You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220520992

Power Asymmetry and Learning in Teams: The Moderating Role of Performance


Feedback

Article  in  Organization Science · April 2010


DOI: 10.1287/orsc.1090.0452 · Source: DBLP

CITATIONS READS
129 1,035

4 authors:

Gerben van der Vegt Simon Barend De Jong


University of Groningen Maastricht University
81 PUBLICATIONS   5,465 CITATIONS    37 PUBLICATIONS   1,415 CITATIONS   

SEE PROFILE SEE PROFILE

J. Stuart. Bunderson Eric Molleman


Washington University in St. Louis University of Groningen
14 PUBLICATIONS   2,919 CITATIONS    93 PUBLICATIONS   3,349 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Multiple team membership View project

Healthcare professionals motivation and quality of care - a systematic review View project

All content following this page was uploaded by Eric Molleman on 23 July 2014.

The user has requested enhancement of the downloaded file.


Organization Science informs ®

Vol. 21, No. 2, March–April 2010, pp. 347–361 doi 10.1287/orsc.1090.0452


issn 1047-7039  eissn 1526-5455  10  2102  0347 © 2010 INFORMS

Power Asymmetry and Learning in Teams:


The Moderating Role of Performance Feedback

Gerben S. Van der Vegt, Simon B. de Jong


Department of Human Resource Management and Organizational Behavior, University of Groningen, 9700 AV Groningen, The Netherlands
{g.s.van.der.vegt@rug.nl, sdejong@deloitte.nl}

J. Stuart Bunderson
John M. Olin School of Business, Washington University in St. Louis, St. Louis, Missouri 63130,
bunderson@wustl.edu
Eric Molleman
Department of Human Resource Management and Organizational Behavior, University of Groningen, 9700 AV Groningen, The Netherlands,
h.b.m.molleman@rug.nl

P ast research suggests that power asymmetry within teams can have a stifling effect on team learning and performance.
We argue here that this effect is contingent on whether power advantages within a team are used to advance individual
or collective interests. This study considers the moderating role of one factor that can influence the individual or collective
orientation of team members—the type of performance feedback that a team receives. We propose that whereas individ-
ual feedback reinforces the negative effects of power asymmetry on team learning, group feedback fosters a collective
orientation within a team that transforms power differences into a stimulus for team learning. Analysis of multisource,
multimethod data obtained from 218 individuals in 46 teams provided support for these hypotheses. Results also suggested
that team learning mediated the relationship between power asymmetry and team performance. These findings suggest that
power asymmetry can be a resource for and not just an obstacle to team learning in power-asymmetric teams.
Key words: power asymmetry; learning; teams; performance feedback; task dependence; field study
History: Published online in Articles in Advance July 2, 2009.

A growing body of research evidence in the organi- reflection and process improvement did not occur when
zation and management literatures suggests that work even one team member had significant power over oth-
teams can differ considerably in the extent to which ers. Similarly, Edmondson (2002) conducted a qual-
they pursue activities related to learning and continu- itative study of learning behavior in 12 teams and
ous improvement and that these differences have impor- found that power differences were negatively associ-
tant implications for team performance (Bunderson and ated with team learning. Other research has suggested
Sutcliffe 2002, 2003; Edmondson 1999, 2002; Gibson that in teams where there is stratification or inequal-
and Vermeulen 2003; Schippers et al. 2003; Van der ity in power relations, team members do not learn from
Vegt and Bunderson 2005; see also Argote et al. 2001). member differences (Bunderson 2003a, 2003b; Pitcher
In today’s market environment, where a firm’s success and Smith 2001). And Eisenhardt and Bourgeois (1988)
is contingent on its ability to improve and adapt more found that power inequality heightens intragroup politics
quickly than its competitors, teams with the capacity to and undermines team self-improvement. Together, these
continually improve processes and approaches to operate studies suggest that power differences can significantly
more quickly, efficiently, and intelligently have become interfere with team learning and that real learning may
a critical competitive advantage. Understanding the fac- only occur in teams where members have equal levels
tors that promote or inhibit these interactive learning of power.
processes within teams has therefore become an impor- This conclusion is problematic for a number of rea-
tant research agenda for management scholars (Argote sons. First, power differences exist in virtually all teams.
1999, Edmondson 1999). In classic formulations, power is defined as a function of
One key factor that may be important for team dependence (Emerson 1962). A has power over B (i.e.,
learning is the configuration of power within a team. dyadic asymmetry in power exists) when B is depen-
A number of researchers have suggested that power dif- dent on A for valued or needed resources (physical, emo-
ferences within a team may stifle team learning behav- tional, informational) (Blau 1964). In work teams, where
iors. For example, Brooks (1994) found that group multiple individuals coordinate differentiated efforts to
347
Van der Vegt et al.: Power Asymmetry and Learning in Teams
348 Organization Science 21(2), pp. 347–361, © 2010 INFORMS

complete some task, resource dependence among group power differences into an asset for both team learning
members is simply a fact of life. Furthermore, although and performance improvement. We tested hypotheses
there may be cases where the power-dependence rela- based on these general propositions in a sample of 218
tions that exist between task group members will balance employees from 46 work teams. Our results underscore
out (i.e., A’s dependence on B is balanced by B’s depen- the importance of a collective orientation for a team’s
dence on A such that neither has a power advantage), it ability to leverage member differences for team learn-
is unreasonable to assume that perfect balance among all ing and improvement (as suggested by Van der Vegt and
group members will be the normal or even a common Bunderson 2005).
state of affairs. In most teams, asymmetries in power-
dependence relations will exist because of differences
in the formal or informal resources controlled by team Theory and Hypotheses
members as a function of different roles, tenures, or nat-
Power Asymmetry and Team Learning
ural endowments (e.g., intelligence or charisma; Ragins
We conceptualize team learning in this paper using a
and Sundstrom 1989).
Moreover, there is evidence to suggest that power group process lens (Edmondson et al. 2007). That is,
differences do not inevitably stifle team learning and we define team learning as “activities by which team
that they may even play a facilitating role. For exam- members seek to acquire, share, refine, or combine
ple, Edmondson (2002) found that the effect of power task-relevant knowledge through interaction with one
differences on team learning depended on how these another” (Van der Vegt and Bunderson 2005, p. 534).
differences were handled and, specifically, on whether Examples of team learning behaviors include experi-
high-power members adopted a participative approach menting with new approaches or ideas, reflecting on
(see also Nembhard and Edmondson 2006). Van der past actions and action-outcome relationships, seek-
Vegt et al. (2006) also found that teams varied in the ing different perspectives, and evaluating alternatives
extent to which high-power members helped lower- (e.g., Edmondson 1999, 2002; Gibson and Vermeulen
power members and that these differences were associ- 2003; Schippers et al. 2007; West 1996). These team
ated with performance. Work by Larson and colleagues learning behaviors are a specific class of “interaction
(1998) suggests that high-power members can play a key processes” in teams— processes involving interactions
role in encouraging the discussion of unique member between team members that play a key role in transform-
information. ing input factors into performance outcomes (Hackman
We therefore suggest that a negative main effects and Morris 1975). We therefore view team learning
model of the relationship between power differences and as conceptually distinct from the outcomes that might
team learning may be too simplistic and that the time result from an engagement in learning-related activi-
has come to articulate and investigate moderating factors ties, outcomes such as more adaptive decisions and
(Chen et al. 2001), i.e., factors that mitigate the nega- actions, improved performance, or, perhaps in some
tive effects of power differences or that transform power cases, decreased efficiency resulting from a misallo-
differences into a resource for learning within teams. cation of effort (see Bunderson and Sutcliffe 2003).
This paper considers one such factor—the type of per- Past research has confirmed that team learning behav-
formance feedback a team receives. iors are distinct from other team interaction processes
Performance feedback plays a critical role in virtu- (Drach-Zahavy and Somech 2001) and that variance in
ally all theories of experiential learning (e.g., Ilgen et al. this specific interaction process across teams explains
1979, Nadler 1979) because it provides the mechanism unique variance in team innovation and improvement
by which an actor assesses the efficacy of past efforts (Drach-Zahavy and Somech 2001, Edmondson 1999,
and identifies areas of needed improvement (Vroom Van der Vegt and Bunderson 2005).
1964). Moreover, past research has suggested that the We conceptualize power asymmetry within a team
type of feedback a group receives and, more specif- as imbalance in the dyadic power-dependence relations
ically, whether a group receives feedback about indi- between and among group members as they perform
vidual and/or group performance, can affect a group’s their separate but interdependent tasks. So a given dyadic
orientation toward individual or collective improvement task relation between two members is power-imbalanced
(Hinsz et al. 1997). Building on this literature, we pro- when member A depends more on member B than B
pose that performance feedback may influence the rela- depends on A for resources (e.g., information, exper-
tionship between power asymmetry and team learn- tise, materials) needed to perform his or her work (see
ing by influencing whether team members are oriented Emerson 1962, Blau 1964). The overall level of power
toward individual or collective goals. More specifically, asymmetry within a team is a simple additive function
we will suggest that whereas individual performance of these dyadic power-dependence imbalances. Power
feedback reinforces the negative effects of power dif- asymmetry is therefore distinct from the construct of
ferences, feedback on group performance transforms power centralization. Although it is true that a team in
Van der Vegt et al.: Power Asymmetry and Learning in Teams
Organization Science 21(2), pp. 347–361, © 2010 INFORMS 349

which power is centralized will have at least some asym- learning behavior. Specifically, we suggest that power-
metric power relations, a team with asymmetric power advantaged members who adopt a collective and devel-
relations need not be at all centralized (e.g., each mem- opmental orientation in their interactions with dependent
ber may be powerful in some relations but dependent in others are able to focus attention, stimulate reflec-
others). Moreover, whereas power asymmetry will often tion, and encourage information exchange in ways that
result from member differences in knowledge, skill, and do not occur as naturally or easily in situations of
experience (i.e., “expertise diversity”; Van der Vegt and balanced power. This argument is supported by past
Bunderson 2005), expertise diversity does not always research examining the effects of power on groups and
lead to power asymmetry, because members may be dyads. For example, research by Larson and colleagues
expert in complementary domains, resulting in balanced (e.g., Larson et al. 1998) suggests that power-advantaged
power relations. group members who adopt a participative (i.e., other-
Past research suggests several compelling reasons why oriented) style encourage more open discussion and
we might expect power asymmetry to dampen team integration of both shared and unshared information
learning efforts. Lower-power members may be con- during group discussion than would otherwise occur.
cerned about negative evaluations from those on whom This is consistent with recent research by Nembhard and
they depend for resources and may therefore become Edmondson (2006), who found that higher-power group
tentative and inhibited in offering opinions and sharing members can prompt learning by inviting and showing
information (see Ridgeway 2001, Keltner et al. 2003). At appreciation for the input of others. Developmental feed-
the same time, higher-power members may simply pay back and questions from power-advantaged members are
less attention to lower-power members (Kipnis 1972, also more likely to focus attention or stimulate reflec-
Fiske 1993, de Jong et al. 2007) or may be unwill- tion among dependent members, because individuals are
ing to acknowledge any dependence on the insights or more attentive and committed to those on whom they are
knowledge of lower-power members because they fear dependent for valued resources (Rusbult and Van Lange
that doing so will undermine their privileged power rela- 2003, Wieselquist et al. 1999). Moreover, past research
tion (Lee 1997). These well-documented reactions to suggests that member differences within a group can
dyadic power differences would all seem to work against stimulate learning as different perspectives and prefer-
the goal of open dialogue among group members about ences interact and recombine (Bantel and Jackson 1989,
how to improve processes and better coordinate member Milliken and Martins 1996, Van der Vegt and Bunderson
efforts. 2005). Power asymmetry signals substantive differences
In contrast, we would argue that there are also between members on dimensions that are truly salient
good reasons to expect that power asymmetry may not to group members, because these differences affect both
always dampen team learning and that it can, in fact, task performance and social standing within the group.
stimulate learning behaviors. Research by Chen et al. These very differences could, therefore, be a stimulus
(2001) suggests that although exchange-oriented individ- for learning if, once again, team members with a power
uals may respond to power advantages in self-serving advantage help create an environment in which those
ways and therefore reinforce the negative effects just differences are not seen in threatening ways.
described, individuals with a communal orientation actu- In other words, power asymmetry within teams can
ally respond to power advantages by attending more either stifle or stimulate team learning, depending on
carefully to the needs and interests of dependent oth- how members respond to intra-team power differences—
ers. They explain this effect by suggesting that the particularly when they find themselves in positions of
effect of power on interpersonal behavior is contin- higher power. If members leverage power advantages
gent on whether an individual associates power with to advance and defend their own interests and status
self- or other-interested goals. Those who associate (an individualistic response pattern), power asymmetry
power with other-interested goals respond to power by is likely to stifle team learning. But if members leverage
assuming greater responsibility for the needs and wel- power advantages to invite reflection and the expression
fare of dependent others; those who associate power and integration of insights from all members (a collec-
with self-interested goals respond by treating others in tivistic response pattern), power asymmetry may actually
instrumental ways (which often means ignoring them, stimulate team learning. This conclusion is consistent
because they have little perceived instrumental value). with Blau’s (1964) analysis of why a power-advantaged
This argument is reminiscent of work by McClelland (i.e., more knowledgeable) member would choose to
(1975), who first distinguished between the personalized share knowledge with a more dependent colleague. Blau
(self-interested, uninhibited) versus socialized (other- argued that this choice will be based on the slope of
interested, restrained) exercise of power. the power-advantaged member’s indifference curve, i.e,
We would suggest that when power advantages are her marginal rate of substitution of the rewards associ-
leveraged in other-interested ways, power asymmetry ated with assisting by the rewards associated with pur-
within a group can become a powerful stimulus for suing her own interests. And this slope, Blau continues,
Van der Vegt et al.: Power Asymmetry and Learning in Teams
350 Organization Science 21(2), pp. 347–361, © 2010 INFORMS

is determined by “the degree to which [she] is oriented A possible explanation for the effects of group feed-
to colleagues as a reference group [i.e., a collectivistic back on group attention and effort observed by DeShon
orientation] rather than to [her] instrumental tasks and et al. (2004) and others (e.g., Barr and Conlon 1994) can
to the superiors who evaluate [her] performance [i.e., an be found in the work of Hinsz et al. (1997). They suggest
individualistic orientation]” (Blau 1964, p. 173). that group feedback “may change the self-attributional
But what factors might influence whether a team mem- focus from the individual to the group” (p. 53). That is,
ber adopts an individualistic or collectivistic response to when feedback is received at the individual level of anal-
power differences? The next section explores one possi- ysis, it directs attention toward the individual as the rele-
ble factor that figures prominently in theories of experi- vant actor and therefore motivates the individual to think
ential learning—the nature of the feedback that members about things that he or she could do to improve his or her
receive about their performance. individual performance. But when feedback is received
at the group level of analysis, attention is directed toward
The Moderating Effects of Performance Feedback the group as the relevant actor and the individual as one
Feedback can be defined as information about the effects piece of that larger collective. As a result, individual
of one’s actions or efforts on some criterion of inter- group members are motivated to think about things that
est (see Herold and Greller 1977, Taylor et al. 1984). they can do as a group to improve performance, e.g., bet-
In its original cybernetic formulation (Wiener 1948), ter coordination, greater information sharing, etc. This
the concept of feedback was used to describe the pro- argument is consistent with a study by Zander and Wolfe
cess by which systems (human or machine) self-regulate. (1964), in which it was found that group feedback led
Feedback about system performance allows a system to to more cooperation and less interpersonal strain within
reflect, adapt, and self-correct until desired performance a group than individual feedback did. This argument is
standards are achieved. Feedback scholars have referred also consistent with the broader literature on social cat-
to this function of feedback as a “cueing” function (see egorization, which has suggested that a shared group
Nadler 1979, Vroom 1964). In human systems, feedback identity increases intragroup cooperation and mitigates
also serves a “motivational” function by reinforcing the intra-group conflict, even when group members are very
promise of a reward and by reinforcing behavior-reward different from one another (Brewer and Miller 1984,
instrumentalities (Annett 1969; Ilgen et al. 1979, p. 361; Gaertner et al. 1996, Van der Vegt and Bunderson 2005).
Vroom 1964). Given these key functions, feedback has In other words, past research on group and individ-
long played a central role in theories of learning, con- ual performance feedback would suggest that feedback
tinuous improvement, and performance achievement. about group performance promotes a collective improve-
Although different types of feedback can be distin- ment orientation within a group (i.e., how are we doing;
guished (Nadler 1979), for the purpose of this study what can we do to improve our performance), whereas
we focus on the important distinction between feed- individual performance feedback promotes an individ-
back that provides information about the performance ual improvement orientation (i.e., how am I doing; what
of individuals within a group (individual feedback) and can I do to improve my performance). Given our ear-
feedback that provides information about the perfor- lier argument, this would imply that members of groups
mance of a group as a whole (group feedback) (see in which feedback is received about group performance
Barr and Conlon 1994, Nadler 1979, Hinsz et al. 1997, will be more likely to use their power advantages in
DeShon et al. 2004). Consider a group of salespeople ways designed to help the group improve. Power advan-
in which each team member covers a different district tages then become very useful for learning, because
within a broader region. In managing and motivating power-advantaged members can direct attention to areas
these team members, team leaders might choose to pro- of needed improvement, invite and encourage the shar-
vide (a) feedback to each salesperson about his or her ing of diverse perspectives, and initiate or facilitate dis-
sales for the year (individual feedback),1 (b) feedback cussions about how to manage member differences and
to the entire team about its aggregate sales (group feed- interdependencies (Larson et al. 1998, Nembhard and
back), or (c) both forms of feedback. Several studies Edmondson 2006). In contrast, members of groups in
have examined the independent and combined effects which feedback is received about individual performance
of individual and group feedback on performance in will be more likely to use their power advantages to
groups. Although findings from these studies have been leverage or improve their own performance (cf. De Jong
inconsistent, some important patterns have emerged. For et al. 2007). Power advantages then become an obsta-
example, DeShon et al. (2004) found that the effects of cle to reflective learning, as power-advantaged members
feedback on effort and performance were homologous focus on improving their own status or resource position,
across group and individual levels such that group or regardless of how this involves or affects their depen-
individual feedback increased attention and effort toward dent colleagues—who are likely to respond by becom-
group or individual goals, which led to higher group or ing more defensive, guarded, and inhibited (Keltner
individual performance. et al. 2003).
Van der Vegt et al.: Power Asymmetry and Learning in Teams
Organization Science 21(2), pp. 347–361, © 2010 INFORMS 351

When power asymmetry is low, however, group mem- Power Asymmetry, Learning, and
bers do not have power advantages with which to either Team Performance
advance their own interests or assist others. As a result, Finally, we would expect that the above effects of power
we would not expect power-balanced groups that receive asymmetry on team learning behavior will have impor-
group feedback to necessarily be more engaged learners tant implications for team performance. The argument
than power-balanced groups that receive individual feed- for a relationship between team learning behaviors and
back. In fact, when power relationships are balanced, team performance rests on the assumption that adap-
it may be difficult for group members to learn from tation and continuous improvement are critical perfor-
feedback about group performance, because there are mance capabilities in all teams, regardless of what they
no power-advantaged members to feel responsible for do or how much innovation is required in their context.
initiating a learning-oriented response. Power-balanced Teams that engage in learning behaviors are more likely
groups may therefore benefit more from individual feed- to learn from their mistakes, integrate new information
back, because individual feedback at least implies clear about their environment, benefit from member expe-
accountability (Goncalo and Duguid 2008). rience, effectively utilize member diversity, and better
Together, these arguments suggest that power asym- coordinate their efforts. Consistent with this reason-
metry will be positively related to team learning in ing, positive relationships between team learning behav-
groups that receive group performance feedback but neg- iors and supervisor ratings of team effectiveness have
atively related to team learning in groups that receive been reported in the pharmaceutical and medical prod-
ucts industry (156 teams; Gibson and Vermeulen 2003),
individual performance feedback. Stated formally:
the oil and gas industry (57 teams; Van der Vegt and
Hypothesis 1. Group performance feedback moder- Bunderson 2005), and the furniture manufacturing in-
ates the relationship between power asymmetry within dustry (51 teams; Edmondson 1999). Furthermore,
a team and team learning; power asymmetry is posi- Bunderson and Sutcliffe (2003) found that learning
tively related to team learning when group performance behaviors in 44 management teams were significantly
feedback is high. associated with business unit performance (unit prof-
itability), although the relationship was curvilinear
Hypothesis 2. Individual performance feedback mod- (teams could engage in too much learning behavior) and
erates the relationship between power asymmetry within moderated by past performance (team learning was more
a team and team learning; power asymmetry is nega- beneficial for lower-performing teams). On the whole,
tively related to team learning when individual perfor- these results provide robust support for learning behav-
mance feedback is high. iors as an important performance capability in teams.
It follows that by influencing team learning, the com-
These hypotheses treat individual and group perfor- bination of power asymmetry and different types of per-
mance feedback separately. But what about cases in formance feedback can have important implications for
which groups receive both individual and group feed- team performance. Specifically, we would expect that
back? We suggest that the combination of high indi- power asymmetry will be positively related to team per-
vidual and high group feedback creates an ambiguous formance when group performance feedback is high and
feedback situation in which it is not obvious to members that this effect will be mediated by team learning behav-
whether power advantages should be used for individ- ior. Similarly, we would expect that power asymmetry
ual or collective interests; both goals are equally possi- will be negatively related to team performance when
ble (see DeShon et al. 2004, Mitchell and Silver 1990, individual performance feedback is high and that this
Saavedra et al. 1993). In such a situation, we would effect will also be mediated by team learning behaviors.
expect that whether a given member adopts an individu- In other words, we would argue that power asymmetry
alistic or collectivistic response to power advantage will can have either positive or negative implications for team
be influenced by other factors, including, for example, performance, depending on the type of feedback that a
the stability of power relations or personality differences group receives; we would also argue that this effect is
(Keltner et al. 2003). It is not our purpose here to the- caused by the contingent effect of power asymmetry on
orize about how these other variables might affect the team learning under different types of performance feed-
relationship between power asymmetry and team learn- back. Stated formally:
ing in ambiguous feedback situations. Rather, we sug- Hypothesis 3. Team learning behavior mediates the
gest only that in cases of ambiguous performance feed- relationship between power asymmetry combined with
back, we would not expect to observe any consistent group performance feedback and team performance;
effects of power asymmetry on team learning and may, power asymmetry is positively related to team perfor-
in fact, observe no relationship. We can therefore offer mance when group feedback is high, because this com-
no hypothesis on this issue. bination results in higher team learning behavior.
Van der Vegt et al.: Power Asymmetry and Learning in Teams
352 Organization Science 21(2), pp. 347–361, © 2010 INFORMS

Hypothesis 4. Team learning behavior mediates the Measures


relationship between power asymmetry combined with
Power Asymmetry. This variable was measured using
individual performance feedback and team perfor-
a peer-rating approach. Consistent with our conceptual-
mance; power asymmetry is negatively related to
team performance when individual feedback is high, ization of power asymmetry, we used the following two
because this combination results in lower team learning items to measure the task dependence of a team mem-
behavior. ber (A) on another team member (B) based on Van der
Vegt et al. (1998): “How dependent are you on B for
materials, means, information, etc. in order to carry out
Method your work adequately?” and “How dependent is B on
Sample and Procedure you for materials, means, information, etc. in order to
The above hypotheses are explicitly concerned with carry out his or her work adequately?” (1 = “not depen-
team-level relationships and were therefore tested at the dent,” 7 = “completely dependent”). In all items, B was
team level of analysis using data obtained from 268 replaced by the name of a specific fellow team member
employees in 46 teams. These teams worked in a vari- (from the list provided by team supervisors).
ety of settings, ranging from the banking sector to the Because we measured A’s dependence on B as well
medical sector and consisted of at least three team mem- as B’s dependence on A from both their perspectives,
bers (M = 583, SD = 368). In most of these teams we were able to examine the relationship between A’s
at least some hierarchical differences and/or differences perception of his or her dependence on B and B’s per-
in degree of specialization between team members were ception of A’s dependence on him or her. This correla-
present. For example, teams from the banking sector tion was 0.31 (p < 0001). The magnitude of this cor-
included a product advisor and an assistant; the med- relation is in line with prior research involving dyads
ical teams consisted of different types of nurses with (e.g., Kenny and Acitelli 2001; see p. 443, Table 3, the
different specializations and/or different levels of author- correlations they reported ranged from 0.20 to 0.47).
ity within the team; and the technical, management, and Additionally, a univariate analysis of variance indicated
consultancy teams consisted of both senior team mem- that within-dyad variance was significantly smaller than
bers and more junior professionals. We reasoned that between-dyad variance (F 1402 1096 = 181, p <
these intrateam differences, which are characteristic of 0001). We therefore used the pooled perspectives of A
the types of member differences that exist in many work and B in each relationship to compute our power asym-
teams, should result in considerable power-dependence metry measure (cf. Borgatti and Everett 1999). More
asymmetries within these teams. specifically, we averaged A’s and B’s ratings of A’s
We approached teams via personal contacts with and dependence on B to obtain a pooled measure of A’s
a presentation to managers about the research project dependence on B, and A’s and B’s ratings of B’s depen-
during a postgraduate MBA course. When a manager dence on A to obtain a pooled measure of B’s depen-
agreed to participate, she or he informed the team and dence on A. Next, we computed the absolute difference
provided the researchers with information including the between A’s dependence on B and B’s dependence on A
names of team members. Two different types of ques- for each pair of team members and averaged all these
tionnaires were subsequently sent to the team: a supervi- scores per team member.3 We then averaged these indi-
sor questionnaire and a team member questionnaire. The vidual scores to obtain a team-level power-asymmetry
supervisor questionnaire was primarily used to collect score. This operationalization of power asymmetry is
team performance data, whereas the team member ques- summarized in the following formula:
tionnaire was used to collect power asymmetry, feed-
back, and learning data.2 We approached 50 supervisors k r 
j=1 i=1 DBAi − DABi /r j
and received 46 usable supervisor questionnaires (92%), 
and 218 team members (of a possible 268) returned their k
questionnaire (81%). Of these respondents, 146 were where DBAi is the mean of A’s and B’s perceptions of
female (67%) and the mean age of the respondents was the task dependence of B on A in relationship i, DABi
36.9 years (SD = 104). Twenty percent of these respon- is the mean of A’s and B’s perceptions of A’s depen-
dents had a high school degree, 44% a vocational degree, dence on B in relationship i, r is the number of relation-
29% a bachelor degree, and 7% a master’s degree or ships that team member j has in the team, and k is the
higher. Educational background also varied within our number of team members. The amount of power asym-
sample: 31% of the respondents had a degree related metry in these teams ranged from 0 to 1.75 (M = 060,
to economics, 21% had a degree related to the medical SD = 042), with higher scores indicating more power
field, 12% had a degree related to engineering, and 10% asymmetry between team members.4
had a degree related to business. Degrees in other fields
(e.g., law, linguistics, social, or natural sciences) were Group Performance Feedback. This variable was
held by less than 10% of the sample. measured with three items based on Van der Vegt et al.
Van der Vegt et al.: Power Asymmetry and Learning in Teams
Organization Science 21(2), pp. 347–361, © 2010 INFORMS 353

(2003): “We receive feedback as a team about the team Control Variables. Given that groups varied in size
performance”; “When we do not perform well, we are and that team size has been shown to relate to team per-
held responsible as a team”; and “We regularly receive formance in past research (e.g., Ancona and Caldwell
feedback about how good or bad we performed as a 1992), we controlled for team size in all of our analyses.
team.” These items were measured on a seven-point Team size information was collected using the super-
scale (1 = totally disagree, 7 = totally agree). The group visor questionnaire. Additionally, past research suggests
performance feedback scores ranged from 2.33 to 6.33, that the overall level of task interdependence might
and Cronbach’s alpha for the individual-level responses be associated with team processes and performance
was 0.84. (e.g., Wageman 1995). To explicitly consider the possi-
ble effects of task interdependence, we also controlled
Individual Performance Feedback. This variable was for this variable in all analyses. Task interdependence
also measured with three items. We reformulated the was measured by computing the mean of A’s pooled
group performance feedback items to reflect a focus on dyadic task dependence on B and B’s pooled dyadic task
individual instead of group performance: “I receive indi- dependence on A and averaging these scores within each
vidual feedback about my own performance”; “When team. Finally, we also controlled for expertise diver-
I do not perform well, I am held responsible as an sity in all our analyses, because research has shown
individual”; “I regularly receive feedback about how that this variable might be related to team learning and
good or bad I performed.” These items were mea- performance (e.g., Van der Vegt and Bunderson 2005).
sured with a seven-point scale (1 = totally disagree, We computed expertise diversity using functional back-
7 = totally agree). The individual performance feedback ground information provided by respondents in the ques-
scores ranged from 2.63 to 6.58, and Cronbach’s alpha tionnaire. We used Blau’s (1977) formula, 1 − pi2 , to
was 0.86. form an aggregate measure of expertise diversity, where
Team Learning. This variable was measured with p represented the proportion of employees of a work
six items adapted from Schippers et al. (2003). The team in the ith category. A higher index score indicates
greater expertise diversity among team members.
Schippers et al. (2003) measure is very similar to other
process measures of group learning (see Edmondson Discriminant and Convergent Validity. We used con-
et al. 2007) and focuses specifically on team interac- firmatory factor analysis to assess the discriminant and
tion processes associated with reflecting on actions and convergent validity of the group feedback, individ-
outcomes—key elements of experiential learning. Spe- ual feedback, and team learning scales. We computed
cific items included: “We talk about different ways in parameter estimates with the LISREL 8.51 computer
which we can reach our objectives”; “In this team the package, using the maximum likelihood method. We first
results of actions are evaluated”; “If things don’t work tested a model in which the group and individual feed-
out as planned, we consider what we can do about back and team learning items loaded on three corre-
it”; “We ask ourselves how effective our procedures for sponding latent constructs. The overall fit of the model
reacting to changes are”; “We regularly discuss whether to the data was adequate ( 2 51 218 = 16665, p <
the team is working effectively”; and “The team often 0001, the standardized root mean square of the residu-
reviews its methods for getting the job done.” These six als [SRMSR] was 0.05, the goodness-of-fit index [GFI]
items were measured on a seven-point scale (1 = totally was 0.89, and the comparative fit index [CFI] was 0.92).
disagree, 7 = totally agree). Cronbach’s alpha was 0.89. The factor loading of each item was significant at the
0.001 level or better.
Team Performance. Because we focused on a rather To further evaluate the discriminant validity of our
diverse sample of teams with different tasks and respon- scales, we computed two alternative models. In the first
sibilities, we used a broad measure of team perfor- model, all feedback items loaded on one latent feedback
mance (Ancona and Caldwell 1992). Specifically, we construct, and the learning items loaded on a separate
asked each supervisor to compare the performance of a learning construct. The fit of this model was significantly
focal team with that of relevant other teams with sim- worse than that of the hypothesized measurement model
ilar tasks and customers. A pilot study revealed that ( 2 2 = 21987, p < 0001, SRMSR = 007, GFI =
the following performance criteria were deemed relevant 077, CFI = 082). The second model contained only one
by supervisors from the different types of work groups: latent variable. Again, the fit of this model was signifi-
efficiency, continuity of the production process, use of cantly worse than that of the original model ( 2 3 =
capacities, speed with which the team produces, control 49607, p < 0001, SRMSR = 011, GFI = 066, CFI =
over the production process, quality of work, and overall 068). Hence, we concluded that the hypothesized three-
team performance. Supervisors rated the team’s perfor- factor measurement model was the most appropriate for
mance using a seven-point scale (1 = far below average, the situation under consideration.
4 = average, 7 = far above average). Cronbach’s alpha A separate confirmatory factor analysis had to be con-
was 0.90. ducted for team performance because the performance
Van der Vegt et al.: Power Asymmetry and Learning in Teams
354 Organization Science 21(2), pp. 347–361, © 2010 INFORMS

ratings were provided by team supervisors. We tested a Analyses


model in which all seven team performance items loaded We used hierarchical multiple regression analyses to
on a single factor, and the overall fit of this model was test our hypotheses. Following the recommendations of
very good ( 2 14 46 = 1395, n.s.). The fit indices were Aiken and West (1991), we standardized all independent
all satisfactory: The SRMSR was 0.043, GFI was 0.92, variables and computed interaction effects by taking the
and CFI was 0.99, and all factor loadings were signifi- product of the respective standardized independent vari-
cant at the 0.001 level or better. ables. We tested four models to isolate the contribution
Interrater Agreement and Reliability. All our con- of different terms. The first model tested the effects of
structs are explicitly conceptualized at the team level. our control variables. In the second model, the main
Nevertheless, because individual and group feedback effects of power asymmetry and group and individual
and team learning were measured from individual feedback were added to the regression model, followed
responses, it is important to evaluate the measure- by the inclusion of the hypothesized two-way interac-
ment assumption that responses from members of the tions between power asymmetry and feedback in the
same team will converge (see Kozlowski and Klein third model. In the case of our performance-dependent
2000). This assumption was investigated by calculating variable, we also examined a fourth model to test the
the interrater agreement coefficient (rwg ; James et al. mediating effect of team learning.
1984). Median rwg values were 0.69 for group feedback
(mean = 070), 0.74 for individual feedback (mean = Results
073), and 0.78 for team learning (mean = 076). These
numbers suggest that team members agreed in their rat- Descriptive Statistics
ings of these variables. The means, standard deviations, and Pearson zero-order
In addition, we would expect variation between teams correlations between variables are presented in Table 1.
in the ratings of feedback and team learning, and we As can be seen, the correlation between group and indi-
would expect ratings of members from the same team vidual feedback was positive and significant (r = 073,
to be more similar to one another than to the ratings p < 0001), indicating that these variables share slightly
of these constructs by members of other teams (see more than 50% of their variance. While this correla-
Bliese 2000). These expectations were investigated by tion is below the 0.75 standard for correlated indepen-
computing the intraclass correlation coefficients (ICC[1] dent variables suggested by Miles and Shevlin (2001),
and ICC[2]; Bliese 2000). One-way analyses of vari- it was high enough to warrant careful examination of
ance suggested that team member ratings of group and variance inflation factors in all regression models. Team
individual feedback and team learning all differed signif- learning was positively related to both types of feedback
icantly (p < 001) between teams. The ICC(1) was 0.21 (r = 058, p < 0001 for group feedback and r = 062,
for group feedback and 0.17 for individual feedback p < 0001 for individual feedback). Finally, perceptions
and learning. These figures indicate that a consider- of team learning were positively associated with super-
able amount of the variance in ratings was due to team visor ratings of team performance (r = 040, p < 001).
membership (Bliese 2000). The reliability of the group
means was examined by calculating the ICC(2) coeffi- Hypothesis Tests: Team Learning
cients. The ICC(2) values were 0.69 for group feedback, Regression results for models with team learning as the
0.63 for individual feedback, and 0.62 for team learn- dependent variable are summarized in Table 2. There
ing. Together, these results support the aggregation of was some evidence for a positive relationship between
individual team member responses to create team-level task interdependence and team learning in Model 1,
variables for group feedback, individual feedback, and although the overall effect of the control variables was
team learning. not significant. The addition of power asymmetry and

Table 1 Descriptive Statistics and Pearson Zero-Order Correlations Among the Study Variables

Standard
Variable Means deviation 1 2 3 4 5 6 7

1. Team size 583 368


2. Expertise diversity 044 025 030∼
3. Task interdependence 373 093 −024 −010
4. Power asymmetry 060 042 006 −019 −012
5. Group performance feedback 437 092 002 −018 031∗ 007
6. Individual performance feedback 467 090 010 023 006 012 073∗∗∗
7. Team learning 461 076 013 024 026∼ 000 058∗∗∗ 062∗∗∗
8. Team performance 477 075 006 002 002 013 −002 014 040∗∗

Notes. N = 46 teams. ∼ p < 010; ∗ p < 005; ∗∗


p < 001; ∗∗∗
p < 0001.
Van der Vegt et al.: Power Asymmetry and Learning in Teams
Organization Science 21(2), pp. 347–361, © 2010 INFORMS 355

Table 2 Results of Regression Analysis for Team Learning

Model 1 Model 2 Model 3

Model Variable B SE B SE B SE VIF

1. Control Team size 012 012 008 009 004 009 119
variables Expertise diversity 068 055 055 051 080 051 174
Task interdependence 024∗ 011 015 010 016∼ 009 130
2. Main Power asymmetry −002 012 −007 011 115
effects Group performance feedback 020 016 034∗ 015 241
Individual performance feedback 031∗ 015 017 014 211
3. Interactions Power asymmetry∗
Group performance feedback 051∗∗ 017 232
Power asymmetry∗
Individual performance feedback −049∗∗ 017 229
2
R 014 049 061
R2 014∼ 035∗∗∗ 012∗

Notes. n = 46 teams. ∼ p < 010; ∗ p < 005; ∗∗


p < 001; ∗∗∗
p < 0001. B = unstandardized regression coefficient; SE =
standard error.

group and individual feedback in Model 2 significantly (positive or negative), 12% of the observed variance in
increased the explanatory power of the model (R2 = team learning could be accounted for by the contin-
035, p < 0001) with a positive and significant coef- gent effect of power asymmetry, i.e., the effect of power
ficient for individual Feedback; groups that reported asymmetry at different levels of group or individual per-
receiving more individual feedback also reported more formance feedback. Variance inflation factors (VIFs) for
team learning behavior. The coefficients for the inter- the third model were all well below 4, suggesting that
actions between group feedback and power asymmetry multicollinearity was not a problem in these analyses
and between individual feedback and power asymme- (e.g., Langfred 2004, Miles and Shevlin 2001).
try, added in Model 3, were both significant at p < 001 We used the unstandardized regression coefficients
(b = 051 for group feedback and −0.49 for individual from the final equation to plot the relation between
feedback), and the addition of these two terms signif- power asymmetry and team learning at high (one stan-
icantly increased the explanatory power of the model dard deviation above the mean) and low (one standard
(R2 = 012, p < 005). These results are consistent deviation below the mean) levels of group and individ-
with Hypotheses 1 and 2 and suggest that whereas ual feedback (see Figures 1(a) and 1(b)). The pattern of
power asymmetry had no direct effect on team learning results depicted in Figures 1(a) and 1(b) is consistent

Figure 1 Power Asymmetry and Team Learning at Different Levels and Types of Feedback

7.0 (a) Group performance feedback 7.0 (b) Individual performance feedback

6.0 6.0

5.0 5.0
Team learning

Team learning

4.0 4.0

3.0 3.0

2.0 2.0

Group performance feedback high Individual performance feedback high


Group performance feedback low Individual performance feedback low
1.0 1.0
Low High Low High
Power asymmetry Power asymmetry
Van der Vegt et al.: Power Asymmetry and Learning in Teams
356 Organization Science 21(2), pp. 347–361, © 2010 INFORMS

with our hypotheses. Simple slope tests suggested that Simple slope tests suggested that power asymmetry
power asymmetry was positively associated with team was positively related to team performance when group
learning when group feedback was high (b = 044, SE = feedback was high (b = 044, SE = 022, p < 005),
019, p < 005) and negatively associated with team but only weakly negatively related to team performance
learning when group feedback was low (b = −059, when group feedback was low (b = −038, SE = 020,
SE = 020, p < 001), consistent with Hypothesis 1 (see p < 010). Simple slope tests for the interaction between
Figure 1(a)). To put these slopes into perspective, team power asymmetry and individual feedback revealed that
learning at high levels of power asymmetry was a full power asymmetry was negatively related to team perfor-
Likert-scale point higher (which is more than one stan- mance when individual feedback was high (b = −061,
dard deviation) when group feedback was high than SE = 029, p < 005) and positively related to team
when it was low. Consistent with Hypothesis 2, simple performance when individual feedback was low (b =
slope tests also suggested that power asymmetry was 063, SE = 024, p < 005). These effects are plotted in
negatively related to team learning when individual feed- Figures 2(a) and 2(b).
back was high (b = −056, SE = 021, p < 005) and Finally, Model 4 allows us to consider the mediated
positively related to team learning when individual feed- effects in Hypotheses 3 and 4 by adding team learning
back was low (b = 042, SE = 018, p < 005). to the regression model. The addition of team learn-
ing to the model resulted in a significant increase in R2
Hypothesis Tests: Team Performance (R2 = 013, p < 005) with a positive and significant
Regression results for models with team performance coefficient for team learning (b = 042, p < 005). Fur-
as the dependent variable are summarized in Table 3. thermore, with team learning in the model, the group and
None of the control variables was significantly related to individual feedback interactions became nonsignificant
team performance in Model 1, and the addition of power (b = 010, n.s., for power asymmetry X group feedback
asymmetry along with both types of feedback in Model 2 and b = −035, n.s. for power asymmetry X individ-
did not increase the explanatory power of the model. ual feedback). These results suggest that team learning
The addition of the two interaction terms in Model 3 did mediated the moderated effects of power asymmetry on
explain significant variance in performance (R2 = 015, team performance, as suggested in Hypotheses 3 and 4.
p < 005), with a positive and significant coefficient for
the group feedback interaction (b = 041, p < 005) and Additional Analysis
a negative and significant coefficient for the individual We argued that the combination of group and individual
feedback interaction (b = −062, p < 001). So whereas performance feedback creates an ambiguous feedback
power asymmetry had no direct effect on performance situation and that we therefore should not observe any
(positive or negative), the contingent effect of power significant relationship between our dependent variables
asymmetry at different levels of performance feedback and either (a) a group by individual performance feed-
accounted for fully 15% of the variance in team perfor- back interaction or (b) a group feedback by individual
mance. Once again, multicollinearity did not appear to feedback by power asymmetry interaction. Follow-up
be a problem, given that VIF scores for the fourth model analyses were consistent with this expectation. Specif-
were all well below 4. ically, the addition of a group feedback by individ-

Table 3 Results of Regression Analysis for Team Performance

Model 1 Model 2 Model 3 Model 4

Model Variable B SE B SE B SE B SE VIF

1. Control Team size 005 012 007 012 004 012 002 011 120
variables Expertise diversity 001 058 −039 066 −054 069 −090 066 186
Task interdependence 003 012 010 013 012 012 003 012 141
2. Main Power asymmetry 010 015 001 015 005 013 117
effects Group performance feedback −031 021 −023 021 −036 020 286
Individual performance feedback 032 020 021 020 011 019 223
3. Interactions Power asymmetry∗
Group performance feedback 041∗ 021 010 023 262
Power asymmetry∗
Individual performance feedback −062∗∗ 023 −035 023 282

4. Mediation Team learning 042 016 243
R2 001 008 023 036
R2 001 007 015∗ 013∗

Notes. n = 46 teams. ∗ p < 005; ∗∗


p < 001.
Van der Vegt et al.: Power Asymmetry and Learning in Teams
Organization Science 21(2), pp. 347–361, © 2010 INFORMS 357

Figure 2 Power Asymmetry and Team Performance at Different Levels and Types of Feedback
7.0 (a) Group performance feedback 7.0 (b) Individual performance feedback

6.0 6.0
Team performance

Team performance
5.0 5.0

4.0 4.0

3.0 3.0

2.0 2.0
Group performance feedback high Individual performance feedback high
Group performance feedback low Individual performance feedback low
1.0 1.0
Low High Low High
Power asymmetry Power asymmetry

ual feedback interaction term to Model 3 in regression be positive at high levels of group performance feedback
models predicting team learning and performance did and negative at high levels of individual performance
not increase the explanatory power of either regression feedback. But we also found that power asymmetry
model or yield a significant coefficient for the interac- was positively associated with team learning and perfor-
tion term (b = 004, SE = 011 for team learning; b = mance when individual performance feedback was low
−019, SE = 012 for team performance). Moreover, the and negatively associated with team learning and per-
addition of a group feedback by individual feedback by formance when group performance feedback was low.
power asymmetry interaction term as Model 4 also did Our theory did not anticipate this result. We expected
not increase the explanatory value of either regression to find that when group or individual performance feed-
model or produce significant coefficients (b = −018, back was low, we would observe no relationship between
SE = 014 for team learning; b = −014, SE = 017 for power asymmetry and either team learning or perfor-
team performance).5 mance, because lack of feedback provides no external
information about how to leverage power advantages.
Discussion On further reflection, we attribute this result to an
Past research suggests that power differences within avoidance dynamic among survey respondents. That is,
teams can stifle team learning. Our goal has been to survey respondents who “strongly disagreed” with state-
explore the thesis that a negative relationship between ments about whether their team receives individual
power differences and learning is not inevitable but is (or group) feedback may be signaling that, in fact,
contingent on whether power advantages within a team
their team deliberately and consciously avoids individual
are used to advance individual or collective interests.
(or group) feedback. In other words, low group feed-
We suggested that when power-advantaged team mem-
back scores may reflect a more individualistic orientation
bers adopt a collective orientation, power differences
can actually be a stimulus for team learning. We fur- within a team, whereas low individual feedback scores
ther proposed that the type of feedback (individual or reflect a more collectivistic orientation. If this is the case,
group) that team members receive will influence whether then the unexpected pattern of results just noted is sim-
they adopt an individualistic or collectivistic orientation ply a function of our measurement approach and, more
and will therefore moderate the relationship between important, is perfectly consistent with our theory. To test
power asymmetry and team learning. Our empirical test this possibility, we looked at the relationships between
of these hypotheses generated results that were clearly power asymmetry and team learning when group or indi-
in line with our expectations, but with some results that vidual performance feedback was at 4.0 on the Likert
bear further consideration. scale (i.e., “neither agree nor disagree”). As expected,
For example, we found support for our central these slopes did not differ significantly from zero.
hypotheses that the relationship between power asym- Second, as we expected, we found no evidence for
metry and both team learning and performance would the combined effects of group and individual feedback
Van der Vegt et al.: Power Asymmetry and Learning in Teams
358 Organization Science 21(2), pp. 347–361, © 2010 INFORMS

on the relationship between power asymmetry and both these learning dynamics in power-asymmetric teams.
team learning and team performance. We attribute this to Specifically, our theory suggested that power-advantaged
the fact that combinations of individual and group feed- team members with a collective mindset were key to
back create an ambiguous feedback situation in which fostering a learning dynamic within these teams. Future
it is not clear to members whether power advantages research should investigate the microdynamics of learn-
should be used for individual or collective interests (see ing in power-asymmetric teams to confirm the facilita-
DeShon et al. 2004, Mitchell and Silver 1990, Saavedra tive role of collectively oriented, power-advantaged team
et al. 1993). In such a situation, team members’ adop- members and to more clearly explicate the behaviors
tion of an individualistic or collectivistic response to through which this facilitation occurs. In-depth analyses
power advantage may be influenced by individual per- of group and/or dyadic interactions in laboratory or field
sonality differences rather than performance feedback groups would be very useful in this regard.
(Keltner et al. 2003), by internal states rather than by Future research might also compare the interactions
external cues. Future research might therefore examine that take place in power-balanced as opposed to power-
the role of variables such as a need for personalized asymmetric groups, with and without a collective orien-
versus socialized power (McClelland 1975), machiavel- tation, to better understand how power differences can
lianism (Christie and Geis 1970), or a social dominance either help or hinder intrateam learning.
orientation (Pratto et al. 1994) in ambiguous feedback Second, this study contributes to the literature on feed-
situations. We should note, however, that another pos- back by confirming that group and individual perfor-
sible explanation for the lack of a significant group-by- mance feedback can have very different implications for
individual feedback effect in this study is that group group process and performance. We found that although
and individual feedback was quite highly correlated group and individual feedback do seem to be corre-
in this sample. Although we found no evidence that lated (suggesting that teams adopting one type of feed-
this compromised our statistical analyses, covariation back are more likely to also adopt the other type of
between these two variables could limit our ability to feedback), these are two separate constructs that exhibit
find a significant interaction between the two, given a markedly different moderating effects on a key input-
to-process relationship. Furthermore, these results sup-
restriction of range in this variable (i.e., too few teams
port the notion that the type of feedback group members
with high-low and low-high group-individual feedback
receive (i.e., individual or group) is important in part
configurations).
because it influences whether group members adopt an
individual versus a collective orientation (Hinsz et al.
Theoretical Implications and Future
1997). Our results therefore contribute to the litera-
Research Directions
ture on team feedback by empirically investigating the
The results of this study suggest several important impli- scarcely researched effects of different types of perfor-
cations for research on teams, learning, and power. First, mance feedback (DeShon et al. 2004).
this study directly challenges the notion that power dif- Third, most research on power differences within
ferences always stifle learning within teams and suggests groups has conceptualized power at the individual and/or
instead that the relationship between power asymmetry the group level; i.e., power is something that each mem-
and team learning is contingent on member responses ber possesses to a greater or lesser degree and that can
to power advantages, which, in turn, are contingent on be evaluated to capture power configurations within a
whether members—and particularly power-advantaged group (e.g., power centralization). But power, in classic
members—adopt an individualistic or collective orienta- formulations, is a relational property that emerges from
tion in their interactions with other members. We iden- the dependence of one actor on resources controlled by
tified one factor that can influence this orientation— another (Emerson 1962, Thibaut and Kelley 1959). It is
group and individual performance feedback. But the therefore possible—and even likely—that a given group
theory presented in this paper would imply that any member will have a power advantage in one relation-
group design factor that encourages a collective mind- ship but not in another. This study explicitly acknowl-
set within a team should have the same effect. So, edges this possibility by conceptualizing power asym-
for example, we might hypothesize that group-based metry within a group as something that emerges from
incentives, outcome interdependence, the existence of a dyadic dependence, i.e., dependence of one dyad mem-
common external threat, or perhaps even strong group ber on task-related resources controlled by another. That
norms could also encourage a collective orientation is, we suggested that to fully represent power asymmetry
within a power-asymmetric team and therefore encour- within a group, one must attend to patterns of asymmet-
age learning-related behaviors. Investigating these vari- ric dependence between each of the dyadic relationships
ous design elements offers one promising direction for within that group. This study therefore introduces an
future research. approach to conceptualizing and measuring power dif-
Future research should also seek to confirm the social ferences in groups that more directly builds on the theo-
psychological processes that we posited as underlying retical underpinnings of power and dependence research.
Van der Vegt et al.: Power Asymmetry and Learning in Teams
Organization Science 21(2), pp. 347–361, © 2010 INFORMS 359

We should note, however, that there may be other Practical Implications


ways of conceptualizing power differences that would Based on the premise that any differences in mem-
be equally important to consider. For example, one ber power frustrate intrateam learning, past research has
might look at power advantages/disadvantages to explic- implied that in cases where team learning is the goal,
itly investigate how performance feedback affects the teams should be constructed in ways that minimize or
attitudes or behaviors of power-advantaged compared eliminate power differences. So, for example, differences
with power-disadvantaged dyad members. One might in tenure or experience should be minimized, author-
also conceptualize power asymmetry in terms of nontask ity relations should be eliminated, etc. We question this
resource dependence—e.g., power differences that arise approach for several reasons. First, attempts to elimi-
from factors such as race, gender, ethnicity, or physical nate or minimize power differences cannot ever succeed
attractiveness—to determine whether power differences entirely, because power and status differences can and
that arise from factors unrelated to a group’s task could will emerge around any number of member differences,
nonetheless be transformed into an impetus for learning. however subtle or intangible (e.g., assertiveness, gen-
Alternatively, one could examine the centralization of der, age or informal coalitions). Second, even if these
power relations within a team rather than, or in addition attempts successfully reduced power differences, they
to, asymmetry in power relations to determine whether come at a high cost, because they reduce the very rich-
power advantages in highly centralized teams might also ness of experience, perspective, contacts, and expertise
be leveraged for group learning. Clearly, the broader that stimulates and facilitates member learning.
issue of power differences within teams and their direct This paper offers an alternative solution to the design
or contingent relationship to team learning can be ana- and operation of teams where team learning is the goal.
lyzed in a number of different ways to address theo- Specifically, we suggest that when power differences
retically significant and practically important research exist between and among group members, it becomes
questions. critical to nurture a collective orientation within the team
so that these power differences can become an asset for
learning rather than a liability. Furthermore, we suggest
Study Limitations that one straightforward means of inducing this collec-
As with any study, certain aspects of this study sug- tive orientation is to provide group rather than individ-
gest the need for caution in how these results should be ual performance feedback to group members. In other
interpreted and/or generalized. For example, this study words, we argue that the solution to dealing with power
adopted a cross-sectional design that limits our ability to differences in teams where learning is a goal is not to
draw firm conclusions about the direction of causality. eliminate those differences but, rather, to manage them
Although alternative arguments based on reverse causal- so that they become an asset rather than an impediment.
ity assumptions are less likely from a theoretical stand-
point (e.g., it is not clear why team learning would inter- Acknowledgment
act with feedback to affect power asymmetry), the fact This research project was facilitated by a Netherlands Organi-
remains that we cannot conclusively eliminate them. sation for Scientific Research grant awarded to S. B. de Jong.
One might also question whether the use of a sur-
vey to collect measures of power asymmetry, feed- Endnotes
1
back, and team learning might have introduced common We expect that team leaders will adopt a consistent approach
method bias into these results (Podsakoff et al. 2003). to managing the performance of team members such that
Given that group measures of power asymmetry were if individual feedback is used to manage one team mem-
ber, it will likely be used with all team members. We are
constructed from dyadic comparisons, it seems highly
therefore interested in individual performance feedback as a
unlikely that correlations between these measures and “shared property” of a team in our theory and analyses (see
the other group process measures would be artificially Kozlowski and Klein 2000, p. 30)—and we explicitly examine
inflated. It is, however, quite possible that relationships this assumption of intrateam convergence empirically.
between the two types of feedback and team learning 2
Supervisors were not included as part of a team for the pur-
could be inflated because of common method variance. poses of this analysis.
3
We explicitly considered the effect of the high correla- When data from only one member of a dyad were available,
tion between our two types of feedback by examining dyadic asymmetry was computed using just those data.
4
variance inflation factors. And any inflated correlations Our measure of power asymmetry is related to but con-
ceptually and mathematically distinct from measures of
between team learning and the two types of feedback
power centralization (e.g., Bunderson 2003b, Freeman 1979).
should not affect the relationships of core theoretical Whereas power centralization assesses stratification in power-
interest in this study. So although we acknowledge the dependence relations, power asymmetry assesses aggregate
possibility of common method variance among some of imbalance in those relations. Power-centralized teams are
our study variables, we do not see this as a major threat therefore power asymmetric, but asymmetric teams are not
to the validity of these results. necessarily centralized. Our measure of power asymmetry also
Van der Vegt et al.: Power Asymmetry and Learning in Teams
360 Organization Science 21(2), pp. 347–361, © 2010 INFORMS

differs from power centralization measures in that it builds Bunderson, J. S., K. M. Sutcliffe. 2002. Why some teams empha-
from dyadic power-dependence ratings (because power is, ulti- size learning more than others: Evidence from business unit
mately, relational). We can, however, compute power central- management teams. H. Sondak (vol. ed.), E. A. Mannix, M. A.
ization from our dyadic ratings as follows: Neale (series eds.). Research on Managing Groups and Teams.
k Vol. 4, Toward Phenomenology of Groups and Group Member-
r
j=1 Pmax − APAj 
ship. Elsevier Science, Oxford, England, 49–84.
i=1 DBAi − DABi 
APAj = PC = 
r k − 1 Bunderson, J. S., K. M. Sutcliffe. 2003. Management team learning
orientation and business unit performance. J. Appl. Psych. 88
where APAj stands for average power (dis)advantage of team 552–560.
member j, PC stands for power centralization, and Pmax is the
Chen, S., A. Y. Lee-Chai, J. A. Bargh. 2001. Relationship orientation
largest power advantage score (APA) in the team. All other as a moderator of the effects of social power. J. Pers. Soc. Psych.
symbols were defined above. The correlation between this cen- 80 173–187.
tralization measure and our measure of power asymmetry was
Christie, R., F. Geis. 1970. Studies in Machiavellianism. Academic
0.80 (p < 0001), and analyses using power centralization gen-
Press, New York.
erated similar, albeit weaker, results to those obtained with
power asymmetry. We used a power asymmetry measure in de Jong, S. B., G. S. Van der Vegt, E. Molleman. 2007. The relation-
this study because our core theoretical interest is in imbalance ships among asymmetry in task dependence, perceived helping
behavior, and trust. J. Appl. Psych. 92 1625–1637.
rather than stratification.
5
Complete results are available on request from the first author. DeShon, R. P., S. W. J. Kozlowski, A. M. Schmidt, K. R. Milner,
D. Wiechmann. 2004. A multiple-goal, multilevel model of feed-
back effects on the regulation of individual and team perfor-
References mance. J. Appl. Psych. 89 1035–1056.
Aiken, L. S., S. G. West. 1991. Multiple Regression: Testing and
Interpreting Interactions. Sage, London. Drach-Zahavy, A., A. Somech. 2001. Understanding team innovation:
The role of team processes and structures. Group Dynamics:
Ancona, D. G., D. F. Caldwell. 1992. Demography and design—
Theory, Research, Practice 5 111–123.
Predictors of new product team performance. Organ. Sci. 3
321–341. Edmondson, A. C. 1999. Psychological safety and learning behavior
in work teams. Admin. Sci. Quart. 44 350–383.
Annett, J. 1969. Feedback and Human Behavior: The Effects of
Knowledge of Results, Incentives and Reinforcement on Learn- Edmondson, A. C. 2002. The local and variegated nature of learn-
ing and Performance. Penguin, Baltimore. ing in organizations: A group-level perspective. Organ. Sci. 13
128–146.
Argote, L. 1999. Organizational Learning: Creating, Retaining, and
Transferring Knowledge. Kluwer, Boston. Edmondson, A. C., J. R. Dillon, K. Roloff. 2007. Three perspec-
Argote, L., D. Gruenfeld, C. Naquin. 2001. Group learning in organi- tives on team learning: Outcome improvement, task mastery,
zations. M. E. Turner, ed. Groups at Work: Theory and Research. and group process. J. P. Walsh, A. P. Brief, eds. The Academy
Lawrence Erlbaum, Mahwah, NJ, 369–411. of Management Annals. Psychology Press, Lawrence Erlbaum,
Mahwah, NJ, 269–314.
Bantel, K. A., S. E. Jackson. 1989. Top management and innova-
tions in banking: Does the demography of the top team make a Eisenhardt, K. M., L. J. Bourgeois. 1988. Politics of strate-
difference? Strategic Management J. 10 107–124. gic decision-making in high-velocity environments: Toward a
midrange theory. Acad. Management J. 31 737–770.
Barr, S. H., E. J. Conlon. 1994. Effects of distribution of feedback in
work groups. Acad. Management J. 37 641–655. Emerson, R. M. 1962. Power-dependence relations. Amer. Soc. Rev.
27 31–41.
Blau, P. 1964. Exchange and Power in Social Life. Wiley, New York.
Fiske, S. T. 1993. Controlling other people: The impact of power on
Blau, P. B. 1977. Inequality and Heterogeneity. Free Press, New York. stereotyping. Amer. Psych. 48 621–628.
Bliese, P. D. 2000. Within-group agreement, non-independence, and Freeman, L. C. 1979. Centrality in social networks: Conceptual clar-
reliability: Implications for data aggregation and analysis. K. J. ification. Soc. Networks 1 215–239.
Klein, S. W. J. Kozlowski, eds. Multilevel Theory, Research, and
Methods in Organizations. Jossey-Bass, San Francisco. Gaertner, S. L., J. F. Dovidio, B. A. Bachman. 1996. Revisiting the
contact hypothesis: The induction of a common ingroup identity.
Borgatti, S. P., M. G. Everett. 1999. Models of core/periphery struc-
Int. J. Intercultural Rel. 20 271–290.
tures. Soc. Networks 21 375–395.
Gibson, C. B., F. Vermeulen. 2003. A healthy divide: Subgroups as a
Brewer, M. B., R. S. Miller. 1984. Beyond the contact hypothe-
stimulus for team learning. Admin. Sci. Quart. 48 202–239.
sis: Theoretical perspectives on desegregation. N. Miller, M. B.
Brewer, eds. Groups in Contact: The Psychology of Desegrega- Goncalo, J. A., M. M. Duguid. 2008. Hidden consequences of the
tion. Academic Press, Orlando, FL, 281–302. group-serving bias: Causal attributions and the quality of group
decision making. Organ. Behav. Human Decision Proc. 107(2)
Brooks, A. 1994. Power and the production of knowledge: Collec-
219–233.
tive team learning in work organizations. Hum. Resources Dev.
Quart. 5 213–235. Hackman, J. R., C. C. Morris. 1975. Group tasks, group interaction
process, and group performance effectiveness: A review and pro-
Bunderson, J. S. 2003a. Team member functional background and
posed integration. Adv. Experiment. Soc. Psych. 8 45–99.
involvement in management teams: Direct effects and the mod-
erating role of power centralization. Acad. Management J. 46 Herold, D. M., M. M. Greller. 1977. Feedback: Definition of a con-
458–474. struct. Acad. Management J. 20 142–147.
Bunderson, J. S. 2003b. Recognizing and utilizing expertise in work Hinsz, V. B., R. S. Tindale, D. A. Vollrath. 1997. The emerging con-
groups: A status characteristics perspective. Admin. Sci. Quart. ceptualization of groups as information processors. Psych. Bull.
48 557–591. 121 43–64.
Van der Vegt et al.: Power Asymmetry and Learning in Teams
Organization Science 21(2), pp. 347–361, © 2010 INFORMS 361

Ilgen, D. R., C. D. Fisher, M. S. Taylor. 1979. Consequences of indi- Pratto, F., J. Sidanius, L. M. Stallworth, B. F. Malle. 1994. Social
vidual feedback on behavior in organizations. J. Appl. Psych. 64 dominance orientation: A personality variable predicting social
349–371. and political attitudes. J. Pers. Soc. Psych. 67 741–763.
James, L. R., R. G. Demaree, G. Wolf. 1984. Estimating within-group Ragins, B. R., E. Sundstrom. 1989. Gender and power in organiza-
interrater reliability with and without response bias. J. Appl. tions: A longitudinal perspective. Psych. Bull. 105 51–88.
Psych. 69 85–98. Ridgeway, C. L. 2001. Gender, status, and leadership. J. Soc. Issues
Keltner, D., D. H. Gruenfeld, C. Anderson. 2003. Power, approach, 57 637–655.
and inhibition. Psych. Rev. 110 265–284. Rusbult, C. E., P. A. M. Van Lange. 2003. Interdependence, interac-
tion, and relationships. Ann. Rev. Psych. 54 351–375.
Kenny, D. A., L. K. Acitelli. 2001. Accuracy and bias in the percep-
tion of the partner in a close relationship. J. Pers. Soc. Psych. Saavedra, R., P. C. Earley, L. Van Dyne. 1993. Complex interdepen-
80 439–448. dence in task-performing groups. J. Appl. Psych. 78 61–72.
Kipnis, D. 1972. Does power corrupt? J. Pers. Soc. Psych. 24 33–41. Schippers, M. C., D. N. Den Hartog, P. L. Koopman. 2007. Learning
in teams: A measure and correlates. Appl. Psych.: An Internat.
Kozlowski, S. W. J., K. J. Klein. 2000. A multi-level approach to Rev. 56 189–211.
theory and research in organizations: Contextual, temporal, and
Schippers, M. C., D. N. Den Hartog, P. L. Koopman, J. A. Wienk.
emergent processes. K. J. Klein, S. W. J. Kozlowski, eds. Mul-
2003. Diversity and team outcomes: The moderating effects of
tilevel Theory, Research, and Methods in Organizations: Foun-
outcome interdependence and group longevity and the mediating
dations, Extensions, and New Directions.. Jossey-Bass/Pfeiffer,
effect of learning. J. Organ. Behav. 24 779–802.
San Francisco, 3–90.
Taylor, M. S., C. D. Fisher, D. R. Ilgen. 1984. Individuals’ reactions
Langfred, C. W. 2004. Too much of a good thing? Negative effects to performance feedback in organizations: A control theory per-
of high trust and individual autonomy in self-managing teams. spective. K. M. Rowland, G. R. Ferris, eds. Research in Person-
Acad. Management J. 47 385–399. nel and Human Resources Management. JAI Press, Greenwich,
Larson, J. R., Jr., P. G. Foster-Fishman, T. M. Franz. 1998. Leadership CT, 81–124.
style and the discussion of shared and unshared information in Thibaut, J. W., H. H. Kelley. 1959. The Social Psychology of Groups.
decision-making groups. Pers. Soc. Psych. Bull. 24 482–495. Wiley, New York.
Lee, F. 1997. When the going gets tough, do the tough ask for help? Van der Vegt, G. S., J. S. Bunderson. 2005. Learning and performance
Help seeking and power motivation in organizations. Organ. in multidisciplinary teams: The importance of collective team
Behav. Human Decision Proc. 72 336–363. identification. Acad. Management J. 48 532–547.
McClelland, D. 1975. Power: The Inner Experience. Irvington, Van der Vegt, G. S., J. S. Bunderson, A. Oosterhof. 2006. Expert-
New York. ness diversity and interpersonal helping in teams: Why those
who need the most help end up getting the least. Acad. Man-
Miles, J., M. Shevlin. 2001. Applying Regression and Correlation: agement J. 49 877–893.
A Guide for Students and Researchers. Sage, London.
Van der Vegt, G. S., B. J. M. Emans, E. Van de Vliert. 1998. Motivat-
Milliken, F., L. Martins. 1996. Searching for common threads: ing effects of task and outcome interdependence in work teams.
Understanding the multiple effects of diversity in organizational Group Organ. Management 23 124–143.
groups. Acad. Management Rev. 21 402–433.
Van der Vegt, G. S., E. Van de Vliert, A. Oosterhof. 2003. Informa-
Mitchell, T. R., W. S. Silver. 1990. Individual and group goals when tional dissimilarity and organizational citizenship behavior: The
workers are interdependent: Effects on task strategy and perfor- role of intrateam interdependence and team identification. Acad.
mance. J. App. Psych. 75 185–193. Management J. 46 715–727.
Nadler, D. A. 1979. Effects of feedback on task group behavior: Vroom, V. H. 1964. Work and Motivation. Wiley, New York.
Review of the experimental research. Organ. Behav. Human Per- Wageman, R. 1995. Interdependence and group effectiveness. Admin.
formance 23 309–338. Sci. Quart. 40 145–180.
Nembhard, I. M., A. C. Edmondson. 2006. Making it safe: The effects West, M. A. 1996. Learning and work group effectiveness: A con-
of leader inclusiveness and professional status on psychological ceptual integration. M. A. West, ed. Handbook of Work Group
safety and improvement efforts in health care teams. J. Organ. Psychology. John Wiley and Sons, Chichester, 555–579.
Behav. 27 941–966. Wiener, N. 1948. Cybernetics or Control and Communication in the
Pitcher, P., A. D. Smith. 2001. Top management team heterogeneity: Animal and the Machine. MIT Press, Cambridge, MA.
Personality, power, and proxies. Organ. Sci. 12 1–18. Wieselquist, J., C. E. Rusbult, C. A. Foster, C. R. Agnew. 1999. Com-
Podsakoff, P. M., S. B. MacKenzie, J. Lee, N. P. Podsakoff. 2003. mitment, pro-relationship behavior, and trust in close relation-
Common method biases in behavioral research: A critical review ships. J. Pers. Soc. Psych. 77 942–966.
of the literature and recommended remedies. J. Appl. Psych. 88 Zander, A., D. Wolfe. 1964. Administrative rewards and coordination
879–903. among committee members. Admin. Sci. Quart. 9 50–69.

View publication stats

You might also like