Science/Vol. 49, No. 6, June 2003 807
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 8/15
KIM
When Private Beliefs Shape Collective Reality
justification (Kappa = 089), dissented with partner
comment (Kappa= 087), confirmed partner statement
(Kappa = 086), asked fact-based question (Kappa =
090), contributed opinion (Kappa= 086), contributedjustification (Kappa = 085), and stated they had no
opinion (Kappa= 074)). A third, blind rater resolved
any differences between these two sets of ratings.
ResultsAll 276 subjects (92 groups) completed the study.
However, audiotapes for three groups could not be
transcribed or coded due to the poor quality of the
recordings. As a result, 89 groups were left for anal-
ysis. Analyses were first conducted to examine the
independent effects of the study variables. Tests of
mediated relationships among these variables werethen explored through a series of hierarchical regres-
sions (Baron and Kenny 1986, James and Brett 1984).
A summary of means, standard deviations, and corre-
lations is presented in Table 1. Means, standard devi-
ations, and sample sizes per condition for the study
variables are presented in Table 2.
Effects on the Discussion Bias
Hypotheses 1a and 2a predicted that whereas percep-
tions of high-coworker competence and motivation
Table 1 Descriptive Statistics and Correlations
Standard Task Critical Noncritical Critical Noncritical Critical/
Means deviations performance mentioned mentioned repeated repeated noncritical clues
1. Task performance 038 049
2. Critical clues mentioned 192 130 036∗∗∗∗∗
3. Noncritical clues mentioned 567 200 037∗∗∗∗∗ 063∗∗∗∗∗∗
4. Critical clues repeated 100 128 014 029∗∗ 045∗∗∗∗∗∗
5. Noncritical clues repeated 112 093 033∗∗∗ 053∗∗∗∗∗∗ 059∗∗∗∗∗∗ 028∗∗
6. Critical/noncritical clues 036 040 009 035∗∗∗∗ −010 037∗∗∗∗∗ −016
7. Discussion behaviors 061a ∗∗∗∗∗∗ 079a ∗∗∗∗∗∗ 086a ∗∗∗∗∗∗ 068a ∗∗∗∗∗∗ 091a ∗∗∗∗∗∗ 021a
Task-relevant phrase 6049 3661 048∗∗∗∗∗∗ 067∗∗∗∗∗∗ 074∗∗∗∗∗∗ 037∗∗∗∗∗ 077∗∗∗∗∗∗ 000
Contribute opinion 1437 857 045∗∗∗∗∗∗ 049∗∗∗∗∗∗ 054∗∗∗∗∗∗ 015 067∗∗∗∗∗∗ −006
Ask for opinion 382 270 019 029∗∗ 019 005 029∗∗ −004
Contribute justification 967 648 043∗∗∗∗∗∗ 049∗∗∗∗∗∗ 057∗∗∗∗∗∗ 027∗ 059∗∗∗∗∗∗ −000
Ask for justification 203 219 005 032∗∗∗ 059∗∗∗∗∗∗ 053∗∗∗∗∗∗ 047∗∗∗∗∗∗ 000
Confirm 570 530 043∗∗∗∗∗∗ 051∗∗∗∗∗∗ 051∗∗∗∗∗∗ 013 036∗∗∗∗∗ 006
Dissent 260 351 033∗∗∗ 041∗∗∗∗∗∗ 038∗∗∗∗∗ 009 057∗∗∗∗∗∗ −004
Fact-based quest ion 404 449 041∗∗∗∗∗∗ 064∗∗∗∗∗∗ 059∗∗∗∗∗∗ 031∗∗∗∗∗ 059∗∗∗∗∗∗ 009
No opinion 080 125 −019 008 019 025∗ −006 003
aCanonical correlation.∗p < 005. ∗∗p < 001. ∗∗∗p < 0005. ∗∗∗∗p < 0001. ∗∗∗∗∗p < 00005. ∗∗∗∗∗∗p < 00001.
would foster a larger discussion bias relative to
perceptions of low-coworker competence and motiva-
tion, respectively, when task information is partially
shared, these effects should not arise and even reverse(for perceived motivation) when information is fully
shared. These predictions were tested by examining
the mentioning and repetition of critical and noncrit-
ical information and the proportion of critical versus
noncritical information discussed.
A multiple analysis of variance (MANOVA) was
conducted to determine whether the experimental
manipulations affected the mentioning and repetition
of critical and noncritical information and the pro-
portion of critical versus noncritical information dis-
cussed. The distribution of information was found
to exert an overall effect on these discussion biasmeasures (F 5 77= 490 p < 0001. Follow-up anal-
yses of variance (ANOVAs) revealed that groups
with full information mentioned more critical infor-
mation (M = 222 sd = 020) than groups with par-
tial information (M = 161 sd = 020) ((F 1 88 =
481 p < 005); (Effect Size ES= 305). No other main
effects or higher order interactions were found ((F s <
322 ns ES < 048)). These results do not support
predictions that perceived coworker competence ormotivation would affect the discussion bias.
808 Management Science/Vol. 49, No. 6, June 2003
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 9/15
KIM
When Private Beliefs Shape Collective Reality
Table 2 Means, Standard Deviation, and Number of Observations by Condition
Partial information Full information
High motivation Low motivation High motivation Low motivationHigh comp. Low comp. High comp. Low comp. High comp. Low comp. High comp. Low comp.
Variable N = 12 N = 11 N = 10 N = 11 N = 11 N = 12 N = 12 N = 10
Task performance 017 018 054 045 067 042 042 018
014 014 014 014 014 014 014 014
Critical clues mentioned 167 164 160 154 218 225 217 230
150 103 135 144 108 136 134 134
Noncritical clues mentioned 625 573 610 600 527 550 558 490
191 210 179 184 200 211 188 264
Critical clues repeated 139 148 045 114 082 067 106 092
166 217 055 134 116 083 107 052
Noncritical clues repeated 125 106 080 128 094 113 141 101
102 065 081 096 081 110 103 106Critical/noncritical 029 037 027 024 037 034 028 077
022 032 028 023 019 030 016 094
Task-relevant phrase 6117 4936 6440 7409 5727 6183 6692 4730
3831 3459 4092 5601 3523 2790 3129 2529
Contribute opinion 1392 991 1730 1836 1309 1408 1742 1060
634 579 1268 1328 699 568 804 432
Ask for opinion 392 309 350 282 554 433 442 270
300 181 264 209 491 192 162 170
Contribute justification 858 691 1050 1354 1054 1092 900 730
650 446 952 705 873 396 544 337
Ask for justification 292 345 150 254 118 108 183 170
211 344 108 284 178 144 175 134
Confirm 600 382 930 882 445 525 492 320
445 322 706 935 294 441 368 132
Dissent 150 091 500 300 191 217 408 240
157 114 518 322 378 310 487 237
Fact-based question 442 300 340 482 445 483 483 220
614 452 232 623 301 551 426 148
No opinion 058 127 040 036 036 142 142 040
116 205 052 050 067 116 178 052
Effects on Discussion Behaviors
Hypotheses 1b and 2b predicted that whereas per-
ceptions of high-coworker competence and motiva-tion would foster less teamwork (assessed by the
discussion behaviors) relative to perceptions of low-
coworker competence and motivation, respectively,
when task information is partially shared, these
effects would not arise and even reverse (for per-
ceived motivation) when task information is fully
shared. The following analyses tested these pre-
dictions as well as the relationship between the
discussion behaviors and the discussion bias.A MANOVA was conducted to examine the overall
effects of the experimental manipulations on the
9 discussion behaviors. Both perceived motivation
(F 9 73= 285 p < 001) and the distribution of infor-
mation (F 9 73 = 332 p < 0005) were found to
affect these discussion behaviors. These main effects
Management Science/Vol. 49, No. 6, June 2003 809
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 10/15
KIM
When Private Beliefs Shape Collective Reality
were qualified, however, by an interaction between
perceived motivation and the distribution of informa-
tion F 9 73= 297 p < 0005.
ANOVAs were then conducted to isolate thespecific variables driving these multivariate effects.
Groups with high-perceived motivation were found
to make fewer dissenting comments M = 162 sd =
050 than groups with low-perceived motivation
M = 362 sd = 052 F 1 88= 758 p < 001ES =
392. Groups with partial information were also
found to ask for justifications and confirm partner
statements more frequently M = 260 sd = 032 and
M = 698 sd = 077, respectively) than groups with
full information M = 145 sd = 032 F 1 88 =
658 p < 005ES = 359 and M = 446 sd = 076
F 1 88 = 544 p < 005ES = 332, respectively).Moreover, the interaction between perceived motiva-
tion and the distribution of information was found
to affect three discussion behaviors: the frequency
with which group members contributed justifications
F1 88= 631 p < 005ES = 361, asked for jus-
tifications F1 88= 399 p < 005ES = 202, and
confirmed partner comments F1 88 = 520 p <
005ES = 229. Groups with high-perceived moti-
vation were found to contribute justifications less
frequently M = 775 sd = 134, ask for justifica-
tions more frequently M = 318 sd = 044, and
confirm partner statements less frequently M =
491 sd = 106 than groups with low-perceived
motivation when information was partially shared
M = 1202 sd = 140M = 202 sd = 046M =
906 sd = 111, respectively), but groups with high-
perceived motivation were found to contribute justi-
fications more frequently M = 1073 sd = 134, ask
for justifications less frequently M = 113 sd = 044,
and confirm partner statements more frequently M =
485 sd = 106 than groups with low-perceived moti-
vation when information was fully shared M =
815 sd=
137M =
177 sd=
045M =
406 sd=
109, respectively).
Canonical correlations were also conducted to
examine the relationship between the discussion bias
and the 9 discussion behaviors. As might be expected,
given that some utterances were coded as both discus-
sion bias and discussion behavior variables, the men-
tioning of critical F 9 79 = 1485 p < 00001 and
noncritical information F 9 79 = 2408 p < 00001
and the repetition of critical F 9 79 = 776 p <
00001 and noncritical information F 9 79 =
4367 p < 00001 were all significantly related tothese discussion behaviors. Univariate correlations
between the discussion bias measures and each of
the discussion behaviors were then examined to iso-
late the particular variables driving these multivari-
ate effects (see Table 1). These univariate correlations
revealed that the mentioning and repetition of critical
and noncritical information were each positively cor-
related with the majority of the discussion behaviors,
indicating that the level of teamwork was positively
associated with the degree to which both critical and
noncritical task-relevant information was discussed.
These results support predictions that perceptionsof coworker motivation affect teamwork (assessed by
the discussion behaviors) and that these relationships
depend on the distribution of task information (i.e.,
Hypothesis 2B was supported). Follow-up univari-
ate analyses, furthermore, suggest that the frequency
with which group members contributed justifica-
tions, asked for justifications, and confirmed partner
comments were driving this multivariate interaction
effect.
Effects on Performance
Hypotheses 1c and 2c predicted that whereas per-
ceptions of high-coworker competence and motiva-
tion would foster lower task performance relative to
perceptions of low-coworker competence and moti-
vation, respectively, when information is partially
shared, these effects should not arise and even reverse
(for perceived motivation) when information is fully
shared. The following analyses tested these predic-
tions as well as the effects of the discussion bias and
discussion behaviors on performance.
An ANOVA was conducted to assess whether
the manipulations affected task performance. Theresults revealed a two-way interaction between per-
ceived motivation and the distribution of infor-
mation. Groups with high-perceived motivation
achieved lower task performance M = 017 sd =
010 than groups with low-perceived motivation
M = 050 sd = 010 when task information was
partially shared, but groups with high-perceived
810 Management Science/Vol. 49, No. 6, June 2003
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 11/15
KIM
When Private Beliefs Shape Collective Reality
motivation achieved higher task performance M =
054 sd = 010 than groups with low-perceived moti-
vation M = 030 sd = 010 when task information
was fully shared F1 91 = 824 p < 001ES =285. These results support predictions that percep-
tions of coworker motivation affect performance and
that these relationships depend on the distribution
of information (i.e., Hypothesis 2C was supported).
Whereas perceptions of high-coworker motivation
lowered performance when information was par-
tially shared, this relationship was reversed when
information was fully shared.
Mediated Regression Results
Hypotheses 3 and 4 predicted that the effects ofperceived coworker competence and motivation on
task performance would be mediated by the dis-
cussion bias (Hypothesis 3) and teamwork (assessed
by the discussion behaviors) (Hypothesis 4). Results
from the prior analyses indicate that the experimen-
tal manipulations affected both the discussion bias
and the discussion behaviors. However, only the dis-
cussion behaviors were significantly related to both
the perceived motivation X information distribution
interaction F 9 73 = 297 p < 0005 and task per-
formance F 9 79 = 507 p < 00001 (see Table 1).
These behaviors, therefore, represent the only factorthat might plausibly mediate the reported relationship
between coworker perceptions and performance. This
possibility was tested through a series of hierarchical
regressions (Baron and Kenny 1986, James and Brett
1984).
Tests of mediation require that four conditions be
met (James and Brett 1984). To demonstrate the medi-
ating effect of variable Y in the relation X → Z:
(1) X must be significantly related to Y , (2) X and Y
must be significantly related to Z, (3) the variance in Z
predicted by X must be nonsignificant after mediator
Y is controlled (the satisfaction of all but this condi-
tion indicates partial mediation), and (4) Y should be
significantly related to Z after X is controlled. Each of
these conditions was examined to determine whether
the discussion behaviors Y mediated the relation-
ship between the perceived motivation X information
distribution interaction X and task performance Z.
A multivariate regression was conducted to deter-
mine whether the perceived motivation X infor-
mation distribution interaction affected the group
discussion variables. This analysis revealed a sig-nificant relationship between the perceived motiva-
tion X information distribution interaction and the
discussion variables, in general F 9 76 = 304 p <
0005, and the frequency with which groups con-
tributed justifications t1 88= 252 p < 005, asked
for justifications t1 88= −203 p < 005, and con-
firmed partner statements t1 88 = 230 p < 005,
in particular. Thus, Condition 1 was satisfied for the
discussion behaviors as a whole and, specifically, for
the frequency with which groups contributed justifi-
cations, asked for justifications, and confirmed part-
ner statements (i.e., the independent variable affected
the potential mediators).
Multiple regressions were conducted to examine
the effects of the perceived motivation X information
distribution interaction and group discussion behav-
iors, respectively, on task performance. These analyses
revealed a significant relationship between the per-
ceived motivation X information distribution interac-
tion and task performance t1 91= 291 p < 0005.
These analyses also revealed significant relationships
between the group discussion behaviors as a whole
and task performance F 9 88= 507 p < 00001, aswell as between the frequency with which groups
contributed justifications t1 88 = 450 p < 00001
and confirmed partner statements t1 88= 440 p <
00005, respectively, and task performance. Thus,
Condition 2 was satisfied for the perceived motiva-
tion X information distribution interaction, the group
discussion behaviors as a whole and, specifically, for
the frequency with which groups contributed justi-
fications and confirmed partner statements (i.e., the
independent variable and potential mediators affected
the dependent variable).
A two-step procedure was performed to determine
whether the variance in task performance predicted
by the perceived motivation X information distribu-
tion interaction was nonsignificant after controlling
for the group discussion behaviors. Task performance
was, first, regressed on the group discussion behav-
iors. The residuals from this first step were then
Management Science/Vol. 49, No. 6, June 2003 811
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 12/15
KIM
When Private Beliefs Shape Collective Reality
regressed on the interaction between perceived moti-
vation and the distribution of information (i.e., after
the effects of the group discussion behaviors were
controlled). This procedure revealed a nonsignificantrelationship between the perceived motivation X
information distribution interaction and performance
t1 88 = 147ns. Analogous procedures were
also conducted on the specific group discussion
behaviors. These procedures revealed nonsignificant
relationships between the perceived motivation X
information distribution interaction and task perfor-
mance when controlling for the frequency with which
groups contributed justifications t1 88= 191ns
andconfirmedpartnerstatements t1 88= 197ns.
Thus, Condition 3 was satisfied for the group discus-
sion behaviors as a whole and, specifically, for the fre-quency with which groups contributed justifications
and confirmed partner statements (i.e., the indepen-
dent variable did not affect the dependent variable
after controlling for the potential mediators).
Finally, a procedure similar to that performed in
Condition 3 was used to determine if the group dis-
cussion behaviors were still significantly related to
task performance after the effects of the perceived
motivation X information distribution interaction was
controlled. Task performance was first regressed on
the perceived motivation X information distribution
interaction. The residuals from this analysis were
then regressed on the group discussion variables
(i.e., after controlling for the perceived motivation
X information distribution interaction). This proce-
dure revealed that the group discussion behaviors,
in general F 9 88 = 409 p < 00005, and the fre-
quency with which groups contributed justifications
t1 88 = 397 p < 00005 and confirmed partner
statements t1 88 = 434 p < 00001, in particular,
still affected task performance. Thus, Condition 4
was satisfied for the group discussion behaviors as a
whole and, specifically, for the frequency with whichgroups contributed justifications and confirmed part-
ner statements (i.e., the potential mediators affected
the dependent variable even after controlling for the
independent variable).
The satisfaction of all four conditions at both the
multivariate and univariate levels supports the notion
that the discussion behaviors, in general, and the
frequency with which groups contributed justifica-
tions and confirmed partner statements, in partic-
ular, completely mediated the relationship between
the perceived motivation X information distributioninteraction and task performance (i.e., Hypothesis 4
was supported). These results, therefore, provide a
clear explanation for the performance differences
observed in this study. The evidence indicates that the
discussion behaviors assessed in this study can play
a critical role in the information-sharing process by
mediating the effects of coworker perceptions on per-
formance. The results also help substantiate, to some
degree, the univariate analyses conducted in the pre-
ceding section. Whereas the effect of the perceived
motivation X information distribution interaction on
any of the 9 discussion behaviors may have occurredby chance, the finding that the frequency with which
groups contributed justifications and confirmed part-
ner statements also met the three other conditions
required for mediation is less likely. More substantive
verification of the role(s) played by these discussion
behaviors, however, will require further investigation.
DiscussionThe purpose of this study was to investigate how
beliefs about coworkers may affect group informa-
tion sharing and performance. In addition, it sought
to identify contexts in which these adverse effects are
more or less likely to arise. Finally, this study inves-
tigated the underlying mechanisms for these find-
ings through an exploratory content analysis of group
discussions.
The evidence from this study suggests that beliefs
about coworkers can exert important effects on group
discussion and performance. The finding that percep-
tions of higher coworker motivation lowered perfor-
mance when task information was partially shared is
consistent with Kim’s (1997) observation that mem-bers’ beliefs about their coworkers can sometimes hin-
der group interactions and outcomes. This study also
extends this insight, however, by suggesting that we
broaden our inquiry beyond mechanisms such as the
“curse of knowledge” to more fully understand these
effects. It may be useful, therefore, to consider how a
range of coworker perceptions may influence group
812 Management Science/Vol. 49, No. 6, June 2003
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 13/15
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 14/15
8/7/2019 When Private Beliefs Shape Collective Reality
http://slidepdf.com/reader/full/when-private-beliefs-shape-collective-reality 15/15
KIM
When Private Beliefs Shape Collective Reality
AcknowledgmentsThe author thanks Thomas Cummings, Daniel Gigone, Terri L.
Griffith, Deborah H. Gruenfeld, Dawn Iacobucci, Claus Langfred,
Margaret A. Neale, Garold Stasser, Melissa Thomas-Hunt, LeighL. Thompson, and Brian Uzzi for their assistance. This research
was supported by grants from the State Farm Foundation and
Northwestern University’s Dispute Resolution Research Center.
ReferencesBaron, R. M., D. A. Kenny. 1986. The moderator-mediator variable
distinction in social psychological research: Conceptual, strate-
gic, and statistical considerations. J. Personality Soc. Psych. 51(6)
1173–1182.
Brauer, M., M. D. Gliner, C. M. Judd. 1995. The effects of repeated
expressions on attitude polarization during group discussions.
J. Personality Soc. Psych. 68(6) 1014–1029.
Butler, J. K. 1991. Toward understanding and measuring condi-tions of trust: Evolution of a conditions of trust inventory.
J. Management 17 643–663.
Camerer, C. F., G. Loewenstein, M. Weber. 1989. The curse of knowl-
edge in economic settings: An experimental analysis. J. Political
Econom. 97 1232–1254.
Fussell, S. R., R. M. Krauss. 1992. Coordination of knowledge in
communication: Effects of speaker’s assumptions about what
others know. J. Personality Soc. Psych. 62(3) 378–391.
Gigone, D., R. Hastie. 1993. The common knowledge effect: Infor-
mation sharing and group judgment. J. Personality Soc. Psych.
65(5) 959–974.
, . 1997. The impact of information on small group choice.
J. Personality Soc. Psych. 72(1) 132–140.
Gruenfeld, D. H., E. Mannix, K. Williams, M. A. Neale. 1996. Group
composition and decision making: How member familiarity
and information distribution affect process and performance.
Organ. Behavior Human Decision Processes 67(1) 1–15.
James, L. R., J. M. Brett. 1984. Mediators, moderators, and tests for
mediation. J. Appl. Psych. 69(2) 307–321.
Jehn, K. A. 1995. A multimethod examination of the benefits
and detriments of intragroup conflict. Admin. Sci. Quart. 40(2)
256–282.
Kim, P. H. 1997. When what you know can hurt you: A study
of experiential effects on group discussion and performance.
Organ. Behavior Human Decision Processes 69(2) 165–177.
, K. A. Diekmann, A. Tenbrunsel. 2003. Flattery may get you
somewhere: The strategic implications of providing positive
vs. negative feedback about ability vs. ethicality in negotiation.
Organ. Behavior Human Decision Processes 90(2) 225–243.
Larson, J. R., Jr, C. Christensen, A. S. Abbott, T. M. Franz.
1996. Diagnosing groups: Charting the flow of information in
medical decision-making teams. J. Personality Soc. Psych. 71(2)
315–330.
Liang, D. W., R. Moreland, L. Argote. 1995. Group versus indi-
vidual training and group performance: The mediating role oftransactive memory. Personality Soc. Psych. Bull. 21(4) 384–393.
. 1999. Transactive memory: Learning who know what in work
groups and organizations. L. Thompson, J. Levine, D. Messick,
eds. Shared Cognition in Organizations: The Management of Knowl-
edge. Erlbaum, Mahwah, NJ, 3–31.
Morgan, B. B., Jr, A. S. Glickman, E. A. Woodward, A. Blaiwes,
E. Salas. 1986. Measurement of team behaviors in a Navy envi-
ronment. NTSC report number 86-014, Naval Training System
Center, Orlando, FL.
Nickerson, R. S., A. Baddeley, B. Freeman. 1987. Are people’s esti-
mates of what other people know influenced by what they
themselves know? Acta Psych. 64 245–259.
Orasanu, J., E. Salas. 1993. Team decision making in complex
environments. G. A. Klein, J. Orasanu, R. Calderwood, C. E.Zsambok, eds. Decision Making in Action: Models and Methods.
Ablex Publishing Corporation, Norwood, NJ, 327–345.
Stasser, G., D. D. Stewart. 1992. Discovery of hidden profiles by
decision-making groups: Solving a problem versus making a
judgment. J. Personality Soc. Psych. 63(3) 426–434.
, , G. M. Wittenbaum. 1995. Expert roles and information
exchange during discussion: The importance of knowing who
knows what. J. Experiment. Soc. Psych. 31 244–265.
Wegner, D. M. 1987. Transactive memory: A contemporary analysis
of the group mind. B. Mullen, G. R. Goethals, eds. Theories of
Group Behavior. Springer-Verlag, New York, 185–205.
Williams, K. D., S. J. Karau. 1991. Social loafing and social compen-
sation: The effects of expectations of coworker performance.
J. Personality Soc. Psych. 61(4) 570–581.Winquist, J. R., J. R. Larson, Jr. 1998. Information pooling: When it
impacts group decision making. J. Personality Soc. Psych. 74(2)
371–377.
Wittenbaum, G. M. 1998. Information sampling in decision-making
groups: The impact of members’ task-relevant status. Small
Group Res. 29(1) 57–84.
, A. P. Hubbell, C. Zuckerman. 1999. Mutual enhancement:
Toward an understanding of the collective preference for
shared information. J. Personality Soc. Psych. 77(5) 967–978.
, G. Stasser, C. J. Merry. 1996. Tacit coordination in anticipation
of small group task completion. J. Experiment. Soc. Psych. 32(2)
129–152.
, S. I. Vaughan, G. Stasser. 1998. Coordination in task-
performing groups. R. S. Tindale, L. Heath, J. Edwards, E. J.
Posavac, F. B. Bryant, Y. Suarez-Balcazar, E. Henderson-King,
J. Myers, eds. Theory and Research on Small Groups. Plenum
Press, New York, 177–204.
Accepted by Linda Argote, former department editor; received March 1, 2001. This paper was with the author 8 months for 3 revisions.
Management Science/Vol. 49, No. 6, June 2003 815